Rheumatoid arthritis
RA patients in the initial cohort (n=541)
Excluded (n=388): 1. Lost to follow-up before Dec 2018ͧ n=49ͨ 2. Did not achieve DAS28- CRP remission before Dec 2018ͧ n=339ͨ Excluded (n=8): having fewer than two DAS28-CRP assessments per year
DAS28-CRP remission (n=153)
Patients included in the final analysis (n=145)
Clinical Deep Remission(n=32)
non-Clinical Deep Remission(n=113)
Sustained remission(n=20)
Sustained remission(n=43)
Relapse(n=12)
Relapse(n=70)
Figure 1 Flow diagram of the selection of the participants. DAS28-CRP, Disease Activity Score in 28 joints using C reactive protein; RA, rheumatoid arthritis.
for data with non-normal distribution; for categorical vari- ables, the χ 2 test was performed, and Fisher’s exact test was employed when cell counts were less than 5. Survival analyses included Kaplan-Meier curves to esti- mate cumulative remission rates, with between-group differences assessed via log-rank test. Cox proportional hazards regression models were employed to assess the impact of various factors on disease relapse. We conducted univariate Cox regression analyses with the following candidate variables: group, tapering status, sex, age, disease duration, smoking status, family history of RA, RF titre, anti-CCP titre, comorbidities and extra- articular manifestations. In the univariable analysis, each candidate variable was examined individually. Variables with a p value <0.1 in the univariable analysis were consid- ered for inclusion in the multivariable model. Because drug tapering occurred at different time points during follow-up, it was modelled as a time-dependent covariate. Specifically, a time-dependent variable, tapering status , was defined as I (t ≥ time of tapering) and included in the Cox model to account for its dynamic effect on relapse risk. This variable took the value 0 before tapering initiation and 1 from the exact time of tapering onward.
Patients who never tapered were assigned a tapering time exceeding their last follow-up. The multivariable Cox model was constructed to eval- uate the following prespecified predictors: group effect (CliDR vs non-CliDR) during periods without tapering, the effect of tapering within the non-CliDR group and their interaction term (CliDR×tapering status), along with family history of rheumatic disease (yes vs no), RF titre (continuous, per 1-unit increase) and comorbidities (yes vs no). The interaction term was included to formally test whether the effect of tapering on relapse risk differed between the CliDR and non-CliDR groups. For all categorical variables, the reference category was the absence of the characteristic (eg, no, absent). Contin- uous variables (RF titre) were modelled as linear terms, with HRs corresponding to a 1-unit increase. Results are presented as HRs with 95% CIs. The propor- tional hazards assumption was assessed using tests based on Schoenfeld residuals. For any variable violating this assumption, additional analyses incorporating a time– covariate interaction term or using stratified Cox models were performed.
3
Li H, et al . RMD Open 2026; 12 :e006387. doi:10.1136/rmdopen-2025-006387
Powered by FlippingBook