The Case-Crossover Design for Drug-Drug Interactions: Considerations for Implementation.

Area of Science:

• Pharmacovigilance
• Observational Study Design
• Health Informatics

Background:

• Electronic healthcare data offers potential for studying drug-drug interactions (DDIs).
• The case-crossover design's utility for DDIs in this data context is not well-established.
• Prior evidence suggests harm from specific drug combinations.

Purpose of the Study:

• To evaluate the case-crossover design for assessing clinical impact of DDIs.
• To compare different modeling approaches within the case-crossover framework.
• To investigate specific drug-drug interactions using US healthcare claims data.

Main Methods:

• Case-crossover analyses were performed on US healthcare claims data (1994-2013).
• Two drug-interaction scenarios were examined: statins with CYP3A4 inhibitors and clopidogrel with SSRIs.
• Analyses utilized a three-parameter model and a six-parameter saturated model, with and without active comparators.

Main Results:

• The three-parameter model yielded results consistent with prior evidence for statin-CYP3A4 inhibitor interactions.
• For clopidogrel-SSRI interactions, the three-parameter model produced unexpected results, but an active comparator improved consistency.
• The saturated model identified significant heterogeneity and confounding, suggesting specific strata yielded less biased estimates.

Conclusions:

• The three-parameter model assumes a uniform DDI effect, while the saturated model can reveal heterogeneity and confounding.
• Addressing within-person confounding may require restricting analyses to specific strata or using active comparators.
• The case-crossover design, with appropriate modeling, can be a valuable tool for DDI evaluation in electronic health records.