Digestive Health, Motility and the Intestinal Barrier

01 What this document is, and five things it is not

This article asks one question of a large commercial category: when a product is sold for digestion, regularity, bloating, or gut health, what has actually been measured, in whom, for how long, and against what comparator?

The category has a split personality, and the split is the reason it is so easy to sell into. At one end, digestive endpoints are among the most concrete in nutritional science. Stool water content is a percentage. Stool output is grams per week. Breath hydrogen is parts per million. Pancreatic lipase activity is United States Pharmacopeia units. Faecal elastase is micrograms per gram, with a numerical threshold below which a diagnosis is warranted (Whitcomb et al., 2023). Gastric half-emptying time is minutes. These are physical quantities, and a trial that moves one of them has moved something real.

At the other end sits the vocabulary the products are actually sold on. Bloating is a sensation, distension is a measurable circumference, and the two are not the same variable (Malagelada et al., 2017). Digestive support corresponds to no quantity at all. Leaky gut is a consumer category built on a research construct whose own leading reviewer states that no disease has ever been cured by normalising it (Camilleri, 2019).

The distance between those two ends is where this article lives. Five things follow immediately.

First, this is not a treatment manual. It recommends no compound, amount, route, or schedule for any person. Several agents discussed here alter motility, luminal water, or gastric emptying, which is a reason for care rather than a dosing instruction.

Second, it is not an ingredient catalogue. The individual subjects have their own articles in this series — Ginger, Peppermint, Dietary-Fibre, L-Glutamine, Zinc, Marshmallow-Root, Turmeric, Aloe-Vera, Digestive-Enzymes — and those remain the deep references. The companion Microbiome-Nutrition title takes probiotics, prebiotics, fermentable fibre, synbiotics and postbiotics; this article hands that literature over deliberately and explains why at the end of section 14. The work of a system article is synthesis: which outcomes matter, which mechanisms are plausible, which claims survive a randomised trial, which work only where a function is genuinely missing, and which combinations exist only because they appear together on a label.

Third, it is not a substitute for diagnosis, and this matters more here than in most of this series. Several conditions here are defined by positive criteria rather than by the failure of a supplement — functional dyspepsia and functional bloating by Rome criteria (Rome Foundation, 2026; Drossman et al., 2026), irritable bowel syndrome by national guidance (National Institute for Health and Care Excellence, 2026), gastroparesis by a four-hour emptying study (Staller et al., 2025), and exocrine pancreatic insufficiency by a stool enzyme assay (Whitcomb et al., 2023). One guideline here states explicitly that improvement on a trial of enzymes is unreliable as evidence that enzymes were needed. That sentence undercuts the most common form of self-diagnosis in this entire category.

Fourth, it is not neutral about the word "support". Where a product's evidence consists of a mechanism, an in-vitro activity, or a single-meal marker, this document says so in the sentence that reports it.

Fifth, it is not neutral about "leaky gut". The intestinal barrier is real, layered, measurable within limits, and genuinely perturbed by identifiable stressors. What is not established is that a consumer can measure their own barrier, that a supplement can repair it in the absence of a provocation, or that repairing it would relieve anything. Sections 20 to 22 set out what the evidence does and does not carry.

02 The physiology, in one page

The gut is a sequential organ, and almost every product here acts at one identifiable station.

Mechanical processing and gastric emptying. Chewing and antral trituration reduce solids to particles small enough to pass the pylorus. The stomach meters delivery of chyme into the duodenum, and the rate of that delivery sets postprandial sensation as much as the meal's composition. Agents that accelerate or delay emptying act here, and so does the sensation of fullness.

Chemical digestion. Gastric acid and pepsin begin protein hydrolysis; bile salts emulsify fat; pancreatic amylase, proteases and lipase perform the bulk of luminal digestion, with brush-border enzymes finishing the job. Pancreatic lipase does the overwhelming majority of triglyceride hydrolysis, which is why fat maldigestion is the signature of pancreatic failure and why enzyme replacement is dosed in lipase units (Domínguez-Muñoz et al., 2025).

Absorption and the colon as a water organ. The small intestine absorbs nutrients across an enormous surface. The colon recovers water and electrolytes and ferments what reaches it. Stool consistency is a water-content phenomenon, and it is the physical variable that most fibre products act on.

Motility, which is neurogenic. The gastrointestinal tract is the only internal organ to have evolved its own independent nervous system. Spencer and Hu review how enteric neurons transduce distension and chemical stimuli into motor patterns, including the mechanotransduction underlying distension-evoked colonic peristalsis, and note a finding that overturned the previous consensus: mucosal serotonin is not required for peristalsis or for colonic migrating motor complexes, although it modulates their characteristics (Spencer and Hu, 2020). Motility is therefore a property of a local neural circuit, not of a single signalling molecule that a supplement might top up.

The barrier, which is four things stacked. From the lumen inwards: a mucus layer, an epithelium sealed by tight junctions, an immune compartment, and the microbiota itself. The mucus is not one layer but two, organised around the heavily glycosylated MUC2 mucin secreted by goblet cells. Johansson, Larsson and Hansson showed that the inner layer is stratified, firmly adherent, approximately 50 µm thick in mouse, and dense enough that bacteria do not penetrate it, keeping the epithelial surface free of bacteria; it is continuously converted into an outer layer roughly 100 µm thick that is the habitat of the commensal flora, expanded by host and probably bacterial proteases and glycosidases, whose MUC2 O-glycans serve both as bacterial nutrients and as attachment sites (Johansson et al., 2011). The functional barrier is therefore partly a secreted gel, which is worth holding in mind when a product claims to "seal" a junction.

One idea organises everything that follows: an oral agent changes a digestive outcome only if it survives to the right segment, acts on a step that is genuinely rate-limiting, and that step is the one limiting the symptom. Most of this review is about how often the third condition fails.

03 The outcome ladder

Seven different things are called "improving digestion", and they are not degrees of one another.

Enzyme or receptor activity in vitro. A hydrolysis rate in a buffer, or a relaxation curve in an isolated tissue bath. Informative about mechanism; silent about everything else.

Luminal chemistry. Breath hydrogen after a substrate load, or a urinary sugar ratio. Measures what happened in the lumen, in the hours after a challenge.

A physical stool property. Water content as a percentage, water weight per bowel movement, total output in grams per week. Cheap, objective, and the endpoint on which the fibre literature rests.

Transit. Colonic transit time, gastric half-emptying time. A rate, not a symptom.

A symptom score. A visual-analogue or Likert instrument for fullness, pain, bloating or flatulence, over days to weeks.

A validated responder definition. A pre-specified dichotomous threshold — a reduction of at least 50 points on the Irritable Bowel Syndrome Severity Scoring System, an increase of at least one complete spontaneous bowel movement per week — which is what regulators and guideline panels treat as a clinical result.

Disease modification or a hard outcome. Mucosal healing, prevented hospitalisation, corrected nutritional status, prevented complication. In this field almost nothing reaches this rung, and the one place it is routinely measured is not a supplement but pancreatic enzyme replacement in diagnosed insufficiency, where the outcome is body weight, muscle function and fat-soluble vitamin status (Whitcomb et al., 2023; Kadaj-Lipka et al., 2025).

A product may sit four rungs below where its packaging implies. Figure 1 sets out which rung each intervention class here actually reached.

04 Replacement, repair, and the third thing that is neither

Three completely different arguments are made for digestive products, and conflating them is the category's central error.

Replacement. A specific function is absent or reduced; supply it exogenously. This is coherent only when the deficit can be identified independently of the response to treatment. Exocrine pancreatic insufficiency qualifies: faecal elastase below 100 µg/g of stool provides good evidence of it, values of 100 to 200 are indeterminate, and — the crucial line — response to a therapeutic trial of pancreatic enzymes is unreliable for diagnosis (Whitcomb et al., 2023). Lactase deficiency qualifies, by breath test against a lactose load (Ojetti et al., 2010). Very little else in the aisle does.

Modification of luminal physics. Do not treat the tissue; change what is in the lumen. Raise viscosity, hold water, resist dehydration, remove a fermentable substrate before the microbiota reaches it, coalesce gas bubbles. These claims are computable from physical properties and testable with objective endpoints, and this is where the strongest non-drug evidence in the article sits (McRorie and McKeown, 2017).

Repair or protection of the mucosa. The tissue is damaged, permeable, or inflamed, and the agent heals or shields it. This is what most of the premium end of the category sells. It is also where clinical demonstration is thinnest: barrier markers have been moved, chiefly against an experimental provocation, and the leading review of the field states that no disease has been cured by normalising barrier function and that it remains unproven that restoring it ameliorates clinical manifestations at all (Camilleri, 2019; Camilleri, 2021).

The commercial distribution is close to inverted. The most-advertised logic is the third; the best-evidenced is the second; and the first has licensed drugs precisely because it is the one place a missing function can be named.

05 Bloating, the hardest word here

More products in this category are sold for bloating than for anything else, and bloating is the endpoint least likely to be the variable a product acts on.

Malagelada, Accarino and Azpiroz separate the terms: bloating is a symptom, distension is a sign, individual patients weight them differently, and complaints range from chronic highly distressing pain to an annoying and unfashionable protrusion of the abdomen (Malagelada et al., 2017). Those are not one clinical problem.

The mechanistic reason matters. Moshiree, Drossman and Shaukat's expert review sets out an evaluation and management framework in fifteen best-practice statements, and two of them reframe the whole product category. Abdominophrenic dyssynergia — a paradoxical descent of the diaphragm with relaxation of the anterior abdominal wall — produces visible girth increase without a change in gas volume, and the treatments named for it are diaphragmatic breathing and central neuromodulators rather than anything that reduces gas (Moshiree et al., 2023). The European consensus reaches the same conclusion, listing the pathophysiology as multifactorial across visceral hypersensitivity, abdominophrenic dyssynergia, intestinal dysmotility and dysbiosis (Melchior et al., 2025).

