Probiotics

Overview

Probiotics are live microorganisms that provide a health benefit when you take enough of them, and "enough" is measured in colony-forming units (CFU) of a specific, named strain. The human evidence is strain-specific and condition-specific: strong for a few strains in a few situations, thin for a healthy adult taking a capsule every day. The catch is that most labels never name a strain, so you cannot match the bottle to a trial.

What is Probiotics?

Key takeaways

• "Probiotic" has a formal definition: live microorganisms that confer a health benefit when taken in adequate amounts (Hill 2014). A yogurt with unnamed cultures and no count is a fermented food, not a probiotic.
• Effects belong to strains, not species. Across 228 trials, efficacy tracked the exact strain and the exact condition (McFarland 2018). "Lactobacillus acidophilus" on a label identifies the species the way "dog" identifies a breed.
• The strongest human evidence is for fewer loose stools during a course of antibiotics: Saccharomyces boulardii cut the risk from 18.7 to 8.5 percent across 21 trials and 4,780 people (Szajewska 2015), and in children any probiotic at 5 billion CFU a day or more cut it from 23 to 8 percent (Guo 2019).
• In healthy adults, seven randomized trials found no change in the diversity or makeup of the gut microbiome from a probiotic (Kristensen 2016), and a 2018 endoscopy study found the strains settled on some people's gut lining and bounced off others' (Zmora 2018).
• More CFU is not better by default. B. longum 35624 worked at 100 million CFU a day and failed at 10 billion (Whorwell 2006); BB-12 did the same thing at 1 billion as at 10 billion (Eskesen 2015). The one compelling dose-response in the field is for antibiotic-associated diarrhea (Ouwehand 2017).
• One third of 52 commercial probiotics tested below their labeled CFU before their expiration date, and only 58 percent were labeled with the correct organisms (Morovic 2016). A count guaranteed through expiration, with a strain code, is the label to buy.
• Taking an 11-strain probiotic right after antibiotics delayed the return of the native microbiome compared with taking nothing (Suez 2018). The "repopulate after antibiotics" story is less settled than the shelf tag says.

The word "probiotic" is a claim, not a category. Under the 2014 consensus from the International Scientific Association for Probiotics and Prebiotics (ISAPP), an organism earns the name only when a specific strain, at a specific dose, has shown a benefit in people (Hill 2014). A strain is a single isolate of a species with its own genome and its own trial record, identified by a code after the species name: GG, BB-12, HN019, 35624, CNCM I-745. Genus, species, strain. Canis, Canis familiaris, your dog. The trials were run on the dog, and a label that only says "Canis" is not telling you much.

Why would a few billion cells matter at all? By the numbers, they mostly should not. An adult body carries about 38 trillion bacteria against roughly 30 trillion human cells, and the bacteria together weigh about 200 grams (Sender 2016). A 20 billion CFU capsule adds one newcomer for every 1,900 residents, and the newcomers do not stay. What a probiotic can do is act on the way through: ferment, acidify, crowd a surface, signal the immune cells that line the gut. The next section covers that. The point here is scale. Nobody is replacing a microbiome with a capsule, and what the resident microbiome actually does is the better place to start if that is the question you came with.

The names on the label are mid-change, too. In 2020 the genus Lactobacillus, which had grown to 261 species, was split into 25 genera on the strength of whole-genome sequencing (Zheng 2020). Lactobacillus rhamnosus GG became Lacticaseibacillus rhamnosus GG; L. plantarum 299v became Lactiplantibacillus plantarum 299v. The umbrella word "lactobacilli" still covers all of them, most labels still print the old names, and this page uses whichever name the cited trial used.

Where the organisms come from is less romantic than the marketing. Commercial strains are grown in fermentation tanks, concentrated, freeze-dried inside a protective sugar matrix, and counted. The unit is the colony-forming unit, CFU: one live, culturable cell that grows into a visible colony on a plate. CFU counts only cells that are alive and willing to grow, which is the right unit for a live probiotic and the wrong one for a pasteurized product; our Akkermansia buyer's guide walks through CFU, AFU, and TFU side by side. Counts fall over a product's life, because freeze-dried cells die slowly in the presence of moisture and heat, which is why the number on a label needs a date attached to mean anything.

Four words get used as if they were interchangeable, and ISAPP has defined each one. A probiotic is the live organism (Hill 2014). A prebiotic is "a substrate that is selectively utilized by host microorganisms conferring a health benefit," in practice a fermentable fiber such as fructooligosaccharides (FOS), inulin, or galactooligosaccharides (Gibson 2017). A postbiotic is "a preparation of inanimate microorganisms and/or their components that confers a health benefit," meaning killed cells or pieces of cells (Salminen 2021). A synbiotic is a probiotic and a prebiotic in one product (Swanson 2020). The clearest postbiotic example is pasteurized Akkermansia muciniphila: in the one randomized human trial, 32 overweight adults finished three months on 10 billion live or pasteurized cells a day, and the pasteurized cells improved insulin sensitivity by about 29 percent against placebo while the live cells did not reach significance (Depommier 2019). Dead, in that case, outperformed alive. How Akkermansia fits alongside a conventional probiotic has its own post.

