Is the birth cohort dimension associated with differences in statin therapy's effectiveness in reducing cardiovascular mortality? Evidence from aggregated time trend analyses

Maarten J. Bijlsma, University of Groningen
Fanny Janssen, University of Groningen
Stijn Vansteelandt, Ghent University
Eelko Hak, University of Groningen

Background: Cardiovascular mortality rates are affected by the birth cohort dimension. Recently, we observed that birth cohort effects also influence the utilization of cardiovascular medication. Objective: We investigated whether the effect of statin therapy on cardiovascular mortality differs between birth cohorts during the study period 1994 to 2010 in ages 50-83 in the Netherlands. Data & methods: Mortality data (myocardial infarction, ischaemic heart disease, cerebrovascular disease) aggregated by age and half study year were received from Statistics Netherlands, whereas data on statin therapy and other cardiovascular medications came from a nationally representative community pharmacy database (IADB.nl). We fit a Poisson regression model with count of cause-specific mortality as the outcome variable and life-years at risk of mortality (estimated from Statistics Netherlands data) as an offset variable. Age, birth cohort, prevalence of cardiovascular medications, and the interaction between prevalence of statin use and birth cohort were used as regressors in a modified version of the age-period-cohort characteristic (APCC) model. Results: We estimated that if statin use is at its mean level (approximately 27 users per 100 individuals), an increase in prevalence of statin use by 1 is associated with a reduction of 1.35% (95% CI 1.14 to 1.56%) in the number of individuals that would die of myocardial infarction two years later. This was 0.67% (CI: 0.45 to 0.89%) for other ischaemic heart disease and 0.90% (CI: 0.69 to 1.12%) for cerebrovascular disease. Birth cohort also significantly added to the models; in general in younger birth cohorts, CVD mortality rates were lower than among older cohorts. The 95% CI’s for the estimated interaction effects for birth cohort with prevalence of statin use generally contained 1, indicating no effect. Furthermore, significant effect sizes were not clinically relevant (< 0.1%). Conclusion: Differences in statin effectiveness by birth cohort were not detected.