Association of origin, performance, and health metrics of beef-on-dairy feedyard cattle with liver abscess at harvest

Abstract

Beef-on-dairy cattle (B × D) exhibit a higher prevalence of liver abscesses (LA) at harvest compared to native beef breeds and Holsteins. The objective was to evaluate associations of origin, performance, and health metrics with LA prevalence and severity at harvest in B × D feedyard cattle. Operational data and liver scores were collected from two sources (Dataset 1, n = 330 cohorts; Dataset 2, n = 386 cohorts). For Dataset 1, livers were scored using Elanco’s liver scoring system with 4 outcomes of interest: Total LA (All scores; D1TA), Severe LA (Scores A+, A + Adhered, A + Open; D1SA), Severe + LA (A + Adhered and A + Open; D1S + A), and Severe as proportion of Total LA (D1ST). For Dataset 2, livers were scored in 3 non-mutually exclusive categories: Abscessed (D2TA), Condemned (D2CL), and Adhered (D2AdL). Generalized linear mixed models estimated probability of LA outcomes with a beta-binomial distribution to account for overdispersion. Fixed effects included sex, days on feed (DOF), shipping weight, head received, average daily gain (ADG), mortality, railers, average daily feed delivery per head, quarter of arrival/harvest, and region of origin. Feedyard was included as a random effect. Multiple risk factors such as sex, DOF, shipping weight, railers, and origin were associated with D1TA, D1SA, D1S + A, and D1ST. Quarter harvested and ADG were associated with D2TA and D2CL. Shipping weight and origin were also associated with D2TA. Sex was the only risk factor associated with D2AdL. Steer cohorts had higher probability (P ≤ 0.05) of D1TA, D1SA, D1S + A, D1ST, and D2AdL than heifer cohorts. Cohorts on feed > 300 days had higher probability (P ≤ 0.05) of D1TA, D1SA, and D1S + A than cohorts with fewer DOF. Cohorts with shipping weights of 680 to 725 kgs had higher risk (P = 0.03) of D1SA than those > 726 kgs. Average daily gain was associated with (P ≤ 0.02) D2TA and D2CL. Cohorts with < 0.5% railed had a lower probability (P ≤ 0.04) of LA outcomes than those with > 0.5% railed. Cattle harvested in quarter 3 had lower risk (P ≤ 0.03) of D2TA and D2CL. Cohorts originating from region 3 had lower risk (P ≤ 0.05) of D1TA, D1SA, D1S + A, and D1ST than those from regions 1 and 4. These results identify associations among B × D cattle demographics, origin, performance, and health metrics with LA prevalence at harvest. Further research is needed to evaluate management strategies to mitigate LA occurrence.

Beef-on-dairy cattle have a higher prevalence of liver abscesses than beef breeds and Holsteins, but contributing factors remain unclear. This study identifies production, health, and origin metrics associated with liver abscess risk, providing insight for targeted management strategies to reduce occurrence.

Introduction

Breeding practices in the dairy industry have shifted to using beef semen to produce crossbred beef-on-dairy (B × D) calves that enter the beef production system. This practice increases the value of the calves produced from dairies as B × D cattle have improved feedyard performance, carcass quality, and meat attributes compared to Holsteins (Foraker et al. 2022). A major issue with B × D cattle is the high incidence of liver abscesses (LA) which is 2 to 3 times greater than crossbred beef cattle (Foraker et al. 2022). Liver abscesses are the most common liver abnormality detected at the time of slaughter, with recent estimates indicating that associated economic losses to the U. S. beef industry may exceed $250 million annually (Taylor et al. 2025). Liver abscesses are thought to occur mainly in the finishing phase due to diets consisting of high concentrations of highly fermentable carbohydrates fed for a long period of time (Bryant and Jennings 2022). Common finishing diets are believed to lower ruminal pH, causing ruminal acidosis that leads to damage of the rumen epithelium acting as a pathway for ruminal bacteria to enter the portal blood supply, and cause infections and abscesses in the liver (Amachawadi and Nagaraja 2022). Previous research has explored the associations between various factors such as demographics (sex, breed, arrival/shipping weight, days on feed (DOF), etc.) and management (roughage level, feedstuff particle size, grain type and inclusion, etc.) with the prevalence of LA in native beef cattle at harvest or necropsy (Nagaraja and Chengappa 1998; Amachawadi and Nagaraja 2016; Broadway et al. 2024; Hales 2024; Champagne et al. 2025). Identifying feedyard variables associated with LA could provide an avenue for producers to attempt to control LA through adjusting management practices. This would be beneficial, as the current method for control and prevention of LA consists of continuous in-feed administration of antimicrobials such as tylosin phosphate (Theurer and Amachawadi 2022; Word et al. 2024; Young et al. 2026). There is growing concern of antimicrobial resistance, so if producers could modify management practices to reduce LA, this could promote more judicious use of in-feed antibiotics.

