What Are Probiotics? Live Cultures, CFU and What the Trials Show

Probiotics are live bacteria and yeasts taken in a measured amount for a health effect; in Great Britain the word itself counts as a health claim with no authorised wording, so UK packs say "live cultures" and list the species and the count instead.^{1, 2} What they do is answered use by use: from the first days of an antibiotic course, live cultures cut diarrhoea from 19% to 8% in 33 children's trials, while for a stomach bug already under way the largest trials found no difference from placebo.^{3, 4, 5} This is the hub of the Wellgard probiotics guide.

Evidence at a glance

• Trials: 14 pooled analyses, the largest 82 trials in 12,127 people.⁴
• Largest single trial: 2,981 UK inpatients aged 65 and over, 21 days.⁶
• Strongest measured outcome: diarrhoea during an antibiotic course, 8% against 19% across 33 children's trials.³
• Not tested: whether a count at manufacture and one at best-before give a different result in anyone.

Sources checked on 16 October 2026.

What are probiotics?

The scientific definition, agreed by an expert panel in 2014, is "live microorganisms which when administered in adequate amounts confer a health benefit on the host".¹ Three words do the work. Live: a pack counts cells that can still grow; dead cells do not count. Adequate amount: the dose a trial used. Health benefit: measured in a trial, not assumed from the species. A prebiotic, the fibre those organisms feed on, is a separate definition: what inulin is. The organisms are lactobacilli, bifidobacteria, a few other bacteria and the yeast Saccharomyces boulardii, sold freeze-dried in capsules, sachets and powders, or alive in yoghurt and kefir.

Why do UK labels say "live cultures" and not "probiotic"?

Because in Great Britain "probiotic" is read as a health claim in its own right, and no live culture has an authorised health claim on the GB register.² The register records that the applications did not meet the regulation's tests on strain identity, on what counts as a beneficial effect and on evidence in the general population; it is not a finding that live cultures do nothing: much of the clinical evidence is in patient groups the regulation treats differently, and most of it postdates the assessments. The one authorised live-culture wording belongs to yoghurt: its live cultures "improve lactose digestion of the product in individuals who have difficulty digesting lactose", conditioned on at least 100 million live starter cells per gram.⁷

Genus, species, strain: how do you read a label?

"Lactobacillus acidophilus" gives a genus and a species; a strain is one cultured line within the species, and it is the strain a trial tests. In 2020 the genus Lactobacillus, then 261 species, was split into 25 genera on whole-genome sequencing, so L. rhamnosus is now Lacticaseibacillus rhamnosus and L. reuteri is Limosilactobacillus reuteri, while L. acidophilus kept its name; the bacteria and their trials are unchanged and "lactobacilli" remains the group word.⁸ "B. lactis" is Bifidobacterium animalis subsp. lactis. Species-only labelling is lawful; a strain code names the exact culture studied, a research aid rather than a promise.

What does "20 billion live cultures" mean, and at what date?

It is a count of colony-forming units (CFU), the cells still able to grow when the pack was tested. UK rules do not say whether the count is at manufacture or at best-before; ask which. Three things reduce the live count in a pack: moisture, oxygen and heat. In laboratory storage one Lactobacillus rhamnosus strain lost about half its cells over 14 months at low moisture and more than 99% at a slightly higher moisture, and in a 12-month study of shop-bought supplements, storage temperature was the main determinant of survival.^{9, 10} So from under half to nearly all of the cells can be lost by the end of shelf life, depending on strain and moisture; a count guaranteed to the best-before date tells you what you will take. A total count says nothing about the split between species, and the law does not require it. Nor is "20 billion" a target: effective doses in the trials ran from 100 million to 100 billion a day by strain and outcome.³ How strong a probiotic should be and whether it needs the fridge take these further.

What do probiotics do for the body?

Only what a trial has measured, and the measurements differ by use.

