Background Methotrexate (MTX) is the disease modifying drug of first choice in the treatment of rheumatoid arthritis (RA). However, response to MTX varies and the ability to predict likely non-response could enable earlier access to alternative drugs and the avoidance of disease progression for some patients.
Objectives To predict non-response to MTX in the Rheumatoid Arthritis Medication Study (RAMS), a cohort of RA patients commencing MTX therapy for the first time.
Methods RAMS is a national multi-centre observational study recruiting RA patients in the UK who are ≥18 years old and starting MTX for the first time. Data on potential predictors of response to MTX was acquired at baseline via questionnaires, case notes and blood samples and used to predict non-response to MTX at six months after commencement. Non-response to treatment was defined as failing to fulfil the EULAR criteria for good response: disease activity score-28 (DAS28) at six months ≤3.2 and reduction in DAS28 from baseline to six months >1.2. Potential predictors were chosen as candidate variables in a multivariable logistic regression model for non-response based on previous univariable analysis. The baseline measures included were body mass index (BMI), current smoking (yes/no), drinking of alcohol (yes/no), symptom duration, DAS28, functional disability (Health Assessment Questionnaire, HAQ), C-reactive protein (CRP), creatinine, anxiety and depression scores (Hospital Anxiety and Depression Scale, HADS) and beliefs about medicines necessity and concerns scores (Beliefs about Medicines Questionnaire, BMQ). All candidate predictors were initially included in a full model before backwards selection was used to remove non-significant terms (p>0.05) and produce a parsimonious final model. The ability of the models to discriminate between responders and non-responders was assessed using the area under the receiver operating characteristic curve (AUC).
Results Data on baseline predictor values and response at six months was available on 781 participants: 546 (70%) female, median age 59 (IQR 48–68) years, median symptom duration 9 (IQR 4–28) months. 549 participants (70%) were non-responders. Multivariable age- and sex-adjusted odds ratios for the predictors in the full and final models are shown in Table 1. The full model had an AUC of 0.70. The final model, containing BMI, current smoking, DAS28 and HADS anxiety score, had an AUC of 0.69.
Conclusions Baseline values of lifestyle, clinical and psychosocial predictors were used to predict non-response to MTX at six months. A model with a parsimonious set of core variables had similar discriminatory performance to a model with a full set of predictors. Of note is the role of participant anxiety on commencing treatment with MTX in predicting response, which may be related to subsequent adherence.
Disclosure of Interest None declared