Effects of monensin supplementation on lactation performance of dairy cows: a systematic review and dose–response meta‑analysis

The aim of this study was to conduct a comprehensive review with meta-analysis to determine the effects of the dose–response relationship between monensin supplementation and dairy cow performance and milk composition. Results from 566 full-text articles and 48 articles with 52 studies were meta-analyzed for pooled estimates. Monensin supplementation up to 23 ppm increased milk production, with the optimal dose being 12.6 ppm. Monensin supplementation at doses ranging from 16 to 96 ppm increased milk production in the prepartum phase (− 28 to 0 day relative to calving). From 60 to 150 DIM, monensin supplementation up to 21 ppm had a significant positive effect on this outcome, while supplementation in the 37 to 96 ppm range caused a decrease in this variable. At 0 to 60 and > 150 DIM, monensin supplementation had no effect on milk yield. At dosages of 22 to 96 ppm, 12 to 36 ppm, and below 58 ppm and 35 ppm, respectively, monensin supplementation resulted in significant decreases in dry matter intake (DMI), milk protein percentage, milk fat percentage, and milk fat yield. Overall, based on the results of this meta-analysis and considering all variables, the recommended optimal dose of monensin could be about 16 ppm.


Materials and methods
Data sources. A literature search was performed in several databases, including Google Scholar, Scopus, Science Direct, and PubMed. The keywords used to search the databases were "dairy cow", "dairy cattle", "monensin", "ionophore", "performance", "milk yield", and "milk composition". A total of 566 studies were identified through database searches: Google Scholar (306), Scopus (158), Science Direct (73), and PubMed (29, Fig. 1). The first round of exclusion excluded studies conducted in goats (n = 30), lambs or sheep (n = 134), and buffaloes (n = 34), as well as duplicate studies (n = 124), leaving 244 studies suitable for the meta-analysis. In the second round of screening, 163 irrelevant articles were excluded and 81 articles remained. In these two stages, the title and, if necessary, the abstracts were reviewed. The final screening was performed to select eligible studies accord- www.nature.com/scientificreports/ ing to the following criteria: (1) the study was conducted between 2000 and 2021, (2) the study used a control group without monensin supplementation, (3) the study reported at least one of the outcome variables listed, and (4) the study reported the mean values and associated error. Finally, 48 articles including 52 studies met the required criteria and were included in this meta-analysis. A summary of the details of the articles used in this meta-analysis is provided in Table 1. Outcome measures examined in our meta-analysis included DMI, milk yield, milk fat, milk protein, milk lactose, milk urea nitrogen (MUN), BCS, and BW.

Statistical analysis.
In the present study, the mean difference was used as the effect size. Because of the superiority of the one-stage method 22 , the one-stage random-effects meta-analysis approach was used to assess the potential non-linear relationship between dietary monensin supplementation and outcome variables. This was done using a restricted cubic spline with three knots at the fixed percentiles (10th, 50th, and 90th). An overall P value was calculated by testing that the two regression coefficients were simultaneously equal to zero. A P value of nonlinearity was calculated by testing that the coefficient of the second spline was equal to 0 21 . Heterogeneity between studies was measured by the variance partition coefficient 21 . When a significant level of heterogeneity was found, subgroup analysis was conducted to examine the potential sources of heterogeneity. Several rules were followed when conducting subgroup analyzes: (1) each subgroup analysis must be based on sound scientific evidence, (2) analyzes were pre-defined, (3) the overall effect of the independent variable was significant, and (4) the subgroup analysis was conducted on data subgroups of the meta-analysis with heterogeneity of > 50% and ≥ 10 studies 22 . Potential factors that may influence dairy cow response to supplemental monensin include parity (multiparous, multiparous and primiparous, and primiparous), decade of article publication (2000-2010 and 2011-2021), the method of using monensin (total mixed ration, TMR), TMR-top dressed, controlled-release capsule, and other), type of diet (TMR, pasture, and others), protein and NDF content of the diet (divided into two categories based on median), number of cows per trial (sample size), number of DIM at the beginning of the trial (day − 28 to 0 (calving), 0 to 60, 60 to 150, and > 150), length of the trial period (divided into two categories based on median), and breed of cows (Holstein, Holstein × Friesian, and others). Publication bias was assessed using Begg's funnel plot and Egger regression asymmetry for meta-analysis datasets with heterogeneity < 50% and ≥ 10 trials 22 .
Sensitivity analysis "leave-one-out approach, location of knots, and removing high monensin supplementation (> 48 ppm)" was performed to evaluate the robustness of the results 22 . In the spline model, the position of the knots may affect the results obtained. Therefore, we tested the sensitivity of the estimated curves to the position of the knots. To do this, we examined alternative knot locations, including various combinations of the 10th, 25th, 50th, 75th, and 90th percentiles of the total dose distributions. The results showed that the estimated curves were not sensitive to the location of the knots. Statistical analyzes were performed using the dosresmeta and metafor packages in R software (https:// www.r-proje ct. org/). A P ≤ 0.05 was considered statistically significant.

