Site-specific articular responses to b/tsDMARDs in rheumatoid arthritis: a longitudinal observational study from the Swiss Clinical Quality Management Registry

Abstract

Objectives

Whether treatment response differs across affected joints in rheumatoid arthritis (RA) remains unclear. We sought to determine whether responses to tumour necrosis factor inhibitors (TNFi) and other biologic/targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs) vary across individual joints in a large cohort of patients with RA.

Methods

Patients with RA from the Swiss Clinical Quality Management in Rheumatic Diseases registry with ≥1 swollen joint at treatment initiation of a b/tsDMARD (abatacept, interleukin-6 receptor inhibitor (IL-6Ri), Janus kinase inhibitor (JAKi), rituximab and TNFi) were included. Time to resolution of swollen joints was evaluated over a 2-year follow-up using mixed effects models accounting for interval-censored data. The analyses compared the speed of improvement of individual joints within the 28-joint count to that of the wrist. The primary analysis focused on bio-naïve patients initiating TNFi treatment, while secondary analyses included all b/tsDMARD classes, irrespective of treatment line.

Results

A TNFi was initiated in 598 patients with ≥1 swollen joint, while 1942 patients with ≥1 swollen joint started a b/tsDMARD (abatacept: 242, IL-6Ri: 317, JAKi: 333, rituximab: 176, TNFi: 874). Compared with the wrist, all joints demonstrated higher rates of resolution, except for the second and third metacarpophalangeal (MCP2/3) joints and, in some analyses, the knee. This pattern of joint-specific response was consistent across all b/tsDMARDs.

Conclusions

The wrist, MCP2 and MCP3 joints resolve more slowly than other joints following treatment with b/tsDMARDs, indicating that a longer evaluation period of treatment response and/or bridging intra-articular steroid injections may be required when these joints are affected.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• A site-specific treatment response pattern has been demonstrated in two large European cohort studies of patients with psoriatic arthritis (PsA) treated with tumour necrosis factor (TNFi) and interleukin-17 inhibitors (IL-17i).

WHAT THIS STUDY ADDS

• Similar to findings in PsA, the wrist and metacarpophalangeal (MCP) 2 and 3 joints show a significantly slower treatment response to TNFi compared with other joints in rheumatoid arthritis (RA).

• This joint-specific response pattern was consistent across all advanced treatment modalities, including abatacept, interleukin-6 receptor inhibitors, Janus kinase inhibitors, rituximab and TNFi.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Evaluation of treatment response in RA may require a longer period if the wrist, MCP 2 and 3 joints are involved.

• Future studies should investigate whether intra-articular steroid injections in these specific joints, while awaiting the full therapeutic effect of an initiated biologic or targeted synthetic disease-modifying antirheumatic drug, could compensate for their delayed response in RA.

Introduction

Rheumatoid arthritis (RA) is a chronic inflammatory polyarticular disease.¹ Although multiple effective therapeutic strategies exist, some joints remain persistently inflamed or develop recurrent inflammation despite treatment.^2–4 Such insufficient local disease control can contribute to the progressive accumulation of joint damage.⁵ Current clinical management follows a treat-to-target approach based on composite disease activity measures that include summarised joint counts.⁶ However, the potential for joint-specific variability in treatment response has not yet been systematically investigated.⁶

Growing evidence suggests that the anatomical location of a joint may affect and potentially interfere with treatment response. In two large European Spondyloarthritis Research Collaboration Network cohort studies in psoriatic arthritis (PsA), joint-specific differences in response to tumour necrosis factor inhibitors (TNFi) and secukinumab were observed: the wrist, metacarpophalangeal (MCP) 2, MCP3 and proximal interphalangeal (PIP) 3 joints showed the lowest rate of synovitis resolution within the 28-joint count.^{7 8} In addition, a post hoc analysis of the ORAL start trial in RA suggested a possible site-specific variation in treatment response within the 66/68 joint count, revealing differences between patients treated with methotrexate (MTX) and those receiving tofacitinib.⁹ Several mechanisms may underline these patterns. Differences in local mechanical stress exposure have been proposed as one factor contributing to joint-specific susceptibility.^10–13 Additionally, emerging data indicate substantial anatomical differences in the epigenomic and transcriptomic profiles of synovial fibroblasts across joints.^14–20

