The prediction of the energy content in commercially prepared dog and cat foods was investigated using three databases [database A: 128 dog digestion trials and food analyses with fibre analysed as crude fibre (CF) including literature data; database B: 107 dog digestion trials, fibre analysed as total fibre (TF) after ENGLYST and CUMMINGS (1988) and database C: 58 cat digestion trials, fibre analysed as CF]. Multiple regression equations with the dependent variable digestible energy (DE) in kJ/100 g dry matter and the proximates including fibre in percentage dry matter as independent variables were calculated. A second approach was based on the prediction of digestibility of energy (ad GE;%) from fibre content in percentage dry matter (for database A: ad GE=91.2 - 1.43 x CF; for database B: ad GE = 93.3 - 1.42 x TF and for database C: ad GE = 87.9 - 0.88 x CF). The DE was then calculated: DE = gross energy(GE) x ad GE/100. The GE in kJ/100 g dry matter was either calculated or analysed by adiabatic bomb calorimetry. The results of all methods correlated significantly with the experimental data (r greater than or equal to 0.88**). The most important source of error was the calculator of GE. The most accurate method for estimation of DE is the measurement of GE by bomb calorimetry and the estimation of ad GE from fibre content. Metabolizable energy (ME) can be estimated by N-correction For dog food 4.34 kJ/g protein and for cat food 3.10 kJ/g protein can be subtracted. A comparison of the correlation between ME as calculated by, the authors' recommended method, or by the NRC equation and the experimentally determined ME in a dataset with 61 prepared dog foods which were not used to develop the method, showed a coefficient of correlation of 0.95** (method of this study) or 0.56** (NRC equation).