Abstract Background Abnormal results of myocardial perfusion SPECT (MPS) or exercise ECG (Ex-ECG) is known to be associated with increased risk of ischemic cardiac events. The predictive value of stress testing however, is influenced by underlying risk factors as well as follow up duration and may vary considerably over time. With traditional Cox proportional hazards analysis the time-varying effects are not sufficiently accounted for. Purpose To evaluate the prognostic longevity of MPS and Ex-ECG using statistical methods that allow hazard ratio changes over time. Methods 908 patients (age 63 years, 49% male, 45% prior ischemic heart disease (IHD)) referred for MPS were prospectively included. Ex-ECG was performed in conjunction with MPS. Registry data were used for follow up. The outcome was a composite of acute myocardial infarction, unstable angina, unplanned revascularization, and cardiovascular death. Cox proportional hazard analysis and smoothed hazard functions with splines were used for risk assessment during follow up. Results Ex-ECG or MPS were abnormal in 198 (22%) or 148 (16%) of patients respectively. The mean follow up time was 8.2±2.6 years and the composite end point was reached in 189 patients (21%). The Kaplan-Meier curves suggest an interaction with time with the risk of event for subjects with abnormal test results being highest in the 2-3 of years after which the curves are approximately parallel (Figure 1). In multivariable models the best fit (lowest Akaike information criterion (AIC) value) was obtained with a logarithmic interaction term for time. With this approach and controlling for known risk factors the hazard ratio (HR) drops to 1 after approximately four years for both Ex-ECG and MPS (Figure 2). Age (HR=1.04, p<0.001), sex (HR=2.0, p<0.001), IHD (HR=1.7, p<0.001) and reduced exercise capacity (HR=0.99, p<0.001) were independent significant predictors of risk. Conclusions The high risk associated with abnormal MPS and Ex-ECG results decrease rapidly over time and after approximately 4 years the predictive value is lost. Thus, the long-term impact on prognosis of stress testing appear questionable. The time-varying effects on risk are substantial and warrant statistical considerations when designing clinical follow up studies of chronic coronary syndrome.
Read full abstract