Developing and Evaluating Relationships of Diet Characteristics with Visceral Organ Mass in Cattle

Simple Summary

Visceral organs such as the heart, liver, kidneys, and gastrointestinal tract account for approximately 40 to 50% of the energy expenditure in cattle. The size of the visceral organs changes in response to animal, management, and dietary characteristics, but previous research has focused on a narrow range of organs and diet types. In this study, results from 38 experiments were combined into a single dataset, and statistical models were used to describe the relationship of organ size with animal, management, and diet characteristics. The models indicated that feed intake, the length of the feeding period, and the amount and physical form of dietary fiber are the main factors linked to changes in organ size, with dietary protein and energy concentrations playing smaller roles overall. For most organs, predicted values closely matched reported values. These equations can be used to improve estimates of the energy cattle need for maintenance under different feeding programs, supporting more precise ration formulation, increased efficiency, and more sustainable beef production.

Abstract

Visceral organ mass is a major determinant of maintenance energy requirements in cattle, suggesting that equations to predict visceral organ mass could increase the accuracy of estimates of energy requirements. The objective of this meta-analysis was to quantify the relationships of visceral organ mass with the chemical composition of the diet, as well as animal and management characteristics. A database of 170 treatment means from 38 studies was assembled from published literature. Mixed-effects models with animal, management, and diet characteristics as fixed effects and study as a random effect were selected based on the lowest corrected Akaike information criterion (AICc) and evaluated via leave-one-trial-out cross-validation. Out of 16 organs, 15 had concordance correlation coefficient (CCC) values over 0.900, and cross-validated coefficient of determination (R²) values ranged from 0.728 to 0.967 across organs. Dry-matter intake, days on feed, and fiber-related diet characteristics (roughage level, neutral detergent fiber, and physically effective neutral detergent fiber) were the most consistently retained predictors, with crude protein and metabolizable energy concentrations being retained less frequently. These equations provide a quantitative basis for more accurate estimation of visceral organ mass in cattle.

1. Introduction

Visceral organs such as the stomach complex, intestines, and liver account for approximately 40 to 50% of whole-animal energy expenditure in cattle, despite representing a much smaller fraction of body mass [1,2,3]. Differences in the size and metabolic activity of these organs are recognized as important contributors to variation in maintenance energy requirements and feed efficiency across cattle types (e.g., beef vs. dairy), diet compositions (e.g., forage vs. concentrate-based), and production systems (e.g., grazing vs. feedlot). Because visceral organ mass is not routinely measured outside of research settings, energy requirement systems such as the NASEM Nutrient Requirements of Beef Cattle [4] and the Cornell Net Carbohydrate and Protein System [3] use a single factor for net energy for maintenance, assuming a constant contribution of visceral organs to maintenance energy requirements across all classes of cattle and diet compositions. Therefore, a quantitative understanding of how visceral organ mass changes with diet composition and management could improve the accuracy of maintenance energy estimates in these systems. These equations could support ration formulation and improve scientist and producer understanding of how diet and management affect visceral organ mass.

Visceral organ mass is influenced by both the amount of feed consumed and the chemical and physical composition of the diet. Increases in feed intake tend to enlarge the reticulorumen and intestines [5,6,7]. Independent of feed intake, changes in time on feed, dietary fiber concentration and physical form, and protein and energy concentrations can shift the size of organs such as the liver, stomach complex, and intestines [1,7,8,9]. Estimates of how strongly dietary and management factors influence visceral organ mass vary across the literature, reflecting differences in sample size, as the range of cattle types and diets evaluated in each study. Most available equations describing these relationships are derived from individual experiments with limited experimental scope, often covering only a few organs or diet types or a single class of cattle.

Integrating data across experiments offers an opportunity to describe relationships of visceral organs with dietary, animal, and management characteristics over a broader range of conditions than is possible within a single trial. The objectives of this study were: (1) to quantify the relationships between dietary, animal, and management characteristics and visceral organ mass for a broad set of visceral organs and (2) to develop and validate mixed-effects prediction equations for these relationships. The hypothesis of this study was that dry-matter intake and dietary characteristics would be positively associated with the mass of organs and that, together, these predictors would account for the majority of the variation in visceral organ mass.

2. Materials and Methods

This analysis was based on data from published literature and involved no live animals; thus, no Institutional Animal Care and Use Committee approval was required.

2.1. Database Creation

A literature search was conducted to identify experiments measuring visceral organ masses and diet composition. Three databases were searched: Google Scholar, PubMed, and AGRICOLA. Multiple combinations of the search terms “cattle”, “cow”, “ruminants”, “visceral organs”, “viscera”, “protein”, and “fat” were used, and all searches were limited to records published on or before January 2025. Google Scholar returned 247 records, PubMed returned 56 records, and AGRICOLA returned 144 records. Duplicate records were removed. Records were screened in three stages: first by title for relevance, second by abstract for measurement of body composition, and third by data tables for the availability of organ weights and diet characteristics. To be included, a record had to report at least one visceral organ mass and either empty body weight or hot carcass weight. Records on post-wean steers, heifers, bulls, and cows were included. Pre-wean calves were excluded because their primary diet is milk, and the dataset lacks the information needed to characterize these diets within the modeling framework used in this study.

