Estimated hazard ratios for death from specific and all causes among twins free of baseline diabetes and coronary heart diseasea.

aHazard ratios and 95% confidence interval were estimated for per 50 mg/dL increment in postload glucose levels at 1 hour after a 50 gram glucose load for the overall and the between-pair effects; or per 50 mg/dL difference between co-twins for the within-pair effects. Frailty survival models were used to obtain the estimates to account for clustering within a pair. Fully adjusted hazard ratios were controlled for age (continuous), smoking (never, past smokers, and current smokers), marital status (never, not married currently, and married currently), years of education (continuous), body mass index (continuous), systolic blood pressure (continuous), low-density lipoprotein cholesterol (continuous), ratio of high density lipoprotein cholesterol to triglyceride (continuous), and antihypertensives (yes/no).bInteraction between within-pair effects and zygosity was tested in the fully adjusted model.