There is an increasing focus on organic and natural beef production in the United States. Conventional feeding systems use pharmaceutical products not allowed in organic systems for finishing cattle and do not use certified organic feedstuffs as defined under the National Organic Program (2008). The recent USDA regulations for natural beef production prohibit the use of growth-promoting hormones and animals that have received antimicrobials. However, the new rule does allow the use of ionophores for the control of coccidiosis (USDA, AMS, 2009). Until recently, there has been no standard definition of natural beef production, which has led to inconsistencies among natural beef production systems. There is very limited research comparing an entire beef cattle production system with pharmaceutical technologies with one that excludes those technologies (Fernandez and Woodward, 1999; Sawyer et al., 2003; Berthiaume et al., 2006), thus limiting the ability to predict the performance differences accurately between cattle in the different systems. Unlike a review article, a meta-analysis provides a method of examining the existing literature critically and in a quantitative manor, accounting for within- and between-trial variance, to provide an overall estimate of effect of a given question(s) based on existing data. The aim of this study was to evaluate the performance effects (ADG, DMI, and G:F) and health effects (morbidity and mortality) of pharmaceutical technologies used in feedlot cattle that may be routinely excluded from nonconventional production systems. This report uses the techniques of forest plot analysis and meta-analysis to evaluate the technologies of steroid implants, monensin and tylosin (Elanco Animal Health, Greenfield, IN), endectocides, and metaphylaxis with any antimicrobial on arrival and their effects on the performance and health of feedlot cattle. In addition, liver abscess risks of cattle receiving tylosin vs. cattle not receiving tylosin were examined.
MATERIALS AND METHODS
Because this report only involved the use of previously published literature, there were no animals used in this research.
The question was, “What is the difference in ADG, G:F, DMI, morbidity, mortality, and liver abscess risks in feedlot cattle with and without pharmaceutical technologies in North America?” Manuscripts were identified through the PubMed and Commonwealth Agricultural Bureau electronic databases from February 2008 through April 2008. A variety of search terms were used to identify articles that possibly contained relevant information about the review questions (Table 1). Each search contained at least one search term from each category. After retrieving the citations, each title and abstract was read by the first author and was retained for further evaluation if it mentioned at least one of the treatments and at least one of the outcomes of interest in beef cattle. Further, the Intervet Schering Plough Animal Health Database (Kenilworth, NJ) and the Texas Tech University North American TBA Implant Database (Texas Tech University Implant Database, 1999) were used to identify studies utilizing steroid implants. The information contained in the database is searchable. The manuscripts and technical bulletins from which the data were sourced are linked from the site. All articles were cross-referenced with abstracts obtained through PubMed and Commonwealth Agricultural Bureau database searches, and technical bulletins from implant database were included as well. These manuscripts and technical bulletins were assessed using the same inclusion criteria, except for being from peer-reviewed sources. Only data from studies using a single implant treatment (i.e., no reimplant) with a contemporary nonimplanted control group were considered relevant. The implant data were further subgrouped into implant studies using heifers and studies using steers to attempt to mitigate the amount of variability in the data sets.
After identifying relevant studies based on the title and abstract, the complete manuscript was obtained and critically evaluated. Studies were retained for further consideration only if the study was conducted in North America, used randomization for allocation to treatment group, used beef breed animals, and contained an untreated control group. Manuscripts that did not meet all the above criteria were excluded from further consideration. Data were then extracted from the remaining studies. Data were extracted by recording the point estimates for ADG, DMI, G:F, mortality, morbidity, the presence of liver abscess and corresponding SE for each treatment group, description of the experimental unit, number of experimental units, and sex of cattle for each of the manuscripts.
