Pharmacoepidemiological approaches for population-based hypothesis testing

Abstract

Pharmacoepidemiology aims to study the use and both the adverse and beneficial effects of drugs and vaccines in the population after market authorization. The efficacy of drugs is assessed in experimental studies before a drug is allowed on the market in a limited and usually selected group of patients. Therefore, after market authorization the focus is on serious and adverse effects in large groups of patients in daily clinical practice. Observational drug research is needed to establish and measure these effects. Observational research faces several challenges to minimize the chance of bias, including confounding by indication, which is caused by selective prescribing of drugs to certain patient groups. A comparison between treated and untreated subjects or between different drug regimens may be biased due to uneven distribution of risk factors for the outcome of interest. Important progress has been made during the past decade in controlling confounding by design and analysis in observational studies. The increasing accessibility of large electronic health record databases has fuelled various international initiatives to analyze multiple databases across countries using common protocols and common data models. Extensive sensitivity analysis across multiple designs, databases, and analytical techniques has provided more insight into causes of variation in results across studies and increases the confidence in findings of observational studies. Transparency of observational drug research through public registration of protocols and detailed reporting of methods should improve reproducibility and thereby reliability of pharmacoepidemiological studies.