What happens after you swallow one is the part the capsule designs argue about. Stomach acid kills a share of most lactobacilli and bifidobacteria, the share depending on the strain, the dose, and whether there is food in the stomach; the yeast S. boulardii and Bacillus spores tolerate acid by their nature. Enough arrive. In the 2018 endoscopy study, healthy volunteers took an 11-strain capsule or placebo while researchers sampled the gut lining directly, and the strains were recovered alive along the intestine (Zmora 2018). Whether they stayed was another matter. Some people's gut lining let the strains settle for a while, others' resisted completely, the pattern was person-specific and region-specific, and stool samples could not tell the two groups apart. Stop taking a probiotic and it leaves. That one finding explains most of what follows: the trials that work are trials of a strain doing something on its way through, every day, for as long as you take it.

Your own residents, meanwhile, do things no capsule does. Gut bacteria make vitamin K2 in the menaquinone forms, a little of it, not reliably enough to skip the food sources. And whether your gut can turn the ellagitannins in pomegranates and walnuts into Urolithin A depends entirely on which bacteria you carry; roughly one adult in ten cannot do it at all. The microbiome is a metabolic organ. A probiotic is a visitor.

Strains with human evidence

The table is the page. Each row is a strain, what it was tested for, the daily dose the trial used, and the grade the evidence earns on the site's scale (Strong, Moderate, Preliminary, Insufficient). A product that names one of these strains at the listed dose can be matched to the trial. A product that names only the species cannot.

Two outcomes are missing from that table on purpose. Immune and mood claims have trials behind them, but this page did not review them; they are strain-specific questions that deserve their own pages, and a grade we have not earned is not a grade. The rows that are there share a pattern worth noticing before you read the mechanism: every outcome is a bowel outcome or an antibiotic outcome, the effect lasts as long as the dosing does, and the one metabolic row belongs to a dead organism.

How it works

Probiotics work on the way through, not by moving in. Four mechanisms have reasonable support, and strains share them unevenly.

Fermentation and acid come first. Lactobacilli and bifidobacteria ferment carbohydrates that reach the colon, and the products are lactic acid and short-chain fatty acids (acetate, propionate, and butyrate, the two- to four-carbon acids that colon cells burn as fuel). Acid lowers the local pH, which favors acid-tolerant residents over many pathogens, and the fatty acids feed the cells of the gut wall. The NIH Office of Dietary Supplements lists inhibiting pathogen growth as the mechanism most strains have in common, and this is the chemistry behind it. It also explains the first-week gas: more fermentation, more gas, until the resident community adjusts.

Competitive exclusion is the second. A strain present in the billions occupies binding sites on the mucus layer and consumes nutrients a pathogen would otherwise use. The 2018 endoscopy study is the clearest human picture of what "present" means: the strains were found alive along the gut lining, in some people far more than others, for as long as dosing continued (Zmora 2018). The resistance the resident community puts up is the same mechanism running in reverse, which is why one capsule does different things in different people.

Barrier signaling is the third, and the one with the most interesting recent data. The gut wall is one cell thick, sealed by tight junctions (protein seams between neighboring cells) and covered by mucus. Certain strains, and certain parts of dead cells, signal those cells to tighten the seams or thicken the mucus. The best-documented example is an Akkermansia outer-membrane protein, Amuc_1100, which binds an immune receptor on gut cells, stays intact at pasteurization temperatures, and improved the gut barrier in mice on its own (Plovier 2017); that is the working explanation for why the pasteurized cells outperformed live ones in the human pilot (Depommier 2019). For most lactobacilli and bifidobacteria the barrier effect is proposed rather than shown: cell culture and animal data, and no human trial that measured gut permeability directly and found a change. That trial is what would prove it.

Immune signaling is the fourth. The gut lining is the body's largest immune surface, and strains differ in which receptors they trigger on the cells beneath it; the NIH fact sheet notes that these effects are the ones most specific to a particular species or strain. The honest test of an immune mechanism is a hard outcome, and the largest randomized test of a probiotic on infection outcomes we could find, 2,653 ventilated intensive-care patients given L. rhamnosus GG, found no reduction in any ICU-acquired infection (Johnstone 2021). That is a sick population, not a healthy adult in cold season, and no trial of that size exists in healthy adults.

Antibiotics change the whole picture. An antibiotic clears a large part of the resident community, and loose stools during the course are the visible result. A probiotic taken through the course reduces that in trial after trial (the table below has the numbers), presumably by fermentation and exclusion, and in the case of S. boulardii, by being a yeast the antibiotic cannot kill. After the course, the question changes: does the probiotic help the residents come back? In 2018 the Weizmann group gave healthy volunteers a course of antibiotics and then either an 11-strain probiotic, a transplant of their own pre-antibiotic stool, or nothing. The probiotic group's native community came back slowest and least completely; the stool-transplant group recovered within days; and in culture, substances secreted by the lactobacilli inhibited the native bacteria (Suez 2018). One trial with invasive sampling, and the most-cited reason the "repopulate after antibiotics" claim is on hold.

The mechanisms predict the outcome list. A strain that ferments and crowds can change stool consistency and transit while you take it, can blunt the collateral damage of an antibiotic while you take it, and cannot be expected to remodel a healthy adult's microbiome. That is what the evidence shows.

Evidence at a glance

What the evidence doesn't support yet

"Restores your gut flora after antibiotics." The benefit during a course is real and well measured: fewer people get loose stools. Restoration is a different claim, and the one trial that looked directly at the gut lining found the probiotic group's native community came back slower and less completely than the group that took nothing (Suez 2018). A bottle that says "replenish" after antibiotics is leaning on the first finding and hoping you do not know about the second. The honest version is "take it during the course, for the stools," and stop there.

