Predicting feed intake and feed efficiency in lactating dairy cows using digesta marker techniques

Abstract

Direc t mea surement of individual animal dry ma tter intake (DMI) remains a funda mental challeng e to assessing dairy feed ef ficiency (FE) . Digest a marker, is currently the most used indirect technique for estimating DMI in production animals. In this meta -analysis we evaluate d the performan ce of ma rker-based estima t es a g ain st direct or ob se r v ed m e a su reme n t s and d e v elope d equatio ns for the prediction of FE (g energy-c orrected milk (ECM)/kg DMI). Data were taken from 29 change-over studies co n sis ting of 416 cow-wit hin period observ ations. Most studies used more than one digesta marker. So, fo r each observed meas urem ent o f DMI, faec al dry matter output (FDM O) an d app arent total tract dry matter dig estibility (DMD), there was one or more correspo nding mark er estimate. There were 924, 409 and 846 observ ations for es timated FDMO (eFDMO) , estim ated appa r en t total tract DMD (eDMD) and estimate d DMI (eDMI), respectively. Th e ex pe rim en t al die ts we re bas ed m ainly on g r a s s sila g e, with soya bean or rape seed meal as prot ein supplemen ts and cereal grains or by-produ cts as energy suppleme nts. Acro ss all diets , average forage to concen trate ratio on a dry matter (DM) basis w as 59 : 41. Va riance compon ent and repeatability estima t es of observ ed and ma rker estimation s w ere deter mined using random factors in mixed procedur es of SAS. Between-c ow CV in obser ve d FDMO, DMD and DMI was, 10.3, 1.69 and 8.04, respectively. Overall, the repea tability estim ates of observ ed variables were greater than their correspo nding mark er-based estimates of repea tability . Reg ression of observed meas urement s on ma rker-bas ed estim ates gav e good relationships ( R 2 = 0.87, 0.68, 0.74 and 0.74 , relative prediction error = 10.9%, 6.5%, 15.4% and 18. 7% for FDMO, DMD, DMI and FE prediction s, respectively). Despite this, the mean and slope biases we re sta tistically signi ficant ( P < 0.001) for all regressions . Mor e than half of the errors in all regres sions were due to mean and slop e biases (5 2.4% 87 . 4%, 82.9% and 85.8% for FDMO, DMD , DMI an d FE, respectively), w hereas the contribu tions of random erro rs were small. Base d on residual varianc e, the best mod el for predicting FE develope d fr om t h e dat a se t was FE (g EC M /k g DMI) = 1179(±54.1) +38.2 (±2.05) × ECM(kg/day) − 0.64(±0.051) × BW (kg) − 75.6(±4.39) × eFDMO (kg/day). Although eDMD w as positiv ely related t o F E , it only showed a tenden cy to reduce the resid ual varianc e. Despite inac curacy in mark er procedur es, eFD MO from external m a r ke rs provided a reliable determina tion for FE mea surement. Ho wever, DMD estim ated by internal markers did not improve predictio n o f FE, probably re flecting small variabilit y.