There has been speculation that early life feeding practices and management may predispose cattle to LA later in life (Machado and Ballou 2022; Hales 2024). Beef-on-dairy cattle are raised similarly to dairy replacement heifers as they are separated from their dams as early as one day after birth. Calves are typically offered a variation of calf starter consisting of either pelleted feeds, a combination of pellets and a roughage source, or a total mixed ration as soon as they are placed in a hutch or pen (Bigelow et al. 2025). This early introduction to grain is believed to potentially compromise the integrity of rumen or small intestine (SI) epithelium, leading to impaired gut function once the calf reaches the feeding phase. During a year-long necropsy trial across 3 commercial calf ranches, only two abscessed livers were recorded out of 243 necropsies, one of which resulted from an umbilical infection (Bigelow et al. 2025). Thus, LA are likely not occurring at young ages, but management practices may affect their occurrence later in life. One of the challenges with identifying associations between management practices at the calf ranch phase, feeding phase, and harvest is the lack of traceability of cohorts through the different production systems. Therefore, the authors opted to explore region of origin as a proxy for calf ranch to identify if there was evidence warranting further exploration of specific management practices. A limitation of this approach was the reality that origin information provided to the feedyard could represent calf ranch origin, backgrounding operation, or an auction barn that calves were marketed through. Therefore, the objective of this research was to evaluate associations of origin, performance, and health metrics with cohort-level prevalence and severity of LA at harvest in B × D feedyard cattle.

Materials and methods

The Kansas State University Institutional Animal Care and Use Committee (IACUC) was contacted regarding the research, and no approval was needed due to data being obtained from operation databases and no procedures were performed on live cattle.

Data collection

Operational closeout records were collected from 11 U.S. feedyards across the High Plains region spanning 2020 to 2025. Closeout records consisted of information such as feedyard ID, cohort ID, sex of the cohort, number of head received, number of head shipped for harvest, dates received/shipped, weight of cohort at arrival/shipping, number of head days, number of deaths and railers, the total pounds of feed delivered to the cohort over the entire feeding period, and origin which typically consisted of a town and state. Harvest data were collected from two different sources from 2021 to 2025. The livers from Dataset 1 (D1) were scored at harvest using a modified version of Elanco’s liver abscess scoring system (n = 343 cohorts) and consisted of cattle from 8 feedyards. These scores were mutually exclusive categories as described in Table 1. For D1, the analyses consisted of four outcomes of interest: Total abscessed livers (D1TA; count of all LA scores/count of cattle without LA), Severe abscessed livers (D1SA; count of A+, A + Adhered (Ad), A + Open (Op), and A + Ad&Op/count of cattle without Severe LA), Severe + abscessed livers (D1S + A; count of A + Ad, A + Op, and A + Ad&Op/count of cattle without Severe + LA), and Severe LA as a proportion of total LA (D1ST; count of A+, A + Ad, A + Op, and A + Ad&Op/count of LA scores A- and A). The outcome D1S + A was of interest because it consists of the most severe LA scores that typically cause loss of trim and cause the chain to stop in the harvesting process which are both large sources of economic loss (Taylor et al. 2025). The outcome D1ST was chosen because the authors expected that the frequency of severe abscesses would increase with overall abscess prevalence; however, the data did not support this expected relationship.

Table 1.

Liver scores used to evaluate livers at harvest in the processing plants.^a

Liver scores by Dataset	Description
Dataset 1
O	Normal liver with no lesions
A−b	1 to 2 small abscesses less than an inch in diameter
Ab	2 to 4 abscesses less than an inch in diameter
A+b,c,d	1 or more large abscesses more than one inch in diameter
A + Adheredb,c,d,e	Abscess present with adhesion(s) to diaphragm or other organs
A + Openb,b,d,e	Abscess present with a ruptured abscess
A + Adhered/A + Openb,c,d,e	Abscess present with both adhesions and a ruptured abscess
Dataset 2
Abscessedf	Any liver abscess present
Condemnedg	Any liver with ruptured abscess
Adheredh	Any liver with an adhesion to diaphragm or other organs

^aLivers from Dataset 1 were scored using a modified version of Elanco’s liver abscess scoring system. Scores were mutually exclusive categories. Livers from Dataset 2 were scored using the processing plant’s internal scoring system. Livers could be classified as abscessed, condemned, and/or adhered. These categories were not mutually exclusive.

^bScores included in Total LA outcome in Dataset 1 (D1TA).

^cScores included in Severe LA outcome in Dataset 1 (D1SA).

^dScores included in Severe LA as a proportion of Total outcome in Dataset 1 (D1ST).

^eScores included in Severe + LA outcome in Dataset 1 (D1S + A).

^fOutcome in Dataset 2 (D2TA).

^gOutcome in Dataset 2 (D2CL).

^hOutcome in Dataset 2 (D2AdL).

For Dataset 2 (D2), three categories were used to score livers per the source’s internal scoring system, but they were not mutually exclusive categories and consisted of cattle from 5 feedyards (n = 409 cohorts). The categories consisted of livers that were abscessed, condemned and/or adhered to the diaphragm and/or other organs, and livers could be counted in no, one or multiple categories. Livers were classified as abscessed when any abscess was present regardless of severity. They were classified as condemned if the liver was condemned and removed from entering the food chain. Livers were classified as adhered if there was any adhesion present to the diaphragm. For D2, the outcomes of interest were total abscessed livers (D2TA; count of abscessed livers/count of cattle without abscessed livers), condemned livers (D2CL; count of condemned livers/count of cattle without condemned livers), and adhered livers (D2AdL; count of adhered livers/count of cattle without adhered livers). Separate outcomes were calculated for D1 and D2 due to the different scoring systems used during collection of liver data.

