Pegbovigrastim treatment resulted in an economic benefit in a large randomized clinical trial in grazing dairy cows

This randomized controlled trial on 4 commercial grazing dairy farms investigated whether pegbovigras-tim (PEG) treatment affected partial net return as calculated from milk revenues and costs for feed, medical treatments [clinical mastitis, uterine disease, and other diseases (i.e., any medical treatment that was not intended for clinical mastitis or uterine disease)], in-seminations, and culling during a full lactation in grazing dairy cows. We also explored the effect of potential interactions of PEG treatment with parity, prepartum body condition score, and prepartum nonesterified fatty acids concentration on partial net return, milk revenues, and the costs mentioned above. Holstein cows were randomly assigned to 1 of the 2 following trial arms: a first PEG dose 9.4 ± 0.3 (mean ± standard error) days before the calving date and a second dose within 24 hours after calving (PEG: primiparous = 342; multiparous = 697) compared with untreated controls (control: primiparous = 391; multiparous = 723). The effect of PEG treatment on the outcomes of interest expressed per year was tested using general linear mixed models. Results are presented as least squares means ± standard error. Overall, PEG treatment increased the partial net return, resulting in an economic benefit per cow per year of $210 ± 100. The cost of treat-ment of clinical mastitis was lower for PEG treated cows compared with control cows ($9 ± 3). The largest nonsignificant difference was seen for the cost of culling; additionally, PEG treatment numerically reduced the cost of culling by $145 ± 77.


INTRODUCTION
Animal diseases associated with the transition period in dairy cows, such as clinical mastitis (CM), uterine disease (UD), and metabolic diseases, have a negative effect on health, welfare, and the economic performance of dairy farming (Overton and Fetrow, 2008;Hogeveen et al., 2019;Steeneveld et al., 2020). Production loss, discarded milk during the withdrawal period after medical treatment, and the actual cost of treatments are important direct economic effects of early lactation diseases Pérez-Báez et al., 2021). Moreover, indirect effects of these diseases can be seen in impaired reproductive performance (Fourichon et al., 2000;LeBlanc et al., 2002;Dolecheck et al., 2019) and culling (Kossaibati and Esslemont, 1997;Bar et al., 2008;Carvalho et al., 2019). Multiple studies have shown that culling makes an important contribution to the costs of clinical disease (Galligan, 2006;Overton and Fetrow, 2008;Heikkilä et al., 2012).
Pegbovigrastim, a long-acting analog of bovine granulocyte colony-stimulating factor (PEG, marketed as Imrestor by Elanco Animal Health), has been used as a tool to improve dairy cows' immune response during the transition period. Pegbovigrastim treatment consistently increases circulating white blood cell counts (McDougall et al., 2017;Van Schyndel et al., 2018;Barca et al., 2021a). Moreover, PEG treatment improves immune function at plasma cytokine profile and white blood cell gene expression level (Lopreiato et al., 2019(Lopreiato et al., , 2020. Several field trials have evaluated the effect of PEG treatment, mainly on early lactation clinical disease. Several these studies also report milk production, fertility, and culling (Canning et al., 2017;Ruiz et al., 2017;Freick et al., 2018;Zinicola et al., 2018;Van Schyndel et al., 2021). A lower early lactation CM incidence due to PEG treatment was reported by Canning et al. (2017) and Ruiz et al. (2017), but 2 studies (Zinicola et al., 2018;Van Schyndel et al., 2021) reported no effects on early lactation CM incidence. Overall, PEG treatment results ranged from a preventive effect for early lactation clinical disease (Canning et al., 2017;Freick et al., 2018) to increased morbidity (Zinicola et al., 2018), particularly for metritis (Ruiz et al., 2017). Recently, based upon results from a large clinical trial, we reported that PEG reduced the occurrence of a first case of CM during the first 30 DIM, particularly in cows with excessive prepartum BCS, and in cows with elevated prepartum nonesterified fatty acids (NEFA) concentrations (Barca et al., 2021b). Moreover, PEG treatment reduced the occurrence of endometritis in cows that had a previous case of metritis in the same lactation. In addition, in cows with elevated NEFA, PEG treatment increased the first insemination rate and counteracted the negative association of early lactation CM and UD [i.e., a cow with a record of retained fetal membranes (placenta), metritis, or both] with the pregnancy rate. Ultimately, in cows with elevated NEFA, PEG treatment decreased the hazard of culling .
An important question to answer is whether the economic benefits of improved health, fertility, and longevity due to PEG treatment outweigh the cost of PEG and its application under field conditions (2 doses). Bio-economic simulation models are often used for this type of economic evaluations. In such bioeconomic models, the available knowledge on the effect of an intervention is implemented in a simulation model of the relevant dairy cow disease or diseases, and the model is consequently used to mimic the use of the intervention and evaluate the economic effect of the intervention compared with a situation without it. This approach was used in recent literature to evaluate the economic impact of different durations of CM treatment (Pinzón-Sánchez and Ruegg, 2011;Steeneveld et al., 2011), the economic benefit of using nonsteroidal anti-inflammatory drugs in the treatment of CM (van Soest et al., 2018), and the economic impact of implementing selective dry cow therapy (Hommels et al., 2021). However, this bio-economic simulation approach does not account for heterogeneity between cows. If sufficient longitudinal data are present, another possible approach is to study the economic performance of control and treated cows in a randomized clinical trial. In such longitudinal studies, the partial net return for each cow in the study may be determined (Burgers et al., 2022). In this report, the hypothesis to be tested is whether PEG treatment increases partial net return as calculated from milk revenues and costs for feed, medical treatments, inseminations, and culling during a full lactation in grazing dairy cows, thereby taking the interaction with BCS and NEFA status of cows into account.
The objective of this study is, therefore, to investigate whether PEG treatment affects partial net return, milk revenues and costs for feed, medical treatments, insemination, and culling during a full lactation. We also explored the effect of potential interactions of PEG treatment with parity, BCS, and NEFA on these outcomes.