If girth can rise without more gas, then a product that removes gas is, in a substantial fraction of patients, acting on a variable that is not causing the sign.

Two further statements from the same review deserve to be quoted rather than paraphrased, because they are the closest thing this field has to a regulator. Gastric emptying studies should not be ordered routinely for bloating and distension. And probiotics should not be used to treat abdominal bloating and distention (Moshiree et al., 2023). The second is an explicit negative recommendation against the single most common ingredient class in bloating products, and it is issued by the professional society whose members see these patients.

06 What already works, and why that sets the bar

No agent here can be judged without knowing what a serious intervention achieves, because the useful question is never "does it do anything" but "does it do anything worth doing".

Constipation. Rao and Brenner screened 1,297 studies and included 41 randomised trials of at least four weeks, grading over-the-counter therapies by level of evidence. Good evidence, grade A, went to two agents: the osmotic laxative polyethylene glycol and the stimulant senna. Moderate evidence, grade B, went to psyllium, magnesium salts, bisacodyl and sodium picosulfate, and fruit-based laxatives including kiwi, mango, prunes and figs. Insufficient evidence, grade I, went to polydextrose, inulin and fructo-oligosaccharide — three of the most heavily marketed "prebiotic fibres" in the aisle. Diarrhoea, nausea, bloating and abdominal pain were common adverse events across the class, with no serious adverse events reported (Rao and Brenner, 2021).

The joint guideline agrees and adds prescription comparators. Chang, Chey and colleagues' panel made strong recommendations for polyethylene glycol, sodium picosulfate, linaclotide, plecanatide and prucalopride, and conditional recommendations for fibre, lactulose, senna, magnesium oxide and lubiprostone (Chang et al., 2023). Fibre is in the guideline, and it is in the conditional tier.

Between the two osmotic agents there is a direct answer. Lee-Robichaud, Thomas, Morgan and Nelson pooled all ten randomised trials then available and found polyethylene glycol better than lactulose for stool frequency, stool form, relief of abdominal pain and the need for additional products, in both adults and children, concluding that polyethylene glycol should be used in preference (Lee-Robichaud et al., 2010). This is established.

The state of the dietary evidence for the same condition is worth stating precisely, because it is the honest backdrop to every claim here. Dimidi, van der Schoot, Whelan and colleagues built the first comprehensive evidence-based dietary guideline for chronic constipation from four systematic reviews covering 75 randomised trials, and generated 59 recommendation statements. Of those 59, eight had moderate-quality evidence, 39 had low, and 12 had very low (Dimidi et al., 2025). That is the best-organised dietary evidence base in this field, assembled by a national professional body, and seven-eighths of it rests on low or very low quality evidence.

Dyspepsia. Black, Paine and colleagues' British Society of Gastroenterology guideline reviewed the evidence, updated a series of pairwise and network meta-analyses, and graded recommendations for a condition affecting roughly 7% of the community and managed mostly in primary care (Black et al., 2022a). Within that framework one intervention has an unusually clean result. Ford, Tsipotis and colleagues pooled 29 randomised trials in 6,781 Helicobacter pylori-positive patients with functional dyspepsia and found eradication therapy superior to control for symptom cure (relative risk of symptoms not being cured 0.91, 95% CI 0.88 to 0.94, number needed to treat 14) and for improvement (0.84, 0.78 to 0.91, number needed to treat 9), with a larger effect where eradication actually succeeded (0.65, 0.52 to 0.82, number needed to treat 4.5) — and with adverse events more than twice as common on treatment (2.19, 1.10 to 4.37) (Ford et al., 2022). The authors call the evidence high quality and the benefit modest. That is what a real, targeted, mechanism-matched intervention looks like in this organ: an eradicable cause, a number needed to treat between 4.5 and 14, and a stated cost in adverse events.

Eusebi, Black, Howden and Ford's network meta-analysis of 15 trials in 6,162 participants ranked test-and-treat first for uninvestigated dyspepsia, with prompt endoscopy performing similarly, and no strategy significantly less effective than test-and-treat — while noting that participants were more satisfied with prompt endoscopy (Eusebi et al., 2019). A first-ranked strategy that is not superior to its alternatives is a recurring pattern in this organ, and it recurs for supplements too.

Irritable bowel syndrome. Black, Staudacher and Ford's network meta-analysis of 13 randomised trials in 944 patients ranked a diet low in fermentable oligosaccharides, disaccharides, monosaccharides and polyols first for global symptoms against habitual diet (relative risk of symptoms not improving 0.67, 95% CI 0.48 to 0.91), first for abdominal pain severity, and first for bloating or distension, where it was superior to standard dietetic advice (0.72, 0.55 to 0.94) (Black et al., 2022b). Nybacka, Törnblom and colleagues then tested diet against drugs head to head in 294 patients: at four weeks, a low-FODMAP diet with traditional advice produced a responder rate of 76%, a fibre-optimised low-carbohydrate diet 71%, and optimised medical treatment selected for the predominant symptom 58%, with a significant difference between groups (Nybacka et al., 2024).

That is the bar. Diet outperformed drug therapy for this condition in a randomised comparison, and the drugs were chosen to match each patient's dominant symptom.

07 What the tests can and cannot tell you

Four measurement problems shape everything that follows, and three of them are invisible in product marketing.

The enzyme deficit is measurable, and the therapeutic trial is not a test. Faecal elastase must be performed on a semi-solid or solid stool specimen; below 100 µg/g provides good evidence of exocrine pancreatic insufficiency and 100 to 200 is indeterminate; the assay can be run while on enzyme therapy; quantitative faecal fat is rarely needed and impractical; cross-sectional imaging cannot identify the condition; and response to a therapeutic trial of pancreatic enzymes is unreliable for diagnosis (Whitcomb et al., 2023). The last clause is the one that matters commercially, because "I took enzymes and felt better, so I must need enzymes" is the reasoning the category runs on, and a professional society has specifically declined to accept it.

Gastric emptying has a right and a wrong protocol. The gastroparesis guideline panel issued a conditional recommendation against two-hour gastric emptying testing and in favour of four-hour testing in suspected gastroparesis, and conditional recommendations against domperidone, prucalopride, aprepitant, nortriptyline, buspirone and cannabidiol as first-line therapies (Staller et al., 2025). A delayed-emptying claim rests on a protocol that half the literature did not use.

Breath hydrogen measures fermentation, not suffering. It is the correct endpoint for a substrate-directed intervention and a poor proxy for a symptom, and the article reports it as what it is throughout.

Permeability measurement disagrees with itself between laboratories. The dual-sugar test is the accepted in-vivo probe, and Lee, Kosek, Lima and colleagues compared platforms on 100 samples from a Peruvian infant cohort. Between two liquid chromatography–tandem mass spectrometry platforms, agreement was high, with Spearman ρ of at least 0.89 for lactulose, mannitol and the ratio. Between high-performance liquid chromatography with pulsed amperometric detection and mass spectrometry, ρ was 0.95 for mannitol, 0.70 for lactulose, and 0.43 for the lactulose:mannitol ratio — with the chromatographic platform overestimating the lowest disaccharide concentrations most (Lee et al., 2014). The ratio is the number the barrier literature reports, and it is the number on which two laboratories agree least.

Camilleri's reviews add the conceptual constraint: assessment of the intestinal barrier requires measurements beyond the epithelial layer, and beyond the measurement of tight junction expression; barrier function is most meaningfully tested in vivo with orally administered probe molecules, with other approaches performed on biopsies or cell layers (Camilleri, 2019; Camilleri, 2021). Rath, Atreya and Neurath, writing from inside inflammatory bowel disease where barrier defects are best characterised, describe barrier assessment as "far away from standard" while arguing that new technologies could make barrier healing clinically assessable (Rath et al., 2023).

Two conclusions follow. A consumer cannot measure their own intestinal barrier. And a trial that reports tight-junction protein expression in a cell line has not measured permeability in a person.

08 Pancreatic enzyme replacement, the licensed case

Exocrine pancreatic insufficiency is defined as a reduction in pancreatic exocrine secretion below the level that allows normal digestion of nutrients. The European guideline, developed jointly by seven societies with systematic review, Oxford and GRADE evidence appraisal and Delphi voting, states that diagnosis should rest on a global assessment of symptoms, nutritional status and a pancreatic secretion test, and that enzyme replacement together with dietary advice is the cornerstone of therapy in insufficiency secondary to pancreatic disease, pancreatic surgery, or other metabolic and gastroenterological conditions (Domínguez-Muñoz et al., 2025).

The dosing is specific and the reasoning is nutritional rather than symptomatic. Enzymes are taken during the meal, with an initial dose of at least 40,000 USP units of lipase per meal in adults and half that with snacks, adjusted for meal size and fat content; non-enteric-coated preparations require acid suppression; all preparations are porcine-derived and equally effective at equivalent doses; and success is measured by reduced steatorrhoea, gain of weight, muscle mass and muscle function, and improved fat-soluble vitamin levels, with baseline and repeat bone densitometry every one to two years (Whitcomb et al., 2023).

That last clause is what a top-rung outcome looks like, and the real-world literature shows what happens when the dose falls short of it. Kadaj-Lipka, Monica and colleagues systematically reviewed 25 observational studies covering 3,818 patients. In 40% of the studies, average doses were below the recommended 40,000 to 50,000 lipase units per meal. Under-dosing produced significant relief of diarrhoea in nearly all of those studies — and improved nutritional status in none of them. Guideline-concordant dosing reduced diarrhoea in most studies and improved or maintained nutritional status (Kadaj-Lipka et al., 2025).

