• Open access
  • Published: 07 September 2020
  • Rheumatoid arthritis
  • Yoshiya Tanaka   ORCID: orcid.org/0000-0002-0807-7139 1  

Inflammation and Regeneration volume  40 , Article number:  20 ( 2020 ) Cite this article

20k Accesses

48 Citations

1 Altmetric

Metrics details

Rheumatoid arthritis is an autoimmune inflammatory disease primarily characterized by synovitis which is accompanied by extra-articular organ involvement, such as interstitial pneumonia, in addition to clinical symptoms including pain, swelling, stiffness of multiple joints, fever, and malaise. Joint destruction progresses soon after the onset, and once the affected joints are deformed, the development of irreversible physical dysfunction is noted. Thus, proper diagnosis and treatment are required from the early stages of the disease. Although palliative therapy with glucocorticoids and anti-inflammatory drugs had been used, disease-modifying antirheumatic drugs (DMARDs) are currently used to suppress immune abnormalities and to control disease activity. DMARDs are classified into different groups, such as conventional synthetic DMARD, targeted synthetic DMARD, and biologic DMARD. The appropriate use of these drugs has allowed remission to be the therapeutic goal in all patients. By maintaining remission, these drugs have also been shown to prevent the progression of joint destruction and physical dysfunction over a long period. The advent of molecular-targeted therapies has allowed for the use of treatments based on pathological mechanisms, and such therapeutic strategies have also been applied to the treatment of various autoimmune inflammatory diseases. In the future, safer and more effective treatments, therapeutic strategies aimed at drug holidays or cure, and the introduction of precision medicine are expected.

Backgrounds

Rheumatoid arthritis is an autoimmune inflammatory disease primarily characterized by synovitis. It commonly affects women in their 30s to 50s, with an incidence of 1 in 150. It is accompanied by multi-organ disorders, in addition to pain, swelling, and stiffness of multiple joints. Joint destruction progresses rapidly after onset, resulting in irreversible physical dysfunction and deformation of the affected joints. Thus, proper diagnosis and treatment are required in the early stages of the disease.

The term rheumatism comes from the 2500-year-old Greek word meaning “flowing current,” indicating the flow of the affected joints in the entire body. This disease has afflicted humanity for a long period of time, and its treatment also has a long history. There is a 2500-year-old record stating that drinking a decoction of European white willow bark alleviates pain. In the nineteenth century, salicin was discovered as a component of the bark. In 1853, Gerhardt first synthesized acetylsalicylic acid, which had superior in vivo stability to salicin, and in 1897, acetylsalicylic acid was marketed as a tablet for arthralgia by Hoffmann in Bayer and is now used other conditions worldwide. In 1949, Dr. Hench reported the first administration of cortisone to patients with rheumatoid arthritis, and its dramatic effect was widely recognized. He was awarded the Nobel Prize in Physiology or Medicine in 1950. This resulted in the use of glucocorticoids and non-steroidal anti-inflammatory drugs for the treatment of rheumatoid arthritis in the twentieth century to alleviate pain and swelling. However, disease control was inadequate, and the progression of joint destruction could not be prevented.

In the late twentieth century, rheumatoid arthritis was recognized as an autoimmune disease primarily characterized by polyarthritis. Immunosuppressive drugs were then used to correct and suppress immune abnormalities and to control disease activity. In 1984, Köhler and Milstein were awarded the Nobel Prize in Physiology or Medicine for their techniques for producing monoclonal antibodies, which were immediately applied to clinical practice. In 1998, the first monoclonal antibody therapy targeting tumor necrosis factor (TNF), a cytokine that plays an important role in rheumatoid arthritis pathogenesis, was approved. Its revolutionary clinical effect caused a paradigm shift in treatment strategies. When immunosuppressive drugs are applied for the treatment of rheumatoid arthritis, they are referred to as disease-modifying antirheumatic drugs (DMARDs). At present, DMARDs are classified into synthetic DMARDs such as methotrexate and biologic DMARDs produced from biological agents.

In the twenty-first century, the appropriate use of DMARDs allowed rheumatologists to aim for clinical remission and to control joint destruction in all patients with rheumatoid arthritis. Such therapeutic strategies are also being applied to the treatment of various autoimmune inflammatory diseases. This article provides an overview of the pathology, clinical features, diagnosis, and treatment of rheumatoid arthritis, from the basic to the latest information.

Genome-wide analyses of single nucleotide polymorphisms in patients with rheumatoid arthritis have identified the human leukocyte antigen D-related B1 gene (HLA-DRB1) as the most relevant disease-susceptible gene and also identified other disease-susceptible genes. These include the protein tyrosine phosphatase non-receptor type 22 (PTPN22), cytotoxic T-lymphocyte antigen-4 (CTLA4), signal transducer and activator of transcription 4 (STAT4), TNF alpha-induced protein 3 (TNFAIP3), C-C motif chemokine ligand 21 (CCL 21), and peptidyl arginine deiminase 4 (PADI4) genes. In Japanese individuals, two haplotypes of the PADI4 gene have been identified that are disease-susceptible and non-susceptible, and the messenger RNA transcribed from the disease-susceptible gene is reported to be stable. Anti-cyclic citrullinated peptide (anti-CCP) antibodies are highly disease specific, and bone or cartilage destruction is more likely to progress in patients positive for anti-CCP antibodies. In contrast, typical environmental factors, including smoking, gingivitis, and intestinal bacterial flora, can cause modulation of the epigenome and the demethylation of histones and DNA, inducing the transcription of proinflammatory cytokines. In rheumatoid arthritis, although no specific autoantigen has been identified, it is understood that the interaction between genetic and environmental factors and the citrullination of extracellular matrix molecules, such as filaggrin and fibrinogen, causes epigenetic modifications, breaking immune tolerance to antigens and inducing autoimmunity [ 1 , 2 , 3 ].

Autoreactive T cells and B cells accumulate in the synovial tissues of patients with rheumatoid arthritis. T cells are immunologically tolerant to autoantigens; however, when self-tolerance is broken, autoreactive T cells are activated, and they stimulate B cells to induce the production of autoantibodies. Autoantibodies form immune complexes with antigens, which are deposited in tissues and activate complements to cause histological damage (type III allergy). Tissues with synovitis are characterized by angiogenesis or vasodilation, proliferation of synoviocytes, and accumulation of lymphocytes. In tissues with diffuse inflammation, the accumulation of memory T cells and B cells can result in the formation of lymphoid follicle-like and germinal center-like structures. Here, co-stimulators and proinflammatory cytokines are highly expressed, and close cellular interactions are observed in these structures [ 1 , 2 , 3 ].

In synovitis lesions, lymphocytes and synoviocytes produce large amounts of inflammatory cytokines, such as TNF, interleukin (IL)-1, and IL-6, which cause synovitis. In addition to systemic symptoms, such as low-grade fever and malaise, extra-articular organ involvement, such as keratoconjunctivitis sicca, sialadenitis, and interstitial pneumonia, is often observed. Furthermore, cytokine-stimulated synoviocytes produce matrix metalloproteinases (MMP), which are released into the synovial fluid. Cartilage is degraded by these enzymes and absorbed. In addition, synoviocytes and lymphocytes express receptor activator of nuclear factor-kappa B ligand (RANKL) to induce the maturation and activation of osteoclasts. Inflammatory granulation tissues containing proliferative and stratified synoviocytes grow until they come in contact with the bones. Multinucleated osteoclasts destroy and absorb bone, causing joint destruction, mainly at the point of contact [ 1 , 2 , 3 , 4 , 5 ].

Clinical features

The characteristic symptoms of rheumatoid arthritis are morning stiffness and polyarticular pain and swelling. Patients often complain of stiffness from the onset of the disease and experience difficulty in moving fingers on awakening, which is often described as having difficulty in forming a fist. Arthralgia is often associated with swelling and limited mobility. These symptoms are likely to appear in the joints of the fingers and toes (e.g., proximal interphalangeal, metacarpophalangeal, and metatarsophalangeal joints), knees, feet, hands, elbows, and cervical spine, among other areas. However, the distal interphalangeal joints are rarely the site of initial onset. In addition, patients often complain of general symptoms such as malaise, fatigue, and fever. Frequently accompanying symptoms include dry eyes associated with keratoconjunctivitis sicca (in approximately 45% of patients), xerostomia due to sialadenitis (40%), subcutaneous rheumatoid nodules on the extensor surface of the forearm (35%), numbness of the hands and feet associated with compressive neuropathy (25%), and shortness of breath on exertion or a dry cough due to interstitial pneumonia (15%).

As for the findings of clinical examinations, visual inspection and palpation tend to reveal tenderness and swelling of articular soft tissues and an accumulation of synovial fluid. Affected joints are characterized by inflammatory findings such as swelling, redness, and hot flashes. In general, multiple joints usually tend to be bilateral, symmetrical, and often mobile. As joint destruction progresses, various patterns of joint deformation are observed, such as the buttonhole deformity and swan-neck deformity of the finger joints. In case of atlantoaxial subluxation, occipital headache and numbness of the hands may occur. When inflammation spreads to the tendons, patients develop carpal tunnel syndrome due to the swelling of the trigger finger or wrist.

In terms of laboratory findings, approximately 80% of patients test positive for rheumatoid factors; however, even healthy individuals or patients with liver disease may be positive for these. Both the sensitivity and specificity of anti-CCP antibodies are 90% or higher, and patients with rheumatoid arthritis develop positivity prior to the onset of symptoms. In patients with high levels of anti-CCP antibodies or rheumatoid factors, the progression of joint destruction is rapid. Findings associated with inflammation include an elevated erythrocyte sedimentation rate and elevated C-reactive protein (CRP) levels, which are both elevated in association with disease activity. Furthermore, elevated white blood cell counts and normocytic hypochromic anemia are observed in association with inflammation. MMP-3 is a protease produced by synovial tissues and is associated with the progression of joint destruction.

Radiographic findings of the joints are important for the diagnosis and assessment of disease progression. Bone erosion localized to the affected joints is useful for diagnosis. Joint destruction is quantitatively assessed based on radiographic findings. The total Sharp score is calculated from multiple radiographs of the wrists, fingers, and toes. On these radiographs, the severity of joint space narrowing (indicating cartilage absorption) and bone erosion (indicating bone destruction) are converted into numerical scores, which are summed. Annual changes in the total Sharp scores are used to evaluate the progression of joint destruction and responses to treatment.

Meanwhile, the survival of patients with rheumatoid arthritis is considered to be shorter than that of the general population by 10 years or more owing to physical dysfunction, organ dysfunction, and adverse drug reactions. In Japanese patients, the causes of death associated with rheumatoid arthritis include respiratory dysfunction and renal failure, in addition to infection. Extra-articular organ involvement, such as interstitial pneumonia, directly affects prognosis.

The rheumatoid arthritis classification criteria published by the American College of Rheumatology (ACR) and the European League Against Rheumatism (EULAR) in 2010 are widely used for diagnosis (Fig. 1 ) [ 6 ]. These criteria, which define rheumatoid arthritis as arthritis that is persistent and may be destructive, were formulated with the aim to differentiate it from other forms of arthritis soon after onset and to allow for prompt initiation of treatment with DMARDs. In the first step, various diseases, such as connective tissue disease accompanied by arthritis of one joint or more, osteoarthritis, spondyloarthritis, and crystal-induced arthritis, are excluded. In the second step, scores of 4 items, arthritis (swelling of the small or intermediate/large joints), serologic test results (rheumatoid factors and anti-CCP antibodies), disease duration (6 weeks or longer), and acute-phase reaction (erythrocyte sedimentation rate and CRP), are weighted and added. A condition with a score of 6 points or higher out of 10 points is classified as definite rheumatoid arthritis. In addition, arthritis affecting one joint or more which is accompanied by typical bone erosion is also classified as rheumatoid arthritis, regardless of the score. When rheumatoid arthritis is comprehensively diagnosed based on the classification criteria, treatment with DMARDs is initiated. This diagnostic process potentially allows for therapeutic intervention prior to joint destruction.

figure 1

The rheumatoid arthritis classification criteria published by the ACR/EULAR in 2010. Modified from reference [ 6 ]

The assessment of disease activity is essential for planning therapeutic strategies. The 28-joint Disease Activity Score (DAS28), which is calculated based on the number of tender or swollen joints among 28 specified joints, 1-h erythrocyte sedimentation rate, and the patient’s global assessment of disease activity with a dedicated formula, is widely used for the objective assessment of disease activity. DAS28 scores are interpreted as follows: > 5.1, high disease activity; 3.2–5.1, moderate disease activity; < 3.2, low disease activity; and < 2.6, remission. Likewise, the Simplified Disease Activity Index (SDAI) and the Clinical Disease Activity Index (CDAI) are also widely used. For the evaluation of physical dysfunction, the Health Assessment of Questionnaire Disability Index, which consists of 20 questions on 8 categories regarding physical function in daily living, is widely used worldwide.

As described above, rheumatoid arthritis is often complicated by extra-articular involvement of the eyes, oral cavity, blood, lungs, heart, skin, nerves, kidneys, and lymph nodes, among others. Lung disorders are important organ dysfunctions that affect prognosis. Chest computed tomography (CT) is used to reveal lung disorders in approximately 70% of patients. Of these patients, approximately 50% are considered to present with nonspecific changes, approximately 30% are considered to present with interstitial pneumonia, and approximately 20% are considered to present with chronic infection or chronic obstructive pulmonary disease. Other pathological conditions that can occur include pleurisy, pulmonary alveolar hemorrhage, and bronchiectasis [ 1 , 2 , 3 ]. In cases of rheumatoid vasculitis, in which progressive arthritis is accompanied by systemic vasculitis in the skin, gastrointestinal tract, heart, lungs, spleen, and pleura, in addition to interstitial pneumonia. Furthermore, as the disease activity of rheumatoid arthritis increases, lymphoproliferative disease may concomitantly occur. Patients may also concomitantly develop various autoimmune diseases, including Hashimoto’s disease, other thyroid diseases, and secondary Sjögren’s syndrome. In all cases, conditions such as organ dysfunction associated with the pathology of rheumatoid arthritis, other comorbidities, concomitant infections with bacteria and viruses, and adverse events caused by drugs should be differentiated.

The basic policy of the treatment of rheumatoid arthritis involves immediate intervention after diagnosis, before the onset of joint destruction, to suppress arthritis and induce remission. Therapeutic strategies should be determined based on a comprehensive assessment of disease activity, imaging findings (such as radiography findings), complications, and comorbidities. Composite objective indices, such as the SDAI, CDAI, and DAS28, are widely used to evaluate disease activity. The therapeutic goal is remission, defined as a clinical condition involving no progression of either joint destruction or dysfunction in the future. Boolean remission and numerical targets, such as an SDAI score ≤ 3.3 and a CDAI score ≤ 2.8, have been statistically established as remission criteria [ 7 ].

In the standard initial treatment after the diagnosis of rheumatoid arthritis, methotrexate, a conventional synthetic DMARD, should be used if it is not contraindicated [ 8 , 9 ]. However, when no improvement is observed within 3 months or when no remission is achieved within 6 months, despite an increase to the full dose of methotrexate, the addition of biological DMARDs or Janus kinase (JAK) inhibitors is recommended. If the therapeutic goal is still not achieved, biological DMARDs or JAK inhibitors should be altered approximately 3–6 months later. Meanwhile, glucocorticoids are recommended for temporary use for up to 3 months as adjunctive therapy to relieve pain and swelling at the time of initial onset or the relapse of arthritis (Fig. 2 ).

figure 2

EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological DMARDs: 2019 update. Modified from reference [ 8 ]

While more than 10 conventional synthetic DMARDs have been approved, methotrexate is recommended as the first choice of standard drug therapy to be administered after the diagnosis of rheumatoid arthritis, if its use is not contraindicated. Methotrexate exerts antirheumatic effects mainly by controlling the proliferation of lymphocytes and synoviocytes in the mitotic phase via antagonistic action against folic acid. It is more effective than any other conventional synthetic DMARD. Adverse reactions to methotrexate include liver dysfunction and gastrointestinal dysfunction. In elderly patients, attention should be paid to myelosuppression, interstitial pneumonia, opportunistic infection, and lymphoproliferative disease. Co-administration of folic acid is useful for reducing adverse reactions. Sulfasalazine and leflunomide are recommended when methotrexate use is contraindicated.

Biological DMARDs are selected when responses to synthetic DMARDs are inadequate. In Japan, TNF-targeting drugs (i.e., infliximab, etanercept, adalimumab, golimumab, and certolizumab), IL-6-targeting drugs (i.e., tocilizumab and sarilumab), and the T cell-selective co-stimulation modulator abatacept can be administered by injection or drip infusion. All of these drugs exert prompt and potent clinical effects. Their use in combination with methotrexate allows the induction of remission in approximately half of the cases. Biological DMARDs can also prevent the progression of joint destruction and dysfunction for long periods of time [ 10 ].

In contrast, inhibitors against JAKs, which are intracellular signaling molecules such as cytokines, are classified as targeted synthetic DMARDs. Tofacitinib, baricitinib, peficitinib, upadacitinib, and filgotinib are used for the treatment of rheumatoid arthritis and differ in their selectivity for different JAK isoforms [ 11 , 12 , 13 , 14 , 15 ]. Although they are all orally administered drugs, they have multi-target effects and exert clinical effects just as promptly as biological DMARDs. JAK inhibitors can be used alone or in combination with methotrexate.

In Japan, when biological DMARDs were used for the treatment of rheumatoid arthritis, post-marketing surveillance was required to verify their safety [ 16 , 17 ]. According to an all-case surveillance study on infliximab administered to 5000 patients over a period of 6 months, adverse reactions occurred in 1401 patients, and serious adverse reactions occurred in 308 patients, including bacterial pneumonia in 108 patients, interstitial pneumonia in 25, pneumocystis pneumonia in 22, and tuberculosis in 14. The risk factors for pneumonia due to the use of biological DMARDs include advanced age, a history of respiratory diseases, and concomitant use of glucocorticoids. Owing to these factors, the use of biological DMARDs requires medical management and treatment for serious adverse reactions such as pneumonia, tuberculosis, and other opportunistic infections, and guidelines for the prevention and treatment of adverse reaction have been established. For example, the prophylactic administration of isoniazid is recommended for patients with risk factors for tuberculosis, and pneumococcal vaccination is recommended for patients with risk factors for pneumonia.

In addition, JAK inhibitors should not be used without careful consideration as they are orally administered drugs with multi-target effects based on the inhibition of intracellular signaling. Screening before their use and monitoring during treatment should be strictly performed. They should be administered by physicians who can perform systemic management in the event of adverse events. JAK inhibitors should not be used in patients with serious infections, liver disorders, renal disorders, or blood cell disorders, and it is necessary to establish evidence on its long-term safety regarding the development of infections such as herpes zoster and malignant tumors such as lymphoma.

In our department, approximately 4000 patients have been treated with biological DMARDs that were introduced into or substituted for other drugs in the FIRST registry since 2003. According to the clinical pathway, these patients were admitted and examined for contraindications and factors requiring caution; then, they were carefully evaluated as to whether they showed indications for the use of these drugs. In addition, the safety and efficacy of the drugs in these patients were rigorously monitored through outpatient visits for over 1 year. In particular, CT from the head to the abdomen detected early lung cancer in 11 patients and non-tuberculous mycobacteriosis in 13 patients from among approximately 2500 patients before they became symptomatic. This demonstrates the importance of in-depth screening.

Development

New therapeutic systems and strategies for rheumatoid arthritis have been applied to other connective tissue diseases and rheumatic diseases. While their indications have been expanded, these systems and strategies have also led to breakthroughs in treatment in each field. Infliximab, a TNF-targeting drug, was initially indicated for rheumatoid arthritis, but this indication has been expanded to the treatment of more than 10 immune diseases, such as Behcet’s disease, Kawasaki disease, psoriasis, psoriatic arthritis, ankylosing spondylitis, Crohn’s disease, and ulcerative colitis. Similar trends have been observed for other TNF-targeting drugs such as adalimumab. Treatment with TNF-targeting drugs prevented the loss of vision due to uveitis in the majority of patients with Behcet’s disease and dramatically reduced the development of fistula formation in Crohn’s disease patients with inflammatory bowel disease. Furthermore, tocilizumab, an IL-6-targeting drug, was found to exert marked effects on juvenile idiopathic arthritis. Its indication has been expanded to include Castleman’s disease and cytokine-release syndrome associated with chimeric antigen receptor T cell therapy, in addition to adult-onset Still’s disease, Takayasu’s arteritis, and giant cell arteritis. Tocilizumab has also been determined to be very effective for all of these conditions.

As various molecular-targeted drugs are used for many autoimmune diseases, it is necessary to develop new therapeutic strategies involving the differential use of drugs. This is especially important for highly diverse autoimmune diseases. Although biological drugs targeting TNF, IL-17, and IL-12/IL-23 have been approved for the treatment of psoriatic arthritis with destructive spondyloarthritis, there is no way to differentiate the use of these drugs. In our department, 8-color flow cytometry is performed to analyze the phenotypes of peripheral blood lymphocytes of patients with psoriatic arthritis registered in the FLOW registry [ 18 , 19 ]. The patients were classified into four groups based on the expression of chemokine receptors: helper T cell (Th) 17-dominant, Th1-dominant, hybrid, and normal type. Patients with Th17-dominant type were treated with an IL-17 antibody, patients with Th1-dominant type were treated with a p40 antibody, and patients with hybrid or normal type were treated with TNF-targeting drugs. The proportion of patients without improvement was reduced to less than 10% in these patients, compared with that in patients who were conventionally treated with biological drugs. Thus, differential use of biological drugs was demonstrated to be highly effective. This result suggested that for diseases in which characteristic cytokines are involved in the pathology, the use of molecular-targeted drugs can be optimized according to the pathology by stratification based on lymphocyte analysis. In other words, the applicability of precision medicine was suggested. These findings are expected to contribute to the development of new therapeutic systems and strategies.

For the treatment of rheumatoid arthritis, safe and favorable maintenance therapy is required over a long period after remission induction with methotrexate and biological DMARDs. However, the burden of medical expenses and medical economic problems due to the long-term continuous use of drugs are urgent issues in Japan and overseas, and the safety of long-term inhibition of targets, such as TNF, is currently unknown. Dose reduction and extension of dosing intervals for biological DMARDs are associated with a lower incidence of relapse than drug withdrawal, but there is a concern that anti-drug antibodies are more likely to be produced in such situations. If biological DMARDs can be withdrawn, adverse events should be preventable. The RRR study and the HONOR study reported the possibility of withdrawal of biological DMARDs after the induction of remission in patients with rheumatoid arthritis [ 20 , 21 ].

At the 2016 International Round-table Conference, study results from Japan and overseas were reviewed, and a consensus was reached regarding the order of drug withdrawal. The drugs should be withdrawn in the following order: glucocorticoids, anti-inflammatory drugs, biological DMARDs, and finally synthetic DMARDs. In addition, four requirements were determined for the withdrawal of DMARDs: fulfillment of the standard remission criteria, maintenance of remission for at least 6 months, maintenance of treatment with the same drugs at the same doses for at least 6 months, and no use of glucocorticoids. Moreover, it was additionally stated that negativity for anti-CCP antibodies, deep remission, and the absence of ultrasound findings of synovitis are all associated with the possibility of remission after withdrawal of DMARDs [ 22 ]. This suggests that if remission can be maintained even after the withdrawal of biological DMARDs, drug-free remission can be subsequently achieved (Fig. 3 ) [ 23 , 24 ]. It is also suggested that if the pathological process is controlled, a cure can be achieved by resetting immune abnormalities while the causes remain in place. The establishment of a new therapeutic system involving a drug holiday is expected to contribute to a reduction in treatment costs and a resolution of medical economic problems.

figure 3

Strategies for the treatment of rheumatoid arthritis. Intensive treatment is required for inducing remission in rheumatoid arthritis, but subsequently maintaining remission with high adherence and safety is a prerequisite for the good long-term outcome. The de-escalation and drug holiday of the DMARDs is an extension of the maintained remission

Conclusions

Rheumatoid arthritis is an autoimmune inflammatory disease pathologically characterized primarily by synovitis. Joint destruction, which is associated with prolonged arthritis, progresses soon after the onset of disease. The deformation of affected joints is irreversible and causes physical dysfunction. Thus, proper diagnosis and treatment are required from the early stages. The classification criteria published by the ACR and EULAR in 2010, which define rheumatoid arthritis as arthritis that is persistent and can be destructive in the future, were formulated with the aim of differentiating it from other types of arthritis soon after onset and to allow for therapeutic interventions prior to joint destruction. For treatment, DMARDs are used to suppress immune abnormalities and control disease activity. DMARDs are classified into conventional synthetic DMARDs (e.g., methotrexate), targeted synthetic DMARDs (e.g., JAK inhibitors), and biologic DMARDs. Appropriate treatment with these drugs has allowed clinicians to aim for remission in rheumatoid arthritis patients. These drug classes have been demonstrated to prevent structural damage to the joints and to prevent the progression of physical dysfunction. The advent of molecular-targeted drugs, such as biological drugs and JAK inhibitors, has allowed for the use of targeted therapies based on pathological mechanisms and the management of autoimmune inflammatory diseases, which were previously considered to be intractable. This can be regarded as revolutionary progress. In the future, safer and more effective treatments, therapeutic strategies aiming at cure, and the introduction of precision medicine are expected. Translation research aimed at developing new therapies and preventative measures may provide motivation for young clinicians and researchers.

Availability of data and materials

Not applicable.

Abbreviations

American College of Rheumatology

Clinical Disease Activity Index

C-reactive protein

Computed tomography

Disease-modifying antirheumatic drugs

European League Against Rheumatism

Janus kinase

Matrix metalloproteinase

Simplified Disease Activity Index

Tumor necrosis factor

Smolen JS, Aletaha D, Barton A, et al. Rheumatoid arthritis. Nat Rev Dis Primers. 2018 8;4:18001.

Smolen JS, Aletaha D, McInnes IB. Rheumatoid arthritis. Lancet. 2016;388:2023–38.

Article   CAS   Google Scholar  

McInnes IB, Schett G. Pathogenetic insights from the treatment of rheumatoid arthritis. Lancet. 2017;389:2328–37.

Tanaka Y, Ohira T. Mechanisms and therapeutic targets for bone damage in rheumatoid arthritis, in particular the RANK-RANKL system. Curr Opin Pharmacol. 2018;40:110–9.

Tanaka Y. Clinical immunity in bone and joint. J Bone Miner Metabolism. 2019;37:2–8.

Aletaha D, Neogi T, et al. 2010 rheumatoid arthritis classification criteria. Arthritis Rheum. 2010;62:2569–81.

Article   Google Scholar  

Felson DT, Smolen JS, Wells G, et al. American College of Rheumatology/European league against rheumatism provisional definition of remission in rheumatoid arthritis for clinical trials. Ann Rheum Dis. 2011;70:404–13.

Smolen JS, Landewé R, et al: EULAR recommendations for the management of rheumatoid arthritis with synthetic and biological disease-modifying antirheumatic drugs: 2019 update. Ann Rheum Dis (in press).

Smolen JS, Breedveld FC, Burmester GR, et al. Treating rheumatoid arthritis to target: 2014 update of the recommendations of an international task force. Ann Rheum Dis. 2016;75:3–15.

Burmester GR, Feist E, et al. Emerging cell and cytokine targets in rheumatoid arthritis. Nat Rev Rheumatol. 2014;10:77–88.

Tanaka Y. The JAK inhibitors: do they bring a paradigm shift for the management of rheumatic diseases? Rheumatology. 2019;58:i1–3.

Tanaka Y, Maeshima Y, Yamaoaka K. In vitro and in vivo analysis of a Jak inhibitor in rheumatoid arthritis. Ann Rheum Dis. 2012;71:i70–4.

O’Shea JJ, Kontzias A, Yamaoka K, Tanaka Y, Laurence A. Janus kinase inhibitors in autoimmune diseases. Ann Rheum Dis. 2013;72:ii111–5.

Taylor PC, Keystone EC, van der Heijde D, Weinblatt ME, del Camen ML, Gonzaga JR, et al. Baricitinib versus placebo or adalimumab in rheumatoid arthritis. New Engl J Med. 2017;376:652–62.

Smolen JS, Pangan AL, Emery P, Eigby W, Tanaka Y, Vargas JI, et al. Upadacitinib as monotherapy in patients with active rheumatoid arthritis and inadequate response to methotrexate (SELECT-MONOTHERAPY): a randomised, placebo-controlled, double-blind phase 3 study. Lancet. 2019;393:2303–11.

Takeuchi T, Tatsuki Y, Nogami Y, et al. Postmarketing surveillance of the safety profile of infliximab in 5000 Japanese patients with rheumatoid arthritis. Ann Rheum Dis. 2008;67:189–94.

Koike T, Harigai M, Inokuma S, et al. Postmarketing surveillance of safety and effectiveness of etanercept in Japanese patients with rheumatoid arthritis. Mod Rheumatol. 2011;21:343–51.

Tanaka Y, Kubo S, Miyagawa I, Iwata S, Nakayamada S. Lymphocyte phenotype and its application to precision medicine in systemic autoimmune diseases. Semin Arthritis Rheum. 2019;48:1146–50.

Miyagawa I, Nakayamada S, Nakano K, Kubo S, Iwata S, Miyazaki Y, et al. Precision medicine using different biological DMARDs based on characteristic phenotypes of peripheral T helper cells in psoriatic arthritis. Rheumatology. 2019;58:336–44.

Tanaka Y, Takeuchi T, Mimori T, Saito K, Nawata M, Kameda H, et al. Discontinuation of infliximab after attaining low disease activity in patients with rheumatoid arthritis, RRR (remission induction by remicade in RA) study. Ann Rheum Dis. 2010;69:1286–91.

Tanaka Y, Hirata S, Kubo S, Fukuyo S, Hanami K, Sawamukai N, et al. Discontinuation of adalimumab after achieving remission in patients with established rheumatoid arthiritis: 1-year outcome of the HONOR study. Ann Rheum Dis. 2015;74:389–95.

Schett G, Emery P, Tanaka Y, et al. Tapering DMARD therapy in rheumatoid arthritis – current evidence and future directions. Ann Rheum Dis. 2016;75:1428–37.

Tanaka Y. DMARD de-escalation — let the patient guide you. Nat Rev Rheumatol. 2017;13:637–8.

Tanaka Y. Intensive treatment and treatment holiday of TNF-inhibitors in rheumatoid arthritis. Curr Opin Rheumatol. 2012;24:319–26.

Download references

Acknowledgements

The author thanks all medical staff at all participating institutions for providing the data, especially Ms. Hiroko Yoshida, Ms. Youko Saitou, and Ms. Ayumi Maruyama for the excellent data management in the FIRST registry. I also thank Dr. Kazuyoshi Saito at Tobata General Hospital, Dr. Kentaro Hanami and Dr. Shunsuke Fukuyo at Wakamatsu Hospital of the University of Occupational and Environmental Health, Dr. Keisuke Nakatsuka at Fukuoka Yutaka Hospital, and all staff members at Kitakyushu General Hospital and Shimonoseki Saiseikai Hospital for their engagement in data collection of the FIRST registry.

This paper is not funded.

Author information

Authors and affiliations.

The First Department of Internal Medicine, School of Medicine, University of Occupational and Environmental Health, Japan, Kitakyushu, 807-8555, Japan

Yoshiya Tanaka

You can also search for this author in PubMed   Google Scholar

Contributions

The author drafted, read, revised, and approved the final manuscript.

Corresponding author

Correspondence to Yoshiya Tanaka .

Ethics declarations

Ethics approval and consent to participate, consent for publication, competing interests.

Y. Tanaka has received consulting fees, speaking fees, and/or honoraria from Daiichi-Sankyo, Eli Lilly, Novartis, YL Biologics, Bristol-Myers, Eisai, Chugai, Abbvie, Astellas, Pfizer, Sanofi, Asahi-kasei, GSK, Mitsubishi-Tanabe, Gilead, and Janssen and has received research grants from Abbvie, Mitsubishi-Tanabe, Chugai, Asahi-Kasei, Eisai, Takeda, and Daiichi-Sankyo.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Tanaka, Y. Rheumatoid arthritis. Inflamm Regener 40 , 20 (2020). https://doi.org/10.1186/s41232-020-00133-8

Download citation

Received : 19 May 2020

Accepted : 01 July 2020

Published : 07 September 2020

DOI : https://doi.org/10.1186/s41232-020-00133-8

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Inflammation and Regeneration

ISSN: 1880-8190

  • Submission enquiries: Access here and click Contact Us
  • General enquiries: [email protected]

research paper on rheumatoid arthritis

  • Research article
  • Open access
  • Published: 18 January 2021

Serum biomarker panel for the diagnosis of rheumatoid arthritis

  • Sora Mun 1   na1 ,
  • Jiyeong Lee 2   na1 ,
  • Mira Park 3 ,
  • Jieun Shin 4 ,
  • Mi-Kyoung Lim 5 &
  • Hee-Gyoo Kang   ORCID: orcid.org/0000-0001-8690-2483 1  

Arthritis Research & Therapy volume  23 , Article number:  31 ( 2021 ) Cite this article

12k Accesses

31 Citations

17 Altmetric

Metrics details

Rheumatoid arthritis (RA) is an autoimmune disease of inflammatory joint damage, wherein C-reactive protein and autoantibodies including rheumatoid factor (RF) and anti-cyclic citrullinated peptide (anti-CCP) are rapidly elevated. These serological factors are diagnostic markers of RA; however, their sensitivity and specificity for prediction warrant improvement for an early and accurate diagnosis.