"Supports gut health" for a healthy adult. Seven randomized trials in healthy adults measured the microbiome with sequencing before and after a probiotic, and none found a change in diversity, richness, or evenness versus placebo (Kristensen 2016). The endoscopy study found that whether a strain even settles depends on the person, not the product (Zmora 2018). No expert body recommends probiotics for healthy people (NIH ODS). If nothing is wrong, there is no trial showing a capsule makes it more right.

"More billions, more benefit." B. longum 35624 worked at 100 million CFU and did nothing at 10 billion (Whorwell 2006). BB-12 at 1 billion matched BB-12 at 10 billion (Eskesen 2015). A review of the field's dose-response data found a compelling dose relationship for antibiotic-associated diarrhea, a weaker one for blood pressure, and none for the other endpoints it examined (Ouwehand 2017). A 100 billion CFU headline is a marketing number until someone runs the 100 billion trial.

"Akkermansia for weight loss." The one human trial reported a 2.27 kg difference against placebo that did not reach significance (p = 0.09) in 32 people (Depommier 2019). Insulin sensitivity moved; weight did not, statistically. Our Akkermansia guide grades that trial outcome by outcome, and we sell a product with Akkermansia in it, which is exactly why this paragraph is here.

Clinical evidence

• Expert consensus document. The International Scientific Association for Probiotics and Prebiotics consensus statement on the scope and appropriate use of the term probiotic — A 2013 ISAPP expert panel reaffirmed the FAO/WHO definition, live microorganisms that, when administered in adequate amounts, confer a health benefit on the host, and clarified that the term applies to defined strains at defined doses with evidence, not to undefined fermented foods or to dead cells. This is the definition every row on this page is graded against. Limitation: a consensus document, not a trial; it sets the standard rather than testing anything.
• Strain-Specificity and Disease-Specificity of Probiotic Efficacy: A Systematic Review and Meta-Analysis — 228 randomized trials from 1970 to 2017 of probiotics with an identifiable strain, grouped by strain and by indication. Efficacy was both strain-specific and disease-specific: 7 of 10 strain groups showed significant efficacy across four preventive indications and 11 of 17 across five treatment indications, with strains inside the same species diverging, and the same strain succeeding in one condition and failing in another. Limitation: a review of reviews across heterogeneous trials, so it proves the principle rather than any one strain's effect size.
• Systematic review with meta-analysis: Saccharomyces boulardii in the prevention of antibiotic-associated diarrhoea — 21 randomized trials with 4,780 children and adults taking antibiotics, S. boulardii versus placebo or no treatment for the length of the course. Antibiotic-associated diarrhea fell from 18.7 to 8.5 percent (RR 0.47, 95% CI 0.38 to 0.57, number needed to treat 10); in adults from 17.4 to 8.2 percent across 15 trials and in children from 20.9 to 8.8 percent across 6. C. difficile-associated diarrhea fell significantly only in children. Limitation: doses and antibiotic regimens varied across trials and were not pooled by dose.
• Systematic review and meta-analysis of Saccharomyces boulardii in adult patients — 27 randomized trials with 5,029 adult patients across 31 treatment arms, 1976 to 2009. S. boulardii was significantly effective and safe in 84 percent of arms; for preventing antibiotic-associated diarrhea the pooled relative risk was 0.47 (95% CI 0.35 to 0.63). The review also covers safety and contraindications for the yeast. Limitation: single-author review that pools trials of different indications; the strongest results are for diarrhea prevention and the rest are labeled promising rather than proven.
• Systematic review with meta-analysis: Lactobacillus rhamnosus GG in the prevention of antibiotic-associated diarrhoea in children and adults — 12 randomized trials with 1,499 participants, L. rhamnosus GG versus placebo or no treatment during antibiotics. Antibiotic-associated diarrhea fell from 22.4 to 12.3 percent overall (RR 0.49, 95% CI 0.29 to 0.83, low quality of evidence), but the effect was significant only in children (5 trials, n=445, RR 0.48, moderate quality); in adults the difference was not significant (6 trials, n=863, RR 0.48, 95% CI 0.20 to 1.15) except in a subgroup on H. pylori eradication therapy. Limitation: low to moderate certainty and wide confidence intervals in adults.
• Probiotics for the prevention of pediatric antibiotic-associated diarrhea — 33 randomized trials with 6,352 children receiving antibiotics, any probiotic versus placebo, active control, or nothing, followed 5 days to 12 weeks. Diarrhea occurred in 8 percent of the probiotic group versus 19 percent of controls (RR 0.45, 95% CI 0.36 to 0.56, number needed to treat 9, moderate certainty); doses of 5 billion CFU a day or more performed better (8 versus 23 percent, RR 0.37) than lower doses (RR 0.68), and diarrhea duration was shorter by about a day. No serious adverse events were attributable to probiotics across 24 trials. Limitation: high risk of bias in 20 of 33 studies and heterogeneity from the mix of organisms used.
• Probiotics for the prevention of Clostridium difficile-associated diarrhea in adults and children — 31 randomized trials with 8,672 adults and children on antibiotics, any probiotic versus placebo or no treatment. C. difficile-associated diarrhea occurred in 1.5 percent of the probiotic group versus 4.0 percent of controls (RR 0.40, 95% CI 0.30 to 0.52, moderate certainty, number needed to treat 42); the benefit was confined to trials where baseline risk exceeded 5 percent (RR 0.30, number needed to treat 12), and detection of C. difficile in stool was not reduced. Adverse events were lower in the probiotic group. Limitation: 27 of 39 studies at high or unclear risk of bias, and missing outcome data in most trials.
• Efficacy of an encapsulated probiotic Bifidobacterium infantis 35624 in women with irritable bowel syndrome — 362 primary-care women with IBS of any bowel-habit subtype, 4 weeks, freeze-dried encapsulated B. infantis 35624 at 1 million, 100 million, or 10 billion CFU versus placebo. Only the 100 million CFU dose beat placebo on the primary outcome of abdominal pain and on the composite score, bloating, bowel dysfunction, and global assessment, where improvement exceeded placebo by more than 20 percent; the 10 billion dose had formulation problems and did no better than placebo. No significant adverse events. Limitation: one 4-week trial in women, and the dose finding has not been replicated at scale.
• Clinical trial: Lactobacillus plantarum 299v (DSM 9843) improves symptoms of irritable bowel syndrome — 214 adults meeting Rome III criteria for IBS, 4 weeks, one capsule of L. plantarum 299v daily versus placebo, double-blind. Pain severity (0.68 versus 0.92) and daily pain frequency (1.01 versus 1.71) were lower on the probiotic, with similar results for bloating, and 78.1 percent rated the effect good or excellent versus 8.1 percent on placebo. Limitation: one 4-week trial, and the CFU per capsule is not stated in the abstract.
• Systematic review with meta-analysis: the efficacy of prebiotics, probiotics, synbiotics and antibiotics in irritable bowel syndrome — 53 randomized trials of probiotics with 5,545 adults with IBS, plus sparse prebiotic and synbiotic data, searched to July 2017. Particular combinations and specific species or strains appeared to improve global symptoms and abdominal pain, but the authors could not draw definitive conclusions about which; adverse events were no more common than placebo. Limitation: heterogeneity across products, doses, and outcome definitions is the finding, which is why this page grades strains individually rather than the class.
