Introduction

The trace element zinc (Zn) is involved in numerous general cellular functions in the body, including replication, transcription, and signal transduction. Furthermore, it is essential in the maintenance of barrier integrity, protection against pathogens, and regulation of immune response. Zinc requirements for post-weaning pigs are stated by the NRC [1] as 100 mg/kg diet during growth from 3- to 10-kg live weight and as 80 mg/kg diet during growth in 10–20-kg live weight.

During the late 1980s, it was discovered that pharmacological concentrations of Zn oxide (ZnO) resulted in reduced diarrhea and increased growth rates in weanling pigs [2]. Zinc toxicity occurred in growing pigs when Zn was supplemented at levels of greater than 2,000 mg/kg diet from Zn carbonate [3] and 1,000 mg/kg diet from Zn lactate [4]. Although Poulson [2] reported that feed intake and growth rate were reduced in post-weaning pigs when 4,000 mg Zn/kg diet was supplemented from ZnO, no reduction of growth was found by Hahn and Baker [5] at levels of 5,000 mg Zn/kg diet. However, when 5,000 mg Zn/kg diet was supplemented from Zn sulfate, weight gain and feed intake decreased.

With Zn poorly absorbed, it can become highly concentrated in manure. To minimize the risk of environmental pollution, European regulations have reduced the maximal Zn concentration authorized in pig diets to 150 mg/kg diet [6]. Among all farm animal species, growth-promoting effects of pharmacological dietary Zn concentrations have only been reported incidentally by some authors in piglets after weaning, and its effects on growth performance remain questionable [7]. This uncertainty resulted in the current meta-analysis across studies to, firstly, identify the existence of effects, if any, of dietary Zn at pharmacological concentrations on growth of post-weaning pigs and, secondly, to quantify the sizes of these effects. Furthermore, independent variables that could have been responsible for variation in results were evaluated. These results could aid in the justification or disapproval of the policy to prohibit pharmacological dietary ZnO concentrations in post-weaning pig diets.

Materials and Methods

Animal Care and Use Committee approval was not obtained for this study because the data were obtained from an existing database.

Selection of Studies and Description of Data Sets

Controlled research studies that have evaluated the effects of pharmacological Zn additions to pig diets on average daily gain (ADG; in gram), average daily feed intake (ADFI; in gram), and gain to feed ratio (G/F) were identified by electronic searches. For inclusion in the data sets, studies that were selected (1) had included a control treatment without any dietary Zn supplementation above physiological requirement, (2) did not apply any treatment that involved restriction or excess of any nutrient in the basal diet, (3) did not present any confounding of Zn supplementation with other feed additives, (4) had presented a measurement of variation, (5) had been written in English, and (6) had been presented in peer-reviewed journals. Experiments were individually coded when more than one experiment was presented in a study or when various basal diets were used within the same experiment. Performance during the post-weaning phase was added as selection criterion because of the lack of an adequate number of studies during further growth phases. Although several studies were identified that supplemented Zn from organic Zn, Zn carbonate, Zn chloride, and Zn sulfate, studies for inclusion in data sets were limited to those that have used inorganic ZnO. The efficacy of growth promotion is not associated with the bioavailability of Zn [8], and Shelton et al. [9] concluded that ZnO is the only form that will improve growth rate in nursery pigs when included at pharmacological concentrations in the diet.

Twenty-six studies, of which most were conducted in the USA, were identified that fulfilled the above criteria (Table 1). Diets were fed ad libitum and exclusively based on maize and soybean meal, except those for Heo et al. [10] and Owusu-Asiedu et al. [11] who used barley–wheat–soybean meal diets. With few exceptions [10, 12], diets were based on recommendations of the NRC [1] or earlier versions, with dietary crude protein (where indicated) varying from 180 to 274 g/kg on an as-fed basis and total lysine from 11.6 to 17.5 g/kg. Two dietary phases were used in studies by Buff et al. [13], Carlson et al. [14], Hill et al. [15, 16], Hollis et al. [17], LeMieux et al. [18], Ragland et al. [19], Smith et al. [20], Williams et al. [21], Wilt and Carlson [22], Woodworth et al. [23], and Yin et al. [24], whereas studies by Han and Thacker [25, 26] and Pérez et al. [27] applied three phases, and Carlson et al. [28], four phases. All basal diets contained mineral mixtures which supplied copper (8–22 mg/kg), iron (60–383 mg/kg), and Zn (75–200 mg/kg) to physiological requirements. In 14 studies, antibiotics were included in the basal diet. Pigs were individually housed in some treatments in studies by Buff et al. [13], Carlson et al. [14], Heo et al. [10], and Li et al. [29], whereas all other treatments kept animals in groups of two or more. Pens were used as experimental units. Age of pigs at the start of experiments varied from 10 to 36 days. Different application levels of Zn, basal diets, time of weaning of pigs, and processes to manufacture ZnO led to 72, 71, and 70 comparisons for ADG, ADFI, and G/F, respectively.

