Prediction equations are often applied to estimate the resting energy expenditure (REE) of end-stage renal disease patients. The accuracy of these equations in elderly on hemodialysis (HD) has not been investigated. We aimed to evaluate the agreement between the REE obtained by indirect calorimetry and three prediction equations in elderly patients on HD.The REE of 72 elderly individuals (57 on HD and 15 with normal renal function), was measured by indirect calorimetry and compared to the prediction equations of Harris & Benedict, Schofield and the World Health Organization 1985 (WHO). The agreement was assessed by the intraclass correlation coefficient (ICC) and by Bland–Altman plot analysis. The ratio predicted/measured REE was used for classifying the agreement as acceptable (ratio: 0.90 to 1.10) or overestimated (ratio: >1.11).The REE estimated by the three equations was significantly higher than that obtained by indirect calorimetry (Harris & Benedict: 1339 ± 245 kcal/day; Schofield: 1358 ± 203 kcal/day; WHO: 1385 ± 225 kcal/day vs. indirect calorimetry: 1245 ± 283 kcal/day; P < 0.05). The ICC was indicative of moderate agreement between indirect calorimetry and the three equations (Harris & Benedict: r = 0.70 (95% confidence interval: 0.54; 0.81); Schofield: r = 0.64 (0.46; 0.77) and WHO: r = 0.62 (0.43; 0.75). Acceptable agreement between the equations and indirect calorimetry was observed in 35% of patients, while overestimation was observed in 50%. Similar results were found in the elderly control group.The three equations showed a moderate degree of agreement with indirect calorimetry and overestimation was the main error observed. These results suggest that these equations do not provide accurate measurements of REE in elderly on HD.