The objectives of the present study were to develop and cross-validate new equations for predicting resting energy expenditure (PREE) in severely obese Italian males, and to compare their accuracy with those of the Harris-Benedict, WHO/ FAO/UNU, Huang, Owen, Mifflin, Livingston, Nelson, Bernstein, and Cunnimgham equations in order to predict resting energy expenditure (REE), using the Bland-Altman method. One hundred and sixty-four severely obese males [mean body mass index (BMI): 45.4 kg/m2; 50.2% fat mass), aged 20 to 65 yr participated in this study. REE was measured by indirect calorimetry and body composition by bioelectrical analysis. Equations were derived by stepwise multiple regression analysis using a calibration group and tested against the validation group. Two new specific equations, based on anthropometric [REE=Weight x 0.048 + Height x 4.655 - age x 0.020 - 3.605 (R2=0.68, SE=1.14 MJ/d)] or body composition parameters [REE=fat free mass (FFM) x 0.081 + fat mass (FM) x 0.049 - age x 0.019 - 2.194 (R2=0.65, SE=1.15 MJ/d)], were generated. Mean PREE were not different from the mean measured REE (MREE) (<1%, p<0.001), REE being predicted accurately (95-105% of MREE) in 66 and 62% of subjects, respectively. The Harris-Benedict, WHO/FAO/UNU, Huang and Owen equations showed mean differences lower than 5% and PREE was accurate in less than 30% of subjects. The Mifflin, Livingston, and Nelson equations showed a mean PREE underestimation >7% (p<0.001) and PREE was accurate in less than 25% of subjects. The Bernstein and Cunningham equations showed a greater PREE underestimation (>22%, p<0.001) in more than 85% of subjects. The new prediction equations allow an accurate estimation of REE in groups of severely obese males and result in lower mean differences and lower limits of agreement between PREE and MREE than the commonly used equations.