A Meta-Analysis of Zilpaterol and Ractopamine Effects on Feedlot Performance, Carcass Traits and Shear Strength of Meat in Cattle

This study is a meta-analysis of the effects of the beta-agonists zilpaterol hydrochloride (ZH) and ractopamine hydrochloride (RAC) on feedlot performance, carcase characteristics of cattle and Warner Bratzler shear force (WBSF) of muscles. It was conducted to evaluate the effect of the use of these agents on beef production and meat quality and to provide data that would be useful in considerations on the effect of these agents on meat quality in Meat Standards Australia evaluations. We conducted a comprehensive literature search and study assessment using PubMed, Google Scholar, ScienceDirect, Scirus, and CAB and identification of other studies from reference lists in papers and searches. Searches were based on the key words: zilpaterol, zilmax, ractopamine, optaflexx, cattle and beef. Studies from theses obtained were included. Data were extracted from more than 50 comparisons for both agents and analysed using meta-analysis and meta-regression. Both agents markedly increased weight gain, hot carcase weight and longissimus muscle area and increased the efficiency of gain:feed. These effects were particularly large for ZH, however, fat thickness was decreased by ZH, but not RAC. Zilpaterol also markedly increased WBSF by 1.2 standard deviations and more than 0.8 kg, while RAC increased WBSF by 0.43 standard deviations and 0.2 kg. There is evidence in the ZH studies, in particular, of profound re-partitioning of nutrients from fat to protein depots. This work has provided critically needed information on the effects of ZH and RAC on production, efficiency and meat quality.


Introduction
Dietary additives, including b-adrenergic agonists (b-AA) are used by the beef industry to enhance the efficiency of gain and modify carcass characteristics and meat quality [1,2]. Both zilpaterol (ZH) and ractopamine (RAC) bind to badrenergic receptors located in the cellular membranes and indirectly lead to decreased lipogenesis (fat synthesis and storage) and increased lipolysis (fat mobilization and hydrolysis) [3,4,5,6]. The magnitude of these changes is influenced by dose and duration of treatment with the b-AA, the type of b-AA, and the species of animal treated [7,5,8]. The b-AA also directly influence protein metabolism. Previous studies have shown that RAC improved growth performance [1,2], with the effects mediated through increased protein synthesis [8] and decreased protein degradation [5,8,6].
A review of the meat science literature [9] concluded that if improvements in animal growth were due in part to decreased protein degradation, this would reduce the rate of post-mortem proteolysis in muscle and result in tougher meat. Given that part of the b-AA effect is likely to be associated with protein degradation rates it's use may produce tougher meat. However the results on meat quality from individual studies appear equivocal with a range of individual studies varying in the magnitude and statistical significance of the responses [6].
The objectives of this study were to evaluate the effects of b-AA on feedlot performance and carcass traits of cattle, using meta-analytic methods. Metaanalysis integrates the results from many studies to provide a more robust estimate of the effects of b-AA (both ZH and RAC) on live and carcass traits. We also examined the heterogeneity or variability of responses in order to better target future research projects and resolve other hypotheses raised about the action of the products and gaps in current knowledge. An understanding of factors that give rise to heterogeneity between studies can also lead to more effective treatments or modifications of management.

Literature search
Our literature searches used PubMed, Google Scholar, ScienceDirect, Scirus, and CAB and identification of other studies from references lists in papers. Searches were based on the following key words: zilpaterol, zilmax, ractopamine, optaflexx, cattle and beef using the terms for the pharmaceutical agent or product brand names separately with the terms cattle and beef. Where possible, studies from theses obtained were also included. A list of published studies on ZH in beef cattle was provided by Meat and Livestock Australia (MLA). The papers were useful in the initial phase of the project, but did not yield additional studies after the study search was completed. In many cases, authors of articles were contacted to provide clarity in regard to results and for additional information.
