Early insights into the at-risk state of RA and terminology

Our current understanding of the RA at-risk state, defined as the state that precedes the first appearance of a swollen joint, is described in figure 1. Notably, we define the state of a swollen joint detected on physical examination as ‘clinical RA’ which may or may not also meet established classification criteria for RA. This term is used in part because that is a standard that many clinicians adopt to diagnose and then recommend disease-modifying antirheumatic drug (DMARD) treatment for RA; furthermore, the presence of a swollen joint is a requirement to fulfil the American College of Rheumatology (ACR)/European Alliance of Associations for Rheumatology (EULAR) 2010 RA classification criteria.9 However, we acknowledge that imaging is increasingly being used to identify joint inflammation—and may be used to make a clinical diagnosis of RA in some cases, even if a swollen joint is not detectable on physical examination. As discussed in more detail below, the use of imaging to identify joint inflammation as a risk factor for future clinical RA, or used to define that clinical RA is present, is both an opportunity as well as a challenge that needs to be addressed.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Disease prevention in the RA at-risk state. The different stages of the at-risk state are illustrated, aligned with potential interception checkpoints, with the choice of intervention reflecting the risk:benefit ratio. Although increasing risk of imminent RA is illustrated going left to right, progression through each stage is not inevitable for all at-risk individuals. It remains to be determined whether immune modulation could be considered for those at lower risk who have yet to develop evidence of autoimmunity. RA, rheumatoid arthritis.

Furthermore, at this time, we have favoured the term for the stage of RA that precedes the first swollen joint and where risk factors for RA are present as ‘at risk of RA’ or an ‘at-risk state’, rather than pre-RA. This is because pre-RA, a term not currently favoured by patient research partners unless used retrospectively in individuals who have developed clinical RA, implies a disease trajectory that is inevitable,10 while findings from cohort studies as well as the clinical prevention trials clearly demonstrate that not all individuals who have risk factors for RA (and may be considered in an at-risk state) progress to clinical RA (although it is acknowledged that some conditions commonly use ‘pre’ to define certain stages of disease for example, pre-diabetes11). Notably, the at-risk state can be further clarified through ‘risk stratification’, where there may be an overall increased risk based on the presence of some risk factors, but there are also varying levels of risk within that—we discuss this important point in more detail below. In addition, we adopt the term ‘imminent disease risk state’ for those individuals in whom risk stratification suggests that development of clinical RA is highly likely to occur within 12 months, although we acknowledge that accurately identifying individuals within this imminent state of developing clinical RA is a challenge that remains to be met, and we additionally discuss this in more detail below. We also wish to highlight that the constellation of symptoms (eg, joint pain, stiffness, swelling and fatigue) that can be identified in part using patient-reported outcomes may constitute an illness that requires intervention, even if clinical RA is not present. Moreover, here, we use the term ‘interception’ to describe an intervention that is used to effect prevention of clinical RA and may use the terms ‘intervention’ and ‘interception’ interchangeably.

Defining, identifying and stratifying risk in individuals who are at risk for RA

Retrospective studies exploiting biobank samples have allowed the identification of biomarkers associated with the at-risk state, including autoantibodies and inflammatory markers such as cytokines and chemokines (figure 1 and table 1).12–16 While these studies have provided important insights into aspects of the natural history and biology of RA, some have limitations in that participants typically were not evaluated in real-time during the ‘at-risk’ period to determine the precise relationships between the evolution of autoantibodies and other markers and signs and symptoms of clinical RA. The case–control nature of many of the studies also limits the ability to develop robust prediction models for future RA. Fortunately, however, prospective studies of individuals at risk for future RA have been especially informative regarding these key points.

