Prediction of Mortality in Coronary Artery Disease: Role of Machine Learning and Maximal Exercise Capacity.

OBJECTIVE: To develop a prediction model for survival of patients with coronary artery disease (CAD) using health conditions beyond cardiovascular risk factors, including maximal exercise capacity, through the application of machine learning (ML) techniques. METHODS: Analysis of data from a retrospective cohort linking clinical, administrative, and vital status databases from 1995 to 2016 was performed. Inclusion criteria were age 18 years or older, diagnosis of CAD, referral to a cardiac rehabilitation program, and available baseline exercise test results. Primary outcome was death from any cause. Feature selection was performed using supervised and unsupervised ML techniques. The final prognostic model used the survival tree (ST) algorithm. RESULTS: From the cohort of 13,362 patients (60±11 years; 2400 [18%] women), 1577 died during a median follow-up of 8 years (interquartile range, 4 to 13 years), with an estimated survival of 67% up to 21 years. Feature selection revealed age and peak metabolic equivalents (METs) as the features with the greatest importance for mortality prediction. Using these 2 features, the ST generated a long-term prediction with a C-index of 0.729 by splitting patients in 8 clusters with different survival probabilities (P<.001). The ST root node was split by peak METs of 6.15 or less or more than 6.15, and each patient's subgroup was further split by age or other peak METs cut points. CONCLUSION: Applying ML techniques, age and maximal exercise capacity accurately predict mortality in patients with CAD and outperform variables commonly used for decision-making in clinical practice. A novel and simple prognostic model was established, and maximal exercise capacity was further suggested to be one of the most powerful predictors of mortality in CAD.

Comparison of the FRIEND and Wasserman-Hansen Equations in Predicting Outcomes in Heart Failure.

Background Percentage of age-predicted peak oxygen uptake (VO2) achieved (ppVO2) has been widely used to stratify risk in patients with heart failure. However, there are limitations to traditional normal standards. We compared the recently derived FRIEND (Fitness Registry and the Importance of Exercise: A National Data Base) equation to the widely used Wasserman-Hansen (WH) ppVO2 equation to predict outcomes in patients with heart failure. Methods and Results A subgroup of 4055 heart failure patients from the FRIEND registry (mean age 53±15 years) was followed for a mean of 28±16 months. The FRIEND and WH equations along with measured peak VO2 expressed in mL/kg-1 per min-1 were compared for mortality and composite cardiovascular events. ppVO2 was higher for the FRIEND versus the WH equation (66±30% versus 58±25%; P<0.001). The areas under the receiver operating characteristic curves were slightly but significantly higher for the FRIEND equation for mortality (0.74 versus 0.72; P=0.03) and cardiac events (0.70 versus 0.68; P=0.008). Area under the receiver operating characteristic curve for measured peak VO2 was 0.70 (P<0.001) for mortality and 0.73 (P<0.001) for cardiovascular events. For each 1-SD higher ppVO2 for the FRIEND equation, mortality was reduced by 18% (hazard ratio, 0.82; 95% CI, 0.69-0.97; P<0.02); for each 1-SD higher ppVO2 for the WH equation, the mortality was reduced by 17% (hazard ratio, 0.83; 95% CI, 0.71-0.97; P=0.02). The corresponding reductions in risk per 1 SD for cardiovascular events for the FRIEND and WH equations were 23 and 21%, respectively (both P<0.001). Conclusions Peak VO2 expressed as percentage of an age-predicted standard strongly predicts mortality and major cardiovascular events in patients with heart failure. The FRIEND registry equation exhibited test characteristics slightly superior to the commonly used WH equation.

A reference equation for maximal aerobic power for treadmill and cycle ergometer exercise testing: Analysis from the FRIEND registry.

Background Maximal oxygen uptake (VO2max) is a powerful predictor of health outcomes. Valid and portable reference values are integral to interpreting measured VO2max; however, available reference standards lack validation and are specific to exercise mode. This study was undertaken to develop and validate a single equation for normal standards for VO2max for the treadmill or cycle ergometer in men and women. Methods Healthy individuals ( N = 10,881; 67.8% men, 20-85 years) who performed a maximal cardiopulmonary exercise test on either a treadmill or a cycle ergometer were studied. Of these, 7617 and 3264 individuals were randomly selected for development and validation of the equation, respectively. A Brazilian sample (1619 individuals) constituted a second validation cohort. The prediction equation was determined using multiple regression analysis, and comparisons were made with the widely-used Wasserman and European equations. Results Age, sex, weight, height and exercise mode were significant predictors of VO2max. The regression equation was: VO2max (ml kg-1 min-1) = 45.2 - 0.35*Age - 10.9*Sex (male = 1; female = 2) - 0.15*Weight (pounds) + 0.68*Height (inches) - 0.46*Exercise Mode (treadmill = 1; bike = 2) ( R = 0.79, R2 = 0.62, standard error of the estimate = 6.6 ml kg-1 min-1). Percentage predicted VO2max for the US and Brazilian validation cohorts were 102.8% and 95.8%, respectively. The new equation performed better than traditional equations, particularly among women and individuals ≥60 years old. Conclusion A combined equation was developed for normal standards for VO2max for different exercise modes derived from a US national registry. The equation provided a lower average error between measured and predicted VO2max than traditional equations even when applied to an independent cohort. Additional studies are needed to determine its portability.

Influence of age in estimating maximal oxygen uptake.

OBJECTIVE: To assess the influence of age on the error of estimate (EE) of maximal oxygen uptake (VO2max) using sex and population specific-equations in cycle ergometer exercise testing, since estimated VO2 max is associated with a substantial EE, often exceeding 20%, possibly due to intrinsic variability of mechanical efficiency. METHODS: 1850 adults (68% men), aged 18 to 91 years, underwent maximal cycle ergometer cardiopulmonary exercise testing. Cardiorespiratory fitness (CRF) was assessed relative to sex and age [younger (18 to 35 years), middle-aged (36 to 60 years) and older (> 60 years)]. VO2max [mL·(kg·min)(-1)] was directly measured by assessment of gas exchange and estimated using sex and population specific-equations. Measured and estimated values of VO2max and related EE were compared among the three age- and sex-specific groups. RESULTS: Directly measured VO2max of men and women were 29.5 ± 10.5 mL·(kg·min)(-1) and 24.2 ± 9.0 mL·(kg·min)(-1) (P < 0.01). EE [mL·(kg·min)(-1)] and percent errors (%E) for men and women had similar values, 0.5 ± 3.2 and 0.4 ± 2.9 mL·(kg·min)(-1), and -0.8 ± 13.1% and -1.7 ± 15.4% (P > 0.05), respectively. EE and %E for each age-group were, respectively, for men: younger = 1.9 ± 4.1 mL·(kg·min)(-1) and 3.8 ± 10.5%, middle-aged = 0.6 ± 3.1 mL·(kg·min)(-1) and 0.4 ± 10.3%, older = -0.2 ± 2.7 mL·(kg·min)(-1) and -4.2 ± 16.6% (P < 0.01); and for women: younger = 1.2 ± 3.1 mL·(kg·min)(-1) and 2.7 ± 10.0%, middle-aged = 0.7 ± 2.8 mL·(kg·min)(-1) and 0.5 ± 11.1%, older = -0.8 ± 2.3 mL·(kg·min)(-1) and -9.5 ± 22.4% (P < 0.01). CONCLUSION: VO2max were underestimated in younger age-groups and were overestimated in older age groups. Age significantly influences the magnitude of the EE of VO2max in both men and women and should be considered when CRF is estimated using population specific equations, rather than directly measured.