Addressing Time-Varying Confounding in Self-Controlled Case Series Using Active Comparators
Overview
This report discusses methods to incorporate active comparators in self-controlled case series (SCCS) designs to address time-varying confounding by indication. Two case studies using UK Clinical Practice Research Datalink data illustrate these methods, focusing on thiazolidinedione use and fractures, and fluoroquinolone use and uveitis. Recommendations are provided to enhance the robustness of pharmacoepidemiologic studies using SCCS with active comparators.
Background
The SCCS design compares risk periods to reference periods within the same individual, inherently controlling for time-invariant confounding. However, it is vulnerable to time-varying confounding, such as changing indications for treatment over time. Active comparators, commonly used in cohort and case-control studies, can help address this by providing a comparison group with similar time-varying confounding patterns. Recent methodological developments allow the incorporation of active comparators into SCCS, which was previously not permitted in the basic method.
Data Highlights
Two case studies were conducted using the UK Clinical Practice Research Datalink: one examining thiazolidinedione use and fractures, and another examining fluoroquinolone use and uveitis. These studies applied methods comparing regression coefficients or using nested regression models to derive active comparator rate ratios within SCCS frameworks, demonstrating practical application of the methods to real-world pharmacoepidemiologic data.
Key Findings
The SCCS design controls for time-invariant confounding but is susceptible to time-varying confounding by indication.
Active comparators can reduce bias from unmeasured time-varying confounders by providing a similar confounding pattern.
Methods to incorporate active comparators in SCCS include comparing regression coefficients or using nested regression models to derive rate ratios.
Application of these methods in two case studies showed their utility in addressing confounding in pharmacoepidemiologic research.
Incorporating active comparators in SCCS enhances the validity and interpretability of causal inferences in drug safety studies.
Clinical Implications
Pharmacoepidemiologists should consider using active comparators within SCCS designs when time-varying confounding by indication is a concern. Employing these methods can improve causal inference by reducing bias, particularly in studies of drug safety where indications and exposures vary over time. This approach supports more robust and reliable evaluation of adverse drug effects.
Conclusion
Incorporating active comparators into SCCS designs offers a valuable strategy to address time-varying confounding by indication, enhancing the robustness of pharmacoepidemiologic studies. The described methods and case study applications provide practical guidance for researchers aiming to improve causal inference in observational drug safety research.
References
Petersen et al. 2023 -- Addressing Time-Varying Confounding in Self-Controlled Case Series Using Active Comparators
