Data from 655 treatments of 116 studies were used in a meta-analysis to determine the daily digestible energy (DE), metabolizable energy (ME) and net energy (NE) intake of Chinese growing-finishing pigs, and to predict feed efficiency responses to change in dietary DE, ME and NE. Three alternative functions (i.e., polynomial, Bridges and asymptotic function) were employed for fitting daily DE, ME or NE intakes to mean body weight. The results showed that the three models from the current study provided reasonable fit (all R2 > 0.83) for the energy intake data. However, under the same energy system, the polynomial function had the smallest Akaike's information criteria (AIC) and residual standard deviation (RSD), followed by Bridges and asymptotic functions. The three model-generated energy intakes of growing pigs were significantly less than that of the Chinese Feeding Standard of Swine, but similar to that of the National Research Council (2012), while the values of finishing pigs were greater than both standards. Compared with those that predict feed efficiency based on DE or ME, the equation with NE as a predictor had the minimized AIC and RSD. It was also found that feed efficiency increased with increasing dietary energy density (DED), but this response varied with pig body weight, and the lighter pigs were more sensitive to DED than heavier pigs.