One study of high research quality design reported that the Mifflin-St. Jeor equation accurately predicted RMR within +/- 10% of measured RMR in 82% of non-obese adults. Of the remaining 18% errors, 10% were overestimations and 8% were underestimations. The individual error range was a maximum overestimate by 15% and underestimate by 18%, indicating a more even distribution and narrower variation (i.e., five and eight percent outside of the defined +/- 10% measure).