while working with the Cooperative Research Centre for Beef Genetic Technologies several projects were funded by Pfizer Animal Health and Pfizer Animal Genetics. He is currently working on a project, in part funded by Elanco Animal Health. Frank R Dunshea has consulted for with Elanco Animal Health as a member of the Swine Nutrition Advisory Panel and has conducted research with ractopamine in swine funded through the Cooperative Research Centre for Internationally Competitive Pork of which Elanco Animal Health is a participating organisation. All three authors are members of the Meat Standards Australia Pathways committee. Meat and Livestock Australia played no role in the study design apart from facilitating and funding the study. It facilitated the study by funding an independent search that identified sufficient studies were present to conduct a meta-analysis. This literature search was made available to the authors but did not provide papers that were not identified by the subsequent systematic literature searches undertaken. The authors have no other competing interests in research funding, employment, patents, products in development or marketing products.

Inclusion and exclusion criteria
Studies were included or excluded in this study based on a series of criteria developed by SBScibus and which are discussed in a review of the use of metaanalysis in animal and veterinary science [12]. Briefly, database searches were augmented with hand searches of library resources for relevant papers, books, abstracts, and conference proceedings. We pursued references in papers and contacted authors active in the field. Quality assessment criteria included randomization of study groups, appropriate analysis of data and comparability of treatment groups at entry to each trial. The extracted data were audited by five reviewers.
References  were evaluated for inclusion or exclusion in the study and details of these studies are provided in File S1. Trials were included in the analysis if they met the following criteria: full manuscripts from peer-reviewed journals, published after 1980, that evaluated use of ZH or RAC supplementation in cattle; had a description of randomization processes; reported the dose of b-AA used, animals studied were cattle; the paper contained sufficient data to determine the effect size for production outcomes (e.g., the number of cattle or pens, carcases or cuts of beef in each treatment and control group); a measure of effect amendable to effect size analysis for continuous data (e.g., standardized mean difference, SMD); a measure of variance (SE or SD) for each effect estimate or treatment and control comparisons. Studies were rejected that could not be adequately interpreted in regard to origin of muscles, that used non-representative sampling methods, or whose authors did not respond to clarify their approach.
Design criteria included the number of animals/pen, animals/group and pens/ group; experimental and analytical unit (animal, pen, carcass, muscle).Experimental details included days on feed, the number of days products were fed, dosage of products fed (per kg DMI or per kg/hd/d), diet and delivery methods of product and withdrawal period (days) of the test products prior to slaughter. Animal details included class of cattle (steers, heifers, bulls or cull cows); vaccination history, use of anthelmintics; other supplements and growth promoting products (hormonal implants); housing and feeding systems.
In many papers muscle, bone and overall maturity were reported as a proportion within the A maturity category, whilst in others it was reported as a continuous variable. As it was not possible to reconcile these units, maturity scores were not included in the current analyses. The ADG, DMI and G:F ratio were reported over the period of supplementation, generally 20-30 days, not over the full feeding period. There was also a lack of detail on sampling methods for sensory samples for a number of ZH and RAC studies and it was considered there were insufficient studies to include this trait in the analyses.

Statistical analysis
We used Stata (Intercooled Stata v.13, USA) to analyze production and carcass data by standardized mean difference (SMD) which is also called effect size (ES) analysis, in which the difference between treatment and control groups means was standardized using the standard deviations of control and treatment groups. The SMD estimates were made for the fixed effect [10], and random effects [11] models that are reported. If the paper reported separate estimates of measure of variance (SE or SD) for each group, these were recorded as such. Many studies reported a common SE or SD and these estimates were used for both control and treatment groups. We also have provided a random effects weighted mean difference (WMD) between treated and control, with the weighting reflecting the inverse of the variance of the studies included according to non-standard method (Stata 13.0 Statacorp, Tx, USA).
Random effects models were conducted for each production outcome to estimate the effect size, 95% confidence intervals, and statistical significance of SMD. We recognize that there is a clustering effect that results from multiple comparisons to a single control group, but have determined that the variance inflation effect will be minor unless there are very large numbers of repeated comparisons. The statistical methods of meta-analytic procedures that were used in this paper have been previously published [12].
