Efficacy of Prebiotics and Probiotics on Growth Performance in Turkeys: A Systematic Review Open Access

In 2022 the total turkey production in the United States was 210 million birds with a value of US$7.1 billion (1). Within this valuable industry, antibiotic supplementation in feed has been widely used for the past 70 yr after the discovery that it improved poultry growth performance (2). However, concerns surrounding a rising incidence of antibiotic resistance attributed to overuse of antibiotics has changed public perception of this practice (3). Changes in U.S. government regulations and consumer demand have been partially responsible for the reduced supplementation of antibiotics in poultry feed (4). The practice of supplementation with prebiotic and probiotic products as alternatives to antibiotics has been widely adopted in the poultry industry (5). The definition of “prebiotic” can vary but is generally considered a supplement that will enhance the resident microbiome (6). Common prebiotics include dietary fibers and carbohydrates, such as oligosaccharides and yeast products (6). The definition of “probiotic” can also vary, but is typically defined as live microorganisms that, when supplemented, provide a health benefit (5). Current probiotics commonly employ Bacillus species and/or lactic acid bacteria such as Lactobacillus (5). Synbiotics are a newer class of products being introduced to poultry producers, and this term refers to a supplement that contains both prebiotics and probiotics (5). Even with the rise in popularity of prebiotic and probiotic products within the poultry industry, studies examining the effect and efficacy of prebiotic and probiotic supplementation on poultry growth performance are limited, and the results are highly variable.

The objective of this study was to conduct a systematic review to assess the following question: Does oral administration of a prebiotic (defined as dietary fiber, dietary carbohydrates, oligosaccharides, or yeast cell wall) or probiotic (defined as single or mixed cultures of living bacteria or fungi) provide beneficial effects on growth performance in turkeys?

This review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement (7). This review was performed according to the protocol deposited at the University of Minnesota Digital Conservancy and in SYREAF (Systematic Reviews for Animals and Food; https://www.syreaf.org/), an online repository for publications of systematic and scoping reviews in food safety, animal health, and animal welfare. Any amendments to the protocol are documented in this paper. The PICO (population, intervention, comparison, and outcome) framework was used to develop the review question and eligibility criteria (8) (Table 1).

All primary research studies assessing the efficacy of prebiotics and/or probiotics in growth performance in vivo in turkeys, regardless of random allocation of treatment and use of microbiological or environmental challenges, were included in the review. Because the outcome of interest cannot be assessed and reported in a nonlongitudinal study design, studies were excluded if a cross-sectional study design was used. Searches were not restricted by language, and translation of eligible studies in a language other than English was attempted, given the resources and personnel available to the reviewers. As such, studies written in Korean, French, and Spanish were included in the review, but studies published in all other languages were excluded.

The following databases were searched for studies relevant to the review question without any limits on language and publication date: PubMed/MEDLINE, Scopus, CAB Abstracts, and AGRICOLA.

The search was conducted on December 27, 2019. It was then updated on September 13, 2021, and November 30, 2022, with modified search strings. The modified search strings looked for studies conducted in turkeys only by removing the “Chicken/growth and development” Mesh term from PubMed/MEDLINE search and “chicken” from Scopus, CAB Abstracts, and AGRICOLA searches. The citations found in the search were uploaded to the reference management software EndNote X9 (Thomson Reuters, Toronto, Canada) where duplicate records were removed. The following original search strings were used.

(“Chickens/growth and development”[Mesh] OR “Turkeys/growth and development”[Mesh] OR “Poultry Diseases/mortality”[Mesh]) AND ((“probiotics”[MeSH Terms] OR “probiotics”[All Fields]) OR (“prebiotics”[MeSH Terms] OR “prebiotics”[All Fields] OR “Dietary Carbohydrates”[All Fields] OR “Dietary Fiber”[All Fields] OR “fructo-oligosaccharides”[All Fields] OR “galacto-oligosaccharides”[All Fields] OR “yeast cell wall”[All Fields])), where “Mesh” was the Medical Subject Heading.

ITLE-ABS-KEY ((chicken* OR turkey*) AND (“feed conversion” or “growth rate” or “body weight” or mortality) AND (probiotic* OR prebiotic* OR “Dietary Fiber” OR “fructo-oligosaccharides” OR “galacto-oligosaccharides” OR “yeast cell wall”)), where “ABS” represents the record’s abstract field, and “KEY” represents the keyword field.