We aimed to identify alternative biomarkers by serum protein profiling using LC-MS/MS. We performed statistical and functional analysis of differentially expressed proteins to identify biomarker candidates complementing conventional serological tests.

Seven biomarker candidates were verified through multiple reaction monitoring-based quantitative analysis, of which angiotensinogen (AGT), serum amyloid A-4 protein (SAA4), vitamin D-binding protein (VDBP), and retinol-binding protein-4 (RBP4) had an area under the curve over 0.8, thus distinguishing RA patients, including seronegative (RF- and anti-CCP-negative) RA patients, from healthy controls.

Conclusions

Therefore, among seronegative RA patients, a four-biomarker panel (AGT, SAA4, VDBP, and RBP4) can prevent false negatives and help diagnose RA accurately.

Rheumatoid arthritis (RA) is an autoimmune disease [ 1 ] with an unknown etiology. However, genetic factors account for 60% of the RA risk factors. Initially, these genetic factors include gene polymorphisms, epigenetic factors including DNA methylation and histone acetylation, and complex factors [ 2 , 3 , 4 ]. Other risk factors include environmental factors such as smoking, oral health, and diet [ 5 ]. The disease is initially characterized by an inflammatory response, followed by autoantibody activation and damage to the synovial membrane and joints. Activation of inflammation increases cytokine, chemokine, and inflammatory reactants such as C-reactive protein (CRP) [ 6 ]. Furthermore, a series of immune responses are triggered with an increase in inflammation. Hence, autoantibodies are overproduced, leading to an increase in immunoglobulin M (RF) and anti-CCP in RA patients [ 7 , 8 ]. When serum peptides are citrullinated or subjected to other posttranslational modifications by various environmental stimuli, the altered peptides are presented to immune cells, including T cells as antigens, and antibodies such as anti-CCP are produced [ 5 , 7 , 8 , 9 , 10 ]. Therefore, CRP, RF, and anti-CCP, representing the inflammatory and immune response of RA, are diagnostic blood biomarkers [ 11 , 12 ].

However, existing biomarkers have limitations concerning RA diagnosis. For example, the sensitivity and specificity of RF are 60–90% and 85%, respectively. To improve the efficiency of RA diagnosis, anti-CCP is used with RF. Anti-CCP has higher ACCP positivity than RF positivity among RA patients; however, the sensitivity and specificity of the two markers do not significantly differ [ 13 ]. Therefore, novel diagnostic biomarkers complementing the existing biomarkers, i.e., RF and anti-CCP, are required. If these new biomarkers could diagnose seronegative (RF- and anti-CCP-negative) RA patients, they could contribute to the accurate diagnosis and treatment of RA [ 14 , 15 ]. Furthermore, current diagnostic biomarkers reflect the status of inflammation and immunity among RA patients. New protein biomarkers potentially detected through serum protein profiling are expected to represent various physiological changes in RA, other than inflammation and immunity.

In most previous proteomics studies, blood samples were pooled for MS analysis [ 16 ]. However, it is important to analyze individual serum samples to reflect individual alterations in serum protein levels [ 12 , 17 ], but it is difficult to analyze individual serum samples. First, it is difficult to obtain an adequate volume of individual serum samples. Second, individual MS analysis is costly and time-consuming. Third, clinical data interpretation is challenging owing to the complexity of the status of RA patients. Finally, it is difficult to control the data processes. Nevertheless, when some patients have a high abundance of certain proteins, these protein expression patterns seem to represent all patients. However, this discrepancy can be eliminated through individual sample analysis. Furthermore, MS analysis of individual serum samples is important for biomarker discovery because it helps validate the pattern of differential protein expression among all RA patients. Moreover, serum samples can be classified and validated under various clinical parameters [ 18 ].

In a recent study, individual samples were classified under clinical parameters to identify diagnostic biomarkers among seronegative (RF- and anti-CCP-negative) RA patients. In addition, individual sample analysis facilitates the separate analysis of the mild-to-moderate and advanced severe cases. As it is important to assess the clinical course of RA patients, analyses based on the disease status of patients are possible through the analysis of individual samples, thus facilitating the classification of patients under the clinical differences and the concomitant identification of novel biomarkers for accurate RA diagnosis and treatment.

Therefore, in this study, we attempted to develop a biomarker panel to distinguish seronegative (RF- and anti-CCP-negative) RA patients by analyzing the serum proteins relative to those of healthy controls. The experimental cohort was further divided into a discovery and validation cohort. Diagnostic biomarkers were selected from among 50 RA patients and 43 healthy controls in the discovery cohort and from among 251 healthy controls and 230 RA patients in the validation cohort (Table  1 ).

Experimental design and statistical rationale

Serum samples of 251 RA patients and 230 healthy controls for biomarker identification were collected from the Eulji University Hospital Institutional Review Board (EMC 2016-03-019, 31 March 2016). Written informed consent was obtained from all subjects. Participants who were diagnosed by rheumatologists fulfilling the ACR were recruited as the patient group, with no restrictions on gender and age. Healthy controls with previous or current disease history (rheumatoid arthritis, myocardial infarction (MI), angina, stroke, high blood pressure, depression, and/or diabetes mellitus) were excluded for recruitment. Blood was collected in an anticoagulant-free vacutainer. After 2 h at 24 °C, blood samples were centrifuged at 4000× g for 5 min to separate the serum. Highly abundant serum proteins including albumin, IgG, antitrypsin, IgA, transferrin, and haptoglobin were depleted using a multiple affinity removal system comprising an LC column (human 6-HC, 4.6 × 50 mm; Agilent Technologies, Santa Clara, CA, USA), as described [ 12 ]. The eluted sample containing low-abundance proteins was concentrated using a Nanosep device with a modified polyethersulfone membrane 3 K (Pall, Zaventem, Belgium) and analyzed using a mass spectrometer(AB Sciex 5600, Framingham, MA, USA) to select significant candidate biomarkers. Candidate biomarkers were validated using multiple reaction monitoring (MRM)-based targeted protein quantification.

Statistical analysis

To select candidate biomarkers, a corrected p value, obtained from the Benjamini-Hochberg test, was used, and differentially expressed proteins with a p value < 0.05 were used for further analysis. We performed unpaired t tests with Welch’s correction using the GraphPad Prism version 8.0 for Windows (GraphPad Software Inc., San Diego, CA) to assess the results of the MRM-based quantification analysis between healthy controls and RA patients; differences with a p value < 0.001 were significant. For predicting the classification accuracy of biomarkers, logistic regression analysis was performed using the SPSS software package version 18.0.0 (SPSS Inc., Chicago, IL, USA).

Determination of protein concentration and tryptic digestion

To determine the serum protein levels for MS analysis, a Bradford assay (Bio-Rad, Hercules, CA, USA) was performed according to the manufacturer’s instructions. Samples containing 100 μg serum proteins were reduced via treatment with 5 mM Tris (2-carboxyethyl) phosphine (Pierce Chemical Company, Rockford, IL, USA) at 37 °C, 300 rpm, for 30 min, followed by treatment with 15 mM iodoacetamide (Sigma-Aldrich, St. Louis, MO, USA) for alkylation at 24 °C, 300 rpm, for 1 h in the dark. Serum proteins were cleaved into peptides, using mass spectrometry-grade trypsin gold (Promega Corporation, Fitchburg, WI, USA) at 37 °C overnight. The cleavage products were cleaned using a C18 cartridge (Waters Corporation, Milford, MA, USA).

OFFGEL fractionation and LC-MS/MS analysis

The serum proteins in each sample were separated into 12 fractions through pH 3–10 isoelectric points, using the OFFGEL fractionator (3100 OFFGEL Low Res Kit, pH 3–10; Agilent Technologies, Santa Clara, CA, USA) according to the manufacturer’s instructions. Twelve fractions were loaded onto an Eksigent nanoLC 400 system and the cHiPLC® (AB Sciex, Concord, ON, Canada) and analyzed, and the proteins were identified using a TripleTOF 5600 mass spectrometer (AB Sciex). Thereafter, for relative analysis, SWATH acquisition was conducted. In each run, 100 μg/mL of samples was injected onto an Eksigent ChromXP nanoLC trap column (350 μm i.d. × 0.5 mm, ChromXP C18 3 μm) at a flow rate of 5000 nL/min. Samples were eluted from the Eksigent ChromXP nanoLC column (75 μm i.d. × 15 cm) at a flow rate of 300 nL/min for 120 min, and mobile phase B buffer was added gradually into the column (5–90%) over a 120-min total run time. The gradient of mobile phase B buffer was (time and % B) 0 min/mobile phase B 5%, 10.5 min/40%, 105.5 min/90%, 111.5 min/90%, 112 min/5%, and 120 min/5%. Mobile phase B and A buffer, and the search parameters are as described [ 12 ].

Synthesis and purification of label-free standard peptides

Seven candidate proteins were determined as putative diagnostic biomarkers. Peptides for absolute quantification through MRM analysis were selected and synthesized using Peptron (Daejeon, South Korea). These criteria were set for peptide selection: (1) peptides without miscleaved sites, (2) unmodified peptides, (3) peptides not comprising Met, (4) peptides with 7–15 residues, and (6) peptides with a low false discovery rate (FDR) (usually zero). After prototypic tryptic peptide standards were synthesized, two-fold serial dilutions were conducted using 1 mM/μL stock peptide standards in 0.1% formic acid or DMSO, following the manufacturer’s protocol.

Label-free quantification through MRM analysis

Skyline was used to determine MRM Q1/Q3 ion pairs from selected peptides, as described (Mun et al. [ 12 ]). Voltage parameters including collision energy (CE), declustering potential (DP), and cell exit potential (CXP) were determined through compound optimization for each transition. AB Sciex Exion LC was used to segregate the samples using ACQUITY UPLC BEH C18 Column (130 Å, 1.7 μm, 2.1 mm × 150 mm) with an ACQUITY UPLC BEH C18 VanGuard Pre-column (130 Å, 1.7 μm, 2.1 mm × 5 mm). Samples of healthy controls and RA patients were analyzed using AB Sciex QTRAP5500. Each sample was loaded onto the LC column with a gradient of 5–90% mobile phase B for a total run time of 30 min. The mobile phase B buffer was gradually introduced in the LC column: (time/% B) 1 min/mobile phase B 5%, 50 min/40%, 21–25 min/90%, and 25.5–30 min/5%. Mobile phase B comprised 0.1% formic acid in HPLC-grade acetonitrile, and mobile phase A comprised 0.1% formic acid in HPLC-grade water. The source parameters for MRM analysis were curtain gas, 206.84 kPa; low collision gas; ion spray voltage, 5500 V; temperature, 400 °C; ion source gas 1, 275.79 kPa; ion source gas 2, 413.69 kPa.

Through a qualitative analysis, performed using SCIEX 5600QTOF, 194 and 111 proteins were uniquely identified from among healthy controls and RA patients, respectively, and 339 proteins were identified in both healthy controls and RA patients. Proteins identified through the IDA method between the two groups were quantified using SWATH acquisition. Principal component analysis (PCA) was performed using the quantification data of each sample. Consequently, healthy controls were distinguished from RA patients (Fig.  1 a).

figure 1

Protein quantification through SWATH acquisition and principal component analysis for group clustering. a Venn diagram of the identified proteins among healthy controls and rheumatoid arthritis (RA) patients. b Clustering analysis of > 2-fold differentially expressed proteins filtered by the p value ( p  < 0.05) on partial least squares discriminant analysis. Differentially expressed proteins by > 2-fold filtered by the p value ( p  < 0.05) through the line plot and volcano plot analysis. c Heatmap analysis for healthy controls and RA patients

First, as indicated in the volcano plot, we selected significantly upregulated or downregulated proteins by > 2-fold ( p  < 0.05) (Fig.  1 b). Heatmap analysis revealed differentially expressed proteins in the two groups (Fig.  1 c). Furthermore, we performed Gene Ontology (GO) analysis of proteins with FC > 2.0 and p  < 0.05 between healthy controls and RA patients. The three most significantly enriched pathways were complement-related pathways associated with immune responses, including the lectin-induced complement pathway, the classical complement pathway, and the alternative complement pathway (Fig.  2 a). Moreover, three pathways were significantly associated with immunity and inflammation, including the complement system, phagosome involvement in antigen presentation, and phagocytosis (Fig.  2 b). On GO analysis of biological processes, the three most significant pathways were antigen processing and presentation of exogenous peptide antigens via MHC class I, antigen processing and presentation of peptide antigens via MHC class Ib, and antigen processing and presentation of endogenous peptide antigens (Fig.  2 c).

figure 2

Process network analysis for differentially expressed proteins between healthy controls and rheumatoid arthritis (RA) patients. a Pathway map, process networks, and Gene Ontology processes associated with differentially expressed proteins between healthy controls and RA patients. b The most significant process networks between healthy controls and RA patients. The process network with the lowest p value was the complement system. c The most significant pathway map between healthy controls and RA patients. The pathway map with the lowest p value was the lectin-induced complement pathway

On protein quantification of individual serum samples, we selected seven candidate biomarkers, which were then subjected to MRM absolute quantification, including angiotensinogen, C-reactive protein, gelsolin, lymphatic vessel endothelial hyaluronan receptor 1, retinol-binding protein 4, serum amyloid A-4, and vitamin D-binding protein (VDBP) (Fig.  3 ). For MRM analysis of the seven selected candidate biomarkers, one peptide was selected per protein, and after optimization, the parameters were selected.

figure 3

Chromatography of selected candidate proteins extracted from RA patients. a – g Extract ion chromatography of the peptides from seven candidate proteins including angiotensinogen, C-reactive protein, gelsolin, lymphatic vessel endothelial hyaluronan receptor 1, retinol-binding protein 4, serum amyloid A-4, and vitamin D-binding protein

Of the seven proteins identified, those with an area under the curve (AUC) > 0.8 were angiotensinogen (AUC = 0.8346), serum amyloid A-4 (AUC = 0.8994), VDBP (AUC = 0.8170), and retinol-binding protein 4 (AUC = 0.9391) (Fig.  4 ), whereas lymphatic vessel endothelial hyaluronan receptor 1, gelsolin, and C-reactive protein revealed AUC values of 0.5309, 0.6794, and 0.5030, respectively.

figure 4

Box and whisker plots of selected biomarker candidates in healthy controls and rheumatoid arthritis (RA) patients. Proteins altered among RA patients relative to the healthy controls were selected. a – d Angiotensinogen, serum amyloid A-4 protein (SAA4), retinol-binding protein-4 (RBP4), and vitamin D-binding protein (VDBP) profiles were compared between healthy controls and RA patients. The number of healthy controls and RA patients was 251 and 230, respectively. Box plots represent the upper quartile, lower quartile, and median (horizontal line). Whiskers enclose the range (min-max value). Independent samples t tests were used to determine the statistical significance ** p  < 0.001. AUC, area under the curve

Furthermore, RA patients were categorized as RF-positive, RF-negative, ACCP-positive, and ACCP-negative. These four biomarker candidates displayed high classification accuracy regardless of the RF-positive or RF-negative status of patients (Fig.  5 ). Furthermore, logistic regression analysis was performed to predict the classification accuracy among healthy controls and RA patients. Consequently, angiotensinogen (AGT) accurately classified 209 healthy controls and 169 RA patients in predicted classes from among 250 healthy controls and 230 RA patients in actual classes. The classification accuracy for healthy controls and RA patients was 83.3% and 73.5%, respectively (Fig.  6 a). Furthermore, serum amyloid A-4 (SAA4) accurately classified 223 healthy controls and 176 RA patients in predicted classes, with a classification accuracy of 88.8% and 76.5% for healthy controls and RA patients, respectively (Fig.  6 a). Retinol-binding protein-4 (RBP4) accurately classified 204 healthy controls and 228 RA patients in predicted classes, with a classification accuracy of 90.8% and 86.0% for healthy controls and RA patients, respectively (Fig.  6 a). Vitamin D-binding protein (VDBP) accurately classified 228 healthy controls and 197 RA patients in predicted classes, with a classification accuracy of 90.8% and 86.0% for healthy controls and RA patients, respectively (Fig.  6 a). Together, the four-biomarker panel accurately classified 234 healthy controls and 210 RA patients with a classification accuracy of 93.2% and 91.7% for healthy controls and RA patients, respectively (Fig.  6 b). The AUC values of the four individual biomarkers were 0.8346, 0.8890, 0.8170, and 0.9430 (Fig.  6 c), and of the four-biomarker panel, 0.9740 (Fig.  6 d).

figure 5

Scatter plots of selected biomarker candidates in healthy controls and RF- and anti-CCP-positive/negative RA. Proteins altered among rheumatoid (RF)- and anti-CCP (ACCP)-positive or -negative RA patients relative to those in the healthy controls were selected. a – d Angiotensinogen, retinol-binding protein-4 (RBP4), serum amyloid A-4 protein (SAA4), and vitamin D-binding protein (VDBP) profiles were compared between healthy controls and RA patients in accordance with the antibody titer. The number of healthy controls and RA patients was 251 and 230 (RF+/ACPA+ n  = 121, RF+/ACPA− n  = 47, RF−/ACPA+ n  = 33, RF−/ACPA− n  = 29), respectively. Plots indicate individual protein abundance in each group. Data are presented as mean ± SD values. Independent samples t tests were used to determine the statistical significance ** p  < 0.001. HC, healthy controls; RA, patients with rheumatoid arthritis

figure 6

Logistic regression analysis for selected biomarker candidates in healthy controls and rheumatoid arthritis (RA) patients. Predictive accuracy of single ( a ) and the four biomarker candidates ( b ). Receiver operating characteristic curve (ROC) analysis of single ( c ) and four biomarker candidates ( d ) was performed. The number of healthy controls and RA patients was 251 and 230, respectively. The plots indicate individual protein abundance in each group. Data are presented as mean ± SEM values. Independent samples t tests were used to determine the statistical significance

In this study, we analyzed the serum proteins in 251 healthy controls and 230 patients with RA to identify diagnostic biomarkers through MS analysis. Analysis of individual serum samples revealed differentially expressed proteins in the two groups, among which, AGT, RBP4, SAA4, and VDBP emerged as novel diagnostic biomarkers on MRM absolute quantification, and their AUC value was over 0.8, indicating a high diagnostic efficiency.

We analyzed proteins significantly upregulated by > 2-fold ( p  < 0.05). A pathway map of the functional analysis revealed that the three main complement pathways were associated with these differentially expressed proteins. Moreover, the complement system was the most significantly associated pathway with RA upon GO analysis of biological processes. In the lectin-induced complement pathway, complement proteins including C3, C3a, C3b, C4, C4a, C4b, C5, C5a, C5b, complement C3d receptor 2 (CD21), ficolin 3, and iC3b were downregulated by > 8-fold among RA patients. Likewise, expression patterns of the proteins involved in the classical and lectin-induced complement pathways were similar in patients with RA. However, the C1s complement protein was upregulated > 9-fold in RA patients in the classical pathway and vitronectin and by > 8-fold in the alternative pathway. Complement C1s protein is an early component of the classical pathway and initiates the complement pathway and is reportedly associated with the degeneration of articular cartilage in RA. Meanwhile, inhibitory protein alpha 1-antitrypsin of the complement pathway inducing inflammation was upregulated in patients with RA [ 19 ]. Furthermore, alpha 1-antitrypsin inhibits thrombin activity and blood coagulation as well as inhibition of inflammation by complement pathway control, suggesting that aberrant blood coagulation initiated in RA can be attenuated through alpha 1-antitrypsin overexpression.

VDBP is mainly produced in the liver. When tissue damage occurred, increased permeability of cells releases polymerization of F-actin, leading to a blocked blood vessel in RA. Along with tissue damage, VDBP was immediately released from the damaged cell [ 20 ]. Increased serum VDBP is suggested to play a role in scavenging actin and inhibiting the negative effects of F-actin. Besides, the VDBP-G-actin complex was involved in neutrophil migration, suggesting VDBP overexpression might allow proteins to act immediately and directly during RA-induced tissue damage [ 20 ]. However, vitamin D activated by VDBP protects against joint tissue damage during RA, owing to its anti-inflammatory effects [ 20 ]. 25-Hydroxyl vitamin D is activated by VDBP and moves to immune cells of several organs, inducing an anti-inflammatory effect [ 21 ]. RA is three times more prevalent in women than in men. The well-established evidence on the prevalence in women is the association between the female sex hormone such as estrogen and RA. The increase in estrogen has been reported to alleviate the onset of RA. Interestingly, VDBP is upregulated by the increase in estrogen, thus playing an important role in the anti-inflammatory activity and tissue recovery [ 12 , 22 ]. Likewise, the role of VDBP in the pathogenesis of RA can be interpreted from both anti- and pro-inflammation.

SAA4 is an acute-phase protein reportedly upregulated in RA [ 23 ]. SAA4 is activated by cytokines such as IL-1, IL-6, and TNF-alpha and has pro-inflammatory effects. Moreover, SAA4 positively correlated with C-reactive protein [ 24 ]. Furthermore, SAA4 had a better efficacy for diagnosis than C-reactive protein being used for the diagnosis of RA [ 24 ]. In this study, SAA4 was also identified as a candidate biomarker for the diagnosis of RA, and the results of MRM absolute quantification were concurrent with those from previous studies.

The renin-angiotensin system is associated with the inflammatory response and helps maintain blood pressure [ 25 ]. Angiotensin II mediates inflammation by stimulating immune cells [ 26 ]. For example, angiotensin II regulates pro-inflammatory transcription factor nuclear factor-κB [ 27 ]. AGT is an angiotensin II precursor [ 28 ]. This study revealed the effects of the renin-angiotensin system on inflammatory reactions. An increase in angiotensinogen in serum samples of RA patients is associated with the renin-angiotensin system comprising AGT and angiotensin II [ 15 ].

RBP4s, called retinol-binding protein (RBP), are transport proteins for retinol (vitamin A 1 ). Retinol is synthesized in the liver and circulated into the blood by RBP [ 29 ]. RBP is associated with insulin resistance, obesity, and cardiovascular disease [ 30 , 31 ]. Previous studies have reported that RBP4 is upregulated in insulin-resistant mice and is upregulated in the serum of patients with obesity or type 2 diabetes, thus affecting insulin signaling [ 32 , 33 , 34 ]. Likewise, RBP4 was reported as the predictor of atherosclerosis in patients with RA [ 35 ]. However, the association between elevated RBP and RA pathology has not been defined. Although, in a previous study on RA biomarkers, it has been reported that RBP4 is a candidate RA biomarker through ELISA [ 22 ].

In correlation analysis between the four biomarkers and conventional blood biomarker, there is no correlation between either the candidate proteins or autoantibody/inflammation markers. As mentioned above, candidate biomarkers are likely to be related with inflammation and dysfunctional immune systems. However, in our study, there is no mechanistic evidence to prove the association between the biomarkers and RA pathology. Thus, further investigation is needed to clarify the mechanism related with the biomarkers for clinical application.

The candidate proteins have a distinction between comparative groups; however, among the four candidate biomarkers, VDBP have a distinction in individual patients and the difference in the mean between the comparison groups. The range of the mean ± SD of the patient and control groups does not overlap. However, individual patients who are not present within ± SD may be confused with the normal group, leading to a false negative. Therefore, to better differentiate RA patients from healthy controls, in further studies, it is necessary to analyze the classified patient group from various clinical perspectives. For example, disease stage reflecting inflammation response, smoking, and estrogen concentration may affect the expression of VDBP [ 36 ]. This classified sample analysis is expected to enable personalized diagnosis and optimal treatment as well as improve diagnostic efficacy.

This study shows that four proteins validated through MRM were analyzed among RF-positive, RF-negative, ACCP-positive, and ACCP-negative RA patients to confirm their potential to distinguish RA patients from healthy controls regardless of the titer of RF and ACCP. RF is an existing RA diagnostic marker; however, it has limitations associated with RA diagnosis, including a low sensitivity of 60% and a specificity of 85%. Furthermore, RF has been detected in non-RA diseases, thus deterring an accurate diagnosis of RA. Therefore, to increase the diagnostic efficiency of RA, anti-CCP is used; however, anti-CCP has a similar or higher specificity and sensitivity than RF. Hence, we identified four candidate biomarkers including angiotensinogen, SAA4, RBP4, and VDBP, which could significantly distinguish RF-positive, RF-negative, ACCP-positive, and ACCP-negative RA patients, and particularly the seronegative (RF- and ACCP-negative) patients. Therefore, a combination of these four markers can diagnose RA with greater accuracy, serving as highly robust biomarkers along with RF and ACCP.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Abbreviations

  • Rheumatoid arthritis

Rheumatoid factor

Anti-cyclic citrullinated peptide

C-reactive protein

Multiple-reaction-monitoring-based

  • Retinol-binding protein-4
  • Serum amyloid A-4 protein
  • Vitamin D-binding protein

Lymphatic vessel endothelial hyaluronan receptor 1

Collision energy

Declustering potential

Cell exit potential

Principal component analysis

Gene Ontology

Complement C3d receptor 2

Area under the curve

Cylwik B, Chrostek L, Gindzienska-Sieskiewicz E, Sierakowski S, Szmitkowski M. Relationship between serum acute-phase proteins and high disease activity in patients with rheumatoid arthritis. Adv Med Sci. 2010;55(1):80–5.

Article   CAS   Google Scholar  

MacGregor AJ, Snieder H, Rigby AS, Koskenvuo M, Kaprio J, Aho K, Silman AJ. Characterizing the quantitative genetic contribution to rheumatoid arthritis using data from twins. Arthritis Rheum. 2000;43(1):30–7.

Gomez-Cabrero D, Almgren M, Sjoholm LK, Hensvold AH, Ringh MV, Tryggvadottir R, Kere J, Scheynius A, Acevedo N, Reinius L, et al. High-specificity bioinformatics framework for epigenomic profiling of discordant twins reveals specific and shared markers for ACPA and ACPA-positive rheumatoid arthritis. Genome Med. 2016;8(1):124–38.

Article   Google Scholar  

Liu Y, Aryee MJ, Padyukov L, Fallin MD, Hesselberg E, Runarsson A, Reinius L, Acevedo N, Taub M, Ronninger M, et al. Epigenome-wide association data implicate DNA methylation as an intermediary of genetic risk in rheumatoid arthritis. Nat Biotechnol. 2013;31(2):142–7.

Smolen JS, Aletaha D, Barton A, Burmester GR, Emery P, Firestein GS, Kavanaugh A, McInnes IB, Solomon DH, Strand V, et al. Rheumatoid arthritis. Nat Rev Dis Primers. 2018;4(18001):1–23.

Google Scholar  

Choy EHS, Panayi GS. Cytokine pathways and joint inflammation in rheumatoid arthritis. N Engl J Med. 2001;344(12):907–16.

Holers VM. Autoimmunity to citrullinated proteins and the initiation of rheumatoid arthritis. Curr Opin Immunol. 2013;25(6):728–35.

Muller S, Radic M. Citrullinated autoantigens: from diagnostic markers to pathogenetic mechanisms. Clin Rev Allergy Immunol. 2015;49(2):232–9.

Konig MF, Abusleme L, Reinholdt J, Palmer RJ, Teles RP, Sampson K, Rosen A, Nigrovic PA, Sokolove J, Giles JT, et al. Aggregatibacter actinomycetemcomitans-induced hypercitrullination links periodontal infection to autoimmunity in rheumatoid arthritis. Sci Transl Med. 2016;8(369):1–17.

Dissick A, Redman RS, Jones M, Rangan BV, Reimold A, Griffiths GR, Mikuls TR, Amdur RL, Richards JS, Kerr GS. Association of periodontitis with rheumatoid arthritis: a pilot study. J Periodontol. 2010;81(2):223–30.

Aletaha D, Neogi T, Silman AJ, Funovits J, Felson DT, Bingham CO 3rd, Birnbaum NS, Burmester GR, Bykerk VP, Cohen MD, et al. 2010 Rheumatoid arthritis classification criteria: an American College of Rheumatology/European League Against Rheumatism collaborative initiative. Arthritis and rheumatism. 2010;62(9):2569–81.

Mun S, Lee J, Park A, Kim H-J, Lee Y-J, Son H, Shin M, Lim M-K, Kang H-G. Proteomics approach for the discovery of rheumatoid arthritis biomarkers using mass spectrometry. Int J Mol Sci. 2019;20(18):4368–81.

Ingegnoli F, Castelli R, Gualtierotti R. Rheumatoid factors: clinical applications. Dis Markers. 2013;35(6):727–34.

Kim D, Mun S, Lee J, Park A, Seok A, Chun YT, Kang HG. Proteomics analysis reveals differential pattern of widespread protein expression and novel role of histidine-rich glycoprotein and lipopolysaccharide-binding protein in rheumatoid arthritis. Int J Biol Macromol. 2018;109:704–10.

Lee J, Mun S, Kim D, Lee YR, Sheen DH, Ihm C, Lee SH, Kang HG. Proteomics analysis for verification of rheumatoid arthritis biomarker candidates using multiple reaction monitoring. Proteomics Clin Appl. 2019;13(3):e1800011.

Lee J, Mun S, Park A, Kim D, Lee YJ, Kim HJ, Choi H, Shin M, Lee SJ, Kim JG, et al. Proteomics reveals plasma biomarkers for ischemic stroke related to the coagulation cascade. J Mol Neurosci. 2020;70(9):1321–31.

Lee J, Joo EJ, Lim HJ, Park JM, Lee KY, Park A, Seok A, Lee H, Kang HG. Proteomic analysis of serum from patients with major depressive disorder to compare their depressive and remission statuses. Psychiatry Investig. 2015;12(2):249–59.

Lee J, Park A, Mun S, Kim HJ, Son H, Choi H, Kim D, Lee SJ, Kim JG, Kang HG. Proteomics-based identification of diagnostic biomarkers related to risk factors and pathogenesis of ischemic stroke. Diagnostics (Basel). 2020;10(5):340–53.

Brien ME, Fee L, Browne N, Carroll TP, Meleady P, Henry M, McQuillan K, Murphy MP, Logan M, McCarthy C, et al. Activation of complement component 3 is associated with airways disease and pulmonary emphysema in alpha-1 antitrypsin deficiency. Thorax. 2020;75(4):321–30.

Kew RR. The vitamin D binding protein and inflammatory injury: a mediator or sentinel of tissue damage? Front Endocrinol. 2019;10:470–80.

Baek JC, Jo JY, Lee SM, Cho IA, Shin JK, Lee SA, Lee JH, Cho MC, Choi WJ. Differences in 25-hydroxy vitamin D and vitamin D-binding protein concentrations according to the severity of endometriosis. Clin Exp Reprod Med. 2019;46(3):125–31.

Mun S, Lee J, Lim M-K, Lee Y-R, Ihm C, Lee SH, Kang H-G. Development of a novel diagnostic biomarker set for rheumatoid arthritis using a proteomics approach. Biomed Res Int. 2018;2018:1–11.

Targońska-Stępniak B, Majdan M. Serum amyloid A as a marker of persistent inflammation and an indicator of cardiovascular and renal involvement in patients with rheumatoid arthritis. Mediat Inflamm. 2014;2014:1–17.

Seok A, Lee HJ, Lee S, Lee J, Mun S, Park A, Chun YT, Lee JH, Lim HJ, Kang HG. Identification and validation of SAA4 as a rheumatoid arthritis prescreening marker by liquid chromatography tandem-mass spectrometry. Molecules (Basel). 2017;22(5):805–17.

Ranjbar R, Shafiee M, Hesari A, Ferns GA, Ghasemi F, Avan A. The potential therapeutic use of renin-angiotensin system inhibitors in the treatment of inflammatory diseases. J Cell Physiol. 2019;234(3):2277–95.

Satou R, Penrose H, Navar LG. Inflammation as a regulator of the renin-angiotensin system and blood pressure. Curr Hypertens Rep. 2018;20(12):100–14.

Ruiz-Ortega M, Lorenzo O, Rupérez M, König S, Wittig B, Egido J. Angiotensin II activates nuclear transcription factor κB through AT 1 and AT 2 in vascular smooth muscle cells: molecular mechanisms. Circ Res. 2000;86(12):1266–72.

Lu H, Cassis LA, Kooi CWV, Daugherty A. Structure and functions of angiotensinogen. Hypertens Res. 2016;39(7):492–500.

Shirakami Y, Lee S-A, Clugston RD, Blaner WS. Hepatic metabolism of retinoids and disease associations. Biochim Biophys Acta. 2012;1821(1):124–36.

Promintzer M, Krebs M, Todoric J, Luger A, Bischof MG, Nowotny P, Wagner O, Esterbauer H, Anderwald C. Insulin resistance is unrelated to circulating retinol binding protein and protein C inhibitor. J Clin Endocrinol Metab. 2007;92(11):4306–12.

Zabetian-Targhi F, Mahmoudi MJ, Rezaei N, Mahmoudi M. Retinol binding protein 4 in relation to diet, inflammation, immunity, and cardiovascular diseases. Advances in nutrition (Bethesda). 2015;6(6):748–62.