Efforts were made to clearly identify the units of interest used in the studies and to clarify the measures of dispersion reported in papers. If there was some uncertainty, authors were contacted to provide clarification of these measures and a number responded. If there was a lack of clarity in regards to the unit of measure, a more conservative measure was used. Specifically, if muscle characteristics were measured and evaluated as the unit of analysis, but the muscles were obtained from pen fed studies, pen was used in our analyses.

Forest plots
The effects of b-AA on production performance of beef cattle are displayed in the forest plots, using the estimated SMD of b-AA using random effects [11] models. Points to the left of the vertical line represent a reduction in the outcome, whereas points to the right of the line indicate an increase in the outcome variable. Each square represents the mean effect size for that study. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval (CI) for the effect size. Forest plots were only presented for selected output variables.
The weighting of a study is estimated by the inverse of the variance of the effect size. Box sizes are proportional to the inverse variance of the estimates. The size of the square box reflects the relative weighting of the study to the overall effect size estimate with larger squares representing greater weight. Boxes draw attention to the studies with the greatest weight. The grey vertical line represents the mean difference of zero or no effect.

Assessment of heterogeneity
Variations among the trial level SMD were assessed using a chi-squared (Q) test of heterogeneity. Heterogeneity in studies reflects underlying differences in clinical diversity of the herds and b-AA used, differences in study design and analytical methods and statistical variation around responses. Identifying the presence and sources of the heterogeneity improves understanding of the responses to b-AA. We used an a level of 0.10 because of the relatively poor power of the x 2 test to detect heterogeneity among small numbers of trials [13]. Heterogeneity of results among the trials was quantified using the I 2 statistic [14], which quantifies the impact of heterogeneity on a meta-analysis, from mathematical criteria, that are independent of the number of studies and the treatment effect metric. I 2 is a transformation of the square root of the x 2 heterogeneity statistic divided by its degrees of freedom and describes the proportion of total variation in study estimates that is due to heterogeneity. Negative values of I 2 were assigned a value of zero, consequently the value I 2 lies between 0 and 100%. An I 2 value greater than 50% indicates substantial heterogeneity.

Meta-regression
Meta-regression analyses were used to explore the source of heterogeneity of response, using the individual SMD for each trial as the outcome and the associated standard error as the measure of variance. Meta-regression can be used to explore sources of heterogeneity even if an initial overall test for heterogeneity is non-significant [14]. This also allows us to quantify the magnitude as a function of the a priori defined covariate changing and exploring reasons for heterogeneity (i.e. possible/probable study-level predictors). Meta-regression is also a technique that can formally test whether there is evidence of different effects in different subgroups of trials [15]. Inclusion of multiple covariates in the meta-regressions [15], and the use of the smoothed within-trial variance estimates were used to improve hypothesis testing with regard to the significance levels. The permutation test approach for assessing the statistical significance of meta-regression methods [16] used the programming methods [17,18] to reduce the risk of type I error.
Meta-regression analysis was conducted by first screening individual variables using a P-value of #0.10. All variables with P-value of #0.10 were entered into a forward stepwise weighted meta-regression, until all remaining variables were significant at P,0.05. Factors identified a priori as factors that might influence responses to treatment were examined, including the unit of study (pen or animal or cut of meat), initial value of the study variable in the control group in each study (eg DMI of control), duration of feeding in the study (days on feed) and period or duration of treatment with each b-AA (eg RAC was fed for 30 days). Data were screened for plausible quadratic relationships for these variables by visual appraisal of univariable scatter plots between the covariate and SMD of each study.

Publication bias
We investigated the presence of publication bias using funnel plots which are a simple scatter plot of the intervention effect estimates from individual studies plotted against study precision. The name 'funnel plot' arises because precision of the estimated intervention effect increases as the size and precision of a study increases. Effect estimates from small studies will scatter more widely at the bottom of the graph and the spread narrows for larger studies. In the absence of bias the plot should approximately resemble a symmetrical (inverted) funnel. If there is bias, for example because smaller studies without statistically significant effects remain unpublished, this will lead to an asymmetrical appearance of the funnel plot and a gap will be evident in a bottom corner of the graph. In this situation, the effect calculated in a meta-analysis will tend to overestimate the intervention effect. The more pronounced the asymmetry, the more likely it is that the bias will be substantial. Individual studies with large or aberrant results were identified from the forest and funnel plots and evaluated for factors that may have influenced the results. This lead to the exclusion of some studies (see Tables 1 and 2).

