Age-period-cohort methodology: Confounding by birth cohort in cardiovascular pharmacoepidemiology

Preventive cardiovascular drugs are of the utmost importance for population health, and hence their information on their utilization and exact effectiveness is crucial. Clinical trials are commonly used for this assessment, but observational studies are also needed as clinical trial populations can differ behaviorally and demographically from end-users. However, observational studies are difficult to correctly perform; unmeasured variables that affect both drug utilization and cardiovascular outcomes, known as confounders, can distort estimates of drug effectiveness. A potentially important confounder is birth cohort; other studies have shown that birth cohort, which represents in what time-period a person is born, affects both drug utilization and affects cardiovascular mortality. If it turns out that birth cohort is indeed a confounder, then it should be taken into account when performing drug effectiveness studies.

It was investigated whether birth cohort was a confounder, both at the population-level and at the patient-level, of the relation between statins (cholesterol lowering drugs) and cardiovascular mortality. This is not easy, as adjusting for birth cohort imposes a statistical identification problem (cohort = period – age). Data that was representative for the Netherlands in the period 1994 – 2012 was used. It was determined that birth cohort can confound population-level estimates of interventions on statin utilization. However, birth cohort was not a confounder of the individual-level effect of statins on cardiovascular mortality. Furthermore, an innovative method determining birth cohort effects was extended and assessed, which can better identify the causal pathways by which age, period and cohort affect an outcome.