Prediction of atherosclerotic cardiovascular death in men using a prognostic score

Abstract

Treadmill and clinical data were gathered prospectively on consecutive patients who underwent exercise testing for evaluation for coronary artery disease in a 1,200 bed Veterans Affairs Medical Center. From 3,609 men referred for exercise testing from 1984 to 1990, 3,134 patients remained after excluding those with significant valvular heart disease and those with prior coronary artery bypass surgery. Of these, 588 were selected for clinical reasons to undergo cardiac catheterization within 3 months of evaluation leaving 2,546 who were not selected. Over 3 years, there were 158 cardiovascular deaths, 99 nonfatal myocardial infarcts and 183 patients who underwent coronary artery bypass surgery. In the total population, the Cox proportional-hazards model demonstrated the following characteristics to be statistically significant independent predictors of time until cardiovascular death: a history of congestive heart failure and/or taking digoxin, exercise-induced ST depression, the change in systolic blood pressure during exercise, and exercise capacity in METs. Using the Cox model coefficients to weight the variables, a simple score (the Veterans Affairs Prognostic Score) was constructed based on these items. Average annual cardiovascular mortality was plotted against the score enabling its estimation for any given patient. In the subgroup selected for cardiac catheterization (n = 588), the mean score was greater, consistent with a poorer prognosis, compared with the total population; 53% (n = 312) had a score <−2 associated with an annual mortality <2%. Thus, in over half of the patients selected for catheterization, the catheterization was unnecessary if performed to lessen their chance of cardiovascular death, since no intervention could improve their prognosis. Similarly, in patients selected for bypass surgery, the mean score was −0.5, which is associated with an estimated annual cardiovascular mortality of 5%. Thirty-five percent of these patients had a score <−2. Thus, variables available from the usual noninvasive workup of patients with known or suspected coronary artery disease can be used to estimate cardiovascular mortality. Exercise-induced ST depression can be falsely excluded from predictive models because of workup bias. This score can estimate cardiovascular mortality in male veterans being evaluated for stable coronary artery disease, obviating the need for cardiac catheterization in many of them.