This is the most useful single finding in the enzyme chapter, and it generalises beyond enzymes. A dose that relieves the symptom is not necessarily a dose that corrects the deficit. The symptom is the cheaper endpoint, it responds first, and stopping there leaves the patient malnourished and satisfied.

Lewis, Rieke and colleagues reviewed 257 articles and found dosing guidance varying widely both within and across disease types, most guidelines framed as initial doses rather than upper limits, most studies focused only on fat digestion, and under-dosing common when compared by total daily dose (Lewis et al., 2024). Grade for enzyme replacement in diagnosed insufficiency: established, with the dose–outcome relationship as part of the finding.

09 The dissolution problem, and why an enzyme blend is not a small version of a drug

The most instructive study here is a laboratory assay of marketed products, and it is what separates a licensed enzyme from a supplement enzyme.

Case, Henniges and Barkin bought nine different pancreatic enzyme products on the open market and tested them by United States Pharmacopeia methods for amylase, protease and lipase content, and for dissolution. Every product's enzyme assay fell within USP limits, and the percentage of label claim was higher than the label stated for all but one batch. Then they ran the dissolution test — one hour in simulated gastric fluid at pH 1.0, then thirty minutes in pH 6 phosphate buffer. Two of three batches of one product did not dissolve at all, and one batch of another released 8% of its lipase activity (Case et al., 2005).

Content on the label is not delivery in the duodenum. Nine products passed the test that measures what is in the capsule, and two failed the test that measures whether it gets out at the right place.

The regulatory consequence is recorded in the same paper: the Food and Drug Administration decreed that all pancreatic enzyme products would require an approved New Drug Application, because differences in pharmaceutical quality had been identified, making product substitution questionable (Case et al., 2005). Pancrelipase is accordingly a drug with an approved application and a dissolution specification (United States Food and Drug Administration, 2026a).

A digestive enzyme supplement has neither. Under 21 CFR 101.93 a structure/function statement on a dietary supplement label requires only a disclaimer that the Food and Drug Administration has not evaluated the statement and that the product is not intended to diagnose, treat, cure or prevent any disease (Office of the Federal Register, 2026), and the agency states that it does not approve dietary supplements before they are marketed (United States Food and Drug Administration, 2026b). The same enzymes, the same USP units on the panel, and no requirement to demonstrate that any of it survives the stomach.

This is not a hypothetical objection. It is the finding of the only study that looked, in the product class where quality was regulated because of it.

10 Lactase, and a deficiency you can actually name

Lactase is the cleanest replacement argument in the aisle, and it shows both what replacement can do and how quickly the answer changes when the population changes.

Ojetti, Gigante and colleagues randomised 60 lactose-intolerant patients to three arms: tilactase immediately before a control lactose hydrogen breath test, placebo before the same test, or Lactobacillus reuteri twice daily for ten days beforehand. Breath-test normalisation was significantly more frequent with tilactase and with the probiotic than with placebo, and significantly more frequent with tilactase than with the probiotic (p < 0.01). Both active arms reduced mean peak hydrogen excretion and improved the mean clinical score against placebo (p < 0.0001), and tilactase was significantly more effective than the probiotic on both (Ojetti et al., 2010). Grade: strongly supported, for a named deficit, a defined challenge, and endpoints measured hours later.

Now change the population. Kozłowska-Jalowska, Stróżyk, Horvath and Szajewska systematically reviewed lactase supplementation for infant colic and found five randomised trials in 391 infants under six months. Three reported reduced crying duration, but one of those only within a compliant subgroup comprising 40.4% of participants. A meta-analysis of the two poolable trials found no difference in crying duration over a week of treatment (mean difference −17.66 min/day, 95% CI −60.8 to 25.5, I² = 68%) and none in fussing time (2.75, −58.2 to 57.2, I² = 80%). Risk of bias was low in one trial, high in three, and raised concerns in the fifth (Kozłowska-Jalowska et al., 2024). The reviewers' conclusion is that the evidence remains inconclusive.

Same enzyme, same substrate, same mechanism. In an adult with a positive breath test and a measured hydrogen peak, it works on the endpoint it was designed for. In an infant whose crying may or may not be caused by lactose, pooled across two trials, it does not.

11 Substrate-directed enzymes, and the enzymes with no missing function

Three further enzyme arguments exist, and they descend in strength.

Alpha-galactosidase, which removes the substrate. Di Stefano, Miceli and colleagues gave eight healthy volunteers a test meal containing 420 g of cooked beans with 300 or 1,200 galactosidase units or placebo, and measured breath hydrogen and symptoms for eight hours. The higher dose significantly reduced both breath hydrogen excretion and flatulence severity; both doses significantly reduced the total symptom score, with reductions apparent across all symptoms considered (Di Stefano et al., 2007). The mechanism is unambiguous: hydrolyse the oligosaccharide in the small bowel so the colonic microbiota never receives it. Grade: emerging. Eight volunteers, one meal, healthy, and no replication at scale in thirty years.

It is worth naming the tension this creates. The mechanism by which alpha-galactosidase relieves flatulence is the deliberate prevention of colonic fermentation — the same fermentation that the prebiotic half of this market sells as the benefit. Both cannot be the goal simultaneously, and which one a person wants depends entirely on which endpoint they care about. Section 14 returns to this, and the companion Microbiome-Nutrition title takes the other side of it.

Gluten-degrading enzymes, which are a research programme. Wei, Helmerhorst, Darwish and colleagues review the field: prolyl endopeptidases from bacterial sources, cysteine proteases, subtilisins from oral Rothia colonisers and engineered synthetic enzymes can cleave the proline- and glutamine-rich, digestion-resistant gluten peptides that activate T cells in coeliac disease, both in vitro and in vivo. To be therapeutic such an enzyme must be stable and active in the acidic stomach and at near-neutral duodenal pH, and must neutralise the antigenic peptides quickly. Their conclusion is that there is no effective or approved treatment for coeliac disease other than a strict gluten-free diet, and that oral enzyme therapy is a promising approach requiring enteric coating and genetic modification to achieve the necessary stability (Wei et al., 2020). Grade: plausible as pharmacology, speculative as a consumer product, and irrelevant to coeliac disease management today.

Lipase in people whose pancreas works, where there is no missing function to replace. Levine, Koch and Koch gave 16 healthy volunteers a capsule containing 280 mg of acid-resistant lipase or placebo immediately before a 355-calorie liquid meal that was 55% fat, in a double-blind crossover, and measured symptoms and electrogastrography for an hour. The meal reliably induced fullness, bloating, nausea and tachygastria (p < 0.01 and p < 0.05). With lipase, reports of stomach fullness were significantly lower than placebo — and there was no effect on gastric myoelectrical activity, bloating, or nausea (Levine et al., 2015). One symptom out of three, in healthy people, after one liquid meal, with the objective measure unchanged. Grade: emerging for postprandial fullness after a high-fat liquid meal; nothing beyond that.

Enzyme	Function it replaces	Is the deficit measurable?	Best human evidence	Regulatory status
Pancrelipase	pancreatic lipase, protease, amylase	yes: faecal elastase <100 µg/g	nutritional status corrected at guideline dose	drug, approved application, dissolution specification
Lactase (tilactase)	brush-border lactase	yes: lactose breath test	breath normalisation and symptom score, adult maldigesters	supplement; also a licensed product in some markets
Alpha-galactosidase	none — removes substrate	no	breath hydrogen and flatulence, 8 volunteers, one meal	supplement
Acid-resistant lipase, healthy adults	none	no	fullness only, 16 volunteers, one liquid meal	supplement
Gluten-degrading proteases	none — degrades antigen	not applicable	in vitro and in vivo peptide cleavage; no approved therapy	investigational
"Digestive enzyme complex" blends	unspecified	no	none as the product sold	supplement, disclaimer only

Table 1. The enzyme ledger. Only the first two rows describe a function that can be shown to be absent before treatment, and only the first has an outcome above the symptom rung. Two of nine marketed pancreatic products failed a dissolution test that supplement enzymes are not required to pass at all.

12 The physics of fibre, and why the soluble-insoluble split is wrong

The most consequential correction here is not about a botanical. It is that the classification scheme printed on fibre supplements does not predict what they do.

McRorie and McKeown set out the physics. In the small bowel, the clinically meaningful benefits — cholesterol lowering and improved glycaemic control — track viscosity, not solubility: high-viscosity gel-forming fibres such as β-glucan, psyllium and raw guar gum produce them, while non-viscous soluble fibres such as inulin, fructo-oligosaccharides and wheat dextrin, and insoluble fibres such as wheat bran, do not. In the large bowel there are exactly two mechanisms that drive a laxative effect: large, coarse insoluble particles that mechanically irritate the mucosa and stimulate water and mucus secretion, and the high water-holding capacity of a gel-forming soluble fibre that resists dehydration. Both mechanisms require that the fibre resist fermentation and still be present in stool. Soluble fermentable fibres — inulin, fructo-oligosaccharide, wheat dextrin — do not provide a laxative effect, and some fibres can be constipating, specifically wheat dextrin and fine, smooth insoluble wheat bran particles (McRorie and McKeown, 2017).

Two of those clauses deserve to be read twice. Fermentability, which the prebiotic market treats as the whole point, is disqualifying for laxation, because a fibre that is fermented away is not in the stool to hold water. And a fibre supplement can cause the symptom it is bought for, depending on particle size.

