Because there are no defined categories for the 6MWT in patients with CHD, participants were divided into quartiles on the basis of 6MWT distance. Baseline participant characteristics across quartiles were compared using analysis of variance for continuous variables, the χ2 test for dichotomous variables, and the Fisher exact test for dichotomous variables with fewer than 5 participants in a category. We compared unadjusted rates of cardiovascular events by quartile using Cox proportional hazards models and the log-rank test. We compared adjusted rates of cardiovascular events, analyzing 6MWT distance as a continuous variable, per 1-SD decrease of 6MWT distance using Cox proportional hazards models adjusted for covariates associated with the 6MWT quartile at P < .10. For any covariates with more than 1% missing data, multiple imputation was performed using iterative chained equations including history of hypertension (3.4% missing), dyslipidemia (5.6%), diabetes mellitus (3.8%), peripheral vascular disease (15.3%), and chronic lung disease (4.9%), as well as ejection fraction (3.4%), diastolic dysfunction (10.6%), log NT-proBNP (4.5%), and log C-reactive protein (4.5%). We tested for interactions between 6MWT distance and age, sex, current smoking, diabetes, BMI, systolic blood pressure, estimated glomerular filtration rate, hemoglobin, and left ventricular ejection fraction. Using a logistic regression model for predicting cardiovascular events on the basis of traditional risk factors (age, sex, current smoking, history of hypertension, history of dyslipidemia, history of diabetes, BMI, systolic blood pressure, diastolic blood pressure, and total cholesterol to high-density lipoprotein cholesterol ratio), we estimated the area under the receiver operating characteristic curve (C statistic), integrated discrimination improvement, and category-free net reclassification improvement for predicting cardiovascular events for the individual addition of continuous measures of the 6MWT, treadmill exercise capacity, NT-proBNP, C-reactive protein, and ejection fraction.27- 29 We compared treadmill exercise capacity to 6MWT distance using the Pearson correlation coefficient. Analyses were performed using Stata, version 12 (StataCorp).