For studies that reported results for different dose concentrations (e.g., monensin at 100, 200, and 300 mg/animal) the average treatment effect was calculated via a calculated mean of the treatment responses from the data at the individual dose concentrations. This data extraction approach was required for 5 studies in which authors were not able to separate out the information and report it separately. The measure for SE of the treatment effects in these studies was a pooled SE, as reported in the manuscript. When studies reported a pooled SE, this was converted to an SD by multiplying the pooled SE of the difference in sample means by the square root of the number of experimental units. The number of experimental units depended on what was defined as the experimental unit, the animal or the pen (i.e., for a 2-arm parallel comparison with 100 animals in 10 pens/arm and treatment allocated at the pen level, then n = 20). The SD was then squared to obtain the variance, which was used in the meta-analysis.
Forest plots using the R-statistical package and the rmeta and meta packages (R Development Core Team, 2008) were used to visually assess whether the effect of the technology on the outcomes of interest (ADG, DMI, G:F) was uniform across studies. For this graphic approach, the production outcome and standard deviations were used to calculate the difference between groups, and these data were used for the forest plot. In a forest plot, each study is listed individually on the left-hand side of the graph. The horizontal line listed next to each study represents a 95% confidence interval (CI) for the difference between groups for continuous outcomes. The size of the shaded box in the middle of the horizontal line represents the relative weight of a study compared with the other studies. The weight is a reflection of the number of experimental units involved in the study (i.e., the greater the number of replicates, the greater the weight). All the studies are then oriented in relation to the large vertical line listed as the zero line or no-effect line. This line represents a situation in which the difference between the 2 treatments equals zero. Studies to the right of the null value indicate a positive value and studies to the left have a negative value. Technologies were considered by the first author to be uniform when 50% of the point estimates (boxes) on the forest plot were to one side of the null effect line.
For the continuous outcomes ADG, DMI and G:F, technologies considered to display a uniform response compared with negative controls were analyzed using general linear mixed models of the MIXED procedure (SAS Inst. Inc., Cary, NC). In brief, each model of continuous outcome variables (ADG, DMI, G:F) contained 1 fixed effect (steroid implants, monensin, tylosin, endectocides, metaphylaxis), a random intercept effect for each study, and a repeated effect to incorporate the within-study variance for each study (St-Pierre, 2001; van Houwelingen, 2002). To define the covariance parameter for the between-study variance (range: 0.01 to 1.0 by 0.01), values for the within-study variance for each study (extracted from the literature) were used to create the profile likelihood function and resultant 95% CI for the between-study variance (van Houwelingen, 2002). The only change from the form described by van Houwelingen (2002) was the addition of the least squares means command to generate model-adjusted means and SE for fixed effects and establish a single overall treatment effect for each technology that showed a statistically significant difference (α <0.10). The model-adjusted means and SE for ADG, DMI, and G:F, when statistically significant, were used to derive a 95% CI for the summary effect and were incorporated into a standardized feedlot breakeven model.
The metaphylaxis morbidity and mortality data and the tylosin liver abscess incidence data were analyzed with generalized linear mixed models with a logit link and binomial distribution using the GLIMMIX procedure of SAS. The outcome variables were modeled in an events/trial format, where the denominator represented the total number of cattle within a group and the numerator included only those with the outcome of interest (morbidity, mortality, liver abscess). A repeated statement was used to account for the multiple observations within studies, and a random intercept term was used to account for the potential correlations among groups within studies. Metaphylaxis was included in the model statement as a fixed effect, and when significant, model adjusted estimates were transformed from the logit form to generate the estimated probability of treatment or death of an animal if receiving metaphylaxis, or having liver abscesses if consuming tylosin, respectively. Therefore, the estimated probability represents a cumulative incidence or risk of these adverse health events occurring at some point during each trial.
A standardized feedlot breakeven model commonly used by feedlot consultants and managers to assess the economics of feeding a pen of cattle with user-defined inputs was used (Thomson, 2008). The economic model assumptions are listed in Table 2. Based on an average of the ADG and G:F reported in Berthiaume et al. (2006), Fernandez and Woodward (1999), and Sawyer et al. (2003), we used an ADG of 1.30 kg/d and a G:F of 0.14 to predict the breakeven values for naturally raised calves. The model was then used to simulate implanting the natural calf, and the estimated differences in ADG and G:F from the meta-analysis were used to estimate a breakeven value. Finally, the model was used to predict feeding the natural calf an organic diet. For the organic breakeven value, the days on feed were increased by 20 d (Fernandez and Woodward, 1999) and the organic feed costs were multiplied by 1.5 (USDA, 2003) to simulate the feedlot performance and feed cost differences of organic cattle. A sensitivity analysis was performed for organic feed prices, ranging from 1.25 to 1.75 of conventional feed prices.
