The objective was to determine digestible fraction values of several tropical forages used in ruminant diets in the tropics through in vivo digestion in order to estimate the metabolizable energy (ME) content of these forages using several methods and predictive equations, and to assess the relationship between in situ neutral detergent fibre (NDF) degradability and in vivo NDF digestibility. In the experiment, 28 confined Nellore heifers 293±46.0kg live weight, were fed chopped sugar cane, sugar cane silage, soybean silage, mombaça silage (Pannicum maximum cv. Mombaça), corn silage, Tifton-85 hay (Cynodon spp.) and chopped elephant grass (Penissetum purpureum cv. Cameroun). Forages with crude protein (CP) below 70g/kg DM were supplemented with urea and ammonia sulphate (9:1). The feeding period lasted 12 days, consisting of 7 days for adaptation and 5 for total fecal collection. The ME estimated from NRC (2001) was based upon chemical component fractions and the same system using digestible NDF (NDF d ) after 48h of in vitro incubation (NRC 48 ) was examined. The UC Davis (UCD) approaches based on several chemical components and an in vitro NDF digestion at 30 and 48h of incubation, respectively, UCD 30 and UCD 48 , were evaluated. The other UCD approach (UCD Gas), which predicts ME from in vitro gas produced after 24h (24h Gas), 48h (48h Gas) and 72h (72h Gas) of incubation, was also examined. The equations developed under tropical conditions by Detmann also calculate digestible energy from the apparent digestible fractions of crude protein (CP ad ), ether extract (EE ad ), non-fibre carbohydrates (NFC ad ) and neutral detergent fibre (NDFad) (Detmann 1). The equations proposed in Detmann 1 were used again, but the estimator was changed to apparently digestible CP (CP ad ) (i.e., Detmann 2). The poor results associated with the descriptive levels of probability compared to the null hypotheses show the inadequacy of the UCD 30 , NRC 48 , 24h Gas and UCD gas models in predicting ME of these forages. The highest accuracy was with the NRC, Detmann 1 and Detmann 2 models as reflected in the lowest estimates of MSPE (6.297; 6.274; 6.236, respectively) and the lowest bias (0.458; 0.040; 0.162, respectively) among all models evaluated. For the CP ad , a similarity was verified between the estimates with the submodel proposed by the NRC (2001), as well as by the Detmann 2 submodel and in vivo observations. The submodels proposed by the NRC (2001) and Detmann diverged from the observed values of digested levels of EE. Similarities between observed and estimated values in the NDF d diet levels occurred. However, a higher predictive adequacy occurred in the Detmann 1 submodel. The same behavior occurred in estimates obtained for digestible fractions of NFC, suggesting that both models satisfactorily estimate the NFC ad , with a higher prediction adequacy associated with the lowest bias in the Detmann 1 submodel. The systems based upon chemical composition, such as NRC (2001) and Detmann allowed accurate estimates of ME but higher precision and accuracy occurred in the Detmann submodels for predicting the ME value of tropical feeds.