The scoring system used in Dataset 1 was completely independent from the system used in Dataset 2. There was no overlap between the two datasets or liver scores/classifications.

Data manipulation

For operational data, DOF was calculated by subtracting the arrival date from the shipping date. Average arrival and shipping weights were calculated by dividing the cohort arrival and shipping weights by head received and head shipped, respectively. Average daily gain with deads included (ADG—deads in) was calculated by subtracting the total cohort arrival weight from total cohort shipping weight and dividing by head days. Average daily gain with deads excluded (ADG—deads out) was calculated by subtracting the average arrival weight per head from the average shipping weight per head and dividing by DOF. Mortality was calculated by dividing the number of deaths by the number of head received and multiplying by 100 to get a percentage of the cohort. Similarly, the percentage of the cohort railed was calculated by dividing the number of head railed by the number of head received and multiplying by 100. Dates of arrival and shipment to harvest were categorized into quarters (Q1: January, February, March; Q2: April, May, June; Q3: July, August, September; Q4: October, November, December). Origin was categorized into regions based on the map used in the USDA’s “Health Management on U.S. Feedyards 2021 Phase 1 Questionnaire” (USDA-APHIS 2020).

Statistical analyses

Continuous variables were categorized and adjusted based on data distributions. Predictors of interest included sex of the cohort, DOF, shipping weight, arrival weight, head received, head shipped, ADG—deads out, ADG—deads in, mortality, railers, average daily amount of feed delivered per head, quarter of arrival, quarter shipped, and region of origin. Correlations between predictor variables were tested using a Pearson correlation matrix. If two variables were strongly correlated (Pearson correlation coefficient (r) > 0.70), one of the variables was removed from the final models and the other was retained. Head received and head shipped were strongly correlated (Pearson correlation coefficient (r) = 0.95), so head received was left in the final models and head shipped was removed. Similarly, ADG—deads in and ADG—deads out had a Pearson correlation coefficient of 0.87. In the final models, ADG—deads out was retained and ADG—deads in was removed. Shipping weight and arrival weight were strongly correlated as well (r = 0.93), so shipping weight was kept and arrival weight was removed from the final models. Variance inflation factors were tested for each final model to ensure there was no multicollinearity between predictor variables.

Outcomes of interest were analyzed in RStudio (Version 2024.04.2 + 764, Boston, MA) using the glmmTMB function from the “glmmTMB” package. Outcomes were analyzed as successes/failures per cohort in generalized linear mixed models (GLMM) using a beta-binomial distribution to account for overdispersion of the data. A GLMM was used to analyze each outcome with the following fixed variables: sex of cohort, DOF, shipping weight, head received, ADG—deads out, mortality, railers, average daily feed delivery per head, quarters of arrival and harvest, and region of origin. Feedyard was included as a random effect. Backwards step-wise regression was used to build the final models where all predictor variables were included initially. The variable with the highest P-value was removed until all remaining variables had a P-value < 0.05. The same model-building process used for D1 analyses was used for D2 analyses.

Results

Descriptives

Operational feedyard data and liver abscess scores at harvest were analyzed for 343 and 409 cohorts from D1 and D2, respectively. Cohorts of mixed sexes and/or cattle other than B × D (Native, Mexican, etc.) were removed from the datasets prior to analyses (D1: n = 13 cohorts; D2: n = 23 cohorts). This resulted in 330 cohorts of 59,210 cattle for D1 and 386 cohorts of 46,884 cattle for D2. Demographic information and liver abscess outcomes are reported in Table 2. For D1, the average D1TA prevalence was 51.2%, and the median was 49.7% with cohorts ranging from 9.8% to 92.4%. The average D1SA prevalence was 32.6%, with a median of 30.0% and a range from 3.5% to 83.5%. The average prevalence of D1S + A was 23.9% with a median of 21.6% and ranged from 2.5% to 71.9%. For D2, the average prevalence of D2TA was 10.6%, with a median of 9.5%, and a range of 0% to 60.2%. Average prevalence of D2CL was 27.9%, and the median was 26.8% ranging from 0% to 57.1%. For D2AdL, average prevalence was 22.6%, with a median of 21.4%, and cohorts ranging from 0% to 73.1%.

Table 2.

Descriptives of cohort-level population demographics and the proportion of liver abscess by score.