The strongest results come from antibiotic courses: across 19 hospital studies, starting within two days of the first antibiotic dose gave a relative risk of 0.32 for C. difficile infection and starting later 0.70, and the largest UK trial, 2,981 inpatients aged 65 and over whose capsule started up to a week into the course, found 10.8% with diarrhoea against 10.4% on placebo.^{18, 6} For IBS, NICE advises anyone who tries them to "take the product for at least 4 weeks while monitoring the effect".¹⁹ Use by use: do probiotics work, IBS, bowel habit, skin and mood.

Are probiotics good for you?

For a healthy adult, they are well tolerated and the measurable benefits are specific rather than general. In pooled trials, side effects were no more common on live cultures than on placebo: relative risk 1.00 across 36 antibiotic trials in 9,312 adults and 1.02 across 23 respiratory trials; the usual early reports are wind, bloating and looser stools (worse before better).^{20, 13} The one healthy-population outcome with a pooled result is colds: a relative risk of 0.76 for at least one upper respiratory infection over more than three months of daily use, low certainty.¹³ "Good for your gut" as a general claim has no authorised wording and no pooled trial. On the heart: no pooled analysis reports harm; the cardiology warning rests on case reports in people with valve disease (when you should not take a probiotic).

Where did the trials find little or no difference?

Weight: across 15 trials in 957 overweight adults, live cultures were associated with 0.6 kg less body weight than placebo, which the authors called small; the claim that L. acidophilus causes weight gain traces to a 2012 analysis that pooled 17 human trials with 65 farm-animal and laboratory studies (SMD 0.15); later adult trials found no weight change (acidophilus and weight).^{17, 21} "Detox" and "leaky gut": no trial, and no diagnosis in UK practice. Eczema that already exists: 39 trials in 2,599 people found symptoms 0.44 points lower on a 0 to 20 scale, little or no difference.¹⁶ A stomach bug already under way: in the low-risk-of-bias trials of a 2020 Cochrane review, diarrhoea lasting 48 hours or more did not differ from placebo.⁴ None is a verdict on live cultures; each is a measured outcome where the trials found little or no difference.

Should I take probiotics?

Treat it as a trial of one. Pick the outcome you can count, choose a product whose label lets you follow what the trials did for it, take the label dose every day, and judge it at the trial's own time point: the antibiotic course itself, four weeks for IBS and bloating, two to eight weeks for constipation, months for colds.^{3, 19, 14, 13} If nothing has changed by twelve weeks, UK guidance for IBS says stop.²² Two things are not reasons to start. "Signs you need a probiotic" is not a diagnosis: there is no test for "low probiotics", no such condition as a "probiotic deficiency", and persistent symptoms need a GP (the symptoms people search for). Antibiotics months ago are not one either: the positive trials dosed during the course. How to choose a probiotic turns this into a label check.

How do you take live cultures with antibiotics?

From the first days of the course, every day until it ends, with the antibiotic taken exactly as prescribed. In the 2021 adult pooling (36 trials, 9,312 people) the benefit held whatever the interval between culture and antibiotic; the authors advise starting "as early as possible".²⁰ Many antibiotics kill bacterial cultures, which is why packs suggest a gap of a few hours; no trial has tested that gap. Waiting until the course has finished is the opposite of what the positive trials did, and in one 2018 gut-lining study an 11-strain product taken after a week of antibiotics delayed the return of volunteers' own bacteria; a single study whose clinical meaning is not known.²³ In full: live cultures with antibiotics.

Foods with live cultures or a supplement?

Both deliver live organisms; only one tells you which and how many. A fermented food carries live cultures only if nothing has killed them since fermenting: yoghurt with live cultures, kefir, raw sauerkraut and kimchi do; sourdough bread, most jarred sauerkraut and pasteurised kombucha do not; fresh fruit never did.²⁴ A jar names no species or count, a capsule states both, and the trials in the table used defined strains at stated doses; this page does not rank the two.

Who should take care, and what about pregnancy?

