What a broiler synbiotic meta-analysis supports, and the one claim it does not

What the study did

What it does and does not establish

The solid part: fifteen trials, all running 40 days or longer, a requirement to report standard deviations, Hedges' g correction for small-sample bias, and conventional and challenged conditions pooled separately. By the standards of the feed-additive literature this is a tidy piece of work, and the three growth metrics point the same way.

Two findings need careful reading:

• The footpad dermatitis result is fragile. It rests on three of the fifteen studies, the upper bound of the confidence interval is 0.98 — touching 1 — and the authors state heterogeneity was substantial. That is a signal worth following, not a number for a product specification.

• Effects were larger under challenge than under conventional conditions. That is mechanistically plausible: there is more room to improve a gut under pressure. It also means the larger numbers in this evidence base come from experimentally induced disease models, not from normal production. A well-managed house with low coccidiosis and enteritis pressure should budget against the conventional stratum, not the challenge stratum.

"As good as antibiotics" is the one claim this analysis cannot support

This is the part worth getting right, because it is the sentence most likely to be lifted into a sales deck.

What the paper reports is that growth performance did not differ significantly between the synbiotic and AGP or coccidiostats in the available trials. Failing to detect a difference and demonstrating equivalence are not the same result.

The defensible statement is: in several trials, no difference in growth performance was detected between this synbiotic and AGP or coccidiostats. The statements that do not follow: that it is equivalent to AGP, that it can replace AGP, or that an antibiotic programme can be adjusted on the strength of this analysis.

This is not merely a wording problem. Antibiotic replacement is among the most closely scrutinised claims in feed, in every market we sell into, and presenting an undetected difference as equivalence does not survive examination under any of them.

The resolution you need depends on what you are claiming

The abstract names all five organisms and gives no strain designations. That is often treated as a defect, but "are there strain numbers?" is not a useful question on its own. The useful question is what the evidence is being asked to support.

Start with what usually gets left out: strain identity is not the only variable, and frequently not the dominant one. Take the same organism into a different host (broilers to weaned piglets), a different age, a different dose or a different challenge pressure, and the result can change entirely. This analysis is itself the evidence for that — the ADG effect under challenge (SMD 1.26) is more than double the conventional one (0.59) with no change of strain at all. What changed was the conditions. Putting "is it the right strain?" ahead of "is it the right host, condition and dose?" gets the order wrong.

With that in place, here is what species composition and formulation can carry:

• Direction of effect: species and formulation are a reasonable unit. The mechanisms at work in this kind of blend are ecological — competitive exclusion, acidification, short-chain fatty acid production, niche occupancy — and those capabilities are conserved to a considerable degree at species level. A blend built from the same five species, at a comparable ratio, with a fermentable substrate, used in the same host under the same conditions, can reasonably be expected to point the same way. That is precisely the value of general formulation research.
• Magnitude: that needs its own data. Direction and size are separate questions. Grams gained and points of FCR track the specific formulation, the viability of the specific organisms, and the production conditions.
• And the weight of strain-level variation differs a great deal between taxa — far more useful than a yes/no rule about strain numbers:
• - Bifidobacterium animalis subsp. lactis is close to genomically monomorphic; strains within the subspecies differ very little. Discussing function at that rank has a biological basis.
• - Lactiplantibacillus plantarum, by contrast, has an open pangenome, with real strain-level variation in both probiotic-associated and safety-related genes — we cite the 1,374-genome comparison in species, subspecies and what each rank guarantees.
• - Worth noting in passing: the paper uses Ligilactobacillus salivarius and Limosilactobacillus reuteri, the current names after the 2020 reclassification of Lactobacillus, rather than the superseded genus. Mixed old and new nomenclature in specifications is a common problem; this paper got it right.

• Safety and registration: here there is no flexibility. This is the one place we hold to strain level, and the reason is regulatory rather than a preference about biology:
• - E. faecium contains both authorised feed additive strains and vancomycin-resistant clinical isolates within the one species. EFSA issued separate safety guidance for E. faecium in animal nutrition, requiring per-strain assessment of acquired (transferable) resistance and virulence determinants, precisely because the within-species variation in this organism sits in safety-related genes.
• - And whatever view one takes of the biology, feed additive authorisations are granted to specific deposited strains, not to species. Registration dossiers and customer audits are built per strain — a separate question from whether a species name carries function.

So the reasonable use of this analysis is: it supports the directional judgement that this class of five-species synbiotic improves growth performance in broilers. It does not support specific numbers for any particular product, and it does not substitute for the safety documentation of any individual strain.

A standardised effect size is not a number you can cost

There is also a practical problem: SMD is dimensionless.

An SMD of 0.74 means a difference of roughly 0.74 standard deviations. It does not tell you how many grams of body weight, or how many points of FCR. The same SMD corresponds to a small absolute difference in a tightly distributed trial and a considerably larger one in a dispersed trial.

A note from an ingredient manufacturer

What this means for formulators and producers

If you are assessing this direction

References

1. A multi-species synbiotic improves growth performance in broiler chickens under conventional and enteric-challenge conditions: a meta-analysis. Poultry Science, 2026. doi:10.1016/j.psj.2026.108031. PubMed record
2. EFSA FEEDAP Panel. Guidance on the safety assessment of Enterococcus faecium in animal nutrition — feed strains in this genus require per-strain assessment of acquired resistance and virulence determinants. EFSA Journal, 2012. doi:10.2903/j.efsa.2012.2682. EFSA record

This note summarises and interprets published research in order to explain how formulation identity, statistical design (the difference between non-inferiority and an undetected difference), effect-size units and challenge conditions bound the applicability of evidence. It is not a claim about the efficacy or safety of any product, is not directed at any company or brand, and is not regulatory advice. Baifude does not supply the product reviewed here, and does not rely on this note to support any efficacy claim for its own products. Permitted categories, approved species lists and labelling requirements for feed additives differ between jurisdictions; the current official notices in the market concerned govern in every case. Findings apply only to the animals, strains, ratios, doses and challenge conditions described in the original source.