In vivo estimation of body composition of mature gilts using live weight, backfat thickness, and deuterium oxide.

Thirty-seven Duroc x (Yorkshire x Landrace) (DYL) and 21 Yorkshire x Landrace (YL) gilts were used to develop equations that predict body composition of replacement-age breeding swine. Before slaughter, gilts were weighed, ultrasonically scanned for 10th rib backfat thickness, and infused with D2O (.25 g/kg live weight). The D2O space (kilograms) was calculated from body water D2O concentration determined at equilibrium (150 and 210 min after infusion). Regression models predicting empty body (Eb) components for DYL and YL groups were fitted using all possible variable combinations (D2O space, live weight, and[or] backfat depth). Variables selected in best-fit models for Eb water, protein, fat, and ash for data from DYL gilts differed from variables selected from data from YL gilts. Average prediction errors (kilograms; [predicted residual sum of squares divided by n]1/2) of best-fit equations were 2.37, 2.03 (Eb weight), 2.36, 1.66 (Eb water), 1.07, .47 (Eb protein), and 2.76, 2.89 (Eb fat) for DYL and YL data sets, respectively. Cross-validation by applying DYL equations to YL data, and vice versa, resulted in larger prediction errors. Likewise, larger errors were obtained when equations published elsewhere were applied to DYL and YL data sets. No cited source provided a set of equations that consistently minimized prediction errors of all Eb components of both DYL and YL gilts. Results indicate that prediction equations using D2O space, live weight, and(or) backfat thickness are accurate in estimating body composition only in animals physiologically resemble the population in which the equations were derived.