Norseen J, Hosooka T, Hammarstedt A, Yore MM, Kant S, Aryal P, Kiernan UA, Phillips DA, Maruyama H, Kraus BJ, et al. Retinol-binding protein 4 inhibits insulin signaling in adipocytes by inducing proinflammatory cytokines in macrophages through a c-Jun N-terminal kinase- and Toll-like receptor 4-dependent and retinol-independent mechanism. Mol Cell Biol. 2012;32(10):2010–9.

Wang Y, Sun L, Lin X, Yuan J-M, Koh W-P, Pan A. Retinol binding protein 4 and risk of type 2 diabetes in Singapore Chinese men and women: a nested case-control study. Nutr Metab. 2019;16(1):3–13.

Takebayashi K, Suetsugu M, Wakabayashi S, Aso Y, Inukai T. Retinol binding protein-4 levels and clinical features of type 2 diabetes patients. J Clin Endocrinol Metab. 2007;92(7):2712–9.

Fath Allah R, El-Refaei K, Osman E, Sabry S, Elbagoury I. Retinol-binding protein 4 as a biomarker of cardiometabolic risk in rheumatoid arthritis. Sci J Al-Azhar Med Fac Girls. 2020;4(2):106–12.

Bouillon R, Schuit F, Antonio L, Rastinejad F. Vitamin D binding protein: a historic overview. Front Endocrinol (Lausanne). 2020;10:1–21.

Download references

Acknowledgements

This research was supported by the Bio & Medical Technology Development Program of the NRF funded by the Korean government, MSIP (GrantNo.2016M3A9B694241).

Author information

Sora Mun and Jiyeong Lee contributed equally to this work.

Authors and Affiliations

Department of Biomedical Laboratory Science, College of Health Sciences, Eulji University, Seongnam, Republic of Korea

Sora Mun & Hee-Gyoo Kang

Department of Biomedical Laboratory Science, College of Health Sciences, Eulji University, Daejeon, Republic of Korea

Jiyeong Lee

Department of Preventive Medicine, School of Medicine, Eulji University, Daejeon, Republic of Korea

Liberal Arts, Woosuk University, Jeonju, Republic of Korea

Division of Rheumatology, Department of Medicine, Eulji University School of Medicine, Daejeon, Republic of Korea

Mi-Kyoung Lim

You can also search for this author in PubMed   Google Scholar

Contributions

S.M. and J.L. wrote the manuscript, analyzed the proteomic data, and created the figures and tables. M.P. and J.S. created the statistical data. M.-K.L. created the clinical information and data. H.-G.K. reviewed the manuscript and administered the project. All authors have read and approved the final manuscript.

Corresponding author

Correspondence to Hee-Gyoo Kang .

Ethics declarations

Ethics approval and consent to participate.

The study was approved by the Eulji University Hospital Institutional Review Board (EMC 2016-03-019, 31 March 2016). We obtained consent from all the individuals.

Consent for publication

We obtained consent for publication from all the individuals whose detailed information was included in this manuscript.

Competing interests

The authors declare no conflict of interest.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Mun, S., Lee, J., Park, M. et al. Serum biomarker panel for the diagnosis of rheumatoid arthritis. Arthritis Res Ther 23 , 31 (2021). https://doi.org/10.1186/s13075-020-02405-7

Download citation

Received : 30 October 2020

Accepted : 22 December 2020

Published : 18 January 2021

DOI : https://doi.org/10.1186/s13075-020-02405-7

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Angiotensinogen

Arthritis Research & Therapy

ISSN: 1478-6362

research paper on rheumatoid arthritis

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • View all journals
  • My Account Login
  • Explore content
  • About the journal
  • Publish with us
  • Sign up for alerts
  • Open access
  • Published: 12 February 2024

Digital health technologies and machine learning augment patient reported outcomes to remotely characterise rheumatoid arthritis

  • Andrew P. Creagh   ORCID: orcid.org/0000-0002-6086-6098 1 , 2 ,
  • Valentin Hamy 3 ,
  • Hang Yuan   ORCID: orcid.org/0000-0001-5944-1925 2 , 4 ,
  • Gert Mertes   ORCID: orcid.org/0000-0003-3155-480X 1 , 2 , 4 ,
  • Ryan Tomlinson 5 ,
  • Wen-Hung Chen 5 ,
  • Rachel Williams 5 ,
  • Christopher Llop 6 ,
  • Christopher Yee 6 ,
  • Mei Sheng Duh 6 ,
  • Aiden Doherty   ORCID: orcid.org/0000-0003-1840-0451 2 , 4   na1 ,
  • Luis Garcia-Gancedo   ORCID: orcid.org/0000-0001-5631-1711 3   na1 &
  • David A. Clifton 1   na1  

npj Digital Medicine volume  7 , Article number:  33 ( 2024 ) Cite this article

2157 Accesses

17 Altmetric

Metrics details

  • Chronic pain
  • Predictive markers
  • Rheumatoid arthritis
  • Translational research

Digital measures of health status captured during daily life could greatly augment current in-clinic assessments for rheumatoid arthritis (RA), to enable better assessment of disease progression and impact. This work presents results from weaRAble-PRO, a 14-day observational study, which aimed to investigate how digital health technologies (DHT), such as smartphones and wearables, could augment patient reported outcomes (PRO) to determine RA status and severity in a study of 30 moderate-to-severe RA patients, compared to 30 matched healthy controls (HC). Sensor-based measures of health status, mobility, dexterity, fatigue, and other RA specific symptoms were extracted from daily iPhone guided tests (GT), as well as actigraphy and heart rate sensor data, which was passively recorded from patients’ Apple smartwatch continuously over the study duration. We subsequently developed a machine learning (ML) framework to distinguish RA status and to estimate RA severity. It was found that daily wearable sensor-outcomes robustly distinguished RA from HC participants (F1, 0.807). Furthermore, by day 7 of the study (half-way), a sufficient volume of data had been collected to reliably capture the characteristics of RA participants. In addition, we observed that the detection of RA severity levels could be improved by augmenting standard patient reported outcomes with sensor-based features (F1, 0.833) in comparison to using PRO assessments alone (F1, 0.759), and that the combination of modalities could reliability measure continuous RA severity, as determined by the clinician-assessed RAPID-3 score at baseline ( r 2 , 0.692; RMSE, 1.33). The ability to measure the impact of the disease during daily life—through objective and remote digital outcomes—paves the way forward to enable the development of more patient-centric and personalised measurements for use in RA clinical trials.

Introduction

Rheumatoid arthritis (RA) patients follow subtle and unpredictable disease courses, patient-to-patient, with a progressive decline in physical function and quality of life and over time—often leading to disability and difficulty to perform many tasks of daily life 1 . RA symptoms include joint pain or tenderness, joint swelling, morning stiffness, reduction in joint range of movement (ROM), muscle pain, and fatigue 1 . Currently, the gold-standard methods to measure the impact of RA on daily life rely on infrequent clinical visits that may often occur every 3–4 months, with assessments depending on a combination of subjective clinician-determined scores 2 and patient-reported outcomes 3 . These have inherent limitations, however, in that they can be subjective and are prone to recall bias 4 , 5 . As such, there is a need to objectively measure the impact of RA on daily life 6 , remotely over a continuous period, rather than restricting assessments to only intermittent physician visits. In recent years, consumer-grade mobile applications (app.) and wearable devices have shown promise to objectively measure participants’ symptoms during daily life 7 ; these digital health technologies (DHT) tools 8 have shown to increase study engagement, improve patient convenience, streamline collection of PROs 9 , and potentially generate more frequent and accurate data that can characterise disease 10 . DHT have been shown to measure RA symptoms and functions, such as range of motion (ROM) and gait-specific metrics during prescribed “active” assessments 11 , 12 . Other studies have shown how “passive” wearable actigraphy sensor-outcome measurements capture differences in RA physical activity (PA) in daily life, compared to healthy controls (HC) 13 , as well as to detect flaring of RA symptoms 14 .

However, there remains a lack of sufficient evidence for how DHT can provide objective insights into the impact of therapies for RA, despite progress made in other disease areas 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 . Particularly, the benefit of sensor-outcomes generated from prescribed active assessments compared with passive monitoring has not yet been explored together. While digitised patient-reported outcomes (PROs) enable a patient the ability to regularly record their “subjective” experience of disease activity in remote settings 23 , it remains unclear how “objective” sensor-outcomes could provide additional insights that can augment PROs to better characterise the impact of RA on daily life. As part of this characterisation, the sensitivity of DHT to measure RA symptoms, such as the volume of remote data required and the number of sensor-outcome measurements needed, will also need to be determined. Finally, the application of DHT sensor-outcomes to monitor RA during daily life remains yet to be validated against standard in-clinic administered assessments of RA impact 24 .

In this study, we therefore aimed to investigate how active and passive sensor-based measurements should be combined using machine learning (ML) to distinguish RA status from healthy controls, to augment traditional patient self-reported outcome (PRO) data, and to estimate standard in-clinic assessments of RA severity. Our work offers the first comprehensive evaluation of how sensor data captured during daily life can characterise RA status and severity, which represents an important first step towards the development of more sensitive and patient-centric measurements for use in RA clinical trials and real-world studies.

In order to investigate the objectives of this study, we performed the following set of analysis and experiments. We first illustrate the variety of sensor-based measurements that can be extracted from daily prescribed (active) smartphone-based assessments and (passive) smartwatch-based activity monitoring in an RA cohort. In this, we evaluate how smartwatch-based daily physical activity patterns can be remotely estimated using our bespoke deep convolutional neural (DCNN), pre-trained using multi-task self-supervised learning (SSL) on a large-scale open-source cohort. We next assess the ability of our sensor-based measurements to identify RA status from healthy controls and to distinguish RA severity levels. As part of our analysis, we also explore the volume of days and number of sensor-outcomes required to remotely distinguish RA status. Finally, we investigated the power of active and passive sensor-outcomes to augment routinely collected patient self-reported outcome (PRO) data to estimate RA severity—as measured by standard in-clinic assessments of RA, such as the RAPID-3 25 .

The GSK weaRAble-PRO study (GSK212295) was a 14-day observational study which investigated how DHT tools could objectively measure the impact of RA on participants’ daily lives. Digital wearable devices—a wrist-worn Apple Watch for passive monitoring and an iPhone, integrated with a bespoke mobile app. which prescribed daily guided assessments—collected high-frequency, objective sensor data in 30 RA patients and 30 matched Healthy Controls (HCs). Figure 1 provides an illustrative overview of the objectives of this study. Sensor-based measures of physical function, mobility, dexterity, and other RA specific symptoms were extracted from daily prescribed (active) iPhone guided tests using a combination of bespoke algorithms and proprietary algorithms developed by Apple ResearchKit, for instance, a wrist-range of motion exercise, a walking assessment, a nine-hole peg test, as well as two pose transition-based mobility exercises, lie-to-stand (LTS) and sit-to-stand (STS). In addition, continuous (passive) actigraphy was recorded from participants’ Apple smartwatch over the study duration in order to characterise daily activity patterns and sleep. In order to illustrate the various characteristics of RA we are interested in assessing, we have grouped measurements in Fig. 1 into four domains: physical function, daytime activity, daily living, and sleep; denoting particular types of measurements which may attribute to each domain. Note: this manuscript details a sub-study of weaRAble-PRO; trial design, feasibility, participant adherence, and other primary related study outcomes are reported in Hamy et al. 26 . Two RA participants withdrew immediately after enroling in the study. Data from these participants were not collected, leaving 28 RA participants, 28 matched HCs, and 2 unmatched HCs for a total of 58 participant

figure 1

The weaRAble-PRO 14-day trial aimed to investigate how digital health technologies (DHT)—a wrist-worn Apple smartwatch and an iPhone device, with bespoke mobile apps.—could augment patient reported outcomes (PRO) to characterise the impact of rheumatoid arthritis (RA) during the daily life of 30 moderate-to-severe RA patients, compared to 30 matched healthy controls (HC). We explore the ability of machine learning (ML) models to (1) estimate categorical RA outcomes, such as identifying RA participants from healthy controls and (2) estimate continuous RA outcomes, such as RA severity, using a combination of PRO and sensor-outcomes.

Assessing smartwatch-based daily physical activity patterns

The daily physical activity of RA participants and healthy controls were estimated with a deep convolutional neural network (DCNN) that was first pre-trained on 100,000 participants in the publicly available UK Biobank, following a multi-task self-supervised learning (SSL) methodology 27 , which was subsequently fine-tuned on the free-living Capture-24 dataset 28 of < 150 participants to determine broad activity patterns of interest {sleep, sedentary, light physical activity, moderate-to-vigorous physical activity (MVPA)} 29 , 30 and fine-grained activity prediction labels {sleep, sitting/standing, mixed, vehicle, walking, bicycling} 28 . In this study, we build upon our previous work by adding a temporal dependency to the “DCNN (SSL)” through a hidden markov model (HMM), which was appended to obtain a more accurate sequence of predicted activities over the continuous study period. It was found that the “DCNN (SSL) + HMM” improved broad activity estimation in Capture-24 ( κ , 0.862 ± 0.088; F1, 0.815 ± 0.103) as compared to a baseline random forest (RF) + HMM approach ( κ , 0.813 ± 0.108; F1, 0.775 ± 0.117) 28 . Next, the fine-tuned “DCNN (SSL) + HMM” model transformed the raw Apple smartwatch sensor data in weaRAble-PRO to determine participants’ daily activity patterns over the 14-day study period, for example, the time spent walking, the frequency of exercise, the length and quality of sleep, and other RA-specific measures, such as morning stiffness. Activity predictions were qualitatively evaluated over the entire RA and HC study population and demonstrated face validity (see Supplementary Figs. 1 and 2 for additional details).

Analysis of sensor-outcomes to distinguish RA status and severity levels

The raw smartphone and smartwatch data recorded during the (active) guided test exercises, and passively during the participants’ daily life, respectively, were summarised as sensor-outcome features. Univariate analysis demonstrated that a total of 153 (93%) sensor-based features (passive, n  = 131 (94%); active, n  = 22 (88%)) displayed significantly different medians (after post-hoc correction for multiple comparisons) between HC and RA severity groups (Kruskal-Wallis H test, p  < 0.05). A further 47 (34%) passive features, compared to 6 (24%) active features, were also significantly different (Mann-Whitney U test, p  < 0.05) between healthy and RA participants. Figure 2 compares the (fortnightly) average feature distributions between healthy controls (HC), RA (moderate) and RA (severe) participants for a selection of examples of passively collected smartwatch features (Fig. 2 a–c) and active guided test sensor features (Fig. 2 d–f) and a selection of patient self-reported outcomes recorded on the smartphone application (Fig. 2 g–i).

figure 2

Comparison of the average feature distributions per participants, between healthy controls (HC), RA (moderate) and RA (severe) groups for: a – c selection of passively collected smartwatch features; d – f selection of guided test collected smartphone features; and g – i selection of patient self-reported outcomes recorded on the smartphone application. For all examples shown, medians were significantly different between HC and RA groups: One-way ANOVA determined from the Kruskal-Wallis H-test, p  < 0.001. deg degrees, HAQ-DI Health Assessment Questionnaire-Disability Index, min minutes, mg mili-gravity acceleration units, MVPA moderate-to-vigorous physical activity, RASIQ GSK RA symptom and impact questionnaire, sed sedentary, sec seconds.

In order to explore the ability of many wearable sensor-outcomes to distinguish symptoms of RA from otherwise healthy individuals, and therefore measure the impact of RA during daily life, we devised a number of multivariate classification-based experiments. First, we investigated the performance of regularised logistic regression (LR) to differentiate RA participants from healthy controls using both passively collected activity monitoring features and guided test exercise features. Comparing model performance between sources (Fig. 3 a), passive activity monitoring-based sensor features better distinguished RA participants using fortnightly averaged features (F1, 0.786) versus active (guided test) features (F1, 0.778). It was found that 12 subjects were misclassified using active-only models and 12 for passive-only, with just 4/12 (33%) of the same subjects incorrectly identified by both sources, 3 of which were the same HC participants. Combining active and passive wearable sensor features yielded in the highest performing models to distinguish RA participants overall, for example, using fortnightly averaged features from both sources (F1, 0.807) (for further expansion of results, see Supplementary Table 4) . It should also be noted that linear logistic regression was found to perform comparatively to non-linear ensembles of decision trees, a Random Forest (RF) model and Extreme Gradient Boosted Trees (XGB)—as such this work subsequently opted to explore simple linear models for further analysis (see Supplementary Table 5) .

figure 3

Comparison of a RA identification (RA vs. HC) performance and b RA severity level estimation (RA (mod) vs RA (sev)), using patient reported outcomes (PRO) and combined PRO (list icon), active (smartphone icon), and passive (smartwatch icon) sensor-based outcomes in the weaRAble-PRO study. auroc area under the receiver operator curve, κ Cohen’s Kappa statistic, F 1 macro-F1 score.

This study next investigated the ability of multiple sensor-based outcomes to augment PRO data in order to stratify RA severity levels. In weaRAble-PRO, participants were denoted as having moderate or severe RA based on baseline clinician-assessed RAPID-3 scores. Following similar procedure to RA identification, LR regularised models were investigated in order to distinguish RA (mod) and RA (sev) as binary classification tasks using fortnightly averaged study data. The benefit of incorporating additional sensor-based outcomes to patient (self-) reported outcomes is presented in Fig. 3 b (expanded in Supplementary Table 6) . It was observed that the linear combination of PRO assessments could accurately stratify RA symptom severity (F1, 0.759). The fusion of PRO data and sensor-based outcomes improved RA severity level estimation further with the addition of active (F1, 0.750) or passive (F1, 0.786) sources. Finally, the amalgamation of PRO outcomes with both active and passive sensor-based outcomes resulted in the most accurate RA severity level estimation (F1, 0.833)—an improvement of 10% compared to PRO outcomes alone (Fig. 3 b). For additional information on the selected PRO + sensor-outcomes, we refer the reader to Supplementary Table 3 .

Estimating the volume of days and number of sensor-outcomes required to remotely distinguish RA status

In weaRAble-PRO, participants performed daily guided test exercises—resulting in daily sensor features—and continuously recorded Apple Watch sensor data were summarised as daily activity monitoring-based features, over the 14-day study period. In this work, we aimed to determine the minimal number of days of sensor data required build a stable and robust estimate of disease status in RA participants compared to HC over the 14-day study period. Figure 4 a represents an experiment exploring the (observation-wise) out-of-sample RA classification performance as a function of varying the number of non-contiguous days of data that are averaged per participant. Evaluated over 500 randomly sampled permutations of non-contiguous days, results (median + IQR) indicated that RA prediction stabilised once more than 7 non-contiguous days of data were used per participant. Furthermore, we found that averaging daily feature values over weekly and fortnightly periods improved model performance. However, it was observed that model performance using weekly-averaged features was often similar to fortnightly averaged (we also refer the reader to Supplementary Table 4) .

figure 4

Comparison of a the minimal amount of days of data needed distinguish RA status, as measured by the F1 score across 5-fold cross validation (CV), between active (smartphone icon), passive (smartwatch icon), and combined (smartphone & smartwatch icons) feature sources; b the feature (test-retest) reliability, as measured by the intraclass correlation coefficient (ICC), between RA participants and HC across the study duration (14 days); F1 scores and ICCs suggest that model performance and feature reliability stabilises once more than 7 days of data are used per participant.

To investigate feature consistency and reproducibility, the intra-class correlation coefficient (ICC) for each feature was evaluated over the study duration (14 days). ICCs were calculated for each feature using n  = [2, 3, …, 14] days of data per participant, individually for HC and RA participants. Higher ICC’s suggest a high degree of similarity on the performance of each task over the course of the study, and lower coefficients mean that participants tended to perform the task differently each day of the study. ICC’s for HCs ranged from 0.582 to 0.854, while those for RA participants ranged from 0.424 to 0.897. Figure 4 b depicts the median + inter-quartile range (IQR) of ICC values for the LR-elastic net retained active + passive features. Intra-rater reliability analyses suggest that feature reliability stabilises to good (ICC=0.75–0.9) and excellent (ICC > 0.9) once more than 7 contiguous days of data were used per participant.

In order to evaluate the number of sensor-outcomes required to remotely distinguish RA status, we compared various feature regularisation techniques, lasso ( ℓ 1 ), ridge ( ℓ 2 ), elastic-net ( ℓ 1 + ℓ 2 ), and sparse-group lasso, using fortnightly (i.e., study duration) averaged features. It was found that introducing sparsity through regularisation improved classification performance. In addition, active and passively recorded sensor-based features could be grouped into domains, based on the guided test they were extracted from, or the perceived functional domain of daily activity they were assumed to assess. Introducing group-wise sparsity with the sparse-group lasso (SG-lasso), regularising on the number of groups (i.e., the feature domains) and the coefficients within each group, resulted in the highest RA participant identification performance (F1, 0.807), compared to lasso ( ℓ 1 , F1, 0.772), ridge ( ℓ 2 , F1, 0.792), and elastic net ( ℓ 1 + ℓ 2 , F1, 0.792) regularisation (for expansion of results, see Supplementary Table 5) . The features and groups selected by each regularisation technique are illustrated in Fig. 5 , represented as the mean LR coefficient value w over CV per each feature and feature domain (coefficient values have been normalised between 0 and 1 to benefit comparison between models). Examining the feature sparsity of elastic-net ( ℓ 1  +  ℓ 2 ) (Fig. 5 a), it was observed that features from multiple domains were selected. In contrast, the SG-lasso, as shown in Fig. 5 b, selected mostly passive activity-based smartwatch features—TVDA with some morning stiffness measures—to distinguish RA status. Group sparsity penalised simultaneously selecting from multiple feature domains, where within group-sparsity regularised the feature coefficient values within the selected domains. Using fewer domains and less features, the SG-lasso was able achieve similar performance to LR elastic-net, even marginally improving performance (F1, 0.807). For further details on the features extracted, and selected, we refer the reader to the Supplementary Methods.

figure 5

Comparison of features selected between regularised logistic regression (LR) models for: a elastic-net (F1, 0.79) and b SG-lasso (F1, 0.81). The SG-lasso promotes group-wise sparsity (i.e., regularising the number of feature domains) and within-group sparsity (i.e., regularising the number of features per domain), achieving a similar performance to LR elastic-net, while selecting a fewer number of domains and features. Feature importance, denoted as the mean LR coefficient value (w) over cross-validation, are illustrated by colour intensity. Feature domains: AF activity fragmentation, DEM demographics, LTS lie-to-stand assessment, MORN morning stiffness, NTR night-time restlessness, PEG 9-hole peg test, STS sit-to-stand assessment, TVDA total volume of daytime activity, WLK walking assessment, WRT wrist assessment.

Estimating in-clinic RA severity scores from PRO and sensor-based outcomes

Rheumatoid arthritis severity levels were denoted by a clinician administered RAPID-3 assessment 25 at baseline in the weaRAble-PRO study. The RAPID-3—a “rapid” and easy to administer questionnaire—is also validated against more exhaustive assessments for RA, such as the disease activity score 28 (DAS28) and clinical disease activity index (CDAI) in clinical trials and clinical care 25 . In this work, we aimed to establish how the combination of PRO and sensor-based outcomes could stratify continuous RAPID-3 RA severity. Note: HC subjects were assigned a RAPID-3 score of zero at baseline. Through multivariate modelling, using LR elastic-net, it was determined that PRO and sensor-based features could accurately estimate RAPID-3 scores to within 1 point ( r 2 , 0.69; MAE, 0.94; RMSE, 1.33), an improvement compared to using PRO measures alone ( r 2 , 0.63; MAE, 1.16; RMSE, 1.45). The association between actual and PRO + sensor-outcome estimated RAPID-3 scores was found to be good-to-excellent ( r  > 0.75), Pearson’s r = 0.60, p  < 0.001; Spearman’s ρ  = 0.83, p  < 0.001.

Participants in weaRAble-PRO were also administered a twice-daily interactive Joint Pain Map (JMAP) questionnaire on their iPhone 11 , in order to more precisely record and localise perceived pain. Participant model-estimated RAPID-3 scores were further interpreted through detailed inspection of the daily smartphone-based patient-reported joint pain map (JMAP) total scores—an external validation measure, which was not included as a predictor in the model—as expanded in Fig. 6 . The JMAP score, defined as the sum of all individual joint pain scores per recording, was intended as a coarse measure to holistically capture participants’ overall level of perceived pain, in addition to validated PRO assessments. Higher JMAP scores indicate higher levels of pain experienced. It was observed that RAPID-3 estimations were reliable and robust, in that they faithfully characterised RA participant’s perceived level of symptoms, through the JMAP. For example, in Fig. 6 , the RA (sev.) participant with consistently the largest reported degree of pain across the 14-day study exhibited the highest actual RAPID-3 score (6.7), which was closely estimated by the model at 7.1. JMAP scores further enabled additional explanation of model performance, especially with respect to RAPID-3 estimations that were not reflective of actual RAPID-3 scores. For instance, the RA (mod) participant with the lowest estimated RAPID-3 score (0.2) actually reported zero pain experienced over the 14-day study duration, despite a RAPID-3 assignment of 3.7 at baseline. Non-zero estimated RAPID-3 scores for some HC could also often be contextualised, due to these participants frequently self-reporting low-levels of pain in their JMAP (i.e., non-zero JMAP entries) over the study period, despite being healthy. As such, it was determined that PRO and sensor-based RAPID-3 estimates could reliably reflect participant’s RA symptoms over the study.

figure 6

Scatter plot of baseline RAPID-3 scores y versus predicted \(\hat{y}\) scores per subject, using elastic net with PRO + sensor-outcomes, over cross-validation (CV). Participant model-estimated RAPID-3 scores can be further interpreted through detailed inspection of the daily smartphone-based patient-reported joint pain map (JMAP) total scores—which was not included as a predictor in the model. Higher JMAP scores indicate higher levels of pain experienced. Additional interpretability, through the JMAP, demonstrated that PRO + sensor-based outcome estimation of the RAPID-3 could reliably reflect patient’s perceived daily RA symptoms. Note: Baseline JMAP total scores, recorded on the same day as the baseline RAPID-3, are denoted in grey; the JMAP y-axis scale is the same among all subplots. HC subjects were assigned a RAPID-3 score of zero at baseline. A black line represents perfect predictions ( r 2 , 0.692; MAE, 0.938; RMSE, 1.333).

Our findings in the weaRAble-PRO study demonstrate how digital health technology (DHT) captured sensor-outcomes, recorded from smartphone-based active tests, and continuously collected passive smartwatch-based monitoring, could characterise meaningful aspects of rheumatoid arthritis (RA) impairment and physical function impacting daily life. Remotely collected wearable sensor-outcomes could distinguish RA status from healthy controls—demonstrating further improved performance when combining the sensor-data from both devices—and how objective sensor-outcomes could augment patient (self-) reported outcomes to remotely estimate RA severity. Furthermore, by the half-way point of the weaRAble-PRO study (day 7), a sufficient volume of data had already been collected to reliably distinguish the characteristics of RA participants. This work provides the first comprehensive evaluation how remote and objective digital sensor-outcomes enrich our ability to understand the impact of RA on daily life between clinical visits.

In this work, we detailed how raw data collected from smartphone and smartwatch sensors can be transformed into sensor-based outcomes that are reflective of disease status. In concurrence with previous studies, many remotely collected smartphone sensor-outcomes distinguished RA participants and RA severity levels. For example, it was observed that joint ROM features differentiated HC and RA groups—a similar finding to our previous work 12 —and that RA participants were less mobile, taking longer to move between positions (as measured during the lie-to-stand exercise)—as previously shown by Andreu-Perez et al. 31 . Continuously collected smartwatch sensor data, known as passive monitoring, allowed the measurement of aspects of RA daily life, such as physical activity, sleep, and other RA specific symptoms, such as morning stiffness, or night-time restlessness. In this study we trained an activity recognition model on the free-living capture-24 dataset to estimate daily activity patterns in the wearable-pro population. Leveraging the latest advances in self-supervised learning (SSL) allowed our model to be pre-trained on 100,000 participants with 700,000 days of diverse, unlabelled wearable sensor data in the uk biobank 27 , which combined with HMM temporal smoothing, significantly improved activity prediction compared to our previous established RF-HMM based methods 28 , 30 . Our SSL DCNN+HMM model enabled a more robust and fine-grained estimation of daily activity patterns beyond traditional acceleration magnitude levels 13 , 14 , which we proposed could allow a richer characterisation of PA and sleep in RA activity monitoring revealed distinct differences distinguishing RA status, for example the daily percent of the day in moderate-to-vigorous physical activity, and similar features, were significantly lower in the RA population compared to healthy controls—a similar finding by Prioreschi et al. 13 , and an observation people with RA regularly self-report 32 . Other specific RA symptom measurements, like morning stiffness or disrupted sleep, were evident in certain RA participants. For example, the mean acceleration value > 30 [mins] after wake-up were lower in RA—also a similar finding to Keogh et al. 33 —or that the number of movement episodes during night-time sleep distinguished some specific RA participants. We also observed that after collecting 7 days of sensor-data in the weaRAble-PRO study, a sufficient volume of data had already been recorded to reliably distinguish RA participants from a healthy population; participant feature reliability (as measured ICC values) stabilised at good-to-excellent levels, maximal identification performance of RA participants plateaued, and that there was no additional benefit to averaging over a fortnight’s worth of data versus a week. Therefore it is recommended that considering at least one week’s worth of sensor data is collected, it might be more beneficial to gather less data from a greater number of participants, rather than greater duration of sensor data from the same participants.

Our work is the first study to combine active smartphone and passive wearable measurements to distinguish RA status and measure variations in RA severity. While models trained on only passive features tended to marginally outperform models trained solely on active guided test features, combining both active + passive features led to the best performance in RA identification for all models investigated. Interestingly, it was found that different subjects were misclassified by active versus passive models. For example, 12 subjects were misclassified using active-only models and 12 for passive-only, with just 4/12 (33%) of the same subjects incorrectly identified by both sources, 3 of which were the same HC participants. In addition, further experiments with the LR-SG-lasso determined that only activity monitoring domain features were mainly needed in order to distinguish RA participants from health controls. This indicates that we sometimes do not need to prescribe all guided test assessments, or to parse all activity feature domains, but that a small number of prescribed assessments can be sufficient to characterise RA status. For example, including only the lie-to-stand assessment rather than also prescribing the similar, and highly correlated, sit-to-stand assessment in future studies; or removing the prescribed walking assessment (shown to have little predictive value in the weaRAble-PRO study), and using passive daily life walking predictions generated from the activity recognition model instead, which could reduce patient burden. Finally, we also found that combining patient-reported outcomes (PRO) and objective sensor-outcomes could better capture RAPID-3-based RA severity at baseline than PROs alone; most estimated RAPID-3 scores correctly stratified participants across severity levels from healthy to moderate to severe RA, suggesting that sufficient information to characterise RA disease severity could be reflected in the remote monitoring outcomes derived in the 14-day weaRAble-PRO study. To the best of the authors knowledge, this offers the first evaluation and insight how remote monitoring outcomes in daily life can estimate in-clinic administered assessments of RA impact.

There are a number of limitations that must be considered in the weaRAble-PRO study. Despite rich individual level measurements, the study recruited a relatively small sample size (HC, n  = 30; RA, n  = 30). As such, a degree of variability and uncertainty existed in constructing cross-validated models to distinguish RA participants, RA severity levels, or estimate the in-clinic RAPID-3 assessment. Extrapolation of results aimed at generalising RA is therefore not possible without the availability of larger cohorts and further external validation. In addition, this study only recruited RA patients with moderate-to-severe levels of disease activity; future studies should also aim to characterise patients with lower levels of disease activity or those in remission. There were also limitations associated with modelling a clinician-administered assessment, or clinical labels formulated from in-clinic assessments. For instance, the RAPID-3 was assessed at baseline, with participants recalling the prior week, yet the PRO and sensor-based features were calculated as averages over subsequent 14-day trial period from baseline. As such, the baseline RAPID-3 may not have precisely reflected the participant’s disease status recorded earlier, due to the underlying mutability and heterogeneity of RA symptoms over short periods of time. The subjectivity of PRO predictors should also considered, for instance, pain or perceived quality of sleep is relative, and some healthy participants recorded experiencing pain or affected sleep in PRO questionnaires. As a result, some PRO values influenced HC RAPID-3 predictions greater than zero, i.e., indicating the presence of RA symptoms—albeit non-zero estimated RAPID-3 predictions for HCs were generally low ( < 2).

The weaRAble-PRO study typifies how continuously collected patient self-reported and sensor-based outcomes may more closely reflect participant perceived and experienced symptoms that impact daily life. While in-clinic assessments are considered the gold-standard means of assessing disease severity in RA, it is clear that remotely collected, continuous, patient-centric measurements generated from PRO and sensor-based outcomes offer promising insights that can undoubtedly augment in-clinic assessments for RA. We believe that our work—the first comprehensive evaluation how remote sensor data can augment traditional PRO measures to estimate clinician-determined RA severity—helps informs future DHT study design to better characterise the impact of RA on daily life, ultimately to expand the use of DHT to develop more sensitive, and patient-centric, endpoints in RA clinical trials and real-world studies.