Table 1 Studies used in the meta-analysis
Full size table

Data Analysis

Differences between means for unsupplemented (\( {{\overline{X}}_{\mathrm{C}}} \)) and Zn-supplemented (\( {{\overline{X}}_{\mathrm{B}}} \)) groups were quantified with the unitless standardized Hedges's g:

$$ g=\frac{{{{\overline{X}}_{\mathrm{B}}}-{{\overline{X}}_{\mathrm{C}}}}}{{s\mathrm{p}}}J $$

Formulas to calculate the pooled standard deviation (s p) and correction factor for bias with small sample sizes (J) included in this effect size and its asymptotic standard error (SE) and precision [95 % confidence intervals (CI)] were detailed in Sales [30], Sales and Glencross [31], and Sales and Homolka [32]. A statistically significance from no effect (0) at the 5 % level (two-tailed) was declared if the 95 % CI did not include zero [33]. Effect sizes were categorized as small, medium, and large at values of 0.2, 0.5, and 0.8, respectively [34].

Fixed and random effects models, as explained in detail by Sales [30], Sales and Homolka [32], and Sales and Jančík [35], were used to calculate mean Hedges's g (\( \overline{g} \)):

$$ \overline{g}=\frac{{\sum\limits_i {{w_i}{g_i}} }}{{\sum\limits_i {wi} }} $$

where w i is the weighing factor for the ith experiment, which is 1/SE2 (within-experiment variance) in a fixed effect model, and 1/(SE2 + v) (within- and between-experiment variances) in a random effects model.

The between-experiment variance (v) was calculated as follows:

$$ v=\frac{Q-n-1 }{z} $$

with Q defined by Hedges and Olkin [36], assuming a weighing factor for a fixed effect model as follows:

$$ Q={{\sum {{w_i}\left( {{g_i} - \overline{g}} \right)}}^2}, $$

and z calculated according to DerSimonian and Laird [37] as follows:

$$ z=\sum\limits_i {{w_i}-} \frac{{\sum\limits_i {{w_i}^2} }}{{\sum\limits_i {{w_i}} }}. $$

Whereas it is assumed in a fixed effect model that there is one true effect that underlies all the studies in the analysis, with all differences in observed effects because of within-study variability (sampling error), a random effects model allowed that the true effect could vary from study to study and included between-study variability (true heterogeneity) as well as sampling error [38].

The presence of true heterogeneity among studies was identified with Cochran's Q chi-square tests and quantified with the I 2 index (a measurement that describes the proportion of total variation across experiments that is due to heterogeneity rather than chance), as described in length by Sales [30] and Sales and Glencross [31]. Comprehensive Meta-Analysis version 2 software [39] was used for the calculations. Weighted fixed effect regression analysis (SAS version 9.2, SAS Institute, Inc., Cary, NC) was applied to identify the contribution of continuous independent variables to heterogeneity according to Lipsey and Wilson [40].

Occurrence of variability (P < 0.10) among the results in the current analysis has prevented the testing for possible publication bias or bias caused by other small-study effects. Under conditions of heterogeneity, tests for publication bias may lead to false-positive claims [41].

Results

Effect sizes for ADG, ADFI, and FE showed significant heterogeneity (Table 2) that could be quantified from low (I 2 = 37 %) to medium (I 2 = 56 %). This implies that true effect sizes were different among comparisons, attributed to the possible influence of a number of factors that could vary from study to study.

Table 2 Mean effect sizes (Hedges's g) and their precision for performance calculated according to frequentist fixed effect models across all studies
Full size table

Zinc supplementation (in milligram of Zn per kilogram diet), initial live weight (in kilogram), final live weight (in kilogram), age at the start of treatment (in days), and duration of treatment (in days) presented continuous independent variables that were consistently reported across treatments and could be used in a meta-regression to explore causes of heterogeneity (Table 3). However, few of the independent variables were related at P < 0.05 to Hedges's g (Zn supplementation on ADG and ADFI, age at start on ADG, and final live weight on ADFI) or showed a trend to significance at P < 0.10 (duration on ADG and ADFI). Furthermore, the residual Q values obtained with meta-regressions indicated that all the included variables explained 11, 14, and 2 % of the variability in Hedges's g obtained for ADG, ADFI, and G/F, respectively.

Table 3 Summary of Weighted fixed effect meta-regression analysis with independent variables that influenced the effects (Hedges's g) of pharmacological zinc supplementation on average daily gain (ADG), average daily feed intake (ADFI), and gain to feed ratio (G/F) in post-weaning pigs
Full size table

As a result of excessive unexplained heterogeneity, random effects models, which included between- and within-study variability, were needed to combine effect sizes. Similar to results when fixed effect models were used to calculate summary statistics, random effects models indicated that Zn supplementation increased ADG (mean Hedges's g = 1.086, 95 % CI = 0.905–1.266), ADFI (mean Hedges's g = 0.794, 95 % CI = 0.616–0.971), and G/F (mean Hedges's g = 0.566, 95 % CI = 0.422–0.710).

