Background Observational studies are critical in assessing medication safety and effectiveness in the real world. Nonrandom assignment can provide insight to how and when medications are prescribed. Since the US introduction of TNF inhibitors (TNFi) in 1998, several newer biologics with varying mechanisms of action are available to patients and physicians.
Objectives To compare baseline characteristics and serious adverse events (serious infections (SI), malignancies) of patients who received abatacept (ABA) compared to those who received other comparison cohorts: conventional DMARDs (cD) and biologic DMARDs (bD) by line of treatment.
Methods Participating patients from 2005–2015 with RA in the National Data Bank for Rheumatic Diseases (NDB) provided treatment and other characteristics through self-reported biannual questionnaires. Initiators of ABA were matched to patients initiating other DMARDs in a 1:3 ratio. All patient-reported outcomes were verified by medical record review. ITT (OT) analyses were performed for malignancies (SI). Incidence rates were calculated using Poisson regression (1000 patients–years (pt-yr)). Marginal structural models were used to estimate the effect of treatment on the outcome by an appropriate control for the effects of time-dependent confounders using stabilized weights and reported as hazard ratios (HR). Variables included sex, employment, smoking, education, income, BMI, insurance, comorbidity index, RA severity, and co-medications.
Results Overall 1496 ABA patients with 2502 (4896) pt-yr of drug exposure (total follow-up time), 1520 cD treated patients with 4330 pt-yr of drug exposure (4816) and 3490 bD treated patients with 9658 pt-yr of drug exposure (11777) were studied (overall mean age 62 years (83% female)). At baseline, the distribution of first line patients in the several cohorts was: 15.2% ABA, for 23.4% bD and 33.8% cD. Switchers were more likely to have longer duration of RA (mean values (yrs): 11.4 for ABA, 8.3 cD; 14.7 bD first line vs. 17.6 ABA; 17.4 cD; 18.0 bD switchers), higher comorbidities, including prior cancer. ABA patients had worse disease compared to cD or bD, both used as first line or as switchers (mean HAQ any line for ABA 1.3 vs 1.0 (1.2) cD or 1.0 (1.3) bD as first line or as switchers). Excluding patients with prior events, incidence rates of malignancies tended to be uniform by line of treatment but SI seemed to increase with higher exposure of prior biologics, in all cohorts (Table). HR estimates by MSM did not show an increased risk for ABA in comparison with both cohorts, irrespectively of the line of treatment considered.
Conclusions Differences in baseline characteristics revealed that channeling bias might be present when comparing ABA with other DMARDs and by line of treatment. After adjustment, no overall risk was found for increased risk of cancer and infections.
Disclosure of Interest S. Pedro Employee of: National Data Bank for Rheumatic Diseases, R. Schumacher Employee of: National Data Bank for Rheumatic Diseases, K. Michaud Employee of: National Data Bank for Rheumatic Diseases