• Dose-response effect of Bifidobacterium lactis HN019 on whole gut transit time and functional gastrointestinal symptoms in adults — 100 adults (mean age 44, 64 percent women) with functional gut symptoms, 14 days, B. lactis HN019 at 17.2 billion or 1.8 billion CFU a day versus placebo. Whole-gut transit time measured by X-ray fell from 49 to 21 hours on the high dose and from 60 to 41 hours on the low dose, with no change on placebo; eight of nine symptoms fell in frequency on the high dose and seven on the low. No adverse events. Limitation: 14 days, industry-sponsored, and the symptom measure was self-reported.
• Effect of the probiotic strain Bifidobacterium animalis subsp. lactis, BB-12, on defecation frequency in healthy subjects with low defecation frequency and abdominal discomfort — 1,248 healthy adults with low defecation frequency and abdominal discomfort, 4 weeks after a 2-week run-in, BB-12 at 1 billion or 10 billion CFU a day in a capsule versus placebo. Average defecation frequency was higher than placebo at every week (P=0.0065) and the odds of responding by at least one extra day a week were 1.55 (95% CI 1.22 to 1.96); the two doses performed the same, suggesting a ceiling at 1 billion, and global relief of discomfort did not differ from placebo. Limitation: manufacturer-run, a narrow primary outcome, and the co-primary well-being outcome was null.
• Probiotic bacteria Lactobacillus acidophilus NCFM and Bifidobacterium lactis Bi-07 versus placebo for the symptoms of bloating in patients with functional bowel disorders: a double-blind study — 60 adults with non-constipation functional bowel disorders (72 percent women, mean age 37), 8 weeks, L. acidophilus NCFM plus B. lactis Bi-07 at a combined 200 billion CFU a day in two doses versus placebo. Bloating severity was lower on the probiotic at 4 weeks (4.10 versus 6.17, P=0.009) and the change held at 8 weeks (P<0.01), with similar results in the IBS subgroup of 33. Limitation: small, single-center, with a dose far above what most products carry.
• Supplementation with Akkermansia muciniphila in overweight and obese human volunteers: a proof-of-concept exploratory study — 40 overweight or obese insulin-resistant adults enrolled and 32 completed, 3 months, 10 billion live or pasteurized A. muciniphila cells a day versus placebo, double-blind, single center. Both forms were safe and well tolerated; the pasteurized cells improved insulin sensitivity by 28.6 percent, cut insulin by 34.1 percent and total cholesterol by 8.7 percent against placebo, while body weight (minus 2.27 kg, P=0.091) and fat mass did not reach significance and the live cells did not produce the same significant changes. Limitation: a 32-person pilot in a metabolically unwell population, which says nothing about healthy adults.
• Personalized Gut Mucosal Colonization Resistance to Empiric Probiotics Is Associated with Unique Host and Microbiome Features — Healthy human volunteers given an 11-strain probiotic combination or placebo, with the gut lining sampled directly by endoscopy before and during dosing, alongside parallel mouse work. The strains stayed viable through the gut, but mucosal colonization was person-, region-, and strain-specific, predictable from baseline host and microbiome features and invisible in stool samples; the effect on the mucosal community and gut gene expression was transient and individualized. Limitation: a small invasive study of colonization, not of symptoms, and PubMed indexes it as a clinical trial without specifying randomization, so this page labels it by its design rather than as an RCT.
• Post-Antibiotic Gut Mucosal Microbiome Reconstitution Is Impaired by Probiotics and Improved by Autologous FMT — Healthy human volunteers and mice given antibiotics, then an 11-strain probiotic, a transplant of their own pre-antibiotic stool, or spontaneous recovery, with endoscopic sampling of the gut lining. Antibiotics increased probiotic colonization of the human mucosa, but the probiotic group showed a markedly delayed and persistently incomplete return of the native stool and mucosal microbiome and host gene expression, while the autologous transplant restored both within days; in culture, Lactobacillus-secreted factors inhibited native bacteria. Limitation: a small invasive study of microbiome composition, not of clinical outcomes, and the clinical cost of slower recovery is unknown.
• Alterations in fecal microbiota composition by probiotic supplementation in healthy adults: a systematic review of randomized controlled trials — Systematic review of 7 randomized placebo-controlled trials in healthy adults that measured overall fecal microbiota composition by sequencing or phylogenetic microarray after probiotic supplementation, searched to August 2015, study quality medium to high. No trial found an effect on alpha-diversity, richness, or evenness compared with placebo, and only one found a shift in overall community structure (beta-diversity). Limitation: seven trials of varied strains and durations, and stool composition is not the same as mucosal activity.
• Effect of Probiotics on Incident Ventilator-Associated Pneumonia in Critically Ill Patients: A Randomized Clinical Trial — 2,653 mechanically ventilated adults in 44 intensive care units in Canada, the United States, and Saudi Arabia, enteral L. rhamnosus GG at 10 billion CFU twice daily versus placebo for a median of 9 days. Ventilator-associated pneumonia occurred in 21.9 percent on the probiotic versus 21.3 percent on placebo (hazard ratio 1.03), with no difference in any of 20 secondary outcomes including other infections, diarrhea, C. difficile infection, or mortality; L. rhamnosus was found in a sterile site or as the predominant organism in a non-sterile site in 15 probiotic patients (1.1 percent) versus 1 control (odds ratio 14.02). Limitation: a critically ill population, so the null result and the harm signal do not transfer to outpatients.
• Probiotic prophylaxis in predicted severe acute pancreatitis: a randomised, double-blind, placebo-controlled trial — 298 patients with predicted severe acute pancreatitis in the Netherlands, a multispecies probiotic preparation or placebo given by tube twice daily for 28 days, starting within 72 hours of symptom onset. Infectious complications were no different (30 versus 28 percent), but 16 percent of the probiotic group died versus 6 percent on placebo (relative risk 2.53, 95% CI 1.22 to 5.25), and nine probiotic patients developed bowel ischemia, eight fatally, versus none on placebo. Limitation: one specific multispecies product in one severe illness; the result is a reason for caution in critically ill patients, not a finding about healthy users.
• Safety assessment of probiotics for human use — A multi-author safety review covering the probiotic field: risk is tied to the vulnerability of the consumer, the dose, duration, and route, and the identity of the organism, with infectivity, in-situ toxin production, transferable antibiotic-resistance genes, and immune effects in infants named as the categories to assess. The authors note the scarcity of trials designed to measure safety against a long history of safe food use, and recommend genome sequencing of candidate strains to exclude genes of concern. Limitation: a review, not a trial, and now 15 years old; the ICU data that followed is in the rows above.