Four groups ask their clinical team before any live-culture product: people with a weakened immune system, people who are seriously ill or in hospital with a central line, premature babies, and anyone whose doctor has said so. The reason is case series, not trials: in a Finnish series of 89 Lactobacillus bloodstream infections, 82% of patients had severe or fatal underlying disease and in 11 the strain matched a widely used commercial strain.²⁵ In a healthy person this is not a documented risk. Pregnancy is neutral because a 2021 Cochrane review of trials that gave live cultures to lower the chance of gestational diabetes found more pre-eclampsia among the women who took them (RR 1.85, high-certainty evidence), so anyone pregnant or breastfeeding speaks to a midwife or GP first.²⁶ The safety canonical: when you should not take a probiotic.

How long do you take them, and what should you expect?

For as long as the outcome needs, and expect them to pass through rather than settle: in 117 people tracked for eight weeks on one bifidobacterium strain, stool levels rose during dosing and approached baseline once it stopped, and in volunteers whose gut linings were sampled during an 11-strain course, the strains were alive after the stomach but settled differently in each person.^{27, 28} No trial reports a rebound or a dependency on stopping. Duration: how long to take probiotics; the mechanics: how to take probiotics; the hour: morning or night.

Every guide in this series

The evidence: do probiotics work, how long they take to work, IBS, bowel habit, do they clean you out, skin, anxiety, psychobiotics, periods and after a colonoscopy. Taking them: how to take probiotics, morning or night, how long, with antibiotics, how strong, acidophilus daily, the fridge, what to avoid and how to choose. Safety: when not to take one, side effects, worse before better, diarrhoea, acid reflux, tiredness, when you don't need them, omeprazole, without asking a doctor and what disease Lactobacillus causes. Species and foods: acidophilus, acidophilus and weight, L. rhamnosus, L. reuteri, foods with L. reuteri, L. reuteri yoghurt, foods high in Lactobacillus, drinks, apple cider vinegar, seven things for gut health, symptoms of low probiotics and psyllium husk. For women: what probiotics do for women, the Wellgard BV guide, discharge and taste.

What we don't know

• No trial has compared a count labelled at manufacture with one guaranteed to the best-before date on any outcome.
• No trial has randomised the day a culture starts against an antibiotic course; the two-day finding is a meta-regression.¹⁸
• No pooled analysis follows an outcome after stopping beyond stool levels of the strain.²⁷

Frequently asked questions

Is it OK to take probiotics every day?

Daily use is the norm in the trials: the respiratory trials ran for more than three months with adverse events no more common than on placebo (RR 1.02), and stool tracking shows the strain is present only while taken.^{13, 27}

What are examples of probiotics?

Lactobacillus species and their 2020 offshoots (L. acidophilus, Lacticaseibacillus rhamnosus, Limosilactobacillus reuteri), bifidobacteria (B. animalis subsp. lactis, B. longum), the yeast Saccharomyces boulardii and the spore-former Bacillus coagulans; among foods, live yoghurt, kefir, raw sauerkraut and kimchi.^{8, 24}

Does the NHS recommend probiotics?

No, and it does not advise against them: the NHS page, last reviewed 24 November 2022 and past its review date, says "there's some evidence that probiotics may be helpful in some cases", names IBS symptoms, and says that for most people "probiotics appear to be safe".²⁹

Can long-term use of probiotics be harmful?

No trial has found harm from long use in healthy adults; one bifidobacterium strain that persisted for six months in 30% of people caused no symptoms and no change to the rest of their gut community.^{13, 30}

Do probiotics detox the body?

No. No trial has measured a toxin leaving the body because of a live culture, and "detox" has no authorised wording.