Studies needed to provide data on visceral organ masses and empty body weight. Not all studies provided data on all visceral organs, but any available data was recorded. If a study did not provide empty body weight but instead provided hot carcass weight, empty body weight was calculated using the equation of Garrett and Hinman [10]: EBW = (1.362 × HCW) + 30.26, where HCW is the hot carcass weight in kilograms. The equation of Garrett and Hinman [10] was chosen because it has been used extensively [3,4] and was recently reported to be highly accurate and precise [11]. Because EBW was estimated from HCW for these studies rather than measured directly, additional uncertainty is introduced into the EBW values, which may propagate into the performance of models that retained EBW as a predictor.

For each included study, feed ingredient composition, the chemical composition of the diet fed prior to harvest, dry-matter intake during the feeding period, and length of time the diet was fed were recorded. When the chemical composition of the diet was not directly reported but the ingredient composition was available, the chemical composition was calculated using values from the Nutrient Requirements of Beef Cattle [4]. Because physically effective NDF could be a factor affecting gastrointestinal tract mass [9], dietary physically effective NDF (peNDF) was calculated for each diet. The eNDF percentage was obtained from the Nutrient Requirements of Beef Cattle [4] and multiplied by the ingredient composition of the reported diet. The term “eNDF” used in NRC tables is treated here as equivalent to peNDF. Because the measured particle size was not available, the resulting peNDF values are approximations and may not fully reflect the physical effectiveness of fiber in the diet.

The final dataset (Supplementary_Dataset.xlsx) contained 170 observations with days on feed (DOF), roughage (% of diet), type of forage, metabolizable energy concentration (MEC, Mcal/kg DM), crude protein concentration (CP, %), neutral detergent fiber concentration (NDF, %), physically effective neutral detergent fiber concentration (peNDF, %), dry-matter intake (DMI, kg/d), cattle breed, cattle sex, empty body weight (EBW, kg), initial weight at start of the feeding period (IW, kg), final weight at the end of the feeding period (FW, kg), and at least one recorded visceral organ mass from 38 studies published between 1985 and 2021. Records were classified by sex/reproductive status as heifers (intact females), steers (castrated males), cows (mature females), or bulls (intact males), and cattle type was established as beef or dairy. Therefore, the dataset captures sex and castration status through this four-level classification but does not include explicit chronological age, as it was not reported in many studies; days on feed, initial weight, and final weight serve as the available indicators of age and physiological stage within the modeling framework. The type of forage was categorized into five groups based on ingredient composition and physical structure: hay, roughage byproducts (RBP), silage, combination of silage and hay, and combination of silage and roughage byproducts. Roughage byproducts included feed ingredients such as cottonseed hulls and corn cobs. Organ masses recorded directly from studies included those of the abomasum, omasum, liver, heart, lungs, pancreas, kidney, duodenum, jejunum, and ileum. Organ masses of the reticulorumen, pluck, stomach complex, small intestine, large intestine, total intestines, gastrointestinal tract, total splanchnic tissues, and total viscera were recorded directly when reported or calculated as the sum of component organs. Dietary variables were roughage, MEC, CP, NDF, and peNDF. Management variables were DOF, DMI, and DMI as a percentage of FW. Animal variables included cattle type, sex, FW, IW, and EBW. DMI as a percentage of FW was calculated using the following equation: (DMI/FW) × 100. Data from studies where the ration energy concentration was reported as digestible energy, net energy for maintenance, or net energy for gain were converted to MEC (Mcal/kg) using the conversion equations published in [4].

The pluck consisted of the heart and lungs. The reticulorumen, when not reported as a single unit, was calculated as the sum of the rumen and reticulum. The stomach complex included the reticulorumen, omasum, and abomasum. The small intestine, when reported in separate segments, was defined as the combined mass of the duodenum, jejunum, and ileum. Total intestines were calculated as the sum of the small and large intestines, while the gastrointestinal tract was recorded as the sum of the stomach complex and total intestines. Splanchnic tissue was defined as the sum of the liver, pancreas, spleen, and gastrointestinal tract. Total viscera were calculated as the sum of the total splanchnic tissues and the pluck.

2.2. Statistical Analyses

All statistical analyses were performed using R statistical software (version 4.4.3; 28 February 2025) with published functions and packages. The outcome variables were the mass of each organ (kg). Data were evaluated for normality using Q-Q plots and non-constant variance using residual plots from the simulateResiduals function in the DHARMa package. The lmer function of the lme4 package was used to model each visceral organ mass (kg) against the dietary, animal, and management variables, with Trial used as a random intercept to account for between-study variation [12]. The dredge function from the MuMIn package was then applied to each full model to evaluate all candidate models and select the model with the lowest AICc. The candidate predictor pool considered by dredge was constrained a priori to a biologically relevant set of dietary, management, and animal variables identified in prior literature as relevant to visceral organ mass.

If assumptions of normality were not met, organ mass was transformed. Normality and constant variance of the transformed data were re-evaluated using DHARMa residual diagnostics. For each organ, all four transformation types (raw, log, square root, and cube root) were fit and evaluated using DHARMa residual diagnostics. Transformations were required to pass three DHARMa tests (uniformity, dispersion, and outliers) at alpha = 0.05 to be considered for selection. When different data transformations (log, square root, or cube root) exhibited equivalent performance across residual diagnostics, the transformation providing the model with the lowest AICc was selected. If no transformation passed all three tests, the two transformations with the highest minimum DHARMa p-values were re-assessed, and the one with the lower AICc was selected. Abomasum, heart, large intestine, omasum, pancreas, pluck, reticulorumen, small intestine, spleen, stomach complex, and total gastrointestinal tract required cube-root transformations. Kidney and liver mass were transformed by square root. A logarithmic transformation was used for total intestines, total splanchnic tissues, and total viscera. Model predictions were back-transformed to the original scale for visualization.

