Empirical performance of the case-control method: lessons for developing a risk identification and analysis system

Area of Science:

• Pharmacovigilance and Pharmacoepidemiology
• Statistical and Epidemiological Methods

Background:

• Large-scale observational healthcare data are crucial for drug safety assessment.
• Various analytical methods, including case-control, cohort, and self-controlled designs, are employed.
• The performance characteristics of these epidemiological methods are not well-established.

Purpose of the Study:

• To evaluate the operating characteristics of the case-control method for drug safety analysis using real-world data.
• To determine the reliability of the case-control approach in large-scale observational studies.

Main Methods:

• Empirical evaluation of the case-control method across 5 real observational healthcare databases and 6 simulated datasets.
• Retrospective analysis of predictive accuracy using 165 positive and 234 negative controls.
• Assessment of performance for four key health outcomes: acute liver injury, myocardial infarction, kidney injury, and upper gastrointestinal bleeding.

Main Results:

• The case-control method demonstrated weak ability to distinguish between true and false drug safety signals.
• Estimates derived from the case-control method were positively biased.
• Confidence intervals associated with the case-control method exhibited inadequate coverage, indicating poor reliability.

Conclusions:

• The case-control method may not be suitable for accurately estimating drug-induced harm in pharmacovigilance.
• Findings suggest caution when using the case-control approach for the four studied outcomes.
• Further research into alternative methods for drug safety surveillance is warranted.