With multiple treatment options available for RA, precision medicine²¹ has the potential to reduce the trial-and-error burden by aligning treatment choices with an individual patient’s disease phenotype. To date, it has not been explored whether the anatomical location of affected joints in RA influences the likelihood of response to specific biologic and targeted-synthetic disease-modifying antirheumatic drugs (b/tsDMARDs)—for example, whether inflammation in the MCP2 joint responds more favourably to a Janus kinase inhibitor (JAKi) than to a TNFi. Demonstrating that joint location or patterns of involvement affect treatment responsiveness would open a new and clinically meaningful avenue toward precision medicine in RA.

The aim of this study was to determine whether the specific joint-response pattern previously identified in PsA is also present in RA, and to compare joint-level responses to advanced immunosuppressive therapeutic agents with different modes of action, including abatacept (ABA), interleukin-6 receptor inhibitor (IL-6Ri), JAKi, rituximab (RTX) and TNFi.

Methods

Study population

Patients with inflammatory rheumatic diseases, including RA, are routinely monitored in the Swiss Clinical Quality Management in Rheumatic Diseases (SCQM) registry through structured assessments that include a standardised evaluation and documentation of joint involvement using the swollen and tender 28-joint count at every visit.²² Follow-up visits are recommended at least 3 months after treatment start or modification, and at least annually for patients with stable disease.

Adult patients from the SCQM registry, who were diagnosed with RA by their treating rheumatologist, were included in the current study if they initiated a b/tsDMARD approved for RA in Switzerland and had at least one swollen joint in the 28-joint count at treatment start (sensitivity tender joint analyses, see below, were performed in patients who had at least one tender joint at treatment start). Patients with treatment initiation between 2005 and 2025 were considered. For agents with different modes of action, the study period varied according to the date of their approval for RA in Switzerland (online supplemental table S1). Co-medication with MTX was allowed, since prior studies in PsA demonstrated that it does not alter the joint-specific response to TNFi.⁷

The choice of treatment courses depended on specific analyses within this study (online supplemental table S1):

• Part 1: For the analysis of joint-specific treatment response to TNFi, only treatment courses with a first TNFi in bio-naïve patients were considered.

• Part 2: For the analysis of joint-specific treatment response to individual b/tsDMARDs treatment courses with ABA, IL-6Ri, JAKi, RTX and TNFi were considered, independently of the number of previously used b/tsDMARDs. Patients with registration of two concomitant b/tsDMARDs were excluded.

• Part 3: For the direct comparison of joint-specific treatment responses across different treatment modalities, only bio-naïve patients were included and compared with TNFi as the reference treatment. The inclusion period was adapted so all medications were available during this time period.

Baseline patient characteristics were collected at the time of treatment start including clinical and demographic information. Job type was defined as blue-collar and white-collar worker according to the following definition: Patients working in transportation, manufacturing and agriculture were subsumed under blue-collar workers. Patients working in the service sector were counted as white-collar workers. For other work sectors including housekeeping and trainees, their indicated work stress level (low/high) decided about the allocation to blue or white-collar worker.²³ Disease activity was assessed using the Disease Activity Score in 28 joints-C reactive protein (3-component formula with 28 joint count). Seropositivity was defined as positive rheumatoid factors and/or anti-citrullinated protein antibodies.

Study outcomes

The primary outcome was time to first resolution of joint swelling during a 2-year follow-up in patients receiving their first TNFi (part 1). As a secondary outcome, time to resolution of joint swelling was assessed in non-bio-naïve patients receiving ABA, IL-6Ri, JAKi, RTX or TNFi (part 2). In part 3, bio-naïve patients receiving ABA, IL-6Ri, JAKi, RTX or TNFi were evaluated for time to resolution of joint swelling. In addition, in parts 1 and 2, time to resolution of joint tenderness was evaluated. Joint swelling and tenderness were assessed clinically without additional imaging. The shoulder was excluded from the analyses due to previous inconsistent results in PsA and the greater clinical challenge to assess synovitis of this joint,²⁴ resulting in a modified 28-joint count (m28-joint count). Only patients with at least one joint assessment during follow-up who remained on the same medication until joint swelling or tenderness resolved were included in the time-to-resolution analyses, resembling a per-protocol approach.