MATERIALS AND METHODS
The experimental protocol (CEUAFVET-PI-162) was evaluated and approved by the Honorary Committee for Animal Experimentation in Uruguay, University of the Republic, Uruguay.

Study Design
This randomized controlled trial was carried out on 4 commercial grazing dairy farms and has recently been described in companion publications (Barca et al., 2021b. In short, 2,153 (farm 1 = 759; farm 2 = 314; farm 3 = 664; farm 4 = 416) Holstein primiparous (animals that were enrolled in the study shortly before their first calving) and multiparous cows (animals that were enrolled shortly before their second or later calving) that calved from February 13 to September 30 of 2018 were included in the study. Twice a week, at −10 to −7 d relative to the expected calving date, cows were randomly assigned to either treatment group or untreated control group based on their unique national ear tag number, which is assigned to cattle at birth. Animals with an even national ear tag number were injected with 15 mg of PEG according to the product label (PEG) and animals with an odd national ear tag number remained as untreated controls (control). Animals assigned to the PEG treatment received a second dose within 24 h after calving; however, only cows that received both doses were included in the study. The national ear tags used for treatment allocation are not related to the large visible ear tags that are used for on-farm identification and management decisions. Research assistants, who did not belong to the farm staff, applied treatments (Barca et al., 2021b. This meant that all people involved in daily farm management were blinded as to which cows had been treated. The resulting study population consisted of 733 primiparous cows (control = 391; PEG 342) and 1,420 multiparous cows (control: 723; PEG = 697). The interval in days between enrollment and calving (in case of PEG this is the interval between PEG doses), did not differ between treatment groups (control = 9.1 ± 0.2; PEG = 9.4 ± 0.3; P = 0.42).
At the time of treatment assignment (−10 to −7 d relative to the expected calving date) BCS was assessed according to Ferguson et al. (1994) and blood was drawn for determination of NEFA (Barca et al., 2021b. Throughout the study, milk yield and fat and protein concentration per cow were determined from monthly test-day samples [fat and protein determinations were performed by Cooperativa Laboratorio Veterinario de Colonia (COLAVECO)]. Clinical diseases were diagnosed and recorded as described before (Barca et al., 2021b, and all treatments were recorded. All farms used AI with estrus detection performed by trained farm personnel. Pregnancy diagnoses were performed by transrectal palpation or ultrasonography by the farm veterinarian. All inseminations and pregnancy diagnoses were recorded by farm personnel or by the farm veterinarian, or both. Dry-off date and date of culling or on-farm death were also recorded. The study ended 305 d after the last recorded calving in the study, which was 529 d after the first recorded calving in the study.