Dosage & forms

• Standard capsule (freeze-dried powder, vegetarian shell):
• Delayed-release or enteric-coated capsule:
• Refrigerated capsule (compared with shelf-stable):
• Powder or sachet:
• Spore-forming Bacillus (B. coagulans, B. subtilis):
• Synbiotic (probiotic plus prebiotic fiber):

Safety & interactions

How much to take

The research dose is a range that spans five orders of magnitude, and the right number depends on which strain you are holding. Bifidobacterium longum 35624 worked at 100 million CFU a day and failed at 10 billion (Whorwell 2006). BB-12: 1 billion, and nothing gained at 10 (Eskesen 2015). HN019 shortened transit at 1.8 billion CFU and shortened it further at 17.2 billion, the one clean dose-response in the strains table (Waller 2011). The pediatric antibiotic trials worked best at 5 billion CFU a day or more (Guo 2019), and the intensive-care trial gave L. rhamnosus GG at 20 billion a day (Johnstone 2021). Most products carry 10 to 30 billion CFU per serving, which sits inside that range for the lactobacilli and bifidobacteria and well above it for 35624. Match the strain to its trial and take the trial's dose; if the label names no strain, the dose is a guess either way.

Timing matters less than consistency. The trials dosed once or twice a day, with or without food, and none compared the two. A capsule with a meal is a reasonable default because food buffers stomach acid, and the yeast and spore-forming products do not care. Expect the first measurable change inside two weeks for transit (Waller 2011) and four weeks for bowel symptoms (Whorwell 2006, Ducrotte 2012, Eskesen 2015); if nothing has changed by week six, the strain is not doing what you bought it for, and a different one might. For slow transit specifically, magnesium citrate's effect on the bowel is the other lever, and the cheaper one. Gas in the first week is fermentation, not failure.

With antibiotics, the trials that worked started the probiotic with the first antibiotic dose and continued through the course (Szajewska 2015; Guo 2019). No trial has tested spacing the two doses apart. Separating them by two or three hours is a reasonable, unproven precaution that keeps the antibiotic and the bacteria from meeting at full strength in the stomach; the yeast Saccharomyces boulardii sidesteps the question because antibiotics do not kill yeast. Whether to keep taking a probiotic after the course ends is the open question the 2018 reconstitution trial raised (Suez 2018), and the honest answer is that the benefit evidence stops when the antibiotic does.

Diminishing returns arrive early. A review of the dose-response data found a compelling relationship for antibiotic-associated diarrhea, a weaker one for blood pressure, and none for the other endpoints it examined (Ouwehand 2017). No regulator has set an upper limit for CFU. The only regulator-reviewed ceiling in the category is the European Food Safety Authority's 34 billion cells a day for pasteurized Akkermansia, set in 2022 and described in our Akkermansia guide. Past the studied dose, more CFU buys more gas and a higher price per day, and that is all the evidence can promise.

Safety profile

For healthy people, the safety record is long and mostly boring. The organisms in most products come from species with decades of use in food, and the 2010 safety assessment by the probiotic field's own experts concluded that the main risks, infection in a vulnerable host and transfer of antibiotic-resistance genes, are tied to who takes the product and how, not to the healthy adult swallowing a capsule (Sanders 2010). The same review noted the scarcity of trials designed to measure safety, which is still true. In the Cochrane review of children on antibiotics, 24 trials with 4,415 participants reported adverse events, none serious and attributable to the probiotic, with 4 percent of probiotic users and 6 percent of controls reporting any event (Guo 2019). In the adult and pediatric review of 32 trials with 8,305 participants, adverse events were lower in the probiotic group than in controls (Goldenberg 2017).

Side effects, by frequency. Common: gas, bloating, and a change in stool consistency in the first one to two weeks, listed in both Cochrane reviews. Uncommon: rash, nausea, constipation, and taste disturbance, also from the Cochrane adverse-event tables. Rare: a probiotic organism in the bloodstream. In the 2,653-patient intensive-care trial, L. rhamnosus turned up in a sterile site or as the dominant organism in a non-sterile one in 1.1 percent of probiotic patients against 0.1 percent of controls (Johnstone 2021), and whole-genome sequencing in a Boston ICU showed the bacteria in patients' blood were the same clone as the capsule (Yelin 2019). Those were ventilated, catheterized patients fed through tubes. The same event in a healthy adult is a case report, not a rate.

Drug interactions

The S. boulardii fungemia literature deserves its own line. A review of 92 invasive Saccharomyces infections found the probiotic subtype accounted for about half of the bloodstream cases, with an intravascular catheter and concurrent antibiotics as the most frequent predisposing factors (Enache-Angoulvant 2005). For a healthy adult taking the yeast through a course of amoxicillin, that is not the population. For a hospitalized relative with a line in, it is.

Who should ask first

Talk to a clinician before starting any live probiotic if you are severely immunocompromised, have a central venous catheter, are in intensive care or recovering from major abdominal surgery, have a damaged gut wall from any cause, or are caring for a preterm infant (the FDA raised concerns about probiotic products in hospitalized preterm infants in 2023, per NIH ODS). Pregnant or nursing people have thin data for most strains and no data for pasteurized Akkermansia, which the European authorization excludes for that reason. Anyone taking an immunosuppressant or an antifungal is in the table above. For at-risk patients, the probiotic experts' own advice is that a pharmacist should see the certificate of analysis before the product is dispensed (Sanders 2016), which is a standard worth applying at home.

There is no established upper limit and no toxicity syndrome. Too much looks like gas and bloating. The one sign that matters is a fever or chills in a vulnerable person while on a live product, and that is a call, not a wait.

If you are a healthy adult with a working immune system and no catheter, the risk is the gas. Everyone else asks first.

Buying guide

CFU honesty: the count through expiration, not at manufacture

The first thing to read on a probiotic label is the date the count applies to. FDA requires the weight of the organisms on a Supplement Facts panel; printing a CFU count is voluntary under its 2018 guidance, and the weight includes dead cells (NIH ODS). ISAPP advises manufacturers to print the CFU, ideally per strain, guaranteed through the expiration date, and advises buyers to avoid products that count "at time of manufacture," because cells die on the shelf. The gap is not theoretical. When 52 commercial probiotics were plated before their expiration dates, 17 of them, one in three, were below their labeled CFU (Morovic 2016). Look for "guaranteed through expiration" or "at expiry." A label with no date qualifier is asking you to trust an overage you cannot see. Our own lab page shows what the proof looks like.

The strain designation, not just the species

A label that reads "Lactobacillus acidophilus, 10 billion CFU" tells you the species and the count and leaves out the only thing that connects a product to a trial. Efficacy is strain-specific: across 228 trials, the strains that worked and the strains that did not sat inside the same species (McFarland 2018). The code after the species name is the designation: GG, NCFM, BB-12, HN019, 35624, 299v, CNCM I-745. Codes are checkable. In the same 52-product test, only 30 products, 58 percent, were labeled with the organisms they actually contained (Morovic 2016), and a strain code is what lets a lab confirm identity by genotyping rather than by guessing from a plate. If the label names no strain, treat the product as a generic culture and price it like one.

A dose inside that strain's studied range

Once you have a strain, the dose question answers itself from the table above: 100 million for 35624, 1 billion for BB-12, a few billion to 20 billion for the lactobacilli and bifidobacteria in the antibiotic trials, 10 billion cells for pasteurized Akkermansia. Fifty billion and 100 billion are counts no trial in the table used. They are not dangerous. They are untested, and they cost more.

Delivery that matches the organism

A plain capsule delivered 35624, 299v, and an 11-strain mix in the trials above, so the shell is not the deciding factor for most lactobacilli and bifidobacteria at studied doses. A delayed-release capsule makes sense for an acid-sensitive strain or a low count, and no trial has shown a clinical outcome improved by the coating itself. Refrigerate only when the label says so; a refrigerated product whose count is not guaranteed through expiration is a cold guess. Spore-forming Bacillus and the yeast need neither coating nor cold. A pasteurized product that asks for refrigeration was labeled by someone who did not read the formulation; the cells are not alive.