(chicken* or turkey*).af. and (feed conversion or feed conversion efficiency or growth rate or body weight or mortality).sh. and (probiotic* or prebiotic* or “Dietary Fiber” or “fructo-oligosaccharides” or “galacto-oligosaccharides” or “yeast cell wall”).af., where “af” represents all fields in the record and “sh” represents “subject heading” in the record.

Three outcome variables of interest were selected: average daily gain (ADG), feed conversion ratio (FCR), and mortality. ADG is defined as an average weight gain per day over a given period of time. It is calculated as body weight gain over the number of days in a growth period, where body weight gain is the difference between the final and initial weight of an animal (9). Feed conversion ratio is defined as a measure of the amount of feed it takes to gain 1 kg of weight and calculated as the cumulative amount of feed intake over the amount of weight gain (10). Mortality is defined as the number of birds that die in a flock during a specified time interval (11).

The selection of relevant studies was performed using a two-stage screening approach. Both screening stages were conducted using a Microsoft Excel spreadsheet (Microsoft Corporation, Redmond, WA) by three reviewers, who independently reviewed records. Prior to each screening process, all reviewers beta tested the spreadsheet study tool with 30 studies that were randomly chosen. The spreadsheet tool was then adjusted accordingly. The first stage, or relevance screening, was performed on all articles by three reviewers (H. H., E. A. M., and A. J.; the 2021 update by H. H. and A. J.; and the 2022 update by A. J. and T. J. J.). They assessed the title and abstract of each paper to identify studies that met the inclusion criteria using the following questions:

1. Does the title and abstract describe a primary research study reported in a peer-reviewed journal?

2. Is the study population commercial (meat-growing) turkey?

3. Does the study describe one or more treatment groups of prebiotics (dietary fiber, dietary carbohydrates, oligosaccharides, and yeast cell wall) and/or probiotics?

Studies that answered “Yes” to all three questions or those that could not be determined from the title and abstract were retained for the second screening, or design screening.

For design screening, full-text articles were retrieved, and the materials and methods section was reviewed by four reviewers (H. H., A. J., I. B., and R. V.; the 2021 update by H. H. and A. J.; and the 2022 update by A. J. and T. J. J.). The following questions were used to screen:

1. Is the article written in English? If not, is translation to English deemed practical and possible (Korean, French, or Spanish)?

2. Did the study use a commercial breed(s) of turkeys?

3. Does the study describe and specify one or more treatment groups of prebiotics (any mention of extracted or purified dietary fiber, dietary carbohydrates, oligosaccharides, or yeast cell wall) or probiotics (defined or undefined single or mixed cultures of living bacteria, fungi, and/or yeast)?

4. Was the treatment(s) administered orally (in feed/water or oral gavaged)?

5. Does the study report an appropriate comparison or control group(s)?

6. Does the study measure any of the three outcome variables (FCR, ADG, or viability/mortality) or include raw and/or relevant data such that these outcomes can be computed?

Studies that met the inclusion criteria (those that answered “Yes” to all six questions) were included for data extraction and risk-of-bias assessment. All disagreements during the screening process were resolved by discussion or by independent assessment by a third reviewer, when necessary.

A standardized data form was created using Qualtrics (Provo, Utah) to extract data. Microsoft Excel was used to synthesize evidence. Prior to the data extraction, the collection form was pretested by reviewers using 10 randomly chosen studies, and improvements were made based on their feedback. The data were extracted by two reviewers (H. H. and A. J.; the 2021 update by H. H. and A. J.; and the 2022 update by A. J. and T. J. J.). If more than one publication reported the same results from the same study population (i.e., the same results published by the same research group first as a conference proceeding and later as a peer-reviewed article), only the publication with the latest publication date was included. Table 2 lists data extracted to characterize studies in this review.

