Background: Recent studies have demonstrated a strong association between exercise capacity and mortality, but potential bias due to differences in the distribution of baseline variables has not been adequately considered.

Aims: To assess the association between exercise capacity and mortality using a propensity score matching method.

Method: A retrospective review of a clinic-based cohort was conducted. Included were male veterans with (n=592) and without (n= 6,167) Type-2 diabetes who underwent maximal exercise testing for clinical reasons between 1986 and 2011. Three fitness levels were established based on peak METs achieved. Propensity scores were used to balance covariate distributions between groups with and without Type-2 diabetes.

Results: Significant predictors for all-cause mortality were: age, hypertension, smoking, Type-2 diabetes, body mass index, and exercise capacity. For each 1 MET increase in exercise capacity in the unmatched group, the adjusted HR for mortality was 0.83 in those with diabetes compared to 0.87 in those without diabetes. Similar trends were observed for the matched dataset: the HR for adjusted mortality was 0.83 for subjects with diabetes compared to 0.88 for those without diabetes.

Discussion and Conclusions: Exercise capacity is a strong predictor of mortality in veterans with and without diabetes. Although the trend in the association between exercise capacity and all-cause-mortality was similar for matched and unmatched data, the mortality risks were relatively inflated when using unmatched data. Such differences suggest that, without minimizing the impact of confounding variables, results derived from an unmatched dataset may be biased.