Collinearity among fixed-effect predictors was assessed using variance inflation factors (VIFs) calculated from the selected models. Predictors with VIF values greater than or equal to 10 were flagged for further evaluation, as VIF values should be evaluated in context rather than as a strict exclusion threshold [13]. Pairwise correlations were computed for flagged predictors, and pairs with an absolute correlation of 0.70 or greater were identified. When at least one predictor in a correlated pair had a VIF of 10 or higher, the predictor with the lower chi-square statistic from the likelihood-ratio test was removed; this rule retains the predictor that contributes more to model fit and discards the predictor whose loss is least costly. When a flagged main effect was also contained in an interaction term, the interaction was removed first; the main effects were then re-assessed for collinearity, and the removal rule described above was only applied if collinearity persisted with the main effect.

In-sample fit metrics and leave-one-trial-out cross-validation (LOTO-CV) metrics were computed as follows. For the in-sample fit metrics, RMSE; conditional, marginal, and relative fixed R² (relative fixed R² = marginal R²/(1 − random R²); and MAE% were computed using the rmse_vec, rsq_vec, and mae_vec functions from the yardstick package. MAE was then expressed as a percentage of the mean observed value. The concordance correlation coefficient (CCC) was computed using the epi.ccc function of the epiR package. For the LOTO-CV, trial-adjusted organ weight was computed by predicting the organ weight using fixed-effect coefficients and adding the overall model residual for each observation, thereby removing the random effect of trial. The trial-adjusted organ weight used in LOTO-CV as a mixed-effects model does not generate a coefficient to predict the withheld trial. The random intercept was dropped using the nobars function of the lme4 package, and an ordinary least-squares model with the final model’s fixed effects as predictors and the trial-adjusted response as the dependent variable was fit on the remaining trials and used to predict the withheld trial. Pooled R², RMSE, and MAE% were computed across all out-of-fold predictions using the rsq_vec, rmse_vec, and mae_vec functions of the yardstick package, and CCC was computed using the epi.ccc function of the epiR package. The R² was computed on the transformed model scale to make it directly comparable with the relative fixed R² from the original model. The remaining metrics were computed on the back-transformed kg scale.

3. Results

3.1. Descriptive Statistics

The descriptive statistics for visceral organ mass are presented in

Table 1. Descriptive statistics for organ masses (kg) recorded from the meta-analysis.

Organ 1	Trials	N	Mean	SD	Min	Max
Reticulorumen	26	106	9.922	3.872	2.390	17.200
Omasum	20	82	3.447	1.706	0.750	9.214
Abomasum	21	84	1.853	1.035	0.650	5.253
Kidney	27	125	0.944	0.273	0.420	2.042
Liver	36	159	5.667	1.764	1.500	9.783
Heart	29	118	2.050	0.573	0.833	3.133
Pluck	27	114	6.072	2.474	1.430	11.310
Spleen	22	98	0.917	0.194	0.460	1.278
Pancreas	11	44	0.500	0.146	0.190	0.857
Stomach Complex	26	117	14.105	5.223	3.810	27.195
Small Intestine	20	88	5.312	1.033	2.700	7.100
Large Intestine	18	84	4.508	1.968	1.300	8.650
Total Intestines	27	134	11.913	6.405	4.000	35.373
Total GI Tract	28	123	25.630	10.571	8.080	52.068
Total Viscera	11	44	42.768	9.667	22.700	60.380
Total Splanchnic	9	40	36.570	9.244	16.330	54.220

¹. Reticulorumen = reticulum + rumen; Pluck = heart + lungs; Stomach Complex = reticulorumen + abomasum + omasum; Small Intestine = recorded directly when reported as a single unit or calculated as the sum of the duodenum, jejunum, and ileum when reported by segment; Large Intestine = reported and recorded as single unit; Total Intestines = small intestines + large intestines; Total GI Tract = Stomach Complex + Total Intestines; Total Splanchnic = liver + spleen + pancreas + Total GI tract; Total Viscera = Total Splanchnic + Pluck.

. Organ weight was heaviest for the total splanchnic tissues and total viscera, as expected, given that these variables are aggregates of multiple component organs, and lightest for the spleen and pancreas. The coefficient of variation was lowest for the small intestine and total viscera (19.4% and 22.6%) and highest for the abomasum and total intestines (55.9% and 53.8%). Pancreas, total viscera, and total splanchnic tissues had considerably lower numbers of observations and trials than the liver, kidneys, and total intestines. This uneven representation reflects the fact that source studies tended to focus on a few organs of primary interest rather than reporting data for all organs, with the pancreas and aggregate measures (total viscera and total splanchnic tissues) being reported less consistently.

3.2. Model Performance and Transformations

Model performance summaries for each organ are presented in

Table 2. Performance metrics for the mixed-effect model predicting visceral organ/organ system mass (kg) in growing/finishing steers, heifers, bulls, and mature cows.