McRorie, Fahey, Gibb and Chey quantified the comparison directly. In patients with chronic idiopathic constipation, non-fermented gel-forming psyllium was 3.4 times more effective than insoluble wheat bran for increasing stool output. Both psyllium and coarse wheat bran increased stool water content, a softening effect — while finely ground wheat bran decreased stool water content, a hardening effect. Their conclusion is that it is a misconception that dietary fibre and all isolated fibres provide a laxative effect, that most isolated fibres are not different from placebo for laxation and several may be constipating, and that treatment guidelines should make specific evidence-based recommendations, because a generic instruction to "increase fiber intake" is akin to a recommendation to "increase pill intake" without regard to therapeutic or adverse effects (McRorie et al., 2020). Grade: established for the mechanism, and it is the rare case here where the mechanism is strong enough to predict which products will fail.

13 Psyllium against its comparators

Psyllium is the one supplement here that appears in a clinical guideline as a therapy, and its evidence is old, physical and boring in the best sense.

Ashraf, Park, Lof and Quigley randomised 22 people with prospectively confirmed idiopathic constipation to psyllium 5 g twice daily or placebo for eight weeks, with a four-week placebo run-in and a four-week washout, and measured colon transit and anorectal manometry at the start and end of each phase. Stool frequency rose from 2.9 ± 0.1 to 3.8 ± 0.4 per week and stool weight from 405.2 ± 75.9 g to 665.3 ± 95.8 g (both p < 0.05); consistency and pain on defecation improved. Colon transit and every anorectal manometry parameter were unchanged, and the placebo group changed on nothing. The authors concluded that the benefit is primarily a facilitation of the defecatory process (Ashraf et al., 1995).

That is a precise mechanistic statement, and it is not the one on the packet. Psyllium does not speed the colon up. It changes the object being passed.

McRorie, Daggy and colleagues ran the head-to-head against the other classic softener: 170 subjects with chronic idiopathic constipation, two weeks of placebo baseline, then two weeks of psyllium 5.1 g twice daily against docusate sodium 100 mg twice daily in a double-blind parallel design with stools collected and assessed. Psyllium increased stool water content by 2.33% against docusate's 0.01% (p = 0.007), stool water weight 84.0 against 71.4 g per bowel movement (p = 0.04), total stool output 359.9 against 271.9 g per week (p = 0.005), and bowel movements 3.5 against 2.9 per week in the second treatment week (p = 0.02), with no difference in week one (McRorie et al., 1998). Grade: established, on objective collected-stool endpoints, against an active comparator.

The pooled picture adds both the dose threshold and the cost. van der Schoot, Drysdale, Whelan and Dimidi meta-analysed 16 randomised trials in 1,251 participants: 311 of 473 fibre-treated participants responded, 66%, against 134 of 329 controls, 41% (relative risk 1.48, 95% CI 1.17 to 1.88, p = 0.001, I² = 57%), with psyllium and pectin the agents carrying significant effects. Higher response appeared only with doses above 10 g/day; stool frequency improved (standardised mean difference 0.72, 0.36 to 1.08) only at higher doses and durations of at least four weeks, with I² of 86%; consistency improved (0.32, 0.18 to 0.46). And flatulence was higher in the fibre groups than controls (0.80, 0.47 to 1.13, p < 0.00001) (van der Schoot et al., 2022).

The dose-limiting adverse effect of the commonest digestive supplement is a symptom that other products in the same aisle are sold to treat.

Chey, Chey, Jackson and Eswaran then tested it against food. Seventy-nine patients with chronic constipation and no more than three complete spontaneous bowel movements per week were partially randomised to two green kiwifruit daily, 100 g of prunes daily, or 12 g of psyllium daily for four weeks. On the primary endpoint — an increase of at least one complete spontaneous bowel movement per week in at least two of four weeks — the proportions of responders were similar across all three. All three raised weekly rates against baseline in weeks three and four (p ≤ 0.003); kiwifruit and prunes improved consistency; all three improved straining; kiwifruit alone improved bloating (p = 0.02). Adverse events were most common with psyllium and least common with kiwifruit, and fewer patients were dissatisfied with kiwifruit than with prunes or psyllium (p = 0.02) (Chey et al., 2021).

Two fruits matched the supplement on the primary endpoint and beat it on tolerability. Both foods carry grade B evidence in the over-the-counter review alongside psyllium (Rao and Brenner, 2021), and both appear in the dietetic guideline's recommendation statements (Dimidi et al., 2025). Ispaghula husk is itself a registered herbal medicinal product in Europe rather than a novel supplement claim (European Medicines Agency, 2026a).

14 Fibre in irritable bowel syndrome, where the same molecule does a different job

Move from constipation to irritable bowel syndrome and psyllium keeps working, but the relevant comparator changes and so does the reason.

Bijkerk, de Wit and colleagues randomised 275 primary-care patients aged 18 to 65 with irritable bowel syndrome to 12 weeks of psyllium 10 g, bran 10 g, or rice-flour placebo 10 g, with adequate symptom relief for at least two weeks in the previous month as the primary endpoint. Responders were significantly more frequent on psyllium than placebo in month one (57% against 35%, relative risk 1.60, 95% CI 1.13 to 2.26) and month two (59% against 41%, 1.44, 1.02 to 2.06). Bran beat placebo only in month three, and not in the worst-case analysis (1.45, 0.97 to 2.16). After three months, symptom severity fell 90 points on psyllium against 49 on placebo (p = 0.03) and 58 on bran (p = 0.61 against placebo). Quality of life did not differ. Early dropout was most common in the bran group, and the main reason was that irritable bowel symptoms worsened (Bijkerk et al., 2009).

Note the shape of the psyllium result: significant in months one and two, and no longer reported as significant in month three. A three-month trial that reports its primary endpoint monthly is showing the reader something about durability, and the honest reading is a real early benefit of uncertain persistence.

Black, Yuan, Selinger and colleagues placed soluble fibre in the wider first-line set with a network meta-analysis of 51 randomised trials in 4,644 patients — of which only 13 were at low risk of bias. On failure to achieve improvement in global symptoms at four to twelve weeks, peppermint oil capsules ranked first (relative risk 0.63, 95% CI 0.48 to 0.83, P-score 0.84) and tricyclic antidepressants second (0.66, 0.53 to 0.83, P-score 0.77). For abdominal pain, tricyclics ranked first (0.53, 0.34 to 0.83, P-score 0.87) — on data from four trials and 92 patients. And critically: there were no significant differences between active treatments on either direct or indirect comparison for either endpoint. Tricyclics were more likely than placebo to cause adverse events (1.59, 1.26 to 2.06). The authors' own interpretation stresses considerable uncertainty given methodological rigour, and that treatment duration of four to twelve weeks leaves long-term relative efficacy unknown (Black et al., 2020).

So soluble fibre, peppermint oil, antispasmodics and tricyclics are, on the available evidence, statistically indistinguishable from one another and distinguishable from placebo. That is a genuinely useful conclusion, and it is not the conclusion any of their marketing draws.

The fermentability boundary. This is the point at which this article hands over to its companion. The property that makes psyllium a laxative is that it resists fermentation and is therefore still present in stool to hold water (McRorie and McKeown, 2017). The property that makes inulin a prebiotic is that it is fermented completely. Those are the same variable with opposite signs, and which sign a reader wants depends on whether the endpoint is stool consistency or microbial substrate. The over-the-counter review found insufficient evidence for inulin, fructo-oligosaccharide and polydextrose as laxatives (Rao and Brenner, 2021) — which is not a criticism of prebiotics, but a statement that they were tested on the wrong endpoint. Fermentable fibre, probiotics, synbiotics, postbiotics and short-chain fatty acids belong to Microbiome-Nutrition, and this article does not restate them.

15 Peppermint oil, the best-supported agent here and still very low certainty

Peppermint oil has the most complete mechanism here, the highest rank in a network meta-analysis of first-line irritable-bowel therapies, and an explicit statement from its own meta-analysts that the quality of evidence is very low. Holding all three at once is the whole discipline of this document.

The mechanism is worked out to the level of an ion channel. Hills and Aaronson studied isolated guinea-pig large intestine and rabbit jejunum with tissue pharmacology and patch-clamp electrophysiology. Peppermint oil relaxed carbachol-contracted guinea-pig taenia coli with an IC50 of 22.1 µg/mL, inhibited spontaneous activity in guinea-pig colon at 25.9 µg/mL and rabbit jejunum at 15.2 µg/mL, and markedly attenuated contractile responses to acetylcholine, histamine, 5-hydroxytryptamine and substance P — that is, to four independent agonists, which places its action downstream of all of them. It reduced contractions evoked by potassium depolarisation and calcium contractions in depolarising solution, and in whole-cell clamp it inhibited potential-dependent calcium currents concentration-dependently, reducing peak amplitude and accelerating current decay, in a manner resembling the dihydropyridine calcium antagonists. Their conclusion: peppermint oil relaxes gastrointestinal smooth muscle by reducing calcium influx (Hills and Aaronson, 1991). The active principle, menthol, is a simple monoterpenoid alcohol (National Center for Biotechnology Information, 2026a).

Chumpitazi, Kearns and Shulman's review adds the other candidate mechanisms and the anatomical reach: smooth-muscle relaxation via calcium blockade or direct enteric nervous system effects, visceral-sensitivity modulation via transient receptor potential cation channels, antimicrobial and anti-inflammatory activity, and modulation of psychosocial distress. Peppermint oil affects oesophageal, gastric, small-bowel, gallbladder and colonic physiology; it is used to facilitate completion of colonoscopy and endoscopic retrograde cholangiopancreatography; and placebo-controlled studies support its use in irritable bowel syndrome, functional dyspepsia, childhood functional abdominal pain and post-operative nausea (Chumpitazi et al., 2018). It is a registered herbal medicinal product in Europe (European Medicines Agency, 2026b).