Economic Calculations, Partial Budgeting
The collected data were used to determine partial net return for each of the 2,153 cows in the trial. Consequently, the cow is the economic unit of interest. Partial net return is defined as combination of milk revenues where all relevant costs are subtracted for each cow i during the experimental days (ED) for cow i. Partial budgeting is widely used to evaluate the economic impact of interventions (for example Rowe et al., 2021) and partial net return as used here was previously defined by Burgers et al. (2022). Equal to a partial budget, the cow-level partial net return only takes into account economic factors that are, in theory, affected by the intervention of interest. Consequently, we based the partial net returns calculation on factors that could be potentially affected by PEG treatment. Briefly, milk revenues R i Milk ( ) were included as the income potentially affected by PEG treatment (Ruiz et al., 2017;Powell et al., 2018;Van Schyndel et al., 2021). Although feed costs in relation to PEG has not been studied, feed intake in relation to PEG treatment has been studied preliminarily, in a CM challenge model, and Powell et al. (2018), reported that post infection, PEG treated cows consumed more feed than untreated control cows. Therefore, costs for feed C i Feed ( ) were included in the partial net return. Pegbovigrastim treatment is associated with the occurrence of diseases and reproductive performance (Canning et al., 2017;Ruiz et al., 2017;Zinicola et al., 2018;Barca et al., 2022). Therefore, expenditures for treatment of CM C i CM ( ) , UD C i UD ( ) ,and treatments for other diseases C i Other ( ; i.e., any medical treatment that was not intended for CM or UD) and cost of inseminations C i Ins ( ) were included. Finally, because diseases are known to be associated with hazard of culling (Carvalho et al., 2019), and that in cows with elevated NEFA, we showed that PEG treatment decreased the hazard of culling . Also, cost of culling C i Cull ( ) was included. As far as we know, there are no indications that other cow-level costs are affected by PEG treatment. Consequently, the partial net return for each individual cow i was calculated as follows: Because ED differed between cows, partial net return i was standardized and expressed per cow per year (Burgers et al., 2022) as follows: Partial net return Partial net return ED

Milk Revenues and Costs Calculations
From the monthly test-day milk samples, milk returns were calculated per individual cow as follows: first, for each cow i and test-day j, the returns for milk Based on the returns per cow per test-day, the total milk returns per cow R i Milk ( ) were calculated for each cow i by multiplying the average milk returns in the period leading to the test-day by the number of days in that period as follows: Barca et al.: PEGBOVIGRASTIM: ECONOMIC EFFECT where DIM ij is the DIM for test-day 1 to n j (n j is total number of test-days of cow i). We assumed a linear change of the daily milk returns between test-days. For the time until the first test-day of a cow, the milk return was taken as the average of the first test-day postpartum (DIM i1 ) and 0, thereby assuming that milk yield started at 0 kg. For the period after the last testday (DIM in ), it was assumed that the milk returns were equal to the milk returns at the last test-day with a period duration from the last test-day to exit from the study for each cow i. Costs for feed supply were estimated per individual cow as follows: where NE Maintenance is net energy for maintenance, considering a 20% increase for grazing activity in mixed systems (NRC, 2001) Milkwithheld is the number of days of treatment and milk withhold, MY ic is milk yield at the time of CM case c for cow i, and Perc ic Prot and Perc ic Fat are percentage protein and fat at the time of a CM case c for cow i, respectively.

C i
Medicine is the price for medicine, C i Labor is the labor price for treatment application, and C i Milkfedtocalves is MY ic by the value of the milk replacer, all for cow i.
Similarly, C i UD and C i Other were calculated for each cow i.
Costs for insemination C i Ins ( ) were calculated per individual cow as follows: where N i Ins is the number of inseminations for each cow i and P Ins is the price per insemination including costs of labor.
The C i Cull was calculated similar to Mostert et al. (2018) and Burgers et al. (2022) as where P Heifer is the price of a replacement heifer and P Slaughter is the revenue of a culled cow at the slaughterhouse, assuming a weight of 550 kg (INALE, 2021). In case of a dead cow, the P Slaughter is equal to 0. The variable L Max is the lactation number of the oldest cow in the experiment and L i is the actual lactation of a culled cow i. More than 5 lactations were considered as 6 lactations; thus, L Max is equal to 6 and L i is an integer number from 1 to 6. Costs of PEG and its application (2 doses) were not included in the economic calculations.