There are multiple approaches for prospective studies of the ‘at-risk’ state in RA (table 2, figure 2). These can include identification and study of individuals with genetic/racial or familial risk factors for RA. For example, individuals from populations who have high prevalence rates (eg, some North American indigenous groups) may be considered ‘at risk’ for RA.17 Furthermore, relatives of an individual with established RA may be considered ‘at risk’ due to the known increased prevalence of RA in families.18 Indeed, prospective studies of North American indigenous populations and family members of patients with RA have yielded important insights into the natural history and risk factors of RA development.19–23

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Emerging features of the at-risk state. The core determinants of risk are highlighted, with emphasis on genes and environmental exposures, identification of immune signatures, symptom complexes and the use of imaging modalities to detect subclinical inflammation. Importantly, we need to understand the interplay between genetic and other host factors, the environment and the initiation and propagation of immune dysregulation and the development of clinical RA so that these processes can be targeted for effective prevention. *EULAR definition. AMPA, antibodies to modified protein antigens; EULAR, European Alliance of Associations for Rheumatology; MRI, magnetic resonance imaging; PET, positron emission tomography; RA, rheumatoid arthritis; SPARRA, the Symptoms in Persons At Risk of Rheumatoid Arthritis; US, ultrasound.

Much of the prospective data relative to the at-risk period has come from approaches which have focused on determining an at-risk status that is based on the presence of musculoskeletal (MSK) symptoms such as arthralgia in combination with biomarkers of autoimmunity (table 2, figure 2). This approach has yielded critically important findings related to symptom complexes, immunobiology and natural history of clinical RA development and has supported the development of models for risk stratification for the development of RA (see below). This is also a method by which most clinical rheumatologists encounter individuals in the at-risk period and therefore provides ‘real-world’ information on the natural history of RA that may be applicable to clinical care settings. Furthermore, individuals at risk for RA can be identified in ‘real time’ by this approach and can be enrolled in observational studies that can be used as at-risk cohorts for recruitment into prevention trials (see below). To generate a definition that can distinguish symptomatic at-risk patients from a broader population with MSK symptoms, an EULAR task force defined the clinical characteristics of patients with arthralgia who are considered at risk of RA by experts, based on their clinical experience. This resulted in the EULAR definition of arthralgia suspicious for progression to RA, which was developed primarily to identify a more homogeneous group of symptomatic individuals for future scientific investigation.24

Other methods of identifying individuals at risk of RA include population-based testing for RA-related biomarkers through activities such as health fairs or population surveys, with testing done through the interest of participants regardless of the a priori presence of symptoms or other risk factors. Since broad-scale biomarker testing to evaluate risk for future RA is not yet standard of care, these approaches are best considered as ‘research only’. However, such approaches may help to identify larger numbers of at-risk individuals in the very earliest stage of risk (figure 1) at a time when the impact of interventions targeting reversible risk factors could be significant. Moreover, future biomarker-based prevention strategies, if approved, could be similar to strategies used now in assessing CVD risk through cholesterol testing, or diabetes risk through autoantibody and glucose testing.

As discussed above, once an individual has been defined as being ‘at risk’ for RA because of the presence of one or more risk factor, the next important challenge is to quantify that level of risk, because within broad definitions of risk discussed above (eg, family history and MSK symptoms), there is still high variability in progression to clinical RA. This approach can be termed ‘prediction’, although the concept of ‘risk stratification’ is perhaps more applicable if a goal is to use a variety of factors to identify an individual’s (or groups of individuals) specific level of risk for developing future RA.25 26 As an example, osteoporotic-related fracture risk can be stratified using approaches such as the ‘FRAX’ score, and preventive treatment for potential fractures started when a certain risk state for a fracture is reached.4 Notably, there is also a potential to include individuals at risk for future clinical RA who have a current clinical syndrome (eg, arthralgia) that warrants a diagnosis and treatment now; that concept will be discussed below in more detail.

