Click on image to view larger version.

Figure 1 Boxes are point estimates of multivariable Cox proportional hazards ratios (lines indicate 95% confidence intervals) for a one-standard deviation increase of the continuous variables (fasting proinsulin, apolipoprotein B/A-I ratio [APOB/APOA-I], beta-carotene, and body mass index) and for occurrence versus non-occurrence of dichotomous variables (hypertension, electrocardiographic left ventricular hypertrophy, and interim myocardial infarction) as predictors of heart failure incidence in middle-aged men free from heart failure and valvular disease at baseline. (A) Variables that remained independently significant in Table 3, with the addition of interim myocardial infarction. (B) Subsample of subjects free from myocardial infarction at baseline and during follow-up.