A pilot study to assess the utility of five established variables to standardize exercise treadmill test reporting

Ashok Krishnaswami, William K.W. Ho, Walter P. Kwan, Christine Tsou, Jamal S. Rana, Matthew D. Solomon, Sheng-Fang Jiang, James J. Jang, Thomas Alloggiamento, Andrew W. Praserthdam

The prognostic utility of 5 established variables (functional capacity, Duke treadmill score, chronotropic response to exercise, heart rate recovery, and premature ventricular contractions) together after routine exercise treadmill testing (ETT) has not been determined. We assessed the combined prognostic ability of 5 established variables for the primary outcome (myocardial infarction [MI], coronary revascularization [CR] or all-cause mortality) and the secondary outcome of unnecessary downstream testing (defined as receipt of further noninvasive imaging without CR, MI, or death) compared with standard methods. Using a retrospective study design, 1857 consecutive patients were enrolled in the year 2014 and followed until December 31, 2015. Optimal discrimination and global fit statistics were assessed from logistic regression models. Classification and regression tree (CART) methodology was used for the final model. The mean [SD] age was 56.0 [12.5]years; median comorbidities (2, IQR 2) with 26% having an equivocal report. Compared to other models, a model with age, sex, and the 5 established variables showed an improvement in discrimination for the primary [c-statistic 0.85 versus (0.69-0.79)] and secondary [c-statistic 0.73 versus (0.65-0.71)] outcomes with substantial improvement in global fit. The final, optimal, 10-fold cross-validated CART model had a c-statistic of 0.78. The utility of the 5-established variables, based on the current study, resides in its ability to decrease unnecessary downstream testing and improve cardiovascular event prognostication. This is accomplished by removing the subjective interpretation of currently used ETT variables that can lead to an equivocal report.