For the descriptive comparison, we identified a total of 5974 RA without concomitant SjD and 337 overlap patients. Figure 1 displays the flow chart of eligible patients and TC after each inclusion step.

Flow chart of eligible patients and treatment courses after each inclusion step. 1Patients with sicca symptoms but without Sjögren’s disease diagnosis were excluded (n = 2417). b/tsDMARD, biologic/ targeted synthetic disease-modifying anti-rheumatic drug; FU, follow-up; JAKi, Janus kinase-inhibitor; OMA, other modes of action; SCQM, Swiss Clinical Quality Management in Rheumatic Diseases; SjD, Sjögren’s disease; TC, treatment course; TNFi, Tumour necrosis factor-inhibitor

Patients with concomitant SjD were more likely to be female, non-smokers, to be rheumatoid factor (RF) positive, to have anti-cyclic citrullinated peptide antibodies, a longer RA disease duration, higher disease activity scores (DAS28, CDAI [5]), worse patient reported outcomes (HAQ or RA Disease Activity Index-5 (RADAI-5) score [19]), more signs of synovitis and a higher erosive burden (Ratingen score [20]) at the time of SCQM inclusion (Table 1). Ultrasound (Swiss Sonography in Arthritis and Rheumatism (SONAR) score [21]) revealed a higher power doppler (PD) but similar grey scale (GS) ultrasound scores in patients with overlap disease. Regarding prior treatment, these patients were more likely to have received previous glucocorticoid, rituximab, or abatacept treatment.

For the assessment of treatment effectiveness, we identified a total of 6781 (3788 TNFi, 2225 OMA, 768 JAKi) eligible TCs from 234 overlap patients and 3287 patients with RA alone. Out of the 6781 TC, 2967 TC were stopped, 1081 TC were stopped because another b/tsDMARD was started before the recorded stop date, and 2733 TC were censored at end of follow up. Stop reasons are documented in 2807 of the 2967 TC (no difference between RA and RA/SjD) and include: ineffectiveness (n = 2807), adverse events (n = 638), remission (n = 275), and other reasons, such as patient preference or pregnancy (n = 660). In 265/553 TC of RA/SjD patients, the SjD diagnosis was subsequent to the initiation of the respective TC. Baseline characteristics at TC-start are displayed in online supplementary Table 2.

Overall, RA/SjD patients displayed shorter drug retention rates as compared to RA patients without concomitant SjD (online supplementary Fig. 4). Median retention time for patients with overlap disease versus without SjD was 518 (95% confidence interval (CI) 391–721) versus 777 (728–866) days for TNFi, 619 (525–979) versus 894 (796–1016) days for OMA and 817 (436–1570) versus 812 (684–949) days for JAKi (Fig. 2). In the unadjusted Cox model (number of events = 4048), there was evidence that overlap patients had a higher hazard for stopping TNFi treatment than patients without SjD (crude HR 1.24 [95% CI 1.06–1.44]). A trend was found for OMA (crude HR 1.16 [0.96–1.39]) and no difference was found for JAKi (crude HR 1.06 [0.71–1.57]). After adjustment (number of events = 2917), the hazard for stopping TNFi treatment was still higher in the overlap group (HR 1.30 [95% CI 1.07–1.60]). No difference was observed for OMA (HR 1.12 [0.91–1.37]) and JAKi (HR 0.97 [0.62–1.53]). Within the overlap group (number of events = 382), there was no evidence for a higher hazard for stopping JAKi compared to TNFi (crude HR 0.88 [95% CI 0.58–1.31]) in the unadjusted Cox model (patient characteristics at TC-start and Kaplan-Meier plot of retention times are displayed in the online supplementary Fig. 1 and online supplementary Table 3). We found a trend for OMA compared to TNFi (crude HR 0.82 [0.66–1.01]), which was not significant. After adjustment (number of events = 260), the hazard for stopping treatment was lower for OMA (HR 0.62 [95% CI 0.46–0.84]) and JAKi (HR 0.52 [0.28–0.96]) when compared to TNFi.

Kaplan-Meier plot of retention times for eligible TNFi, OMA, and JAKi treatment courses. JAKi, Janus kinase-inhibitor; OMA, other modes of action; TC, treatment course; TNFi, Tumour necrosis factor-inhibitor

The drug retention rates of overlap patients did not differ between etanercept as compared to the other TNFi and IL-6Ri as compared to rituximab and abatacept (Fig. 3). Median retention time was 405 (95% CI 301–919) days for etanercept and 521 (95% CI 410–902) days for the other TNFi (crude HR 1.036 [95% CI 0.76–1.41]) and 659 (95% CI 423–1197) days for IL-6Ri and 619 (95% CI 518–1166) days for rituximab and abatacept (crude HR 1.156 [0.81–1.65]), respectively. Drug retention of overlap patients differed between rituximab as compared to TNFi (online supplementary Fig. 5). We observed no differences in drug retention between overlap patients as compared to patients without SjD in model 2 including adjustment for DAS28-CRP levels (TNFi: 1.14 [95% CI 0.81–1.60], OMA: 1.10 [0.81–1.50], JAKi: 1.22 [0.67–2.22]) and in model 3 including adjustment for HAQ levels (TNFi: 1.24 [0.93–1.65], OMA: 0.97 [0.69–1.35], JAKi: 0.83 [0.46–1.49]).

Kaplan-Meier plot of retention times for patients with rheumatoid arthritis and concomitant Sjögren’s disease. A, Eligible abatacept, rituximab, and IL-6Ri treatment courses. B, Eligible etanercept and other tumour necrosis factor-inhibitor treatment courses. IL-6i, Interleukin-6 receptor-inhibitor

For the secondary outcome, we assessed response at one year. The subset of eligible TCs included in the PP and RTR response analysis is displayed in the online supplementary Fig. 2.

At one year, overlap patients receiving TNFi, OMA, or JAKi were less likely to reach DAS28 remission as compared to patients without concomitant SjD in the PP analysis (unadjusted model). After adjustment, overlap patients had inferior remission rates with JAKi and OMA, but not with TNFi treatment in the PP dataset (online supplementary Fig. 3 and online supplementary Table 4). The RTR analysis revealed that overlap patients were less likely than patients without associated SjD to reach DAS28 remission with TNFi (crude HR 2.12 [95% CI 1.35–3.34]), OMA (HR 1.71 [1.15–2.55]), or JAKi (HR 2.81 [1.16–6.83], Fig. 4). This was also true after adjustment (TNFi HR 2.0 [95% CI 1.22–3.28]; OMA HR 1.77 [1.17–2.7]; JAKi HR 3.49 [1.18–10.26]). Online supplementary Fig. 6 displays individual DAS28-CRP scores over time.

DAS28-CRP score at baseline and 1 year follow-up visit (response tolerance remission analysis). The distributions of DAS28-CRP values per group and timepoint are displayed as violin plots (density from lowest to highest value, median indicated by a horizontal line). A, Comparison of DAS28-CRP values between rheumatoid arthritis patients with and without concomitant Sjögren’s disease. B, Comparison of DAS28-CRP values between different treatment modalities within the overlap group. Grey: in remission (DAS28 < 2.6). BL, baseline; CRP, C-reactive protein; DAS28, Disease Activity Score-28; FU, follow-up; JAKi, Janus kinase-inhibitor; OMA, other modes of action; TNFi, Tumour necrosis factor-inhibitor

Within overlap patients, no difference was observed in reaching DAS28 remission for JAKi or OMA as compared to TNFi (PP and RTR analysis, unadjusted and adjusted model, Fig. 4, online supplementary Fig. 7 and online supplementary Table 5).