Results
Pre-analysis superficial review. Figure 2 and Supplemental Table S1 provide general information about the effect of monensin supplementation on outcome variables. The corresponding studies examined various dosages of monensin supplementation ranging from 0 to 96 ppm. The figure also shows that most of the monensin dosages studied had no effect on the outcome variables. Monensin supplementation appeared to have the greatest effect on DMI and milk yield, while other outcomes were less affected.
Effect on DMI. Forty-seven eligible articles, including 51 studies, were included in the analysis to evaluate the effects of monensin supplementation on DMI. The results showed a significant effect of monensin supplementation on DMI (P = 0.04). Dry matter intake was significantly decreased by increasing the monensin dose from 22 to 96 ppm (Fig. 3A). Leave-one-out analysis did not reveal any influential study that changed the results of the model. In addition, the overall trend of the association did not change when the monensin dose was removed above 48 ppm and the location of the knots. These tests of sensitivity analysis indicate that the results are statistically robust. No significant heterogeneity was observed for this outcome measure (heterogeneity < 50%). In addition, publication bias was not significant (P > 0.10, Fig. 4A).
Effect on milk yield. Forty eligible articles, including 43 studies, examined the effects of monensin supplementation on milk yield. A significant non-linear association was found between monensin supplementation and milk yield (P < 0.01). Increasing monensin supplementation dose up to 23 ppm linearly increased milk yield, and supplementation from 24 to 38 ppm monensin had no significant effect on this variable. In contrast, monensin supplementation above 38 ppm caused a significant decrease in milk yield (Fig. 3B). The results suggest that the optimal dose of monensin supplementation to maximize milk yield was 12.61 ppm. The sensitivity analysis of "leave one out, remove doses above 48 ppm, and knots location" did not change the model result. There was high heterogeneity among studies in terms of milk yield (heterogeneity > 50%) and therefore, to explore the causes, subgroup analysis was performed for decade of publication, parity, method of using monensin, type of feed, protein and NDF content of the diet, length of experimental period, and breed of cow. Subgroup analysis revealed that of all these potential factors, only the effect of DIM was significant (P < 0.001). Accordingly, monensin supplementation at doses ranging from 16 to 96 ppm significantly increased milk yield in the pre-partum phase (− 28 to 0 day relative to calving). From 60 to 150 of DIM, monensin supplementation up to 21 ppm had a significant positive effect on this outcome; dosages of 22 to 36 ppm monensin were not effective, while its supplementation in the 37 to 96 ppm range reduced milk yield. No significant effect of monensin supplementation was observed at 0 to 60 and > 150 DIM. No publication bias was observed for milk yield (P > 0.10, Fig. 4B). www.nature.com/scientificreports/ Effect on milk fat percentage and yield. Forty-seven articles with 51 studies were included in the metaanalysis on the effect of monensin on milk fat. A relatively linear association was observed between monensin supplementation and milk fat (P = 0.001). The percentage of milk fat decreased significantly when monensin was added to the diet up to 51 ppm, and its effect at the doses from 52 to 96 ppm was not significant (Fig. 3C). A relatively linear association was also found between the addition of monensin and milk fat yield (P = 0.009; Fig. 3D). The addition of monensin at doses 21-31 significantly reduced this result, and outside this range the effect of monensin was not significant. No significant heterogeneity (heterogeneity < 50%) and no publication bias (P > 0.10) were observed for milk fat percentage (Fig. 4C) and yield (Fig. 4D). Leave-one-out analysis, removal of a high dose of monensin, and knots location did not alter the outcomes of the models with respect to these results.
Effect on milk protein percentage and yield. Forty-seven articles with 51 studies were included in this meta-analysis. This meta-analysis revealed a relatively linear relationship between monensin supplementation and milk protein content (P = 0.01). Monensin supplementation at doses ranging from 12 to 36 ppm resulted in a significant decrease in milk protein percentage (Fig. 3E), and outside this range, the effect was not significant.
Leave-one-out analysis did not identify an influential study. Excluding the data point above 48 ppm monensin from the analysis and the location of the knots also did not change the results. Low heterogeneity was found for this result (heterogeneity < 50%). For milk protein percentage, the funnel plot and Egger's regression test indicated significant publication bias (P = 0.04), suggesting that the results should be interpreted with caution (Fig. 4E). www.nature.com/scientificreports/ A J-shaped relationship was observed between supplemental administration of monensin and milk protein yield (P < 0.001; Fig. 3F), and the estimated optimal dose of monensin was 13.50 ppm. Moreover, the effect of monensin was significant up to a dose of about 24 ppm, above which it was not significant. No significant heterogeneity (heterogeneity < 50%) and publication bias (P > 0.10) were observed for milk protein yield (Fig. 4F). Based on leave-one-out analysis, study No. 5 23 was a highly influential study. Their exclusion changed the model result; all doses of monensin addition resulted in increased milk protein yield. Our results also suggest that an 80% dose (13.87 ppm) is sufficient to maximize milk protein yield. However, removal of doses above 48 ppm did not affect the model result for this outcome.
Effect on milk lactose percentage. Thirty-eight eligible articles with 41 studies were included in this analysis. Although there was a relatively linear relationship between monensin supplementation and milk lactose percentage, only monensin supplementation at doses ranging from 16 to 96 ppm resulted in increase (trend, P = 0.06) in milk lactose percentage (Fig. 3G). Based on the results of the leave-one-out analysis, study No. 41 24 with a dose of 96 ppm monensin was identified as an influential study. When this study was excluded from the analysis, the overall P value of the model changed from < 0.001 to 0.0552, indicating that monensin supplementation tends to increase lactose content (Fig. 3G). Heterogeneity for milk lactose was low (heterogeneity < 50%). No publication bias was detected in this analysis (P > 0.10, Fig. 4G).