Organ 1	RMSE 2	MAE%	CCC (95% CI)	Marginal R2	Random R2	Conditional R2	Rel. Fixed R2
Reticulorumen	0.591	4.537	0.988 (0.983, 0.992)	0.704	0.267	0.971	0.960
Omasum	0.378	8.417	0.974 (0.960, 0.983)	0.418	0.507	0.925	0.847
Abomasum	0.293	10.316	0.955 (0.933, 0.970)	0.548	0.344	0.892	0.835
Small Intestine	0.301	4.486	0.954 (0.931, 0.969)	0.550	0.375	0.924	0.879
Large Intestine	0.401	6.766	0.978 (0.967, 0.986)	0.393	0.579	0.972	0.933
Liver	0.351	4.952	0.980 (0.972, 0.985)	0.711	0.250	0.961	0.948
Spleen	0.084	7.155	0.893 (0.846, 0.926)	0.781	0.026	0.808	0.802
Pancreas	0.060	9.048	0.905 (0.834, 0.946)	0.873	0.000	0.873	0.873
Pluck	0.333	4.091	0.991 (0.987, 0.994)	0.378	0.601	0.979	0.947
Heart	0.093	3.324	0.986 (0.981, 0.991)	0.614	0.354	0.967	0.950
Kidney	0.083	5.570	0.949 (0.929, 0.964)	0.456	0.438	0.895	0.812
Total GI Tract	1.445	4.347	0.990 (0.986, 0.993)	0.413	0.561	0.974	0.940
Stomach Complex	0.979	5.382	0.982 (0.974, 0.987)	0.654	0.301	0.955	0.936
Total Intestines	0.686	3.964	0.994 (0.992, 0.996)	0.315	0.668	0.983	0.949
Total Splanchnic	1.481	2.983	0.987 (0.975, 0.993)	0.595	0.372	0.967	0.948
Total Viscera	1.533	2.692	0.987 (0.976, 0.993)	0.456	0.517	0.973	0.944

¹ Reticulorumen = reticulum + rumen; Pluck = heart + lungs; Stomach Complex = reticulorumen + abomasum + omasum; Small Intestine = recorded directly when reported as a single unit or calculated as the sum of the duodenum, jejunum, and ileum when reported by segment; Large Intestine = reported and recorded as single unit; Total Intestines = small intestines + large intestines; Total GI Tract = Stomach complex + Total Intestines; Total Splanchnic = liver + spleen + pancreas + Total GI Tract; Total Viscera = Total Splanchnic Tissues + Pluck; ² RMSE = root mean squared error; MAE% = mean absolute error expressed as a percentage of mean observed value (100 × MAE/mean of observed values); CCC = concordance correlation coefficient with 95% CI; Rel. Fixed R² = fixed-effects share of non-random-effect variance.

. Over half of the organs (n = 9) exhibited a mean absolute error (MAE%, expressed as a percentage of mean observed value) below 5.0%; six organs had MAE% values between 5.0 and 10.0%; and the abomasum was marginally higher, at 10.3%. Concordance correlation coefficients (CCCs) for most organs (n = 15) exceeded 0.900, indicating excellent agreement between observed and predicted values and strong predictive precision of diet composition, management characteristics, and animal characteristics. Only the spleen (0.893) showed a CCC value below 0.900. Fixed effects representing diet composition, management, and animal characteristics accounted for little of the variation in pluck and total intestine mass (marginal R² = 0.378 and 0.315, respectively); for these organs, the random effect of study accounted for the majority of variation explained. For the spleen and pancreas, fixed effects accounted for nearly all the modeled variation, but total variation explained was lower (conditional R² = 0.808 and 0.873, respectively) because the random effect contributed little. Conditional R² values exceeded 0.800 for all organs. The relative fixed R² was highest for the reticulorumen (0.960) and heart (0.950), indicating that after accounting for random study variation, the fixed effects accounted for a large amount of the remaining variation. For the pancreas, the random effect of study contributed no additional variance beyond the fixed effects (random R² = 0.000), indicating that between-study clustering had no impact on model fit for this organ.

3.3. Predictor Effects

3.3.1. Dry-Matter Intake and Days on Feed

Dry-matter intake was retained for 12 organs, and DMI as a percentage of final weight (DMI % FW) was retained for 7 organs, with at least one intake-related predictor retained in 15 of 16 models (

Table 3. Coefficients for predictors (±SE) in the final models to predict reticulorumen, omasum, abomasum, small intestine, and large intestine mass (kg) in growing/finishing steers, heifers, bulls, and mature cows.

Predictor 1	Reticulorumen 2	Omasum	Abomasum	Small Intestine	Large Intestine
Intercept	1.4394 ± 0.1401 *	0.7700 ± 0.3469 *	0.3412 ± 0.3944	1.4731 ± 0.0487 *	0.9790 ± 0.1239 *
DMI	0.0476 ± 0.0120 *	0.0404 ± 0.0057 *	0.0459 ± 0.0070 *	0.0224 ± 0.0047 *	0.0318 ± 0.0087 *
DMI % FW	−0.1449 ± 0.0620 *	-	−0.1332 ± 0.0569 *	-	-
DOF × 100	-	-	0.0930 ± 0.0320 *	-	0.0230 ± 0.0220
EBW × 100	0.0690 ± 0.0260 *	-	-	0.0240 ± 0.0140	0.0740 ± 0.0270 *
CP	0.0115 ± 0.0045 *	0.0009 ± 0.0060	-	-	0.0016 ± 0.0060
MEC	-	0.1127 ± 0.1073	0.1755 ± 0.1184	-	-
NDF	0.0049 ± 0.0014 *	0.0027 ± 0.0025	0.0039 ± 0.0023	-	0.0020 ± 0.0015
peNDF	−0.0015 ± 0.0015	−0.0012 ± 0.0015	-	-	−0.0059 ± 0.0020 *
NDF × peNDF	−0.00004 ± 0.00002 *	-	-	-	-
Roughage	-	-	0.0002 ± 0.0013	−0.0011 ± 0.0004 *	-
Forage Type: RBP 3	-	-	-	-	-
Forage Type: Silage	-	-	-	-	-
Forage Type: Silage/Hay	-	-	-	-	-
Forage Type: Silage/RBP	-	-	-	-	-

