Prediction of microbial crude protein flow from the rumen of dairy cattle by means of dietary characteristics - a meta-analysis.

Protein supply to ruminants relies mainly on the flow of microbial crude protein (MCP) from the rumen, which is commonly assumed to primarily depend on energy supply. This study evaluated this assumption with recent data and tested if ruminally fermented organic matter (FOM) was a better predictor of MCP flow than total-tract digestible organic matter (DOM) and if more variables could improve the prediction of MCP flow. A previously published data set was extended by additional studies resulting in a data set of 139 studies including 407 treatment means, typical to Central European rations. Either DOM or FOM had to be reported and estimates of MCP were all based on gastrointestinal measurements. Dietary treatments were restricted to a maximum concentrate proportion of 0.6 and a CP concentration from 10% to 20% of dietary dry matter (DM). Treatments with more than 1% of animal by-products were excluded. Mixed models with "study" as random effect were used to test the ratios between MCP and DOM or FOM. Dietary characteristics including DM intake, OM digestibility (OMD), OM fermentability (OMF), and ruminal N balance and dietary concentrations of CP, rumen-degraded CP (RDP), rumen-undegraded CP (RUP), neutral detergent fibre, starch, and ether extract were evaluated as additional variables to improve prediction accuracy. Regression of MCP flow against DOM ( = 324) or FOM ( = 349) revealed estimates of 139 g ± 36 g and 164 g ± 40 g of MCP for each kg of DOM or FOM, respectively. The best-fitting mixed model estimating MCP/DOM [g/kg] was 312-2.75 · OMD - 0.229 · RUP + 0.290 · RDP whereas MCP/FOM [g/kg] may be calculated as 294-2.80 · OMF + 0.401 · RDP, with OMD and OMF in [%] and RDP and RUP in [g/kg DM]. Microbial CP flow [g/d] was most accurately described as 1896 + 110 · DOM - 23.7 · OMD - 0.164 · RUP + 0.153 · RDP, with DOM in [kg], OMD in [%] and RDP and RUP in [g/d]. Predictions based on DOM were at least as precise as FOM-based predictions and comparison with published models revealed that the presented models were of similar accuracy as the ones from literature.