Studies identified and information extracted
A comprehensive literature search using 5 search engines identified relevant peerreviewed papers, theses and proceedings that were published on ZH and RAC supplementation between 2000 and 2013. For both interventions, details were recorded, but are not provided, on sex of cattle, breed, treatments including vaccinations, parasite control, hormonal implants, feeding regimes, country in which the study was conducted and distance cattle were transported to slaughter.
A total of more than 50 studies on ZH, published between 2000 and 2013, were identified. Of these, 31 papers (with 83 sub-trials or comparisons on different outcomes) met the selection criteria. The required data and information from those studies that met the inclusion criteria were extracted and were tabulated and presented in S1 to S5 Tables in S1 File. The data included information on animal performance, carcass characteristics and meat quality. S6 Table in S1 File provides details of studies that were excluded for various reasons.
Similar search, extraction and reporting methods were used for RAC and a total of 31 papers were identified with 68 sub-trials or comparisons on different outcomes. These trials are presented in S7 to S10 Tables in S1 File. Contact with workers in the field provided valuable detail on studies and allowed re-analysis of some data. Again, three reviewers assessed, extracted and validated the data and 8 papers were excluded (S11 Table in S1 File). There were insufficient studies reporting some variables, especially those reporting sensory traits of meat, for meta-analytical evaluation.
The methods of meta-analysis used in this study have been described and published previously [12,103,104]. A critical consideration in this study was to ensure that appropriate estimates of evaluative units (eg pen, animal or carcase) were used for the analyses and that the estimates of standard deviation were also appropriate. Our decisions were influenced by advice received on design and analysis of pen studies that the analyses using SAS would report standard errors of the difference that were appropriate to providing estimates of standard deviation [105] Tempelman, pers com 2013). Studies that did not clearly provide the unit of analysis were evaluated at the highest level of unit identified, ie pen. This may provide a conservative bias in the analysis. Similarly, only random effects models were used, as previous work concluded that when there was uncertainty in the evaluative units caused by clustering of observations the random effects model was appropriate [106]. The effect of reported and assessed evaluative unit was examined by testing of this in meta-regression. Only three outcomes, indicated that the evaluative unit significantly influenced the results of meta-regression, LMA, marbling and WBSF for ZH studies (Table 1). However, in all cases the raw mean difference, an unweighted measure, differed for the outcomes for the pen and individual animal studies, suggesting that there may be a biological reason the differences in effect. In all cases, also, the expected effect of ZH, i.e. increased LMA, lower marbling and increased WBSF was accentuated in the group fed animals. The lack of other significant findings in regard to unit of evaluation in the study overall suggest the possibility that whether cattle had access to individual feeders or were fed in pens influenced the LMA, marbling and WBSF for ZH treated cattle, but had little effect on other output variables. It is possible that such results may reflect the differences between single access to feed and competing in the group-fed situation. Competition can increase stress and agonistic behaviours [107] which may heighten responsiveness to endogenous and exogenous adrenergic stimulation. While effects of group size on feed intake are variable, feed efficiency is generally poorer and a poorer nutritional state is also associated with heightened adrenergic responsiveness [108]. We conclude that the evaluative units were probably correctly identified and, consequently, had little influence on results.