¹ DMI = dry-matter intake (kg/d); DMI % FW = dry-matter intake as a percentage of final live weight; DOF = days on feed; EBW = empty body weight (kg × 100); CP = crude protein concentration (% DM); MEC = metabolizable energy concentration (Mcal/kg DM); NDF = neutral detergent fiber concentration (% DM); peNDF = physically effective neutral detergent fiber concentration (% DM); RBP = roughage byproduct. Coefficients and SE are on the transformed scale used to fit each model; all organs are on the cube root scale. ² Reticulorumen = reticulum + rumen; Small Intestine = recorded directly when reported as a single unit or calculated as the sum of the duodenum, jejunum, and ileum when reported by segment; Large Intestine = reported and recorded as a single unit; ³ Forage type: Hay was used as the baseline category, with coefficients for other forage types represented as deviations from the baseline. * Regression coefficients are significantly different from zero at p ≤ 0.05. “-“ indicates the predictor was not retained in the final model.

,

Table 4. Coefficients for predictors (±SE) in the final models to predict liver, spleen, pancreas, pluck, heart, and kidney mass (kg) in growing/finishing steers, heifers, bulls, and mature cows.

Predictor 1	Liver	Spleen	Pancreas	Pluck 2	Heart	Kidney
Intercept	0.9096 ± 0.1182 *	0.4504 ± 0.1250 *	0.8554 ± 0.1064 *	1.3701 ± 0.0914 *	1.0121 ± 0.0342 *	0.5810 ± 0.0629 *
DMI	0.0527 ± 0.0070 *	-	-	0.0023 ± 0.0039	0.0084 ± 0.0025 *	0.0122 ± 0.0046 *
DMI % FW	-	0.0078 ± 0.0112	-	-	−0.0134 ± 0.0056 *	−0.0070 ± 0.0114
DOF × 100	-	−0.0380 ± 0.0090 *	-	-	-	−0.0240 ± 0.0150
EBW × 100	0.1830 ± 0.0220 *	0.0480 ± 0.0050 *	0.0500 ± 0.0050 *	0.0820 ± 0.0140 *	0.0540 ± 0.0080 *	0.0510 ± 0.0120 *
CP	0.0231 ± 0.0056 *	0.0058 ± 0.0020 *	-	0.0015 ± 0.0031	-	0.0085 ± 0.0033 *
MEC	-	0.0840 ± 0.0345 *	−0.0967 ± 0.0315 *	-	-	-
NDF	-	0.0009 ± 0.0008	−0.0029 ± 0.0008 *	-	-	-
peNDF	−0.0035 ± 0.0012 *	−0.0015 ± 0.0008	-	−0.0020 ± 0.0007 *	−0.0024 ± 0.0007 *	-
Forage Type: RBP 3	-	0.0420 ± 0.0240	0.0414 ± 0.0193 *	0.3344 ± 0.2072	-	-
Forage Type: Silage	-	−0.0004 ± 0.0100	0.0436 ± 0.0120 *	0.1564 ± 0.0813	-	-
Forage Type: Silage/Hay	-	0.0327 ± 0.0158	0.0535 ± 0.0273	0.1026 ± 0.0891	-	-
Forage Type: Silage/RBP	-	−0.0504 ± 0.0206 *	0.0764 ± 0.0175 *	0.1700 ± 0.0852	-	-

¹ DMI = dry-matter intake (kg/d); DMI % FW = dry-matter intake as a percentage of final live weight; DOF = days on feed; EBW = empty body weight (kg × 100); CP = crude protein concentration (% DM); MEC = metabolizable energy concentration (Mcal/kg DM); NDF = neutral detergent fiber concentration (% DM); peNDF = physically effective neutral detergent fiber concentration (% DM); RBP = roughage by product. Coefficients and SE are on the transformed scale used to fit each model; liver and kidney are on the square-root scale, and spleen, pancreas, and pluck are on cube-root scale. ² Pluck = heart + lungs. ³ Forage type: Hay was used as the baseline category, with coefficients for other forage types represented as deviations from the baseline. * Regression coefficients are significantly different from zero at p ≤ 0.05. “-“ indicates the predictor was not retained in the final model.

and

Table 5. Coefficients for predictors (±SE) in the final models to predict stomach complex, total intestine, total gastrointestinal tract, total splanchnic, and total viscera mass (kg) in growing/finishing steers, heifers, bulls, and mature cows.

