Our objective was to predict composition (% fat, % protein, % ash, and % water) as % of empty body weight (EBW) in heifers and cows using only data from Holsteins. We conducted mixed regressions analysis on 129 treatment means from 24 peer-reviewed publications. Methods to determine composition were direct analysis of EBW, analysis of carcass or rib, and dilution of deuterium oxide or urea. Means and standard deviations were 10 ± 5% fat, 19 ± 2% protein, 4 ± 1% ash, 66 ± 4% water, 118 ± 81 kg EBW, and 0.74 ± 0.25 kg daily gain (ADG) for growing animals and 20 ± 6% fat, 17 ± 2% protein, 5 ± 0.7% ash, 60 ± 3% water, 493 ± 64 kg EBW, and 2.9 ± 0.4 body condition score (BCS) for mature animals. We first used a mixed model with random effects of study and method to adjust dependent variables based on EBW, ADG, BCS, and possible 2-way interactions, with studies weighted by inverse of their variance. Second, adjusted compositions were modeled as functions of EBW, the natural logarithm of EBW (lnEBW), or EBW to the power of 0.75 (MEBW) to generate a base BW response. Finally, we examined effects of including ADG for heifers and BCS for cows. Cross-validation within the data set was used to select best fitting models based on concordance correlation coefficient (CCC), root means square error of prediction (RMSEP), and its decomposition. Including ADG for heifers and BCS for cows improved the overall prediction of body composition, and using lnEBW improved predictions of EBFat and EBWater. The proposed heifer models were: % Fat = −7.8 + 2.41 × lnEBW − 7.96 × ADG + 3.13 × lnEBW × ADG (CCC = 0.88, RMSEP = 1.77); % Protein = 27.5 − 1.62 × lnEBW − 1.20 × ADG (CCC = 0.43, RMSEP = 1.43); % Ash = 5.1 + 0.091 × lnEBW − 1.85 × ADG (CCC = 0.36, RMSEP = 0.79); and % Water = 90 − 3.41 × lnEBW + 0.990 × ADG − 1.74 × lnEBW × ADG (CCC = 0.69, RMSEP = 1.60). The proposed cow models were: % Fat = −24 + 6.43 × lnEBW + 1.82 × BCS (CCC = 0.54, RMSEP = 2.55); % Protein = 27 − 1.62 × lnEBW − 0.451 × BCS (CCC = 0.37, RMSEP = 0.98); % Ash = 5 + 0.091 × lnEBW − 0.340 × BCS (CCC = 0.30, RMSEP = 0.27); and % Water = 92.9 − 5.15 × lnEBW − 0.973 × BCS (CCC = 0.36, RMSEP = 2.34). These models could be further used to determine composition of BW gain.