Prices
Input values regarding price levels (Table 1) were based on prices for 2018, available at the governmental national institute for dairy production (INALE, 2021). Prices for medicines and milk replacer to feed calves were based on information supplied by one of the biggest medicine suppliers for dairy in Uruguay (PROLESA, 2021). Beef price at the slaughterhouse was based on available information at the Asociación de Consignatarios de Ganado (ACG, 2021). Prices were given in local currency and transformed to US dollars using the average currency exchange rate of 2018 (INALE, 2021).

Statistical Analysis
Data were analyzed using SAS software (SAS University Edition, SAS Institute Inc.).
The effect of PEG treatment on partial net return, R Milk , and on the costs expressed per cow per year was tested using general linear mixed models (PROC MIXED). The following were considered as class variables: parity (primiparous/multiparous), BCS (under: <3; acceptable: 3 to 3.5; over: >3.5; Roche et al., 2009), NEFA (low ≤0.3; high >0.3 mM, Overton et al., 2017), treatment (control/PEG), and calving month (6 classes: February/March, April, May, June, July, and August/September). Farm, also as a class variable, was included as a random effect. Two-way interactions between parity, BCS, NEFA, and treatment were checked for significance.
The general model was then After the initial full model lay-out, a backward variable selection process was performed. Parity and treatment were forced into all models. All other variables or their 2-way interaction with treatment remained in the model when P ≤ 0.10 (significance level to stay, SL-STAY). Exceptionally, variables or an interaction term remained in the model when removal of the variable resulted in an important change in the treatment effect. Such variables would be considered potential confounders. Statistical significance was decided at a level of P ≤ 0.05. The result of variables that remained in the final models are presented as least squares means (LSM) ± standard error. All P-values of pair-wise comparisons Barca et  of LSM were corrected with the Tukey-Kramer adjustment.

Sensitivity Analysis
To assess the effect of changes in revenues and costs on the partial net return per cow per year, a sensitivity analysis was performed. First, considering the variations observed in prices of protein and fat from 2012 to 2020 (INALE, 2021), milk revenues were based on either the average price for protein and fat for 2018, or on a 20% lower or higher price (Table 1). Second, it was assumed that milk withheld due to medical treatments would or would not be fed to calves. Third, we assumed either the normal or the lowest beef price at slaughterhouse. The lowest beef price is the price paid for cows in an inferior health condition (ACG, 2021; Table 1). Fourth, we varied the price of a replacement heifer from the normal price to a 20% higher price (INALE, 2021; Table 1). Table 2 presents an overview of ED, end-of-study status, productive performance, and economic factors for control and PEG cows. During the study, 4.4% of the cows died (control = 4.6%; PEG = 4.2%), 14.5% were culled (control = 15.1%; PEG = 13.9%), 53.1% were dried-off being pregnant (control = 53.3%; PEG = 53.0%), 10.5% were dried-off being not pregnant (assumed culled at that date; control = 9.7%; PEG = 11.3%), 8.0% ended the study period being pregnant (control = 8.9%; PEG = 7.0%), and 9.4% ended the study period being not pregnant (assumed culled at the end-of-study date; control = 8.4%; PEG = 10.5%).

Descriptive Statistics
Economic descriptive results, presented in Table 2, showed that R Milk per year were $2,401 ± 27 and $2,429 ± 26, in control and PEG treated cows respectively. The biggest cost was C Feed per year, which represented 36.9% of the R Milk per year in the control group and 36.4% in PEG cows. The C Cull per year represented 16.1% of the R Milk per year in control cows and 14.0% in PEG cows. A descriptive evaluation of C Cull per year showed some extremely high values (right skewed) that made statistical analysis difficult. Consequently, cows with a C Cull per year > mean + 2 × standard deviation were treated as outliers and not included in the final data set (Burgers et al., 2022) as presented in Table  2. Thus, control = 12 (death = 11, cull = 1), PEG = 13 (death = 12; cull = 1) cows, which were defined as outliers, were not considered for the final statistical analyses, resulting in several remaining cows of control Barca et  Cows that were not pregnant at the end-date were assumed to be culled at that date.
2 R Milk = returns for milk; C Feed = costs for feed; C Cull = cost of culling; C CM = cost of clinical mastitis; C UD = cost of uterine disease (cost of treatment for retained placenta, metritis or both); C Other = cost of any medical treatment recorded that was not intended for clinical mastitis, retained placenta, or metritis; C Ins = cost for insemination.
= 1,102 and PEG = 1,026 (Table 2). Descriptive data of cows treated as outliers is provided in Supplemental Table S2 (https: / / doi .org/ 10 .17632/ 7n77zzn2p2 .3; Barca, 2022). For these cows, the ED were 6 ± 2 and 5 ± 1, and the C Cull per year was $72,333 ± 23,731 and $42,446 ± 5,157, in control and PEG cows, respectively. After removing these 25 cows, C Cull per year still showed a relatively large variation (Table 2), but in the original data set, with a much higher mean value, this variation was considerably larger (control = 1,162 ± 334; PEG = 867 ± 162). The C CM , C UD , C Other , and C Ins per year represented 2% or less of the R Milk per year each.