A total of 14,311 citations were identified by the initial electronic search. After examining the titles and abstracts for possible relevance and removing duplicate citations, 140 manuscripts were retrieved for quality assessment and data extraction. After quality assessment and application of the inclusion criteria, 91 treatment to control comparisons were identified from 51 manuscripts (Table 3). The 3 most frequent reasons studies were excluded were failure to include an untreated control group, failure to report variation of the outcome either as an SD or SE, and failure to use randomization to allocate animals to treatment groups.
Based on visual assessment of the forest plots, the use of endectocides (figure not shown), steroid implants (Figures 1 and 2), monensin (figure not shown), and metaphylaxis (Figure 3) showed a performance advantage for treated cattle relative to cattle in the negative control groups (i.e., more than 50% of point estimates to the right of the null). Tylosin studies did not show a consistent advantage in treated cattle relative to control cattle with respect to ADG, G:F, and DMI. An insufficient number of studies met the inclusion criteria to conduct a meta-analysis comparing endectocides, monensin, or tylosin. Therefore, a summary effect measure was calculated only for the metaphylaxis and implant data sets.
Meta-analysis and Breakeven
Average daily gain in feeder cattle receiving metaphylaxis and a variety of antibiotics used on arrival was 0.11 kg/d (95% CI = 0.10 to 0.13, P < 0.01) relative to cattle that did not receive metaphylaxis on arrival (Table 4). The use of implants in heifers was associated with increased ADG by 0.08 kg/d compared with nonimplanted controls (95% CI = 0.01 to 0.15, P = 0.09). The use of implants in heifers was not associated with differences in G:F (P = 0.14) or DMI (P = 0.44). The use of implants in steers was associated with 0.25 kg/d greater ADG (95% CI = 0.23 to 0.27, P < 0.01) and 0.53 kg/d greater DMI (95% CI = 0.45 to 0.61, P < 0.01) relative to nonimplanted control steers. The use of implants was also associated with increased G:F in steers relative to nonimplanted steers, by 0.02 (0.17 vs. 0.15; implanted vs. control, 95% CI = 0.018 to 0.022, P < 0.01).
The point estimates of differences in ADG and G:F for implanted and nonimplanted steers were incorporated into the breakeven model. The model suggests that implanted steers were associated with a $77/animal lower cost of production than nonimplanted steers fed similar diets. In addition, implanted steers fed a nonorganic diet had a $349/animal lower cost of production than nonimplanted cattle fed an organic diet, assuming cattle were being sold on the same market. The sensitivity analysis was performed on the assumption of organic feed price being 1.5 of conventional feed prices. These results indicated that the cost of production for organic beef, as well as conventional beef, is highly sensitive to feed price. For each 10% increase in the price of organic feed, the breakeven estimate increased by approximately $54/animal. The simulations did not incorporate morbidity and mortality effects.
Morbidity, Mortality, and Liver Abscesses
For cattle that received metaphylaxis and with a variety of antibiotics used on arrival at the feedyard, morbidity was estimated at 29% (95% CI = 21.2 to 38.58%) compared with 55% (95% CI = 44.46 to 65.14%) in the cattle that did not receive metaphylaxis (P < 0.01; Figure 4). For cattle receiving metaphylaxis on arrival at the feedyard, mortality was estimated to be 1.8% (95% CI = 1.05 to 3.12%) compared with 3.8% (95% CI = 2.30 to 6.50%) for cattle not receiving metaphylaxis (P < 0.01; Figure 4). Cattle not fed a ration containing tylosin had an estimated 30% (95% CI = 18.62 to 44.77%) risk of having liver abscesses compared with 8% (95% CI = 4.43 to 14.07%) in cattle consuming tylosin (P < 0.01; Figure 4).
