New models were developed for predicting the resting metabolic rate (RMR) w
ith sufficient accuracy for use in epidemiologic studies and for weight con
trol of individuals. For this purpose, the RMR of 213 women and 76 men was
measured, and physical measurements were taken. The RMR was regressed on co
rrelates of RMR, avoiding harmful degrees of collinearity by rejecting inte
rregressor correlations exceeding r = 0.5. For women, the best model (R-2 =
0.71) included the regressors age, race, weight, pulse rate, smoking, and
body temperature. The best model for males (R-2 = 0.81) included age, race,
weight, blood pressure, smoking, time (of day the RMR was measured), and w
hether subjects had a meal prior to calorimetry. The models were cross vali
dated internally and also validated using an external database. In both cas
es, die mean estimated RMR did not differ significantly from the measured R
MR. The accuracy of the models was compared with four models reported in th
e literature, three of which overestimated the RMR by up to 17%. In conclus
ion, improved RMR prediction models have been developed, more accurate than
existing models, rendering them suitable for application to epidemiologica
l databases and for individual weight control programs.