Statistical analysis

Baseline characteristics are reported as mean with SD for normally distributed continuous variables, median with IQR for non-normally distributed continuous variables and percentages for categorical variables.

The time to resolution of joint swelling and tenderness was evaluated within a timeframe of 2 years, as schematically depicted in figure 1. For the analysis, we used mixed-effects Cox proportional hazard transformation models^{25 26} able to account for interval-censored data, including a patient-level random effect. An interval-censored approach was chosen because the exact timing of joint swelling and tenderness resolution was unknown and could only be assessed at the reported visits. The interval-censored model assumes that the time of resolution of joint swelling is during the period between the last visit when the joint was swollen and the subsequent visit when it had resolved. The wrist served as the reference joint in comparison with other joints of the m28-joint count. The wrist was selected as the reference joint because it was the most frequently affected joint and to ensure consistency with previous analyses in PsA.⁸ A higher hazard compared with the wrist (HR>1) indicates faster resolution in the comparator joint, whereas a lower hazard (HR<1) suggests slower resolution.

Figure 1.

Time to resolution analysis illustrated using two hypothetical patient cases. Illustration created with BioRender (https://biorender.com). Two example patients illustrate the follow-up visit structure used in the time-to-resolution analysis. The exact time point at which joint swelling resolves between visits is unknown; hence, the data are interval-censored. For example, in patient 1, the right wrist resolves at some time point between baseline and second visit. For the primary analysis of time to resolution of joint swelling, both patient 1 and patient 2 are included in the model comparing the elbow and knee to the wrist. For the additional pairwise comparison analysis, both patient 1 and patient 2 are included in the comparison of the elbow to the wrist, while only patient 2 is included in the comparison of the knee to the wrist, since only in this patient both comparator joints are affected. TNFi, tumour necrosis factor inhibitors.

The models were adjusted for age and sex in the analyses in part 1 of the study, additionally for the number of previously used b/tsDMARDs in the analysis in part 2 and for MTX co-therapy and seropositivity in part 3 of the study. To compare joint-specific treatment responses across different modes of action (part 3 of the study), we included an interaction term between treatment type and joint site in the model. The different b/tsDMARDs were compared with TNFi as the reference treatment. Only the wrist and MCP2/3 joints were included in this analysis to reduce the number of comparisons. These joints were chosen because they were the most frequently affected ones.

To evaluate the effect of body side (left vs right), which may arise from asymmetric mechanical loading (eg, more frequent use of the dominant hand, typically the right), body side was included as a covariate in a sensitivity analysis. The following covariates were added in sensitivity analyses to adjust for their possible influence on location-dependent treatment response: body mass index (BMI), co-medication with MTX and job type (blue-collar work vs white-collar work). BMI was included as a confounder because it may influence the resolution of joint swelling in weight-bearing joints. Job type was also considered, as differences in mechanical demands between blue-collar and white-collar work could affect time to resolution in mechanically stressed joints. Similarly, body side was included in the models to account for potential effects related to right-handedness. An additional analysis (pairwise comparison) was restricted to patients in whom both comparator joints were affected, with separate analyses performed for each joint comparison (figure 1). Because both joints were swollen in the same patient, this approach reduces the likelihood that observed differences in time to resolution of joint swelling are attributable to patient-level characteristics.

All analyses were performed on patients with available data (complete case analysis) using R (V.4.5.2.)²⁷

Results

Treatment with first TNFi

A total of 2071 patients initiating TNFi after inclusion into SCQM had at least one swollen joint at treatment start. Out of these patients, 598 initiated their first TNFi. Patients were predominantly female (71%), 71% were seropositive, and the mean (SD) age was 55.8 (±13.2) years. Many patients had a co-therapy with MTX (72%) (table 1). Out of these 598 patients, 254 patients remained under TNFi treatment until their joint swelling was resolved (online supplemental table S2).

