Sources of heterogeneity in case-control studies on associations between statins, ACE-inhibitors, and proton pump inhibitors and risk of pneumonia.

The heterogeneity in case-control studies on the associations between community-acquired pneumonia (CAP) and ACE-inhibitors (ACEi), statins, and proton pump inhibitors (PPI) hampers translation to clinical practice. Our objective is to explore sources of this heterogeneity by applying a common protocol in different data settings. We conducted ten case-control studies using data from five different health care databases. Databases varied on type of patients (hospitalised vs. GP), level of case validity, and mode of exposure ascertainment (prescription or dispensing based). Identified CAP patients and controls were matched on age, gender, and calendar year. Conditional logistic regression was used to calculate odds ratios (OR) for the associations between the drugs of interest and CAP. Associations were adjusted by a common set of potential confounders. Data of 38,742 cases and 118,019 controls were studied. Comparable patterns of variation between case-control studies were observed for ACEi, statins and PPI use and pneumonia risk with adjusted ORs varying from 1.04 to 1.49, 0.82 to 1.50 and 1.16 to 2.71, respectively. Overall, higher ORs were found for hospitalised CAP patients matched to population controls versus GP CAP patients matched to population controls. Prevalence of drug exposure was higher in dispensing data versus prescription data. We show that case-control selection and methods of exposure ascertainment induce bias that cannot be adjusted for and to a considerable extent explain the heterogeneity in results obtained in case-control studies on statins, ACEi and PPIs and CAP. The common protocol approach helps to better understand sources of variation in observational studies.