There are several published studies in which formal risk stratification models for future clinical RA have been presented (table 3). These approaches have used a variety of risk factors including family history of RA, symptoms, autoantibodies, inflammatory markers (eg, erythrocyte sedimentation rate), genetic tests and imaging findings of ‘subclinical inflammatory arthritis’ (which can be defined as imaging findings of joint/tendon inflammation without a swollen joint on physical examination). In general, for anticitrullinated protein antibody (ACPA)(+) individuals who were found on clinical evaluation to have MSK symptoms, rates of development of clinical RA within 1–5 years range between 20% and 40%; however, with additional factors, such as subclinical inflammation defined by imaging, positive predictive values (PPVs) can be in the region of 70%–80% (table 3). Nonetheless, the strongest risk factors for RA are the presence of autoantibodies, and, not unexpectedly, dual positivity of ACPA and rheumatoid factor is the strongest autoantibody predictor.27–29 Detection of high levels of glycosylation of the ACPA IgG variable domain adds to risk when evaluated over time, is quantifiable and may be reversible.30 31

Risk stratification criteria for use in research are being developed by an EULAR/ACR task force, taking advantage of data from 10 arthralgia at-risk cohorts and the expertise from a group of rheumatology experts, patient partners and health professionals from Europe and North America. After a data-driven approach, consensus was obtained. This initiative resulted in criteria that incorporated clinical, serological and imaging variables which could be adapted for use in the presence or absence of imaging to detect subclinical joint inflammation (AUCs (area under the curve) of 0.80 and 0.87, respectively). Endorsement of these criteria by EULAR and ACR and a formal publication are underway; furthermore, an additional task force that will explore risk stratification in individuals identified in population-based approaches (eg, ACPA testing of FDR or at health fairs) is also underway.

These risk stratification models are highly informative and indeed have underpinned the design of clinical prevention trials described below. However, there are some caveats. First, these models have largely been developed in individuals who initially presented to clinical care providers with MSK symptoms. As such, we know less about prediction in individuals who may be found to be at risk through population-based approaches and not initially identified as being at risk for RA because they presented to healthcare programmes because of MSK symptoms/arthralgia. To this point, emerging data suggest that ACPA positivity, even if identified through population-based approaches that include ACPA testing at health fairs or of FDR is still associated with PPVs of developing RA over longer periods of time of ~20%–50%.22 23 29 This finding suggests that population-based approaches may identify individuals earlier in the at-risk period; furthermore, these approaches may achieve a higher priority if clearly effective treatment were available for people at risk. Second, these models do not include estimates of severity or persistence of clinical arthritis that may develop. This is important because individuals with MSK symptoms, or those found through population-based approaches may have a milder form of RA than those who are identified with clinical RA through referral to specialist centres (which could represent time length bias). Furthermore, given the growing understanding of ‘palindromic rheumatism’, it is possible that these approaches identify individuals who will not develop persistent arthritis. Indeed, it is emerging that findings such as imaging evidence of ‘subclinical inflammation’ may resolve over time without DMARD therapy, even in autoantibody-positive individuals deemed to be at risk.32 33 Importantly, some clinicians may use subclinical inflammation to diagnose clinical RA and initiate treatment34 which based on these findings may result in potential overtreatment35 36 (and a concept to reduce overmedicalisation is termed ‘quaternary prevention’7). To address this point, going forward the field will need to develop clear guidelines on how to incorporate imaging into clinical trials as inclusion factors and outcomes, as well as how to use imaging in clinical care to establish a diagnosis of clinical RA. Third, risk stratification/prediction models do not clearly define specific biological states that may represent an individual’s unique ‘endotype’ for developing RA—and, in particular, specific biological pathways to target. This is key because ideally risk stratification should both identify an individual’s risk for future clinical RA but also provide insights into what factors should be targeted to reduce that risk—in essence, more personalised prevention akin to individually treating cholesterol or blood pressure abnormalities in CVD prevention. This becomes especially important in at-risk individuals in whom synovitis is absent. Finally, as discussed below, clinical prevention trials have enrolled individuals in part based on the risk stratification models presented in table 3. In some of the clinical trials, the rates of progression to RA within the placebo arms are lower than would be expected using the published models that were developed in observational studies. This may be because individuals who participate in clinical trials are somehow ‘healthier’, perhaps due to lifestyle factors, than the at-risk populations included in the observational studies. Clearly, prediction models need refinement, and as mentioned above there are efforts underway sponsored by EULAR and the ACR to develop risk stratification models in both arthralgia-based and population-based approaches, and these projects should yield important results that will address many of the challenges in risk stratification that are discussed above.

