Evaluation of Conventional and New Maximum Heart Rate Prediction Models for Individuals

Abstract

The purpose of this study was to develop a regression model to predict maximum heart rate (HRmax) from basic sociodemographic variables and to compare it with the 220-age rule of thumb. Data were obtained from 635 adults of all ages, gender, and physical activity levels, rigorously tested for maximum aerobic capacity. HRmax was found to be significantly correlated (p<.05) to age, tobacco use in the past, current tobacco use, and self-reported physical activity. There was no evidence of a difference in HRmax between males and females (p=.997). Several significant (p<.00005) linear regression models involving these variables were developed, but their ability to explain the variation in HRmax was only slightly better than a model that relied on age alone. Based on R2 values, the age model was able to account for 44.9% of the variation in HRmax, compared to 48% when using the most complicated model. The 220-age rule of thumb also gave an r2 =.449 (44.9%), but the average estimate was biased (-8 beats per minute [b * min~l]). Individual estimates were highly inaccurate: 50.5% of the predicted values were off by 10 b * min-1 or more, compared to 27.6% with our simplest model based on age alone. Furthermore, both the 220-age rule and our regression models were very poor predictors when applied to ten-year age subgroups. It was concluded that in spite of a significant correlation between HRmax and other variables, regression models based on these variables are highly inaccurate in the prediction of individual HRmax values. Therefore, the practice of relying on them for individualized exercise prescription and as a criterion for graded exercise test termination is not warranted.