Instead of using the risk-of-bias assessment tool described in the review protocol, the revised Cochrane risk-of-bias 2.0 tool for randomized trials was used with modifications (Table 3) (12). The risk-of-bias assessment tool contains 19 items on potential biases distributed in five domains (domain 1: bias due to randomization process; domain 2: bias due to deviations from the intended interventions; domain 3: bias due to missing outcome data; domain 4: bias in measurement of outcome; and domain 5: bias in selection of the reported result). Each domain was scored as having low risk, some concerns, or high risk, and an overall risk-of-bias score was assigned using the criteria listed in Table 3. Two reviewers independently assessed the risk of bias of each paper, and the result was resolved by a third reviewer if there was disagreement between the two reviewers (H. H., A. J., and I. B.; the 2021 update by H. H., A. J., and I. B.; and the 2022 update by A. J., T. J. J., and I. B.). The quality of the study was scored on each outcome variable (ADG, FCR, and mortality) independently. Risk-of-bias decisions were captured using an Excel spreadsheet, and disagreements were resolved by a third reviewer in weekly meetings. The figures used for displaying risk of bias were created in R version 4.3.1 (13) using the package “ggplot2.”

Risk ratio (RR) was used to compare mortality between studies. Studies where only percent mortality was reported were converted to an actual number using the number of birds used in the study per group provided in the paper. RR and 95% confidence intervals were calculated in R with the package “meta” using the DerSimonian-Laird estimator (14). For FCR, the studies were compared using standardized mean difference (SMD). SMD and 95% confidence intervals were calculated with the R package “meta” using the Glass method (14). ADG was compared using absolute mean difference.

After evaluating the entire collection of prebiotic and probiotic studies, it was determined that a meta-analysis that compared all treatment groups would not be appropriate because of the wide variety of probiotic and prebiotic products used in the studies. Additionally, variability between the different treatment group types within both prebiotics and probiotics would make comparing all prebiotic or all probiotic treatment groups an inappropriate comparison. However, a forest plot was created using the R package “meta” to display the results of prebiotic and probiotic treatments and their impact on FCR and mortality. ADG was not compared using a forest plot because it was not feasible to determine variability measurements for most of the studies. The results for ADG were displayed using a boxplot created in R. One-way ANOVAs were used to compare differences between treatment groups for the boxplots displayed. Post hoc comparisons were performed using a Tukey HSD test.

Subgroups of different treatment groups were compared in a meta-analysis. Based on the available data, a meta-analysis was performed comparing treatments using probiotics from the subgroup “Bacillus” for both the mortality and FCR outcomes, and a separate meta-analysis was performed comparing the subgroup “Lactobacillus” for both the mortality and FCR outcomes. Additionally, a meta-analysis was performed comparing the use of prebiotics under the subgroup “oligosaccharide” for the mortality and FCR outcomes. All meta-analyses were performed using the R package “meta.” RR was used to compare treatment effects of meta-analyses considering mortality. A random effects model using DerSimonian-Laird estimation to calculate variability between studies was used for the mortality calculation (8). A continuity correction by adding an increment of 0.5 in all cells that were zero was used for mortality RR calculation (14). FCR was compared in the meta-analysis by SMD which was calculated using the Glass method (14). A random effects model using a restricted maximum likelihood estimator was used to calculate variability between studies for FCR (14). For some of the studies used in these meta-analyses, a unit of analysis problem arose. This was because some treatment groups compared in the same meta-analysis were from a single study using the same control group. This would cause some controls to be double counted in the meta-analysis. This issue was addressed by splitting the sample size of the shared control group (14).

A funnel plot was created in the R package “meta” to visualize reporting bias for both FCR and mortality. Not enough information was available to visualize reporting bias for ADG. An Eggers test was also performed to assess bias using the R package “meta” (15).

For the meta-analyses, any groups indicating significant effects on the outcome(s) were assessed individually by treatment group for heterogeneity and size of pooled effects. I² tests to measure heterogeneity were performed using the R package “meta.” If the I² test indicated heterogeneity, the individual effects of the studies were evaluated with a Baujat plot using the R package “dmetar” (14,16).

The original aim of this review was to assess prebiotic and probiotic supplementation effects on growth performance in both chickens and turkeys. The original search yielded 3064 papers. After following each screening step as described in the published protocol (17), 712 papers were remaining that could be used for data extraction. In further refinement of our inclusion criteria we chose to focus on papers that evaluated prebiotic/probiotic supplementation effects on turkey growth performance only. After updating the search screening process twice to include newer publications, the total number of papers included in the turkey-focused review was 38 (Fig. 1).