Category	Count	Percent total	Category	Count	Percent total
Dataset 1 (n = 330 cohorts, 59,210 head)			Dataset 2 (n = 386 cohorts, 46,884 head)
Sex			Sex
Heifer	178	53.9%	Heifer	203	52.6%
Steer	151	45.8%	Steer	183	47.4%
Days on feed categories			Days on feed categories
< 200	13	3.9%	< 200	53	13.7%
200–249	92	27.9%	200–249	124	32.1%
250–299	155	47.0%	250–299	130	33.7%
≥ 300	70	21.2%	≥ 300	79	20.5%
Shipping weight categories (kgs)			Shipping weight categories (kgs)
< 589	13	3.9%	< 589	67	17.4%
589–634	62	18.8%	589–634	124	32.1%
635–679	123	37.3%	635–679	106	27.5%
680–725	95	28.8%	680–725	81	21.0%
≥ 726	37	11.2%	≥ 726	8	2.1%
Average daily gain, kgs/day			Average daily gain, kgs/day
< 1.13	19	5.8%	< 1.13	35	9.1%
1.13–1.24	60	18.2%	1.13–1.24	58	15.0%
1.25–1.35	106	32.1%	1.25–1.35	97	25.1%
1.36–1.46	72	21.8%	1.36–1.46	104	26.9%
≥ 1.47	73	22.1%	≥ 1.47	92	23.8%
Percent of the cohort raileda			Percent of the cohort raileda
< 0.5	116	35.2%	< 0.5	226	58.5%
0.5–0.99	70	21.2%	0.5–0.99	47	12.2%
1.0–1.49	56	17.0%	1.0–1.49	47	12.2%
1.49–1.99	34	10.3%	1.49–1.99	15	3.9%
≥ 2.0	54	16.4%	≥ 2.0	21	5.4%
Region of origin			Region of origin
1	98	29.7%	1	3	0.8%
2	41	12.4%	2	259	67.1%
3	81	24.5%	3	44	11.4%
4	107	32.4%	4	66	17.1%
5	3	0.9%	5	14	3.6%
Total livers evaluated	59,210		Total livers evaluated	46,884
Total LAb	30,332	51.2%	Abscessed liversc	4985	10.6%
Severe LAd	19,285	32.6%	Condemned liversc	13,069	27.9%
Severe + LAe	14,156	23.9%	Adhered liversc	10,605	22.6%

Operational data was pulled from feedlot databases and harvest data was collected in the processing plant between March 2024 and June 2025.

^aA “railed” animal is sent to harvest prior to the rest of its cohort because it is not expected to finish with its penmates.

^bTotal LA consists of all LA scores: A-, A, A+, A + Adhered, A + Open, A + Adhered/A + Open.

^cLiver score categories from Dataset 2 are not mutually exclusive.

^dSevere LA consists of A+, A + Adhered, A + Open, A + Adhered/A + Open.

^eSevere + LA consists of A + Adhered, A + Open, A + Adhered/A + Open.

Model results: Dataset 1

Backwards step-wise regression was used to generate the final GLMM that evaluated the probabilities of four different LA outcomes (D1TA, D1SA, D1S + A, and D1ST) being associated with sex of cohort, DOF, shipping weight, head received, ADG—deads out, mortality, railers, daily feed delivered per head, quarters of arrival and harvest, and region of origin. For D1TA, the final model included sex of cohort, DOF, shipping weight, railers, and region of origin. The D1SA model included sex, DOF, shipping weight, railers, and region of origin. Similarly, the D1S + A included all of these variables as well as quarter of harvest. For D1ST, only sex of the cohort and region of origin remained in the final model. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for predictors retained in the final models (Supplemental File 1, available as supplementary data at Translational Animal Science online).

Steer cohorts had a greater risk (p ≤ 0.01; Fig. 1) compared to heifer cohorts for D1TA (52.4 ± 4.5% vs. 46.1 ± 4.3%, respectively), D1SA (35.4 ± 3.5% vs. 29.6 ± 4.3%), D1S + A (22.9 ± 3.6% vs. 18.4 ± 2.9%), and D1ST (71.6 ± 4.7% vs. 65.7 ± 5.2%). Days on feed was associated with (P ≤ 0.02) D1TA, D1SA, and D1S + A scores where cattle on feed for more than 300 days had a greater risk than cattle on feed 250 to 299 days, 200 to 249 days, and less than 200 days (Fig. 2). Shipping weight was associated with (P ≤ 0.04) D1TA, D1SA, and D1S + A scores with cohorts between 680 and 725 kgs having a greater risk (D1TA: 53.3 ± 4.4%; D1SA: 37.6 ± 4.9%; D1S + A: 23.6 ± 3.6%) of such scores than cohorts with shipping weights of 726 + kgs (D1TA: 45.4 ± 4.7%; D1SA: 29.5 ± 4.6%; D1S + A: 18.8 ± 3.2%).

Figure 1.

The association between cohort sex and various LA outcomes was analyzed for two datasets: Dataset 1 (n = 330 cohorts) and Dataset 2 (n = 386 cohorts). Each panel shows the probability of a LA outcome: A = Total LA from Dataset 1 (D1TA), B = Severe LA from Dataset 1 (D1SA), C = Severe + LA from Dataset 1 (D1S + a), D = Severe LA as a proportion of Total LA (D1ST), and E = Adhered livers from Dataset 2 (D2AdL). The white numbers represent the probability of that outcome, and the black bars represent the standard error of the mean. Columns within each graph with different letters denote statistical differences (P ≤ 0.05).

Figure 2.