The clinical result, stated with its own caveats. Ingrosso, Ianiro and colleagues updated their meta-analysis specifically because recent studies had cast doubt on peppermint oil's role, and identified 10 randomised trials in 1,030 patients. Peppermint oil was more efficacious than placebo for global symptoms (relative risk of not improving 0.65, 95% CI 0.43 to 0.98, number needed to treat 4, 95% CI 2.5 to 71) and for abdominal pain (0.76, 0.62 to 0.93, number needed to treat 7, 4 to 24). Adverse events were significantly more frequent than placebo (1.57, 1.04 to 2.37). Their conclusion is that peppermint oil was superior to placebo, adverse events were more frequent, the quality of evidence was very low, and adequately powered trials of peppermint oil as first-line treatment are needed (Ingrosso et al., 2022).

A number needed to treat whose confidence interval runs from 2.5 to 71 is not a precise quantity. Grade: strongly supported that peppermint oil beats placebo for global symptoms; emerging as to how much.

The adverse effect is the mechanism. A calcium-channel-blocking smooth-muscle relaxant delivered orally relaxes the lower oesophageal sphincter along with everything else, which is why reflux is the characteristic complaint and why the trial formulations are enteric-coated capsules rather than oil (Hills and Aaronson, 1991; Chumpitazi et al., 2018; Ingrosso et al., 2022). Efficacy and adverse effect are the same pharmacology; a formulation that eliminated one would eliminate the other.

The fixed combination. Madisch, Frieling and colleagues reviewed Menthacarin, a proprietary peppermint-and-caraway-oil combination, finding five randomised trials in 580 patients and pooling three functional-dyspepsia trials with 249 patients: pain intensity fell with a standardised mean difference of 0.80 (95% CI 0.39 to 1.21) and the Clinical Global Impression change item favoured treatment with a risk ratio of 2.65 (1.81 to 3.87) (Madisch et al., 2023). Grade: emerging for the specific preparation, which is what was tested.

16 Ginger: three indications, three different answers

Ginger is the clearest demonstration here that a mechanism which moves is not a symptom which resolves, because the demonstration happened inside a single experiment.

The pharmacology sets a low ceiling. Songvut, Nakareangrit and colleagues characterised the interconversion kinetics of the major constituents in rats after single intravenous and oral doses. [6]-Gingerol (National Center for Biotechnology Information, 2026b), [6]-shogaol and zingerone were rapidly but partially absorbed, extensively distributed and heavily biotransformed, limiting oral bioavailability of each pure compound to below 2%; gingerol converted to shogaol and to zingerone, with irreversible metabolic clearance significantly greater than the reversible interconversions, and glucuronide conjugates eliminated largely renally (Songvut et al., 2024). Rat data, single dose, pure compounds — but it constrains what a systemic mechanism can be asked to explain, and it is the same constraint that applies to any luminally active agent whose claimed target is the gut wall itself.

Gastric emptying: the mechanism moved. Hu, Rayner and colleagues studied 11 patients with functional dyspepsia twice in randomised double-blind fashion. After an eight-hour fast they took three capsules containing 1.2 g of ginger or placebo, then 500 mL of low-nutrient soup an hour later, with ultrasound measurement of antral area, fundus dimensions and antral contraction frequency, visual-analogue symptom scores, and plasma glucagon-like peptide-1, motilin and ghrelin. Gastric emptying was faster on ginger: median half-emptying time 12.3 minutes (range 8.5 to 17.0) against 16.1 (8.3 to 22.6), p ≤ 0.05, with a trend toward more antral contractions (p = 0.06). Fundus dimensions did not differ. Gastrointestinal symptoms did not differ. None of the three gut peptides differed (Hu et al., 2011).

Eleven patients, one experiment, one visit each way. The stomach emptied a quarter faster and the symptoms did not move. This is the reference case for the whole article: if a product's claim is that it accelerates emptying, this trial supports the claim and refutes the implication.

Nausea: a different endpoint, a real answer. Viljoen, Visser, Koen and Musekiwa meta-analysed 12 randomised trials in 1,278 pregnant women. Ginger significantly improved nausea against placebo (mean difference 1.20, 95% CI 0.56 to 1.84, p = 0.0002, I² = 0%) and did not significantly reduce vomiting episodes (0.72, −0.03 to 1.46, p = 0.06, I² = 71%), with subgroup analyses favouring daily doses below 1,500 mg for nausea relief. It posed no significant risk of spontaneous abortion against placebo (relative risk 3.14, 0.65 to 15.11, p = 0.15, I² = 0%) or against vitamin B6 (0.49, 0.17 to 1.42), and no significant risk of heartburn or drowsiness. The authors describe potential benefit, bearing in mind the limited number of studies, variable outcome reporting and low quality of evidence (Viljoen et al., 2014).

Two features are worth naming. Zero heterogeneity on the nausea endpoint across 12 trials is unusual here and argues for a real effect. And the lower dose subgroup did better, which is the opposite of the dose-response assumption on which supplement labelling is built. Grade: strongly supported for nausea in pregnancy; emerging for vomiting episodes, where the pooled estimate did not reach significance.

Irritable bowel syndrome: placebo won. van Tilburg, Palsson, Ringel and Whitehead randomised 45 patients to placebo, 1 g of ginger, or 2 g of ginger daily for 28 days, with a responder defined as at least a 25% reduction in the Irritable Bowel Syndrome Severity Scale. Responders: 57.1% on placebo, 46.7% on 1 g, 33.3% on 2 g. Adequate relief was reported by 53.3% on placebo and 53.3% across both ginger groups combined. Side effects were mild and reported by 35.7% on placebo and 16.7% on ginger. Their conclusion is that ginger was well tolerated but did not perform better than placebo (van Tilburg et al., 2014).

Placebo beat both doses, with an inverse dose trend, in a trial the authors correctly label a pilot. This is also the number to keep beside section 21: a 57.1% placebo response rate is what an irritable-bowel trial ordinarily produces.

In combination. Giacosa, Guido and colleagues randomised 126 patients with functional dyspepsia to a ginger and artichoke-leaf extract supplement or placebo, two capsules daily before lunch and dinner, for four weeks. The supplement group improved significantly by day 14 while placebo did not, and the advantage persisted to the end of the study, with significant item-level advantages for nausea, epigastric fullness, epigastric pain and bloating (Giacosa et al., 2015). Ginger is a registered herbal medicinal product in Europe (European Medicines Agency, 2026c). Grade for the combination: emerging, in the specific preparation tested, and the artichoke component has its own single-agent trial in section 17.

17 Fixed herbal combinations, artichoke, and the celandine problem

The best combination evidence here belongs to a licensed multi-plant pharmaceutical, and so does the most serious documented harm.

STW 5. Melzer, Rösch, Reichling and colleagues identified STW 5 as the descriptor in six randomised controlled trials, reanalysed and pooled the raw data of the three placebo-controlled studies meeting selection criteria, and used a fourth reference-controlled study for safety, covering 199 treated and 198 control patients. Pooled, the preparation was more effective than placebo on the severity of the most bothersome gastrointestinal symptom (odds ratio 0.22, 95% CI 0.11 to 0.47, p = 0.001); a fourth trial showed no significant difference from the prokinetic cisapride; adverse events were similar to placebo with no serious events (Melzer et al., 2004). Grade: strongly supported for the preparation, in functional dyspepsia.

Aguilar, Alcala-Gonzalez and colleagues then went after the mechanism with a physiology experiment rather than a symptom diary. Thirty-two patients with functional dyspepsia and bloating underwent a gas challenge — continuous gastric gas infusion at 100 mL/min for 15 minutes with simultaneous nutrient perfusion — after two weeks of the reformulated preparation or placebo, with gas evacuation, symptom perception and abdominal distension registered for 90 minutes. Rectal gas evacuation was significantly accelerated on treatment at 20 minutes from infusion start (319 ± 81 mL against 80 ± 39 mL, p = 0.015), but that difference declined and was gone by 90 minutes (final retention 470 ± 160 against 662 ± 179, p = 0.431). Symptom perception, mainly bloating, rose less on treatment (increment 0.8 ± 0.4 against 2.1 ± 0.4, p = 0.036). Abdominal distension did not differ between groups (Aguilar et al., 2025).

Read that against section 05. The sensation changed and the sign did not — exactly what a model in which bloating and distension are different variables predicts. And the transit advantage was a difference in timing rather than in total.

And now the hazard. Leroy, Perrin and colleagues report a 32-year-old woman with no previous medical history presenting with abdominal pain, jaundice and pruritus, whose blood tests showed acute severe hepatitis with a marked rise in direct bilirubin. Other aetiologies of acute liver injury were excluded and the diagnosis retained was drug-induced liver injury associated with the preparation; symptoms resolved spontaneously five weeks after stopping it. The authors note that the formulation comprises nine medicinal plant extracts including greater celandine, that safety and tolerability were extensively evaluated in randomised placebo-controlled studies with only rare and usually mild adverse symptoms reported, and that occasional cases of drug-induced liver injury have nonetheless been documented (Leroy et al., 2022). Greater celandine appears by name in the standard review of herbal hepatotoxicity, alongside kava, green tea, black cohosh, chaparral, germander and pyrrolizidine alkaloids (Bunchorntavakul and Reddy, 2013).

This is the cleanest available illustration of a general point about blends. A nine-component preparation has a nine-component hazard profile, one component was already known to injure the liver, and the aggregate trial programme — which was substantial and positive — was not powered to detect a rare idiosyncratic injury. A blend is not a safety average of its parts. It is the union of their risks.