Zilpaterol
The period of feeding cattle was 143.8¡56.1 days and the average period that cattle were exposed to ZH was 26.6¡9.0 days. The results of the number of studies analysed, raw mean differences for variables, SMD, weighted mean differences, I 2 estimated heterogeneity and results of meta-regressions evaluating the effects of ZH are provided in Table 1. These results showed that supplementation of cattle with ZH during the last period of feeding resulted in a substantial impact on live and carcass traits of the animal. The weighted mean differences in BW and ADG between ZH and control groups were substantial being of the order of 0.4 and 0.9 standard deviations, respectively. This effect was equivalent to approximately 8 kg BW and 0.15 kg ADG gain more than the controls. Cattle fed ZH also had a lower DMI of the order of 0.5 standard deviations which was equivalent to approximately 0.1 kg/hd/day. Given that ZH resulted in increased BW and ADG and decreased DMI there was a substantial improvement of 1.4 standard deviations in the G:F ratio which was equivalent to 0.02 improvement in G:F.
The increase in HCW due to ZH supplementation was1.3 standard deviations above controls and equivalent to an extra 15 kg of HCW, almost twice the increase in final BW ( Table 1). The disproportionate increases in BW and HCW resulted in an improvement of dressing percentage of 2.2 standard deviations, which was equivalent to an increase of 1.7% in dressing percentage.
ZH acts as a repartitioning agent where there is a substantial redirection of nutrients from fat to protein deposition [6]. The magnitude of the metabolic changes in protein deposition was reflected in the large increase in LMA of 2.3 standard deviations, which was equivalent to an increase of ca. 8 cm 2 in LMA. This effect was not significantly explained by the difference in HCW between treatment and controls or the HCW of the treatment group. Similarly the reduced fat deposition was reflected in a decrease of 0.9 and 0.7 standard deviations in USDA marbling score and fat depth respectively. These differences were equivalent to a decrease in marbling score of 14 units and a decrease of 1 mm in fat depth. There were few data provided in the papers which examined the physiological and health impacts of this very substantial repartitioning of protein and fat in the treated cattle. It is possible that the profound effect of these agents on repartitioning of nutrients is associated with increased death rates reported for cattle treated with ZH and RAC [110]. Loneragan et al. [109], found that the cumulative risk and incidence rate of death was 75 to 90% greater in cattle fed the bAA compared to contemporaneous controls and that during the exposure period, 40 to 50% of deaths among groups fed the bAA were attributed to administration of the drug.
Forest Plots for HCW, LMA, marbling score and WBSF variables are presented in Figs. 1, 2, 3 and 4 respectively. These Forest Plots demonstrate the consistency of the ZH responses in key commercial carcass and quality traits. Funnel plots were produced, but are not presented. These plots did not show any marked asymmetry with the possible exception of gain:feed. Given the lack of evidence of missing studies in other variables derived from many of the same studies, we conclude that there is little or no evidence of publication bias. Table 1 also reported an estimate of the heterogeneity for the output traits. Most traits had small estimates of heterogeneity with the exception of LMA and D%. Investigation of the sources of this heterogeneity using meta-regression identified that significant heterogeneity in LMA was attributable to the period of feeding of ZH, further emphasising the causal basis of the effect. However, the unit of evaluation, in this case animal or pen indicated that the SMD of LMA was greater in the pen fed studies. As shown in Fig. 5 this difference was evident in the raw, unweighted data, showing that the effect was not mediated through a difference in weighting of studies. It is therefore likely that this result and that for marbling in which the unit of measure was also a significant covariate and that similarly had a difference between pen and animal studies in the raw means reflect differences in expression of effect of ZH between pen fed animals and individually fed animals. The period of feeding of ZH was significant in increasing LMA and approached significance in reducing marbling, again indicating the biological significance of treatment. It is possible that the sensitivity to the ZH is increased under the possibly more stressful group fed conditions in which epinephrine levels may be higher.
