SIBO: what the breath test shows, and why the order of steps matters more than the remedy

In brief

SIBO is that rare diagnosis which is both missed and made too readily. The same breath test is behind each: it is available, it is cheap, and it gets read as though it saw straight through the gut.

So — what it actually measures, where overgrowth comes from, and why in this subject the order of steps matters more than the choice of remedy.

What overgrowth is

The colon is densely populated with bacteria, as it should be. The small intestine is nearly sterile: its contents move fast, bile and acid suppress growth, and between meals a clearing wave of contractions sweeps whatever is there downstream.

SIBO is bacteria settling where they do not belong. They start intercepting food ahead of their host: fermenting carbohydrate into gas, hence the bloating; damaging the villi, hence poor absorption; splitting bile acids, hence trouble with fats.

The symptoms are unhelpfully unspecific — bloating, gas, unstable stool, heaviness after eating. Exactly what a dozen other conditions produce. Which is why the conversation is meaningless without a test.

How it is confirmed

The direct way is to take fluid from the small intestine and culture it. More accurate, but it requires endoscopy, and in ordinary practice it is rarely done.

The practical way is the hydrogen breath test. The logic is simple: you drink a sugar solution, and if bacteria live in the small intestine they ferment it before it reaches the colon. The products are absorbed and breathed out — gas that should not be in the breath is the trace of bacteria in the wrong place.

The criteria were standardised by the 2017 North American Consensus (PMID 28323273^[1]) — 17 specialists, 28 statements, agreement reached on 26. From it:

And a caveat the consensus makes itself: the test is not suitable for assessing how fast food moves through the gut. It is often used for exactly that — an error recorded in the primary source.

A second caveat from the same place: SIBO must be excluded before testing for carbohydrate intolerance, or you get a false positive — the gas comes from bacteria and gets blamed on lactose or fructose.

Methane is not bacteria at all

The most underrated detail in the subject. Methane is produced by archaea, not bacteria — a separate form of life, neither bacterium nor fungus. So the methane pattern differs from the hydrogen one not in shade but in substance.

It differs clinically too: methane goes with constipation more often, while the hydrogen pattern goes with bloating and looser stool. And it responds differently, because antibacterial agents do not act on archaea as they do on bacteria.

Hence the practical point: if the test is methane-positive and the treatment is aimed at ordinary SIBO, the result will predictably disappoint.

Where it comes from

The small intestine is defended three ways: by acid, by movement, and by the valve at the entrance to the colon. Overgrowth appears where one of the three has failed.

Acid suppression

The commonest and most discussed cause. A 2025 meta-analysis (PMID 40649078^[2]) pooled 29 studies: 3682 people on proton pump inhibitors against 2907 controls.

SIBO was found in 36.8% of those taking them against 19.9% of controls — an odds ratio of 2.14 [1.45–3.18]. Roughly double the risk.

And the key finding is the dependence on duration: each additional month of therapy added 4.3 percentage points to the probability of SIBO (p = 0.0024). Heterogeneity between studies was high, so the figure reads as a direction rather than a standard.

The conclusion is not «acid suppressants are harmful». It is that they are prescribed for two weeks and taken for years, and that gap between the prescription and the practice is what creates the problem.

Disturbed movement

That clearing wave between meals only starts on an empty stomach. Constant snacking cancels it — perhaps the most everyday factor of all. Alongside it: diabetes with neuropathy, scleroderma, the aftermath of surgery.

Anatomy

Adhesions, diverticula, a removed ileocaecal valve, blind loops after operations — anything that creates a place where contents sit still.

Probiotics: why «adding bacteria» can make it worse

This is the most counter-intuitive part of the subject, and there is a specific study behind it.

In 2018 (PMID 29915215^[3]) thirty people with unexplained brain fog, bloating and gas were assessed — with an intact gut, normal endoscopy and normal imaging. They were compared with eight patients without the fog.

The mechanism: some bacteria produce the D-form of lactic acid, which humans handle poorly. When there are too many such bacteria and they live in the wrong place, D-lactate accumulates — hence the fog.

The authors' conclusion, verbatim: symptoms improved with antibiotics and stopping probiotics.

It does not follow that probiotics are harmful. What follows is different: bacteria are not automatically good. Where overgrowth already exists, adding bacteria is reinforcement sent to the wrong side.

Hence an order that matters more than any particular remedy: look first, take second. Not the other way round.

What is used to treat it

A 2025 network meta-analysis (PMID 41394885^[4]) brought together 30 trials, 1552 participants, 12 interventions and ranked them by probability of being best.

• Uncomplicated SIBO — berberine, a plant alkaloid, ranked first rather than the antibiotic. What that first place means, and what it does not, is covered separately.
• With functional gastrointestinal disorders — rifaximin combined with a prokinetic, ranking probability 89%.
• With chronic liver disease — a prokinetic alone, 79.6%.

This needs reading carefully, and here is why. A ranking probability is not the same as winning a head-to-head comparison: it says how often an intervention came out on top in the model, not that it was compared face to face with the rest. Three of the thirty trials were judged by the authors to carry a high risk of bias.