The overall characteristics of the 38 studies included in our review are summarized in Supplemental Dataset S1. Seven of the 38 studies included more than one trial. The publication years of the studies ranged from 1990 to 2021. The largest number of studies were published between 2007 and 2008 (n = 8, 21%). The studies represented in this review were conducted in 12 unique countries, with the largest number from the United States (n = 19, 50%) and Poland (n = 6, 16%). All studies used an experimental design with 25 (65%) reporting to have conducted randomization. Pen trials were the most common experimental unit (n = 24, 63%). Cage trials represented 13% (n = 5) and house trials represented 5% (n = 2) of the experimental units. Seven studies (18%) did not report an experimental unit.

The most used companies for sourcing genetic lines of turkeys were “British United Turkeys of America (BUTA)” (n = 11, 26%), “Hybrid” (n = 11, 26%), and “Nicholas” (n = 11, 26%). Nine studies (21%) did not report which strain of turkey was used. Because some published studies included multiple trials using different genetics company sourced turkeys, the total number of genetic companies used to calculate percentage was 42, not 38. Male turkeys were most used (n = 17, 45%), and females were used 29% of the time (n = 11). Sixteen percent (n = 6) of studies used mixed sexed turkeys for their experiments, while 11% (n = 4) did not report which sex was used. The starting age of birds for the studies ranged from day of hatch to 70 days of age. Since the definition of day of hatch can vary between 0 and 1 days, for the purposes of this study, day of hatch will be considered day 1. One day old was the most common starting age (n = 30, 79%). Two studies (5%) did not report a starting age. Experimental length ranged from 1 to 20 wk, with 18 wk being the most common length. Three studies (8%) conducted trials using birds exclusively given challenges. One of the studies, Casas et al. (1998a) challenged all the birds with cold stress, cycling the temperature in the room the birds were housed in between 27 and 35 C for the entire duration of the trial. Another study, Rahimi et al. (2019), challenged birds orally with 10⁵ CFU/ml of either Salmonella enterica serovar Heidelberg or Campylobacter jejuni on day 7 of the trial. The study by Slizewska et al. (2020) challenged all the birds with ochratoxin A–contaminated feed at an average concentration of 328 µg/kg for the entire duration of the trial. In all three studies, the challenge was performed on the nonsupplemented control group and dietary additive group(s) equally. Forty-six unique experimental trials were conducted within the 38 studies.

Risk of bias was assessed based on each unique trial (n = 46) and not on each study (n = 38) since experimental methods sometimes varied between trials within a published study. Additionally, risk of bias was assessed for each outcome (ADG, FCR, mortality) separately giving 89 individual risk-of-bias assessments. When the final score of risk of bias was broken down by outcome, ADG had the largest percentage of trials classified as “high risk” at 49% (18/37) (Fig. 2). “Some concerns” were identified in 43% of the trials (16/37), and 8% (3/37) trials were classified as “low risk.” FCR had 46% of trials (16/35) categorized as “high risk” (Fig. 2). “Low risk” was the most highly represented group with 49% of trials (17/35) falling into that category. Only 6% of trials (2/35) fell into the “some concerns” category for FCR. Mortality had the largest percentage of studies falling into the “low risk” category at 65% (11/17) (Fig. 2). Based on mortality, 24% of trials (4/17) fell under “high risk,” and 12% of trials (2/17) fell under “some concerns.”

Domain 3 investigates missing outcome data. When investigating the risk of bias by individual domain (all outcomes included together), domain 3 appeared to be the most often responsible for trials falling into the “high risk” category (Fig. 2). When comparing risk scores in domain 3 across all outcomes, 43% of trials (38/89) fell into the “high risk” category, and 57% of trials (51/89) were categorized as “low risk.” No outcomes of any trials fell into the “some concerns” category in this domain. The high number of trials falling into the “high risk” category in domain 3 was most often because of disproportionate numbers of birds within experimental groupings no longer being able to be used in the entirety of the experiment. The birds were often no longer able to be studied because of death from either illness or necropsying for other outcomes investigated in the experiments. Domain 1, which investigated randomization, and domain 4, which investigated outcome measurement, both had 100% categorization of “low risk” across all trials and all outcomes (Fig. 2). Domain 2, which investigated deviations from the proposed intervention, had 2% of trials (2/89) fall into the “some concerns” category and the other 98% (87/89) fell into the “low risk” category. Both experiments categorized into the “some concerns” group in domain 2 were two different outcomes in the same published study. Domain 5b, which investigated selection of reported results, had 64% of experiments (57/89) falling into the “low risk category” and 36% (32/89) falling into the “some concerns” category. The high proportion of studies classified into this category was most commonly because of the lack of reporting variability measurements or statistical comparisons between groups.