Summary of RA interception trials

The ability to identify individuals who are at risk for RA, as well as estimate the rates of progression to RA, have underpinned the execution of published clinical trials. The chronology of these studies is depicted in figure 3, with details summarised in table 4. Importantly, in the design of these trials, the inclusion criteria were developed to identify individuals with a range of PPVs for future clinical RA of 30%–50%. This is not a 100% ‘risk’ of developing clinical RA; however, ethical review boards and regulatory agencies have considered that these PPVs are sufficient, given the known potential risks and benefits, to approve the conduct of these trials.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Milestones in the history of the RA at-risk state. Clinical trials in individuals at-risk of RA are illustrated, underpinned by consensus guidelines and points to consider led by task forces of EULAR and the ACR. ACR, American College of Rheumatology; APIPPRA, abatacept in individuals at high risk of rheumatoid arthritis; ARIAA, abatacept inhibits inflammation and onset of rheumatoid arthritis in individuals at high risk; EULAR, European Alliance of Associations for Rheumatology; HCQ, hydroxychloroquine; PRAIRI, prevention of clinically manifest rheumatoid arthritis by B-cell directed therapy in the earliest phase of the disease; RA, rheumatoid arthritis; RCT, randomised clinical trial; STAPRA, statins to prevent rheumatoid arthritis; StopRA, strategy to prevent the onset of clinically-apparent rheumatoid arthritis; TREAT EARLIER, intervention with methotrexate in patients with arthralgia at risk of rheumatoid arthritis to reduce the development of persistent arthritis and its disease burden.

In terms of interception, corticosteroids alone,37 a single dose of rituximab (with corticosteroids),38 atorvastatin,39 methotrexate (with single-dose corticosteroids and noting 23% ACPA positivity)40 and hydroxychloroquine41 did not reduce overall rates of progression to clinical RA within the study periods. However, in these trials, rituximab delayed the onset of clinical RA, and methotrexate was associated with both improved function and diminished evidence of inflammation seen on MRI of the joints. In addition, in two trials, compared with placebo, abatacept-treated participants had reduced rates of progression to clinical RA within the trial period (although within both arms a proportion of participants did not develop clinical RA).42 43 Moreover, in APIPPRA (Abatacept in individuals at high risk of rheumatoid arthritis, ultrasound detected synovitis and power Doppler scores were reduced by abatacept.42 In ARIAA (Abatacept inhibits inflammation and onset of rheumatoid arthritis in individuals at high risk), abatacept reduced MRI inflammation,43 findings consistent with the imaging outcomes reported in the ADJUST study which trialled abatacept in individuals with undifferentiated IA at baseline.44 Interestingly, while not specifically designed for RA prevention, VITAL (Vitamin D and omege 3 trial) tested the effects of 5 years of supplementation with vitamin D, omega-3 fatty acids, both, or neither in men over 50 and women over 55 years of age at enrolment on the prevention of several prespecified outcomes, including a composite of all incident autoimmune diseases. Vitamin D, with or without omega-3 supplementation, significantly reduced rates of development of overall autoimmune disease, with signals for reduction of several individual autoimmune diseases, including RA.45 Furthermore, in additional follow-up 2 years beyond the initial 5 years of randomised intervention in the VITAL trial, omega-3 fatty acid supplementation was shown to have significantly lowered the rate of autoimmune disease development, while that for RA remained low but not significantly so, suggesting a longer-term effect of this agent.46