Figs. 5 and 6 show the difference in SMD and raw mean difference for individually fed animals and pen fed groups for LMA and for USDA marbling score, respectively. In studies where animals were individually fed, the response in LMA was smaller than for pen fed groups. In contrast, the reverse was evident for USDA marbling score where individually fed animals showed little response to ZH supplementation compared with group fed animals which showed a large decrease in marbling score. Obviously behavioural changes or competition effects between animals are driving these effects. The authors are not aware of these effects being A Forest plot of the effect size or standardized mean difference (standardized using the z-statistic) and 95% confidence interval of the effect of zilpaterol treatmenton hot standing carcase weight. The weights that each study contributed are in the right hand column and are indicated by the size of the box. The larger the box, the greater the study contribution to the overall estimate. The solid vertical grey line represents a mean difference of zero or no effect. Points to the left of the line represent a reduction in hot carcase weight, while points to the right of the line indicate an increase. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval for the effect size. The overall pooled effects size and 95% confidence interval is indicated by the diamond at the bottom. This effect was heterogenous as indicated by the I 2 of 55.9%. reported previously. The effect of feeding system on the magnitude of the responses highlights the need to evaluate products under the commercial conditions in which they will be used.
The effect of ZH on WBSF was large being increased by SMD 1.2 and a weighted mean difference increase of more than 0.8 kg in WBSF (Table 1). There Fig. 2. Forest plot of Longissimus muscle area (cm 2 ) responses for Zilpaterol studies. A Forest plot of the effect size or standardized mean difference (standardized using the z-statistic) and 95% confidence interval of the effect of zilpaterol treatmenton Longissimus muscle area. The weights that each study contributed are in the right hand column and are indicated by the size of the box. The larger the box, the greater the study contribution to the overall estimate. The solid vertical grey line represents a mean difference of zero or no effect. Points to the left of the line represent a reduction in Longissimus muscle area, while points to the right of the line indicate an increase. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval for the effect size. The overall pooled effects size and 95% confidence interval is indicated by the diamond at the bottom. This effect was heterogenous as indicated by the I 2 of 70.8%. were several significant interactions identified by meta-regression. The magnitude of the ZH effect on WBSF decreased with days that the meat was aged. (Fig. 7).If dietary supplements or implants lower protein degradation rates in the live animal, these are likely to reduce post-mortem ageing rates of the carcase and produce tougher meat [9]. As the improvement in tenderness is exponential, it Fig. 3. Forest plot of Standardised USDA marbling score responses for Zilpaterol studies. A Forest plot of the effect size or standardized mean difference (standardized using the z-statistic) and 95% confidence interval of the effect of zilpaterol treatmenton USDA Marbling Score. The weights that each study contributed are in the right hand column and are indicated by the size of the box. The larger the box, the greater the study contribution to the overall estimate. The solid vertical grey line represents a mean difference of zero or no effect. Points to the left of the line represent a reduction in Standardised USDA marbling score, while points to the right of the line indicate an increase. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval for the effect size. The overall pooled effects size and 95% confidence interval is indicated by the diamond at the bottom. This effect was relatively homogenous as indicated by the I 2 of 37.3%.  A Forest plot of the effect size or standardized mean difference (standardized using the z-statistic) and 95% confidence interval of the effect of zilpaterol treatmenton Warner-Bratzler Shear Force. The weights that each study contributed are in the right hand column and are indicated by the size of the box. The larger the box, the greater the study contribution to the overall estimate. The solid vertical grey line represents a mean difference of zero or no effect. Points to the left of the line represent a reduction in Warner-Bratzler Shear Force, while points to the right of the line indicate an increase. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval for the effect size. The overall pooled effects size and 95% confidence interval is indicated by the diamond at the bottom. This effect was heterogenous as indicated by the I 2 of 61.0%.   follows that as meat ages differences between ZH supplemented and control carcasses will decrease.
The type of study design, that is, whether the study was conducted at the animal level or the pen level also influenced WBSF. Fig. 8 showed an increase in the magnitude of the ZH effect for experiments where pen was the experimental unit. The implications drawn about the unit of evaluation for WBSF are difficult because animal was used as the unit of evaluation in some studies where carcases were randomly selected within a group of cattle fed in pens. However, all studies were in the same direction, indicating an increase in WBSF. The heterogeneity identified, however, is only in degree of increase of WBSF, reflecting quite substantial differences in study design among the studies displayed in Fig. 8.