Effect on MUN.
For MUN, twenty-nine articles with 32 studies were included in the meta-analysis. According to the results, monensin supplementation linearly increased MUN content (P < 0.001, Fig. 3H). In the leaveone-out analysis, studies No. 4 23 , 11 25 , 14 26 , and 27 27 were found to be influential. Exclusion of these studies from the analysis did not change the statistical significance of the original model, but did change the estimated effect sizes. Monensin supplementation at 13 to 30 ppm dose range increased MUN, and other doses of monensin had no significant effect. Because of the lack of studies with doses greater than 48 ppm, the sensitivity analysis for high doses (> 48 ppm) of monensin supplementation was not performed. The heterogeneity index for MUN was low (heterogeneity < 50%), and there was no significant publication bias (P > 0.10, Fig. 4H).

Effect on BCS and BW.
Eighteen studies were used for the meta-analysis by BCS. This analysis found no association between additional monensin intake and BCS (Fig. 3I). No influential study was identified in the leave-one-out analysis, indicating the consistency of the results obtained. No study examined the effects of a high dose (> 48 ppm) of monensin on BCS. There was no evidence of significant heterogeneity (heterogeneity < 50%) and publication bias (P > 0.10) for this result (Fig. 4I).
Twenty-one articles involving 22 studies were included in the meta-analysis to evaluate the effect of monensin addition on BW. The results showed that the addition of monensin had no significant effect on BW of dairy cows (Fig. 3J). Leave-one-out analysis did not change the estimates. There was no study with a higher dose than 48 ppm, so the sensitivity analysis for high doses of monensin (> 48 ppm) was not performed. No significant heterogeneity (heterogeneity < 50%) and no publication bias (P > 0.10) were observed for this result (Fig. 4J). www.nature.com/scientificreports/ www.nature.com/scientificreports/ www.nature.com/scientificreports/

