Introduction: Individual participant data (IPD) meta-analyses involve participant-level data from multiple cohort studies. However, these cohorts have different periods (years) of follow-up, target regions, and distributions of risk factors (including patient age). It remains unclear if these variations affect the heterogeneity of absolute/relative measures of mortality in cardiovascular disease (CVD), stroke, and coronary heart disease (CHD) among cohorts.
Hypothesis: There is diverse heterogeneity in absolute measures of mortality, but negligible heterogeneity in relative measures among cohorts in IPD meta-analyses.
Methods: The Evidence for Cardiovascular Prevention from Observational Cohorts in Japan (EPOCH-JAPAN) study is an IPD meta-analysis of cardiovascular epidemiology. This project comprises 14 cohort studies with 105,945 Japanese subjects (total CVD deaths: 5,314). First, we examined the correlation between the follow-up periods of the baseline surveys and multivariate-adjusted mortality rates (CVD, stroke, and CHD) among the cohorts. Next, we estimated the cohort-specific mortality rates that adjusted for the stated follow-up periods, regions, age, and other risk factors using Poisson regression. Finally, we explored the heterogeneity of multivariate-adjusted mortality rates, mortality rate ratios, and rate ratios of 10-mmHg increases in systolic blood pressure using Higgins’s I2, which measures heterogeneity in meta-analyses.
Results: High correlations were observed between the stated follow-up periods of the cohorts and their mortality rates (CVD [men, -0.70; women, -0.79], stroke [men, -0.65; women, -0.73], CHD [men, -0.24; women, -0.89]). In the multivariate-adjusted mortality rates, we observed clear heterogeneity in mortality rates among the cohorts (CVD [I2: men, 98.6%; women, 99.3%], stroke [I2: men, 98.5%; women, 98.3%], and CHD [I2: men, 98.2%; women 92.4%]). In the rate ratio comparison of 10-mmHg increases in systolic blood pressure, no heterogeneity was detected among the cohorts (CVD [I2: men, 0.0%; women, 17.9%]). Our results indicated that the ratio measure, which shows the magnitude of each risk factor, was stable even in the heterogeneity of absolute measures.
Conclusions: A clear heterogeneity in mortality was observed in absolute measures, but not in relative measures, among cohorts after adjusting for the periods of follow-up, regions, and other risk factors.