Table 1.

Baseline characteristics at initiation of tumour necrosis factor inhibitor in bio-naïve patients with at least one swollen joint from the m28-JC

Parameter	N=598	TNFi (bio-naïve)
Sex, female N (%)	598	425 (71.1)
Age, years	598	55.8 (13.2)
BMI, kg/m2 (median (IQR))	551	25.7 (22.7, 29.2)
Symptom duration, years (median (IQR))	577	3.5 (1.5, 10.4)
Seropositivity*, N (%)	564	400 (70.9)
CRP, mg/L (median (IQR))	555	8.0 (3.4, 15.0)
DAS-28 CRP	551	4.3 (1.1)
Smoker (current/former), N (%)	274	125 (45.6)
Job type, blue-collar N (%)	524	144 (27.5)
Comedication with MTX†, N (%)	598	429 (71.7)
Swollen joints (out of m28-JC) (median (IQR))	598	6.0 (3.0, 9.8)
Tender joints (out of m28-JC) (median (IQR))	540	6.0 (3.0, 11.0)
Follow-up time in months (median (IQR))	598	15.9 (12.4, 21.6)
Follow-up visits, N (median (IQR))	598	3.0 (2.0, 4.0)

Presented statistics are means with SDs, except where otherwise indicated.

*Rheumatoid factors and/or anticitrullinated protein antibody.

†Treatment with other csDMARDs was excluded (see the Methods section).

BMI, body mass index; CRP, C reactive protein; csDMARDs, conventional synthetic disease-modifying antirheumatic drugs; DAS-28, Disease Activity Score in 28 joints; JC, joint count; m28-JC, modified 28-joint count (shoulder joints excluded); MTX, methotrexate; TNFi, tumour necrosis factor inhibitor.

The most frequently swollen joint at baseline in patients initiating their first TNFi was the wrist (62%), MCP2 (63%), MCP3 (63%) and PIP3 (42%) joints (figure 2).

Figure 2.

Distribution of swollen joints at baseline in bio-naïve patients starting tumour necrosis factor inhibitor treatment. The values above the plots indicate the number of patients with affected joints among the 598 included patients. IP, interphalangeal joint; MCP, metacarpophalangeal joint; PIP, proximal interphalangeal joint.

Most patients had swollen joints on both body sides (left and right), while the right body side was more often swollen compared with the left body side (online supplemental figure S1). Most joints resolved during the first 6 to 12 months of follow-up (online supplemental figure S2). The most often recurring joints were the wrist, MCP2/3, PIP2/3 and the knee with an overall low recurrence rate between 8% and 11% (online supplemental figure S3).

Using the wrist as reference, the hazard of resolution of joint swelling was significantly higher in all other joints of the m28-joint count, except for MCP2 (HR=0.85, 95% CI 0.66 to 1.11) and MCP3 (HR=1.06, 95% CI 0.82 to 1.37) (figure 3). Both MCP and PIP joints of the first, fourth and fifth showed a tendency for higher rates of synovitis resolution than the corresponding joints of digits 2 and 3. Within the PIP joints, PIP3 showed the lowest rate of joint swelling resolution compared with the wrist joint. Resolution at the knee occurred at a rate only slightly higher than that of the wrist (HR=1.46, 95% CI 1.02 to 2.10). The exact number of joints resolved during follow-up can be found in the online supplemental table S3.

Figure 3.

HRs of resolution of joint swelling (HR and 95% CI) after start of tumour necrosis factor inhibitor in bio-naïve patients estimated using mixed-effects Cox proportional hazard models accounting for interval-censored data and adjusted for age and sex. Numbers in parentheses indicate the number of swollen joints and the number of contributing patients for each joint site. IP, interphalangeal joint; MCP, metacarpophalangeal joint; NJ, overall number of swollen joints; NP, overall number of patients; PIP, proximal interphalangeal joint.

An analysis in the whole cohort with available follow-up visits, including those that discontinued their treatment (resembling an intention-to-treat approach), revealed a similar joint response pattern (online supplemental figure S4).

