Published predictive equations overestimate measured resting metabolic rate in young, healthy females

To compare resting metabolic rate (RMR) measured by indirect calorimetry versus RMR predicted by several published formulas in a sample of healthy young women.

RMR was measured using indirect calorimetry and predicted using 6 commonly used equations (Nelson, 1992; Mifflin, 1990; Owen, 1986; Schofield(Weight), 1985; Schofield(Weight and Height), 1985; Harris-Benedict, 1919) in 47 reportedly healthy young females (age = 22.8 ± 2.9 years; body mass index = 21.8 ± 2.1 kg/m(2)). Comparisons between measured versus predicted RMR were conducted using paired t tests, and agreement using Pearson's correlation coefficient, analysis of variance, and the method of Bland-Altman.

All 6 equations overestimated measured RMR by 140-738 kcal/d (all p < 0.001). The proportion of subjects for whom measured versus predicted RMR differed by ±10% ranged from 74% (Nelson) to 100% (Harris-Benedict). The adjusted coefficients of determination (R(2)) between measured and predicted RMR ranged from 0.13 to 0.19 (all p < 0.05). Bland-Altman analysis R(2) values ranged from 0.03 (p = 0.233; Harris-Benedict) to 0.72 (p = 0.000; Owen). Given its continued popularity, we modified the Harris-Benedict equation (RMR(modified Harris-Benedict) (kcal/d) = 738 / (RMR(Harris-Benedict) - 738)). Doing so reduced the mean difference between measured and predicted RMR from +738 kcal/d to -0.53 kcal/d (p = 0.984).

No equation performed well, and none should be used interchangeably with measured RMR. We recommend that a new equation be validated for, and prospectively tested in, young women. In the interim, RMR should be measured in this population or predicted using the modified Harris-Benedict equation that we developed.