Effect of Pegbovigrastim on Partial Net Return Per Cow Per Year
Final model results showed that PEG treatment affected partial net return per year (P = 0.036). Prepartum BCS was not associated with the partial net return per year (P = 0.24), but the BCS by treatment interaction remained in the model (P = 0.10). Table  3 presents treatment LSM for partial net return per year, overall, and by BCS class. Overall, PEG treatment increased the partial net return per year by $210 ± 100. Figure 1 illustrates PEG treatment effect on the partial net return per year, overall and by BCS class.

Effect of Pegbovigrastim on Milk Revenues and Costs Per Cow Per Year
Final model results of C Cull per year show that PEG treatment effect did not reach statistical significance (P = 0.061). Prepartum BCS was not associated with the C Cull per year (P = 0.65), whereas the removal of the BCS by treatment interaction (P = 0.12) resulted in an important change in the treatment effect and it was therefore maintained in the model. Table 3 presents treatment LSM for C Cull per year, overall, and by BCS class. Overall, PEG treatment numerically reduced the C Cull per year by $145 ± 77. Differentiated by parity, the C Cull per year was $453 ± 68 and 336 ± 53 (P = 0.07) in primiparous and multiparous cows, respectively.

Sensitivity Analysis
Supplemental Table S3 (https: / / doi .org/ 10 .17632/ 7n77zzn2p2 .3; Barca, 2022) provides an overview of the final regression models for partial net return per year assuming different input levels as used for the sensitivity analyses.
Least squares means and the treatment effect on partial net return per year by different variable levels as used in the sensitivity analyses are presented in Table 4. Compared with the price levels assumed in this study, a 20% lower price for milk protein and fat resulted in a 7% lower economic benefit of PEG treatment, whereas a 20% higher price showed a 6% higher economic benefit. Assuming that milk withheld due to medical treatments would not be fed to calves resulted in a 2% higher economic benefit of PEG treatment. Assuming the lowest beef price at slaughter resulted in a 10% higher economic benefit of PEG treatment. Assuming a 20% higher price of a replacement heifer showed a 22% higher economic benefit of PEG treatment. Barca