Testing you can see

Our probiotics are made in our own cGMP facility in Mt. Angel, Oregon. Every batch is tested in our in-house lab for identity, potency by plate count, and purity, and the certificate of analysis for each lot is published on our lab-testing page, including potency through shelf life. That is the proof stack. It is also the one to demand from any brand, including us: a per-lot certificate with the plate count at release, the stability data behind the expiration claim, and a strain identity method. "Third-party tested" with no document attached is a badge. How to read a certificate of analysis and what to ask a brand for covers the general method, and what we found testing 44 supplements bought on Amazon is the reason the method exists.

Price per day for the real dose

Divide the bottle price by the number of servings, then ask whether the serving is a studied strain at a studied dose. A $12 bottle of 30 unnamed-species capsules is not cheaper than a $40 bottle of named strains if the first one is at 30 percent of label by month six; it is more expensive per live, identifiable cell. Our own two products land at $0.33 to $1.40 a day; the named-strain competitors in our Akkermansia guide ran $0.83 to $2.33 a day on the day we checked them. Above that range you are paying for count, a fridge, or a brand.

Red flags

A proprietary blend that gives one total and hides the per-strain counts. "Clinically proven" with no strain and no study named. A count printed as "cells" with no unit, or a TFU count on a product sold as alive. "At time of manufacture." A 100 billion headline with no trial behind it. A medical condition printed on the bottle as something the product treats, which is a drug claim a supplement cannot make.

Our two probiotics against this checklist

3-in-1 Probiotic Fusion prints 13 cultures with a CFU count for each one, no proprietary blend, totaling about 20 billion CFU per capsule; 100 mg of FOS prebiotic; 5 billion TFU of pasteurized Akkermansia muciniphila, labeled as a postbiotic in the correct unit; and 25 mg of champignon mushroom extract. Two cultures carry strain codes, L. rhamnosus LRa-05 at 5 billion and L. reuteri LR08 at 4 billion, and the other 11 are listed at species level. On the second criterion that is a partial pass, not a full one, and we would rather say so here than have you find it on the label. The 100 mg of FOS is a token dose next to a bowl of oats. The Akkermansia is half the human-trial dose. It is shelf-stable with no refrigeration, potency is tested through shelf life, and the lot certificate is on the lab page. Price: $41.99 for 30 capsules, $1.40 a day, or $71.99 for 60, $1.20 a day, one capsule daily. It supports digestive balance and a healthy gut barrier as part of a daily routine.* Our head-to-head with Seed DS-01 covers how it compares with a higher-count synbiotic, and if Akkermansia is the reason you are here, the Akkermansia guide ranks it against four competitors.

Good Flora is the simpler, older formula: Bifidobacterium longum at 900 million, Bifidobacterium bifidum at 450 million, and Lactobacillus acidophilus at 750 million per capsule, 2.1 billion in total, with the same 100 mg of FOS and 25 mg of champignon extract, in a capsule that contains whey. It fails the second criterion outright: no strain codes. It also prints its counts as "Million" without the CFU unit, which the first criterion asks for. Those two misses are why it is not in the product card on this page. What it is, honestly, is an inexpensive daily culture of three common species: $19.99 for 60 capsules at one to three a day, $0.33 to $0.67 a day, made and tested in the same facility. It supports healthy levels of good bacteria from mouth to colon as part of everyday digestion.* The story behind its 2 billion count is in its own post, and if your actual complaint is meals sitting heavily, a digestive enzyme is a different tool for a different job.

Next reads

• What the gut microbiome does and how diet moves it: the residents a probiotic visits, and why fiber matters more than any capsule for most people.
• Seed DS-01 versus our 3-in-1, strain list against strain list: the comparison question answered with both labels on the table.
• Best probiotics with Akkermansia, ranked and priced: five products against the one human trial, with the count units decoded.

The bottom line

If you are starting a course of antibiotics, a named strain with trial evidence, Saccharomyces boulardii or L. rhamnosus GG at 5 billion CFU a day or more, taken from the first antibiotic dose through the end of the course, is the one use with strong evidence behind it. If your complaint is slow transit or bloating, pick a strain from the table at its studied dose and give it four weeks. If you are healthy and nothing is wrong, no trial shows a capsule will make you healthier, and the money is better spent on the fiber that feeds the residents you already have. If you are immunocompromised or have a line in, ask first.

Frequently Asked Questions

What is the difference between a probiotic, a prebiotic, and a postbiotic?

A probiotic is a live microorganism that provides a health benefit at an adequate dose; a prebiotic is a substrate, usually a fermentable fiber such as FOS or inulin, that your resident microbes use for a benefit; a postbiotic is a preparation of killed microorganisms or their parts that provides a benefit. All three definitions come from ISAPP consensus statements (Hill 2014, Gibson 2017, Salminen 2021). A synbiotic combines a probiotic and a prebiotic in one product. The clearest postbiotic example is pasteurized Akkermansia muciniphila, which outperformed the live organism in its one human trial. The live ones visit; the fiber feeds what already lives there; the dead ones work through their parts.

Which probiotic strains have human evidence, and for what?

Saccharomyces boulardii CNCM I-745 and Lactobacillus rhamnosus GG for preventing antibiotic-associated diarrhea, with the yeast's evidence stronger (21 trials, 4,780 people) and LGG's convincing in children but not in adults. Bifidobacterium longum 35624 at 100 million CFU and L. plantarum 299v for IBS pain and bloating, each on one good trial. B. lactis HN019 for whole-gut transit time and B. lactis BB-12 for stool frequency in adults with sluggish bowels. Pasteurized Akkermansia muciniphila for insulin sensitivity in one 32-person pilot. Every one of those is a specific strain at a specific dose, and none of the evidence transfers to a label that names only the species.