Ractopamine
Cattle were fed on average 114.0¡44.8 days in the lot and the average period that cattle were exposed to RAC was 30.8¡5.3 days. The results of the studies, which pertain to the RAC exposure period, include the raw mean differences for variables, SMD, weighted mean differences, I 2 estimated heterogeneity and results of meta-regressions evaluating the effects of RAC and are provided in Table 2. Animals supplemented with RAC showed an increase of 0.4 and 0.8 standard deviations in BW and ADG, respectively, which were equivalent to about 8 kg and 0.19 kg/day. In contrast to ZH, the increase in weight gain with RAC was not mediated through increased DMI. The point direction was very close to a zero effect and the overall data showed a marked variation around the mean with some studies showing significant increases or significant decrease in DMI. There was no evidence in this case of missing data from the funnel plot and none of the covariates were significant in explaining the effects of the DMI.
While the increase in ADG during the period of supplementation was substantial, some caution should be expressed in regard to this result as there is some evidence of missing data on the funnel plot (Fig. 9). It appears that small negative studies may not be present in this particular data set. The covariates examined did not explain any of the variation in effect size. The gain:feed was significantly improved and quite similar in effect size to the response in regard to the ADG. This is unsurprising given the lack of significant increase in DMI observed. It clearly indicates that there are efficiency gains with treatment and the funnel plot was quite symmetric. It is possible, therefore that the funnel plots in the ADG data may be an artefact, reflecting the large number of investigations, of funnel plots in this study. Again none of the sources of variation investigated by meta-regression significantly influenced the gain to feed efficiency.
Forest plots for HCW, USDA marbling score and WBSF are presented in Figs. 10, 11 and 12. The Forest Plots for RAC supplementation show that the magnitudes of the RAC responses were generally smaller than those for ZH. Although not presented, funnel plots were calculated and showed little evidence of missing studies and consequent publication bias. However, there was substantial   evidence of small study effects with these studies sometimes having quite large effects that differed from the larger studies.
The pH data were sparse and not significant. Evidence of repartitioning of nutrients was present in the carcase data. The SMD for LMA and dressing percentage was higher for the treated cattle and the SMD for marbling was significantly lower for the RAC treated cattle, as was the yield grade. The rib fat thickness was not significantly lower for RAC. These findings are consistent with the physiological action of RAC [5]. These results, in combination indicate that the response was to increase muscle, as indicated by increased LMA and HCW, but to decrease lipid as indicated by the marbling scores. The WBSF was also significantly increased by RAC treatment. The carcase data were less heterogenous than the feedlot data with only the HCW and LMA I 2 approaching or exceeding 50, respectively.

Conclusions
These findings support the previously identified physiological roles of the b-AA and provide strong evidence for producers and others to examine and consider the effects of ZH and RAC on beef cattle production. A large number of reviews and basic physiological studies have not clearly identified the mechanisms by which the actions of the specific b-AA are exerted, however, this study clearly demonstrates that the repartitioning effects are rapid, marked and highly integrated.  A Forest plot of the effect size or standardized mean difference (standardized using the z-statistic) and 95% confidence interval of the effect of ractopamine treatmenton hot standing carcase weight. The weights that each study contributed are in the right hand column and are indicated by the size of the box. The larger the box, the greater the study contribution to the overall estimate. The solid vertical grey line represents a mean difference of zero or no effect. Points to the left of the line represent a reduction in hot carcass weight, while points to the right of the line indicate an increase. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval for the effect size. The overall pooled effects size and 95% confidence interval is indicated by the diamond at the bottom. This effect was heterogenous as indicated by the I 2 of 46.5%.