Remotely collected smartphone and smartwatch sensor data was obtained from the GSK study title: Novel Digital Technologies for the Assessment of Objective Measures and Patient Reported Outcomes in Rheumatoid Arthritis Patients: A Pilot Study Using a Wrist-Worn Device and Bespoke Mobile App. (212295, weaRAble-PRO) 26 . This observational study followed 30 participants diagnosed with moderate-to-severe RA and 30 matched HCs over 14 days. The population demographics, in-clinic, and relevant patient self-reported outcomes, as assessed at baseline, are reported in Table 1 . RA participants were denoted as displaying moderate disability, RA (mod), or severe disability, RA (sev), as determined by their baseline RAPID-3 score. Note: Two RA participants withdrew immediately after enroling in the study. Data from these participants were not collected, leaving 28 RA participants, 28 matched HCs, and 2 unmatched HCs for a total of 58 participants. All study information, informed consent, study questions and instructions for conducting the guided tests were first drafted in the form of a survey instrument. The survey instrument was then programmed into the mobile app. All documentation including the study protocol, any amendments, and informed consent procedures, were reviewed and approved by Reliant Medical Group’s IRB. All participants provided written informed consent before any study procedures were undertaken. The study was conducted in accordance with the International Committee for Harmonisation principles of Good Clinical Practice and the Declaration of Helsinki. We refer the reader to Hamy et al. 26 for further study details. In addition, participant requirement and data collection are outlined in the accompanying Supplementary Methods material.

Sensor-based data collection

The Apple Watch and iPhone were used to collect high frequency raw sensor data from predefined, (active) guided tests on a daily basis. Participants were prescribed daily to perform five iPhone-based assessments: WRT, a wrist range of motion (ROM) exercise 12 ; WLK, a 30-second walking exercise 12 ; PEG, a digital 9-hole peg test 34 ; STS, a sit-to-stand transition exercise 31 , 35 ; and LTS, a lie-to-stand transition exercise 31 , 35 . A brief overview of the guided tests prescribed in weaRAble-PRO are presented in Supplementary Table 8 . In addition, the Apple Watch was used to continuously collect background sensor data (denoted passive data), as the participants went about their daily activities. Participants were asked to maintain a charge on both the Apple Watch and the iPhone, so that interruptions to monitoring and data transfer were kept to a minimum. Since night-time activity was also monitored, while participants were asleep, it was requested that charging should be done during the day, in a way that fit the participants’ schedules (e.g., charging in the morning while getting ready for the day). For more details on the activity monitoring features, see Supplementary Table 9 .

Patient-reported outcomes

Patient-reported outcomes (PRO), most often self-report questionnaires, were administered to assess disease activity, symptoms, and health status and quality of life from the patients’ perspective 36 , 37 . The weaRAble-PRO study administered a selection of validated PRO measures for RA in complement to bespoke digital PRO assessments—that are validated in clinical trials, where the questions, response options, and the general approach to assessment were standardised for all participants. PROs were recorded on days 1, 7, and 14 of data collection. The PRO assessments administered to participants are outlined in Supplementary Table 7 .

Smartwatch-based estimation of daily life patterns

In order to generate unobtrusive measures characterising physical activity and sleep in RA participants during daily life, the raw Apple Watch actigraphy (i.e., accelerometer) sensor data was transformed through a human activity recognition (HAR) sensor processing and deep convolutional neural network (DCNN) pipeline. Figure 7 illustrates how a deep convolutional neural network (DCNN) can transform raw Apple smartwatch sensor data to estimate a participant’s daily activity patterns in the weaRAble-PRO study using self-supervised learning (SSL). The construction of this pipeline yielded unobtrusively measured summary features of physical activity and sleep for RA participants, computed daily during normal life.

figure 7

Continuous (passive) actigraphy was recorded from patients' Apple smartwatch over the study duration. Deep convolutional neural networks (DCNN) were pre-trained on 700,000 person days in the publicly available UK Biobank using self-supervised learning—and fine-tuned with the Capture-24 dataset—to estimate participant’s daily activity patterns in the weaRAble-PRO study. Physical activity (PA) metrics of daily-life, for example, the time spent walking, the frequency of exercise, or the length and quality of sleep were investigated as markers to characterise symptoms of disease in people with RA compared to HC.

A deep convolutional neural network (DCNN) with a ResNet-V2 architecture was first pre-trained following a multi-task self-supervised learning (SSL) methodology on 100,000 participants, each participant contributing 7 days yielding roughly 700,000 person days of data, in the open-source UK biobank 27 . The SSL pre-trained model was then fine-tuned to perform activity recognition as a downstream task in the Capture-24 dataset.

The Capture-24 study is a manually labelled, free-living dataset—that is reflective of real-world environments—and is available for training an activity recognition model to be applied to the weaRAble-PRO study. In Capture-24, actigraphy data was collected for 24-h from 132 healthy volunteer participants with a Axivity AX3 wrist-worn device as they went their normal day. Activity labels provided by photographs automatically captured roughly every 30 seconds by a wearable camera for each participant. Capture-24 was labelled with 213 activity labels, standardised from the compendium of physical activities 29 . Activity labels were then summarised into a small number of free-living behaviour labels, defining activity classes in Capture-24.

There are two major labelling conventions used within Capture-24 that the model was trained to predict, defined as broad activity: {sleep, sedentary, light physical activity, moderate-to-vigorous physical activity (MVPA)} 29 , 30 ; and fine-grained activity: {sleep, sitting/standing, mixed, vehicle, walking, bicycling} 28 .

HAR model predictions are essentially independent—meaning that the sequence of activities over each 30 s epoch incorporates no temporal information epoch-to-epoch, for instance how the previous epoch prediction affects the current, or next, activity prediction. In order to add temporal dependency to the “DCNN (SSL)” model, a Hidden Markov Model (HMM) was implemented in a post-processing step to obtain a more accurate sequence of predicted activities over the continuous 14-day data collection period as per Willetts, et al. 28 .

This Capture-24 fine-tuned “DCNN (SSL) + HMM” model was then implemented to estimate daily activities in weaRAble-PRO study data. For additional information of the HAR deep network, SSL, and other related information, we refer the reader to our previous work 27 . Further results relating to the “DCNN (SSL)” models are outlined in the Supplementary Table 1 . The sensor processing pipeline developed for the Apple Watch in the weaRAble-PRO study is outlined in Supplementary Fig. 5 and within the accompanying Supplementary Methods.

Extraction of sensor-based outcomes

Wearable sensor-based features were derived from the smartphone during the active guided tasks and passively from the smartwatch during daily life. “Active” features, extracted from smartphone sensor-based measurements during the prescribed guided tests, aimed to capture specific aspects of RA physical function, related to pain, dexterity, mobility and fatigue 12 . In addition “passive” features were extracted from smartwatch sensor-based measurements, collected continuously in the background over the 14-day period. Daily activity predictions from the ML SSL model were summarised into general features measuring activity levels, period, duration and type of activity, as well as sleep detection and sleeping patterns. Furthermore, devised under the guidance of Rheumatologists, additional activity monitoring features specifically aimed at characterising well-known RA symptoms were also developed, such as morning stiffness and night-time restlessness.

The Supplementary Methods also detail algorithms used to extract active and passive features in the weaRAble-PRO study. For a full list of extracted sensor-based features in weaRAble-PRO, we refer the reader to Supplementary Table 9 .

Statistical analysis

Univariate testing.

Pair-wise differences groups between groups, for example HC vs. RA, or RA (mod) vs. RA (sev) were analysed for the equality in population median using the non-parametric Mann-Whitney U test (MWUT) 38 , 39 , 40 . One-way analysis of variance (ANOVA) tests were also used to assess differences between medians of multiple groups, for example HC vs. RA (mod) vs. RA (sev) were assessed using the Kruskal-Wallis (KWt) test by ranks 41 . The Brown-Forsythe (BF) test by (absolute deviation) of medians was used to investigate if various groups of data have been drawn with equal variances 42 .

Correlation analysis

Correlation analysis was utilised to determine the association or dependence between sets of random variables, such as the dependence between features, or to assess a features’ clinical utility by measuring the association to an established clinical metric. This study investigated the (linear) Pearson’s r correlation and the (non-linear) Spearman’s Rho ρ correlation between features, between features and PROs, and between clinical assessments to determine levels of association. The strengths of the correlations were classified as good-to-excellent ( r  > 0.75), moderate-to-good ( r  = 0.50–0.75), fair ( r  = 0.25–0.49) or no correlation ( r  < 0.25) 43 .

Feature reliability

Intra-rater (i.e., test-retest) reliability was determined using intra-class correlation coefficient (ICC) values 44 , which were used to assess the degree of similarity between repeated features over the course of the study for each patient. In this work, the I C C (3,  k ) was calculated 45 –which considers the two-way random average measures with k repeated measurements—for the 14-day session across subjects, where the raters k are the study days. Reliability was categorised as either poor (ICC < 0.5), moderate (ICC=0.5–0.75), good (ICC=0.75–0.9), or excellent (ICC > 0.9) 46 .

Correcting for multiple hypothesis testing

Multiple hypothesis testing was performed due to the large volume of features by controlling the false discovery rate (FDR) at level α using the linear step-up procedure introduced by Benjamini and Hochberg (BH) 47 , 48 .

Machine-learning estimation of RA status and severity

This work explored how state-of-the art machine learning (ML) models characterise the impact of RA during the daily life of participants in the 14-day weaRAble-PRO study. Multivariate modelling aimed to explore the ability of active, passive, and PRO measures to (1) distinguish RA participants from healthy controls (HC), and (2) to estimate RA disease severity: between RA participants with moderate symptoms (RA mod) and severe symptoms (RA sev) as binary classification tasks. Expansions of this analysis subsequently investigated how the in-clinic RAPID-3 assessment, a continuous measure of RA severity, could be estimated from the combination of PRO and sensor-based outcomes.

Overview of models

This analysis compared both linear and non-linear ML models to transform PRO and sensor-based outcomes to capture RA status and severity. Regularised linear regression (LR) models, with combinations of ℓ 1 and ℓ 2 priors, such as LR-lasso ( ℓ 1 ), LR-ridge ( ℓ 2 ), and LR-elastic-net ( ℓ 1 + ℓ 2 ) were compared to yield predictive, yet sparse model solutions 49 . Further regularisation extensions were also investigated using the sparse-group lasso (SG-lasso)—an extension of the lasso that promotes both group sparsity and within group parameter-wise ( ℓ 2 ) sparsity, through a group lasso penalty and the lasso penalty—which aims to yield a sparse set of groups and also a sparse set of covariates in each selected group 50 , 51 .

Linear regression regularised models were also compared to decision tree (DT) based non-linear models, for instance the off-the-shelf Random Forest (RF) 52 and Extreme Gradient Boosted Trees (XGB) 53 . Both LR- and DT-based models can intrinsically perform regression or classification depending on the task required. In the LR case, classification is denoted as logistic regression (though a logit-link function). NOTE: in this analysis LR can refer to both linear regression for continuous outputs or logistic regression for classification outputs. In the DT case, the mean prediction of the individual trees creates a continuous output for regression. For further details on the models employed in this study, we refer the reader to the Supplementary Methods.

Model evaluation

To determine the generalisability of our models, a stratified subject-wise k-fold cross-validation (CV) was employed. This consisted of randomly partitioning the dataset into k=5 folds, which was stratified with equal class proportions where possible. Participant data remained independent between training, validation, and testing splits. One set was denoted the training set (in-sample), and the remaining 20% of the dataset was then denoted testing set (out-of-sample) on which predictions were made.

Feature-wise and prediction-wise aggregation

In this work, we experimented with feature-wise and prediction-wise aggregation. In feature-wise aggregation, features were computed either as: daily feature values over the 14-day study period; the average daily feature value over a 7-day period (weekly); the average daily feature value over a 14-day period (fortnightly). Predictions could then be evaluated for each day (denoted observation-wise ) or aggregated over all days through majority voting each individual prediction per subject (denoted subject-wise ). For example, daily and weekly averaged features result in daily, or weekly predictions (i.e., observation-wise ), which were summarised into subject-wise outcomes by majority voting over the repeated predictions.

Evaluation metrics

Multi-class classification metrics were reported as the observation-wise median and interquartile (IQR) range over one CV, as well as the subject-wise outcome for that CV, using: auroc, area under the receiver operating characteristic curve; k , Cohen’s kappa statistic 54 , 55 ; F 1 , F1-score. The coefficient of determination, r 2 , the mean absolute error (MAE), and root-mean squared error (RMSE) were used to evaluate modelling the (continuous) in-clinic RAPID-3 scores 56 .

Data availability

Anonymised individual participant data that support the findings of this study are available from the corresponding author, upon reasonable request and subject to GSK’s approval.

Code availability

Apple Watch sensor processing was performed using a bespoke version of the biobankAccelerometerAnalysis toolkit, found at: https://github.com/OxWearables/biobankAccelerometerAnalysis . Deep networks were built using Python v3.7 through a PyTorch v1.7 framework. Our self-supervised learning activity prediction code and trained models are publicly available at: https://github.com/OxWearables/ssl-wearables , including pre-trained models on 100K participants in the UK Biobank. Some guided test exercises and health metrics calculated are proprietary to Apple ResearchKit ( http://researchkit.org/ ) and Apple HealthKit ( https://developer.apple.com/documentation/healthkit ) which we refer the reader for more details. Statistical and machine learning analysis was developed using scikit-learn v1.1.1. Further analysis code can be made available from the corresponding author upon reasonable request.

Grassi, W., De Angelis, R., Lamanna, G. & Cervini, C. The clinical features of rheumatoid arthritis. Eur. J. Radiol. 27 , S18–S24 (1998).

Article   PubMed   Google Scholar  

Banderas, B., Skup, M., Shields, A. L., Mazar, I. & Ganguli, A. Development of the rheumatoid arthritis symptom questionnaire (rasq): a patient reported outcome scale for measuring symptoms of rheumatoid arthritis. Curr. Med. Res. Opin. 33 , 1643–1651 (2017).

Lubeck, D. P. Patient-reported outcomes and their role in the assessment of rheumatoid arthritis. Pharmacoeconomics 22 , 27–38 (2004).

Campbell, R., Ju, A., King, M. T. & Rutherford, C. Perceived benefits and limitations of using patient-reported outcome measures in clinical practice with individual patients: a systematic review of qualitative studies. Quality Life Res. 1–24 (2021).

Gossec, L., Dougados, M. & Dixon, W. Patient-reported outcomes as end points in clinical trials in rheumatoid arthritis. RMD Open 1 , e000019 (2015).

Article   PubMed   PubMed Central   Google Scholar  

Flurey, C. A., Morris, M., Richards, P., Hughes, R. & Hewlett, S. It’s like a juggling act: rheumatoid arthritis patient perspectives on daily life and flare while on current treatment regimes. Rheumatology 53 , 696–703 (2014).

Piga, M., Cangemi, I., Mathieu, A. & Cauli, A. Telemedicine for patients with rheumatic diseases: systematic review and proposal for research agenda. In Seminars in Arthritis and Rheumatism , Vol. 47, 121–128 (Elsevier, 2017).

Taylor, K. I., Staunton, H., Lipsmeier, F., Nobbs, D. & Lindemann, M. Outcome measures based on digital health technology sensor data: data-and patient-centric approaches. NPJ Digital Med. 3 , 1–8 (2020).

Article   Google Scholar  

Yun, H. et al. Assessing rheumatoid arthritis disease activity with patient-reported outcomes measurement information system measures using digital technology. Arthritis Care Res. 72 , 553–560 (2020).

Munos, B. et al. Mobile health: the power of wearables, sensors, and apps to transform clinical trials. Ann. New York Acad. Sci. 1375 , 3–18 (2016).

Article   ADS   Google Scholar  

Crouthamel, M. et al. Using a researchkit smartphone app to collect rheumatoid arthritis symptoms from real-world participants: feasibility study. JMIR mHealth uHealth 6 , e9656 (2018).

Hamy, V. et al. Developing smartphone-based objective assessments of physical function in rheumatoid arthritis patients: the PARADE study. Digital Biomarkers 4 , 26–44 (2020).

Prioreschi, A., Hodkinson, B., Avidon, I., Tikly, M. & McVeigh, J. A. The clinical utility of accelerometry in patients with rheumatoid arthritis. Rheumatology 52 , 1721–1727 (2013).

Gossec, L. et al. Detection of flares by decrease in physical activity, collected using wearable activity trackers in rheumatoid arthritis or axial spondyloarthritis: an application of machine learning analyses in rheumatology. Arthritis Care Res. 71 , 1336–1343 (2019).

Pourahmadi, M. R. et al. Reliability and concurrent validity of a new iphone® goniometric application for measuring active wrist range of motion: a cross-sectional study in asymptomatic subjects. J. Anatom. 230 , 484–495 (2017).

Pratap, A. et al. Evaluating the utility of smartphone-based sensor assessments in persons with multiple sclerosis in the real-world using an app (elevateMS): observational, prospective pilot digital health study. JMIR mHealth uHealth 8 , e22108 (2020).

Webster, D. E. et al. Clinical validation of digital biomarkers and machine learning models for remote measurement of psoriasis and psoriatic arthritis. medRxiv (2022).

Omberg, L. et al. Remote smartphone monitoring of Parkinson’s disease and individual response to therapy. Nat. Biotechnol. 40 , 480–487 (2022).

Article   PubMed   CAS   Google Scholar  

Creagh, A. P. et al. Smartphone- and smartwatch-based remote characterisation of ambulation in multiple sclerosis during the two-minute walk test. IEEE J. Biomed. Health Inf. 25 , 838–849 (2021).

Creagh, A. et al. Smartphone-based remote assessment of upper extremity function for multiple sclerosis using the draw a shape test. Physiol. Measur. 41 , 054002 (2020).

Article   ADS   CAS   Google Scholar  

Lipsmeier, F. et al. Reliability and validity of the Roche PD mobile application for remote monitoring of early parkinson’s disease. Sci. Rep. 12 , 1–15 (2022).

Lipsmeier, F. et al. A remote digital monitoring platform to assess cognitive and motor symptoms in huntington disease: cross-sectional validation study. J. Med. Internet Res. 24 , e32997 (2022).

El Miedany, Y. et al. Toward electronic health recording: evaluation of electronic patient-reported outcome measures system for remote monitoring of early rheumatoid arthritis. J. Rheumatol. 43 , 2106–2112 (2016).

Coravos, A., Khozin, S. & Mandl, K. D. Developing and adopting safe and effective digital biomarkers to improve patient outcomes. NPJ Digital Med. 2 , 1–5 (2019).

Google Scholar  

Pincus, T., Yazici, Y. & Bergman, M. J. Rapid3, an index to assess and monitor patients with rheumatoid arthritis, without formal joint counts: similar results to das28 and cdai in clinical trials and clinical care. Rheum. Dis. Clin. 35 , 773–778 (2009).

Hamy, V. et al. Patient-centric assessment of rheumatoid arthritis using a smartwatch and bespoke mobile app in a clinical setting. Sci. Rep. 13 , 18311 (2023).

Article   ADS   PubMed   PubMed Central   CAS   Google Scholar  

Yuan, H. et al. Self-supervised learning for human activity recognition using 700,000 person-days of wearable data. arXiv preprint arXiv:2206.02909 (2022).

Willetts, M., Hollowell, S., Aslett, L., Holmes, C. & Doherty, A. Statistical machine learning of sleep and physical activity phenotypes from sensor data in 96,220 uk biobank participants. Scientific reports 8 , 1–10 (2018).

Article   CAS   Google Scholar  

Ainsworth, B. E. et al. 2011 compendium of physical activities: a second update of codes and met values. Med. Sci. Sports Exerc. 43 , 1575–1581 (2011).

Walmsley, R. et al. Reallocating time from device-measured sleep, sedentary behaviour or light physical activity to moderate-to-vigorous physical activity is associated with lower cardiovascular disease risk. MedRxiv (2020).

Andreu-Perez, J. et al. Developing fine-grained actigraphies for rheumatoid arthritis patients from a single accelerometer using machine learning. Sensors 17 , 2113 (2017).

Article   ADS   PubMed   PubMed Central   Google Scholar  

Sokka, T. et al. Physical inactivity in patients with rheumatoid arthritis: data from twenty-one countries in a cross-sectional, international study. Arthritis Care & Research: Official Journal of the American College of Rheumatology 59 , 42–50 (2008).

Keogh, A. et al. A thorough examination of morning activity patterns in adults with arthritis and healthy controls using actigraphy data. Digital Biomarkers 4 , 78–88 (2020).

Mathiowetz, V., Weber, K., Kashman, N. & Volland, G. Adult norms for the nine hole peg test of finger dexterity. The Occupational Therapy Journal of Research 5 , 24–38 (1985).

Bohannon, R. W. Sit-to-stand test for measuring performance of lower extremity muscles. Perceptual and motor skills 80 , 163–166 (1995).

of Health, U. D. et al. Guidance for industry: patient-reported outcome measures: use in medical product development to support labeling claims: draft guidance. Health and Quality of Life Outcomes 4 , 79 (2006).

Mercieca-Bebber, R., King, M. T., Calvert, M. J., Stockler, M. R. & Friedlander, M. The importance of patient-reported outcomes in clinical trials and strategies for future optimization. Patient Related Outcome Measures 9 , 353 (2018).

Mann, H. B. & Whitney, D. R. On a test of whether one of two random variables is stochastically larger than the other. The Annals of Mathematical Statistics 50-60 (1947).

Hollander, M., Wolfe, D. A. & Chicken, E. Nonparametric statistical methods , Vol. 751 (John Wiley & Sons, 2013).

Gibbons, J. D. & Chakraborti, S. Nonparametric Statistical Inference: Revised and Expanded (CRC press, 2014).

Kruskal, W. H. & Wallis, W. A. Use of ranks in one-criterion variance analysis. J. Am. Stat. Assoc. 47 , 583–621 (1952).

Brown, M. B. & Forsythe, A. B. Robust tests for the equality of variances. J. Am. Stat. Assoc. 69 , 364–367 (1974).

Portney, L. G. & Watkins, M. P. Foundations of clinical research: applications to practice, vol. 892 (Pearson/Prentice Hall Upper Saddle River, NJ, 2009).

Weir, J. P. Quantifying test-retest reliability using the intraclass correlation coefficient and the sem. J. Strength Condit. Res. 19 , 231–240 (2005).

Shrout, P. E. & Fleiss, J. L. Intraclass correlations: uses in assessing rater reliability. Psychol. Bull. 86 , 420 (1979).

Koo, T. K. & Li, M. Y. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J. Chiropr. Med. 15 , 155–163 (2016).

Shaffer, J. P. Multiple hypothesis testing. Ann. Rev. Psychol. 46 , 561–584 (1995).

Benjamini, Y. & Hochberg, Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc.: Ser. B (Methodological) 57 , 289–300 (1995).

MathSciNet   Google Scholar  

Hastie, T., Tibshirani, R. & Friedman, J.The elements of statistical learning: data mining, inference, and prediction (Springer Science & Business Media, 2009).

Friedman, J., Hastie, T. & Tibshirani, R. A note on the group lasso and a sparse group lasso. arXiv preprint arXiv:1001.0736 (2010).

Simon, N., Friedman, J., Hastie, T. & Tibshirani, R. A sparse-group lasso. J. Comput. Graph. Stat. 22 , 231–245 (2013).

Article   MathSciNet   Google Scholar  

Breiman, L. Random forests. Mach. Learn. 45 , 5–32 (2001).

Chen, T. & Guestrin, C. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining , 785–794 (2016).

He, H. & Garcia, E. A. Learning from imbalanced data. IEEE Trans. Knowl. Data Eng. 21 , 1263–1284 (2009).

Cohen, J. A coefficient of agreement for nominal scales. Educ. Psychol. Meas. 20 , 37–46 (1960).

Rao, C. R. Linear statistical inference and its applications, vol. 2 (Wiley New York, 1973).

Download references

Acknowledgements

We are grateful to all the study participants and their families for their time and dedication to this study. The authors would also like to thank Priyanka Bobbili PhD, Julien Bendelac BSc, Jessica Landry MSc, Maral DerSarkissian PhD, Mihran Yenikomshian MBA, and Med Kouaici (MEng) from Analysis Group (MA, USA) for their support in app. design & development and data collection, and to Elinor Mody from Reliant Medical Group (MA, USA) for patient recruitment. The weaRAble-PRO study was funded and sponsored by GSK Plc. The research described in this paper was funded by GSK Plc. This research also acknowledges support from the National Institute for Health Research (NIHR) Oxford Biomedical Research Centre (BRC). Aiden Doherty is supported by the Wellcome Trust [223100/Z/21/Z].

Author information

These authors jointly supervised this work: Aiden Doherty, Luis Garcia-Gancedo, David A. Clifton.

Authors and Affiliations

Institute of Biomedical Engineering, Department of Engineering Science, University of Oxford, Oxford, UK

Andrew P. Creagh, Gert Mertes & David A. Clifton

Big Data Institute, University of Oxford, Oxford, UK

Andrew P. Creagh, Hang Yuan, Gert Mertes & Aiden Doherty

Value Evidence and Outcomes (VEO), GSK, UK

Valentin Hamy & Luis Garcia-Gancedo

Nuffield Department of Population Health, University of Oxford, Oxford, UK

Hang Yuan, Gert Mertes & Aiden Doherty

Value Evidence and Outcomes (VEO), GSK, US

Ryan Tomlinson, Wen-Hung Chen & Rachel Williams

Analysis Group (AG), Boston, MA, USA

Christopher Llop, Christopher Yee & Mei Sheng Duh

You can also search for this author in PubMed   Google Scholar

Contributions

A.P.C. conceptualised the data analysis, designed methodology, software, and interpretation. V.H., H.Y., and G.M. contributed software applications for analysis. V.H., R.T., W.-H.C., R.W., L.G.-G. contributed to the design of the study and towards the data analysis and interpretation. C.L., C.Y., and M.S.D. were involved in the design of the study, data collection, and software for data acquisition. A.D., L.G.-G., and D.A.C. jointly supervised. A.P.C. wrote the manuscript; all other authors: review & editing.

Corresponding author

Correspondence to Andrew P. Creagh .

Ethics declarations

Competing interests.

A.P.C, H.Y, G.M, A.D, D.A.C are employees of the University of Oxford. A.P.C is a GSK postdoctoral fellow and acknowledges the support of GSK. D.A.C received research funding from GSK to conduct this work. In addition, A.D., H.Y., and G.M. acknowledge the support of Novo Nordisk plc. A.D. AD is supported by the Wellcome Trust [223100/Z/21/Z]. V.H, W-H.C, R.T, R.W and L.G-G are employees of GSK and own stock and or shares. C.L, C.Y, M.S.D are employees of Analysis Group, which received research funding from GSK to conduct the study.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary material rendered, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Creagh, A.P., Hamy, V., Yuan, H. et al. Digital health technologies and machine learning augment patient reported outcomes to remotely characterise rheumatoid arthritis. npj Digit. Med. 7 , 33 (2024). https://doi.org/10.1038/s41746-024-01013-y

Download citation

Received : 16 November 2022

Accepted : 18 January 2024

Published : 12 February 2024

DOI : https://doi.org/10.1038/s41746-024-01013-y

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

  • Explore articles by subject
  • Guide to authors
  • Editorial policies

Sign up for the Nature Briefing: Translational Research newsletter — top stories in biotechnology, drug discovery and pharma.

research paper on rheumatoid arthritis

  • Search Menu
  • Advance articles
  • Editor's Choice
  • Supplement Archive
  • Article Collection Archive
  • Author Guidelines
  • Submission Site
  • Open Access
  • Call for Papers
  • Why Publish?
  • About Nutrition Reviews
  • About International Life Sciences Institute
  • Editorial Board
  • Early Career Editorial Board
  • Advertising and Corporate Services
  • Journals Career Network
  • Self-Archiving Policy
  • Journals on Oxford Academic
  • Books on Oxford Academic

Issue Cover

Article Contents

Introduction, acknowledgments.

  • < Previous

Rheumatoid arthritis and dietary interventions: systematic review of clinical trials

ORCID logo

  • Article contents
  • Figures & tables
  • Supplementary Data

Elena Philippou, Sara Danuta Petersson, Carrie Rodomar, Elena Nikiphorou, Rheumatoid arthritis and dietary interventions: systematic review of clinical trials, Nutrition Reviews , Volume 79, Issue 4, April 2021, Pages 410–428, https://doi.org/10.1093/nutrit/nuaa033

  • Permissions Icon Permissions

The impact of various dietary interventions on rheumatoid arthritis (RA), characterized by immune-inflammatory response, has been subject to increased attention.

A systematic review was conducted to update the current knowledge on the effects of nutritional, dietary supplement, and fasting interventions on RA outcomes.

Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, with prespecification of all methods, Medline and Embase were systematically searched for relevant articles.

Data were extracted by 2 independent reviewers.

A total of 70 human studies were identified. Administration of omega-3 polyunsaturated fatty acids at high doses resulted in a reduction in RA disease activity and a lower failure rate of pharmacotherapy. Vitamin D supplementation and dietary sodium restriction were beneficial on some RA outcomes. Fasting resulted in significant but transient subjective improvements. While the Mediterranean diet demonstrated improvements in some RA disease activity measures, outcomes from vegetarian, elimination, peptide, or elemental diets suggested that responses are very individualized.

Some dietary approaches may improve RA symptoms and thus it is recommended that nutrition should be routinely addressed.

Rheumatoid arthritis (RA) is a disease characterized by systemic immune-inflammatory responses, causing global disability. 1 In addition to joint damage, bone erosion, and severe pain, which significantly impact patients’ physical function, emotional state, and quality of life, 2 RA also results in increased risk for cardiovascular disease and a reduction in life expectancy. 3 The management of RA consists of conventional synthetic disease-modifying antirheumatic drugs (DMARDs), but also biological DMARDs and other recently discovered therapies such as targeted synthetic DMARDs. During the acute stages, steroids are used to control active disease. With a rapidly expanding therapeutic armamentarium, remission in RA is now a possible target, with reduction in pain, joint tenderness, and functional improvements. 4 Nevertheless, severe impairment in health is still often experienced by a substantial proportion of patients, who often find that their “own and their physicians’ treatment goals, expectations, and aspirations are seldom met.” 5 Additionally, medications can be associated with significant side effects, including weight gain and insulin resistance, further exacerbating cardiovascular disease risk and overall morbidity. 6

It is thus perhaps not surprising that in an attempt to relieve their symptoms, patients often seek dietary therapies with various outcomes and placebo effects. 7 The most popular dietary approaches include a vegetarian or gluten-free diet, Mediterranean diet (MD), elemental diet, elimination diet, or fasting, while dietary supplements, especially omega-3 polyunsaturated fatty acids (PUFAs), are also commonly used. In a Cochrane review on dietary interventions in RA published in 2009, it was concluded that effects were still uncertain as the existing studies were small. Moreover, the review included mostly single trials with moderate-to-high risk of bias, and high dropout rates and weight loss in the diet group, indicating that the potential adverse effects of dietary interventions for RA should not be ignored. 8 Since evidence is ambiguous, specific dietary guidelines for RA do not exist, with some countries such as Sweden recommending RA patients should follow healthy eating guidelines. 9

Recently, there has been a higher level of interest in the role of diet in RA, with rheumatologists urging clinicians not to neglect nutrition in these patients. 10 To the best of our knowledge, a systematic review examining all nutritional, dietary supplement, and fasting interventions along with RA outcomes has not been published to date, and thus we aimed to inform the literature to gain better insights into the role of diet in various aspects of the disease.

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines 11 were used to conduct this review ( Appendix S1 ; please see the Supporting Information online ), with prespecification of all methods prior to the literature search. A literature search was carried out with the support of an experienced librarian (C.R.) for studies assessing adherence to whole diet or dietary supplementation, whole-diet interventions or interventions with dietary supplementation, or fasting as exposures, and with outcomes examined comprising risk of developing RA, prevalence of RA, progression or regression of RA, or clinical outcomes for RA progression or regression. The comparison groups were no diet or dietary interventions, or other dietary supplements or combinations of supplements or nonfasting. A systematic search for relevant articles (yielding 1946 articles; 08 October 2018) was performed on Medline and Embase via the OvidSP platform using the following terms: rheumatoid arthritis or arthritis, rheumatoid or inflammatory arthritis and diet or diet therapy or dietary supplement(s) or dietary intake or nutrition or nutritional supplements or nutrient or nutrients or nutrient uptake or vitamin or vitamin uptake or fasting or diet restriction or “Diet, Food and Nutrition.” In addition, the reference lists of systematic and other review articles were handsearched. The protocol can be found in the Supporting Information online.

The PICOS (population, intervention, comparison, outcome, and study design) criteria for inclusion and exclusion are described in Table 1 . Inclusion criteria were as follows: human studies that examined the effects of adherence to, or intervention with, a specific diet or dietary supplementation, or fasting, on a defined indicator/indicators of RA prevalence or progression; randomized controlled trials; and prospective and retrospective observational studies. There was no restriction on sample size or participants’ age, sex, or health status. Studies examining individuals with or without signs of RA at baseline were also included. Exclusion criteria were as follows: systematic reviews, meta-analyses, narrative reviews, abstracts, conference reports, letters, commentaries, and opinions; studies with insufficient information to evaluate the effect of the diet or dietary supplementation on RA progression and/or prevalence; studies assessing only intake without relating it to RA risk or outcomes; studies assessing only serum concentration of micronutrients without reporting intake; studies not published in English; animal studies; and in vitro studies.

PICOS criteria for inclusion of studies

Abbreviation: RA, rheumatoid arthritis.