What does CFU mean, and does a higher count matter?

CFU is colony-forming units: the number of live cells in a dose that can grow into a colony on a culture plate, which makes it a count of viable, culturable organisms and not a count of total cells. A higher count is not better by default. B. longum 35624 worked at 100 million CFU and failed at 10 billion (Whorwell 2006), BB-12 did the same at 1 billion as at 10 billion (Eskesen 2015), and the only outcome with a clear dose-response is antibiotic-associated diarrhea, where 5 billion CFU a day or more beat lower doses in children (Guo 2019). The number that matters more than the size of the count is the date it is guaranteed through.

Do probiotics survive stomach acid, how do capsule designs differ, and should a probiotic be refrigerated?

Enough survive. In a 2018 study that sampled the gut lining by endoscopy, an 11-strain probiotic in a plain capsule was recovered alive along the intestine (Zmora 2018), and 35624 and 299v were tested in plain capsules. Acid kills a share of most lactobacilli and bifidobacteria, which formulators cover with an overage; Saccharomyces boulardii and spore-forming Bacillus tolerate acid on their own. A delayed-release capsule opens past the stomach and suits an acid-sensitive strain, though no trial shows the coating itself improved an outcome. Refrigerate only if the label says so: cold slows cell death on the shelf and does nothing for survival in the gut, and a through-expiration count at room temperature is the proof a shelf-stable product worked.

How long does it take for a probiotic to work, and what should you notice?

Two to four weeks for the outcomes that have been measured. Whole-gut transit time shortened within 14 days on B. lactis HN019 (Waller 2011); bowel symptoms such as pain, bloating, and stool frequency changed over 4 weeks in the 35624, 299v, and BB-12 trials. During a course of antibiotics the effect is immediate and invisible: fewer people develop loose stools, which you only notice by not getting them. What you should notice in week one is some extra gas, which is fermentation settling in. What you should not expect is a change in energy, mood, skin, or weight, because no strain in the evidence table was tested for those. If nothing has changed by week six, switch strains or stop.

Can you take probiotics with antibiotics, and how far apart?

Yes, and the trials that showed benefit started the probiotic with the first antibiotic dose and continued through the course (Szajewska 2015; Guo 2019). No trial has tested spacing. Separating the two by two or three hours is a reasonable, unproven precaution so the antibiotic and the bacteria do not meet at full strength in the stomach; Saccharomyces boulardii avoids the problem entirely because it is a yeast and antibiotics do not kill yeast. The open question is what to do after the course ends: in a 2018 trial, an 11-strain probiotic taken after antibiotics delayed the return of the native microbiome compared with taking nothing (Suez 2018), so the evidence for continuing stops when the antibiotic does.

Is a multi-strain probiotic better than a single strain?

Not by default. The best-documented outcomes in the evidence table belong to single strains: S. boulardii, L. rhamnosus GG, B. longum 35624, L. plantarum 299v, B. lactis HN019 and BB-12. Multi-strain products have strong pooled evidence for antibiotic-associated diarrhea as a class, but the 2012 JAMA meta-analysis that pooled 63 trials noted the strains were poorly documented (Hempel 2012), so the class result does not transfer to any particular bottle. A mixture is a sensible bet when it names each strain with a count and at least one of those strains has its own trial; a mixture that lists twelve species and one total count is twelve guesses in one capsule.

Does the strain designation on the label matter, and what should a trustworthy probiotic label show?

It is the whole game. Across 228 trials, efficacy tracked the exact strain, not the species (McFarland 2018), so "Lactobacillus rhamnosus GG" can be matched to its trials and "Lactobacillus rhamnosus" cannot. A trustworthy label shows the genus, species, and strain code for each organism; a CFU count per strain, or at least per product, guaranteed through the expiration date rather than at manufacture; the storage condition; and a lot number that leads to a certificate of analysis with a plate count. Red flags are a proprietary blend with one total, "clinically proven" with no study named, a count with no unit, and any disease name on the bottle.

Who should ask a doctor before taking a probiotic?

Anyone severely immunocompromised, anyone with a central venous catheter or feeding tube, anyone in intensive care or recovering from major abdominal surgery, anyone with a damaged gut wall, and the parents of a preterm infant. In those groups the capsule's own organism has been traced into the bloodstream by genome sequencing (Yelin 2019), the largest ICU trial found no benefit and a higher rate of the organism in sterile sites (Johnstone 2021), and a trial in severe acute pancreatitis found higher mortality (Besselink 2008). Anyone on an immunosuppressant or an antifungal should ask too, and pregnant or nursing people have thin data for most strains. A healthy adult with a working immune system does not need permission.

Do I need a probiotic at all?

Probably not, if nothing is wrong. No expert body recommends probiotics for healthy people (NIH ODS), seven randomized trials in healthy adults found no change in gut microbiome composition from a probiotic (Kristensen 2016), and the endoscopy study found that whether a strain even settles depends on the person rather than the product (Zmora 2018). The cases with real evidence are specific: a course of antibiotics, slow transit, or IBS-type pain and bloating, each matched to a named strain at its studied dose. If you eat fiber from plants most days, you are already feeding the residents, which is the intervention with the broadest effect. Buy a probiotic for one of the specific reasons, not as insurance.

Related Mt. Angel Vitamins Products

• Recommended for Probiotics

  3-in-1 Probiotic Fusion – Pre, Pro & Postbiotic Gut Support | 30 or 60 Capsules

  $41.99

  Shop Now