This study suggests performance effects of several modern technologies on North American beef cattle production. Studies that compared a single implant with no implant in steers indicated an improvement in ADG by approximately 0.25 kg/d and improved G:F by 0.02. Over a 210-d feeding period, this could result in a 52.5-kg difference in BW between implanted and nonimplanted steers. This is approximately a 17% improvement in ADG and a 9% improvement in G:F, which is in agreement with other reports of expected implant performance when compared with nonimplanted cattle (Bartle et al., 1992; Preston, 1999; Duckett and Andrae, 2001).
Science-based performance and economic expectations can be achieved by using the meta-analysis performance results in a breakeven model. These data can provide an accurate indication of the premiums necessary to offset increased costs of producing beef in alternative systems compared with producing beef in conventional beef production systems. The use of metaphylaxis on arrival resulted in an estimated 53% reduction in subsequent morbidity treatments and an estimated 27% reduction in mortality losses compared with cattle not receiving metaphylaxis.
Feeding tylosin to feedlot cattle reduced the liver abscess risks from 30 to 8% in the studies examined. This study did not look at the severity of liver abscesses and relate it back to subsequent performance. This might explain why there was no significant difference in ADG between tylosin treatment groups. Nagaraja and Lechtenburg (2007), in their review of liver abscesses, reported significant variation in the performance effects of abscessed livers and stated that it was likely a function of severity. The small or mild liver abscesses likely have less of a negative impact on performance than do larger or more severe abscesses.
In general, the use of these technologies has increased the amount of beef produced per animal and has produced that beef more efficiently and economically (Lawrence and Ibarburu, 2006). The benefits of using these modern technologies are not limited to only the 3 performance indicators we measured. Additional impacts, such as control of bloat, coccidiosis, and external parasites, occur with the use of technologies but were not included in this analysis because of a dearth of published data. Still other technologies that were not addressed in this study are commonly used to control the reproductive cycling in heifers or increase the amount of HCW in the final period of feeding. As technologies are integrated into conventional feeding systems, their effects on economic and biological efficiencies in beef cattle production, as well as their effect on environmental and animal welfare issues, will need to be examined.
This meta-analysis was broader in scope than some of the previously published meta-analyses in beef cattle production (Van Donkersgoed, 1992; Wellman and O’Connor, 2007). Most meta-analyses try to answer the question, “Is one treatment more favorable than another in terms of a single or small number of outcomes?” McPhee et al. (2006) demonstrated the use of meta-analysis for applications examining multiple treatments with multiple outcome effects. The current study was designed to examine the effects of one production practice that has multiple treatments and multiple potential effects, as compared with another production practice that excludes all the treatments. This, by default, introduced a large amount of heterogeneity into the meta-analysis, which would preclude reporting any summary of effect statistics used in traditional meta-analyses. No test for homogeneity was done on the data sets. The existence of heterogeneity was assumed, and was the basis for the use of the random effects model, as was described previously. van Houwelingen (2002) stated that “heterogeneity might be present and should be part of the analysis even if the test for heterogeneity is not significant.” St-Pierre (2001) also remarked that the use of the random effects mixed model provides a more accurate estimate of values by allowing for the random effect of different studies.
One major source of bias is the approach used to identify studies. Published studies and technical reports are strongly subject to publication bias. There is a potential likelihood that studies that failed to show an effect have a decreased opportunity to be published. Therefore, the source of papers may represent a biased subset of papers and, if this is the case, may provide a best-case scenario difference. Unfortunately, it is not possible to estimate the magnitude of publication bias that may or may not exist. Further, it is feasible that for “truly effective treatments,” publication bias may not occur for different reasons (i.e., if a treatment is truly effective, the need to continue to publish the same outcome many times is diminished). Consumers of the review must make their own decisions about the potential for publication bias and its impact.