When comparing the different risk categories by domain broken down by outcome, ADG and mortality were the two outcomes represented in these analyses that found “some concerns” in domain 2 (Supplemental Fig. S1). Domain 3 “high risk” categories were found in 49%, 46%, and 24% in ADG, FCR, and mortality, respectively. “Low risk” categories represented 51%, 54%, and 76% for ADG, FCR, and mortality respectively. Domain 5 “low risk” categories represented 30%, 89%, and 88% of ADG, FCR, and mortality respectively. The “some concerns” category represented 70%, 11%, and 12% of ADG, FCR, and mortality, respectively. The high representation of “some concerns” being found in the ADG outcomes was because studies often did not report ADG in their results and therefore did not report variability measurement of summary statistics. Therefore, a large number of the ADG data were calculated in this study based on beginning and ending weight measurements and experimental duration.

Overall, 86 different experimental supplementations were identified across the 38 studies included in this review (Supplemental Datasets S2: ADG data, S3: FCR data, and S4: Mortality data). Seven of the studies included more than one experimental trial in the study. Prebiotics were experimentally assessed in 23% (20/86) of the experiments, probiotics were experimentally investigated 56% (48/86) of the experiments, and synbiotics were investigated in 21% (18/86) of the experiments. Thirty-four percent (13/38) of studies tested only one probiotic, prebiotic, or synbiotic at one concentration in one experimental trial. Twenty-one percent (8/38) of studies tested the same probiotic, prebiotic, or symbiotic at different concentrations. Thirty-two percent of studies (12/38) investigated more than one type of prebiotic, probiotic, or synbiotic. Of the 12 studies investigating multiple types of supplements, eight of those studies included a combination of different types of supplements (prebiotic, probiotic, and synbiotic). Forty-three unique individual or mixtures of prebiotic, probiotic, and synbiotics were tested. The commercial prebiotic Bio-Mos mannan oligosaccharide was tested the most (7/38 studies). Bacillus subtilis was the probiotic species most often tested (5/38 studies). The commercial probiotic BioPlus2B (which contains a mixture of Bacillus subtilis and Bacillus licheniformis) was evaluated in an additional four studies. Prebiotics or probiotics were supplemented in feed in 87% (33/38) of experiments. A combination of water and feed was used in 8% (3/38) of studies, and water and oral gavage were each used in one singular study (3% for each).

ADG was either directly measured or calculated for 76% of studies (29/38) included in this review (Supplemental Dataset S2). Sixty-one different prebiotic or probiotic supplements in different combinations and concentrations were compared across those 29 studies. ADG results ranged from +10 to −3 g/d mean difference compared to nontreatment controls. The average was +2 g/d mean difference from the controls.

Feed conversion ratio was measured or calculated for 79% of studies (30/38) included in this review (Supplemental Dataset S3). Sixty different prebiotic and probiotic supplements in different combinations and concentrations were compared across 30 different studies. FCR results ranged from +0.18 to −0.24 points mean difference compared to nontreatment controls. The average was −0.038 points mean difference from the controls.

Percent mortality was measured or calculated for 13/38 (34%) of studies included in this review (Supplemental Dataset S4). Twenty-three different prebiotic and probiotic supplements in different combinations and concentrations were compared across 13 different studies. Percent mortality results ranged from +3.61 to −7 percent difference compared to nontreatment controls. The average was −2 percent mortality difference from the controls.

Because of limited variability in the data, the ability to synthesize and visualize ADG data was restricted. Because of these limitations the results for mean difference of ADG are only displayed in boxplots (n = 61; Fig. 3). The results of ADG were compared by treatment type. No statistically significant differences were found between supplementation type (P = 0.102). The prebiotic group had the numerically highest effect on ADG with a mean difference of +3.4 g/d (n = 14). The probiotic group had a mean difference of +1.4 g/d (n = 36), and the synbiotic group had a mean difference of +2.5 g/d (n = 11).