A similar pattern was found when time to resolution of joint tenderness was chosen as an outcome (online supplemental figure S5).

After additional adjustment for BMI, MTX co-therapy and job-type the response pattern remained similar (online supplemental figure S6). In a further model adjusted for age, sex and body side (right vs left), the joints of the right body side were associated with a slower rate of resolution (HR=0.77, 95% CI 0.66 to 0.89) (figure 4).

Figure 4.

HR of resolution of joint swelling (HR and 95% CI) after start of tumour necrosis factor inhibitor in bio-naïve patients estimated using interval-censored mixed-effects Cox proportional hazard models adjusted for age, sex and body side (right, left). Numbers in parentheses indicate the number of swollen joints and the number of contributing patients for each joint site. IP, interphalangeal joint; MCP, metacarpophalangeal joint; NJ, overall number of swollen joints; NP, overall number of patients; PIP, proximal interphalangeal joint.

Pairwise comparisons between two joint locations, conducted among patients with both joints affected, yielded comparable results (online supplemental figure S7 and table S4). The wider CIs reflected the smaller number of patients included in these analyses.

Treatment with ABA, IL-6Ri, JAKi, RTX and TNFi

To investigate whether the documented pattern of response based on joint localisation was specific to TNFi, we repeated the analysis on treatment with other b/tsDMARDs (part 2 of analysis). While allowing previous b/tsDMARD treatment, we restricted the analysis to the first agent of the respective therapeutic modality used after inclusion of the patient into SCQM. A total of 1942 of these patients had a swollen joint at treatment initiation (ABA: 242, IL-6Ri: 317, JAKi: 333, RTX: 176, TNFi: 874). The baseline characteristics in the different groups were comparable with a high percentage of female patients (74%–81%), mostly seropositive (72%–85%) and a mean age between 56 and 62 years (table 2). Patients with TNFi were more often bio-naïve and had a shorter symptom duration compared with the other treatment approaches (table 2). Out of these 1942 patients, 990 patients (ABA: 133, IL-6Ri: 163, JAKi: 183, RTX: 115, TNFi: 396) remained under their respective treatment until their joint swelling resolved (online supplemental table S5).

Table 2.

Baseline characteristics at initiation of biologics in not exclusively bio-naïve patients with at least one swollen joint from the m28-JC

	ABA N=242		IL-6Ri N=317		JAKi N=333		RTX N=176		TNFi N=874
Sex, female N (%)	242	185 (76.4)	317	236 (74.4)	333	269 (80.8)	176	136 (77.3)	874	643 (73.6)
Age, years	242	62.0 (12.3)	317	57.3 (13.3)	333	58.5 (13.5)	176	59 (12.1)	874	55.7 (13.3)
BMI, kg/m2 (median (IQR))	215	25.7 (22.1, 29.6)	284	25.5 (22.9, 29.4)	280	25.4 (22.5, 29.4)	160	25.7 (22.4, 29.3)	792	25.5 (22.5, 29.1)
Symptom duration, years (median (IQR))	233	9.10 (3.3, 17.9)	314	7.8 (3.7, 15.9)	328	10.1 (4.8, 18.6)	169	10.7 (5.8, 19.7)	850	5.5 (2.0, 12.9)
Seropositivity,* N (%)	236	187 (79.2)	310	258 (83.2)	332	253 (76.2)	171	146 (85.4)	833	597 (71.7)
CRP, mg/L (median (IQR))	226	6.0 (2.7, 12.0)	292	8.0 (3.3, 21.0)	320	5.0 (2.0, 10.0)	153	8.0 (3.0, 20.0)	805	8.0 (3.0, 15.0)
DAS-28 CRP	223	4.1 (1.1)	289	4.3 (1.1)	318	3.9 (1.1)	153	4.2 (1.2)	797	4.2 (1.1)
Smoker, current or former (%)	118	47 (39.8)	152	67 (44.1)	160	82 (51.2)	93	36 (38.7)	415	199 (48.0)
Job, blue-collar N (%)	201	48 (23.9)	287	84 (29.3)	272	70 (25.7)	144	37 (25.7)	755	199 (26.4)
b/tsDMARD treatment line	242		317		333		176		874
1st line (%)		82 (33.9)		88 (27.8)		90 (27.0)		28 (15.9)		598 (68.4)
2nd line (%)		70 (28.9)		86 (27.1)		69 (20.7)		66 (37.5)		153 (17.5)
>3rd line (%)		90 (37.2)		143 (45.1)		174 (52.3)		82 (46.6)		123 (14.1)
MTX co-therapy,† N (%)	241	144 (59.8)	317	147 (46.4)	333	126 (37.8)	176	93 (52.8)	874	580 (66.4)
Swollen joints (out of m28-JC) (median (IQR))	242	5.0 (3.0, 10.0)	317	6.0 (3.0, 9.0)	333	5.0 (3.0, 8.0)	176	5.0 (2.0, 9.3)	874	5.0 (3.0, 9.0)
Tender joints (out of m28-JC) (median (IQR))	223	5.00 (2.0, 10.0)	282	6.00 (3.0, 10.0)	301	5.0 (2.0, 9.0)	160	6.0 (2.0, 12.0)	785	6.0 (3.0,11.0)
Follow-up time in months (median (IQR))	242	16.8 (11.6, 22.9)	317	18.0 (12.0, 22.8)	333	17.3 (12.4, 22.5)	176	17.9 (11.4, 23.0)	874	18.3 (12.5, 23.1)
Follow-up visits, N (median (IQR))	242	3.0 (2.0, 5.0)	317	4.0 (2.0, 6.0)	333	4.0 (2.0, 6.0)	176	4.0 (2.0, 6.0)	874	3.0 (2.0, 4.0)