There are several important considerations when evaluating the results of these trials, and in particular, when contemplating the development of future prevention studies in RA. First, most of these studies recruited individuals who had first presented to healthcare providers with MSK symptoms; exceptions are StopRA (Strategy to prevent the onset of clinically-apparent rheumatoid arthritis) where one-third of participants were identified through population-screening approaches (eg, ACPA testing at health fairs) (unpublished data), and the VITAL trial which enrolled participants from the general population. This MSK symptom-driven approach allowed recruitment activities to leverage existing clinical ‘pipelines’ where individuals with MSK symptoms are evaluated for additional risks for RA. It also likely enriches for a subset of at-risk individuals with a prolonged prodrome of symptoms sufficiently severe to warrant seeking medical care, and excluding those with disease that is more subtle, or in contrast, explosive in onset, or that has waxing and waning or palindromic symptoms. Second, most of the studies used medications that have already been approved for use in clinical RA. This was based on assumptions that since these medications have proven efficacy in clinical RA they may offer benefit during the ‘at-risk’ stage. It is conceivable that the use of treatments approved for RA facilitated both approval by ethics committees as well as the willingness of at-risk individuals to enroll. Third, most of the clinical trials adopted fixed period dosing of the Investigational Medicinal Product under investigation (and in the case of rituximab, a single infusion), with a drug-free follow-up period to evaluate durability of any response. This design has the advantage of improving drug adherence, safety and costs while addressing the notion head-on that interception may have durable effects even after cessation of therapy. However, future studies may need to consider longer duration of therapy, intermittent dosing regimens or combinations of agents to suppress inflammation and restore immune tolerance. Fourth, these studies had differing inclusion criteria. For example, several studies required the presence of ‘arthralgia’ for inclusion; however, while there are formal EULAR-approved criteria for arthralgia that are suspicious for progression to RA,47 these criteria were not available at the inception of many studies that predated them. As such, the type of MSK symptoms leading to inclusion are likely to differ and may have influenced rates of progression. In addition, the primary endpoints differed between studies, although the majority were based on the development of, or time to developing clinical arthritis or RA detected by physical examination. Finally, as discussed above, the overall rates of progression to clinical RA within the placebo arms of these studies varied from as low as 14% to as high as 57%. This demonstrates, unambiguously, that risk over time varies between studies and highlights the need for more robust risk stratification tools and application of interventions that more closely align with that level of risk.

Are we equipped for the next phase of RA interception trials?

It is exciting that the field has advanced to the point when several randomised clinical trials in RA prevention have been completed. In particular, while TREAT EARLIER did not reduce rates of progression to RA, it is encouraging that sustainable improvement in subclinical joint inflammation, physical function, work productivity and symptoms in methotrexate-treated individuals highlights a clinical trajectory of RA modulated by interventions in the at-risk period40; it is also encouraging that a limited course of abatacept reduced rates of progression to clinical RA within the trial period, reduced multiple aspects of the symptom burden and reduced subclinical joint inflammation (APPIPRA and ARIAA).42 43 Nonetheless, time-limited interventions were not sufficient to conclusively ‘prevent’ RA, indicating that challenges remain to further advance prevention in RA. These are summarised in box 1 and addressed in more detail below. Importantly, these challenges have been, or are being, addressed by a number of EULAR/ACR task forces, building on the existing framework for studies in the at-risk period summarised in figure 3.

