Background Although the presence of rheumatoid factor (RF) may be a risk factor for the onset and progression of rheumatoid arthritis (RA), sufficient literature does not exist to support the clinical relationship between RF positivity and the effects of treatment with biologic disease-modifying antirheumatic drugs (bDMARDS). This multicenter study aimed to explore the association of RF positivity with the effects of bDMARDS treatment in bio-naïve RA patients using a linear mixed-effect model.
Objectives In a multicenter study, patients are clustered within institutions, therefore results of adjustment models are likely to be biased by random, unobserved between-institution differences. Such bias could lead to inaccurate prediction and interpretation of outcomes. We used a linear mixed-effect model including between-institution variation as a random effect, which would improve the performance of this multicenter study.
Methods In total, 625 bio-naïve RA patients registered in the Tsurumai Biologics Communication Registry (TBCR), which comprises Nagoya University and 15 affiliated institutions in Japan, who received bDMARDS treatment during the study period (2006–2016) were eligible for inclusion. Demographic information and disease characteristics were assessed at baseline. DAS28 using erythrocyte sedimentation rate was recorded at baseline and following 24 weeks of therapy. In order to predict DAS28 improvement at 24 week, a linear mixed-effect model including between-institution variation as a random effect, controlling for RF positivity, age, sex, stage, methotrexate (MTX) use, prednisolone (PSL) use, tumor necrosis factor inhibitor (TNFi) or non-TNFi, and DAS28 at baseline, was developed.
Results Of the 625 patients, 513 showed RF positivity and 112 were antibody negative. Mean ± SD age at baseline was 56.9±14.0 years; 509 patients were women (81.4%). The mean ± SD DAS28 score at baseline was 5.19±1.24. Proportion of MTX and PSL use were 79.3% and 58.1%, respectively. Following adjustment for relevant covariates, RF positivity was associated with a decrease biologic treatment effect (β = −0.33±0.12, p<0.05). In another model including an additional interaction term of RF status and TNFi or non-TNFi, the influence of RF status on treatment effect was persistent (β = −0.26±0.14, p<0.1). These two models had comparable AIC. A model excluding RF positivity term had larger AIC than these two models, suggesting that RF positivity is crucial for predicting the effect of bDMARDS treatment.
Conclusions In our multicenter study using a linear mixed-effect model including between-institution variation as a random effect, RF positivity, in addition to some well-known variables, was found to be independently associated with decreased effects of bDMARDS treatment in bio-naïve RA patients.
Disclosure of Interest None declared