Continuous variables are presented as means ± standard deviation, and dichotomous variables as percentages. Participants were divided into groups based on TnT levels: undetectable (n = 917) and detectable (n = 39). Those with detectable levels were further divided into 2 groups: low (<0.03 ng/ml) and high (≥0.03 ng/ml), corresponding to the cut point suggested by current clinical guidelines, which recommend that TnT levels greater than the 99th percentile of a healthy population, in which imprecision is ≤10% (0.03 ng/ml), are considered abnormal (38). Participants were also grouped by NT-proBNP level into low (<450 pg/ml, n = 758) and high (≥450 pg/ml, n = 199). This cut point corresponds to the recommended decision threshold for the general age group represented in the Rancho Bernardo Study (36). Differences in baseline levels of risk factors and clinical characteristics between participants with and without detectable TnT or high NT-proBNP were analyzed with t tests and chi-square tests; the Fisher exact test was used as appropriate. High-density lipoprotein (HDL), triglycerides, blood urea nitrogen (BUN), and NT-proBNP were not normally distributed and were log transformed for analyses; geometric means are reported. The association between NT-proBNP and age was examined using Spearman rank-order correlation. The NT-proBNP levels and creatinine clearance levels were compared in participants with undetectable, low, and high TnT levels using analysis of variance and Tukey post-hoc tests. Multivariate covariates of elevated TnT and NT-proBNP were identified by logistic regression including variables with significant univariate associations; covariates that remained significant at p < 0.05 were retained in the final model. Multivariate Cox proportional hazards regression models were used to determine the association of TnT and/or NT-proBNP with all-cause and CVD mortality. Results were expressed as hazard ratios (HRs) with 95% confidence intervals (CIs). For each analysis, 3 sequential Cox regression models were run. Model 1 adjusted for age and gender. For Model 2, life-style, risk factor, and laboratory covariates from (Table 1) were analyzed, and univariate predictors of all-cause mortality were identified if significant at p < 0.10. This process yielded age, gender, hypertension, BMI, heart rate, systolic blood pressure, diastolic blood pressure, diabetes, physical activity, BUN, eCrCl, and logHDL, low-density lipoprotein, and total cholesterol as potential predictors. Backward stepwise Cox regression analysis was performed using these 14 covariates; those that remained significant at p < 0.05 (age, gender, systolic blood pressure, BMI, heart rate, physical activity, eCrCl, and total cholesterol) were retained in the final Model 2. Forward stepwise analysis yielded the same 8 covariates. The influence of prevalent CHD was tested by adjusting for a history of CHD in Model 3, and also by excluding those with CHD at baseline and repeating Models 1 and 2. Participants who were alive were censored at the date of their last follow-up. For survival analyses with CVD death as the outcome, subjects who died of non-CVD causes were censored at date of death.