Further considerations for study design and interventions

The points highlighted above underscore the premise that RA interception should be commensurate with the at-risk state, taking into account the risks and benefits of any particular intervention. Another factor influencing the choice of intervention relates to what at-risk individuals would find acceptable and how this relates to perception of risk, familiarity with RA as well as the drug safety profile.48–50 Lifestyle modifications, such as cessation of smoking, might be considered more acceptable for asymptomatic individuals with genetic and environmental risk factors (figure 1), especially if it could be demonstrated that those interventions are appropriate for the biological stage of disease51; for example, if we know that tobacco smoke is key to propagating RA-related autoimmunity at a certain stage, then smoking cessation for an individual who is a smoker may be a ‘precision-medicine’ preventive approach. Furthermore, qualitative studies have demonstrated that some lifestyle interventions would be acceptable to individuals who are at risk for RA.52 53 Notably, lifestyle and dietary factors such as exercise and healthy diets are associated with decreased risk of RA in observational studies,54 and specific interventions including whole-food plant-based diet, physical activity and stress management have demonstrated benefits in improving disease activity in clinical RA.55 56 As such, these types of approaches may be primary interventions for prevention; however, they can be challenging to implement in clinical trials as isolated interventions in part due to adherence as well as potentially long periods of time required to see biological effect (eg, observational studies suggest smoking cessation may be associated with RA risk reduction in 10 or more years57). As such, lifestyle/dietary interventions may need to be included in trials as additions to other preventive interventions. In addition, the presence of disease-specific serum autoantibodies or other markers of autoimmunity (ie, autoantigen-specific T cell reactivity) could justify interventions such as antigen-specific immune tolerance adopting cell and/or peptide-based approaches.58 Furthermore, novel approaches may be instituted that mitigate mucosal-associated processes in gums, gut, respiratory or reproductive tracts that, over time, drive initiation and propagation of immune dysregulation, with candidates for specific interventions being identified through mucosal testing. Notably, while some interventions may be biologically appropriate and acceptable to individuals who are in an ‘asymptomatic’ stage of RA development, the presence of autoimmunity and immuno-inflammatory responses, combined with symptoms and signs of joint inflammation, is a stage where there would be more justification for therapies that target the immune dysregulation suspected of driving these early events. For example, a biological agent that inhibits the expression and function of inflammatory mediators such as TNF (tumor necrosis factor) or IL-6 (interleukin-6) may suppress but not truly modulate the early, asymptomatic at-risk phase but would be more likely to be of benefit at a point when subclinical joint inflammation is detectable by imaging modalities such as MRI or ultrasonography. In contrast, therapies that target the earliest phase of adaptive immune responses may be more plausible options as soon as evidence of adaptive immunity becomes manifest, and these pathways are active. These concepts, supported in part by animal models,59 highlight the need for further interrogation of the molecular and cellular pathways that underpin each phase of the at-risk state.

Regardless of the specific intervention, each approach needs to be not only clinically impactful but also have a favourable risk-to-benefit ratio. Given that in the at-risk state the requirements for assessing ‘activity’ will be quite distinct from those adopted for the assessment of established RA, the ability to monitor cellular and molecular signatures of a therapeutic response will become a priority, aligned with evidence of restoring immune homoeostasis, such as seroconversion. In this case, the risk, such as ACPA, and the state of immunoglobulin variable domain glycosylation might be deemed targetable therapeutically. Notwithstanding issues around monotherapy and immunogenicity, the success of any clinically impactful intervention will require adequate adherence to the intervention, and that in turn will be determined by pragmatic routes of administration and dosing schedule deemed acceptable to an at-risk population. Linked to this are the uncertainties surrounding the choice of dosing period, and ongoing risk monitoring to determine the need for retreatment. While treatment with hypertensive agents and lipid-lowering drugs may be lifelong, the possibilities of fixed period dosing, or intermittent dosing (eg, based on the APPIPRA and ARIAA trials perhaps intermittent dosing of abatacept, especially if there are biomarkers to identify individuals who may respond to additional therapy), remain appealing, not least because of the positive impact in terms of safety. In such scenarios, monitoring the perturbations of host immunity or durability of immune tolerance becomes all the more important.