Sources

1. Hill C, et al. Expert consensus statement on the term probiotic. 2014. PubMed 24912386.
2. DHSC. Great Britain nutrition and health claims register. Published 31 December 2020, last updated 19 May 2026. gov.uk.
3. Guo Q, et al. Cochrane review, 33 randomised trials, 6,352 children. 2019. PubMed 31039287.
4. Collinson S, et al. Cochrane review, 82 trials, 12,127 people with acute diarrhoea. 2020. PubMed 33295643.
5. Goodoory VC, et al. Pooled analysis, 82 randomised trials, 10,332 people with IBS. 2023. PubMed 37541528.
6. Allen SJ, et al. PLACIDE. Randomised trial, 2,981 UK inpatients aged 65 and over, 21 days; NIHR funded. 2013. PubMed 23932219.
7. EFSA NDA Panel. Scientific opinion on live yoghurt cultures and improved lactose digestion. EFSA Journal, 2010. doi:10.2903/j.efsa.2010.1763.
8. Zheng J, et al. Taxonomic study of the genus Lactobacillus, 261 species. 2020. PubMed 32293557.
9. Vesterlund S, et al. Laboratory storage study, one Lactobacillus rhamnosus strain, 14 months. 2012. PubMed 22663981.
10. Brousilova T, et al. Storage study of commercial multi-strain supplements, 12 months; no competing interests declared. 2026. PubMed 42824401.
11. Goodman C, et al. Pooled analysis, 42 randomised trials, 11,305 adults. 2021. PubMed 34385227.
12. Esmaeilinezhad Z, et al. Cochrane review, 38 randomised trials, 13,179 people. 2025. PubMed 40931979.
13. Zhao Y, et al. Cochrane review, 23 randomised trials, 6,950 people. 2022. PubMed 36001877.
14. Dimidi E, et al. Pooled analysis, 14 randomised trials, 1,182 adults with constipation. 2014. PubMed 25099542.
15. Liu RT, et al. Pooled analysis, 34 controlled trials. 2019. PubMed 31004628.
16. Makrgeorgou A, et al. Cochrane review, 39 randomised trials, 2,599 people with eczema. 2018. PubMed 30480774.
17. Borgeraas H, et al. Pooled analysis, 15 randomised trials, 957 overweight adults, 3 to 12 weeks. 2018. PubMed 29047207.
18. Shen NT, et al. Meta-regression, 19 studies, 6,261 hospital adults. 2017. PubMed 28192108.
19. NICE. Irritable bowel syndrome in adults (CG61), recommendation 1.2.1.6. Published 23 February 2008, last updated 4 April 2017. nice.org.uk/guidance/cg61.
20. Liao W, et al. Pooled analysis, 36 randomised trials, 9,312 adults. 2021. PubMed 33234881.
21. Million M, et al. Pooled analysis, 17 human trials and 65 animal studies. 2012. PubMed 22634320.
22. Vasant DH, et al. British Society of Gastroenterology guidelines on irritable bowel syndrome. 2021. PubMed 33903147.
23. Suez J, et al. Invasive sampling study, healthy volunteers, 11-strain product. 2018. PubMed 30193113.
24. Marco ML, et al. Expert consensus statement on fermented foods; several authors declare industry fees. 2021. PubMed 33398112.
25. Salminen MK, et al. Case series, 89 Lactobacillus bloodstream infections, Finland. 2004. PubMed 14679449.
26. Davidson SJ, et al. Cochrane review, 7 randomised trials, 1,647 pregnant women. 2021. PubMed 33870484.
27. Charbonneau D, et al. Placebo-controlled stool-tracking study, 117 adults, 8 weeks; authors employed by the product's maker. 2013. PubMed 23549409.
28. Zmora N, et al. Endoscopic sampling study, healthy volunteers, 11-strain product. 2018. PubMed 30193112.
29. NHS. Probiotics. nhs.uk/tests-and-treatments/probiotics. Page last reviewed 24 November 2022 (review overdue).
30. Maldonado-Gómez MX, et al. Human persistence study, one Bifidobacterium longum strain, 6 months. 2016. PubMed 27693307.

Spotted something out of date? Email support@wellgard.co.uk with the page address and the sentence.