The association between DOF and various LA outcomes was analyzed for two datasets; however, only outcomes from Dataset 1 (n = 330 cohorts) were significantly associated with DOF. Each panel shows the probability of a LA outcome: A = Total LA, B = Severe LA, and C = Severe + LA. The white numbers represent the probability of that outcome, and the black bars represent the standard error of the mean. Columns within each graph with different letters denote statistical differences (P ≤ 0.05).

Percentage of cohort railed was associated (P ≤ 0.01) with D1TA, D1SA, and D1S + A. For D1TA and D1S + A, cohorts with 0.5% to 0.99% (D1TA: 50.5 ± 4.5%; D1S + A: 21.6 ± 3.4%), 1.0% to 1.49% (D1TA: 51.1 ± 4.5%; D1S + A: 22.2 ± 3.5%), and 1.5% to 1.99% (D1TA: 50.2 ± 4.6%; D1S + A: 21.7 ± 3.5%) of the arrival head railed had a greater risk of LA outcomes compared to cohorts with less than 0.5% of the arrival head railed (D1TA: 44.9 ± 4.2%; D1S + A: 17.5 ± 2.8%). Cohorts with 0.5 to 0.99 (33.2 ± 4.7%), 1.00 to 1.49 (34.3 ± 4.9%), 1.50 to 1.99 (34.2 ± 4.9%), and greater than 2% (32.3 ± 4.7%) of arrival head railed had greater risk of D1SA compared to cohorts with fewer than 0.5% railed (28.1 ± 4.2%).

Quarter of harvest was associated with (P = 0.03) D1S + A. Cohorts harvested in quarter 4 had a lower risk (17.0 ± 3.1%) of D1S + A compared to cohorts harvested in quarter 2 (21.9 ± 3.2%) or quarter 3 (23.4 ± 3.5%). Region of origin was associated (P ≤ 0.05) with D1TA, D1SA, D1S + A, and D1ST where calves originating from region 3 (AZ, NM, OK, TX) had a lower risk of LA outcomes (D1TA: 43.6 ± 4.1%, D1SA: 27.4 ± 4.0%, D1S + A: 18.1 ± 2.8%, D1ST: 64.2 ± 5.2%) compared to calves from regions 1 (CA, ID, NV, OR, WA; D1TA: 49.8 ± 4.1%, D1SA: 33.0 ± 4.4%, D1S + A: 21.3 ± 3.1%, D1ST: 68.2 ± 4.9%), 4 (IA, IL, IN, MI, MN, MO, OH, WI; D1TA: 52.2 ± 4.1%, D1SA: 34.4 ± 4.5%, D1S + A: 22.6 ± 3.2%, D1ST: 69.7 ± 4.7%) and region 2 (CO, KS, MT, ND, NE, SD, UT, WY) for D1SA (32.7 ± 4.7%; Fig. 4).

Figure 4.

The association between region of origin and various LA outcomes was analyzed for two datasets: Dataset 1 (n = 330 cohorts) and Dataset 2 (n = 386 cohorts). Each panel shows the probability of a LA outcome: A = Total LA from Dataset 1, B = Severe LA from Dataset 1, C = Severe + LA from Dataset 1, D = Severe LA as a proportion of Total LA from Dataset 1, and E = Abscessed livers from Dataset 2. No cohorts originated from Region 6 (the Northeastern states). Regions within each graph with different letters denote statistical differences (P ≤ 0.05).

Feedyard was included as a random effect in the model for each outcome, and the marginal and conditional R² report how much of the variance was explained by fixed effects or fixed and random effects, respectively. For D1TA, the marginal R² was 0.29, and the conditional R² was 0.84. This means that including feedyard as a random effect improved the explanation of the variance by 55.2 percentage units. For D1SA, the marginal R² was 0.22 and conditional R² was 0.81, meaning the inclusion of feedyard as a random effect explained 58.9 percentage units more variation than the inclusion of fixed effects alone. For D1S + A, the marginal R² was 0.24 and the conditional R² was 0.74. Including feedyard as a random effect improved the explanation of the variance by 49.6 percentage units. The effect of including feedyard as a random effect was more pronounced for D1ST with a marginal and conditional R² of 0.06 and 0.97, respectively. These values indicate that there were differences between feedyards that were not adequately captured within the collected operational data.

Model results: Dataset 2

Outcomes of interest for D2 were proportion of abscessed livers, condemned livers, and adhered livers which were assessed using the same model-building process used for D1. For D2TA, shipping weight, ADG, quarter of harvest, and region of origin were retained in the final model. For D2CL, head received, ADG, and quarter of harvest made up the final model. The final model for D2AdL included only sex of the cohort. Odds ratios (OR) and 95% confidence intervals (CI) were calculated for predictors retained in the final models (Supplemental File 1, available as supplementary data at Translational Animal Science online).

