Background: The clinical and exercise test variables chosen for predicting prognosis vary in the available studies. This could be due to the effect of age of the patients tested and the choice of outcomes used as end points in these follow-up studies.
Objective: To evaluate the effect of age and end points on exercise test variables chosen as significantly and independently associated with time to death.
Methods: Analyses were performed on the first treadmill test performed on consecutive male veterans at the Palo Alto and Long Beach Veterans Affairs Medical Centers since 1987. After removal of patients with congestive heart failure, coronary interventions, left bundle-branch block, atrial fibrillation, myocardial infarction and/or Q wave, and digoxin use, 3,745 male subjects remained. The outcomes were cardiovascular and all-cause mortality. The study population was divided into subsets according to age; exercise test and clinical variables were analyzed within the age subsets using the Cox hazard model.
Results: The mean age at the time of testing was 57 +/- 12 years (+/- SD) and they were followed up for a mean of 6.6 years. There were 544 all-cause deaths, with 206 of the deaths being due to cardiovascular causes (38%). When the study group was classified into subsets based on age, exercise capacity (in metabolic equivalents [METs]) was chosen by the Cox hazard model most consistently in the age groups using either end point. Even when age was added to the Duke treadmill score, prediction of death did not improve in those > 70 years of age because of the nonlinear relationship between age, the exercise test variables, and time to death. The most important age cut points for clinically important differences in exercise test predictors appeared to be 70 years and 75 years of age. In the patients 70 to 75 years of age, peak METs was the only variable predictive of all-cause mortality, and exercise-induced ST-segment depression was the only predictor of cardiovascular death; in the patients > 75 years of age, none of the exercise test responses were predictive of either death outcome.
Conclusion: Both age and the outcome selected as an end point affect the exercise test responses chosen for scores to predict prognosis. Differences in age of the subjects tested and/or the outcome selected as the end point can explain the differences in the studies using exercise testing to predict prognosis.