Artichoke alone. Holtmann, Adam and colleagues randomised 247 patients with functional dyspepsia to a commercial artichoke leaf extract at 2 × 320 mg three times daily or placebo for six weeks, analysing 244 by intention to treat. The primary endpoint, the summed weekly rating of overall change in dyspeptic symptoms, favoured the extract (8.3 ± 4.6 against 6.7 ± 4.8, p < 0.01), as did the Nepean Dyspepsia Index quality-of-life score (−41.1 ± 47.6 against −24.8 ± 35.6, p < 0.01) (Holtmann et al., 2003). It is a registered herbal medicinal product in Europe (European Medicines Agency, 2026d). Grade: emerging — one adequately sized positive trial, on a subjective primary endpoint, and see section 18 on what happens to unreplicated botanical positives.

18 Aloe, turmeric, and a literature that does not replicate

Aloe is the most instructive negative here, because it was positive first.

Hong, Chun and colleagues meta-analysed randomised trials of Aloe vera against placebo in irritable bowel syndrome and found three, totalling 151 patients. The improvement in symptom score favoured aloe (standardised mean difference 0.41, 95% CI 0.07 to 0.75, p = 0.02); by intention-to-treat, response rates favoured aloe (pooled risk ratio 1.69, 1.05 to 2.73, p = 0.03); heterogeneity was absent (p = 0.90, I² = 0%); and no aloe-related adverse events were found in the included studies. The paper's title states the conclusion: aloe is effective and safe in short-term treatment of irritable bowel syndrome (Hong et al., 2018).

Ahluwalia, Magnusson, Simrén and colleagues then ran a single trial larger than that entire meta-analysis. One hundred and sixty patients meeting Rome III criteria, all subtypes, received an Aloe barbadensis extract or an inulin control for four weeks, with severity scored and faecal samples taken for microbiota and metabolomic profiling. Overall severity fell in both arms (p < 0.001 within each) with no difference between them (p = 0.62). Responders numbered 33 of 85 on aloe, 39%, and 34 of 75 on control, 45% (p = 0.49). Faecal microbiota and metabolite profiles did, however, differ between aloe responders and non-responders both before and after treatment, which the authors offer as a possible predictive signal in subsets (Ahluwalia et al., 2020).

Nothing was miscalculated in the meta-analysis. A conclusion was stated at a confidence that 151 patients across three trials could not carry, and a properly controlled trial of 160 patients superseded it. The correct handling is to report the later, larger, active-comparator trial as governing.

Hawrelak, Wohlmuth, Pattinson and colleagues supply the structural fact that should govern how every botanical result here is read. Their systematic review of double-blind placebo-controlled trials of Western herbal medicines in irritable bowel syndrome identified 33 trials evaluating 18 different preparations — 17 of the 33 evaluating peppermint essential oil. Meta-analyses supported peppermint essential oil as efficacious and well tolerated in short-term management, and supported aloe and asafoetida for global symptoms. Then the sentence that matters: with the exception of peppermint essential oil, aloe and asafoetida, none of the positive trials have been replicated, and this lack of replication limits the capacity to make definitive statements of efficacy (Hawrelak et al., 2020).

Eighteen preparations, three with any replication, and half the trial literature is about one oil. That is the shape of the field.

Turmeric. Thavorn, Wolfe and colleagues screened 1,136 citations and retained 26 eligible studies of turmeric across inflammatory bowel disease, irritable bowel syndrome, dyspepsia, reflux disease and peptic ulcer. Most studies were assessed at high risk of bias and many had methodological limitations; meta-analysis was not conducted because heterogeneity was too high. Descriptively, turmeric appeared safe with possible efficacy in inflammatory bowel disease or irritable bowel syndrome and inconsistent effects for other conditions, and the authors conclude that efficacy remains unclear (Thavorn et al., 2024). Grade: emerging at best, and see section 25: turmeric is simultaneously the most-consumed botanical in the United States among six with documented liver liability (Likhitsup et al., 2024).

19 The demulcents: marshmallow root and slippery elm

This is the honest floor of the article, and it is treated at length rather than omitted because both agents are common in digestive formulations and neither has the evidence its shelf position implies.

Marshmallow root. Althaea officinalis root has a registered entry as a herbal medicinal product in Europe (European Medicines Agency, 2026e). That registration category matters and is routinely misread: a traditional-use registration rests on plausibility and a long record of use at a stated preparation and dose, explicitly not on efficacy demonstrated in controlled trials. The proposed mechanism is physical and entirely reasonable — high-molecular-weight mucilage polysaccharides hydrate into a viscous film that could plausibly act as a demulcent on an irritated mucosal surface. It is testable in principle. It has essentially not been tested in the digestive tract.

The nearest controlled pharmacology available is Hage-Sleiman, Mroueh and Daher's evaluation of an aqueous extract in rats — and the details of what they actually studied are the point. The extract was of the flower, not the mucilage-rich root, which the authors state explicitly as their reason for the study. Over a month of intake in drinking water, 50 mg/kg raised serum HDL cholesterol with no effect on liver enzymes; anti-inflammatory activity against carrageenan- and formalin-induced inflammation and antiulcerogenic activity against ethanol-induced gastric ulcer were significant at 50, 100 and 250 mg/kg; platelet aggregation was inhibited in vitro at 500 µg/mL. And at 500 mg/kg the extract caused a significant decrease in stool water content (Hage-Sleiman et al., 2011).

Animal model, wrong plant part, a gastric-ulcer model rather than a digestive-comfort endpoint, and one finding — drier stool — that runs against the way demulcent products are positioned. Grade: plausible on mechanism, with no controlled human digestive evidence.

Slippery elm. Ulmus rubra is worse served still. Ahuja and Ahuja's critical appraisal of popular remedies for oesophageal symptoms covers dietary manipulation, apple cider vinegar, melatonin, acupuncture and herbal products including slippery elm, licorice, the peppermint oil already discussed and the STW 5 preparation of section 17, and concludes that a substantial gap persists between anecdotal and empirical understanding of the majority of non-pharmacologic remedies for oesophageal symptoms, while noting that the landscape raises interesting mechanistic hypotheses and opportunities for research (Ahuja and Ahuja, 2019).

Both agents are therefore handled here as short evidence dossiers folded into a system article rather than promoted to standalone titles with nothing to report — which is the disposition the gap map records for them, and the reason it does.

Agent	Endpoint that moved	Best evidence	Grade	Where it failed or stops
Pancrelipase	steatorrhoea; nutritional status	25 studies, 3,818 patients; society guidelines	established	needs a diagnosis; under-dosing fixes symptoms only
Psyllium	stool water, output, frequency	170-patient active-comparator trial; 16-trial meta-analysis	established	>10 g/day needed; flatulence rises; matched by kiwifruit
Peppermint oil	global IBS symptoms; pain	10 trials, 1,030 patients; ranked first of five classes	strongly supported	NNT 95% CI 2.5 to 71; certainty very low; reflux
Lactase	breath hydrogen; symptom score	60-patient three-arm randomised trial	strongly supported	null for infant colic across 5 trials
Ginger	nausea in pregnancy	12 trials, 1,278 women, I² = 0%	strongly supported	no effect on FD symptoms; placebo won in IBS
STW 5	most bothersome FD symptom	3 pooled placebo-controlled trials	strongly supported	documented acute liver injury; distension unchanged
Artichoke leaf	dyspepsia symptom sum; quality of life	247-patient trial	emerging	single positive trial, subjective endpoint
Zinc carnosine	indomethacin-induced permeability	10-volunteer crossover with provocation	emerging	no symptom endpoint; unreplicated since 2007
Glutamine	IBS-SS and permeability, PI-IBS	106-completer trial; 10-trial meta-analysis null	emerging	implausible placebo rate; no confirmatory trial
Alpha-galactosidase	breath hydrogen; flatulence	8 volunteers, one bean meal	emerging	never scaled; opposes the prebiotic rationale
Acid-resistant lipase	postprandial fullness	16 volunteers, one liquid meal	emerging	bloating, nausea, electrogastrography unchanged
Turmeric	none consistently	26 studies, unpoolable, high risk of bias	emerging	efficacy unclear; liver liability at population scale
Aloe vera	IBS symptom score	3-trial meta-analysis, then a 160-patient null	superseded	larger active-comparator trial found no difference
Marshmallow root	gastric ulcer area, in rats	flower extract, animal model	plausible	wrong plant part; reduced stool water content
Slippery elm	none	narrative appraisal only	plausible	substantial gap between anecdote and evidence

Table 2. Agent against endpoint. The grade column is about the strength of the demonstration, not the size of the effect, and the right-hand column is where each agent stops. Only three rows describe an endpoint above the symptom rung of Figure 1.

20 What "leaky gut" is, and what the term is being asked to carry

The consumer construct and the research construct share a name and almost nothing else.

The research construct is precise. Camilleri's reviews describe the intestinal barrier as a mucosal barrier of considerable complexity, with the epithelial layer only one of several defences, functioning as a critical interface between what the body absorbs and what it excludes. Barrier dysfunction is documented in inflammatory bowel disease and coeliac disease, in irritable bowel syndrome, and in association with non-steroidal anti-inflammatory drugs, alcohol, stress and enteric infection. Assessment requires measurements beyond the epithelial layer and beyond tight-junction expression, and is most meaningfully performed in vivo with orally administered probe molecules (Camilleri, 2019; Camilleri, 2021).

Then come the two sentences that should be printed on every product in this part of the aisle. No disease has been cured by normalising intestinal barrier function. And it remains unproven that restoring barrier function ameliorates clinical manifestations at all (Camilleri, 2019). This is not a sceptic writing from outside the field. It is the author of its standard review.

Rath, Atreya and Neurath, working in inflammatory bowel disease where the barrier defect is best characterised, hold both halves honestly: barrier impairment is a well-documented feature of the disease and a plausible therapeutic target, novel technologies could make barrier healing clinically assessable, and clinical assessment of the barrier remains far away from standard practice (Rath et al., 2023). A target can be real and simultaneously unavailable as a clinical endpoint. That is the actual state of the field.