Eligibility assessment of the identified studies was conducted independently by 2 reviewers (E.P. and S.D.P.). If eligibility was disputed, this was discussed between the two reviewers and subsequently a third (E.N.) until a consensus was reached. Data extraction was carried out by 2 reviewers (E.P. and S.D.P.). The following data were extracted from each study: title, authors, year of publication, sample at baseline, sample completing study, percentage excluded or dropped out and reasons, mean age, percentage female, treatment at baseline, setting, method of participant recruitment, inclusion and exclusion criteria, control group characteristics at baseline (if different), severity of illness at baseline, comorbidities, study design, primary and other aims of study, intervention, control group intervention, duration of intervention, time(s) of assessment, diet assessment, RA outcomes assessed, other outcomes/comorbidities assessed, recorded dietary changes, outcome of intervention on RA outcomes, outcome of intervention on other outcomes/comorbidities, and authors’ conclusions. For the purpose of this publication, only data from clinical studies will be presented since they are more clinically relevant and represent the highest level of evidence. Data from case-control and cohort studies were systematically reviewed and recently published separately. 12 The PICOS criteria for inclusion and exclusion of clinical trials are described in Table 1 .

Quality assessment of clinical trials was based on the Cochrane Collaboration Risk of Bias 2.0 tool for randomized, parallel-group trials and the additional considerations for crossover trials where appropriate. 13 , 14 The following were considered: selection bias by assessing random sequence generation (including period effects for crossover trials only) and allocation concealment; performance bias by assessing blinding of participants and personnel (including washout period for crossover trials only); detection bias by assessing blinding of outcome assessment; attrition by assessing incomplete outcome reporting; and reporting bias by assessing selective or incomplete outcome reporting.

Study selection

As shown in Figure 1, 1 964 articles were identified through the Medline and Embase databases and handsearching relevant journals. After an initial review of titles and abstracts, 1844 articles were excluded, leaving 120 articles to be reviewed in full. Of these, 70 articles met the inclusion criteria and were progressed to full review and data extraction.

Flow diagram of the literature search process

Flow diagram of the literature search process

Study characteristics

The study characteristics are shown in Table S1 ( please see the Supporting Information online ) and arranged by type of intervention. It is noted that the earliest publication was in 1982. The number of participants was highly variable, ranging from 10 to 139, with the vast majority of studies involving 20–60 participants. Patient dropout or withdrawal also varied, being up to 20% in most studies, but reaching up to 48% in Beri et al’s study (1988) 15 and 62% in Kavanagh et al’s study (1995). 16 The mean participant age was 50–60 years, with the great majority of participants being female, as expected for RA patients. In the majority of studies, treatment for RA included steroids, nonsteroidal anti-inflammatory drugs, and conventional synthetic DMARDs (medication not shown in Table S1 in the Supporting Information online). All studies included RA patients diagnosed according to set criteria of the American College of Rheumatology (ACR) or the European League Against Rheumatism, and most studies were carried out in outpatient clinics or the community, with a few using an in-patient design.

The primary aims and inclusion and exclusion criteria of the included studies are shown in Table S2 ( please see the Supporting Information online ), and Table S3 in the Supporting Information online shows the design and intervention characteristics of the included studies. The duration of interventions varied from 7 days to 13 months, with most studies (n = 26) lasting more than 3 months. The great majority of supplement studies were double-blind randomized controlled trials, while the studies reporting on fasting or dietary interventions were either single-blind or non-blinded trials, and only 6 studies 17–22 had a crossover design.

Results of included studies in relation to rheumatoid arthritis laboratory parameters, clinical parameters, and parameters of remission

↓ Significant decrease.

/ No significant result.

↓↓/↑↑ Result was significantly stronger vs placebo as well as vs other intervention group.

≠ Morning stiffness longer in the control group; no change in the experimental group.

Only within group, ie compared with baseline.

Midway assessment.

Abbreviations: AID, anti-inflammatory diet; ALA, alpha-linolenic acid; CRP, C-reactive protein; DAS28, Disease Activity Score-28; DHA, docosahexaenoic acid; EPA, eicosapentaenoic acid; EPO, evening primrose oil; ESR, erythrocyte sedimentation rate; FO, fish oil; GLA, gamma-linolenic acid; IU, international units; SJC,wollen joint count; TJC, tender joint count; LA, linoleic acid; MS, morning stiffness count; NSAID, nonsteroidal anti-inflammatory drug; P, polyunsaturated; PUFA, polyunsaturated fatty acid; S, saturated; SFA, saturated fat; VAS, visual analog scale; WD, Western diet

Most supplement studies investigated the effects of fish oils (n = 17 studies) 21 , 23–38 and omega-3 fatty acid supplements (n = 9 studies), 19 , 22 , 39–45 7 studies investigated other supplements (evening primrose oil [n = 2], 46 , 47 vitamin D [n = 2], 48 , 49 vitamin K [n = 1], 50   or selenium [n = 1], 51 or a combination of various supplements [n = 1] 52 ), and 1 study assessed sodium restriction. 53 The dosage of supplementation used was highly variable, reaching up to 10 g/d for fish oil supplements. The remainder of the studies (n = 36) investigated fasting or dietary modifications. Fasting or calorie restriction, specifically, was investigated by 5 studies, 17 , 18 , 20 , 54 , 55 fasting and ketogenic diets by 1 study, 56 and Ramadan fasting by 1 study. 57 Several studies investigated the effects of whole-diet changes; these comprised a short period of fasting followed by a longer period of vegetarian feeding (n = 7), 58–64 a vegetarian diet (n = 2), 65 , 66 a vegan diet (n = 3), 67–69 or a gluten-free vegan diet (n = 2). 70 , 71 Dietary restrictions and/or dietary challenges based on possible allergenic foods were investigated by 6 studies 7 , 15 , 72–75 and elemental or peptide diets by 4 studies. 16 , 76–78 Finally, the effects of an MD were investigated by 4 studies 79–82 and a mixed dietary intervention by 1 study. 83 While some studies did not use a control group, where a control was used the focus was on other oils (eg, olive oil), or placebo supplements in supplement studies, or normal/usual diet in feeding studies.

Dietary assessment and compliance

In supplementation studies, compliance was assessed using pill counting and/or biomarkers, while dietary intervention studies used mostly diet diaries, 24-hour recalls, and/or dietetic consultations. In 14 studies, no dietary assessment was carried out or no information was provided. 7 , 15 , 17 , 18 , 26 , 30 , 34 , 39 , 48 , 49 , 51 , 57 , 64 , 78

Study outcomes

A general observation across all studies reviewed was that improvements in RA disease activity were most notable for the more subjective rather than objective components of the disease ( Table 2 23–83 ; Table S4 – please see the Supporting Information online ). Following is a description of the study outcomes in order of appearance in the tables, starting with dietary supplementation studies.

In studies involving administration of omega-3 fatty acids, fish oil, or other oils, there appeared to be a reduction in RA disease activity; however, improvements were most notable in subjective rather than objective components of the disease (eg, tender joint count, patient and physician global assessment, and pain scores). For example, significant decreases in Disease Activity Score-28 (DAS28) in the fish oil group, 25 and the groups taking n-3 PUFA and evening primrose oil, were demonstrated, as well as a reduction in the number of painful joints and pain assessed by visual analog scales. 38 Some studies reported a significant difference in morning stiffness between treatment and control groups, 19 , 34 and in a well-designed study by Proudman et al (2015b), 32 a reduction in the rates of DMARD failure or higher rates of remission according to ACR criteria were demonstrated. Specifically, in a time-to-event analysis, the same study demonstrated a lower failure rate of triple DMARD therapy and decreased time to first ACR remission in the fish oil group, compared with the control group. 32

Higher fish oil doses (eg, 130 mg of n-3 PUFA/kg/d, as used in the study by Kremer et al 1995 27 ) demonstrated improvements in inflammatory biomarkers, including erythrocyte sedimentation rate and C-reactive protein (CRP) but also neutrophil B and macrophage interleukin-1 production. At these higher doses, some studies reported the ability of patients to reduce or discontinue nonsteroidal anti-inflammatory drugs or analgesics, even in studies where no statistically significant differences between groups in patient-reported and other disease outcomes were observed. 25 , 29 , 34 , 35 , 46 , 47 In high-dose fish oil groups, significant improvements in physician evaluation of global arthritis activity were sustained up to 24 weeks, and with a carryover effect significant improvement at 30 weeks was observed. 27 In a study by Geusens et al (1994), 40 significant improvements in patient and physician global assessments were demonstrated only in those taking 2.6 g/d of n-3 PUFA, while lower fish oil doses (<30 mg of n-3 PUFA/kg/d, as used by Kremer et al 27 ) were most beneficial only on subjective measures of disease.

In other supplementation studies, which were investigated by only 1 or 2 studies in each case, only vitamin D supplementation and dietary sodium restriction showed some possible improvements in RA outcomes. In particular, Buondonno et al (2017) 48 demonstrated a role of vitamin D in RA, with significantly lower serum 25-hydroxy vitamin D concentration in the patients with persisting active disease compared with patients with low disease activity or remission at baseline. Patients with low vitamin D also appeared to have prolonged disease duration, although no correlation was seen between duration of illness and vitamin D concentration. Importantly, vitamin D supplementation in these patients resulted in significant improvements in parameters of disease activity, findings that were in line with the study by Chandrashekara and Patted, 49 which also assessed vitamin D supplementation.

Sodium restriction was assessed in only one study included in this review and it resulted in a reduction in the pro-inflammatory response, as shown by significant reductions in the cytokines transforming growth factor beta-1 and interleukin-9. 53

Contrarily, supplementation with antioxidants, as studied by Bae et al (2009), 52 did not result in any improvements in inflammatory biomarkers or disease severity in RA patients, with the authors suggesting that this may be due to disease activity being mild at baseline and/or low bioavailability of the antioxidants used. Lastly, vitamin K or selenium supplementation, as studied by Shishavan et al (2016) 50 and Tarp et al (1985), 51 respectively, had no effect on RA outcomes.

Studies addressing different types of fasting or calorie restriction suggested overall improvements in global patient assessments, morning stiffness, and number of painful joints, with some contradictory evidence on the effects on inflammation markers such as erythrocyte sedimentation rate. Different types of fasting elicited different responses, as shown in Table 2 . “Subtotal fasting” – providing 800–1260 kJ/d for 7–10 days – had a positive impact through the reduction in number of swollen joints, pain score, Ritchie articular index, and tender joints, all of which were sustained by a vegetarian diet for 1 year. 58 “Short subtotal fast” followed by a vegetarian diet for longer periods (eg, over 1 year) appeared to reduce not only pain and morning stiffness but also inflammatory biomarkers such as leukocyte counts, immunoglobulin M, complement levels, and rheumatoid factor. 60

Trials involving vegetarian and vegan diets, with the latter tested in several studies, suggested that responses are very individualized and mainly depend on food intolerances/allergies. Reductions in DAS28 and morning stiffness were noted, as were reductions in concentrations of CRP and rheumatoid factor. 67 In the Hafström et al study, 71 a vegan and gluten-free diet led to reductions in immunoglobulin G against gliadin and also β-lactoglobulin. Moreover, in the study by Elkan et al (2008), 70 DAS28 and Health Assessment Questionnaire scores were significantly reduced at both 3 and 12 months in the vegan group compared with baseline, and CRP was reduced at 12 months compared with baseline. In line with these findings, in the Hafström et al (2001) 71 study the response rate as defined by the ACR 20% improvement criteria, for individual patients was higher in the vegan diet group than in the nonvegan diet group.

In the case of elimination diet or dietary restriction studies, whereby the most commonly eliminated foods were dairy, egg, meat, fish, refined sugars, wheat, corn, nuts, citrus fruits, and coffee, among others, there appeared to be improvements in the number of tender joints but also in inflammatory biomarkers during the elimination phase, such as in erythrocyte sedimentation rate, CRP, tumor necrosis factor, and interleukin-1 beta. 7 , 74 Food challenges appeared to increase inflammatory biomarkers, although this observation seemed more pronounced in prick-positive groups or those with a history of allergies/intolerances. 72 , 73

With regards to elemental and peptide diets (ingestion of liquid nutrients in an easily assimilated form), studies revealed reduced subjective symptoms including morning stiffness and pain using the visual analog scale; improvements, however, were only transient. No improvements in objective measures, such as inflammatory biomarkers, were seen. 16 , 65 , 75 , 77 , 78 It is also not clear whether these effects were due to the dietary intervention itself or to the resulting associated weight loss, since a statistically significant correlation in the diet group between weight loss and grip strength was observed.

Studies on the Mediterranean diet demonstrated improvements in subjective measures of RA disease activity such as patient global assessment, pain score, early morning stiffness, and composite DAS28 80–82 (the latter likely driven by improvements in the subjective components of the score). Improvements in physical function and vitality have also been shown, as well as in objective measures of inflammation such as CRP, although the latter was a limited finding with no universal improvement in laboratory measures across all studies. Generally, improvements in stable and modestly active RA were seen, although evidence was limited – with only 3 MD-related studies included in this review. 80–82

Owing to the heterogeneity of the included studies, a meta-analysis could not be conducted.

Sources of bias

The studies eligible for inclusion in this systematic review were very heterogeneous and had various sources of bias. Table S1 ( please see the Supporting Information online ) shows participant characteristics, including patient age and sex of participants (were not clearly described in 8 studies 7 , 15 , 31 , 39 , 44 , 64 , 71 , 73 ; treatment at baseline – including medication dosage and percentages receiving each treatment – was not clear in many studies (especially in older publications; not cited). Additionally, the majority of participants (in most studies >70%) were women, possibly reflecting some selection bias suggesting that more women may show interest, or be willing to take part, in dietary intervention studies. Also, some publications report different outcomes for the same intervention. For example, Kjeldsen-Kragh et al (1994, 1995a, 1995b, 1996), 59–61 , 66 Haugen et al (1994), 76 and Peltonen et al (1994) 62 noted either different outcomes or outcomes assessed at different time-points in the Kjeldsen-Kragh et al (1991) 58 study, Elkan et al 70 observed different outcomes for the intervention described by Hafström et al, 71 and Karatay et al (2006) 73 noted different outcomes to those observed by Karatay et al (2004). 72 Thus, of the 70 papers, the above 8 papers refer to the same interventions, so the review included only 62 unique interventions.

Concerning dietary assessments or dietary compliance assessments, 14 of the 70 studies 7 , 15 , 17 , 18 , 26 , 30 , 34 , 39 , 48 , 49 , 51 , 57 , 64 , 78 either did not provide any information or did not carry out any assessment, and in many studies only subjective assessment methods were used, such as capsule counting, patient reporting intake, or diet diaries, thus introducing a high risk of bias.

With regard to the control groups, 9 studies 19 , 21 , 24 , 27 , 35 , 37 , 40 , 44 , 47 – most of which were published >25 years ago – used olive oil, corn oil, or coconut oil capsules as “placebo” treatments, which were possibly inappropriate since such oils exert anti-inflammatory effects, while 10 studies 15 , 36 , 49 , 53 , 56 , 67 , 72 , 73 , 75 , 79 did not include a control group. Additionally, as is true in all dietary intervention studies, it was not possible to blind participants taking part in studies assessing diet or fasting.

A quality and bias risk assessment is shown in Supplementary Table S5 ( please see the Supporting Information online ). In most of the studies, there was insufficient information on how randomization to treatment groups was carried out, resulting in unclear selection bias. Blinding of participants and personnel was highly variable and depended on the intervention. Most dietary supplement studies had a double-blind design and thus a low risk of bias, while dietary intervention, elimination, or fasting studies were highly biased owing to the inability to blind participants. Even though this source of bias could have been reduced by blinding personnel and the outcome assessment, this was not attempted in most studies. Attrition was moderate in most studies, with a dropout rate of about 20%, but was inappropriately high in 2 studies. 15 , 16 Nevertheless, most studies reported on (valid) reasons for dropouts. Additionally, several studies had moderate-to-high reporting bias as they inappropriately emphasized within-group rather than between-group analysis of outcomes and/or did not account for patients who dropped out or were withdrawn from the study. Finally, many studies reported on outcomes of the patients retained in the study rather than using an intention-to-treat analysis.

To the best of our knowledge, this is the first systematic review to include all dietary, supplement, and fasting interventions on RA and assess their effects on outcomes. Although the studies assessed were highly heterogeneous with multiple sources of bias, some key conclusions can be drawn.

Trials on the MD – characterized by an abundance of plant foods such as unrefined cereals, fruit, vegetables, legumes, and extra-virgin olive oil; moderate consumption of poultry, dairy products, and eggs; and low consumption of sweets and red meat – demonstrated overall modest improvements in stable and modestly active RA. In a recently published review of clinical trials on the MD and omega-3 PUFA supplementation and RA outcomes, 84 it was reported that consumption of the MD supplemented with omega-3 PUFAs seems promising for reducing inflammation and reduces risk of comorbidities such as cardiovascular disease, thereby lessening future disease-related complications. 85

With regards to other dietary interventions, elemental and peptide diets – typically composed of amino acids, fats, sugars, vitamins, and minerals – were tested in limited, low-quality trials, of short duration, with small study numbers and high dropout rates. The results thus do not generally encourage the use of elemental diets, although such diets may be suitable for people with specific dietary allergies (supporting the theory that RA may be exacerbated by food antigens).

Studies of vegetarian/vegan diets resulted in improvements in a combination of both subjective measures (eg, morning stiffness) and objective measures (eg, CRP) of the disease. The benefits may relate to improvements in the fecal flora, reduced immune responses to food allergens in gut microflora (reduced antigenic challenge), or reduced meat-induced gut inflammation.

Fasting studies show evidence that calorie restriction can ameliorate RA activity, with more pronounced effects on subjective disease symptoms. Total fasting for specified periods was the most effective type of fast; however, such fasting is unsustainable or even impossible in some cases – and thus we would not recommend it. Improvements noted with periods of fasting followed by a vegetarian diet or an elimination diet may have reflected changes in the intestinal flora as a key underlying mechanism. Indeed, as recently reviewed by our group, 86 dietary factors may act as environmental triggers in genetically susceptible individuals, leading to the development of RA. The resulting inflammatory process encompasses a cascade of events and includes the production of cytokines and chemokines by joint tissue cells, leading to joint damage and CRP-enhanced inflammation. 87 Red meat, salt, or even excessive food intake may act as triggers, as mentioned above, while in contrast, fish, fruit, and vegetables act as “moderators,” reducing inflammation. 88 , 89 It should be noted that dietary therapy may evoke a placebo response that leads to improved well-being and a decrease in psychological distress. Yet, the effects of elimination diets may also be associated with increased permeability of the intestinal mucosa to allergens in RA, inducing an inflammatory response. In fact, the so-called “unknown triad” of “diet, microbiota, and gut permeability” is the focus of recent reviews such as that of Guerreiro et al, 90 who proposed that chronic inflammatory response induced by gut dysbiosis can critically contribute to the development of rheumatic diseases. Thus, it may be possible that diets eliminating food groups that trigger an allergic or inflammatory response result in reduced inflammation by altering the gut microbiota population. In line with these outcomes, probiotics have been shown to lower pro-inflammatory cytokines. 91

With regards to omega-3 PUFA supplementation, low doses were not generally found to be of clinical significance in the study by Philippou and Nikiphorou, 86 but higher doses (≥2.7 g/d) – as supported by a meta-analysis conducted by Lee et al 92 – could help reduce the daily requirement for nonsteroidal anti-inflammatory drugs and concomitant analgesic consumption.

The observed associations between low vitamin D concentration and more active disease, as well as the benefits of vitamin D supplementation, were not surprising, as previously demonstrated in the study by Nikiphorou et al. 93 Correction of vitamin D concentration in those deficient is imperative owing to the central role of vitamin D in calcium homeostasis, bone mineralization, and general preservation of musculoskeletal health, and thus should be advocated. Lastly, the observed benefits of sodium restriction on RA inflammation were also as expected, since sodium chloride activates pro-inflammatory macrophages and T helper 17 cells and decreases regulatory T cells, as recently reviewed. 10

This systematic review has several strengths. To the best of our knowledge, this is the first study to include all dietary, supplement, and fasting interventions on RA and assess effects on outcomes. Prior to this, a Cochrane review published in 2009 8 included only 15 studies, and several other published systematic reviews or meta-analyses on diet and RA concentrated on only one specific dietary intervention or supplement. 85 , 94

There are also limitations to our review that need to be addressed. Firstly, many of the presented studies have methodological flaws. Several studies are based on only a small sample size and/or have a high dropout rate and did not use an intention-to-treat analysis. Secondly, the duration of the clinical trials, with some exceptions, was generally short and did not allow important clinical outcomes to be studied. Furthermore, the biological plausibility of the potentially expected findings from the dietary intervention and true (nonplacebo) clinical effects could not be disentangled. Placebo effects were high in many studies, while in some studies inappropriate ‘placebo’ treatments such as olive oil were used, which themselves have anti-inflammatory properties. Additionally, several studies reported greater effects on the more subjective elements of the disease such as TJS, with less evidence on objective features of active RA such as laboratory markers or radiographic progression. Nevertheless, it is of note that subjective measures of the disease are of particular importance to RA patients, especially when in terms of overall well-being and pain levels. Objective measures, however – traditionally, inflammatory markers – are of greater importance in terms of “intervention success.” Lastly, a meta-analysis of findings could not be conducted owing to the heterogeneity of the included studies.

To sum up, based on the available evidence, and as also suggested in our previous work, 84 we advocate to use the MD and fish oil supplements in conjunction (but not as a replacement) of other non-dietary interventions and pharmacotherapy, as and where indicated. There is uncertainty and moderate-high risk of bias with regards to the use, effectiveness, safety or even need of other dietary interventions such as fasting, vegetarian, elemental or elimination diets and other dietary supplements which might lead to unnecessary restrictions or dietary costs or even social isolation without objective benefits, with the exception of potential benefits to some individual patients. In this case, it is recommended that any elimination of foods is done under dietetic supervision to reduce the risk of an unbalanced or restricted macro- or micronutrient intake. By contrast, the MD is a well-balanced, non-restrictive diet with anti-inflammatory, antioxidant and cardio-protective properties.

Based on the evidence reviewed, we also recommend a reduction in salt intake to <5 g/d, as supported by the World Health Organization’s Healthy Eating Guidelines, 95 and routine assessment of vitamin D3 concentration as well as supplementation where necessary. There remains the need for further large, well-controlled randomized controlled trials on plausible and realistic dietary interventions, such as the MD with additional “fatty” fish or omega-3 PUFA supplements, in order to assess both objective and subjective RA outcomes. Evidently, diet may improve the symptoms of RA by reducing inflammation, either by removing pro-inflammatory foods or increasing anti-inflammatory foods and altering the gut microbiota. It is thus recommended that nutrition should be routinely addressed in RA patients by referral to registered dietitians, who are experts in identifying and addressing nutrition-related issues.

Author contributions . The authors’ responsibilities were as follows—all authors designed the research; C.R. conducted all searches, E.P. and S.D.P.: Extracted all data with the input of E.N. and C.R. were necessary; E.P. and E.N. critically reviewed the data and wrote the manuscript and all authors read and approved the final paper.

Funding . No financial support has been provided.

Declaration of interest . The authors declare no conflicts of interest.

Supporting information

The following Supporting Information is available through the online version of this article at the publisher’s website.

Appendix S1   PRISMA checklist

Table S1   Participants’ characteristics and sample completing study

Table S2   Primary aim and inclusion and exclusion criteria of the included studies

Table S3   Design and intervention characteristics of included studies

Table S4   Results of supplementary assessments (if any)

Table S5   Quality and bias risk assessment

Cross M , Smith E , Hoy D , et al. . The global burden of rheumatoid arthritis: estimates from the global burden of disease 2010 study . Ann Rheum Dis.   2014 ; 73 : 1316 – 1322 .

Google Scholar

Malm K , Bergman S , Andersson ML , et al. . Quality of life in patients with established rheumatoid arthritis: a phenomenographic study . SAGE Open Med . 2017 ; 5 : 205031211771364 .

Lindhardsen J , Ahlehoff O , Gislason GH , et al. . The risk of myocardial infarction in rheumatoid arthritis and diabetes mellitus: a Danish nationwide cohort study . Ann Rheum Dis.   2011 ; 70 : 929 – 934 .

Scott DL , Wolfe F , Huizinga T.   Rheumatoid arthritis . Lancet . 2010 ; 376 : 1094 – 1108 .

Taylor PC , Moore A , Vasilescu R , et al. . A structured literature review of the burden of illness and unmet needs in patients with rheumatoid arthritis: a current perspective . Rheumatol Int.   2016 ; 36 : 685 – 695 .

Radhakutty A , Mangelsdorf BL , Drake SM , et al. . Effects of prednisolone on energy and fat metabolism in patients with rheumatoid arthritis: tissue‐specific insulin resistance with commonly used prednisolone doses . Clin Endocrinol . 2016 ; 85 : 741 .

Darlington L , Ramsey N , Mansfield J.   Placebo-controlled, blind study of dietary manipulation therapy in rheumatoid arthritis . Lancet . 1986 ; 327 : 236 – 238 .

Hagen KB , Byfuglien MG , et al. . Dietary interventions for rheumatoid arthritis . Cochrane Database Syst Rev ; 2009 . Available at: https://doi.org//10.1002/14651858.CD006400.pub2 . Accessed June 18, 2019.

Livsmedelsmerket National Food Agency. Find your way to eat greener. http://www.fao.org/3/a-az854e.pdf . Accessed April 12, 2018.

Cutolo M , Nikiphorou E.   Don’t neglect nutrition in rheumatoid arthritis!   RMD Open . 2018 ; 4 : e000591 .

Moher D , Liberati A , Tetzlaff J , et al. . Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement . Ann Intern Med.   2009 ; 151 : 264 – 269 .

Philippou E , Petersson S , Erodotou S , et al. . Dietary intake, dietary interventions, nutrient supplements and rheumatoid arthritis: systematic review of the evidence . Proc Nutr Soc . 2019 ; 78 :E21.

Higgins J. Revised Cochrane risk of bias tool for randomized trials (RoB 2.0): additional considerations for cluster-randomized trials. https://www.riskofbias.info/welcome/rob-2-0-tool/archive-rob-2-0-cross-over-trials-2016 . Accessed April 4, 2019.

Higgins J. Revised Cochrane risk of bias tool for randomized trials (RoB 2.0): additional considerations for cross-over trials. https://www.riskofbias.info/welcome/rob-2-0-tool/archive-rob-2-0-cross-over-trials-2016 . Accessed December 8, 2018.

Beri D , Malaviya AN , Shandilya R , et al. . Effect of dietary restrictions on disease activity in rheumatoid arthritis . Ann Rheum Dis . 1988 ; 47 : 69 – 72 .

Kavanagh R , Workman E , Nash P , et al. . The effects of elemental diet and subsequent food reintroduction on rheumatoid arthritis . Br J Rheumatol . 1995 ; 34 : 270 – 273 .

Trang LE , Lövgren O , Bendz R , et al. . The effect of fasting on plasma cyclic adenosine-3′, 5′-monophosphate in rheumatoid arthritis . Scand J Rheumatol . 1980 ; 9 : 229 – 233 .

Uden AM , Trang L , Venizelos N , et al. . Neutrophil functions and clinical performance after total fasting in patients with rheumatoid arthritis . Ann Rheum Dis . 1983 ; 42 : 45 – 51 .

Kremer J , Jubiz W , Michalek A , et al. . Fish oil fatty acid supplementation in active rheumatoid arthritis . Ann Intern Med.   1987 ; 106 : 497 – 503 .

Hafström I , Ringertz B , Gyllenhammar H , et al. . Effects of fasting on disease activity, neutrophil function, fatty acid composition, and leukotriene biosynthesis in patients with rheumatoid arthritis . Arthritis Rheum . 1988 ; 31 : 585 – 592 .

van der Tempel H , Tulleken JE , Limburg PC , et al. . Effects of fish oil supplementation in rheumatoid arthritis . Ann Rheum Dis . 1990 ; 49 : 76 – 80 .

Dawczynski C , Schubert R , Hein G , et al. . Long-term moderate intervention with n-3 long-chain PUFA-supplemented dairy products: effects on pathophysiological biomarkers in patients with rheumatoid arthritis . Br J Nutr.   2009 ; 101 : 1517 – 1526 .

Adam O , Beringer C , Kless T , et al. . Anti-inflammatory effects of a low arachidonic acid diet and fish oil in patients with rheumatoid arthritis . Rheumatol Int.   2003 ; 23 : 27 – 36 .

Cleland LG , French JK , Betts WH , et al. . Clinical and biochemical effects of dietary fish oil supplements in rheumatoid arthritis . J Rheumatol.   1988 ; 15 : 1471 – 1475 .

Galarraga B , Ho M , Youssef H , et al. . Cod liver oil (n-3 fatty acids) as an non-steroidal anti-inflammatory drug sparing agent in rheumatoid arthritis . Rheumatology (Oxford) . 2008 ; 47 : 665 – 669 .

Kolahi S , Ghorbanihaghjo A , Alizadeh S , et al. . Fish oil supplementation decreases serum soluble receptor activator of nuclear factor-kappa B ligand/osteoprotegerin ratio in female patients with rheumatoid arthritis . Clin Biochem . 2010 ; 43 : 576 – 580 .

Kremer JM , Lawrence DA , Jubiz W , et al. . Dietary fish oil and olive oil supplementation in patients with rheumatoid arthritis: clinical and immunologic effects . Arthritis Rheum . 1990 ; 33 : 810 – 820 .

Kremer JM , Lawrence DA , Petrillo GF , et al. . Effects of high‐dose fish oil on rheumatoid arthritis after stopping nonsteroidal antiinflammatory drugs: clinical and immune correlates . Arthritis Rheum . 1995 ; 38 : 1107 – 1114 .

Lau C , Morley K , Belch J.   Effects of fish oil supplementation on non-steroidal anti-inflammatory drug requirement in patients with mild rheumatoid arthritis – a double-blind placebo controlled study . Br J Rheumatol . 1993 ; 32 : 982 – 989 .

Magaro M , Altomonte L , Zoli A , et al. . Influence of diet with different lipid composition on neutrophil chemiluminescence and disease activity in patients with rheumatoid arthritis . Ann Rheum Dis . 1988 ; 47 : 793 – 796 .

Nielsen G , Faarvang K , Thomsen B , et al. . The effects of dietary supplementation with n‐3 polyunsaturated fatty acids in patients with rheumatoid arthritis: a randomized, double blind trial . Eur J Clin Invest . 1992 ; 22 : 687 – 691 .

Proudman SM , Cleland LG , Metcalf RG , et al. . Plasma n-3 fatty acids and clinical outcomes in recent-onset rheumatoid arthritis . Br J Nutr.   2015 ; 114 : 885 – 890 .

Proudman SM , James MJ , Spargo LD , et al. . Fish oil in recent onset rheumatoid arthritis: a randomised, double-blind controlled trial within algorithm-based drug use . Ann Rheum Dis.   2015 ; 74 : 89 – 95 .

Rajaei E , Mowla K , Ghorbani A , et al. . The effect of omega-3 fatty acids in patients with active rheumatoid arthritis receiving DMARDs therapy: double-blind randomized controlled trial . Glob J Health Sci.   2015 ; 8 : 18 – 25 .

Sköldstam L , Börjesson O , Kjällman A , et al. . Effect of six months of fish oil supplementation in stable rheumatoid arthritis: a double-blind, controlled study . Scand J Rheumatol . 1992 ; 21 : 178 – 185 .

Sperling RI , Weinblatt M , Robin J , et al. . Effects of dietary supplementation with marine fish oil on leukocyte lipid mediator generation and function in rheumatoid arthritis . Arthritis Rheum . 1987 ; 30 : 988 – 997 .

Tulleken JE , Limburg PC , Muskiet FA , et al. . Vitamin E status during dietary fish oil supplementation in rheumatoid arthritis . Arthritis Rheum . 1990 ; 33 : 1416 – 1419 .

Veselinovic M , Vasiljevic D , Vucic V , et al. . Clinical benefits of n-3 PUFA and ɤ-linolenic acid in patients with rheumatoid arthritis . Nutrients . 2017 ; 9 : 325 .

Espersen G , Lervang H , Nielsen G , et al. . Decreased interleukin-1 beta levels in plasma from rheumatoid arthritis patients after dietary supplementation with n-3 polyunsaturated fatty acids . Clin Rheumatol.   1992 ; 11 : 393 – 395 .

Geusens P , Wouters C , Nijs J , et al. . Long‐term effect of omega‐3 fatty acid supplementation in active rheumatoid arthritis . Arthritis Rheum . 1994 ; 37 : 824 – 829 .

Kremer J , Michalek A , Lininger L , et al. . Effects of manipulation of dietary fatty acids on clinical manifestations of rheumatoid arthritis . Lancet . 1985 ; 325 : 184 – 187 .

Leventhal LJ , Boyce EG , Zurier RB.   Treatment of rheumatoid arthritis with gammalinolenic acid . Ann Intern Med.   1993 ; 119 : 867 – 873 .

Mirtaheri E , Pourghassem Gargari B , Kolahi S , et al. . Effects of alpha-lipoic acid supplementation on inflammatory biomarkers and matrix metalloproteinase-3 in rheumatoid arthritis patients . J Am Coll Nutr . 2015 ; 34 : 310 – 317 .

Nordström D , Friman C , Konttinen Y , et al. . Alpha-linolenic acid in the treatment of rheumatoid arthritis. A double-blind, placebo-controlled and randomized study: flaxseed vs. safflower seed . Rheumatol Int.   1995 ; 14 : 231 – 234 .