Lessons learnt

The growing portfolio of RA interception studies has been underpinned by input from patients and individuals who are at-risk for RA, and combined with experiences of enrolling at-risk individuals into interception studies (key features are illustrated in figure 4). This must continue, with added emphasis on access to care. More consideration should also be given to how physicians communicate risk for developing RA in the clinical setting, learning from other disciplines, such as CVD and clinical genetics, as well as explaining the benefits of enrolling in prevention studies. Our collective experience has highlighted the advantage of at-risk individuals being included in such studies, reducing the burden on clinicians while providing a referral pathway to healthcare systems for prompt treatment when RA develops. Importantly, one of the major challenges in executing the clinical trials for RA prevention was finding at-risk individuals who were further willing to participate in trials. Learning from the success of enrolment for numerous clinical prevention trials for type 1 diabetes mellitus (T1DM) which relies on international networks to identify individuals who are at risk for disease,60 we need networks in the rheumatology community that can identify and follow individuals who are at risk for RA. From such individuals we can learn more about the natural history of RA and the pathophysiology of disease evolution as well as their preferences through research codesign. These are also important cohorts for recruitment for relevant clinical trials. Several regions including The Netherlands and the UK have established clinical referral networks,27 28 61 but we will need international efforts for representative codesign and to enhance recruitment feasibility for trials.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Recruitment and retention in RA interception studies. Core features of study operations are depicted, from screening, participant enrolment through to measures that can be adopted to support retention. The knowledge gained, including input from patient experts, feeds forward into risk screening programmes and refining trial design. IMP, investigational medicinal product; MSK, musculoskeletal; RA, rheumatoid arthritis.

The development of fit-for-purpose instruments for assessing the symptom burden will become paramount.62 Several important studies, informed by patients as well as individuals who are at risk for RA, have investigated the wide constellation of symptoms associated with the at-risk state. Besides pain and fatigue, these include physical and emotional well-being, weakness and loss of strength, anxiety and depression, sleep disturbance and work instability, among many others.63–65 These instruments require further validation in the context of at-risk states and RA prevention studies, not only to better quantify the burden of symptoms on quality of life, but to evaluate reversibility following intervention and hence clinical impact. The development of such clinical disease activity tools (such as a pre-RA-validated disease activity measure akin to the Clinical Disease Activity Index that is used as an outcome in clinical RA66) will benefit from the emerging RA prevention trial datasets. Furthermore, while imaging with US and MRI appear sensitive to change in at-risk individuals with subclinical inflammation,40 42 43 experience from the APIPPRA and TREAT EARLIER (Intervention with methotrexate in patients with arthralgia at risk of rheumatoid arthritis to reduce the development of persistent arthritis and its disease burden) studies suggests that plain radiographs of hands and feet may not be sensitive enough to capture short-term radiographic outcomes.40 42

Finally, there have been no safety signals of concern reported between the placebo and treatment arms from RA prevention studies to date. This may reflect the relatively small trial cohort sizes, safety of the individual drugs used in these studies, inclusion/exclusion criteria that optimise safety as well as the trial designs that include a shorter duration of dosing than in trials of participants with clinical RA, and ‘monotherapy’ with study drug (or combination with only a short course of corticosteroid). Other factors that influence safety, especially compared with trials including clinical RA, are differences in age, the burden of comorbidities in those at risk, as well as the fact that at-risk individuals have not been exposed to multiple immune-modulating therapies at the time of interception. Nonetheless, evaluating the balance between risk and benefit and minimising unnecessary exposure to immune-modulating drugs remains paramount, given that often more than 50% of at-risk individuals recruited to the placebo arms of prevention trials to date do not progress to RA within the time frame of the trials.

Importantly, when thinking of the feasibility of making prevention actionable in RA, it is relevant to note that prevention strategies are already undertaken in some autoimmune diseases such as the use of antibiotics to treat streptococcal pharyngitis, which prevents the subsequent development of myocarditis, glomerulonephritis or other autoimmune manifestations.67 In addition, in T1DM, the presence of multiple autoantibodies to certain antigens (eg, insulin, GAD65, ZnT8A and IA-2A) are highly predictive of future continued loss of insulin production and the onset of symptomatic glucose elevations (a state-designated stage 3 T1DM).68 The predictive power of autoantibodies for future stage 3 T1DM has prompted an increasing number of clinical trials (which as mentioned above have leveraged international networks of at-risk individuals for enrolment). Indeed, the immune intervention teplizumab (a humanised anti-CD3 monoclonal antibody immune intervention) is now approved by the Food and Drug Administration in the USA for individuals who have elevations of two or more T1DM-related antibodies and abnormal glucose tolerance testing (a state called ‘stage 2 T1DM’) in order to prevent or delay transition to stage 3 disease.69 70 Since the approval of teplizumab in stage 2 T1DM, various programmes for islet autoantibody screening have been established for siblings of children with T1DM, and population-based childhood screening is becoming increasingly available. We can envision similar approaches in RA once preventive interventions have been approved.