Three consequences follow for reading any barrier product.

First, the provocation is part of the finding. Almost every positive human barrier result here was obtained by administering something that damages the barrier — indomethacin, running in heat — and showing that the agent blunted the resulting rise. That is a real experiment. It is not a demonstration that the agent lowers permeability in a person who has not been provoked, and it says nothing about how such a person feels.

Second, the marker is not the symptom. Section 07 established that two laboratories agree on the lactulose:mannitol ratio at ρ = 0.43 (Lee et al., 2014). A product cannot be titrated against a number that unstable, and no consumer has access to it.

Third, an in-vitro barrier result is a cell-biology result. Migration and proliferation in a wounded monolayer are informative about mechanism and silent about whether an oral dose reaches an epithelium at the necessary concentration.

21 Glutamine, and a placebo arm that cannot be right

Glutamine is the most-sold barrier ingredient and the site of the largest internal contradiction in the digestive literature. It is a conditionally essential amino acid and the principal respiratory fuel of the enterocyte (National Center for Biotechnology Information, 2026c), so the mechanism is not in question. What the trials show is.

The trial everyone cites. Zhou, Verne and colleagues randomised adults who had developed diarrhoea-predominant irritable bowel syndrome with increased intestinal permeability following an enteric infection to glutamine 5 g three times daily or placebo for eight weeks; 54 and 52 completed. The primary endpoint — a reduction of at least 50 points on the Irritable Bowel Syndrome Severity Scoring System — occurred in 79.6% on glutamine and 5.8% on placebo, a fourteen-fold difference. Every secondary endpoint moved: severity score 301 to 181, daily bowel movements 5.4 to 2.9, Bristol stool form 6.5 to 3.9, permeability ratio 0.11 to 0.05, all p < 0.0001. Hyperpermeability normalised in the glutamine group and not the control group. Adverse events and discontinuations were low and similar, with no serious events. The authors' own conclusion calls for large randomised trials to validate the findings (Zhou et al., 2019).

The number to interrogate is not 79.6%. It is 5.8%. Placebo response rates in irritable bowel syndrome trials sit in the range of 30 to 50% — section 16's ginger pilot recorded 57.1% (van Tilburg et al., 2014), the fibre meta-analysis recorded 41% (van der Schoot et al., 2022), the aloe trial 45% (Ahluwalia et al., 2020). A placebo arm in which 3 of 52 patients improved is an order of magnitude below what this condition ordinarily produces in a blinded trial, and any explanation for a fourteen-fold treatment difference has to account for it. The finding may be real; a single-centre result with an anomalous comparator arm and an explicit call for validation is not a settled one. Grade: emerging, and it has stood unreplicated at that scale for seven years.

The pooled evidence on the mechanism itself is null. Abbasi and colleagues systematically reviewed clinical trials of glutamine supplementation on intestinal permeability and found 10 studies from 1998 to 2014 with 352 participants of varying health status. Overall, glutamine did not significantly affect intestinal permeability (weighted mean difference −0.00, 95% CI −0.04 to 0.03). A high-dose subgroup showed a reduction, on which the authors themselves call for further investigation — and the reported subgroup figures are not internally consistent, since the published abstract gives the threshold in grams in one sentence and in milligrams in another, and quotes a confidence interval that does not contain its own point estimate. No dose threshold is carried forward from it here (Abbasi et al., 2024).

Ten trials found no overall effect on the marker that is the mechanism of the claim. That is the single most important sentence in the barrier chapter, and it sits beside the trial that reported a fourteen-fold symptom benefit.

Where the marker does move, the symptom does not. Pugh, Sage and colleagues had ten recreationally active men complete four trials — placebo and glutamine at 0.25, 0.5 and 0.9 g per kg of fat-free mass, two hours before a 60-minute treadmill run at 30 °C. The serum lactulose:rhamnose ratio was likely lower at 0.25 and 0.5 g/kg and very likely lower at 0.9 g/kg, dose-dependently. Gastrointestinal symptoms were typically low and there was no effect of supplementation on them; the authors state it remains unclear whether the permeability change will lead to reductions in symptoms (Pugh et al., 2017).

This is the cleanest instance in the article of a marker responding in a dose-dependent, mechanistically coherent way while the thing a person would notice does not. And Akobeng and colleagues' earlier study of glutamine supplementation and intestinal permeability in Crohn's disease sits in the same tradition of testing the marker directly rather than a symptom (Akobeng et al., 2000).

The honest summary: the mechanism is real biochemistry, the marker moves against a provocation and does not move on pooled analysis, one trial reports a spectacular symptom benefit against an implausible placebo arm, and its own authors asked for validation that has not arrived.

22 Zinc carnosine, and NSAID-provoked injury as a model system

Zinc carnosine — polaprezinc, a zinc–L-carnosine chelate (National Center for Biotechnology Information, 2026d) — has the best-designed single barrier experiment here, and the design is precisely what limits the conclusion.

Mahmood, FitzGerald and colleagues ran three layers of experiment. In vitro, zinc carnosine stimulated migration and proliferation dose-dependently, with maximal effects at 100 µmol/L in HT29 cells and an approximately threefold increase in both. In rodents it reduced gastric damage by 75% and small-intestinal villus shortening by 50%, both p < 0.01. Then the human layer: ten healthy volunteers in a randomised crossover, given indomethacin 50 mg three times daily for five days with either zinc carnosine 37.5 mg twice daily or placebo. Indomethacin tripled gut permeability in the control arm — lactulose:rhamnose 0.35 ± 0.035 before, 0.88 ± 0.11 after, p < 0.01 — while no significant increase in permeability was seen when zinc carnosine was co-administered. The authors conclude that it stabilises gut mucosa at concentrations likely to be found in the gut lumen, and that further studies are warranted (Mahmood et al., 2007).

The provocation is doing the work, and that is a strength of the design and a limit on the claim. Zinc carnosine prevented a drug-induced rise in a permeability marker in ten healthy people over five days. There was no symptom endpoint, because healthy volunteers on a short indomethacin course have no digestive symptom to relieve. Nineteen years later the study remains unreplicated at scale. Grade: emerging, for prevention of NSAID-provoked hyperpermeability.

Hwang and colleagues place it in the wider field with a systematic review of natural products, functional foods and probiotics for non-steroidal anti-inflammatory drug-induced small intestinal injury, searched to January 2026 and restricted to randomised studies assessed by capsule endoscopy or permeability testing. Twenty-two studies were included, 21 randomised and one quasi-randomised. Geranylgeranylacetone showed protection in three of four capsule-endoscopy studies; lactoferrin, zinc carnosine and fish protein hydrolysate reduced indomethacin-induced hyperpermeability; probiotic effects were outcome-dependent, with more consistent benefit on capsule-endoscopy mucosal injury than on permeability. Three trials were at low risk of bias, 15 raised some concerns, three were at high risk. Heterogeneity prevented meta-analysis. The authors' conclusion is that the certainty of evidence was generally low or very low and the findings should be read as hypothesis-generating (Hwang et al., 2026).

Two structural observations follow. The barrier literature's positive results cluster in a population defined by taking a drug that damages the barrier — which is a coherent clinical question and not the population the products are sold to. And in the one place where two endpoint families were compared directly, they disagreed: probiotics looked better on visualised mucosal injury than on permeability, which means the choice of endpoint is choosing the answer.

Zinc's systemic nutritional role is a separate matter with its own established requirements and upper limits (National Institutes of Health Office of Dietary Supplements, 2026), and the deep reference is the Zinc article. Nothing in the barrier literature establishes that a person with adequate zinc status and no NSAID exposure has a permeability problem for polaprezinc to prevent.

23 Simethicone, and what an antifoaming agent can and cannot do

Simethicone is the oldest product in this category and the one whose mechanism is least mysterious: it lowers surface tension so gas bubbles coalesce and can be passed. Its evidence is instructive because the strongest data come from a setting where the outcome is objective and measured by someone other than the patient.

Moolla, Poonja and colleagues meta-analysed 16 randomised trials in 5,630 patients undergoing bowel preparation before colonoscopy. Polyethylene glycol with simethicone improved colon cleansing over polyethylene glycol alone (odds ratio 1.48, 95% CI 1.11 to 1.97, p = 0.008), an effect present for single dosing (1.83, 1.20 to 2.79) and not for split dosing (1.32, 0.72 to 2.43, p = 0.38). Adenoma detection was unaffected overall (1.22, 0.81 to 1.83) but improved with single dosing (1.96, 1.22 to 3.16). For bloating, the polyethylene-glycol-alone group had 2.33 times the odds of experiencing it (1.70 to 3.20, p < 0.00001), while nausea, vomiting and abdominal pain did not differ (Moolla et al., 2019). Grade: established, for bloating during bowel preparation, in a population with a very large and identifiable gas load.

Note the conditionality. The benefit disappeared with split-dose preparation, which is the modern standard. An adjunct that helps when the primary intervention is done the older way is a real finding about a specific protocol, and it is a long way from a general claim about abdominal comfort.

Petrisor, Iancu and colleagues then used simethicone as the active comparator in functional abdominal bloating and distension — a Rome-defined condition rather than a procedure. Eighty-eight patients were randomised to a xyloglucan-and-pea-protein product or simethicone, three times daily for 20 days, with safety as the primary outcome. No adverse events or serious unexpected adverse reactions occurred in either arm. The test product showed faster onset and significant reductions in bloating and abdominal pain against simethicone, and at day 20 reduced abdominal girth by an average of 4.7 cm against 1.8 cm, with a significant fall in breath hydrogen from baseline (Petrisor et al., 2024).