Discussion
Monensin is an ionophore produced by Streptomyces cinnamonensis 28 . It is often classified as an antibiotic because it interferes with ion transport across cell membranes, resulting in alteration of the rumen bacterial population through selective bacteriostatic action 28 . It has long been known that monensin improves digestive efficiency and energy metabolism by inhibiting Gram-positive bacteria rather than Gram-negative bacteria 29 . This change in rumen bacterial populations has several implications for ruminant metabolism. The ability of the molecule to alter rumen fermentation to increase the production of propionic acid and decrease the molar fractions of butyric acid, acetic acid, and carbon losses in the form of carbon dioxide and methane improves feed efficiency 28,29 .
Monensin supplementation in lactating dairy cows has been studied extensively. However, Arnqvist et al. 30 noted that each study alone is of little value unless it is compared with similar studies. Meta-analysis involves summarizing the results of appropriate studies on a topic using statistical methods 31 . This meta-analysis summarizes the results of published articles from 2000 to 2021 on the effects of monensin supplementation in lactating dairy cows. The present meta-analysis showed that supplemental administration of monensin altered the performance parameters of dairy cows. There was a J-shaped association between monensin supplementation and milk yield, and supplementation as low as 23 ppm had a significant positive effect on milk yield. Monensin supplementation in the range of 24 to 38 ppm had no statistically significant effect, and milk yield decreased linearly with doses above 38 ppm. The optimal dose was estimated to be 12.61 ppm. A significant degree of heterogeneity was found in this result. Subgroup analysis indicated that DIM could influence the effectiveness of monensin.
A J-shaped relationship was found between monensin supplementation and milk fat percentage. In addition, there was a linear relationship between monensin supplementation and the percentage of milk protein and lactose. Milk fat and protein percentage decreased linearly with monensin addition in the ranges of 0 to 58 and 12 to 36 ppm, respectively. However, as the monensin dose was increased from 16 to 96 ppm, the percentage of milk lactose tended to increase. The results also showed that the addition of monensin resulted in an increase in milk protein yield, while the addition of monensin at a dosage of 21 to 31 ppm reduced milk fat yield. The relationship between monensin addition and DMI was also J-shaped, with monensin addition at doses above 22 ppm reducing DMI.
The literature indicates that dairy cow responses to monensin supplementation were not consistent. Addition of 8 to 48 ppm monensin to the diet of dairy cows resulted in changes in milk yield ranging from − 1.3 to 2.8 kg/ day. Addition of monensin at 15 to 18 ppm resulted in maximum milk yield, and cows receiving the highest monensin dose gave the lowest milk yield 9,23,[32][33][34][35][36] . Meta-analysis of 77 studies found that supplemental monensin increased milk yield by only 0.7 kg/day 10 .
The effects of monensin supplementation on milk protein percentage and milk yield were also inconsistent in previous studies. According to Akins et al. 35 , the addition of monensin decreased milk protein percentage, but milk protein yield remained unchanged. However, Benchaar et al. 32 reported that neither milk protein yield nor percentage changed with monensin supplementation. A previous study of 3577 dairy cows showed that supplemental monensin increased milk protein yield while protein percentage remained unchanged 36 . In contrast, the meta-analysis by Duffield et al. 10 found that the addition of monensin decreased milk protein percentage.
Similar to this study, inclusion of low to moderate levels of monensin in the diets of dairy cows was found to decrease milk fat yield and milk fat percentage in several previous studies 23,32,33,37,38 . Interestingly, milk fat yield remained unchanged at moderate to high monensin inclusion, and milk fat percentage remained unchanged or decreased 23,35,36 . This meta-analysis confirms the results of the meta-analysis by Duffield et al. 10 , who reported that dietary monensin supplementation resulted in a lower milk fat percentage. This decrease in milk fat percentage could be due to the dilution effect resulted from increased milk volume production 32 .
Monensin has been shown to increase the production of propionate, a glucogenic precursor, and therefore may contribute to a reduction in energy loss during feeding, which could increase glucose and lactose synthesis, leading to improved milk yield 9 . In addition, monensin decreased the deamination of amino acids in the rumen, which decreased the NH3 concentration in the rumen and increased the escape and absorption of gluconeogenic amino acids, which in turn increased the availability of glucose for milk production and lactose 39,40 .
