Multi-strain probiotics: is more actually better?

Most nutritionists now accept that a well chosen probiotic earns its place in a pig or poultry diet. That was last month's conclusion. The market agrees, and it has moved in one direction: more. Single strain products are giving way to blends of three, four, five organisms, sold almost without exception as innovative and synergistic.

The assumption is that adding strains adds benefit. Yet the research is far less clear. Some trials favour the blend, some the single strain, and plenty find no difference. Partly that's the usual suspects: different strains, doses and challenges. But there is a more awkward reason as well, and it runs through everything that follows. A blend always carries a higher total dose than any one of its components, so when a blend wins, you often can't tell whether it won on diversity or simply on numbers.

This is not an academic point. At 50 g per tonne a probiotic is a real line on the formulation, bought to hold the gut together and stand in for the growth promoters and the zinc oxide you can no longer use. A blend that quietly cancels itself out costs you twice. So does adding strains actually add anything, and if it does, when? Let's look at the research and find out.

To know whether a blend beats the strains inside it, you have to run the blend against those same strains on their own, at the same total dose, in the same trial. That sounds obvious. It is almost never done. Chapman's review put the question directly, are mixtures better than single strains, and could not find the trials to answer it(1). That review was published back in 2011, which tells you how long the gap has been sitting there.

The best attempt comes from human medicine, where the research money is. Lynne McFarland, at the University of Washington, trawled randomised trials in people from 1973 to 2019 and found 65 that qualified, covering more than ten thousand patients across eight conditions(2). Out of all that literature, only three strains had ever been tested both on their own and in a matching blend.

That is three strains out of an entire body of clinical research. And when the comparison was finally made, it cut both ways: sometimes the single strain won, sometimes the blend won, and most often there was no measurable difference at all, in which case the extra organisms are adding cost and nothing else.

Our own literature is no better. One Iranian trial fed broiler breeders a four strain product, a single strain product, both together, or nothing(3). No probiotic improved egg production, body weight or FCR. But one arm stood out: both products together, which gave higher TLR2 and TLR4 expression (the gut wall's bacteria-sensing receptors) than either product alone, and the lowest yolk cholesterol of any group.

At first glance that looks a great deal like synergy. Look more closely, though. The authors calculated that their hens were underdosed, taking in around 1.7 × 10⁷ organisms a day against the 10⁸ to 10⁹ they say is needed, and feeding two products together doubles the dose. In other words, in a trial where every bird was short of bugs, the arm that received twice as many did best.

What's odd is how rare that trial is, because it ought to be easy here: uniform birds, a controlled diet, feed conversion as a cheap, hard endpoint. It mostly doesn't get run, and it's worth remembering who would be paying for it if it did.

The clearest assessment of the question comes from a 2018 review in the Journal of Clinical Gastroenterology, which set out to test the case for blends directly(4). Its lead author was then with DuPont Nutrition & Health, a probiotics supplier.

Their conclusion, in plain language: the assumptions behind multi-strain products, more chances of success, a broader spectrum, additive or synergistic effects, are not supported by convincing evidence. They are fair about it, adding that there's no strong evidence the assumptions are wrong either. The honest position, they say, is that we simply don't know.

A team of American researchers at Penn State, with a co-author at Ohio State, did exactly that, publishing in BMC Research Notes(5). The work received no external funding, and the authors declare no competing interests.

Their subject was a proprietary Bacillus based probiotic sold for poultry. Bacillus is the workhorse genus for good reason: the organisms form spores, which survive pelleting and low pH, so they reach the bird alive(6). The blend held four species: B. coagulans, B. licheniformis, B. pumilus and B. subtilis. Of those, B. subtilis is much the best characterised against Clostridium perfringens, the organism behind necrotic enteritis(7).

The design was a co-culture: grow the pathogen alongside the probiotic and count what survives. Rather than testing only the finished product, they took it apart, running each strain alone and then various combinations, at the in-feed dose the manufacturer recommends. Everything ran in triplicate, against a pure C. perfringens control.

The complete four strain product had no measurable effect on C. perfringens at all. The pathogen count was statistically indistinguishable from the untreated control (P = 0.499).

Then they tested the strains individually, and they were not equal. Not remotely. The B. subtilis strain in that product, on its own, produced a six log reduction, taking counts from 10⁸ down to 10². The other three each managed a one log reduction. Useful, but an order of magnitude is an order of magnitude.

B. subtilis (BS); B. coagulans (BC); B. pumilus (BP); B. licheniformis (BL); C. perfringens (CP). CFU: colony-forming units. Different superscripts in the same column denote statistical differences.