Predictor 1	Stomach Complex	Total Intestines	Total GI Tract 2	Total Splanchnic	Total Viscera
Intercept	1.6000 ± 0.3461 *	0.4985 ± 0.3518	1.8995 ± 0.0999 *	3.1688 ± 0.3643 *	3.0149 ± 0.1513 *
DMI	-	0.0299 ± 0.0060 *	-	0.0416 ± 0.0064 *	0.0311 ± 0.0065 *
DMI % FW	0.0739 ± 0.0310 *	-	0.0981 ± 0.0337 *	-	-
DOF × 100	-	-	-	-	-
EBW × 100	0.1570 ± 0.0140 *	0.1570 ± 0.0200 *	0.1850 ± 0.0160 *	-	0.0570 ± 0.0240 *
CP	0.0061 ± 0.0062	-	-	-	0.0119 ± 0.0056 *
MEC	−0.0373 ± 0.1078	0.2283 ± 0.1001 *	-	0.0466 ± 0.1076	-
NDF	0.0047 ± 0.0027	0.0089 ± 0.0023 *	0.0034 ± 0.0025	-	-
peNDF	−0.0049 ± 0.0016 *	0.0056 ± 0.0024 *	-	−0.0087 ± 0.0040 *	-
Roughage	-	-	−0.0013 ± 0.0011	-	−0.0025 ± 0.0008 *
NDF × peNDF	-	−0.00011 ± 0.00003 *	-	-	-
Forage Type: RBP 3	-	−0.0881 ± 0.2917	-	-	-
Forage Type: Silage	-	0.1278 ± 0.0903	-	-	-
Forage Type: Silage/Hay	-	−0.0204 ± 0.0560	-	-	-
Forage Type: Silage/RBP	-	0.0109 ± 0.1023	-	-	-

¹ DMI = dry-matter intake (kg/d); DMI % FW = dry-matter intake as a percentage of final live weight; DOF = days on feed; EBW = empty body weight (kg × 100); CP = crude protein concentration (% DM); MEC = metabolizable energy concentration (Mcal/kg DM); NDF = neutral detergent fiber concentration (% DM); peNDF = physically effective neutral detergent fiber concentration (% DM); RBP = roughage byproduct. Coefficients and SE are on the transformed scale used to fit each model: stomach complex and total GI tract are on the cube-root scale, and total intestines, total splanchnic, and total viscera are on the logarithmic scale. ² Stomach Complex = reticulorumen + abomasum + omasum; Total GI Tract = Stomach Complex + Total Intestines; Total Intestines = small intestines + large intestines; Total Splanchnic = liver + spleen + pancreas + Total GI Tract; Total Viscera = Total Splanchnic Tissues + Pluck. ³ Forage type: Hay was used as the baseline category, with coefficients for other forage types represented as deviations from the baseline. * Regression coefficients are significantly different from zero at p ≤ 0.05. “-“ indicates the predictor was not retained in the final model.

). Dry-matter intake had positive coefficients for all organs in which it was retained in the final model, with the largest positive coefficients being retained for the liver and reticulorumen. Dry-matter intake as a percentage of final weight had positive coefficients for most organs in which it was retained but was negative for the reticulorumen, abomasum, and heart. Days on feed was rarely retained (n = 4) and had an even split of positive and negative coefficients, with the strongest positive coefficient for the abomasum and the strongest negative coefficient for the spleen.

3.3.2. Roughage, Physically Effective Neutral Detergent Fiber, and Neutral Detergent Fiber

The dietary roughage level was infrequently retained (n = 4) and had a negative effect for most organs (Table 3, Table 4 and Table 5). The largest negative coefficients were for total viscera and the gastrointestinal tract. Additionally, peNDF had a mostly negative correlation with visceral organ mass, with the strongest negative coefficient for total splanchnic tissues and the largest positive coefficient for total intestines. Neutral detergent fiber was retained for nine organs and had positive coefficients for all organs except the pancreas, with the largest positive coefficients for the total intestines and reticulorumen.

3.3.3. Forage Type

Forage type was retained in the model for four organs (Table 3, Table 4 and Table 5). For the spleen, only the silage/RBP category differed significantly from the baseline of hay, yielding a negative coefficient, indicating lower spleen mass for silage/RBP diets; the other forage types did not differ significantly from hay. All forage types were associated with greater pancreas mass than the baseline of hay.

3.3.4. Crude Protein and Metabolizable Energy Concentration

Crude protein concentration was retained as a predictor in the model for nine organs and had a positive coefficient for most organs (n = 8) (Table 3, Table 4 and Table 5). The largest positive CP coefficients occurred for the liver and total viscera, with the only negative coefficient occurring for the omasum. Metabolizable energy concentration was retained in the model for seven organs and exhibited a positive coefficient for all organs except the pancreas. The spleen and total intestines had the largest positive coefficients for MEC.

3.3.5. Animal Characteristics

Empty body weight was the most frequently retained predictor (n = 13) across all models (Table 3, Table 4 and Table 5). Empty body weight had a positive coefficient with the mass of all organs for which it was retained in the final model, with the largest positive coefficients being with the gastrointestinal tract and liver. Sex and breed type were evaluated by likelihood-ratio tests and were found to be not significant (p > 0.05) in any model.

3.4. Cross-Fold Validation Performance

Cross-fold validation results for each organ model can be found in

Table 6. Cross-fold validation of final models for visceral organ/organ system mass (kg) in growing/finishing steers, heifers, bulls, and mature cows.

Organ 1	R2	RMSE	MAE%	CCC (95% CI)
Reticulorumen	0.967	0.590	4.756	0.980 (0.971, 0.986)
Omasum	0.867	0.370	9.075	0.917 (0.876, 0.945)
Abomasum	0.854	0.262	10.726	0.901 (0.854, 0.934)
Small Intestine	0.899	0.300	4.615	0.941 (0.912, 0.961)
Large Intestine	0.943	0.367	7.147	0.964 (0.945, 0.976)
Liver	0.958	0.368	5.108	0.976 (0.967, 0.982)
Spleen	0.728	0.101	8.581	0.833 (0.761, 0.885)
Pancreas	0.784	0.070	11.798	0.832 (0.707, 0.907)
Pluck	0.952	0.325	4.157	0.972 (0.959, 0.981)
Heart	0.951	0.100	3.581	0.972 (0.960, 0.980)
Kidney	0.808	0.085	5.927	0.888 (0.845, 0.920)
Stomach Complex	0.945	0.977	5.532	0.969 (0.956, 0.978)
Total GI Tract	0.951	1.437	4.612	0.972 (0.960, 0.980)
Total Intestines	0.958	0.592	4.190	0.974 (0.964, 0.981)
Total Splanchnic	0.951	1.609	3.260	0.968 (0.941, 0.983)
Total Viscera	0.944	1.608	3.026	0.965 (0.937, 0.981)