Similar to our findings, Ipharraguerre et al. 41 reported a 1.5% decrease in DMI in 14 ionophore experiments, while Santos et al. 23 reported an 18% or 3.58 kg/day decrease in DMI in lactating dairy cows supplemented with high levels (48 mg/kg DM) of monensin. The addition of monensin to a TMR diet at concentrations ranging from 8 to 33 ppm also decreased DMI 42 . Conversely, the addition of monensin had no effect on DMI in light lactating dairy cows 32 . However, in early lactating dairy cows, the addition of monensin increased DMI by 5% 43 . Benchaar et al. 32 and Ipharraguerre et al. 41 indicate that the effect of monensin on DMI depends on the stage of lactation, the status of energy balances, the level of monensin administration, and the number of animals used in the study.
The findings from our meta-analysis suggest that monensin supplementation could increase milk urea concentration. However, Duffield et al. 44 , reported no effects of monensin supplementation on milk urea in a metaanalysis study. Mullins et al. 45 also found an increase in MUN in multiparous Holsteins when 400 mg/day monensin was administered through the feed. Mullins et al. 45 did not provide a clear explanation for the higher concentration of MUN after monensin administration. Mammi et al. 29 reported higher MUN with monensin supplementation and attributed this to reduced microbial degradation and higher flux of undegraded proteins into the intestine and the resulting higher contribution of dietary AA absorbed from the small intestine to the profiles of milk AA. This explanation was also supported by the higher blood urea levels 43,45 , and lower ammonia levels in the rumen fluid of monensin-treated cows 46 . Feeding monensin to dairy cows has been shown to reduce deamination from 27 to 17 nmol ammonia mg/protein/min 47 . Blood urea nitrogen distributes freely throughout body fluids, including milk. Thus, this proposed mechanism may explain why MUN was higher in cows receiving monensin 45  www.nature.com/scientificreports/ It has been reported that the addition of 27 to 33 ppm monensin to the diet of dairy cows causes an increase in CP digestion in the small intestine by inhibiting the ruminal protein degradability 42,48 . Moreover, McGuffey et al. 3 found that supplementation with monensin reduced protein degradation, ammonia accumulation, and microbial nitrogen in rumen fluid in vitro. These results indicate that more dietary protein reaches the small intestine and consequently more amino acids are absorbed by the small intestine when dairy cows' diets are supplemented with monensin. The higher rate of AA absorption may contribute to increased gluconeogenesis from nonessential AA, possibly increasing milk protein synthesis, lactose synthesis, and mammary gland efficiency 49 .
Moreover, Mammi et al. 29 found that urea synthesis in the liver increased with monensin treatment because the lower lipid accumulation in hepatocytes improved liver functionality. Lower triglycerides accumulation in the liver of monensin-treated cows in early and mild lactation has been noted 16,43 . These results are supported by the higher mRNA abundance of carnitine palmitoyl transferase 1 in the liver, the slower accumulation of triglycerides in the liver 45 , and the better conversion rate of propionate to glucose in monensin-supplemented cows 43 .
The results of this meta-analysis showed no significant effect of monensin supplementation on BW and BCS. Consistent with our results, feeding monensin premix did not alter BCS 35,50 . Also, according to Vasquez et al. 51 , neither total BW nor postpartum BCS was affected by diet type or monensin supplementation. However, the meta-analysis conducted by Duffield et al. 10 found an improvement in BCS and BW with monensin supplementation.

Conclusions
This comprehensive review and meta-analysis examined the existing literature on the effects of monensin supplementation on performance and milk composition of dairy cows. The results showed that supplementation with monensin up to 23 ppm increased milk production, and the estimated optimal dose of monensin was 12.61 ppm. The response of milk production to monensin supplementation varied according to DIM. In addition, monensin supplementation significantly decreased DMI, milk protein, milk fat, and milk fat yield at doses of 22 to 96 ppm, 12 to 36 ppm, and below 58 ppm and 35 ppm, respectively. All monensin doses increased milk protein yield, and an 80% dose (13.87 ppm) is sufficient to maximize milk protein yield. In addition, supplementation with monensin at doses ranging from 13 to 30 ppm increased MUN. It tended to increase the lactose content of milk and had no effect on BCS and BW. Overall, the optimal dose of monensin could be 16 ppm.

Data availability
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request. www.nature.com/scientificreports/