Steer cohorts had a greater (P < 0.01; Fig. 1) risk of D2AdL (27.4 ± 1.5%) compared to heifer cohorts (20.5 ± 1.3%), but sex was not associated with D2TA or D2CL. Shipping weight was associated with (P < 0.01) D2TA where cohorts with an average shipping weight less than 589 kgs had a lower probability of LA (5.0 ± 1.7%) than cohorts between 680 and 725 kgs (9.1 ± 1.5%) and those between 635 and 679 kgs (10.0 ± 1.4%). The number of head received was associated with (P = 0.03) D2CL where cohorts of more than 250 head had a lower risk (16.8 ± 4.6%) compared to cohorts between 200 and 249 head (30.0 ± 2.9%), 150 and 199 head (32.0 ± 2.9%), and less than 150 head (30.5 ± 2.7%). Average daily gain was associated with (P ≤ 0.02) D2TA and D2CL. For D2TA, cohorts with an ADG between 1.36 and 1.46 kg/d had a lower risk (6.6 ± 1.1%) compared to cohorts between 1.13 and 1.24 kg/d (9.1 ± 1.6%), and were not different from cohorts with ADG between 1.25 and 1.36 kg/d (7.6 ± 1.3%), less than 1.13 (6.6 ± 1.3%), and equal to or greater than 1.47 kg/d (7.8 ± 1.3%; Fig. 3). For D2CL, cohorts with ADG between 1.36 and 1.46 kg/d had a lower risk (24.4 ± 2.8%) compared to cohorts with an ADG between 1.13 and 1.24 kg/d (29.2 ± 3.3%) and between 1.25 and 1.35 kg/d (28.3 ± 3.1%).

Figure 3.

The association between ADG and various LA outcomes was analyzed for two datasets; however, only outcomes from Dataset 2 (n = 386 cohorts) were significantly associated with ADG. Each panel shows the probability of a LA outcome: A = Abscessed livers (D2TA), and B = Condemned livers (D2CL). The white numbers represent the probability of that outcome, and the black bars represent the standard error of the mean. Columns within each graph with different letters denote statistical differences (P ≤ 0.05).

Quarter of harvest was associated with (P ≤ 0.01) D2TA and D2CL. Cohorts shipped to harvest in quarter 3 had a lower risk of D2TA (6.3 ± 1.1%) compared to cohorts harvested in quarter 4 (8.9 ± 1.4%) For D2CL, cattle shipped to harvest in quarter 3 had the lowest probability of the outcome (21.0 ± 2.6%), followed by quarters 2 (25.3 ± 2.9%) and 4 (28.3 ± 3.1%), and cohorts shipped in quarter 1 had the greatest risk of condemnation of livers (33.4 ± 3.5%).

Region of origin was associated with (P ≤ 0.01) D2TA where cohorts originating from regions 1 (CA, ID, NV, OR, WA) and 4 (IA, IL, IN, MI, MN, MO, OH, WI) had lower risks of the outcome (2.3 ± 1.6% and 7.6 ± 0.8%, respectively) compared to cohorts from regions 2 (CO, KS, MT, ND, NE, SD, UT, WY; 10.2 ± 0.08%), 3 (AZ, NM, OK, TX; 10.2 ± 1.1%) and 5 (AL, AR, FL, GA, KY, LA, MS, NC, SC, TN, VA, WV; 12.1 ± 2.0%).

In D2, random effect of feedyard contributed little to the explained variation compared to D1 models. For D2TA and D2AdL, variation among feedyards was negligible, with essentially no difference in marginal and conditional R². The marginal R² for the D2TA model was 0.30, and the reported conditional R² was 0.30 due to the extremely small effect of feedyard as a random effect. The marginal R² for the D2AdL model was 0.15, and the conditional R² was 0.20 indicating that feedyard did not help explain variation in the data for this outcome. For D2CL livers, the inclusion of feedyard as a random effect did improve the amount of variation explained; marginal R² was 0.26 and conditional R² was 0.53, resulting in an additional 26.3% of the variation explained.

Discussion

As the number of B × D cattle harvested continues to grow, the economic losses due to liver abscesses can be expected to increase as well. Identifying potential risk factors that are associated with LA outcomes may provide producers with management practices to lower prevalence of LA at harvest. The results of this research indicated that sex, DOF, shipping weight at harvest, the number of head received upon arrival to the feedyard, ADG, railers, quarter of harvest and region of origin were associated with various LA outcomes from D1 and D2. Dataset 1 consisted of mutually exclusive liver scores which were used to calculate four outcomes of interest (D1TA, D1SA, D1S + A, and D1ST). The scoring system used for D2 had 3 non-mutually exclusive categories that were used as outcomes (D2TA, D2CL, D2AdL). Dataset 1 had a larger prevalence of LA compared to D2; however, this is most likely due to the differences in scoring systems used.

Dataset 1 consisted of cohorts from 8 feedyards located in the High Plains which were included as a random effect in analyses. Dataset 2 consisted of cohorts from 5 feedyards located in the High Plains. The differences in marginal and conditional R² provided by feedyard as a random effect varied between the outcomes and the data sets (D1 vs D2). In D1, the difference in the marginal and conditional R² values ranged from 49.6 to 91.0 percentage units. This range of difference between the R² values indicated that inclusion of feedyard as a random effect accounted for substantial additional variation in the outcomes beyond that explained by the fixed effects. This suggests that important sources of variation in liver abscess outcomes may exist among feedyards that were not captured by the operational variables evaluated in this study. Such variation could reflect differences in management practices, cattle populations, feeding programs, or other feedyard-level characteristics that were not measured or included in the models.

