Time-related biases in pharmacoepidemiology

Area of Science:

• Pharmacoepidemiology
• Health Informatics
• Biostatistics

Background:

• Computerized healthcare databases are increasingly used for drug repurposing research.
• Observational studies in this area often report significant drug effectiveness.
• However, these findings can be compromised by various time-related biases.

Purpose of the Study:

• To describe and illustrate common time-related biases in observational drug studies.
• To demonstrate the impact of these biases using a cohort of patients with chronic obstructive pulmonary disease (COPD).

Main Methods:

• Utilized Quebec healthcare databases to identify 124,030 patients (≥50 years) with COPD (2000-2015).
• Examined exposures like inhaled corticosteroids (ICS) and long-acting bronchodilators.
• Investigated outcomes including lung cancer, myocardial infarction, and death, applying biases like protopathic, latency, immortal, time-window, depletion of susceptibles, and immeasurable time.

Main Results:

• Protopathic bias inflated lung cancer rates in the first year of bronchodilator use.
• Misclassification of latency and immortal time led to an apparent decreased lung cancer risk with ICS (HR 0.32), which changed to 0.50 and then 0.96 after corrections.
• Other time-related biases similarly impacted study findings.

Conclusions:

• Many observational studies on drug repurposing report exaggerated effectiveness due to avoidable time-related biases.
• Proper study design and analysis are essential to mitigate these biases.
• Future research must address these temporal issues to prevent severely biased conclusions in drug repurposing.