¹ Reticulorumen = reticulum + rumen; Pluck = heart + lungs; Stomach Complex = reticulorumen + abomasum + omasum; Small Intestine = recorded directly when reported as a single unit or calculated as the sum of the duodenum, jejunum, and ileum when reported by segment; Large Intestine = reported and recorded as a single unit; Total Intestines = small Intestine + Large Intestine; Total GI Tract = Stomach Complex + Total Intestines; Total Splanchnic = liver + spleen + pancreas + Total GI tract; Total Viscera = Total Splanchnic Tissues + Pluck; RMSE = root mean square error.

. The highest performing model was the reticulorumen model (R² = 0.967; RMSE = 0.590; MAE% = 4.756). The lowest performing model was the spleen model (R² = 0.728; RMSE = 0.101; MAE% = 8.581). The observed vs. predicted values for each organ for the original and cross-validation models are presented in Figures S1–S16.

4. Discussion

4.1. Overall Finding and Model Performance

This meta-analysis quantified relationships of visceral organ mass with diet composition, management characteristics, and animal characteristics in cattle across 38 studies. Across the 16 organ models, the average relative fixed R² was 0.906, indicating that diet composition, management characteristics, and animal characteristics explain substantial variation in organ mass. These findings align with reports by Sainz and Bentley [6] and Johnson et al. [1] that alterations in dietary energy sources and fiber content influenced total stomach-complex mass and with more recent work characterizing visceral organ contributions to between-animal variation in feed efficiency [2,5], indicating that visceral organs adjust to changes in nutrient supply and physical bulk.

4.2. Effects of Dry-Matter Intake and Days on Feed

Higher DMI was associated with enlargement of the total intestines and total viscera, consistent with the idea that intake drives hypertrophy of tissues that process digesta [1]. These increases in the mass of the gastrointestinal tract and total intestines align with the findings of McLeod et al. [14], who reported that DMI levels were systematically increased, with the weights of the gastrointestinal tract increasing as a result. Recent work in finishing steers also supports the broader pattern of visceral mass adjustment to intake-related variation in feed efficiency [5].

Dry-matter intake expressed as a percentage of final weight (DMI % FW) had positive coefficients for the stomach complex and total gastrointestinal tract, suggesting that animals that consume more feed relative to their body size have larger digestive and fermentative organs. This is consistent with the idea that intake drives hypertrophy of the forestomach and associated structures [1]. However, DMI % FW had negative coefficients for the heart. The negative association with heart mass reflects the dilution effect, as higher relative intake drives growth of digestive tissues and the heart scales sub-proportionally with intake [15].

More days on feed (DOF) were linked to decreases in the mass of the spleen. This negative association between DOF and spleen mass has not been identified in the current literature and warrants further investigation.

4.3. Empty Body Weight

Empty body weight had a positive coefficient with the mass of all organs in which it was retained, indicating that larger animals have greater visceral organ mass. This is consistent with allometric growth, whereby visceral organ mass increases as animals grow [1,16].

4.4. Roughage, Physically Effective Neutral Detergent Fiber, and Neutral Detergent Fiber

Roughage (% of diet) had a negative coefficient with the mass of the small intestine. Heightened roughage levels are characterized by lower energy density [17] and slower rates of passage [4], which reduces the flow of digestible nutrients to the small intestine [7] and limits the physical stimulus needed for intestinal growth [18]. This is consistent with the dependence of small intestinal mass on the quantity and quality of nutrients presented for absorption [6,19]. The negative association with total viscera reflects the cumulative effect of reduced nutrient availability across multiple organs in cattle consuming diets with higher roughage levels.

Physically effective NDF had a negative coefficient with the mass of most organs in which it was retained, including the large intestine, liver, and stomach complex, with the total intestines being the only organ to exhibit a positive association. The mostly negative relationship between peNDF and organ mass reflects the fact that diets high in peNDF reduce the energy density and digestibility of diets [7,18], which limits the supply of nutrients available to support organ growth and maintenance [14]. The positive association between peNDF and total intestinal mass is consistent with the role of dietary bulk and particle size in stimulating intestinal hypertrophy [1]. The intestines respond to the physical form of the diet [20]. Physically effective NDF levels are correlated with particle size and the amount of bulk [21], which are known to stimulate intestinal growth [20].

Neutral detergent fiber had a positive coefficient with the reticulorumen and total intestines but had a negative coefficient with pancreatic mass. High concentrations of NDF have positive effects on the reticulorumen by influencing the functional workload associated with the muscular handling of bulky feed and stimulating physical processes [22]. The positive relationship between NDF and the reticulorumen is consistent with the reticulorumen’s responsiveness to dietary fiber content [19]. Mcleod and Baldwin [19] reported that increased dietary fiber content caused hyperplasia and an increase in reticulorumen mass. The increase in total intestinal mass is driven by the functional workload that feedstuffs higher in NDF impose on the gastrointestinal tract [20]. The negative relationship between NDF and the mass of the pancreas is consistent with the mass of the pancreas being responsive to the functional workload of metabolizing a high nutrient supply [14]. Feedstuffs with higher NDF levels are often correlated with reduced nutrient availability [23], which has been associated with reduced pancreas mass [14]. Therefore, if high NDF correlates with reduced nutrient and energy availability, the functional workload and subsequent mass of the pancreas will decline. These intake- and fiber-driven shifts in visceral organ mass have also been observed in lambs recovering from nutritional restriction, with dietary energy density and rumen-undegradable protein supplementation found to alter viscera mass [23].