Presented statistics are means (SDs), except where otherwise indicated with median (IQR).

*Rheumatoid factors and/or anticitrullinated protein antibody.

†Treatment with other csDMARDs was excluded (see the Methods section).

ABA, abatacept; BMI, body mass index; b/tsDMARDs, biologic or targeted-synthetic disease-modifying anti-rheumatic drug; CRP, C reactive protein; csDMARDs, conventional synthetic DMARDs; DAS-28, Disease Activity Score in 28 joints; IL-6Ri, interleukin-6 receptor inhibitor; JAKi, Janus kinase inhibitor; JC, joint count; m28-JC, modified 28-joint count; MTX, methotrexate; RTX, rituximab; TNFi, tumour necrosis factor inhibitor.

The distribution of swollen joints at start of treatment was comparable between the individual therapeutic modalities (online supplemental figure S8).

Rates of resolution of synovitis at different joint locations are shown for each treatment modality (ABA, IL-6Ri, JAKi, RTX, TNFi) in figure 5. A response pattern similar to that observed in TNFi-treated patients was also evident across the other treatment modalities, with higher rates of joint swelling resolution in most joints compared with the wrist, except for the MCP2 and MCP3 joints, as well as the knee (online supplemental table S6 for exact resolution rates). The response pattern in the hand, characterised by a tendency for higher rates of resolution in the first, fourth and fifth digits, was also consistently observed. The number of joints resolved during follow-up can be found in online supplemental table S7.

Figure 5.

HR of resolution of joint swelling (HR and 95% CI) after start of ABA, IL-6 receptor inhibitor, JAKi, RTX and TNFi estimated using interval-censored mixed-effects Cox proportional hazard models adjusted for age, sex and previous treatment. ABA, abatacept; IL-6Ri, interleukin-6 receptor inhibitor; IP, interphalangeal joint; JAKi, Janus kinase inhibitor; MCP, metacarpophalangeal joint; PIP, proximal interphalangeal joint; RTX, rituximab; TNFi, tumour necrosis factor inhibitor.

The resolution of joint tenderness showed a similar pattern (online supplemental figure S9 and table S8).

Following additional adjustments for BMI, MTX co-therapy and job-type, the response pattern remained unchanged (online supplemental figure S10 and table S9). For the pairwise comparison of swollen joints (online supplemental figure S11 and table S10), the response remained similar.

