The CFU Number Is the Wrong Thing to Compare

Walk down a supplement aisle and the probiotic shelf reads like a bidding war. Ten billion. Fifty billion. One hundred billion. The implied logic is obvious and almost never stated out loud: more organisms, more effect.

The published evidence does not organize itself that way.

What a CFU counts

CFU stands for colony forming unit, and it is a viability count. Laboratories arrive at it by culturing the product and counting what grows, using the plate count methods applied in the label-accuracy studies below. Because it counts only cells capable of forming a colony, it is not a measure of mass, and on its own it does not establish which organisms are present. That is why those same studies had to add genotyping or mass spectrometry alongside the plate counts to check identity.⁴⁵

The most widely used definition of a probiotic comes from a 2013 expert panel convened by the International Scientific Association for Probiotics and Prebiotics, which reaffirmed the earlier FAO and WHO wording: live microorganisms which when administered in adequate amounts confer a health benefit on the host. The panel’s stated purpose was to make use of the term more precise so that clinicians and consumers could differentiate the diverse products on the market.¹

Note the phrase “adequate amounts”. The definition requires an adequate dose. It does not say what adequate is, or that it is the same number for every organism and every outcome.

What dose-response research has found

A 2017 review in Beneficial Microbes gathered the human studies that had actually varied the dose, either as meta-analyses or as trials comparing two or more doses of the same product.²

The picture it describes is uneven rather than linear. For antibiotic associated diarrhea, both meta-analyses and dedicated dose-response studies observed a positive relationship between dose and risk reduction, which the author called compelling. For Clostridium difficile associated diarrhea, no dose-response was observed. For necrotising enterocolitis, atopic dermatitis, and slow intestinal transit, meta-analyses identified no dose-response relation. For relief of irritable bowel syndrome, none was determined. For blood pressure, one meta-analysis observed that doses above 10¹¹ CFU were more effective than lower doses.²

Two qualifications in that review matter more than any single finding. First, the author states plainly that these findings do not allow for extrapolation: a dose relationship demonstrated for one organism and one outcome cannot be carried across to a different organism or a different outcome. Second, the absence of a clear dose-response for most endpoints does not prove that none exists. It means the available data do not support drawing the conclusion either way.²

That is narrower than “more is better” and narrower than “dose does not matter”. It is also the accurate one.

When dose has been tested directly

Dose-ranging trials here are rare, which is part of why the question stays open. One better-powered example ran 307 adults with diarrhea-predominant irritable bowel syndrome, allocated one to one to one, to placebo or to Lactiplantibacillus plantarum Lpla33 at either 1 x 10⁹ or 1 x 10¹⁰ CFU per day for eight weeks.³

Both active groups recorded larger reductions in total IBS severity score than placebo, and the higher dose group recorded a greater absolute reduction than the lower dose group. Stool consistency responders made up 62.5 percent of the lower dose group and 88.4 percent of the higher dose group, against 26.3 percent on placebo. No significant shifts in microbial diversity were observed.³

That is a real dose-ranging signal, measured rather than assumed. It is also a signal for one strain, at one pair of doses, over eight weeks, in one patient group. The study was funded by the strain’s manufacturer and three of the four authors were its employees, which the paper discloses.³ None of it transfers automatically to a different organism on a different shelf.

Strain specificity is the reason. NCCIH states that different types of probiotics may have different effects, and that a benefit shown for one specific kind of Lactobacillus does not necessarily mean another kind of Lactobacillus would do the same thing.⁶ A CFU count compares quantity across products whose contents are not interchangeable in the first place.

Whether the number on the label is even in the bottle

The comparison also assumes the label is accurate. Two independent laboratory analyses have tested that assumption, one in Pakistan and one in the United Kingdom. Neither sampled a US shelf, which is worth holding onto before either set of figures is read as a description of the bottle in your hand.

A 2022 analysis in Probiotics and Antimicrobial Proteins, carried out at a national probiotic laboratory in Pakistan on what its authors describe as the local market, cultured and genotyped marketed probiotic supplements intended for human consumption. Almost none of the products tested met their label claim. Eleven percent had no viable cells at all. Twenty-seven percent showed significant inter-batch variation. Around a quarter of the total microbial content consisted of organisms that were not the ones claimed, and half the products contained antibiotic-resistant strains. The authors described the market they sampled as essentially unregulated and called for regular market surveillance.⁴

A smaller 2025 study in Microorganisms quantified viable bacteria in twelve commercial probiotics sold in the United Kingdom, seven for poultry and five for human use, using selective plate counts and MALDI-TOF mass spectrometry against declared values. All twelve deviated significantly from their label claims, although eleven of the twelve fell within a practical acceptance range of plus or minus 0.5 log CFU. The human products were the more variable group in both species composition and stated count, and one of them contained no detectable viable bacteria. One labelled species, Bifidobacterium bifidum, was absent from a human product, and some Bacillus isolates turned out to be different species than labelled.⁵

Both found that the printed CFU figure and the measured CFU figure were not the same number, and both found identity errors as well as count errors. They reached different verdicts on severity, which is unsurprising given they sampled different markets under different regulatory regimes.

The provenance should be stated plainly rather than left to be inferred. Read literally, the 11 percent with no viable cells describes products bought in Pakistan, and the one product with no detectable viable bacteria describes products bought in the United Kingdom.⁴⁵ What the two papers establish is that a printed CFU count is a claim rather than a measurement, and that independent testing has found the printed and measured figures to diverge in more than one market. What they do not establish is a rate of divergence on US shelves. Comparable published US shelf-testing of this kind is not what either paper provides.

What to compare instead

A CFU figure is only interpretable once three other things are known.

The full strain designation, not just the genus and species. Lactiplantibacillus plantarum Lpla33 (DSM34428) is a specific organism with specific published data behind it.³ “Lactobacillus” on its own identifies a group whose members are not expected to behave alike.⁶

The outcome. Dose relationships in the published record are outcome-specific, and a relationship observed for antibiotic associated diarrhea was not observed for several other endpoints.²

The dose that was actually studied for that strain and that outcome, rather than a larger round number chosen for the front of a box. Where a dose comparison has been run, as in the eight week IBS-D trial above, the doses tested were 10⁹ and 10¹⁰ CFU per day.³

One broader caution belongs here. In a meta-analysis of 25 case-control studies, breath testing classified 35.5 percent of patients with irritable bowel syndrome as having small intestinal bacterial overgrowth, and 29.7 percent of controls the same way.⁷ A microbial framing is not automatically a precise one.

The practical read

The CFU count is a viability measurement, and a useful one, but it is a within-product specification rather than a between-product ranking. Comparing 50 billion of one organism against 100 billion of another compares two numbers that were never on the same scale, before accounting for the finding that, in the two markets where this has been tested independently, the printed figure and the measured figure differed.⁴⁵ Persistent digestive symptoms are worth discussing with a clinician rather than escalating through the CFU ladder.