Remans P , Sont J , Wagenaar L , et al. . Nutrient supplementation with polyunsaturated fatty acids and micronutrients in rheumatoid arthritis: clinical and biochemical effects . Eur J Clin Nutr.   2004 ; 58 : 839 – 845 .

Belch JJ , Ansell D , Madhok R , et al. . Effects of altering dietary essential fatty acids on requirements for non-steroidal anti-inflammatory drugs in patients with rheumatoid arthritis: a double blind placebo controlled study . Ann Rheum Dis . 1988 ; 47 : 96 – 104 .

Brzeski M , Madhok R , Capell H.   Evening primrose oil in patients with rheumatoid arthritis and side-effects of non-steroidal anti-inflammatory drugs . Br J Rheumatol . 1991 ; 30 : 370 – 372 .

Buondonno I , Rovera G , Sassi F , et al. . Vitamin D and immunomodulation in early rheumatoid arthritis: a randomized double-blind placebo-controlled study . PloS One.   2017 ; 12 : e0178463 .

Chandrashekara S , Patted A.   Role of vitamin D supplementation in improving disease activity in rheumatoid arthritis: an exploratory study . Int J Rheum Dis.   2017 ; 20 : 825 – 831 .

Shishavan NG , Gargari BP , Kolahi S , et al. . Effects of vitamin K on matrix metalloproteinase-3 and rheumatoid factor in women with rheumatoid arthritis: a randomized, double-blind, placebo-controlled trial . J Am Coll Nutr . 2016 ; 35 : 392 – 398 .

Tarp U , Overvad K , Thorling EB , et al. . Selenium treatment in rheumatoid arthritis . Scand J Rheumatol . 1985 ; 14 : 364 – 368 .

Bae S , Jung W , Lee E , et al. . Effects of antioxidant supplements intervention on the level of plasma inflammatory molecules and disease severity of rheumatoid arthritis patients . J Am Coll Nutr . 2009 ; 28 : 56 – 62 .

Scrivo R , Massaro L , Barbati C , et al. . The role of dietary sodium intake on the modulation of T helper 17 cells and regulatory T cells in patients with rheumatoid arthritis and systemic lupus erythematosus . PloS One.   2017 ; 12 : e0184449 .

Iwashige K , Kouda K , Kouda M , et al. . Calorie restricted diet and urinary pentosidine in patients with rheumatoid arthritis . J Physiol Anthropol Appl Human Sci.   2004 ; 23 : 19 – 24 .

Sköldstam L , Jorfelt L , Lindell B , et al. . Specific plasma proteins as indices of inflammation during a modified fast in patients with rheumatoid arthritis . Scand J Rheumatol . 1983 ; 12 : 161 – 165 .

Fraser D , Thoen J , Dioseland O , et al. . Serum levels of interleukin-6 and dehydroepiandrosterone sulphate in response to either fasting or a ketogenic diet in rheumatoid arthritis patients . Clin Exp Rheumatol . 2000 ; 18 : 357 – 362 .

Al-Dubeikil KY , Abdul-Lateef WK.   Ramadan fasting and rheumatoid arthritis . Bahrain Med Bull . 2003 ; 25 : 1 – 7 .

Kjeldsen-Kragh J , Borchgrevink C , Laerum E , et al. . Controlled trial of fasting and one-year vegetarian diet in rheumatoid arthritis . Lancet . 1991 ; 338 : 899 – 902 .

Kjeldsen-Kragh J , Mellbye O , Haugen M , et al. . Changes in laboratory variables in rheumatoid arthritis patients during a trial of fasting and one-year vegetarian diet . Scand J Rheumatol . 1995 ; 24 : 85 – 93 .

Kjeldsen-Kragh J , Hvatum M , Haugen M , et al. . Antibodies against dietary antigens in rheumatoid arthritis patients treated with fasting and a one-year vegetarian diet . Clin Exp Rheumatol . 1995 ; 13 : 167 – 172 .

Kjeldsen-Kragh J , Sumar N , Bodman-Smith K , et al. . Changes in glycosylation of IgG during fasting in patients with rheumatoid arthritis . Br J Rheumatol . 1996 ; 35 : 117 – 119 .

Peltonen R , Kjeldsen-Kragh J , Haugen M , et al. . Changes of faecal flora in rheumatoid arthritis during fasting and one-year vegetarian diet . Br J Rheumatol . 1994 ; 33 : 638 – 643 .

Sköldstam L , Larsson L , Lindström FD.   Effects of fasting and lactovegetarian diet on rheumatoid arthritis . Scand J Rheumatol . 1979 ; 8 : 249 – 255 .

Sundqvist T , Lindström F , Magnusson K , et al. . Influence of fasting on intestinal permeability and disease activity in patients with rheumatoid arthritis . Scand J Rheumatol . 1982 ; 11 : 33 – 38 .

Haugen MA , Kjeldsen-Kragh J , Førre O.   A pilot study of the effect of an elemental diet in the management of rheumatoid arthritis . Clin Exp Rheumatol . 1994 ; 12 : 275 – 279 .

Kjeldsen-Kragh J , Haugen M , Borchgrevink C , et al. . Vegetarian diet for patients with rheumatoid arthritis – status: two years after introduction of the diet . Clin Rheumatol.   1994 ; 13 : 475 – 482 .

McDougall J , Bruce B , Spiller G , et al. . Effects of a very low-fat, vegan diet in subjects with rheumatoid arthritis . J Altern Complement Med . 2002 ; 8 : 71 – 75 .

Nenonen MT , Helve TA , Rauma AL , et al. . Uncooked, lactobacilli-rich, vegan food and rheumatoid arthritis . Br J Rheumatol . 1998 ; 37 : 274 – 281 .

Peltonen R , Nenonen M , Helve T , et al. . Faecal microbial flora and disease activity in rheumatoid arthritis during a vegan diet . Br J Rheumatol . 1997 ; 36 : 64 – 68 .

Elkan A , Sjöberg B , Kolsrud B , et al. . Gluten-free vegan diet induces decreased LDL and oxidized LDL levels and raised atheroprotective natural antibodies against phosphorylcholine in patients with rheumatoid arthritis: a randomized study . Arthritis Res Ther.   2008 ; 10 : R34 .

Hafström I , Ringertz B , Spångberg A , et al. . A vegan diet free of gluten improves the signs and symptoms of rheumatoid arthritis: the effects on arthritis correlate with a reduction in antibodies to food antigens . Rheumatology (Oxford).   2001 ; 40 : 1175 – 1179 .

Karatay S , Erdem T , Yildirim K , et al. . The effect of individualized diet challenges consisting of allergenic foods on TNF-α and IL-1β levels in patients with rheumatoid arthritis . Rheumatology (Oxford) . 2004 ; 43 : 1429 – 1433 .

Karatay S , Erdem T , Kiziltunc A , et al. . General or personal diet: the individualized model for diet challenges in patients with rheumatoid arthritis . Rheumatol Int.   2006 ; 26 : 556 – 560 .

Sarzi-Puttini P , Comi D , Boccassini L , et al. . Diet therapy for rheumatoid arthritis: a controlled double-blind study of two different dietary regimens . Scand J Rheumatol.   2000 ; 29 : 302 – 307 .

van de Laar MA , van der Korst JK.   Food intolerance in rheumatoid arthritis. I. A double blind, controlled trial of the clinical effects of elimination of milk allergens and azo dyes . Ann Rheum Dis . 1992 ; 51 : 298 – 302 .

Haugen MA , Kjeldsen-Kragh J , Bjervea KS , et al. . Changes in plasma phospholipid fatty acids and their relationship to disease activity in rheumatoid arthritis patients treated with a vegetarian diet . Br J Nutr.   1994 ; 72 : 555 – 566 .

Holst-Jensen M , Pfeiffer-Jensen M , Monsrud M , et al. . Treatment of rheumatoid arthritis with a peptide diet: a randomized, controlled trial . Scand J Rheumatol . 1998 ; 27 : 329 – 336 .

Podas T , Nightingale JM , Oldham R , et al. . Is rheumatoid arthritis a disease that starts in the intestine? A pilot study comparing an elemental diet with oral prednisolone . Postgrad Med J . 2007 ; 83 : 128 – 131 .

Abendroth A , Michalsen A , Lüdtke R , et al. . Changes of intestinal microflora in patients with rheumatoid arthritis during fasting or a Mediterranean diet . Forsch Komplementmed.   2010 ; 17 : 307 – 313 .

Hagfors L , Leanderson P , Sköldstam L , et al. . Antioxidant intake, plasma antioxidants and oxidative stress in a randomized, controlled, parallel, Mediterranean dietary intervention study on patients with rheumatoid arthritis . Nutr J.   2003 ; 2 : 5 .

McKellar G , Morrison E , McEntegart A , et al. . A pilot study of a Mediterranean-type diet intervention in female patients with rheumatoid arthritis living in areas of social deprivation in Glasgow . Ann Rheum Dis . 2007 ; 66 : 1239 – 1243 .

Sköldstam L , Hagfors L , Johansson G.   An experimental study of a Mediterranean diet intervention for patients with rheumatoid arthritis . Ann Rheum Dis . 2003 ; 62 : 208 – 214 .

Hansen G , Nielsen L , Kluger E , et al. . Nutritional status of Danish rheumatoid arthritis patients and effects of a diet adjusted in energy intake, fish-meal, and antioxidants . Scand J Rheumatol . 1996 ; 25 : 325 – 333 .

Petersson S , Philippou E , Rodomar C , et al. . The Mediterranean diet, fish oil supplements and Rheumatoid arthritis outcomes: evidence from clinical trials . Autoimmun Rev . 2018 ; 17 : 1105 – 1114 .

Forsyth C , Kouvari M , D’Cunha NM , et al. . The effects of the Mediterranean diet on rheumatoid arthritis prevention and treatment: a systematic review of human prospective studies . Rheumatol Int.   2018 ; 38 : 737 – 747 .

Philippou E , Nikiphorou E.   Are we really what we eat? Nutrition and its role in the onset of rheumatoid arthritis . Autoimmun Rev . 2018 ; 17 : 1074 – 1077 .

Demoruelle MK , Deane KD , Holers VM.   When and where does inflammation begin in rheumatoid arthritis?   Curr Opin Rheumatol.   2014 ; 26 : 64 – 71 .

Oliviero F , Spinella P , Fiocco U , et al. . How the Mediterranean diet and some of its components modulate inflammatory pathways in arthritis . Swiss Med Wkly.   2015 ; 145 : W14190 .

Manzel A , Muller DN , Hafler DA , et al. . Role of “Western diet” in inflammatory autoimmune diseases . Curr Allergy Asthma Rep.   2014 ; 14 : 404 .

Guerreiro CSG , Calado ÂCA , Sousa JSJ , et al. . Diet, microbiota and gut permeability – the unknown triad in rheumatoid arthritis . Front Med . 2018 ; 5 : 349 .

Mohammed AT , Khattab M , Ahmed AM , et al. . The therapeutic effect of probiotics on rheumatoid arthritis: a systematic review and meta-analysis of randomized control trials . Clin Rheumatol.   2017 ; 36 : 2697 – 2707 .

Lee Y , Bae S , Song G.   Omega-3 polyunsaturated fatty acids and the treatment of rheumatoid arthritis: a meta-analysis . Arch Med Res . 2012 ; 43 : 356 – 362 .

Nikiphorou E , Uksila J , Sokka T.   A cross-sectional study of vitamin D levels in a large cohort of patients with rheumatic diseases . Clin Rheumatol.   2018 ; 37 : 803 – 810 .

Senftleber N , Nielsen S , Andersen J , et al. . Marine oil supplements for arthritis pain: a systematic review and meta-analysis of randomized trials . Nutrients . 2017 ; 9 : 42 .

World Health Organization. Healthy diet. https://www.who.int/news-room/fact-sheets/detail/healthy-diet . Accessed June 19, 2019 .

  • rheumatoid arthritis
  • mediterranean diet
  • omega-3 fatty acids
  • dietary supplements
  • vegetarianism
  • science of nutrition

Supplementary data

Email alerts, citing articles via.

  • Recommend to your Library

Affiliations

  • Online ISSN 1753-4887
  • Print ISSN 0029-6643
  • Copyright © 2024 International Life Sciences Institute
  • About Oxford Academic
  • Publish journals with us
  • University press partners
  • What we publish
  • New features  
  • Open access
  • Institutional account management
  • Rights and permissions
  • Get help with access
  • Accessibility
  • Advertising
  • Media enquiries
  • Oxford University Press
  • Oxford Languages
  • University of Oxford

Oxford University Press is a department of the University of Oxford. It furthers the University's objective of excellence in research, scholarship, and education by publishing worldwide

  • Copyright © 2024 Oxford University Press
  • Cookie settings
  • Cookie policy
  • Privacy policy
  • Legal notice

This Feature Is Available To Subscribers Only

Sign In or Create an Account

This PDF is available to Subscribers Only

For full access to this pdf, sign in to an existing account, or purchase an annual subscription.

  • International edition
  • Australia edition
  • Europe edition

Woman's hands are examined by a doctor

Trial offers hope for millions that jab could prevent rheumatoid arthritis

An existing drug for the chronic disease could slow or stop its progression, researchers say

Scientists have discovered a jab that could prevent rheumatoid arthritis (RA), a development experts say could offer hope to millions at risk of the disease.

RA is a chronic disease that causes inflammation in the body and triggers pain in the joints. About 18 million people globally are affected by the condition, which can lead to heart, lung or nervous system problems, according to the World Health Organization.

It most commonly begins in middle age, but much younger people can also be affected. No treatment exists that can prevent the disease.

Now researchers have found that an existing drug for RA, which patients can inject into their stomach or thigh, could help slow its progression in those with early symptoms or stop it in its tracks altogether.

A clinical trial found abatacept to be “effective in preventing the onset” of RA. Researchers said the results, published in the Lancet , were “promising” and could be “good news for people at risk of arthritis”.

Abatacept is prescribed to people who already have RA, but a team led by King’s College London explored whether it could prevent it in those deemed at risk.

The drug – administered in hospital through a drip or at home with weekly injections – works by targeting the cause of inflammation.

Two hundred and thirteen patients were recruited from 28 hospital-based early arthritis clinics in the UK and the Netherlands. They were all evaluated to be at early risk of RA by researchers.

Of the total, 110 were given abatacept and the remainder assigned to a placebo group. The estimated proportion of patients remaining arthritis-free at 12 months was 92.8% in the abatacept group and 69.2% in the placebo group.

After two years, 27 (25%) members of the abatacept group had progressed to RA compared with 38 (37%) in the placebo group.

Prof Andrew Cope, of King’s College London, said: “This is the largest rheumatoid arthritis prevention trial to date and the first to show that a therapy licensed for use in treating established rheumatoid arthritis is also effective in preventing the onset of disease in people at risk.

“These initial results could be good news for people at risk of arthritis as we show that the drug not only prevents disease onset during the treatment phase but can also ease symptoms such as pain and fatigue.”

He added: “There are currently no drugs available that prevent this potentially crippling disease. Our next steps are to understand people at risk in more detail so that we can be absolutely sure that those at highest risk of developing rheumatoid arthritis receive the drug.”

The trial also showed other outcomes of using abatacept, such as lower pain scores and higher quality of life measurements among patients.

One patient, Philip Day, 35, of Eltham, south-east London, was enrolled in the trial in 2018 and prescribed abatacept. Joint pain had prevented the once-keen footballer from taking part in the sport.

Day described the trial as a “ray of hope at a dark time”. He added: “Within a few months I had no more aches or pains and five years on I’d say I’ve been cured. Now, I can play football with my three-year-old son and have a normal life.”

Prof Lucy Donaldson, of the charity Versus Arthritis, welcomed the findings, adding: “This research highlights how important it is to spot the early signs of arthritis to give us a chance at stopping it in its tracks, offering hope to thousands of people living with – or at risk of developing – rheumatoid arthritis.”

  • Medical research

Most viewed

ScienceDaily

Clinical trial shows rheumatoid arthritis drug could prevent disease

A drug used to treat rheumatoid arthritis could also prevent the disease in individuals deemed to be at risk.

Results from a Phase 2b clinical trial, published today in The Lancet by researchers led by King's College London, provides hope for arthritis sufferers after it was shown that the biologic drug abatacept reduces progression to this agonising chronic inflammatory disease.

Rheumatoid arthritis affects half a million people in the UK and develops when the body's immune system attacks itself, causing joint pain, swelling and significant disability. The disease most commonly begins in middle age, but much younger age groups can be afflicted, and until now there is no cure or prevention.

Abatacept is currently used as an effective second or third line treatment for people living with established rheumatoid arthritis and is given by weekly injections at home or in hospital via a drip.

Researchers from King's College London recruited 213 patients at high risk of the disease to understand whether a year-long treatment of the biologic drug could be used to prevent progression to rheumatoid arthritis.

They recruited men and women over the age of 18 with early symptoms such as joint pain but no joint swelling, and treated half with the drug and half with a placebo every week for a year. The study drug was then stopped, and study participants monitored for a further 12 months.

After twelve months of treatment, 6% of patients treated with abatacept had developed arthritis compared to 29% in the placebo arm. By 24 months, the differences were still significant, with a total of 25% progressing to rheumatoid arthritis in the abatacept arm compared to 37% in the placebo arm.

Professor Andrew Cope, from King's College London, said: "This is the largest rheumatoid arthritis prevention trial to date and the first to show that a therapy licensed for use in treating established rheumatoid arthritis is also effective in preventing the onset of disease in people at risk. These initial results could be good news for people at risk of arthritis as we show that the drug not only prevents disease onset during the treatment phase but can also ease symptoms such as pain and fatigue. This is also promising news for the NHS as the disease affects people as they age and will become more expensive to treat with a growing aging population."

Secondary outcomes for the trial showed that abatacept was associated with improvements in pain scores, function and quality of life measurements, as well as lower scores of inflammation of the lining of joints detectable by ultrasound scan.

Philip Day, a 35-year-old software engineer and founder of FootballMatcher from Eltham, was at high-risk for rheumatoid arthritis. A keen football player, Philip's joint pain deterred him from playing and affected his day-to-day life. He was enrolled in the trial in 2018, at the age of the 30, and was prescribed abatacept.

He said: "The pain got so terrible I stopped going to football, and I got lazier and felt progressively worse physically and mentally. The pain was unpredictable, it would show up in my knees one day, my elbows the next, and then my wrists or even my neck. At the time, my wife and I wanted to have children and I realised my future was pretty bleak if the disease progressed. I'd always wanted to be the kind of dad that played football with his son and I knew the pain would stop me from realising that dream.

"Enrolling in the trial was a no-brainer; it was a ray of hope at a dark time. Within a few months I had no more aches or pains and five years on I'd say I've been cured. Now, I can play football with my three-year-old son and have a normal life."

One year's treatment with abatacept costs the NHS about £10,000 per patient and is not without risk. Side effects include upper respiratory tract infections, dizziness, nausea and diarrhea, but these are generally mild.

Professor Cope added: "There are currently no drugs available that prevent this potentially crippling disease. Our next steps are to understand people at risk in more detail so that we can be absolutely sure that those at highest risk of developing rheumatoid arthritis receive the drug."

Rheumatologist Professor Sir Ravinder N Maini FRS FMedSci FRCP, who was not involved in the research, said: "Professor Cope and colleagues from King's College, London, in collaboration with researchers in the UK and Netherlands, have published the results of an exciting clinical trial in The Lancet, which demonstrates that it is now possible to prevent the onset of RA, a disease that remains incurable despite great advances in its treatment in the recent past.

"The results clearly show that during the treatment period almost all individuals receiving the biologic drug showed no symptoms or signs of RA compared with the control population amongst many more developed RA. In the follow up period of 1 year off treatment, it is interesting to note that some appeared to go into remission.

"Prevention of disease is of course a highly desirable goal in preventing the ravages of disabling RA, which is associated with a significant social and financial burden. Many further questions arise from this important study. For example, will this preventive approach be safe and cost effective if continued long term or can the selection of suitable populations be refined so that only those likely to benefit most are treated with a short course of treatment?"

  • Joint Health
  • Pharmacology
  • Chronic Illness
  • Diseases and Conditions
  • Healthy Aging
  • Pain Control
  • Rheumatoid arthritis
  • Personalized medicine
  • COX-2 inhibitor
  • Drug discovery

Story Source:

Materials provided by King's College London . Note: Content may be edited for style and length.

Journal Reference :

  • Andrew P Cope, Marianna Jasenecova, Joana C Vasconcelos, Andrew Filer, Karim Raza, Sumera Qureshi, Maria Antonietta D'Agostino, Iain B McInnes, John D Isaacs, Arthur G Pratt, Benjamin A Fisher, Christopher D Buckley, Paul Emery, Pauline Ho, Maya H Buch, Coziana Ciurtin, Dirkjan van Schaardenburg, Thomas Huizinga, René Toes, Evangelos Georgiou, Joanna Kelly, Caroline Murphy, A Toby Prevost, Sam Norton, Heidi Lempp, Maria Opena, Sujith Subesinghe, Toby Garrood, Bina Menon, Nora Ng, Karen Douglas, Christos Koutsianas, Faye Cooles, Marie Falahee, Irene Echavez-Naguicnic, Anurag Bharadwaj, Michael Villaruel, Ira Pande, David Collins, Suzannah Pegler, Sabrina Raizada, Stefan Siebert, George Fragoulis, Jesusa Guinto, James Galloway, Andrew Rutherford, Theresa Barnes, Helen Jeffrey, Yusuf Patel, Michael Batley, Brendan O'Reilly, Srivinisan Venkatachalam, Thomas Sheeran, Claire Gorman, Piero Reynolds, Asad Khan, Nicola Gullick, Siwalik Banerjee, Kulveer Mankia, Deepak Jordan, Jane Rowlands, Mirian Starmans-Kool, James Taylor, Pradip Nandi, Ilfita Sahbudin, Mark Maybury, Samantha Hider, Ann Barcroft, Jeremy McNally, Jo Kitchen, Muhammad Nisar, Vanessa Quick. Abatacept in individuals at high risk of rheumatoid arthritis (APIPPRA): a randomised, double-blind, multicentre, parallel, placebo-controlled, phase 2b clinical trial . The Lancet , 2024; DOI: 10.1016/S0140-6736(23)02649-1

Cite This Page :

  • Anchors Holding Antarctic Land-Ice Shrinking
  • Compound Vital for All Life and Life's Origin
  • How Competition Can Help Promote Cooperation
  • Giant New Snake Species Identified in the Amazon
  • Chemists Synthesize Unique Anticancer Molecules
  • Neutron Star at Heart of Supernova Remnant
  • The Search for More Temperate Tatooines
  • Steering Light With Supercritical Coupling
  • Dramatic Improvements in Crohn's Patients
  • Record-Breaking Quasar Discovered

The immunology of rheumatoid arthritis

Affiliations.

  • 1 Department of Medicine, Stanford University School of Medicine, Stanford, CA, USA. [email protected].
  • 2 Department of Medicine, Veterans Affairs Palo Alto Health Care System, Palo Alto, CA, USA. [email protected].
  • 3 Department of Medicine, Stanford University School of Medicine, Stanford, CA, USA.
  • 4 Department of Medicine, Veterans Affairs Palo Alto Health Care System, Palo Alto, CA, USA.
  • PMID: 33257900
  • PMCID: PMC8557973
  • DOI: 10.1038/s41590-020-00816-x

The immunopathogenesis of rheumatoid arthritis (RA) spans decades, beginning with the production of autoantibodies against post-translationally modified proteins (checkpoint 1). After years of asymptomatic autoimmunity and progressive immune system remodeling, tissue tolerance erodes and joint inflammation ensues as tissue-invasive effector T cells emerge and protective joint-resident macrophages fail (checkpoint 2). The transition of synovial stromal cells into autoaggressive effector cells converts synovitis from acute to chronic destructive (checkpoint 3). The loss of T cell tolerance derives from defective DNA repair, causing abnormal cell cycle dynamics, telomere fragility and instability of mitochondrial DNA. Mitochondrial and lysosomal anomalies culminate in the generation of short-lived tissue-invasive effector T cells. This differentiation defect builds on a metabolic platform that shunts glucose away from energy generation toward the cell building and motility programs. The next frontier in RA is the development of curative interventions, for example, reprogramming T cell defects during the period of asymptomatic autoimmunity.

Publication types

  • Research Support, N.I.H., Extramural
  • Research Support, Non-U.S. Gov't
  • Research Support, U.S. Gov't, Non-P.H.S.
  • Arthritis, Rheumatoid / etiology
  • Arthritis, Rheumatoid / immunology*
  • Autoimmunity
  • Inflammation / immunology
  • Self Tolerance
  • Synovitis / immunology
  • T-Lymphocytes / immunology

Grants and funding

  • R01 AI129191/AI/NIAID NIH HHS/United States
  • R01 HL142068/HL/NHLBI NIH HHS/United States
  • I01 BX001669/BX/BLRD VA/United States
  • R01 AI108906/AI/NIAID NIH HHS/United States
  • R01 AI108891/AI/NIAID NIH HHS/United States
  • P01 HL129941/HL/NHLBI NIH HHS/United States
  • U19 AI057266/AI/NIAID NIH HHS/United States
  • R01 AR042527/AR/NIAMS NIH HHS/United States
  • R01 AG045779/AG/NIA NIH HHS/United States
  • R01 HL117913/HL/NHLBI NIH HHS/United States

U.S. flag

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

  • Publications
  • Account settings
  • Advanced Search
  • Journal List
  • Front Med (Lausanne)

Etiology and Risk Factors for Rheumatoid Arthritis: A State-of-the-Art Review

Vasco c. romão.

1 Rheumatology Department, Hospital de Santa Maria, Centro Hospitalar Universitário Lisboa Norte, Lisbon Academic Medical Centre and European Reference Network on Rare Connective Tissue and Musculoskeletal Diseases Network (ERN-ReCONNET), Lisbon, Portugal

2 Rheumatology Research Unit, Instituto de Medicina Molecular João Lobo Antunes, Faculdade de Medicina, Universidade de Lisboa, Lisbon, Portugal

João Eurico Fonseca

Rheumatoid arthritis (RA) is the most common systemic inflammatory rheumatic disease. It is associated with significant burden at the patient and societal level. Extensive efforts have been devoted to identifying a potential cause for the development of RA. Epidemiological studies have thoroughly investigated the association of several factors with the risk and course of RA. Although a precise etiology remains elusive, the current understanding is that RA is a multifactorial disease, wherein complex interactions between host and environmental factors determine the overall risk of disease susceptibility, persistence and severity. Risk factors related to the host that have been associated with RA development may be divided into genetic; epigenetic; hormonal, reproductive and neuroendocrine; and comorbid host factors. In turn, environmental risk factors include smoking and other airborne exposures; microbiota and infectious agents; diet; and socioeconomic factors. In the present narrative review, aimed at clinicians and researchers in the field of RA, we provide a state-of-the-art overview of the current knowledge on this topic, focusing on recent progresses that have improved our comprehension of disease risk and development.

Introduction

Rheumatoid arthritis (RA) is a chronic immune-mediated multisystemic disease that mainly localizes to the joints. It is the most common systemic inflammatory rheumatic disease ( 1 , 2 ) and it is associated with considerable morbidity and disability, as well as increased mortality ( 3 ). In the last decades, the prognosis of RA patients has been dramatically improved by the expansion of knowledge on the etiology and pathophysiology of the disease that paved the way for the development of a number of currently available effective drugs ( 4 ). However, despite these advances, there is still a substantial gap in the management of the disease, with many patients failing to attain profound and sustained clinical responses, ultimately demonstrating modest long-term outcomes. Even more strikingly, the actual impact on the prevention or delay of the disease in subjects at high-risk has overall been marginal.

Extensive efforts have been devoted in the last decades to investigate the epidemiological association of several factors with the risk of developing RA, as well as its course and prognosis. Tremendous progress has, nonetheless, been made. Although a precise etiology is yet to be determined, it is apparent that RA is a multifactorial disease, with a complex interplay between the host and the environment determining the overall risk of disease susceptibility, persistence and severity ( 5 , 6 ). In fact, this intricacy is well-illustrated in the European League Against Rheumatism (EULAR) definition of the stages that unfold until the development of RA ( Box 1 ) ( 7 ). The insights from the wealth of epidemiological studies available, though not directly proving causality, are important for understanding the etiology and pathogenesis of RA. Additionally, they can provide support to patient advice on modifiable risk factors.

Stages preceding the development of RA as proposed by EULAR.

  • a. Genetic risk factors for RA
  • b. Environmental risk factors for RA
  • c. Systemic autoimmunity associated with RA
  • d. Symptoms without clinical arthritis
  • e. Unclassified arthritis
  • a . to e . may be present simultaneously at a given moment (e.g., a+c+d)
  • ‘Pre-RA' should only be applied retrospectively to patients with RA, do describe a phase where any of a . to e . are present, individually or in combination.

Adapted from Gerlag et al. Ann Rheum Dis 2012 ( 7 ).

Risk factors for developing RA can be generically divided into host- and environment-related ( Figure 1 ). Host factors that have been associated with RA development may be further grouped into genetic; epigenetic; hormonal, reproductive and neuroendocrine; and comorbid host factors. In turn, environmental risk factors include smoking and other airborne exposures; microbiota and infectious agents; diet; and socioeconomic factors. Herein, we provide a state-of-the-art overview of the current knowledge on this topic, aimed at clinicians and researchers in the field of RA. We specifically focus on recent progresses that have improved our comprehension of disease risk and development.

An external file that holds a picture, illustration, etc.
Object name is fmed-08-689698-g0001.jpg

Summary of risk factors for the development of rheumatoid arthritis. Factors that are associated with decreased risk are represented in brackets. Factors for which evidence is conflicting and uncertainty remains are followed by a question mark. AI, aromatase inhibitors; COPD, chronic obstructive pulmonary disease; EBV, Epstein-Barr virus; FLS, fibroblast-like synoviocyte: GI, gastrointestinal; GU, genitourinary; HCV, hepatitis C virus; HDAC, histone deacetylases; HLA, human leukocyte antigen; HPA, hypothalamic-pituitary-adrenal; IBD, inflammatory bowel disease; lncRNA, long non-coding RNA; miRNA, micro RNA; MS, multiple sclerosis; OC, oral contraceptives; PBMCs, peripheral blood mononuclear cells; PTSD, post-traumatic stress disorder; SE, shared epitope.

Host Factors

As with many other immune-mediated diseases, the host is closely linked to the risk for developing RA ( Figure 1 ). This includes, first and foremost, genetic factors, which account for a major proportion of disease risk. More recently, epigenetic mechanisms have been identified to be directly involved in RA pathogenesis, modulating the risk of disease development. Notably, they can be influenced by the environment, linking extrinsic and intrinsic factors. Hormonal, reproductive and neuroendocrine factors have long been proposed as contributing to RA, given the observed female preponderance of the disease. Finally, a number of concomitant pre-existing conditions have been proposed to increase the risk of incident RA. We detail the available evidence concerning each of these groups in the following sections.

Genetic Factors

Data supporting a genetic component in RA first arose from familial and twin studies. In fact, the risk of a monozygotic twin of an RA patient for developing RA is 9–15%, which is up to 4-fold that seen for dizygotic twins, and much higher than the general population [relative risk (RR) 25–35] ( 8 – 10 ). Likewise, first-degree relatives have a RR of RA that varies from 2 to 5 and is similar in men and women ( 11 – 14 ). In addition, the risk of RA is also increased by 1.5–3-fold in offspring of parents with other immune-mediated inflammatory diseases, such as systemic lupus erythematosus, Sjögren's syndrome, ankylosing spondylitis, or Hashimoto thyroiditis ( 12 ). These studies have allowed the estimation of the heritability of RA, that is, the quantification of the genetic contribution of the disease, which was found to be 50–65% ( 15 ). Interestingly, it has been recently shown to be higher in ACPA-positive RA (50%), compared to ACPA-negative disease (20%) ( 11 ).

Currently, over 100 loci have been associated with increased RA risk in trans-ethnic populations ( 16 , 17 ). Due to its importance for the immune system, the major histocompatibility complex (MHC ) locus was the first to be identified and remains the most studied region, accounting for around a third of the disease genetic susceptibility ( 18 ). Certain human leukocyte antigen (HLA) loci such as HLA-DRB1 have been found to be strongly associated with RA in most populations ( 19 ). However, the risk varies according to specific alleles and ancestry, being higher for HLA-DRB1 * 0101/ * 0401/ * 0404 in Caucasians, HLA-DRB1 * 0405/ * 0901 in Asians, and HLA-DRB1 * 1402 in Native American Indians ( 19 , 20 ). A major breakthrough was the identification of a sequence of five amino acids in residues 70–74 (QKRAA, QRRAA, RRRAA) in the third hypervariable region of the DRß1-chain, encoded by the HLA-DRB1 gene, that was highly preserved in risk alleles for RA and was therefore named shared epitope (SE) ( 19 , 21 ). The SE hypothesis proposed that this given sequence enabled binding of a specific peptide to the HLA molecule of antigen-presenting cells (APCs), which was recognized by T cells, eventually leading to an autoimmune response. While such an antigen has not been unequivocally identified so far, and the role of the SE-containing alleles seems less important in some populations ( 19 ), the SE hypothesis was vital in providing an etiological support to the aforementioned epidemiological observations. Yet, this theory fails to explain the differential risk conferred by SE alleles (higher with * 0401 and * 0405 vs. * 0101 and * 0404, respectively) suggesting that other genetic factors may be involved ( 22 , 23 ). Moreover, the presence of the SE is more strongly associated with ACPA-positive than ACPA-negative RA ( 23 – 25 ), and has been linked with more severe disease ( 23 , 26 ), extra-articular manifestations ( 23 , 27 ) and radiographic damage ( 28 ).