Looking to the future

Looking ahead, it is apparent that additional therapeutic targets and strategies, as well as improved risk assessment tools, will have to be developed to achieve the desired clinical outcomes in RA prevention; this will be especially important for seronegative RA. One major limitation is incomplete knowledge of the mechanisms by which the RA disease process is initiated during the at-risk phase and the molecular drivers of the initial break-in tolerance, epitope spreading, IgG V-domain glycosylation and immune targeting of the joint itself.71 72

To that point, recent studies by a number of groups support a ‘mucosal origins of RA hypothesis’.73 This hypothesis posits that the initial systemic break in immune tolerance to self-antigens occurs in association with chronic inflammation and/or dysbiosis at mucosal sites, such as the lungs, the periodontium and the gut. The local drivers are likely to be microbial in origin and to use diverse mechanisms.74–77 In the absence of resolution, these localised immune processes transition into a systemic process that targets the joints, either by direct effects of microbiota, molecular mimicry and/or immune amplification. This, in turn, leads to inappropriate engagement of a range of effector mechanisms in both synovium and periarticular sites.

Thus, similar to T1DM the at-risk stage of RA could be considered as a continuum, progressing through a number of ‘stages’ during which immune checkpoints that limit progression to autoimmunity fail (figure 5).72 Applying this concept, one could envision first identifying the stage at which an individual sits within this at-risk continuum and then applying the most appropriate prevention strategies. It follows from this that such an approach would have to be personalised not only in terms of the magnitude of risk, but also the underlying pathogenic mechanisms at the affected mucosal site, lymphoid organs and synovium.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5

Demonstration of potential checkpoints that fail during the natural history process leading to the development of clinical RA. Considering that the ultimate development of RA follows failed mechanisms that should in principle stop the pre-RA disease process, shown are failed checkpoints that could in principle be therapeutically targeted in at-risk individuals. Some of the mechanisms may also provide benefits for patients with clinical RA. RA, rheumatoid arthritis; RF, rheumatoid factor.

This staged/checkpoint approach to prevention, therefore, suggests a number of possibilities for prevention.72 78 With regard to causal microbiome factors, approaches could include probiotics, prebiotics and dietary intervention, treatment of periodontal disease that may drive dysbiosis, specific antibiotic treatment, faecal microbiome transplantation, use of strain-specific bacteriophages, alterations of antibodies active in the gut lumen, modulation of microbial processes such as metabolites/lipids that could affect the host or use of commensal bacterial strains that could abrogate effects of ‘pathogenic’ organisms. With regard to the mucosal barrier and local immune responses, known modulators of inflammation could constitute therapeutic targets, including TNF, IL-12/IL-23, IL-17, sphingosine-1-phosphate (S1P), JAKs (Janus kinases) and lymphocyte homing receptors such as α4β7. Locally delivered therapies are also possible, such as corticosteroids and DNAse, as well as pro-resolving modulators such as omega-3 fatty acids. Finally, with regard to modulation of a systemic immune response, an attractive series of drugs are being developed. Beyond utilisation of CTLA4-Ig as in the APIPPRA and ARIAA trials, targeting of the CD40-CD40L pathway, immune modulators of T cells such as teplizumab (which was successful in delaying onset of T1DM69) and modulation of T follicular helper (Tfh) and/or T peripheral helper (Tph) differentiation, migration or activity are worth consideration. Likewise, restriction of antigen presentation of citrullinated targets by peptidyl arginine deiminase inhibitors could provide benefit, as would induction of antigen-specific tolerance.