Prediction of portal and hepatic blood flow from intake level data in cattle.

Abstract

Interest is growing in developing integrated postabsorptive metabolism models for dairy cattle. An integral part of linking a multi-organ postabsorptive model is the prediction of nutrient fluxes between organs, and thus blood flow. The purpose of this paper was to use a multivariate meta-analysis approach to model portal blood flow (PORBF) and hepatic venous blood flow (HEPBF) simultaneously, with evaluation of hepatic arterial blood flow (ARTBF; ARTBF=HEPBF - PORBF) and PORBF/HEPBF (%) as calculated values. The database used to develop equations consisted of 296 individual animal observations (lactating and dry dairy cows and beef cattle) and 55 treatments from 17 studies, and a separate evaluation database consisted of 34 treatment means (lactating dairy cows and beef cattle) from 9 studies obtained from the literature. Both databases had information on dry matter intake (DMI), metabolizable energy intake (MEI), body weight, and a basic description of the diet including crude protein intake and forage proportion of the diet (FP; %). Blood flow (L/h or L/kg of BW0.75/h) and either DMI or MEI (g or MJ/d or g or MJ/kg of BW0.75/d) were examined with linear and quadratic fits. Equations were developed using cow within experiment and experiment as random effects, and blood flow location as a repeated effect. Upon evaluation with the evaluation database, equations based on DMI typically resulted in lower root mean square prediction errors, expressed as a % of the observed mean (rMSPE%) and higher concordance correlation coefficient (CCC) values than equations based on MEI. Quadratic equation terms were frequently nonsignificant, and the quadratic equations did not outperform their linear counterparts. The best performing blood flow equations were PORBF (L/h)=202 (±45.6) + 83.6 (±3.11) × DMI (kg/d) and HEPBF (L/h)=186 (±45.4) + 103.8 (±3.10) × DMI (kg/d), with rMSPE% values of 17.5 and 16.6 and CCC values of 0.93 and 0.94, respectively. The residuals (predicted - observed) for PORBF/HEPBF were significantly related to the forage % of the diet, and thus equations for PORBF and HEPBF based on forage and concentrate DMI were developed: PORBF (L/h)=210 (±51.0) + 82.9 (±6.43) × forage (kg of DM/d) + 82.9 (±6.04) × concentrate (kg of DM/d), and HEPBF (L/h)=184 (±50.6) + 92.6 (±6.28) × forage (kg of DM/d) + 114.2 (±5.88) × concentrate (kg of DM/d), where rMSPE% values were 17.5 and 17.6 and CCC values were 0.93 and 0.94, respectively. Division of DMI into forage and concentrate fractions improved the joint Bayesian information criterion value for PORBF and HEPBF (Bayesian information criterion=6,512 vs. 7,303), as well as slightly improved the rMSPE and CCC for ARTBF and PORBF/HEPBF. This was despite minimal changes in PORBF and HEPBF predictions. Developed equations predicted blood flow well and can easily be used within a postabsorptive model of nutrient metabolism. Results also suggest different sensitivity of PORBF and HEPBF to the composition of DMI, and accounting for this difference resulted in improved ARTBF predictions.