Boxplots comparing FCR by supplementation type were also created using all available FCR data (n = 60; Fig. 4). Standardized mean differences in FCR points were compared between supplementation types (Fig. 5). No statistically significant differences were found between treatment and nontreatment controls (P = 0.82). The prebiotic group had the most beneficial effect on FCR with a SMD of −0.048 points (n = 19). The probiotic group had a SMD of −0.033 points (n = 29), and the synbiotic group had a SMD of −0.034 points (n = 12).

Percent mortality by supplementation type was examined using all available percent mortality data (n = 23) (Fig. 6). Difference in percent mortality was compared between supplementation types (Supplemental Fig. S2). There were no statistically significant differences in percent mortality between supplementation types (P = 0.058). The synbiotic group had the numerically best effect on mortality with a difference of −4.0 percent mortality (n = 4). The probiotic group had a difference of −2.4 percent mortality (n = 13), and the prebiotic group had a difference of +0.06 percent mortality (n = 6).

Using probiotic species in the same genus or prebiotics with similar sugar compositions and the available data for FCR and mortality, meta-analyses were performed on a subset of the prebiotic and probiotic supplementation groups. The genera with the largest number of representatives with similar experimental methods were Bacillus and Lactobacillus. Additionally, the effects of mannooligosaccharides and fructooligosaccharides prebiotics were compared based upon FCR and mortality.

For the outcome FCR, 11 representative groups were identified in the Bacillus genus, eight representative groups in the Lactobacillus genus, and 13 representative groups in the oligosaccharide supplementation comparison (Fig. 7). No statistically significant effects were identified for any of the treatment groups (Bacillus group: P = 0.17; Lactobacillus group: P = 0.24; oligosaccharide group: P = 0.11). The Bacillus group had an SMD range of −1 to +0.24 points with an average of −0.15 points. The Lactobacillus group had an SMD range of −0.52 to +0.21 points with an average of −0.04 points. The oligosaccharide group had an SMD range of −0.58 to +0.37 points with an average of −0.10 points.

For the outcome mortality, five representative groups were identified in the Bacillus genus, five representative groups in the Lactobacillus genus, and five representative groups in the oligosaccharide supplementation comparison (Fig. 8). There were no statistically significant effects of the treatment groups Bacillus genus group and oligosaccharide group on the risk of a mortality event (Bacillus group: P = 0.75; oligosaccharide group: P = 0.43). Both of these groups had a mean RR >0.90 (but <1.0), indicating they had a very marginal reduction in the risk of a mortality event. There was an observed significant effect within the Lactobacillus genus group (P = 0.02) with a RR of 0.73 indicating that supplementation with Lactobacillus significantly reduces the risk of a mortality event. However, when evaluating heterogeneity, the I² value was high at 89% (below 50% is acceptable, above 75% is considered highly heterogeneous), and heterogeneity also had a statistically significant P value (P < 0.01). This indicates that the effect seen in this meta-analysis is most likely influenced by heterogeneity (13). A Baujat plot (Supplemental Fig. S3) determined that the very large study “Casas et al. – b Commercial” was not significantly affecting heterogeneity but had a large influence on pooled effect measure. The group “Casas et al. – b Preliminary” was very skewed and had a large influence on heterogeneity but did not substantially influence the pooled effect measure. Even after removing the treatment group identified as impacting heterogeneity (Casas et al. – b Preliminary), the meta-analysis effect result was still heavily influenced by one very large treatment group (Casas et al. – b Commercial). Since the study group sizes were so different between studies it may not be appropriate to perform a meta-analysis comparing these Lactobacillus treatment groups.

Funnel plots for the outcomes FCR and mortality (Fig. 9) showed that the mortality funnel plot did not appear to be asymmetrical, and an Egger’s test confirmed the absence of plot asymmetry (P = 0.52). When evaluating the funnel plot for FCR, asymmetry was observed (P = 0.0079). Publication of FCR results appeared to be skewed toward lower FCR. Specifically, more studies were published that had low FCR effects, especially in the statistically significant range.