So one strain was doing something powerful, and in the finished product that power had vanished. The third experiment explains why. This time they paired B. subtilis with each of its companions in turn. On its own it cut the pathogen by nearly five log. Paired with B. licheniformis it still worked, but the reduction fell to three and a half. Paired with B. pumilus, or with B. coagulans, the effect was no longer statistically distinguishable from feeding nothing at all. Every companion damaged it, and two of them, on their own, wiped it out. The authors' own summing-up is blunt. The anticlostridial properties of these strains, they write, were "negatively affected when combined with other Bacillus spp. strains." Which is a careful way of saying that the organisms in that bag were fighting each other.

One in vitro study proves nothing on its own. If this were the only evidence of strains cancelling each other out, I'd tell you to file it and wait. It isn't.

Remember Chapman, whose 2011 review asked whether mixtures beat single strains and couldn't find the trials to answer it? The following year, his group at Reading went and ran the experiment. They tested 14 probiotic strains against one another using two separate assays. Every single strain inhibited at least some of the others, and they concluded that strains may well be inhibiting each other inside a multi-strain product(8). The work was funded by a probiotics company, which had no role in the design or the analysis.

Three independent groups, using three different methods, have arrived at the same conclusion: strains in a blend interfere with one another, and it is not a rare event.

There are good biological reasons why strains in a blend work against one another.

The first is bacteriocins, antimicrobial peptides that bacteria make to kill competitors. Bacillus species are prolific producers(11). But here's the problem. A bacteriocin is not fussy about what it kills. It takes out other species in its own genus just as readily as a pathogen. So the compound that makes your B. subtilis useful against C. perfringens can knock back the B. licheniformis sitting next to it in the bag, and the other way round.

The second is quorum sensing. Bacteria switch genes on and off depending on who else is in the neighbourhood, and research in the Proceedings of the National Academy of Sciences found that the interactions shaping B. subtilis behaviour are driven mainly by members of its own genus(12). The organism most likely to change how your B. subtilis behaves is another Bacillus. The third is plainer still: four strains chasing the same substrate will suppress one another, or shift the effective one's metabolism away from the compounds it would otherwise make.

There is a fourth, and it is not biological at all. We'll come to it at the end, because it is the one you can check on a datasheet.

So how much weight should this carry? Three things limit how far it travels, and to their credit the authors flag them themselves:

In plain English: a warning light, not a verdict. The authors say so themselves and call for in vivo work. They also note the blend was never sold as a C. perfringens control, so it is hardly surprising it didn't act like one. Fair enough. But the interference is the finding here, not the failure, and interference doesn't care what the product was marketed for.

When individual studies conflict, the meta-analyses should help, and here the picture is now unusually clear.

A 2025 systematic review in Frontiers in Animal Science screened 338 broiler papers and sorted 807 experimental groups by genus: Bacillus, Lactobacillus, or a multi-genus mix(16). Only the in-feed groups were pooled, which is the route that concerns us. All three worked: weight gain rose and feed conversion improved in every category. Then the authors tested the number of strains in a mixture as a moderator of the results. It explained nothing. Neither did dose. And on the numbers that pay the bills, the mixes were beaten. Lactobacillus delivered around 222 g of extra body weight and cut the feed needed for a kilo of gain by 170 g, against 197 g and 140 g for the mixes. Their closing line is blunt for a meta-analysis: multi-genus mixtures remain the industry's most commonly used option, and that is a practice which should be reconsidered.

Read that carefully, because it's the whole argument in one sentence. The strain matters. The number of strains doesn't.

Two things before you act on the Lactobacillus result. It is broiler data, so don't carry it straight into a pig diet. And Lactobacillus does not form spores: a vegetative cell struggles to get through a pellet mill alive, which is exactly why Bacillus dominates the market. So if the pooled data says Lactobacillus is the stronger performer, it is also telling you that collecting the benefit means mash, post-pellet application or encapsulation. The best organism on paper isn't always the one that survives your process.

None of this is a verdict against blends. Bacillus blends have been reported to support gut barrier integrity under challenge in ways single strains don't(17), the mechanisms that let strains interfere could, in principle, let them cooperate, and a recent review of poultry probiotics weighs both sides and concludes the question is still open(18). But nobody has run the head-to-head at a matched dose that would settle it. And when someone finally pulled a poultry product apart, one strain was carrying the whole effect while its own companions cancelled it out. In short, there is very little evidence that adding more organisms adds more benefit.

So the message here isn't "avoid blends." Here is what that means when you sit down to choose one: