Previous investigators reported that peak oxygen uptake (VO2peak) could be accurately predicted from nonexercise test variables, and that this score would be suitable for categorizing cardiorespiratory fitness (CRF) within epidemiological studies. However, the accuracy of these models has varied considerably. The purposes of this study were: 1) assess the accuracy of predicting VO2peak with a new nonexercise model, and 2) assess the utility of the predicted VO2peak for categorizing CRF in epidemiological studies. Subjects included 2,350 men and women. Cross-validated multiple regression models revealed that age, sex, resting heart rate, body weight, percentage body fat, smoking, and physical activity were significant predictors (P < 0.001) of VO2peak. The multiple regression model for relative VO2peak (ml.kg-1.min-1) had R2 = 0.733 (SEE = 5.38), whereas the model for absolute VO2peak (l.min-1) had R2 = 0.773 (SEE = 0.425). The 95% confidence intervals for the predicted VO2peak were large (+/- 10.6 ml.kg-1.min-1 and +/- 0.833 l.min-1). These results support the notion that VO2peak can be predicted from a multiple regression model devoid of exercise test variables. However, due to the extreme variability in the predicted scores, the regression models were unable to effectively distinguish CRF categories. Therefore, despite statistical success in predicting VO2peak for the nonexercise test regression models, we conclude that such models fail to provide the accuracy needed for categorizing CRF within large epidemiological cohorts where the purpose is to assess mortality risk.