The effect of feedyard was not as strong for D2 ranging from negligible to 26.3 percentage units, indicating the feedyard-level factors did not contribute as strongly to variation in these data. This could be due to several differences between the datasets. First, the datasets included different methods of data collection which consisted of non-mutually exclusive groups of abscessed, condemned, and/or adhered livers in D2 compared to mutually exclusive liver scores in D1. Second, D1 and D2 consisted of data from different feedyards, and therefore represented different sources of between-feedyard variation. Finally, D2 had a lower average prevalence and narrower range of liver abscess outcomes than D1 which may have reduced the amount of variation between feedyards captured by the random effect. Either way, these data indicate considerable differences among feedyards that was not captured in the predictor variables evaluated in this study and warrants further research.

Similar to the results of the current study, previous literature has reported similar findings that Native and Holstein steers are more likely to have LA than heifers (Smith 1998; Amachawadi and Nagaraja 2016; Grimes et al. 2024; Champagne et al. 2025). However, the previous findings were reported for Native, Holstein, and B × D cattle whereas the present study consisted of only B × D cattle. The increased risk for steer cohorts is thought to be related to increased dry matter intake (DMI) as well as longer DOF compared to heifers as steers eat greater quantities of feed for a longer period of time, there would be more opportunities for rapid fermentation of large quantities of carbohydrates to occur leading to frequent acidotic events (Amachawadi and Nagaraja 2016). This could contribute to increased ruminal epithelial insult allowing for bacteria to escape and translocate to the liver to form LA (Nagaraja and Lechtenberg 2007). However, we did not find an association between average daily feed delivered per head and any LA outcome.

The finding that cattle on feed for longer had a higher risk of LA outcomes is supported by previous literature. Champagne et al. reported in feedyard mortalities that cattle on feed for longer than 100 days had a higher probability of LA with cattle ranging from 1 to 226 DOF (Champagne et al. 2025). In the present data, DOF ranged from 176 to 402 for D1 and 148 to 421 for D2. Dairy and B × D cattle may arrive at a feedyard weighing less than traditionally raised beef animals and thus must be on a high-grain finishing diet for a longer duration (Reinhardt and Hubbert 2015; Amachawadi and Nagaraja 2016). Thus, the total number of days B × D cattle are fed high-concentrate diets may result in greater ruminal wall insult due to prolonged exposure (Nagaraja and Lechtenberg 2007). Chronic acidosis and rumenitis could lead to more LA formation due to their role in the pathogenesis of LA as discussed above. Shipping weight was also significantly associated with the risk of multiple LA outcomes; however, the probability of LA was higher for cohorts between 680 and 725 kgs compared to 726 + kgs. This was surprising, as one would expect the cattle heaviest at shipping to have the highest risk of LA due to prolonged DOF and higher intake. But there were fewer cohorts with an average shipping weight of 726 + kgs and thus should be interpreted carefully.

The number of head in a cohort was associated with D2CL where cohorts of fewer cattle had higher risk than those with more. The biological relevance of this finding is unknown, and to the authors’ knowledge, there is no literature assessing the relationship of cohort size and LA outcome. An explanation could be lack of competition for bunk space in smaller cohorts. This would allow cattle to spend more time at the bunk and potentially increase feed intake which could contribute to more subacute rumen acidosis events. However, previous literature has not found differences in feed intake when comparing various bunk space allowance (Harrison and Oltjen 2021; Duncan et al. 2022).

Average daily gain was also found to be associated with D2TA and D2CL (Fig. 3). Cohorts with an ADG between 1.13 and 1.24 kgs/d had a higher risk of D2TA than cohorts with an ADG between 1.36 and 1.46 kgs/d and cohorts with an ADG less than 1.13 kgs/d. For D2CL, cohorts with an ADG between 1.13 and 1.35 kgs/d had a higher risk than those between 1.36 and 1.46 kgs/d. Fox et al. (2009) reported the presence of liver abscess at harvest tended to decrease ADG. Similarly, another study reported reduced ADG with severe LA (A+) and severe rumenitis lesions found at harvest (Rezac et al. 2014). Due to the observational nature of this dataset, it is impossible to determine whether the presence of LA reduced ADG or if a reduction in ADG caused LA. Additionally, there is no way to know when LA develop or resolve prior to harvest.

The percentage of the cohort that was railed was associated with D1TA, D1SA, and D1S + A scores. Cohorts with fewer than 0.5% of cattle railed had a lower risk of these outcomes. A cohort with increased railers likely indicates increased incidence of BRD and potentially subclinical BRD. Cohorts previously suffering from BRD may be less capable of combating pathogenic bacteria entering the gastrointestinal epithelium leading to liver abscess. The authors hypothesized that cohorts with a greater burden of liver infection or abscesses may experience greater overall disease burden, potentially contributing to increased susceptibility to other diseases such as bovine respiratory disease (BRD). Additionally, a higher prevalence of BRD within a cohort may contribute to an increased proportion of animals being railed, as animals exhibiting clinical disease may be removed from the cohort rather than risk mortality.

