Abatacept and non-melanoma skin cancer in patients with rheumatoid arthritis: a comprehensive evaluation of randomised controlled trials and observational studies
Teresa A. Simon, Lixian Dong, Samy Suissa, Kaleb Michaud, Sofia Pedro, Marc C. Hochberg, Maarten Boers, Johan Askling, Thomas Frisell, Anja Strangfeld, Yvette Meißner, Vadim Khaychuk, Alyssa Dominique, Michael A. Maldonado
Abstract
OBJECTIVES: This study aims to evaluate non-melanoma skin cancer (NMSC) risk associated with abatacept treatment for rheumatoid arthritis (RA). METHODS: This evaluation included 16 abatacept RA clinical trials and 6 observational studies. NMSC incidence rates (IRs)/1000 patient-years (p-y) of exposure were compared between patients treated with abatacept versus placebo, conventional synthetic (cs) disease-modifying antirheumatic drugs (DMARDs) and other biological/targeted synthetic (b/ts)DMARDs. For observational studies, a random-effects model was used to pool rate ratios (RRs). RESULTS: ~49 000 patients receiving abatacept were analysed from clinical trials (~7000) and observational studies (~42 000). In randomised trials (n=4138; median abatacept exposure, 12 (range 2-30) months), NMSC IRs (95% CIs) were not significantly different for abatacept (6.0 (3.3 to 10.0)) and placebo (4.0 (1.3 to 9.3)) and remained stable throughout the long-term, open-label period (median cumulative exposure, 28 (range 2-130 months); 21 335 p-y of exposure (7044 patients over 3 years)). For registry databases, NMSC IRs/1000 p-y were 5-12 (abatacept), 1.6-10 (csDMARDs) and 3-8 (other b/tsDMARDs). Claims database IRs were 19-22 (abatacept), 15-18 (csDMARDs) and 14-17 (other b/tsDMARDs). Pooled RRs (95% CIs) from observational studies for NMSC in patients receiving abatacept were 1.84 (1.00 to 3.37) vs csDMARDs and 1.11 (0.98 to 1.26) vs other b/tsDMARDs. CONCLUSIONS: Consistent with the warnings and precautions of the abatacept label, this analysis suggests a potential increase in NMSC risk with abatacept use compared with csDMARDs. No significant increase was observed compared with b/tsDMARDs, but the lower limit of the 95% CI was close to unity.