Safety of Abatacept Compared With Other Biologic and Conventional Synthetic Disease-Modifying Antirheumatic Drugs in Patients With Rheumatoid Arthritis

Data From an Observational Study

Gulsen Ozen; Sofia Pedro; Rebecca Schumacher; Teresa A. Simon; Kaleb Michaud


Arthritis Res Ther. 2019;21(141) 

In This Article


In this US-wide observational cohort study of patients with RA, abatacept was associated with low IRs of malignancies, infections, autoimmune diseases, infusion/injection reactions and mortality, with no new safety signals. The incidence of infusion/injection-site reactions was lower with abatacept than with other bDMARDs. The risks of overall malignancies, infections and autoimmune diseases were comparable across treatment groups; however, the risk of hospitalized infection was lower with abatacept than with other bDMARDs.

Our study showed a low IR of new malignancies with abatacept. To date, only two observational studies (Sweden national cohort and CORRONA registry) have examined the comparative risk of malignancy associated with abatacept; they reported an increased risk of NMSC with abatacept compared with csDMARDs (HR [95% CI] 2.15 [1.31, 3.52])[23] and methotrexate (15.3 [2.05, 114.0]),[28] whereas the comparative risks of haematological and solid malignancies were similar.[23,28] In contrast to our analysis, both studies included patients with a prior history of malignancy, which was reflected in the higher rates of malignancy observed. The power restraints and inadequate follow-up duration of RCTs for the detection of malignancies and the scarcity of observational data suggest further research is warranted to improve understanding of the risk of malignancy associated with abatacept, particularly in comparison with other bDMARDs.

The incidence of hospitalized infections with abatacept in this analysis was lower than that reported in other observational studies, possibly due to differences in the characteristics of the study populations, such as patients' sex and co-morbidities,[20,25] or in the definitions used for infections.[22]

Following adjustment for clinical factors influencing infection risk, we found that the overall risk of hospitalized infections with abatacept was lower than that of other bDMARDs and csDMARDs. When limited by line of therapy, the risk difference in hospitalized infections between abatacept and other bDMARD groups decreased, possibly due to infections from prior drugs precluding further use of bDMARDs. Although the comparative data currently available are limited and inconsistent, our findings are consistent with an overall trend toward a lower risk of hospitalized infections or SIs with abatacept compared with other bDMARDs.[19,22,25,34] Observational findings from administrative datasets have shown that compared with other bDMARDs, abatacept was associated with a lower risk of SIs (19, 25). Conversely, other observational studies found that SI risk with abatacept was comparable with that of rituximab[22] and etanercept.[34] Although the risk of SIs with abatacept compared with other bDMARDs observed in our analysis was lower, the risk estimate was smaller. This difference could be due to variables such as the mean age of the study population, treatment episodes (incident/prevalent use), treatment line, comparator groups, covariates included and methods used to address channelling bias. The hospitalized infection risk with abatacept treatment compared with csDMARDs in the present study is consistent with an epidemiologic assessment of data from seven clinical trials that showed the IRs of SIs with abatacept were comparable with those in patients with RA treated with csDMARDs.[35] However, we observed a lower hospitalized infection risk with abatacept treatment compared with csDMARDs despite being not significant. At baseline, csDMARD initiators were slightly older, had higher smoking frequencies and had a slightly higher dose of glucocorticoid use; these are potentially associated with infection risk. Although we used an appropriate methodology, MSM, to address the channelling bias, as an observational study there may be residual channelling due to unmeasured confounders.

In this analysis, the IRs of most autoimmune diseases were similar across treatment groups. Psoriasis was the most frequently occurring autoimmune disease in all treatment groups; however, the IRs were within the ranges reported elsewhere.[13,17] Patients with a pre-existing autoimmune disease are known to be at an increased risk of developing another.[36–38] Psoriasis, a T cell-mediated autoimmune disease, is known to coexist with RA more frequently than with osteoarthritis.[38] As expected, we did not observe an altered risk of psoriasis with abatacept treatment compared with other bDMARDs and csDMARDs.

We observed low IRs of infusion/injection reactions with abatacept treatment. In this analysis, use of IV abatacept was associated with a slightly higher IR of severe reactions than the SC formulation, possibly due to the chronological availability of the two formulations, leading to the inclusion of fewer patients treated with SC abatacept and correspondingly shorter follow-up periods than for those treated with the IV formulation. A phase IIIb noninferiority study to compare the efficacy and safety of SC versus IV abatacept reported similar incidences of infusion and injection reactions (2.2% vs 2.7%, respectively).[39] In this analysis, lower IRs of infusion/injection reactions were observed with abatacept versus TNFis and other bDMARDs, which is consistent with the ATTEST and AMPLE trials that showed infusion and injection reactions were more frequent with infliximab and adalimumab than with abatacept.[11,16] Furthermore, an examination of healthcare data indicated the IRs of hypersensitivity reactions were lower for abatacept than for other injectable bDMARDs.[24]

Although early, aggressive treatment strategies and the use of bDMARDs have improved treatment outcomes in RA,[1,40] safety concerns have been raised.[3] Research on the comparative risks and benefits of treatment using observational data have become critical, as comparative safety data from RCTs are scarce. Conducting an appropriate analysis to address channelling bias in observational studies is challenging; however, the use of MSM methodology to adjust for time-varying confounders by weighting for the treatment groups where the baseline characteristics were balanced allowed us to obtain less biased estimates.[32] This methodology also addresses the attrition bias between treatment groups by weighting for the censoring distribution and balancing the characteristics of patients lost to follow-up with those who were followed up.

Despite using a strong methodology, our study had some limitations. First, the numbers of events, and accordingly incidence rates, reported were lower than in previous observational studies, which precluded the construction of models to assess the risk of different malignancies and hospitalized infections. It is possible that the inclusion of different study populations and the application of the strict validation process in FORWARD, although increasing the accuracy of the events reported, may be among the reasons for lower incidence rates of safety events than in previous observational studies. In addition, patients who are possibly in better health may be more likely to participate in FORWARD than those who are frail and at higher risk of infection and cancer. This participation bias can also explain the low incidence rates of the events. Moreover, due to the self-reported nature of the data, some events may not have been recorded. Second, our sample size was inadequate to examine the safety risks associated with abatacept compared with individual bDMARDs or classes of bDMARDs. Furthermore, the analysis population included both biologic-naïve and biologic-experienced patients, for whom safety risks could differ. In order to address this, we added the number of previous bDMARDs to the model. We also performed a subgroup analysis on biologic-naïve and biologic-experienced patients (Table 2); however, the sample size was not adequate to see trends in safety risks. There is the potential for time-lag bias due to the 10-year follow-up period, particularly as patients were allowed to contribute to different treatment groups and drug indications may change over time. However, calendar matching will have minimized this potential bias.[41] The matching did not take into consideration that patients may switch treatments; however, continuing to include switchers in the study increased the number of patients included. Lastly, although our analyses included several clinical confounders including disease severity measures, unmeasured confounders could be present in this observational study.