Discussion

This meta-analysis presented conclusive results that pharmacological levels of ZnO increased ADG, ADFI, and G/F in post-weaning pigs. Enhanced feed intake appeared to account for a part of the improvement in growth. According to Yin et al. [24], ZnO stimulated secretion of ghrelin, a gastrointestinal hormone which acts on the small intestine and brain to promote feed intake. However, increased growth without an associated increase in feed intake [8, 20, 24, 29] indicated that the effects of ZnO on growth involve a more complex mechanism than simply enhancing feed intake. Despite several hypotheses, the exact mechanism whereby dietary Zn improves growth of post-weaning pigs is yet to be elucidated [9, 10]. Antimicrobial properties of ZnO were illustrated by changes in the gastrointestinal ecosystem of the piglet [42–44], leading to the assumption that high levels of dietary ZnO enhanced the growth of weaned pigs by controlling pathogenic bacterial scours. Carlson et al. [28] suggested a systemic effect via the blood rather than a direct influence on the gastrointestinal tract. Conversely, the effectiveness of ZnO despite its relative low availability compared to other sources of Zn indicated a local effect in the intestine [27].

Although the current meta-analysis indicated that the effects of dietary Zn on ADG and ADFI increased linearly when dietary Zn concentrations increased, a limited range and unequal distribution of application levels prevented further evaluation to establish an optimal Zn concentration for growth. Variable results were obtained in individual experimental studies. Data from nine university research stations with 1,978 weanling pigs showed that growth responses to added ZnO reached a plateau at dietary concentrations of 1,500 to 2,000 mg Zn/kg [16]. Similarly, Zhang and Guo [45] reported that ADG and ADFI did not further increase as dietary Zn concentration increased from 2,250 to 3,000 mg/kg diet. However, Shelton et al. [9] and Mavromichalis et al. [46] found linear growth responses up to 3,000 mg Zn/kg diet. According to Windisch et al. [47], a clear pharmacological effect based on ZnO requires dietary Zn concentrations of at least 3,000 mg/kg diet, which is equivalent to almost 4,000 mg ZnO/kg diet. Differences in the optimum level of pharmacological dietary Zn concentrations might be due to environmental variations and weanling age of piglets [45]. Although Hill et al. [16] reported that growth responses in early-weaned pigs (<15 days) on ZnO supplementation were greater than those in pigs weaned at 20 days or later, effects in the current evaluation increased with an increasing age at the start of experiments, which can be assumed as the weaning age. Furthermore, standardized effect sizes (Hedges's g) and random effects models were used to account for differences among studies.

The current meta-analysis did not reveal any significant influence of the duration of supplementation on growth. However, effect sizes for both ADG and ADFI showed a trend to increase when the period of supplementation increased from 7 to 35 days. It was stated by Carlson et al. [28] that feeding of 3,000 mg Zn/kg diet from ZnO during the first 2 weeks after weaning resulted in an ADG equivalent to that achieved when fed for 28 days. In their study, supplementation only during the first week after weaning, or feeding solely during weeks 2 and 3, did not result in any better growth compared to no supplementation.

When pharmacological concentrations of Zn are fed for longer than 10 days to nursery pigs, tissues become loaded and homeostatic mechanisms excrete excess Zn [48]. With most of the dietary supply being excreted, manure highly concentrated in Zn may concentrate in top soil and cause toxicity to plants and microorganisms. To reduce environmental pollution, the practice of supplementing pig diets during the first 2 weeks post-weaning with pharmacological levels of 1,500 to 3,000 mg Zn/kg diet to stimulate growth is prohibited in the European Economic Community [49]. However, ZnO remains a valid alternative to in-feed antibiotics [50] and is still used in many other countries. According to Broom et al. [51], the omission of pharmacological levels of ZnO and antibiotics from starter diets has a detrimental effect on post-weaning and subsequent (unsupplemented) pig growth performance, which increase the days to slaughter. Jondreville et al. [49] suggested that a better understanding on dietary factors and mechanisms that affect Zn availability, such as the use of microbial phytase to improve Zn availability and Zn complexed with organic compounds, is needed to safely reduce Zn supplies. However, results obtained by Augspurger et al. [52] with post-weaning pigs suggested that pharmacological concentrations of Zn chelate the phytate complex, thereby decreasing its availability for hydrolysis by phytase. Furthermore, a level of 500 mg Zn/kg diet from several organic sources was found to be inferior to 2,000 mg Zn/kg diet from ZnO in stimulating growth, feed intake, or feed efficiency in nursery pigs [17].

It can be concluded from this study that ZnO at pharmacological dietary levels presented a viable option to increase growth of pigs during the post-weaning phase. Any lowering of levels because of the potential environmental hazard of Zn will have to be counteracted by alternative dietary components or management techniques.