Records from treadmill maximal graded exercise tests (GXTs) for 61 patients from the Virginia Tech Intervention Center were screened for changes in systolic blood pressure. These blood pressure responses were standardized according to exercise demand (ΔSBP/MET) between three different levels of the exercise test. Subject records were chosen on the basis that they did not reflect a physician diagnosis of coronary artery disease (CAD) and were not taking antihypertensive medications. The ΔSBP/MET responses were stratified as follows: low to moderate (ΔBP/MET1) = difference between a systolic blood pressure at a moderate intensity stage minus the first stage systolic blood pressure, adjusted for the corresponding changes in metabolic demand (MET); moderate to high (ΔBP/MET2) = difference between systolic blood pressure at the maximal stage minus the moderate-intensity stage per MET change; and low to high (​​ΔBP/METS) = difference between systolic blood pressure at the maximal stage minus the first stage per MET change. Subjects were separated (STΔ and NoSTΔ) according to whether or not they had exercise-induced ST segment shift of 1 mm (≥ 0.1 mV) at maximal exercise. The two groups were similar in physical characteristics, except the NoSTA group had a significantly higher BMI (Body Mass Index), were a few years younger and exhibited a lower RPP at maximal effort. Discriminant Function Analysis was used to predict group classification of individual patients (STA or NoSTA). Based on predictions using physical characteristics alone, (age, BMI, TC), age, BMI and TC (Total Blood Cholesterol) could correctly predicted classification in 66% of the cases. The set of age, BMI, TC and ΔBP/METS3 (low to high) generated a prediction with 77% correct classification. Thus, ΔBP/MET level alone was not the primary variable to explain predictive accuracy for clinically important ST changes in exercise testing. However, in accordance with the Bayesian principle, this hemodynamic exercise response is adjusted for overall metabolic demand in the test and coupled to markers of pre-test coronary risk, the ability to predict ST response is improved.