Purpose The present study aims a) to assess the agreement between the measured resting metabolic rate (RMR) using indirect calorimetry and different predictive equations (predicted RMR), and b) to propose and cross-validate two new predictive equations for estimating the RMR in high-level athletes. Methods The RMR of 102 athletes (44 women) was assessed using indirect calorimetry, whereas the body composition was assessed using skinfolds. Comparisons between measured and predicted RMR values were performed using one-way ANOVA. Mean difference, root mean square error (RMSE), simple linear regression, and Bland-Altman plots were used to evaluate the agreement between measured and predicted RMR. The accuracy of predictive equations was analyzed using narrower and wider accuracy limits (+/- 5% and +/- 10%, respectively) of measured RMR. Multiple linear regression models were employed to develop the new predictive equations based on traditional predictors (equation 1) and the stepwise method (equation 2). Results The new equations 1 and 2 presented good agreement based on the mean difference (3 and -15 kcal center dot d(-1)), RMSE (200 and 192 kcal center dot d(-1)), and R-2 (0.71 and 0.74), respectively, and accuracy (61% of subjects between the limit of +/- 10% of measured RMR). Cunningham's equation provided the best performance for males and females among the existing equations, whereas Jagim's equation showed the worst performance for males (mean difference = -335 kcal center dot d(-1); RMSE = 386 kcal center dot d(-1)). Compared with measured RMR, most predictive equations showed heteroscedastic distribution (linear regression's intercept and slope significantly different from zero; P <= 0.05), mainly in males. Conclusions The new proposed equations can estimate the RMR in high-level athletes accurately. Cunningham's equation is a good option from existing equations, and Jagim's equation should not be used in high-level male athletes.