DISCUSSION
In this study, we report for the first time the economic result of using PEG in dairy cows. Based on the economic performance of cows during a full lactation, we determined the partial net return per cow per year. The main finding was that PEG treatment resulted in an overall economic benefit, as it increased the partial net return per cow per year. Interestingly, although numerical differences could be noticed in the descriptive statistics of the underlying economic factors of the partial net return, only the C CM per year was individually statistically significant, whereas all other factors represented in the partial net return (i.e., R Milk , Ins , and C Cull per year) were not significant. However, the total effect of all these economic factors, represented in the partial net return per year, resulted in a statistically significant economic benefit for PEG treated compared with control cows.
The method we used is relatively novel. Previous studies on the economics of cow-level health interventions used bio-economic simulation models [e.g., working on Meloxicam (van Soest et al., 2018) or working on Cabergoline ]. In such bioeconomic simulation modeling studies, specific decisions need to be made on what treatment effects need to be considered. Effects of treatment on occurrence of udder health were derived from clinical trials. In contrast, effects on reproduction or culling were modeled mechanistically using studies other than clinical trials. Such bio-economic simulation studies do not allow to account for heterogeneity between cows. In this study, we used longitudinal data from each enrolled cow during the full lactation. It allowed us to evaluate the combined effect of all cow production factors that may influence the economic performance of that cow. By using the data of all individual cows, we could evaluate the overall economic effect of PEG treatment, account-ing for heterogeneity between cows. We consider our current approach as a more reliable method of evaluating the true economic effects of cow-level interventions.
Almost 70% of the increased partial net return per cow per year due to PEG treatment was explained by the reduced C Cull per cow per year. Culling has previously been identified as one of the main contributors to the cost of clinical disease (Galligan, 2006;Overton and Fetrow, 2008;Heikkilä et al., 2012). Rollin et al. (2015) attributed more than 40% of the cost of CM during the first 30 DIM to costs of culling and replacement. van Soest et al. (2018) reported that reducing the proportion of culling had the highest economic impact on the net economic benefit of using nonsteroidal anti-inflammatory drugs in the treatment of CM. In a novel approach, Denis-Robichaud et al. (2021) reported that PEG treatment as an adjunct therapy increased survival after 30 d in cows with severe CM. It may, therefore, be hypothesized that PEG treatment reduces early culling as a consequence of disease, which would be particularly expensive due to the short time that those cows remain in lactation. In our trial, a first case of CM during the first 30 DIM was associated with an almost 2-fold increase in the hazard of culling . Pegbovigrastim treatment reduced the occurrence of a first case of CM during the first 30 DIM (Barca et al., 2021b) and, in multiparous cows, counteract the negative association of a first case of CM during the first 30 DIM with the hazard of culling . We also reported that PEG treatment reduced the hazard of a first case of CM and the rate of total cases of CM (Barca et al., 2021b). Hertl et al. (2018) reported that more CM cases in early lactation resulted in an increased rate of CM cases during the lifetime of a cow and that these CM cases also increased the hazard of culling. Thus, the reduced C Cull per year in PEG treated cows may be explained by the preventive effect on CM in combination with a lower Barca et al.: PEGBOVIGRASTIM: ECONOMIC EFFECT  For partial net return per year, the BCS by treatment interaction remained in the final model, suggesting that the economic benefit of using PEG depends on BCS. Again, this would be mainly explained by the C Cull per year. For the C Cull per year, the removal of the BCS by treatment interaction resulted in an important change in the PEG treatment effect. Therefore, BCS acted as a modifier for the effect of PEG treatment on the C Cull per year. Numerically, PEG treatment increased partial net return per year in under and over BCS cows, but not in cows with an acceptable BCS. Simultaneously, PEG treatment numerically reduced C Cull per year in under and over BCS cows, and not in cows with an acceptable BCS. These results are in line with our earlier results, which suggest PEG treatment reduced the occurrence of a first case of CM during the first 30 DIM in under (only numerically) and over BCS cows, and not in cows with an acceptable BCS (Barca et al., 2021b). It is relevant that, in our study, most of the cows that classified as under BCS were multiparous cows (data not shown). In multiparous cows, a first case of CM during the first 30 DIM was associated with a more than 2.5fold increase in the hazard of culling. As mentioned, in multiparous cows, PEG treatment counteracted the negative association between a first case of CM during the first 30 DIM and the hazard of culling (Barca et al., 2022). This contributes to explain the numerical differences particularly in under BCS cows.
Pegbovigrastim treatment reduced the C CM per year. This further supports the findings related to the reduced C Cull per year in PEG treated cows. As mentioned, PEG treatment reduced the occurrence of a first case of CM during the first 30 DIM (Barca et al., 2021b), consistent with previous reports (Canning et al., 2017;Ruiz et al., 2017). Moreover, we reported that PEG treatment reduced the hazard of a first case and the rate of total cases of CM during the full lactation (Barca et al., 2021b). All these effects explain the reduction of the C CM per year. Additionally, Powell et al. (2018) reported that PEG treatment reduced the severity of a CM case. In an experimental mastitis challenge, it was reported that PEG treated cows exhibited a less severe milk yield drop (Powell et al., 2018). A less pronounced milk drop would result in an economic benefit. Moreover, reduced severity of a CM case would affect C Medicine and C Labor per year for a CM treatment. To study whether PEG treatment affects severity of CM or whether the CM cases in control and PEG treated cows differ (e.g., causal pathogen, duration, bacteriological cure, and so on) warrants further research.
