Self-perceived general health at start of TNFi therapy predicts therapeutic response in patients with rheumatoid arthritis: analysis from the ATTRA registry

Abstract

Background: Patient-reported outcomes (PROs) have been shown to predict various disease outcomes. One of the most widely used PRO instruments is the Short Form (SF) 36 questionnaire which evaluates the patient's health status. Our goal was to evaluate the association between therapeutic response and patients' self-perceived general health status at TNFi initiation based on answers to two selected questions (Qs) in the SF-36 questionnaire.

Methods: We included two separate datasets with RA patients (pts) initiating the first-line TNFi within the period 01/01/2001-31/12/2017 (primary dataset) and 01/01/2018-01/01/2020 (validation dataset) with at least one-year follow-up and filled SF-36 questionnaire at baseline. Patients were grouped according to their response ('definitely/mostly yes' vs 'definitely/mostly no') to Q11A and Q11C at baseline. The primary outcome was remission (REM) according to DAS28-ESR (<2.6) at the 12-month visit. REM rates were compared across patients' groups with Pearson's chi-squared test. Using logistic regression, crude and adjusted (to baseline DAS28-ESR and HAQ) odds ratios (ORs) were computed. Drug retentions were obtained through the Kaplan-Meier method. We repeated the analysis on propensity score-matched patients at baseline as a sensitivity analysis.

Results: Within the primary dataset (648/792 pts answering positively/negatively to Q11A; 730/580 pts answering positively/negatively to Q11C), patients answering 'yes' to Q11A/Q11C had 1.5/1.4 times higher odds for REM at 12-month visit than patients answering 'no'. The odds remained significantly different even after accounting for baseline DAS28-ESR and HAQ and within propensity score-matched datasets. Further, patients answering 'yes' to Q11A had a 1.3 times higher risk of TNFi discontinuation than patients answering 'no'. The validation dataset analysis (216/254 pts answering 'yes'/'no' to Q11A; 231/201 answering 'yes'/'no' to Q11C) gave similar results. Patients answering 'yes' to Q11A/Q11C had 1.7 times higher odds of reaching REM at the 12-month visit than patients responding 'no'. Even after accounting for baseline disease activity and functional status and within PS-matched datasets, the odds remained significantly higher. However, there was no statistically significant difference in drug retentions.

Conclusions: We provide strong evidence that self-perceived general health at TNFi initiation predicts reaching remission at 12 months in pts with RA.