Treatment with bio-naïve ABA, IL-6Ri, JAKi, RTX and TNFi

To directly compare site-specific treatment effects across drugs with different mechanisms of action, we included an interaction term between treatment type and joint location and adjusted for age, sex, seropositivity and concomitant MTX use. TNFi served as the reference treatment. The analyses were performed in bio-naïve patients treated with the respective treatment modality as a first b/tsDMARD, during a time period where all the treatment modalities were available. Since there were almost no bio-naïve RTX patients, this medication was excluded. The baseline characteristics of the patients included in this analysis, which focused on the wrist and MCP2 and 3 joints are described in the online supplemental table S11. The results are shown in figure 6. As the CIs for the resolution rates across treatment modalities at each joint location overlap, no statistically significant differences in time to synovitis resolution could be detected between the various treatments.

Figure 6.

HR of resolution of joint swelling (HR and 95% CI) after start of abatacept, IL-6 receptor inhibitor, JAKi and TNFi estimated using interval-censored mixed-effects Cox proportional hazard models with an interaction of treatment with joint site, adjusted for age, sex, seropositivity and methotrexate co-therapy. TNFi is the reference treatment. b/tsDMARDs, biologic/targeted synthetic disease-modifying antirheumatic drug; IL-6Ri, interleukin-6 receptor inhibitor; JAKi, Janus kinase inhibitor; MCP, metacarpophalangeal joint; TNFi, tumour necrosis factor inhibitor.

Discussion

Our observational analysis of a large national cohort of patients with RA initiating advanced targeted therapies under real-life conditions highlights several findings. First, location of the affected joint matters in RA. Involvement of the wrist, MCP2 and MCP3 joints and potentially the knee was associated with the lowest likelihood of achieving a clinical response. Second, the pattern of joint-specificity observed in RA closely resembled that previously demonstrated for TNFi and secukinumab in PsA, suggesting that joint-intrinsic factors may influence joint-specific therapeutic outcomes more strongly than the overarching pathogenic pathways of different rheumatic diseases. Although components of the JAK-signal transducer and activator of transcription pathway have been shown to be differentially expressed across synovial tissues from different articular sites,^{18 19} targeting this pathway with JAKi and IL-6Ri did not result in a statistically significant distinct pattern of response compared with other therapeutic mechanisms.

The methodology used in this study to estimate the hazard (or rate) of synovitis resolution at specific articular sites identifies joints that, when affected, tend to respond more slowly. Given that the majority of individuals are right-handed, the slower resolution of right-sided joints suggests that differences in mechanical strain are a major driver of joint-specific treatment responses. The wrist, which contributes to three different arches that stabilise the hand, is among the joints exposed to the highest mechanical stress.¹² Likewise, the second and third fingers—functionally the strongest fingers of the hand—experience substantial strain.²⁸ Notably, these joints were also among the most frequently affected at baseline. Mechanical strain has been shown to influence signalling pathways in synovial fibroblasts, in part through regulation of the expression of Cadherin-1.¹⁰ In addition, a cohort study demonstrated that mechanical stress in the context of subclinical joint inflammation was associated with the subsequent development of RA, further underscoring the pathogenic relevance of biomechanical load.¹¹ These same joints were also the most frequently affected by recurrent synovitis during follow-up.²⁹ While new/recurrent synovitis occurred in our study <12% of joints not involved at baseline, the wrist and the joints of the second and third fingers were most frequently affected by recurrent synovitis.

Beyond differences in mechanical strain, the specific pattern observed in the hand with a slower resolution in the middle fingers, and a faster resolution in the outer fingers, may be related to the spatial expression of developmental genes such as homeobox genes. These genes are differentially expressed along the body axes including the posterior–anterior axis extending from the ulnar to the radial side of the hand.¹⁴

The results regarding the response at the level of the knee, the only joint of the lower extremities included in the 28-joint count, were inconsistent across the different treatment modalities. We speculate that intra-articular steroid injections, although infrequently administered during follow-up, may have disproportionately involved this joint. However, because the SCQM registry documents only the number of intra-articular injections per year and not the specific joints injected, we were unable to substantiate this hypothesis.