For the C Other per year, NEFA interacted with treatment, in High NEFA cows, PEG treatment numerically reduced the C Other per year. In this trial, all treatments were recorded. However, we did not evaluate the effect of PEG on the occurrence of diseases other than CM and UD. Further research analyzing the effect of PEG on the occurrence of other diseases would add information about the effect of PEG treatment on the C Other per year. Elevated prepartum NEFA concentrations were associated with an increased risk of disease (Ospina et al., 2013). Previously, we reported that PEG treatment reverted the negative association between prepartum NEFA concentration and postpartum circulating neutrophil counts (Barca et al., 2021a), potentially improving the immune responsiveness in PEG treated cows.
We decided to perform our analyses parametrically, as, despite the usual skewness in the distribution of costs, in pragmatic randomized trials like this one, it is the arithmetic mean that is the most informative measure, and comparisons between treatment groups are reliable if skewness is not too extreme (Thompson and Barber, 2000). We excluded 12 control and 13 PEG cows, due to their extremely high C Cull per year that made statistical analyses using parametric linear models difficult, if not impossible. The extremely high C Cull per year was a consequence of the very few ED (≤15 DIM) of these cows. This same approach with regard to outliers in C Cull per year calculations was recently reported by Burgers et al. (2022). Because of the calculation method and the need to express the economic performance of cows during an equal time period (a year in this study), extremely high costs or returns per year may occur when cows have very few ED. For these 25 cows, the C Cull per year was considerably lower in PEG treated cows compared with control cows. Hence, including these cows in the analyses would only further favor the economic outcome toward the PEG treatment. We acknowledge that our method to calculate the C Cull per year has shortcomings, as it resulted in some extremely high values which were excluded from the final analyses. However, we believe that this method allows us to achieve a realistic estimation in most cows, accounting for the future value of a cow, as it calculates the C Cull of each cow according to its actual lactation while accounting for the ED that each cow remained in the experiment. We consider this to be the best approach with the available data, as any revenue or cost of each cow should be spread out over the ED of each cow.
Sensitivity analyses showed that different variable levels did not have large effects on the economic consequences of PEG treatment. This indicates that the results are robust and not particularly sensitive with regard to differences in inputs (mainly price levels). The most influential variable with regard to the economic benefit of using PEG was the price of a replace-ment heifer. In this study, we used a conservative price because prices of replacement heifer are higher in most scenarios (INALE, 2021).
As mentioned, costs of PEG and its application (2 doses) were not included in the economic calculations. From an economic point of view, the calculated differences between control and PEG treated cows may be considered as the maximum amount of money that can be spent on PEG treatment. The costs of PEG treatment consist of the expenditure on the product, as well as the time it takes to administer treatments as part of farm transition cow management. It would consist of the time to identify and restrain the cow, and administer PEG. This may be highly dependent on the cost of labor and the ease with which cows can be identified and restrained. Thus, it could vary in different regions or herds with different management and facilities.
Overall PEG treatment resulted in a net economic benefit per cow per year. However, PEG treatment appears to be particularly beneficial in cows that are metabolically challenged (Barca et al., 2021a(Barca et al., ,b, 2022. Our results suggest that targeting PEG treatment to cows at a higher risk of disease due to a metabolic challenge and lower immune competence (Roche et al., 2009(Roche et al., , 2015Ingvartsen and Moyes, 2015) would be the more efficient option from an economic point of view. Thus, it would be of great value to develop predictive models to identify high-risk animals to further target the use of PEG. Loss of BCS would be a reliable indicator of metabolic challenge (Sheehy et al., 2017;Barletta et al., 2017). Under and over BCS cows may be identified by routine BCS scoring protocols or by using commercially available automated sensors (e.g., Mullins et al., 2019), whereas big data, metabolomics, and the use of automatic sensors to predict metabolic health (Overton et al., 2017) are promising technologies to identify cows that would benefit most from PEG treatment.

CONCLUSIONS
Based on data from a large randomized clinical trial, we conclude that PEG treatment resulted in an overall economic benefit, as it increased the partial net return per cow per year by $210 ± 100. Although the difference detected in the partial net return was statistically significant, the individual components of partial net return were not significantly different between control and PEG treated cows. Only the cost of treatment of CM was significantly lower for PEG treated cows compared with control cows ($9 ± 3). The largest numerical difference was seen for the cost of culling: PEG treatment reduced the cost of culling by $145 ± 77.

ACKNOWLEDGMENTS
The cooperation of farmers and farm personnel is gratefully acknowledged. The authors thank to Jorge Artagaveytia from INALE (Montevideo, Uruguay) and Gastón Moroni from PROLESA (Montevideo, Uruguay) for providing price inputs. Funding was provided by Elanco Animal Health (Greenfield, IN) and the University of the Republic, Uruguay. The authors have not stated any conflicts of interest.