P282 ANA seroconversion during prior anti-TNF therapy abolishes anti-CCP antibody positivity as a predictor of abatacept retention in rheumatoid arthritis

Abstract

Abstract Background/Aims Few long-term drug survival analyses have been reported for patients with rheumatoid arthritis (RA) treated with abatacept. Here we report a machine learning approach to investigate drug-survival of abatacept over 5 years in RA patients. Methods We performed a retrospective observational study on a tertiary hospital dataset of RA patients who started abatacept between January 2008 and December 2020. Time to abatacept discontinuation over 5 years was estimated using Kaplan-Meier survival analyses. A multivariate cox-proportional hazard model to predict abatacept discontinuation was chosen by random survival forest and partial least regression. Results A total of 112 RA patients (81% female, mean age 58.1 [SD 13.5] years) received abatacept. Mean disease duration was 18.3 months (SD 13.6). More than half (65/112) were co-prescribed at least one conventional synthetic DMARD (csDMARD). Methotrexate was the most frequently concomitant csDMARD (n = 37), followed by hydroxychloroquine (n = 23), sulfasalazine (n = 15), and leflunomide (n = 7). 42 (37.5%) patients were treated with glucocorticoids (intermittent or continuous) with abatacept. Abatacept was mostly used as 4th (n = 29) and 3rd line (n = 24) bDMARD, but 19 patients received abatacept as their first line bDMARD. 75 (67%) patients were rheumatoid factor (RF) positive and 73 (65.2%) were anti-citrullinated protein antibody (ACPA) positive. Anti-nuclear antibody (ANA) was positive (≥1:80) in 25 patients (pre-biologics) and 18 seroconverted to previously exposed anti-TNF (ANA seroconversion group). Abatacept was discontinued in 54 patients (48.2%); 19 (35.2%) due to an adverse event and 35 (64.8%) due to loss of efficacy. Overall, the median time to discontinuation of abatacept was 3.8 years. Multivariate cox-proportional hazard model revealed that ACPA positivity was associated with reduced risk of abatacept discontinuation with HR of 0.56 (95% CI 0.32-0.95,p=0.03) compared to the ACPA-negative group. In contrast, the ANA seroconversion group demonstrated worse retention of abatacept with HR of 2.67 (95% CI 1.39-5.16, p = 0.0033) compared to ANA-negative patients. Kaplan-Meier analysis demonstrated that in the ANA seroconversion group, retention of abatacept was significantly inferior to pre-anti-TNF ANA positive patients (p = 0.0084) and ANA-negative patients (p = 0.0041). ACPA positivity was associated with better survival only in the non-ANA seroconverted group with HR of 0.54 (95% CI 0.29-0.92, p = 0.041), after adjusted by propensity score. Combining abatacept with any csDMARDs reduced the risk of abatacept discontinuation with HR of 0.57 (95% CI 0.33-0.98, p = 0.043) after adjusting by propensity score. No statistical difference was found between first-line, second-line, or any subsequent chronology of use of abatacept. Conclusion Our data suggest that previous ANA seroconversion occurring whilst on anti-TNF therapy reduces subsequent abatacept retention and cancels out the protective effect of anti-CCP positivity. The mechanism through which this occurs is unclear and further prospective and mechanistic studies are needed to validate these findings. Disclosure J. Kimpton: None. M. Shipa: Grants/research support; MS is funded by Versus Arthritis. S. Yeoh: Grants/research support; SY is funded by the Royal College of Physicians, Rosetrees Trust, NIHR University College London Hospitals Biomedical Research Centre, UCLH Charities, and Versus Arthritis. E. Hawkins: None. M. Ehrenstein: Grants/research support; ME is supported (in part) by the University College London Hospital Biomedical Research Centre.