Subsequent studies have complemented the SE hypothesis and determined that almost all of the RA risk conferred by the MHC region is explained by six amino acids in four HLA molecules: HLA-DRB1 (positions 11/13, 71 and 74), HLA-B (position 9), HLA-DPB1 (position 9), and HLA-A (position 77) ( 17 , 29 , 30 ). It should be noted that despite most of these amino acids being located outside the SE region, all are found in the peptide binding grooves of HLA, strengthening the importance of antigen presentation to T cells [both cluster of differentiation (CD) 4 + and CD8 + ] for the pathogenesis of RA. Importantly, these and other recent studies also confirmed the association of HLA genes, including HLA-DRB1 SE alleles, with seronegative RA, although with a smaller effect size and a differential pattern from ACPA-positive disease ( 17 , 31 – 33 ). Finally, it was proposed that a 16-category hierarchy was adopted for RA susceptibility, instead of the SE-positive/negative approach, based on the positions 11/13, 71 and 74 of the HLA-DRB1 alleles ( 29 ). These categories were also recently confirmed to be associated with radiographic damage and mortality ( 34 , 35 ).

Given that only 30% of the genetic risk for RA is explained by the MHC region, significant effort has been devoted to studying non-HLA genes . This has been done through candidate gene association studies or large-scale genome-wide association studies. Until now, more than 100 risk loci have been validated across multiple populations ( 16 , 17 , 36 , 37 ). Some important conclusions have emerged: (i) each allele confers a small risk increase [odds ratio (OR) usually <1.5–2.0] and multiple susceptibility genes interact to determine the occurrence of disease; (ii) the identified genes have contributed to better understand the mechanisms of disease, as most are directly linked to the immune system; (iii) only 20% of risk loci include coding variants, with the remaining 80% being linked to other processes such as gene expression regulation, non-coding ribonucleic acids (RNAs) or post-transcriptional changes; (iv) despite all of the advances, non-HLA alleles explain only 5–6% of genetic variation ( 16 , 17 , 36 , 37 ).

Some of the most studied non-HLA genetic variants are related to immune cell function and therefore deserve mention. A single nucleotide polymorphism (SNP; R620W) in the PTPN22 gene, which encodes a protein tyrosine phosphatase involved in antigen receptor signaling of B and T cells, has been the first to be widely replicated and remains the second stronger genetic risk factor for RA, with an OR just under 2 ( 38 ). This gain-of-function variant alters T and B cells activation thresholds, leading to changes in clonal selection and emergence of autoreactive cells ( 38 ). The R620W SNP has only been associated with seropositive RA, and support for this observation comes from studies showing that it also results in enhanced peptide citrullination ( 33 , 39 ). Interestingly, this SNP is absent in East Asian (e.g., Japanese) populations, that instead present common genetic variants of PADI4 , a gene encoding peptidyl-arginine deiminase (PAD, a peptide citrullination enzyme), that are associated with increased risk of RA (OR 1.31/allele copy) ( 38 , 40 , 41 ). Other loci and genes involved in inflammatory pathways that have been implicated in RA with a modest effect size include CTLA4 (negative regulator of T cell activation) ( 41 ), STAT4 (transcription factor involved in intracellular cytokine signaling) ( 42 ), TNFAIP3 [inhibitor of nuclear factor κ-light chain enhancer of activated B cells (NF-κB) signaling; it is required for termination of tumor necrosis factor (TNF)-induced signals] ( 43 ), TRAF1-C5 (locus including TRAF1 , which encodes a negative regulator of intracellular TNF signaling that binds to TNFAIP3, and C5 , which encodes complement factor C5) ( 44 ), IL2/21 [encoding interleukin (IL)-2 and IL-21] ( 45 ), CD40 (surface receptor present in various immune cells, including B cells, where it is crucial for B-T cell interaction) ( 46 ), IL2RA/IL2RB (encoding IL-2 receptor alpha and beta chains, respectively), ( 47 , 48 ) IL6R [encoding the IL-6 receptor (IL-6R)] ( 49 ) or CCL21 (lymphocyte chemokine) ( 46 ), among many others ( 16 , 50 , 51 ). While some genes, such as STAT4 and CTLA4 , have not been shown to be of genome-wide significance, this is a rapidly expanding field with large-scale studies identifying or confirming novel associations, most often in loci associated with immune function and regulation ( 17 , 49 – 51 ). In addition, novel genes studied in other fields such as oncology, are emerging in RA. For instance, the human RNASET2 tumor suppressor gene has been associated with RA development in Asian populations ( 52 ). This gene encodes for ribonuclease T2, an enzyme implicated in cancer development, and recently shown to regulate the innate immune response (e.g., macrophage function), which is implicated in RA development ( 53 , 54 ). Importantly, besides disease susceptibility, some of the non-HLA genes have also been associated with severity ( 55 , 56 ) and differences in seropositive and seronegative RA ( 31 , 57 ).

Epigenetic Factors

In the last decade, the role of epigenetics in RA development has started to be unraveled. Epigenetic mechanisms induce heritable variations in gene expression without actual changes in the deoxyribonucleic acid (DNA) sequence ( 58 , 59 ). In this way, they may help to explain the low concordance rate observed between monozygotic twins (9–15%) ( 8 – 10 ) and the incomplete contribution of genetic factors to the disease ( 15 ). Indeed, a recent large epigenome-wide association study found differentially variable methylation signatures in monozygotic twin pairs discordant for RA ( 60 ). Additionally, because epigenetic modifications can be induced by external stimuli (e.g., drugs, smoke, diet), they might provide the link between genome and environment interactions ( 61 ). The major epigenetic changes include DNA methylation, post-translational histone modifications and non-coding RNAs, all of which have been shown to contribute to RA susceptibility ( 58 , 59 ).

Differential DNA methylation signatures have been described in peripheral blood mononuclear cells (PBMCs) and fibroblast-like synoviocytes (FLSs) of RA patients ( 62 – 65 ). These include global hypomethylation of these cell populations ( 62 , 64 ), as well as hypo- or hypermethylation of specific promoter regions, which lead to an increase or decrease, respectively, in the transcription of pro- (e.g., IL-6, IL6-R, CXCL12, CD40L ) or anti-inflammatory [e.g., CTLA4 in T regulatory (Treg) cells] genes ( 58 , 59 , 63 , 66 , 67 ). Interestingly, treatment with methotrexate (MTX) has been shown to reverse the global hypomethylation of B cells, T cells and monocytes ( 64 ), as well as to restore Treg cell function through demethylation of the FOXP3 locus ( 68 ), highlighting the reversible nature of epigenetic changes and its potential as a therapeutic target. Furthermore, recent epigenome-wide studies have reported several differentially methylated genes in blood samples of RA patients ( 69 ). A large study identified two clusters located within (nine genes) or in close proximity (one gene) to the HLA locus, suggesting that the genetic risk for RA conferred by this region is, at least partially, mediated by DNA methylation ( 65 ). Likewise, other studies applied the same approach in RA and control [osteoarthritis (OA)] FLSs and were able to identify different methylation patterns in key genes involved in RA pathogenesis ( 70 – 73 ). Surprisingly, within RA patients, two separate groups recently identified joint-specific methylome and transcriptome signatures that varied across several joint locations (e.g., hip, knee, MCP), providing a possible explanation for the distinctive articular pattern of the disease ( 74 , 75 ).

Substantially less is known about the role of histone modification in RA. Histones can be modified by processes such as acetylation, methylation, phosphorylation, citrullination, and others, resulting in alterations of chromatin structure and, consequently, gene transcription ( 58 , 59 ). Most studies have focused on acetylation status in blood and synovial tissue of RA patients, by measuring the expression and balance of the histone-modifying enzymes histone deacetylases (HDACs) and histone acetyl transferases. Both decreases ( 76 ) and increases ( 77 – 79 ) in synovial tissue activity and expression of HDACs have been reported, although the latter are likely to be more important and have also been reported in blood ( 80 ). Different, sometimes antagonistic, effects of specific HDACs or HDAC classes may contribute to complexity ( 81 , 82 ). Nonetheless, various studies have demonstrated the impact of HDAC inhibitors in suppressing inflammation [notably, IL-6 and type-I interferon (IFN) responses], angiogenesis, and function and survival of FLSs and macrophages ( 79 – 84 ). Importantly, it was recently demonstrated that smoking, a major environmental risk factor for RA, increases the levels of two key HDACs [sirtuin (SIRT) 1 and SIRT6], again reinforcing the importance of epigenetics for gene-environment interface ( 85 , 86 ).

Non-coding RNAs are yet another mode of epigenetic regulation and include microRNAs (miRNAs, around 22 nucleotides) and long non-coding RNAs (lncRNAs, over 200 nucleotides), both of which have been extensively studied in RA susceptibility, severity and treatment ( 58 , 87 , 88 ). miRNAs are non-coding RNAs that bind messenger RNA (mRNA), leading to its destruction or blocking its translation. Due to this regulation effect on gene expression, they have been the object of significant attention in areas like oncology, metabolic diseases and inflammatory arthritides ( 87 ). A wealth of miRNAs have been studied in RA, of which the most established in terms of relevance for RA pathogenesis include miRNA-155, miRNA-146a, miRNA-223 and miRNA124a ( 58 ). The earliest and best-studied are miRNA-155 ( 89 ) and miRNA-146a ( 90 ) which were shown to be increased in many cells (B, T and NK cells, macrophages, FLSs) and tissues (blood, synovial tissue/fluid) of RA patients and to have pleiotropic, but opposite, roles in promoting or suppressing, respectively, several inflammatory, proliferating and bone erosion pathways ( 58 , 87 ). Specific gene targets of these miRNAs have been validated, confirming their role in RA pathogenesis ( 58 ). miRNA-223 and miRNA-124a have also been shown to be increased and decreased, respectively, in RA synovial tissue/fluid, FLSs and blood cells, directly contributing for regulation of osteoclastogenesis, FLS proliferation, and T cell and macrophage-mediated inflammation ( 58 , 91 , 92 ). In recent years, lncRNAs have started to be investigated in RA, due to their function as nuclear and cytoplasmic regulators of gene transcription and mRNA translation ( 58 , 88 ). Several dozen lncRNAs have been reported to be differentially expressed in RA, and in just over ten the functional role was shown to involve regulation of inflammation and matrix degradation pathways in FLSs, T cells and monocytes ( 58 , 88 ). The most characterized lncRNA is HOTAIR, which represses the expression of matrix metalloproteinase (MMP)-2 and MMP-13 ( 93 ). It was found to be increased in PBMCs of RA patients ( 94 ), as well as in FLSs from lower vs. upper extremity joints ( 74 ), implicating it as another mechanism involved in RA joint patterning ( 93 ).

Hormonal, Reproductive and Neuroendocrine Factors

Considering the female preponderance in the distribution of RA , hormonal and sex-related factors have long been investigated as predisposing to the disease ( 95 ). The sex imbalance is commonly attributed to estrogens, generally described as being pro-inflammatory, in opposition to the anti-inflammatory effects of progesterone and androgens, which are decreased in male and female RA patients ( 96 ). However, their actions are far more complex and, in fact, estrogens also possess anti-inflammatory properties in a number of cells and tissues ( 96 , 97 ). The global net effect is likely dependent on other factors such as serum and tissue concentration, predominant cell types and estrogen receptors involved, as well as the reproductive stage ( 95 , 97 ). These mechanistic aspects are important to understand the conflicting findings reported for a number of hormonal and reproductive factors in the risk of RA. Parity, breastfeeding, pregnancy loss, early menarche, age at first pregnancy, OCs and hormone replacement therapy have all been associated with increased, unchanged or decreased likelihood of development of RA ( 95 ). Although in some instances the evidence points slightly more strongly in one direction [e.g., breastfeeding ( 98 ) as protective factor in recent meta-analyses], in other cases [e.g., OCs ( 99 ) and parity ( 100 , 101 )] there is controversy even between separate meta-analyses. Thus, the true effect of such aspects is currently not fully understood, as additional factors such as varying estrogen dosage, differential impact in seropositive/seronegative disease, or interactions with other reproductive or environmental factors may contribute to heterogeneity. On the other hand, situations associated with an abrupt decline in global estrogen load, such as early menopause ( 102 ), post-menopausal stage ( 103 ), puerperium ( 104 ) and anti-estrogen agents [selective estrogen receptor modulators ( 105 ) and aromatase inhibitors ( 105 , 106 )], have been more consistently identified as risk factors for RA ( 95 ). However, a recent robust study failed to confirm an association with tamoxifen or aromatase inhibitors with the risk of incident RA in women with breast cancer ( 107 ). Finally, pregnancy itself has been historically associated with both reduced incidence and major clinical improvement of RA during the gestational period, an observation that still stands today and is attributed to the profound hormonal (sharp rise in estrogen and progesterone) and immunological [T helper (Th) 1-to-Th2 shift] maternal changes ( 104 , 108 ).

Although with more limited evidence, other related risk factors for RA include pre-eclampsia ( 109 ) and both low and high birth weight ( 110 , 111 ). The former was hypothesized to be explained by fetal microchimerism (i.e., the exchange and persistence of fetal cells in maternal circulation), which is increased both in women with pre-eclampsia ( 112 ) and with RA ( 113 ), where it is thought to mediate maternal acquisition of the SE. On the other hand, birth weight has been shown to have a U-shaped association with decreased adult hypothalamic-pituitary-adrenal (HPA) function ( 114 ). As RA patients have decreased cortisol levels and responsiveness ( 115 ), birth weight extremes were proposed to contribute to RA through downregulation of the HPA axis, setting it to function at a reduced level ( 110 ).

Indeed, ever since the first anecdotal reports of major clinical improvement induced by incident Cushing's disease ( 116 ) or focal neurological lesions ( 117 ), disturbances in the neuroendocrine system and its relation with the immune system have been implicated in RA ( 118 , 119 ). During systemic inflammation, both the HPA axis and the autonomic nervous system are physiologically activated centrally, in an attempt to suppress peripheral inflammation. This is achieved, respectively, through anti-inflammatory hormones (cortisol, adrenal androgens) and neurotransmitters [norepinephrine (ß receptors), adenosine (A2 receptors) and endogenous opioids (μ receptors)] ( 118 ). Several studies have shown that these pathways are dysregulated in RA, leading to a proinflammatory ambient at the joint level and, consequently, to synovitis ( 120 ).

Hypothalamic hyporesponsiveness to stress and inflammation was first suggested ( 115 ), but a state of relative adrenal insufficiency with inappropriately low levels of cortisol and adrenal androgens upon chronic inflammation has since been established as the key pathogenic mechanism ( 118 , 121 , 122 ). Changes in the circadian rhythm of secretion of cortisol and of proinflammatory hormones melatonin and prolactin, also likely play a role ( 115 , 123 , 124 ). At the synovial tissue level, there is impaired ability to reactivate inactive cortisone ( 125 ); increased levels, local synthesis and action of melatonin ( 126 ); upregulated aromatase activity with enhanced androgen-to-estrogen conversion and high estrogen-to-androgen ratio ( 126 , 127 ); and a preponderance of estrogen receptor-ß over estrogen receptor-α ( 120 , 128 ).

These endocrine imbalances are paralleled by changes in the sympathetic and sensory nervous systems that also contribute to RA ( 118 , 119 ). Loss of synovial sympathetic nerve fibers, probably due to the production of nerve repellent factors by macrophages and FLSs, and preservation of sensory nerve fibers (at a 1: 10 ratio) have been described in the rheumatoid synovium ( 120 , 129 ). This leads to a proinflammatory environment, due to the predominant effects of the sensory neuropeptide substance P and a shift in sympathetic signaling from anti-inflammatory ß and A2 receptors to inflammatory α and A1 receptors ( 118 ). Moreover, there is an increased systemic sympathetic tone, which is thought to be a physiological response to the decreased HPA axis function and that results in the uncoupling of these two systems ( 119 , 130 ). Overall, defects in physiological neuroendocrine mechanisms contribute to immune system dysregulation and lead to the establishment and perpetuation of RA. This is best exemplified by the effect of psychological stress ( 131 ), including childhood traumatic stress ( 132 ) and post-traumatic stress disorder ( 133 – 135 ), in predisposing for or aggravating RA ( 136 ).

Another important endocrine mediator with immunomodulatory properties is vitamin D . The close relationship with the immune system has been revealed in the last 20 years and, currently, vitamin D is known to exert pleiotropic anti-inflammatory effects through direct action in several immune cells (macrophages, dendritic cells, lymphocytes, FLSs) that express the vitamin D receptor ( 137 ). This fact, together with the replicated observation that reduced vitamin D levels and vitamin D deficiency were common in RA patients ( 138 ), led to the investigation of vitamin D as a potentially protective factor for RA ( 139 ). The present situation is still equivocal as large prospective cohort studies have either found ( 140 ) or failed to find ( 141 ) an inverse association of RA incidence and vitamin D intake. These studies were later meta-analyzed and a 24% risk reduction was observed in high vs. low vitamin D intake ( 142 ), although the main issue is that dietary questionnaires are likely not the best method to assess vitamin D status, which can be affected by time fluctuations, sun light exposure and other confounding factors ( 137 ). Interestingly, a subsequent study of the same cohort that had negative results for vitamin D intake, reported that higher cumulative ultraviolet-B light exposure, which is the primary source of vitamin D, significantly decreased RA risk ( 143 ). Further suggestions of a link between vitamin D and RA come from studies reporting associations with SNPs from the vitamin D pathway genes VDR ( 144 ) (encoding vitamin D receptor), GC ( 145 ) (encoding vitamin D-binding protein) and CYP27B1 ( 146 ) (encoding 25-hydroxy-vitamin D-activating enzyme 1α-hydroxylase) ( 137 , 139 ). Nevertheless, the true role of vitamin D as a protective factor for RA remains unclear and a large randomized controlled trial (RCT) failed to demonstrate a preventive effect of daily calcium and vitamin D supplementation on RA risk ( 147 ). Instead, more robust evidence supports the association of vitamin D deficiency and a poor prognosis in RA patients, determined by increased disease activity, functional impairment and poor HRQoL ( 137 – 139 ).

Finally, evidence supporting obesity as a risk factor for RA is mounting ( 6 , 148 – 151 ). Two recent meta-analyses including up to 13 studies have confirmed a positive association of obesity with RA (pooled RR 1.21–1.31) and a dose-response effect of body mass index (BMI) on RA risk ( 149 , 150 ). There are also some indications that this association is stronger in women and, potentially, seronegative disease, although in the latter case the two meta-analyses have conflicting results ( 149 , 150 ). These findings are in accordance with the secular rise in BMI and obesity prevalence at RA presentation observed in the last two decades ( 148 , 152 ). While obesity may be regarded as a surrogate for other environmental and lifestyle risk factors (see below), its predisposing effect for RA may be explained by metabolic and endocrine mechanisms. Plausible hypotheses include increased adipocyte secretion of proinflammatory cytokines and adipokines, as well as perturbation of sex hormone metabolism, with increase in estrogen levels due to enhanced aromatase-mediated conversion in the adipose tissue ( 153 , 154 ). Moreover, hyperlipidemia, which is linked to obesity, has been reported to be increased in individuals who develop RA, particularly in women ( 155 – 157 ). In line with this, higher statin use has been shown to be protective of incident RA in large cohort ( 158 ) and nested case-control studies ( 157 , 159 ) (more than 500,000 participants in one study that found a 23% decrease in risk) ( 159 ). This effect has been hypothesized to be linked both to lower lipid blood levels and to the anti-inflammatory actions of statins. However, the role of statins as risk factors is still not entirely understood. A case-control study found a contrasting increased risk of RA (though with no cumulative dose trend) ( 160 ) and other two very large [ n > 1,000,000 ( 161 ) and n > 2,000,000] ( 162 ) prospective cohort studies failed to detect an association with incident RA. In one of these studies there was even a potential increase of RA risk in the first year after statin commencement, that progressively returned to normal ( 161 ). Discrepancies are likely due to methodological and definition issues, but clearly a definite conclusion cannot yet be drawn.

Comorbid Host Factors

There have been epidemiological associations of other, apparently unrelated, concurrent diseases with increased future risk of developing RA. This is different from comorbidities affecting established RA patients (e.g., cardiovascular disease, infection, lymphoma, osteoporosis), which occur at higher rates than the general population and have a significant impact in the prognosis of the disease (see below) ( 2 , 163 ).

Psychiatric conditions appear to have a particularly interesting link with RA. As mentioned above, an association between post-traumatic stress disorder and an increased risk of RA has been described in both men ( 133 ) and women ( 134 ). This has been hypothesized to be related to the previously described dysregulated neuroendocrine-immune mechanisms induced by chronic stress ( 136 ). Most recently, depression, a well-known common RA comorbidity found in 15–17% of patients, was also proposed as a risk factor for RA, suggesting a bidirectional relationship ( 164 – 166 ). This observation follows the publication of three large longitudinal cohort studies that identified depression to confer a 28–65% increase in the risk of developing RA ( 167 – 169 ). Interestingly, one of the studies showed that the use of antidepressants among depressed patients was protective of RA development [hazard ratio (HR) 0.74, 95% CI 0.71–0.76] ( 168 ), whereas another study found it to be associated with subsequent seronegative RA (HR 1.75, 95% CI 1.32–2.32) ( 169 ). Novel insights into the pathogenesis of depression, suggesting prominent systemic inflammatory mechanisms, were proposed as a possible explanation for the association with RA ( 165 ). Similar, recently identified relations with other rheumatic (psoriatic arthritis), gastroenterological (inflammatory bowel disease) or dermatological (alopecia areata, vitiligo) immune-mediated diseases, further support this hypothesis ( 165 ).

In contrast, a puzzling consistent negative association has been recognized with schizophrenia for over eight decades now ( 170 , 171 ). These epidemiological data were revisited in two updated meta-analyses also including the latest studies, which confirmed the significant protective effect of schizophrenia for the development of RA (OR 0.48–0.65) ( 172 , 173 ). A possible infectious cause was first proposed as an explanation for this intriguing observation, but recent data have strengthened the genetic-immunologic theory ( 170 , 174 , 175 ). Studies have demonstrated a negative SNP-genetic correlation between the two conditions ( 174 ) and identified pleiotropic SNPs in established HLA risk genes that differentially contributed for RA and schizophrenia (i.e., based on the specific allelic variant within the same gene) ( 175 ). Although still a matter of debate, these interesting complex genetic mechanisms help to explain how a given disease can reduce the chance of developing another seemingly unconnected condition.

Atopy and allergic diseases (e.g., asthma, allergic rhinitis, atopic dermatitis) were initially suggested to be negatively associated with the risk of RA, an observation that was hypothesized to be related to the predominance of Th2-dependent pathways, as opposed to a Th1 phenotype in RA ( 176 ). Since then, several epidemiological studies have rebutted this view and reported an increase in RA incidence in allergic populations ( 177 – 180 ). Although the literature is controversial, most high-quality population-based cohort studies point toward a positive, rather than negative, association between atopy and RA ( 177 ). This has been reinforced by large recent studies ( 181 – 183 ), and linked to possible shared genetic (e.g., HLA-DRB1, IL-6R, CD40L ), immunological [e.g., natural killer (NK), Th1 and Th17 cells, TNF] and environmental (e.g., smoking) mechanisms ( 16 , 177 ). In addition, respiratory diseases , both acute and chronic, and of the upper or lower airway tract, have recently been associated with an increased risk of seropositive and seronegative RA ( 184 ). The association, though, was limited to non-smokers, suggesting possible separate, or complementary, pathogenic pathways in smoking and respiratory disease. Other studies have also confirmed the association of chronic obstructive pulmonary disease with subsequent RA ( 179 , 183 , 185 , 186 ).

Several other diseases have been identified as risk factors for incident RA, most notably other non-rheumatological immune-mediated diseases , such as autoimmune thyroid disease (i.e., Hashimoto thyroiditis and Graves' disease) ( 187 , 188 ), type 1 diabetes mellitus ( 187 , 189 ), alopecia areata ( 190 ), vitiligo ( 191 ), inflammatory bowel disease ( 192 ) and, possibly, multiple sclerosis ( 193 ) (less robust evidence) ( 187 ). Various common genetic risk determinants have been identified and other host and external factors are also likely to be important ( 16 , 17 , 38 , 45 , 50 , 61 , 188 , 189 ). Interestingly, despite being an important RA comorbidity, one population-based case-control study has associated type 2 diabetes mellitus with increased risk of incident RA, ( 194 ) although no effect had been reported in an earlier similar study ( 189 ).

Finally, large longitudinal cohort studies have recently suggested that sleep disorders ( 195 ), including both obstructive sleep apnea ( 195 , 196 ) and non-apnea sleep disorders ( 195 , 197 ), could be additional risk factors for RA. The influence of sleep disturbances on immune dysregulation has also been proposed as a possible explanation for the 91% increased risk of RA found in patients with migraine ( 198 ), highlighting the complex multifactorial nature of the disease.

Environmental Factors

Although the data above support a large impact of host on the development of RA, the environment also plays a fundamental role in determining the ultimate risk of disease. In fact, extrinsic factors have been identified to interact with at-risk subjects and confer a multiplicative increase in the likelihood of developing RA ( Figure 1 ). Environmental factors can be roughly grouped into four categories: airborne exposures, notably including smoking; microbiota and infectious agents; diet; and socioeconomic factors, including occupational and recreational exposures. Extensive data are available directly implicating these numerous aspects in the etiology of RA.

Smoking and Other Airborne Exposures

The recognition of the lung as a major site of early pathogenic events has been one of the great breakthroughs in the understanding of the disease and is well-exemplified by the strong association of several airborne noxious agents with RA ( 199 , 200 ).

Smoking is the most important of such exposures and has been established as one of the main risk factors for the development of RA ( 199 ). Its association with RA has been extensively replicated since the first description more than three decades ago ( 201 ), and smoking is currently thought to explain 20–25% of overall RA risk and up to 35% of ACPA-positive RA ( 202 , 203 ). A meta-analysis of 16 studies estimated an overall OR for ever smoking of 1.40 (95% CI 1.25–1.58) ( 204 ). The effect was stronger for RF-positive RA (OR 1.66, 95% CI 1.42–1.95) and in men (OR 1.89, 95% CI 1.56–2.28), with a multiplicative interaction (RF-positive men: OR 3.02, 95% CI 2.35–3.88) ( 204 ). Moreover, there is a clear dose-response relationship, with significantly higher risks for current or heavy vs. past or light smokers, respectively, and a linear increase in risk with smoking pack-years ( 203 , 204 ). Accordingly, smoking cessation was shown to progressively decrease the risk of RA development, returning to that of never smokers after a period of 20–30 years ( 202 , 203 , 205 ). This is, therefore, a practical advice that should be routinely given to patients, especially to those at higher risk of developing the disease. Passive smoking should also be avoided, as studies have additionally suggested a possible link with prenatal ( 111 , 206 , 207 ), childhood ( 207 , 208 ) and adult ( 203 ) passive exposure to cigarette smoke.

A key discovery was the gene-environment interaction of smoking with SE alleles and seropositive RA, with a multiplicative dose effect of both smoking load (e.g., OR 6.3, 12.0, 24.6, and 37.6 in homozygotic SE carriers with 0, <10, 10–19 or ≥20 pack-years, respectively) and SE risk alleles (e.g., RR 21.0, 6.5, and 1.5 in ever smokers carrying two, one or no copies of SE genes, respectively) ( 199 , 202 , 209 ). All of these epidemiological observations have provided the basis for a novel model of RA pathogenesis, in which smoking is responsible for in situ protein citrullination (i.e., the conversion of arginine to citrulline) in the lungs of SE-positive individuals, with the subsequent generation of ACPAs and, eventually, RA ( 199 , 209 ). Subsequent demonstration that both activity and expression of the citrullination enzyme PAD2 are increased in the bronchioalveolar compartment of smokers ( 210 ) and that, after citrullination, peptides such as vimentin or fibrinogen bind specifically to SE-containing HLA molecules and induce the emergence of autoreactive T cells ( 211 ), further supported this hypothesis ( 209 ). In line with it, cigarette smoke, rather than tobacco or nicotine per se , seems to be crucial in the process, as it leads to chronic airway inflammation ( 200 ), whereas non-inhaled moist snuff tobacco does not increase the risk of RA ( 212 ). Moreover, the RA risk associated with smoking has been shown to be influenced by SNPs and deletions in genes encoding enzymes involved in the detoxification of smoke carcinogens [e.g., glutathione S-transferases ( 213 , 214 ), N-acetyltransferases ( 215 ) or heme-oxygenase ( 214 )], again demonstrating the importance of smoke-induced changes in the respiratory airway. Finally, carbamylation (i.e., the conversion of lysine to homocitrulline) has also been reported to be associated with smoking, generating novel autoantigens that are targeted by the RA immune-specific response ( 216 ). Indeed, anti-carbamylated protein antibodies have been described in the serum of pre-RA subjects, and predict the development of RA ( 217 ).

Several other inhaled agents are thought to exert similar harmful effects and increase the risk of RA ( 200 ). The first to be recognized and best studied is silica , a common occupational exposure (e.g., mining, construction or ceramic industries) that is independently associated with RA (OR around 2–3) ( 218 , 219 ). Similarly to smoking, it is specifically associated with ACPA-positive RA and both exposures have an additive effect (OR 7.36 in silica-exposed smokers), which increases with pack-years of smoking ( 220 , 221 ). Additionally, textile dust (OR 2.8, 95% CI 1.6–5.2, similar for ACPA-positive and ACPA-negative RA, but with an interaction with HLA-SE in the former: OR 39.1, 95% CI 5.1–297.5) ( 222 ) and inorganic dust (e.g., asbestos, cement) ( 223 ) were also linked to RA ( 224 ). In contrast, despite a few contradicting reports ( 225 ), ambient air pollution does not seem to be a consistent risk factor for RA ( 226 ).

Microbiota and Infectious Agents

The “infectious hypothesis” has long been proposed as a likely explanation for the development or triggering of RA ( 227 ). The decline in RA incidence observed in various populations following the improvement in health and sanitary conditions was one of the main facts indirectly supporting this possibility ( 228 – 230 ). It was reinforced by epidemiological and translational studies directly involving specific viruses, bacteria and other microbial agents, that could contribute to RA through non-specific immune activation, molecular mimicry or other mechanisms ( 227 , 231 ). However, after decades of research, no single infectious agent has been consistently identified to be the cause or to increase the risk of RA ( 231 ). Notably, in the last years, the aforementioned mucosal immunity, together with oral/intestinal dysbiosis and chronic infections have been closely implicated in the etiology and pathogenesis of RA ( 199 , 232 ).

Periodontitis , the main of such factors, results from dysbiosis of the oral microbiota and has been associated with increased risk of RA ( 233 ). The relationship between both diseases is bidirectional (i.e., RA patients also have higher likelihood to develop periodontitis) and profound, as they share similar genetic (e.g., HLA-SE alleles) and environmental (e.g., smoking, nutrition) risk factors and both lead to chronic inflammation, bone erosion and tissue destruction ( 233 , 234 ). Porphyromonas gingivalis ( P. gingivalis ) is a major cause of periodontitis and the most important agent that has been specifically associated with RA ( 233 ). Its involvement is not just circumstantial but an etiologic role has been proposed through mechanisms similar to those previously described for smoke ( 235 ). In fact, P. gingivalis is unique in that it has its own PAD, that can cause chronic citrullination of bacterial and host proteins ( 236 ), leading to breach of immune tolerance, ACPA production and, eventually, through molecular mimicry and/or epitope spreading, culminating in RA ( 233 , 235 ). This theory is supported by several studies demonstrating PAD activity, protein citrullination and ACPA generation in the inflamed periodontium ( 237 – 239 ); gingival citrullination patterns in periodontitis that mirror those seen in the rheumatoid joint ( 239 ); increased ACPA titers in periodontitis patients ( 240 ); cross-reactivity between RA-specific anti-human α-enolase ACPAs and P. gingivalis ( 241 ); and increased antibodies against P. gingivalis in at-risk subjects and in patients with periodontitis, pre-RA, RA and ACPA-positive RA, that additively interacted with smoking and HLA-DRB1 ( 233 , 242 – 244 ). Of note, a recent study further demonstrated that oral dysbiosis, enriched in P. gingivalis , was present even in periodontally healthy sites of ACPA-positive at-risk individuals ( 245 ). In addition, another pathogenic periodontitis agent, Aggregatibacter actinomycetemcomitans ( A. actinomycetemcomitans ), has been directly implicated in RA through different mechanisms involving neutrophil-mediated citrullination ( 239 ). Nonetheless, despite all of the rationale and evidence supporting the periodontitis-RA link, it should be noted that the more robust, largest, population-based prospective studies have either failed to demonstrate an association of periodontitis with incident RA ( 246 – 248 ) or did so with a small effect size (OR 1.16–1.17) and without adjustment for major confounding factors such as smoking ( 249 , 250 ). Whether this is the result of methodological and disease-definition issues or actually reflects a smaller role of periodontitis in RA risk is unclear.