Summary effects were reported in this study for 2 main reasons. First, the heterogeneity in the meta-analysis was limited as much as possible for such a broad question through the critical manuscript review and inclusion criteria, and the remaining heterogeneity reflects the reference population the models seek to represent. There is a significant amount of variation among different segments of the beef cattle industry in terms of cattle management and production system goals. One producer may use a single implant, whereas others may use up to 4 in the lifetime of a calf. Second, it is hoped this study will stimulate more quantitative research on the health and performance effects of natural and organic beef production.
None of our summary statistics disagreed with smaller related reports in the literature. In addition, our summaries seemed to agree with anecdotal responses seen in the field. Based on a study of only 54 cattle, Fernandez and Woodward (1999) reported that a 39% greater selling price would be required to compensate for the performance reductions incurred with organic beef production, as compared with conventionally produced steers. Our simulated breakevens indicated that a $0.62/kg of BW premium would be required for an organically raised animal to generate the equivalent net return compared with a conventionally raised animal. Fernandez and Woodward (1999) also reported a 0.03 decrease in G:F and a 16% (0.26 kg/d) decrease in ADG in organically raised steers compared with conventionally fed steers. Our study also found that the difference in feed cost makes up the majority of the difference in costs between the 2 systems. In a 40-animal study, Berthiaume et al. (2006) found that nonimplanted cattle had a 16% reduction in HCW, had a 31% reduction in quality grade, and would require a 15% greater premium over implanted cattle to remain equivalent. Our simulated breakeven model suggests that a naturally raised steer would have to generate a premium of at least a $0.14/kg of BW to generate a return equivalent to that of a conventionally raised steer. Sawyer et al. (2003) also demonstrated similar results in a 64-steer study. Lawrence and Ibarburu (2006, 2008) also used a meta-analysis to examine the effects of removal of all pharmaceutical technologies from all segments of beef production. There were no statements regarding article selection or the process for examining the evidentiary value of the articles. Nonetheless, they estimated that the effect of removing pharmaceutical technologies from the feedlot phase of production would be $155/animal, which is similar to the estimate calculated from this study. Our model examined only the effects of a single implant and used only the difference attributable to that single implant in the analysis. Lawrence and Ibarburu (2006, 2008) attributed $71 of the $155/animal difference to implants, whereas the current analysis estimated a $77/animal benefit. There are likely benefits in ADG and G:F for multiple-implant programs that could explain the remainder of the difference between our estimates and those of Lawrence and Ibarburu (2006, 2008).
The primary purpose of a meta-analysis is to evaluate the existing literature critically and provide an overall summary of effects. This report was able to provide an overall summary of effects for a limited number of the technologies. The secondary purpose of a meta-analysis is to highlight reasons why studies were not found, were not included in the analysis, or both. For this report, the secondary purpose may be as important as the primary purpose. We found that there has been a shift away from including untreated controls in many studies in an effort to compare one technology with another. As the natural and organic industries continue to grow, it will be important to evaluate the effect of various technologies on beef production efficiency. It is also important for members of the beef industry to conduct further field trials comparing natural or organic systems directly with conventional systems. It is equally important to report the data more thoroughly by reporting measures of variation between treatment groups and between measured outcomes, and by accurately describing methods used for blinding because we had to exclude several studies for these reasons. It is possible the analysis presented here overestimates the direct impact of these technologies because of publication bias. However, it is also likely that these products are highly effective because they are widely adopted in conventional production systems.
|Cattle||Prophylaxis||Average daily gain|
|Heifer||Monensin||Dry matter intake|
|Calf||Rumensin||Grain to feed|
|In BW||250 kg|
|Laid in price||$2.42/kg|
|Feed cost||$209.00/909 kg|
|Days on feed||210 d|
|Yardage||$0.30/animal per day|
|Out BW||568 kg|
|Fat cattle price||$2.11/kg|
|Metaphylaxis||0.11 ± 0.02***||—||—|
|Implanted heifers||0.08 ± 0.04†||0.01 ± 0.003‡||—|
|Implanted steers||0.25 ± 0.01***||0.02 ± 0.001***||0.53 ± 0.05***|