4.5. Forage Type

Forage type had significant effects on pancreas mass. Relative to the baseline of hay, the RBP, silage, and silage/RBP categories each had greater pancreatic mass. The greater mass with diets containing silage is consistent with silage being more energy-dense than hay, increasing the demand for the synthesis and excretion of pancreatic enzymes and, therefore, pancreatic mass [24].

4.6. Crude Protein and Metabolizable Energy Concentration

Crude protein concentration had a positive coefficient for total viscera and liver. Previously published research has specifically demonstrated that increasing the dietary CP concentration increases the mass of viscera [8]. Increased CP intake raises the amount of absorbed nutrients presented to the visceral organs, increasing their workload [6]. For the liver, increased CP availability is believed to increase liver mass, as the organ must enlarge to compensate for the heightened functional workload to metabolize amino acids and ammonia to urea [1]. These directional effects of energy and protein supply on visceral mass are consistent with more recent reports in ruminants linking dietary energy density and protein supply to viscera-to-carcass partitioning [23].

Metabolizable energy concentration had a positive coefficient for the mass of the spleen and total intestines and a negative coefficient for the pancreas. Elevated MEC levels necessitate an increase in portal blood flow to the gastrointestinal tract to process nutrients, forcing the spleen to increase its mass to expand its blood storage capacity [16]. For the total intestines, elevated MEC delivers a greater nutrient load to post-stomach tissues, which requires the intestines to increase in mass to physically accommodate digestion and absorption [6]. As MEC increases, cattle are able to meet their nutrient requirements while consuming a lower volume of DMI [25]. Concurrently, the mass of the pancreas decreases under elevated MEC levels, as reduced DMI reduces the need for the pancreas to secrete digestive enzymes, ultimately lowering its functional workload and reducing its mass [14,24].

4.7. Summed Components vs. Composite Prediction

Two approaches for predicting gastrointestinal tract mass were compared: the summing of predictions from the individual component models (reticulorumen, omasum, abomasum, small intestine, and large intestine) versus the use of a single composite model fit directly to total gastrointestinal tract mass. Based on fixed-effect coefficients, where random effects from trial were removed and predictions relied exclusively on diet and management fixed effects, the approach where each individual component was predicted, then summed outperformed the composite gastrointestinal tract model. The R² was 0.757, compared with 0.741, and the CCC was 0.846, compared with 0.810, for the component and composite gastrointestinal tract models, respectively. The RMSE was 3.606 kg, compared with 4.074 kg, and the MAE% was 12.3%, compared with 14.5%, for the component and composite gastrointestinal tract models, respectively. Both approaches showed moderate to strong performance when used as prediction equations (R² = 0.757 and 0.741); however, the component-summing approach performed better across all four metrics, suggesting that the strategy of aggregating component organ predictions is preferable relative to the modeling of a composite organ. The likely reason for this is that subtle differences in the mass of individual components are masked when trying to predict the composite organ mass, leading to a more generalized prediction rather than specificity for different combinations of diet, management, and animal characteristics. This is consistent with previously published literature in which steers managed under different winter grazing programs exhibited varying treatment effects on the relative mass of individual gastrointestinal tract components, with the omasum being most responsive and the large intestine unaffected [26].

4.8. Study Limitations

The number of observations varied considerably among organs, with organs such as the liver (n = 159), kidneys (n = 125), and total intestines (n = 134) having significantly more observations than the pancreas (n = 44), total viscera (n = 44), and total splanchnic tissues (n = 40). Future research expanding the database for under-represented organs would improve prediction accuracy and broaden the range of conditions under which these equations can be applied.

Where chemical composition of the diet was not directly reported, values from the Nutrient Requirements of Beef Cattle [4] were used to estimate composition from ingredient lists; these values represent population averages for feedstuffs and may differ from the actual composition of the specific feedstuffs used in individual studies, introducing additional uncertainty in the diet-related predictors. The same caveat applies to physically effective NDF, which was calculated from NRC ingredient values. Thus, model coefficients may tend to predict toward the average values for feedstuffs, potentially resulting in prediction error for diets with extreme values.

5. Conclusions

Empty body weight, dry-matter intake, days on feed, and fiber-related diet characteristics (roughage, neutral detergent fiber, and physically effective neutral detergent fiber) were the most consistently retained predictors across organs, with crude protein concentration and metabolizable energy concentration being retained less frequently. In-sample agreement between observed and predicted values was strong for most organs, with the spleen showing the lowest agreement (the only organ with a CCC below 0.900). Models performed well under cross-validation, with leave-one-trial-out R² values ranging from 0.728 to 0.967 across organs. Within the range of conditions represented in this dataset, the equations provide a quantitative basis for estimating visceral organ mass and should support more refined estimates of maintenance energy requirements. Future research should expand the dataset for under-represented organs, evaluate these equations in independent datasets, and link predicted changes in visceral organ mass to direct measurements of maintenance energy requirements in cattle.