Heterogeneity of the phenotypic definition of coronary artery disease and its impact on genetic association studies

Abstract

Background: Variability in phenotypic characterization of coronary artery disease (CAD) may contribute to the heterogeneity of genetic association studies, and more consistency in phenotype definitions might improve replication of genetic associations. We assessed the extent of phenotypic heterogeneity and quantified its impact in a large literature sample of association studies.

Methods and results: We searched for large (≥15 studies) meta-analyses of genetic associations and reviewed all studies included therein. From each primary study, we extracted phenotypic definitions, demographics, study design characteristics, and genotypic data. For each association, we assessed the magnitude and heterogeneity of genetic effects within and across CAD phenotypes, using meta-analytic methodologies. A total of 965 individual studies investigating 32 distinct variants in 22 genes were included, from which we grouped CAD phenotypes into 3 categories: acute coronary syndromes (ACS) (426 [44%] studies); angiographically documented disease (323 [34%] studies); and broad, not otherwise specified CAD (216 [22%] studies). These clinical phenotypes were overlapping. Subgroup meta-analyses by phenotype showed discordant results, but phenotypic classification generally explained small proportions of between-study heterogeneity. Differences between phenotypic groups were minimized for associations with robust statistical support. No CAD phenotype was consistently associated with larger or more homogeneous genetic effects in meta-analyses.

Conclusions: Substantial phenotypic heterogeneity exists in CAD genetic associations, but differences in phenotype definition make a small contribution to between-study heterogeneity. We did not find a consistent effect in terms of the magnitude or homogeneity of summary effects for a specific phenotype to support its preferential use in genetic studies or meta-analyses for CAD.

Figure 1

Flow chart of selected meta-analyses/genetic association studies and studies excluded, with specification of reasons.

Figure 2

Venn diagram showing the volume of studies belonging to each one of the phenotypic sampling strategies (ACS: acute coronary syndrome, n=426; angiographic definition, n=323; broad definition, n=216) and the extent of conceptual overlap (in terms of patients suffering from ACS) between them. The extent of overlap between the angiographic and broad definitions (i.e. patients with qualifying angiographic lesions in the broad definition) is not known.

Figure 3

Forest plots indicating meta-analyses summary results (odds ratios and 95% confidence intervals) for all studies included and for phenotypic definition subgroups. The symbols of the point estimates indicate the major pathway where each gene belongs (hollow square: lipid metabolism; filled square: thrombosis-hemostasis; hollow circle: endothelial dysfunction; filled circle: inflammation; x mark: unknown mechanism)

Figure 4

Forest plots indicating the relative odds ratios (RORs) and the corresponding 95% confidence intervals for the between subgroup comparisons. The symbols of the point estimates indicate the major pathway where each gene belongs (hollow square: lipid metabolism; filled square: thrombosis-hemostasis; hollow circle: endothelial dysfunction; filled circle: inflammation; x mark: unknown mechanism)

Figure 5

Bar diagram showing the results of the Q-partitioning analyses for the major phenotypic subgroups. The length of each bar corresponds to the Q-statistic observed in the all-inclusive meta-analysis. Each bar is partitioned to the within subgroup Q statistics (Q_ACS, Q_broad, Q_angiographic) and the remaining between subgroup heterogeneity (Q_phen). Statistically significant results for the Q_phen are depicted with asterisks.