In this review we evaluated the available literature comparing prebiotic, probiotic, and synbiotic supplementation effects on turkey growth and performance, using ADG, FCR, and mortality. A wide range of different prebiotic, probiotic, and synbiotic supplements were identified as being tested in poultry, specifically turkeys. This review found that a large proportion of studies investigating the use of these products in turkeys yielded results that were not statistically or even numerically different. More than 30% of studies that performed statistics on the outcomes investigated in our study reported a statistically significant result. Study variability appeared to be large and contributed to this finding. Even when comparing subsets of supplements based on similarity (Bacillus, Lactobacillus, and oligosaccharides), the results still maintained substantial variability. These data put a spotlight on the lack of consistency in efficacy when utilizing probiotic and prebiotic supplementation in turkeys. The variability in results between studies could be influenced by a variety of factors, one of which was the lack of consistency in concentration of supplements given. There was wide variability in the supplement dose, even between different studies using the same commercial product. Bird age when treatment began, delivery method, and supplement frequency were similar in many of the studies. However, study length and turkey genetic line used differed more than the aforementioned parameters and could also have contributed to the variability in results. Prebiotic supplementation had the most beneficial effects when assessing ADG and FCR, which were of greater magnitude than symbiotic or probiotic products. It can be noted that in all outcomes, the probiotic group had considerably larger numbers of treatments tested. This could indicate less publishing discretion compared to the other treatment group types; however, publishing bias based on treatment type was not investigated in this study. Despite beneficial results in ADG and FCR, prebiotic supplementation had a slightly negative effect on mortality, with the prebiotic supplementation group having a mean difference of +0.6% mortality compared to controls. This contrasts with the synbiotic and probiotic groups, which had mean differences of −4.0% and −2.4% mortality compared to controls, respectively. In the meta-analyses groups, all treatments had a beneficial effect on both outcomes examined. Though those effects were sometimes numerically small, such changes in performance can have large financial impacts for a producer. Lactobacillus supplementation initially appeared to have a large influence on decreased mortality, in contrast with the marginal effects of Bacillus and oligosaccharides. Notably, though, these results appear to be skewed by two very large trials published in the same paper with notably high percent mortality in their control groups (Supplemental Dataset S4).

One limitation of this study was that weight gain, a significant aspect of poultry growth performance measurements, was not able to be adequately and thoroughly assessed. We used ADG to measure this, but most studies did not report this metric. In the process of this review, we were able to calculate ADG using starting and final weights and experimental duration, but the factors and data needed to be able to perform a formal meta-analysis, create forest plots, and assess publication bias were not available. Being able to compare weight gain more adequately between studies would have been beneficial. We opted to focus on commercial turkey production and exclude broiler production because turkey production research is dwarfed in comparison to chicken production research (highlighted by the differences in search string returns in the beginning of this study), and there is general consensus by the turkey industry that broiler performance trials cannot be used to infer possible benefits in turkey production. As such, we had a limited number of studies to investigate. A robust meta-analysis was not possible based upon limited data provided by these articles. As such, it was difficult to parse out small effects that potentially could be found if more studies were included. This contrasts with research on prebiotic and probiotic supplementation on chicken growth performance, which has significantly more related research published.

The large number of different commercially available supplements tested in turkeys alone is reflective of the interest in their use as antibiotic alternatives for poultry production. The changes in use of antibiotics in poultry is clearly fueling the need for alternatives to antibiotics and contributing to the steady increase in development of prebiotics and probiotics (4). This review illustrates the variability of these studies in their approaches, experimental design, and reporting. That, coupled with variability of results and reporting biases, leaves questions regarding the true efficacy of these different supplements. Prebiotic and probiotic use has clearly gained traction within turkey production based on promising research in mammals and other poultry showing intestinal health benefits. However, these effects and their subsequent influence on growth performance cannot be understood if consistent methodologies and reporting are not utilized.

This review highlights the overall relatively small amount of robust peer-reviewed research for prebiotic and probiotic products in commercial turkey production, as well as the large variability in results. These results are most likely magnified by the lack of uniform methods used to measure and investigate their effects on growth performance, and this review was hampered by studies that lacked the required effect measures. To be able to accurately investigate which products are in fact effective, steps need to be taken to standardize experimental methods and reporting in this research area. Furthermore, the clear reporting biases observed here underscore the problem of utilizing such data to inform on-farm decisions regarding product choices.

Supplemental data associated with this article can be found at https://doi.org/10.1637/aviandiseases-D-24-00034.s1.

This work was supported by Agricultural and Food Research Initiative (AFRI) grant 2018-68003-27464 from the USDA National Institute of Food and Agriculture.

ADG =

average daily gain;

FCR =

feed conversion ratio;

PICO =

population, intervention, comparison, and outcome;

RR =

risk ratio;

SMD =

standardized mean difference