Quarter of harvest was associated with outcomes from both datasets: D1S + A, D2TA, and D2CL. For D1S + A, the risk was higher for those harvested in quarters 2 and 3 compared to quarter 4. Risk of D2TA and D2CL was highest for cohorts harvested in quarter 1 or 4, intermediate for quarter 2, and lowest for quarter 3. Previous literature reports mixed results regarding seasonality of LA. Brown and Lawrence (2010) did not observe a relationship between harvest month and abscess prevalence on an individual-animal basis, but other studies have found that LA and/or other abnormalities were highest in quarters 2 and 3 (Harman et al. 1989; Brown and Lawrence 2010; Grimes et al. 2024). Grimes et al. (2024) reported total liver abscess risk on an individual-animal basis was lowest in January (Quarter 1) for both datasets analyzed and highest in August (Quarter 3) and April (Quarter 2; database 1) and May and June (Quarter 2; database 2). Another study had similar findings with highest liver abnormalities reported in April and June (Quarter 2) on a cohort-level basis (Harman et al. 1989). The risk for condemned livers was highest in January (Quarter 1) which disagrees with Grimes et al.’s findings; however, this could be due to population differences. The present data includes data from 2021 to 2025 that was analyzed as success/failures on a cohort-level and consists only of B × D cattle across 3 processing plants and 11 feedyards, whereas Grimes’ et al. data were collected on an individual-animal basis representing 4 breed types (Native, B × D, Holstein, and Mexican origin) across 26 processing plants and 138 different feedyards. Based on the present data and previous literature, there does not appear to be a consistent pattern for recorded liver abnormalities found at harvest; however, this could be due to seasonality of calf placement depending on breed. Beef-on-dairy calves enter feedyards year-round whereas native beef calves are largely placed seasonally depending on calf age and backgrounding. This inconsistency highlights the need for additional research to elucidate the underlying risk factors and mechanisms driving these findings.

Region of origin was associated with D1TA, D1SA, D1S + A, D1ST, and D2TA (Fig. 4). For all D1 outcomes, cohorts of cattle that originated from region 3 (AZ, NM, OK, TX) had a lower risk than cohorts from regions 1 (CA, ID, NV, OR, WA) and 4 (IA, IL, IN, MI, MN, MO, OH, WI). Regions 2 and 5 were not different from the other regions for D1TA, D1S + A, and D1ST, but region 2 (CO, KS, MT, NE, ND, SD, UT, WY) had higher risk than region 3 for DISA. Conversely, for D2TA, cohorts originating from regions 2, 3, and 5 had higher risk compared to those from regions 1 and 4. This variable was included as a proxy for commercial calf ranch of origin; however, the origin recorded in the operational data could have corresponded to a calf ranch, backgrounding operation, or auction market. Thus, the results should be interpreted with caution. To the authors’ knowledge, there is no previous literature that describes LA based on the region where the cattle were raised prior to the feedyard. Grimes et al. (2024) analyzed LA by processing plant and feedyard location and attributed differences to variation in grain type and processing method as well as environmental factors. In the present data, all the processing plants and feedyards were in the High Plains region, thus the authors would not expect to see differences associated with locations where cattle were finished and harvested. The authors believe these data provide evidence that management of calves prior to arrival to the feedyard (ie at calf ranches, backgrounding operations, and/or auction market systems) may influence outcomes later in an animal’s life and warrants further research.

Conclusion

Liver abscesses are a major economic loss to the beef industry, particularly in beef-on-dairy cattle due to higher prevalence. The cause of increased prevalence is not entirely known, and there is speculation that management of beef-on-dairy calves throughout their production lifecycle may be contributing to increased LA compared to native beef breeds. This study identified associations between LA prevalence at harvest and cohort-level factors such as cattle demographics, origin, and performance and health metrics during the finishing phase. A large amount of variation among feedyards was also observed in these data and provide evidence that there may be variables that are currently not measured or recorded in standard operational data that contribute to increased prevalence of LA. There is still a large need for further exploration of management strategies to mitigate the risk of LA. A more comprehensive understanding of both early-life and finishing-phase management practices will be essential to develop effective, evidence-based strategies for reducing live abscess prevalence in beef-on-dairy production systems.

Glossary

List of abbreviations

Ad

Adhered

ADG

average daily gain

B × D

Beef-on-dairy

D1TA

Dataset 1 Total Abscesses

D1SA

Dataset 1 Severe Abscesses

D1S + A

Dataset 1 Severe + Abscesses

D1ST

Dataset 1 Severe as proportion of Total Abscesses

D2TA

Dataset 2 Liver Abscesses

D2CL

Dataset 2 Condemned Livers

D2AdL

Dataset 2 Adhered Livers

DOF

days on feed

LA

liver abscess(es)

Op

Open

SI

small intestine

Supplementary data

Supplementary data is available at Translational Animal Science online.

Funding

This research was funded by the International Consortium for Antimicrobial Stewardship in Agriculture—Foundation for Food and Agriculture Research (ICASA- FFAR), grant number 22-000448.

Author contributions

All authors contributed to the study in at least two of the various categories: conceptualization, funding acquisition, investigation, methodology, data collection, validation, project administration, supervision, writing–original draft, writing–reviewing and editing, and/or resources.

Conflicts of interest

None declared.