doi:10.1371/journal.pone.0115904.g010 Fig. 11. Forest plot of Standardised USDA marbling score for Ractopamine studies. A Forest plot of the effect size or standardized mean difference (standardized using the z-statistic) and 95% confidence interval of the effect of ractopamine treatmenton USDA marbling score. The weights that each study contributed are in the right hand column and are indicated by the size of the box. The larger the box, the greater the study contribution to the overall estimate. The solid vertical grey line represents a mean difference of zero or no effect. Points to the left of the line represent a reduction in Standardised USDA marbling score, while points to the right of the line indicate an increase. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval for the effect size. The overall pooled effects size and 95% confidence interval is indicated by the diamond at the bottom. This effect was homogenous as indicated by the I 2 of 0.8%. Once these results have been critically reviewed by others, they can be immediately applied and used to formulate strategies to make best use of agents that markedly improve the efficiency of production. Registration is being sought for these products in countries including Australia. The adoption of these technologies will necessarily involve a consideration of the benefits in production costs and associated environmental benefits of improved efficiencies of resource use against the effects on shear force.
This work provided information on the effects of ZH and RAC on production, efficiency and meat quality. The meta-analysis provided more precise and robust estimates of the effects of the b-AA on efficiency of production and carcase quality measures. We identified, using meta-regression, that the method of feeding cattle Fig. 12. Forestplot of Warner-Bratzler Shear Force for ractopamine studies. A Forest plot of the effect size or standardized mean difference (standardized using the z-statistic) and 95% confidence interval of the effect of ractopamine treatmenton Warner-Bratzler Shear Force. The weights that each study contributed are in the right hand column and are indicated by the size of the box. The larger the box, the greater the study contribution to the overall estimate. The solid vertical grey line represents a mean difference of zero or no effect. Points to the left of the line represent a reduction in Warner-Bratzler Shear Force, while points to the right of the line indicate an increase. The upper and lower limit of the line connected to the square represents the upper and lower 95% confidence interval for the effect size. The overall pooled effects size and 95% confidence interval is indicated by the diamond at the bottom. This effect was homogenous as indicated by the I 2 of 0%.

Acknowledgments and Funding
We are particularly grateful to Professor Mike Gaylean for provision of raw data. We also sincerely thank Professor Robert Tempelman for discussions on the unit of interest, error estimates and study design. We acknowledge the contributions of the staff at SBScibus to the project.
Supporting Information S1 Checklist. PRISMA Checklist. doi:10.1371/journal.pone.0115904.s001 (DOC) S1 File. Contains the following files: Table S1. Authors, experimental unit, number of cattle and pens per group, and means for control (Con) and treatment (ZH) groups for final body weight (BW), average daily gain (ADG),dry matter intake (DMI) and Gain/Feed (G/F) ratio. Table S2. Authors, number of cattle per group, experimental unit, number of cattle and carcasses per group for analysing carcass characteristics in experiments with control (Con) and treatment (ZH) groups. Table S3. Authors and study means for control (Con) and treatment (ZH) groups for hot carcass weight (HCW), longissimus muscle area (LMA), ultimate pH and muscle colour score. Table S4. Authors and study means for control (Con) and treatment (ZH) groups for USDA marbling score, fat thickness (mm) and dressing percentage (D%). Table S5. Author, experimental unit, days aged, number of animals, number of pens, number of muscles and study means for control (Con) and treatment (ZL) groups for Warner Bratzler shear force (WBSF)and cooking loss percentage (CL%). Table S6. Zilpaterol studies that did not meet the selection criteria, had insufficient data or data provided were not appropriate. Table S7. Authors, experimental unit, number of cattle and pens per group, and means for control (Con) and ractopamine treatment (RAC) groups for final body weight (BW), average daily gain (ADG), dry matter intake (DMI) and Gain/Feed (G:F) ratio. Table S8. Authors, \and means for control (Con) and ractopamine treatment (RAC) groups for hot carcass weight (HCW), longissimus muscle area (LMA) and ultimate muscle pH. Table S9. Authors and means for control (Con) and ractopaminetreatment (RAC) groups for USDA marbling score, fat thickness (mm) and dressing percentage (D%). Table S10. Authors, experimental unit, days aged, number of animals, number of pens, number of muscles and means for control (Con) and ractopamine treatment (RAC) groups for Warner -Bratzler shear force and cooking loss. Table S11. List of ractopamine studies that did not meet the selection criteria, insufficient data or data provided were not appropriate.