But one idea from that analysis holds without the ranking: the choice depends on what else the person has. With a functional disorder it matters more to get the gut moving; with liver disease, more still. Treating the same test with the same regimen means ignoring half the picture.

From our own catalogue the item that belongs to this subject is Cobra Gut — herbal support for an irritated stomach and gut. Support, not treatment of overgrowth: it does not replace the order set out above.

What rifaximin does well

The strongest evidence is not about SIBO itself but about irritable bowel syndrome without constipation. Two identically designed phase 3 trials, TARGET 1 and TARGET 2 (PMID 21208106^[5]): rifaximin 550 mg three times daily for just 14 days, followed for a further ten weeks.

A difference of about nine percentage points — roughly one person in eleven treated gets relief that placebo would not have given. Modest, but steady and confirmed twice.

A third trial, TARGET 3 (PMID 27528177^[6]), answers the next question: what to do when symptoms come back. 44.1% responded to an open-label course, but symptoms returned in 64.4% of them — and a repeat course worked: 38.1% against 31.5% (P=0.03).

And here is the detail almost nobody quotes. The repeat course helped pain — 50.6% against 42.2% (P=0.018) — while for stool consistency there was no difference at all: 51.8% against 50.0%, P=0.42. What comes back is part of the benefit, not all of it.

And what it does less well than is assumed

All of the above is about symptoms. On eradicating the overgrowth itself the picture differs, and it is worth having whole.

A 2021 systematic review (PMID 34767484^[7]) pooled 26 studies and 874 patients: eradication 59% [50–69] by intention to treat and 63% [53–72] per protocol. Heterogeneity was enormous, I² about 90%. A 2026 meta-analysis (PMID 41883799^[8]) confirms the same 59% and adds that doses from 1200 mg a day work slightly better than lower ones.

Convincing — right up to the point where the same review looks separately at the five studies that had a control group. Verbatim: no significant difference was found between rifaximin and placebo or active controls — a risk ratio of 1.14, 95% CI 0.59–2.19, P=0.15.

This needs putting plainly, because the number matters. 59% is the share of people whose test turned negative after treatment. But in the studies with something to compare against, just as many turned negative without rifaximin. The test fluctuates on its own, and part of that «success» is its fluctuation rather than the drug.

Hence the honest conclusion: rifaximin demonstrably helps symptoms in IBS without constipation, while its ability to clear the overgrowth as such is not confirmed by controlled data.

The methane form is treated differently

If the test is methane-positive, rifaximin alone is weaker — archaea are not bacteria. A 2026 review of managing IBS subtypes puts it directly: the methane-predominant form with constipation responds better to a rifaximin–neomycin combination, while in the diarrhoea subtype rifaximin itself gives strong relief (PMID 42310284^[9]). This is a review rather than a single trial — but its direction matches the mechanism.

And a separate finding, this one about the test itself. In a 2026 study (PMID 42509595^[10]) 188 people with constipation underwent glucose breath testing, and methane was read not by a present-or-absent threshold but by the shape of the curve. Three distinct patterns emerged, and they are not equivalent.

The first pattern is associated with slow movement through the colon independently of everything else — an odds ratio of 5.67 [2.22–14.50]. And it is more precise than the usual threshold: specificity 92.4% against 76.5% for the familiar «methane ≥ 10 ppm».

The point here is the one running through the whole subject: a single threshold number discards the shape of the curve, and the shape carries information. «Methane positive» on its own says less than how the methane behaved across the test.

On safety, and on the limits

A common worry is Clostridioides infection after an antibiotic. It was checked in a large cohort: 19,597 people in each arm after matching, infection within 60 days in 0.21% of those given rifaximin against 0.15% of those not. But P=0.152 — no significant difference was found, and the authors themselves warn that events were too few and the study may simply have missed one (PMID 42355617^[11]). Repeat courses added no risk: 0.20% against 0.21%.

And the frame without which the picture is incomplete. In Europe rifaximin for SIBO is prescribed off-label. In Germany it is authorised for travellers' diarrhoea and for preventing hepatic encephalopathy — there is no SIBO indication; in Italy the wording is broader and covers intestinal infections; in the United States there is a separate indication for IBS with diarrhoea. This is not a prohibition: off-label use is lawful and ordinary, but the decision is the physician's and it is theirs to carry.

The bottom line

SIBO is a real condition with a real test, but the test sees less than is asked of it. It does not measure the intestine directly, it catches gas in the breath; it does not assess transit, though it is used for that; and it does not tell bacteria from archaea until methane is read separately.

So the practical sense of the whole subject comes down to order: test first, treat second, and treat the person rather than the test — with their acid suppressants, their snacking, their surgical history.

And one thing worth keeping even if the rest is forgotten: with bloating and unstable stool, adding bacteria blind can make it worse. This is the case where taking a step without checking is worse than taking none.

The breath test is one of the four checks in our guide to four systems, where the other three are set out.

This article is for information. A diagnosis follows investigation, and treatment — antibiotics and herbal agents alike — is prescribed by a physician.