Understanding the Utility of Zero Coronary Calcium as a Prognostic Test

Abstract

Coronary artery calcium (CAC) is nearly pathognomonic for the presence of coronary atherosclerosis, although not all coronary plaque is calcified.1,2 As such, cardiac CT scans for CAC are highly specific but less sensitive for the presence of some coronary atherosclerosis. In contrast, these tests are highly sensitive but less specific for detecting obstructive atherosclerotic coronary artery disease (CAD). A Bayesian approach is critical for discerning the value of new tests. When used in the intended population, a test with high sensitivity is excellent for ruling out disease (ie, highly sensitive tests produce a high negative predictive value [NPV] when negative). For this reason, there is interest in using “zero” CAC scores to predict excellent prognosis—and potentially conserve medical resources—in the large population of patients at intermediate midterm risk of developing clinical cardiovascular disease. The efficiency of any test relies on the frequency of disease in the population tested (the prior probability distribution). For example, take the d-dimer test for deep venous thrombosis and pulmonary embolism. If a d-dimer is checked in all outpatients, a low positive predictive value (PPV) will result given the low frequency of thrombosis in this scenario. If a d-dimer is tested exclusively in patients with acute dyspnea and a swollen lower extremity, there will be a low NPV given the high incidence of pulmonary embolism in such patients. Stated in mathematical terms, the likelihood of thrombosis (A) given a positive d-dimer (B) depends not only on the conditional probability of B given the presence or absence of A (the test characteristics), but also on the marginal probability of A independent of B (the prior probability distribution). This principle is summarized in probability theory by Bayes theorem (Figure 1). Knowingly or unknowingly, Bayes theorem underlies essentially all clinical decision making. …