Growing interest has been devoted in the last decade to gut microbiota and its role in immune homeostasis and the development of disease ( 251 ). Intestinal dysbiosis has been linked to a number of inflammatory rheumatic diseases, including RA ( 251 , 252 ). Evidence supporting this association firstly stems from animal studies demonstrating that microbial flora is indispensable for the development and aggravation of experimental arthritis ( 232 , 251 , 253 ). In addition, human studies reported changes in the composition of gut microbiota of RA patients, with decreased microbial diversity, enrichment of Prevotella copri, Lactobacillus spp. and Clostridium spp. and decrease in Bacteroides spp. and Haemophilus spp ( 232 , 251 , 253 – 257 ). Interestingly, a comprehensive metagenome-wide association study demonstrated remarkable concordance between fecal and oral dysbiosis patterns, which could differentiate RA patients from controls with remarkable accuracy [area under the curve (AUC) 0.94 and 0.87, respectively] ( 232 ). It is important to mention that all clinical association studies have a cross-sectional design and, thus, do not clearly imply an etiologic role. Nevertheless, a direct causal effect has been suggested by recent studies showing that RA gut dysbiosis induces activation of autoreactive T cells ( 253 ) and that HLA-DR-presented Prevotella copri peptides can generate RA-specific Th1 and Th17 responses ( 258 ). In line with this, Prevotella spp has been shown to be enriched in subjects at risk for RA, further implicating intestinal dysbiosis in the pathophysiology of RA ( 257 , 259 ). Moreover, the fact that oral and gut microbiome dysbiosis patterns of new-onset untreated RA patients were correlated with disease activity and improved with conventional synthetic DMARD (csDMARD) treatment, also provides indirect evidence of pathogenicity ( 232 ).

As mentioned, external infectious agents have been implicated as risk factors for RA for decades ( 227 ). Besides the discussed relation with the decrease in RA incidence over time, putative mechanisms linking infection and RA include molecular mimicry, epitope spreading, B cell amplification/proliferation, non-specific inflammatory activation and superantigens ( 260 ). Several studies supported this concept by demonstrating increased prevalence of microbial-specific antibodies in the serum of RA patients and identifying bacterial/viral proteins or genetic material in rheumatoid synovial fluid and/or tissue ( 231 ). Commonly reported agents include Epstein-Barr virus (EBV), cytomegalovirus (CMV), parvovirus B19, rubella virus, mycoplasma, Proteus mirabilis, Escherichia coli , hepatitis B/C virus, Borrelia burgdorferi (Lyme's disease), Chikungunya virus, and others ( 5 , 227 , 231 , 260 ). However, the epidemiological associations are inconsistent and the overall study quality is poor ( 260 ). This was revealed in a recent meta-analysis of 48 studies that concluded that only parvovirus B19 (OR 1.77, 95% CI 1.11–2.80), hepatitis C virus (OR 2.82, 95% CI 1.35–5.90) and, possibly, EBV [anti-VCA (OR 1.5, 95% CI 1.07–2.10) and anti-EA (OR 2.74, 95% CI 1.27–5.94) but not anti-EBNA antibodies], but not hepatitis B, CMV or other viruses were associated with RA ( 260 ). A previous meta-analysis of 23 studies of EBV seroprevalence had failed to demonstrate an increased risk of RA, especially when only higher quality studies were considered ( 261 ). These findings are in disagreement with extensive experimental research implicating EBV in RA through potent B cell stimulation, molecular mimicry (including cross-reactivity with citrullinated host antigens and recognition of anti-EBV antibodies by HLA-DR), increased blood and synovial tissue viral load and impaired viral-specific T cell response ( 231 , 262 ). The association therefore remains equivocal for most, if not all, infectious agents. An important confounding issue is the fact that these infections per se can cause RA-like polyarthritis in the acute phase (e.g., parvovirus B19, rubella), chronic polyarthralgia/polyarthritis following the disease (e.g., Chikungunya, Lyme's disease) and even generate ACPAs or RF (e.g., hepatitis C virus), thus making it harder to ascertain case definition ( 231 , 260 ). Moreover, limited available evidence failed to demonstrate temporal or spatial clustering of incident RA that could be related to a common infectious exposure ( 263 ). Surprisingly, recent self-reported bacterial urogenital and gastrointestinal, but not respiratory, infections were found to be associated with a decreased risk of incident RA in a large population-based case-control study ( 264 ). This effect was hypothesized to be due to antibiotic- and/or infection-induced changes in the microbiota that could, in this case, be protective.

Nonetheless, taking all the data into account, it is likely that, as with other environmental factors, infectious agents have some kind of role in priming, triggering or potentiating disturbed immune mechanisms in a genetically predisposed host. Novel evidence supporting this concept comes from an important study that elegantly demonstrated a mechanism mediated by protective/predisposing HLA alleles, through which infection (and microbiota) can influence the risk of ACPA-positive RA ( 265 ). Cross-recognition by CD4 + T cells of an epitope (DERAA) that is found both in synovial citrullinated vinculin and in several microbes (including gut bacteria such as Lactobacillus , enriched in RA) ( 232 , 254 , 255 ) and which is presented by predisposing HLA-DQ molecules (HLA-DQ5, DQ-7.3 and DQ8, all in tight linkage disequilibrium with HLA-DR SE alleles), leads to increased ACPA production and, eventually, RA ( 265 ).

The epidemiological association of dietary factors with RA has been extensively studied. In spite of the difficulties in accurately assessing patient nutritional behavior before RA onset and isolating the effect of a given food, drink or nutrient, some findings are consistent. Importantly, they also provide clues on RA etiology and pathogenesis, as shown by the influence of the modulation of the intestinal microbiome by diet on the risk of developing RA ( 251 , 252 ).

Low-to-moderate alcohol consumption is protective of RA development. A meta-analysis that included only cohort or nested case-control studies ( n = 8; 195, 029 participants), reported a 14% decrease in the risk of RA (RR 0.86, 95% CI 0.78–0.94) ( 266 ). The effect was dependent on dose (J-shaped non-linear trend, with greater benefit for 9g/day vs. 3 or 12g/day), time (17% reduction if consistent intake for ≥10 years) and sex (19% reduction in women) and unrelated to beverage type ( 266 ). Alcohol-induced downregulation of the immune response and proinflammatory cytokine production has been proposed as an explanation for this observation ( 267 ). Most recently, a population-based study confirmed the protective nature of alcohol in both ACPA-positive and ACPA-negative RA and reported an additive interaction with HLA-SE and smoking for development of the former (OR 25.3, 95% CI 17.7–36.2 for never-drinkers, ever smokers, HLA-SE-positive patients) ( 268 ). The mechanism justifying this interaction is currently unclear.

Similarly, general healthy eating behaviors have also been associated with decreased risk of RA ( 251 , 252 ). Long-term adherence to a healthier diet (assessed through a standard dietary quality score) in women from the Nurses' Health Study ( n = 169, 989) was protective of younger-onset (≤ 55 years-old) seropositive RA development (HR 0.60, 95% CI 0.51–0.88) ( 269 ). Subgroup analysis revealed that lower red/processed meat (OR 0.58, 95% CI 0.43–0.79) and sodium (OR 0.65, 95% CI 0.44–0.98) consumption were associated with a significant decrease in RA risk ( 269 ). A previous analysis of the same study had also associated daily consumption of sugar-sweetened soft drinks with a 63% increased risk of seropositive RA, which was higher in RA starting after the age of 55 (HR 2.64, 95% CI 1.56–4.46) ( 270 ). Similar findings implicating high red meat intake as a risk factor for RA had been reported in another prospective cohort ( 271 ). Interestingly, sodium has been shown to interact with smoking to increase RA incidence only in smokers (OR 2.26, 95% CI 1.06–4.81), particularly that of ACPA and/or HLA-SE-positive disease ( 272 ).

In contrast, an important component of a healthy diet is the consumption of food rich in polyunsaturated oils (e.g., omega-3 fatty acids), such as fish and olive oil. Most evidence supports the protective role of fish ( 273 , 274 ), omega-3 and omega-6 fatty acids ( 274 , 275 ), and olive oil ( 251 , 276 ). Although a few studies failed to find a beneficial association with these foods, compelling mechanistic evidence has been recently provided by a nested-case control study that demonstrated an inverse relation between RF- (OR 0.27, 95% CI 0.10–0.79) and ACPA-positivity (OR 0.42, 95% CI 0.20–0.89) in SE-positive individuals at risk for RA ( 251 , 277 ). Moreover, RCTs have demonstrated that fish oil improves pain outcomes and clinical response to csDMARDs, corroborating a possible role in RA pathogenesis ( 251 , 278 ).

Fruits and vegetables are a hallmark of a balanced diet and a major source of fiber and antioxidant elements such as vitamin C. Both have been associated with a decreased risk of RA in robust prospective studies, ( 276 , 279 , 280 ) although data from another large cohort could not confirm these findings ( 269 ). Nonetheless, the overall picture is clear in pointing toward a trend for a protective role of healthier nutritional behaviors in RA development. A Mediterranean diet, usually richer in fruits, vegetables, olive oil and fish, has been proposed as a possible explanation for the North-to-South gradient of RA seen in Europe, complementary to other genetic and environmental (e.g., sun exposure) factors ( 6 , 251 , 252 ). As an intervention, it has been shown to improve RA inflammation, pain and function ( 251 ). However, a positive effect of adherence to a Mediterranean diet could not, thus far, be demonstrated ( 281 ). This could be due to statistical and epidemiological issues, as, for example, a follow-up analysis of one of such negative studies, with a larger sample size and longer follow-up did demonstrate a benefit of healthier diet, close to the Mediterranean diet definition, in RA risk ( 269 ).

Finally, coffee, tea and caffeine have been inconsistently associated with RA. A meta-analysis including five studies (2 cohort and 3 case-control) and 134,901 participants found an increased risk with total coffee consumption (RR 2.43, 95 % CI 1.06–5.55) and no association with tea intake ( 282 ). Subgroup analyses investigating cohort studies, caffeinated-decaffeinated coffee, caffeine dose or seronegative RA were not significant, but a homogeneous modest association was seen with seropositive RA (RR 1.33, 95% CI 1.16–1.52). More recently, another large prospective cohort study ( n = 76,853) found no increased RA incidence with coffee (caffeinated or decaffeinated) consumption, whereas caffeinated tea intake conferred a 40% increase in risk (HR 1.40, 95% CI 1.01–1.93) ( 283 ). This is in disagreement with the meta-analysis and most previous studies and, as such, the link between coffee, tea and RA remains equivocal.

Socioeconomic and Other Environmental Factors

A lower socioeconomic status seems to increase RA risk, although it probably has a stronger link with poor disease outcome ( 5 , 6 ). Several studies have demonstrated that a lower level of education is independently associated with RA, particularly with seropositive disease ( 111 , 284 – 286 ). Although earlier reports did not find a significant effect of education or other socioeconomic factors on RA risk ( 287 , 288 ), the associations observed in the positive studies could not be attributed to smoking or other known socioeconomic or lifestyle factors. Moreover, low childhood (i.e., parental) household education and other poor early life socioeconomic status (food insecurity, young maternal age) have also been linked to greater development of adult RA, further supporting the positive epidemiologic observations ( 111 ). These studies suggest that socioeconomic deprivation may be an identifiable risk factor for RA, possibly through unmeasured environmental exposures (e.g., infections, low quality diet).

Additionally, a low socioeconomic status may also be more common in individuals with manual, blue collar jobs , which, likewise, have been associated with increased RA incidence ( 185 , 224 , 285 , 289 , 290 ). This association is also seen with paternal occupation ( 110 ) and may be related to various factors. First, many blue collar jobs are associated with exposure to silica, inorganic dust, textile dust and other respiratory harmful agents that, as previously discussed, are important RA risk factors ( 218 , 222 – 224 ). Second, other blue collar professionals such as auto mechanics and farmers frequently deal with mineral oil ( 289 ) and pesticides ( 291 , 292 ), respectively, both of which have been associated with RA. Interestingly, in two large women cohorts, common direct and indirect (i.e., by others, such as the spouse) pesticide exposure have been reported in adulthood ( 292 ), as well as during childhood ( 293 ), with a dose-response effect. Third, prolonged repetitive physical workload, typical of blue collar jobs, was recently revealed to be associated with an increased risk of both ACPA-positive and ACPA-negative RA, with an interaction with HLA-SE in the former ( 290 ). Fourth, a curious complementary study demonstrated that working in a cold environment increased the odds of developing RA by 50%, both ACPA-positive (60%) and ACPA-negative (40%), also with a dose-response relationship (for indoor work) and an additive interaction with another environmental work-related factor, repetitive hand/finger movements ( 294 ). Finally, other work-related factors such as work stress, conflict at work and shift work, have also been shown to increase the risk of RA ( 124 , 131 ).

In contrast to professional activity, two large separate prospective cohort studies [ n = 30,112 ( 295 ) and n = 113,366 ( 296 )] have recently shown that higher levels of recreational physical activity significantly decrease the risk of incident RA by up to 33–35% ( 295 , 296 ). This effect was found to be cumulative throughout life and was only partially explained by a decrease in BMI, being possibly linked to the anti-inflammatory properties of exercise ( 295 , 296 ).

Overall, the body of evidence regarding the etiology of, and risk factors for, incident RA that is available thus far is robust and allows for better understanding of the disease. RA can be regarded as the prototypical multifactorial immune-mediated disease. Multiple susceptibility genes, subjected to expression regulation via epigenetic mechanisms, significantly modulate the individual risk of disease. Concomitant hormonal and neuroendocrine determinants, together with comorbid conditions, further determine the likelihood of incident RA. Subsequently, throughout a subject's life, external environmental factors continuously interact with the predisposed host, slowly adding additional breaches to the immune tolerance barrier. Over time, this complex interplay gives rise to multiple cellular and molecular pathophysiological changes that culminate in the crumbling of the entire defense structure against self-aggression. Ultimately, a particular event triggers the final steps that lead to clinically evident overt disease.

A special word is warranted for the impact of cohort studies in the identification of such risk factors. Indeed, robust large-scale longitudinal prospective long-term cohort studies, such as the Nurses' Health Study ( 297 ) or the Women's Health Initiative ( 298 ), have been instrumental in the unraveling of the relationship between host and environmental factors and incident RA. Although difficult to develop and carry out, requiring extensive investment of time, funds, and effort, these are the most powerful studies to investigate risk factors of a given disease, greatly limiting potential biases of other, more accessible, and feasible study designs, such as cross-sectional, case-control or retrospective cohort studies. As an example, among many other findings, the Nurses' Health Study (NHS) has provided high-quality evidence of the association between diet ( 299 , 300 ), smoking ( 205 , 207 , 299 ), UV light ( 143 ), hormonal/reproductive factors ( 103 , 301 ), obesity ( 302 ), physical activity ( 296 ) or depression ( 169 ), and RA development. Just as the landmark Framingham Heart Study was pivotal in determining the major risk and protective factors of heart disease (e.g., diet, smoking, exercise, aspirin), saving countless lives at a global level, so too can similar large-scale prospective studies accurately inform on the risk of RA.

Besides providing valuable clues into the etiology of RA, additional insights can be gained from such epidemiological studies. The realization that genetic factors contribute to around one half to one third of RA risk, and that most of the genes implicated are directly linked to the immune system puts immune dysregulation at the core of disease pathogenesis. Moreover, the prominence of the MHC region as the main genetic risk factor, and the importance of HLA-SE molecules as key determinants of disease, confirm that antigen presentation-driven mechanisms are pivotal early pathophysiological steps. The synergic interaction between these genetic variants and environmental factors such as smoking, microbiota and diet bring additional clarity to the events taking place outside the joint. Epigenetic changes, possibly induced by external stimuli, regulate the expression of crucial genes and further determine the final risk of disease. Notably, modulating gene expression through epigenetic regulators is an attractive groundbreaking therapeutic avenue being currently pursued.

On a different level, this knowledge and understanding paves the way for an entirely novel field in RA, which is that of disease prevention ( 303 ). In fact, we are currently in the early steps of the path toward what may become, on a short-to-medium term, a new paradigm in medicine in general and rheumatology in particular: being able to correctly identify high-risk individuals at a broad population-level, and institute concrete non-pharmacological and pharmacological interventions aimed at preventing, or delaying, the onset of RA. In this regard, the first step is to stratify subjects according to risk of progression to clinically evident RA ( 304 ). Pinpointing those at high risk of developing RA, such as first-degree relatives of patients with RA, ACPA/RF-positive asymptomatic individuals, or patients with clinically suspicious arthralgia, is particularly relevant. Based on the evidence discussed above on the modifiable factors that are strongly associated with RA development, there is rationale to make specific recommendations to these subjects, which may hamper the risk of disease. These high-risk individuals may be advised to stop smoking, maintain a proper oral health, address concomitant conditions such as periodontitis, depression and obesity, and promote a healthy lifestyle, focused on a balanced Mediterranean diet, regular exercise, and stress-limiting activities. This requires the involvement of both rheumatologists and family physicians, who play a central role in health promotion and patient education. Unfortunately, it seems that this is not performed in clinical practice as commonly as desirable ( 305 ). A RCT has demonstrated that a personalized education for the risk of RA is more effective in improving healthy behaviors than standard patient education ( 306 ). However, it should be noted that these recommendations are mostly supported by epidemiological evidence, which is limited by nature. For instance, an analysis of the NHS has found that smoking cessation was associated with a decreased trend for incident RA ( 205 ). Also, while obesity has been associated with RA development, and weight loss is therefore recommended, a prospective study did not show an effect of bariatric surgery in the risk of incident RA. Unfortunately, RCTs of lifestyle interventions, which are needed to firmly establish preventive strategies, are mostly missing. An alternative approach involves pharmacological intervention in at-risk patients. This idea has recently been tested in several clinical trials. The PRAIRI study provided proof-of-concept that such a strategy could be useful, by demonstrating that a single infusion of B cell-depleting rituximab delayed the onset of arthritis in at-risk individuals ( 307 ). Other studies are investigating a similar effect with treatments such as abatacept, methotrexate or hydroxychloroquine ( 303 , 308 ). On the opposite direction, a RCT of ACPA-positive subjects with inflammatory arthralgia did not demonstrate a protective effect of two dexamethasone administrations in progression to RA ( 309 ). Similar findings were reported by a systematic literature review and meta-analysis investigating the impact of glucocorticoids, csDMARDs or biologic DMARDs for the prevention of RA in at-risk individuals without arthritis ( 310 ). Hopefully, as more data accrues, we will be able to provide additional counseling that can have a potential impact in reducing RA incidence or, at least, delaying its onset.

In summary, over the past decades there has been tremendous progress concerning the etiology and risk factors for the development of RA. While several questions remain unanswered, there is now a clearer notion of the dynamic between host and environmental factors, which sets the key pathogenic events in motion and eventually leads to a step-wise preclinical phase, where mucosal breach of tolerance is followed by systemic autoimmunity and inflammatio, ultimately targeting the articular compartment. This so-called early arthritis stage represents a window of opportunity where it may still be possible to intervene and prevent, or delay, the onset of overt disease. Novel study paths in this field have recently emerged and are likely to bring relevant contributions in the future, improving our understanding of the etiology, risk, and pathogenesis of RA. Ultimately, this will translate into better preventive and therapeutic strategies that will improve the lives and outcomes of patients with RA.

Author Contributions

Both authors have contributed to study conception and design, collected and analysed the data and drafted the manuscript. Both authors have critically reviewed the manuscript for important intellectual content, and have read and approved its final version.

VCR work was funded by Fundação para a Ciência e Tecnologia (Interno Doutorando Bursary reference SFRH/SINTD/95030/2013); European League Against Rheumatism (EULAR Scientific Training Bursary 2014); and Sociedade Portuguesa de Reumatologia (Fundo de Apoio à Investigação da SPR 2015 & 2016, to VCR).

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Commentary on the factors with liver fibrosis in rheumatoid arthritis patients treated with methotrexate

  • LETTER TO THE EDITOR
  • Published: 23 February 2024

Cite this article

  • Difeng Chen 1 &
  • Junwu Zhang   ORCID: orcid.org/0000-0002-8536-2080 2  

The Original Article was published on 30 December 2023

Avoid common mistakes on your manuscript.

Dear Editor

We read Slouma’s paper [ 1 ] with great interest. In this paper, the author aims to determine the frequency of liver fibrosis in rheumatoid arthritis patients treated with methotrexate and to identify its associated factors. Then they found that cumulating more than 3 g of methotrexate was associated with liver fibrosis in rheumatoid arthritis (RA) patients. Having a metabolic syndrome, hypoalbuminemia, higher age, and elevated alkaline phosphatase levels were also likely to be independently associated with liver fibrosis. Despite definite results, in this letter, we raise some concerns about some of the details in the article.

In this study, we noted that 20 factors were analyzed for the univariate logistic regression analysis to analyze the factors associated with liver fibrosis in RA patients treated with methotrexate, which included age ≥ 60 years, male gender, alcoholic consumption, BMI ≥ 25, abnormal WC, comorbidities, metabolic syndrome, disease duration ≥ 10 years, erosive disease, positive RF, positive ACPA, DAS28 CRP > 2.6, DAS28 ESR > 2.6, ESR > 20 mm, CRP > 8 mg/l, presence of fatty liver, steroids use, cumulated dose of MTX > 3 g, hypoalbuminemia, increased ALP, and increased AST. However, these 20 factors were all included in the univariate analysis for a total of 68 RA patients, and only 9 RA patients had liver fibrosis. However, the sample size of univariate and multivariate linear regression analysis is at least 15 times that of the analyzed variable factor [ 2 , 3 ]. The more variable factors are analyzed, the more example sizes are required. Therefore, more sample sizes are required to make the univariate logistic regression analysis in this liver fibrosis in RA patients treated with methotrexate. Without considering this important point, it could result in unreliable results.

Despite these comments, we extend our congratulations to Slouma et al. for their outstanding work.

Slouma M, Lahmar W, Mohamed G, Dhrif O, Dhahri R, Bellali H, Gharsallah I, Ebdelli N (2023) Associated factors with liver fibrosis in rheumatoid arthritis patients treated with methotrexate. Clin Rheumatol. https://doi.org/10.1007/s10067-023-06847-7

Article   PubMed   Google Scholar  

Yadav SK, Singh S, Gupta R (2019) Univariate logistic regression: theoretical aspects. In: Yadav SK, Singh S, Gupta R (eds) Biomedical statistics : a beginner's guide. Springer Singapore, Singapore, pp 219–222. https://doi.org/10.1007/978-981-32-9294-9_28

Ranganathan P, Pramesh CS, Aggarwal R (2017) Common pitfalls in statistical analysis: logistic regression. Perspect Clin Res 8(3):148–51. https://doi.org/10.4103/picr.PICR_87_17

Download references

Author information

Authors and affiliations.

Department of Clinical Laboratory, Yuyao People’s Hospital of Zhejiang Province, Ningbo, Zhejiang, China

Difeng Chen

Department of Clinical Laboratory, Wenzhou TCM Hospital of Zhejiang Chinese Medical University, 9 Jiaowei Road, Wenzhou, 325000, China

Junwu Zhang

You can also search for this author in PubMed   Google Scholar

Corresponding author

Correspondence to Junwu Zhang .

Ethics declarations

Ethical approval.

Not applicable.

Consent from all authors

All authors reviewed this manuscript and agreed to submit this manuscript.

Disclosures

Additional information, publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Chen, D., Zhang, J. Commentary on the factors with liver fibrosis in rheumatoid arthritis patients treated with methotrexate. Clin Rheumatol (2024). https://doi.org/10.1007/s10067-024-06896-6

Download citation

Received : 17 January 2024

Revised : 17 January 2024

Accepted : 29 January 2024

Published : 23 February 2024

DOI : https://doi.org/10.1007/s10067-024-06896-6

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

  • Find a journal
  • Publish with us
  • Track your research

IMAGES

  1. (PDF) Vitamin D in rheumatoid arthritis

    research paper on rheumatoid arthritis

  2. (PDF) ABC of rheumatology. Rheumatoid arthritis--I: Clinical features

    research paper on rheumatoid arthritis

  3. ⛔ Rheumatoid arthritis research paper. Rheumatoid Arthritis Research

    research paper on rheumatoid arthritis

  4. (PDF) Strategies toward rheumatoid arthritis therapy; the old and the new

    research paper on rheumatoid arthritis

  5. Aspects of Rheumatoid Arthritis

    research paper on rheumatoid arthritis

  6. (PDF) A Research Paper on "Effect of Yogic Package on Rheumatoid Arthritis"

    research paper on rheumatoid arthritis

VIDEO

  1. Rheumatology Medicine 2

  2. 1893) Rheumatoid Arthritis

  3. International Rheumatoid Arthritis day

  4. RHEUMATOID ARTHRITIS OVERVIEW

COMMENTS

  1. Rheumatoid Arthritis: A Brief Overview of the Treatment

    Rheumatoid arthritis (RA) is a chronic, symmetrical, inflammatory autoimmune disease that initially affects small joints, progressing to larger joints, and eventually the skin, eyes, heart, kidneys, and lungs. Often, the bone and cartilage of joints are destroyed, and tendons and ligaments weaken [ 1 ].

  2. Rheumatoid arthritis: pathological mechanisms and modern pharmacologic

    The prevalence rate reported in 2002 ranged from 0.5% to 1% of the population and had regional variation. 1 RA primarily affects the lining of the synovial joints and can cause progressive disability, premature death, and socioeconomic burdens.

  3. Rheumatoid Arthritis

    Rheumatoid arthritis (RA) is a systemic autoimmune disease characterized by inflammatory arthritis and extra-articular involvement. It is a chronic inflammatory disorder caused in many cases by the interaction between genes and environmental factors, including tobacco, that primarily involves synovial joints.[1] It typically starts in small peripheral joints, is usually symmetric, and ...

  4. Rheumatoid arthritis: pathological mechanisms and modern ...

    Rheumatoid arthritis (RA) is a chronic systemic autoimmune disease that primarily affects the lining of the synovial joints and is associated with progressive disability, premature death, and...

  5. Browse Rheumatoid Arthritis

    In this phase 2 placebo-controlled trial, peresolimab, a humanized monoclonal antibody that stimulates the programmed cell death protein 1 inhibitory pathway, showed efficacy in adults with...

  6. Rheumatoid arthritis

    Rheumatoid arthritis is an autoimmune disease that is characterized by inflammation of the joints and the subsequent destruction of cartilage and erosion of the bone. Patients with rheumatoid...

  7. Rheumatoid arthritis

    Abstract Rheumatoid arthritis is an autoimmune inflammatory disease primarily characterized by synovitis which is accompanied by extra-articular organ involvement, such as interstitial pneumonia, in addition to clinical symptoms including pain, swelling, stiffness of multiple joints, fever, and malaise.

  8. Diagnosis and Management of Rheumatoid Arthritis: A Review

    10.1001/jama.2018.13103 Rheumatoid arthritis (RA) occurs in about 5 per 1000 people and can lead to severe joint damage and disability.

  9. A Review on Rheumatoid Arthritis Interventions and Current ...

    Rheumatoid arthritis is a chronic autoimmune disorder characterized by inflammation, swelling, and joint destruction primarily affecting the peripheral joints. In recent years, RA has become an alarming concern affecting more than 1.5% of the population worldwide. The majority of the drugs in clinic …

  10. Rheumatoid Arthritis

    Rheumatoid arthritis (RA) is a common systemic inflammatory autoimmune disease characterized by painful, swollen joints that can severely impair physical function and quality of life. The presenting symptoms of musculoskeletal pain, swelling, and stiffness are common in clinical practice, so familiarity with diagnosing and managing RA is crucial.

  11. Top 100 cited articles on rheumatoid arthritis

    Table 1: General information related to top 20 of the top 100 cited articles on rheumatoid arthritis (sorted by total citations). Interestingly, we found that the top-cited article, which was related to the classification criteria of RA and published by the American Rheumatism Association in 1988, [20] had a total number of citations of 11,922.

  12. Serum biomarker panel for the diagnosis of rheumatoid arthritis

    Background Rheumatoid arthritis (RA) is an autoimmune disease of inflammatory joint damage, wherein C-reactive protein and autoantibodies including rheumatoid factor (RF) and anti-cyclic citrullinated peptide (anti-CCP) are rapidly elevated. These serological factors are diagnostic markers of RA; however, their sensitivity and specificity for prediction warrant improvement for an early and ...

  13. Digital health technologies and machine learning augment ...

    Digital measures of health status captured during daily life could greatly augment current in-clinic assessments for rheumatoid arthritis (RA), to enable better assessment of disease progression ...

  14. Abatacept in individuals at high risk of rheumatoid arthritis (APIPPRA

    Therapeutic intervention during the at-risk phase of rheumatoid arthritis is feasible, with acceptable safety profiles. T-cell co-stimulation modulation with abatacept for 12 months reduces progression to rheumatoid arthritis, with evidence of sustained efficacy beyond the treatment period, and with no new safety signals.

  15. Rheumatoid arthritis and dietary interventions: systematic review of

    Rheumatoid arthritis (RA) is a disease characterized by systemic immune-inflammatory responses, causing global disability. 1 In addition to joint damage, ... . 72 Thus, of the 70 papers, the above 8 papers refer to the same interventions, so the review included only 62 unique interventions. ...

  16. Management of Rheumatoid Arthritis: An Overview

    Rheumatoid arthritis (RA) is defined as a systemic autoimmune pathology associated with a chronic inflammatory process, which can damage both joints and extra-articular organs, including the heart, kidney, lung, digestive system, eye, skin and nervous system [ 1, 2 ].

  17. (PDF) Rheumatoid arthritis

    Rheumatoid arthritis is an autoimmune inflammatory disease primarily characterized by synovitis which is accompanied by extra-articular organ involvement, such as interstitial pneumonia, in...

  18. Advancements of next generation sequencing in the field of Rheumatoid

    Rheumatoid arthritis (RA) is an autoimmune and chronic inflammatory disorder associated with the joint deformities and bone erosion in the case of an individual. ... The review articles and research papers published by various reputed publishers such as Elsevier, Springer, and Taylor & Francis imprints and Hindawi were considered as the data ...

  19. Trial offers hope for millions that jab could prevent rheumatoid arthritis

    Prof Lucy Donaldson, of the charity Versus Arthritis, welcomed the findings, adding: "This research highlights how important it is to spot the early signs of arthritis to give us a chance at ...

  20. Literature review of rheumatoid arthritis in India

    Rheumatoid arthritis (RA) is an autoimmune disease with a worldwide prevalence of approximately 0.5% to 1% among adults. 1, 2 RA investigators have noted that prevalence in North America and Europe may be higher than prevalence in Asia. 3, 4 It is unclear if this geographical variability exists due to genetic, environmental or study design diffe...

  21. The Prevalence of Rheumatoid Arthritis: A Systematic Review of ...

    Objective: To estimate the prevalence of rheumatoid arthritis (RA) from international population-based studies and investigate the influence of prevalence definition, data sources, classification criteria, and geographical area on RA prevalence. Methods: A search of ProQuest, MEDLINE, Web of Science, and EMBASE was undertaken to identify population-based studies investigating RA prevalence ...

  22. FGF23 facilitates IL‐1β synthesis in rheumatoid arthritis through

    Rheumatoid arthritis (RA) is a well-known autoimmune disorder related with joint pain, joint swelling, cartilage and bone degradation as well as deformity. Fibroblast growth factor 23 (FGF23) is an endocrine factor of the FGF family primarily produced by osteocytes and osteoblasts, involves an essential effect in pathogenesis of RA.

  23. Medical imaging in rheumatoid arthritis: A review on deep learning

    1. Introduction Arthritis is a type of bone disease that affects the joints in the body and can be debilitating. It is most commonly found in the hand, knee, and finger joints [ 1 ]. Arthritis cannot be cured, but it can be managed at any stage of the disease. Pain, stiffness in the morning, and edema are the first signs of the disease [ 2 - 5 ].

  24. Radiological Progression and Predictors of Radiological Progression in

    Rheumatoid arthritis (RA) is a systemic inflammatory disorder characterised by symmetrical polyarthritis that affects both small and large joints. This leads to cartilage destruction and bone erosion, especially in seropositive patients. Uncontrolled diseases can lead to deformities and functional impairments.

  25. Clinical trial shows rheumatoid arthritis drug could prevent disease

    A drug used to treat rheumatoid arthritis could also prevent the disease in individuals deemed to be at risk. Your source for the latest research news. Follow: Facebook X/Twitter Subscribe: RSS ...

  26. The immunology of rheumatoid arthritis

    PMC8557973 10.1038/s41590-020-00816-x The immunopathogenesis of rheumatoid arthritis (RA) spans decades, beginning with the production of autoantibodies against post-translationally modified proteins (checkpoint 1).

  27. Etiology and Risk Factors for Rheumatoid Arthritis: A State-of-the-Art

    Rheumatoid arthritis (RA) is a chronic immune-mediated multisystemic disease that mainly localizes to the joints. It is the most common systemic inflammatory rheumatic disease ( 1, 2) and it is associated with considerable morbidity and disability, as well as increased mortality ( 3 ).

  28. Commentary on the factors with liver fibrosis in rheumatoid arthritis

    In this paper, the author aims to determine the frequency of liver fibrosis in rheumatoid arthritis patients treated with methotrexate and to identify its associated factors. Then they found that cumulating more than 3 g of methotrexate was associated with liver fibrosis in rheumatoid arthritis (RA) patients.