The clinical implications of our findings are manifold. The observation that the wrist and the MCP2 and MCP3 joints require longer to respond suggests that treatment response should be assessed at a later time point when these joints are involved. In contrast, the other joints, which showed HRs of approximately 2–3 relative to the wrist, appeared to resolve at a substantially faster rate. However, the wide CIs indicate considerable uncertainty regarding the precise magnitude of this effect. The observed response pattern aligns with current RA treatment recommendations, which stipulate that clinical improvement should be evident by 3 months and the treatment target achieved by 6 months.⁶ In real-life practice, however, treatment switches may occur earlier, particularly in countries with higher national socioeconomic indicators.³⁰ While bridging strategies are often used when initiating conventional synthetic DMARDs—typically with oral glucocorticoids or intra-articular steroid injections—such approaches are less frequently applied when starting b/tsDMARDs, given their generally faster onset of action.³¹ Our data suggest that intra-articular steroid injections could also be considered early during treatment with b/tsDMARDs when the wrist and MCP2 and MCP3 joints are affected, as these joints may require more time to respond to systemic therapy.

Strengths of our investigation include the prospective study design, the standardised and regular clinical assessments, and the application of statistical methods that account for both the uncertainty of the exact time point at which synovitis resolves between visits and the nested data structure of multiple joints per patient. A previous preliminary analysis that had suggested potential joint-specific responses to different modes of action had important methodological drawbacks⁹ including a purely descriptive study design that lacked a robust framework for evaluating the statistical significance of observed differences and the use of a combined tender and swollen joint score that disproportionately emphasised the joints most frequently affected.

Our study has several limitations. First, ultrasonographic or MRI data to confirm the presence of synovitis were not available in our dataset. Second, restricting the analysis to patients who remained on their initial treatment may have introduced selection bias by excluding non-responders and patients who discontinued therapy because of adverse events. However, a sensitivity analysis that included all patients with available follow-up visits, regardless of treatment persistence, yielded similar results. Third, patients with slower resolution of joint swelling may have been more likely to be lost to follow-up, potentially resulting in informative censoring bias. Such bias could lead to an underestimation of the time to resolution. Fourthly, although we sought to minimise indication bias across treatment groups, residual confounding cannot be fully excluded. By restricting comparisons to bio-naïve patients and adjusting for potential confounders influencing treatment choice, such as seropositivity, we were able to reduce this bias. Finally, joints of the feet—which represent an important site of disease involvement in RA³²—are not included in the 28-joint count. Analyses based on the 44-joint count or higher were not feasible, as only the 28-joint count is routinely reported in the SCQM registry. The findings presented here could be further strengthened by the inclusion of radiographic outcomes, which were not available in sufficient quantity in the current study population. Radiographic data in future studies may help to distinguish erosive damage resulting from persistent inflammation in specific joints from other structural changes, such as osteoarthritis related to mechanical overuse. Such changes may cause joint symptoms and could therefore confound the interpretation of clinical joint swelling outcomes in the present analysis.

In conclusion, the wrist, MCP2 and MCP3 joints respond more slowly than other joints following initiation of a b/tsDMARD, indicating that overall treatment response should be evaluated at a later time point when these joints are involved. In addition, bridging intra-articular steroid injections targeting these specific joints may be considered at treatment initiation of a b/tsDMARD to help manage persistent local inflammation.

Data availability statement

Data may be obtained from a third party and are not publicly available. This study is based on data collected by the SCQM Foundation (Swiss Clinical Quality Management in Rheumatic Diseases), which operates a national registry for inflammatory rheumatic diseases. Access to the data is subject to restrictions and requires approval from the SCQM Foundation in accordance with the SCQM Rules of Research and Collaboration (https://www.scqm.ch/en/research/research-with-scqm-data/). Interested parties may contact the SCQM Foundation (research@scqm.ch) to request access to the data for research purposes.

Ethics statements

Patient consent for publication

Not applicable.

Ethics approval

This study involves human participants and was approved by the Ethics Committee of the Canton of Zurich. Participants gave informed consent to participate in the study before taking part.