Abstract
BACKGROUND: Due to the paucity of direct evidence, we aimed to evaluate whether the association between postload plasma glucose levels (ppGlucose) and long-term risk of mortality from coronary heart disease was independent of or attributable to genes and common environment. METHODS AND FINDINGS: From the prospective National Heart, Lung, and Blood Institute (NHLBI) Twin Study, we included 903 middle-aged male twins, who were nondiabetic, free of coronary heart disease at baseline (1969-1973), and followed for up to 38 years for coronary heart, cardiovascular, and all-cause mortality. Frailty survival models were used to estimate hazard ratio (HR) for various associations: overall (equivalent to singleton population association), within-pair (controlling for genes and environment common to co-twins), and between-pair association (reflecting influences of common factors). Overall associations were statistically significant for coronary heart and cardiovascular but not all-cause deaths after controlling for known risk factors. The associations were not statistically significant in within-pair analyses. The within-pair association was not statistically different by zygosity for specific and all-cause death risk. After the full adjustment for known risk factors, HR (95% confidence interval) for within-pair association was 1.07 (0.90, 1.28), 1.06 (0.94, 1.19), and 0.99 (0.94, 1.05) for coronary heart, cardiovascular, and all-cause mortality, respectively. The fully adjusted between-pair associations were statistically significant for specific and all-cause death risk: a 50 mg/dL increase in the mean value of ppGlucose for a twin pair was associated with a raised death risk [HR (95% confidence interval) 1.15 (1.02, 1.30), 1.10 (1.02, 1.20), and 1.05 (1.01, 1.09) for coronary heart, cardiovascular, and all-cause mortality, respectively]. Between-pair association was significant in dizygotic but not in monozygotic twins. CONCLUSION: The positive association between ppGlucose and long-term coronary heart mortality risk is largely explained by factors shared between co-twins, including familial factors; however, within-pair effects cannot be absolutely excluded.