Improving consistency in findings from pharmacoepidemiological studies: The IMI-protect (Pharmacoepidemiological research on outcomes of therapeutics by a European consortium) project

Abstract

Background: Pharmacoepidemiological (PE) research should provide consistent, reliable and reproducible results to contribute to the benefit-risk assessment of medicines. IMI-PROTECT aims to identify sources of methodological variations in PE studies using a common protocol and analysis plan across databases. In addition, differences by design, applied to a same drug-adverse event (AE) pair in different databases are examined. Results from PE studies will be evaluated on seven drug-AE pairs (i.e. (1) antibiotics and acute liver injury; (2) antidepressants and hip fracture; (3) benzodiazepines and hip fracture; (4) anticonvulsants and suicide/suicide attempts; (5) calcium channel blockers and malignancies; (6) inhaled long-acting b2 agonists and acute myocardial infarction; (7) a negative control study: antibiotics and acute myocardial infarction) conducted in eight European and one US electronic databases. These are: the UK (CPRD), UK (THIN), the Danish national registries, the Dutch Mondriaan project (NPCRD, AHC), the Spanish BIFAP, the German Bavarian Claims (KVB) database (only descriptive), PGRx and US InVision Datamart (formerly LabRx). In order to maintain the blinding of investigators from one another's results, these results will only be disclosed during the ICPE conference. Objectives: To review and understand the methodological issues encountered in these studies and to draw conclusions about their relevance for future PE research. Description: We will present data on association studies in the various databases using different designs with a focus on cohort, but also case-control, case-crossover, and self-controlled case series for some drug-AE pairs. The major methodological issues such as choice of study design, analytical methods to control for confounding, variation in operational definitions of exposure, outcome and confounders across databases with different coding systems will be discussed.