6F-06 - A Case Study Examining Ranolazine and Seizure Risk to Compare Potential Bias when Using Self-Controlled Study Designs

Background: Prior research suggests that ranolazine, used for the treatment of chronic angina, is associated with an increased risk of seizures immediately following initiation. Because ranolazine is typically prescribed only after patients fail first- and second-line angina therapies, the absence of a suitable active comparator motivates the use of self-controlled study design to assess the safety of this medication. Thus, we aimed to compare estiamtes of seizure risk immediately following ranolazine using self-controlled case series (SCCS) and self-controlled-risk-interval (SCRI) designs.

Objectives: To compare estimates of seizure risk immediately following ranolazine using SCCS and SCRI designs, where time-varying bias is a concern.

Methods: We used SCCS and SCRI study designs to investigate the risk of seizure associated with ranolazine initiation among Medicare beneficiaries > 65 years old between 2011-2020. The exposure risk window was defined as the 30 days after ranolazine initiation, and further subdivided into three 10-day intervals (high-risk: days 1-10; medium-risk: days 11-20; low-risk: days 21-30). The control period for the SCCS design consists of all observable time outside the risk windows, while that for the SCRI design consists of the 30 days immediately following the risk window. We further conducted an extension of the SCCS design based on a pseudo-likelihood approach which has been proposed to better handle violations of the assumption of event-dependent exposures.

Results: We found consistently larger incidence rate ratios (IRRs) using the SCCS design compared to the SCRI design. Using the SCCS design, we showed elevated seizure risk in the high- and medium- risk periods only (IRR [95% CI]; high-risk: 3.89 [2.68-5.65]; medium-risk: 2.01 [1.21-3.36]; low-risk: 1.21 [0.62-2.33]). Using the SCRI design, we showed elevated seizure risk in the high-risk period only (IRR [95%CI]; high-risk: 2.49 [1.52-4.07]; medium-risk: 1.29 [0.70-2.35]; low-risk: 0.77 [0.37-1.61]). In our sensitivity analysis limiting the SCCS post-exposure control period to 30 days following ranolazine initiation, there was markedly better agreement with SCRI results (IRR [95% CI]; high-risk: 2.63 [1.74-3.98]; medium-risk: 1.36 [0.79-2.33]; low-risk: 0.82 [0.42-1.60]).

Conclusions: Our results suggest that unmeasured time-varying confounding and violation of the event-dependent exposure assumption may lead to biased estimates of seizure risk following ranolazine exposure, particularly with the standard SCCS design. We suggest careful consideration regarding assumption violations and choice of control risk periods when using self-controlled methods to assess medication safety.