Two cautions belong with that result. The trial was powered for safety, with efficacy as a secondary outcome, so the comparison is exploratory by the investigators' own design. And a mucosal film-forming agent outperforming an antifoaming agent on girth is consistent with section 05: if a substantial fraction of visible distension is abdominophrenic dyssynergia rather than gas volume, then a product whose entire mechanism is gas coalescence is aimed at the wrong variable, and being beaten is what one would predict. Grade: emerging for the combination product, in the specific 20-day protocol tested.

24 The boundary with the microbiome, drawn on purpose

Two findings mark the edge of this review and belong to its companion.

Wauters, Slaets and colleagues randomised 68 patients with Rome IV functional dyspepsia to eight weeks of Bacillus coagulans MY01 and Bacillus subtilis MY02 or placebo, twice daily, stratified by proton-pump-inhibitor status, with a decrease of at least 0.7 in the Leuven Postprandial Distress Scale as the primary endpoint. Clinical responders numbered 12 of 25, 48%, on probiotics against 6 of 30, 20%, on placebo (relative risk 1.95, 95% CI 1.07 to 4.11, p = 0.028). Adverse events were similar. The authors label the study exploratory and a pilot, and describe accompanying immune and microbial changes as possible future predictors or targets (Wauters et al., 2021). Grade: emerging, for two named strains, in one condition, in 68 patients at one centre.

That result should be read directly against section 05's negative recommendation. The American Gastroenterological Association's expert review states that probiotics should not be used to treat abdominal bloating and distension (Moshiree et al., 2023). Both statements can be true, because they concern different conditions, different strains and different endpoints — which is exactly the point. Strain, condition and endpoint are not interchangeable, and a product that lists a genus is not the intervention that was tested.

The FODMAP literature marks the same boundary from the other side. A low-FODMAP diet ranked first for global symptoms, abdominal pain and bloating in a network meta-analysis of 13 trials, superior to standard dietetic advice for bloating specifically, with the authors noting that most trials were in secondary or tertiary care and did not study reintroduction and personalisation (Black et al., 2022b). The intervention works by removing fermentable substrate. The prebiotic aisle works by adding it. Both appear in the same shopping basket, and the article that adjudicates fermentation as a benefit is Microbiome-Nutrition.

25 Where the real risks are, and they are not where the warnings are

The safety profile of this category is unusual: the agents with the most human data are the least hazardous, and the risk concentrates in botanicals sold on the softest evidence.

Hepatotoxicity, at population scale. Likhitsup, Chen and Fontana analysed National Health and Nutrition Examination Survey data for adults enrolled between January 2017 and March 2020, covering 9,685 people. Overall herbal and dietary supplement use was 57.6%; use of six botanicals with documented liver liability — turmeric or curcumin, green tea extract, Garcinia cambogia, black cohosh, red yeast rice and ashwagandha — was 4.7%, with turmeric the most common by a wide margin. Users were older, more educated and more likely to have arthritis. An estimated 15.6 million United States adults had consumed at least one of the six within the past 30 days — comparable to the estimated number prescribed simvastatin (14.0 million) or non-steroidal anti-inflammatory drugs (14.8 million). The commonest stated reason for taking turmeric or green tea was to improve or maintain health (Likhitsup et al., 2024).

Turmeric appears in section 18 with 26 studies, mostly at high risk of bias, too heterogeneous to pool, and efficacy unclear (Thavorn et al., 2024). It is simultaneously the most widely consumed botanical with a documented liver signal in the United States. Weak efficacy evidence and population-scale exposure is the worst possible combination, and it is the actual situation.

Bunchorntavakul and Reddy's review supplies the clinical shape: incidence rates are largely unknown, presentation ranges from mild hepatitis to acute hepatic failure requiring transplantation, causality scoring systems for drug-induced liver injury have not been validated for herbal injury, and patients under-report use while physicians under-recognise its magnitude. Greater celandine appears by name in their list alongside kava, green tea, black cohosh, chaparral, germander and pyrrolizidine alkaloids (Bunchorntavakul and Reddy, 2013) — which is where section 17's case of acute liver injury from a nine-plant dyspepsia preparation containing celandine belongs (Leroy et al., 2022).

Interactions, stated precisely rather than dramatically. McEwen's review of herbal medicine effects on platelet function and coagulation finds reduced platelet aggregation reported for feverfew, garlic, ginger, ginseng, motherwort, St John's wort and willow bark, with in-vitro reductions for a longer list including turmeric; potential warfarin interactions for cranberry, danshen, dong quai, Ginkgo, ginseng, green tea and St John's wort; St John's wort with clopidogrel; and danshen with aspirin. The practical recommendation is repeat testing of platelet function and coagulation after excluding herbal products that could have affected initial results (McEwen, 2015).

Then the specific test. Jiang, Williams and colleagues gave 12 healthy men a single 25 mg dose of warfarin alone or after seven days of pretreatment with recommended doses of Ginkgo or ginger from products of known quality, continuing the herb for seven days afterwards, in an open-label three-way crossover with platelet aggregation, international normalised ratio, enantiomer protein binding, plasma enantiomer concentrations and urinary S-7-hydroxywarfarin all measured. Neither herb affected the international normalised ratio or platelet aggregation alone. Apparent clearance ratios for S- and R-warfarin were essentially unity with both herbs; the urinary S-7-hydroxywarfarin excretion ratio indicated no effect on CYP2C9; volumes of distribution and protein binding were unchanged. Conclusion: at recommended doses, neither affects clotting status or warfarin pharmacokinetics or pharmacodynamics in healthy subjects (Jiang et al., 2005).

That is how a class-level caution should be resolved — a designed interaction study, with a named product of known quality, at a stated dose, reporting a negative result with confidence intervals. Twelve healthy men on a single warfarin dose does not license unlimited co-administration in anticoagulated patients, and the class caution stands for the herbs McEwen names. But the specific fear that ordinary ginger destabilises warfarin was tested and not supported.

The mechanism-derived cautions. Peppermint oil relaxes smooth muscle by blocking calcium influx, which includes the lower oesophageal sphincter, and its trials report significantly more adverse events than placebo (Hills et al., 1991; Ingrosso et al., 2022). Fibre's dose-limiting adverse effect is flatulence, significantly greater than control across 16 trials (van der Schoot et al., 2022), and the over-the-counter review found diarrhoea, nausea, bloating and abdominal pain common across laxative classes with no serious events (Rao and Brenner, 2021). Bran caused enough symptom worsening to drive the highest dropout rate in a 275-patient trial (Bijkerk et al., 2009). Ginger carried no significant risk of spontaneous abortion, heartburn or drowsiness across 12 pregnancy trials (Viljoen et al., 2014). Adverse events on Helicobacter pylori eradication were more than twice as common as control — in the intervention with the best efficacy evidence here (Ford et al., 2022).

Product quality is a safety domain, not a consumer-affairs one. Two of nine marketed pancreatic enzyme batches failed dissolution, one releasing 8% of its lipase, in products whose label content all passed (Case et al., 2005). Under 21 CFR 101.93 a structure/function claim requires a disclaimer and no premarket demonstration (Office of the Federal Register, 2026), and the agency does not approve supplements before marketing (United States Food and Drug Administration, 2026b). Where a preparation is registered — ispaghula husk, peppermint oil, ginger rhizome, artichoke leaf, marshmallow root (European Medicines Agency, 2026a; 2026b; 2026c; 2026d; 2026e) — the registration fixes the preparation and the dose, which is a quality assurance and not an efficacy claim.

Domain	What the evidence actually says	Practical weight
Hepatotoxicity, turmeric	15.6 million US adults exposed in 30 days; liver signal documented; efficacy unclear across 26 studies	highest population-level concern here
Hepatotoxicity, multi-plant blends	acute liver injury from a nine-extract dyspepsia preparation containing greater celandine; resolved on withdrawal	a blend carries the union of its components' risks
Peppermint oil	adverse events significantly more frequent than placebo; reflux is the mechanism	inseparable from the efficacy
Fibre	flatulence significantly greater than control; bran drove the highest dropout in a 275-patient trial	dose-limiting, and it is the target symptom
Warfarin, ginger and Ginkgo	designed crossover: no effect on INR, aggregation, clearance or CYP2C9	specific fear not supported; class caution stands for other herbs
Antiplatelet class caution	aggregation reduced by feverfew, garlic, ginger, ginseng, St John's wort, willow bark; warfarin interactions for seven	retest coagulation after excluding herbal use
Enzyme product quality	2 of 9 batches failed dissolution; one released 8% of lipase; all passed content	why pancrelipase became a drug and supplements did not
Regulatory floor	structure/function claim needs only a disclaimer; no premarket approval	label content is not delivered activity
Self-diagnosis	response to a therapeutic enzyme trial is unreliable for diagnosing insufficiency	the category's core reasoning is explicitly rejected
Missed diagnosis	Rome criteria, four-hour emptying study, faecal elastase, NICE guidance define these conditions positively	a supplement trial is not a diagnostic test

Table 3. The safety ledger. The largest quantified risk here attaches to the botanical with the least clear efficacy, and the second largest attaches to the structure of blend products rather than to any single ingredient.

26 What the reader is actually buying

Six statements follow from everything above, and each is traceable to a specific study rather than to a general scepticism.

A digestive enzyme supplement is not a small pancreatic enzyme. Pancrelipase is a drug with an approved application and a dissolution specification, because pharmaceutical quality differences were found in marketed products and two of nine batches did not adequately dissolve (Case et al., 2005; United States Food and Drug Administration, 2026a). A supplement version carries the same USP units on the panel with no requirement to demonstrate that any of it emerges at the right pH.

Feeling better on enzymes is not evidence that enzymes were needed. A professional society states that response to a therapeutic trial is unreliable for diagnosing exocrine pancreatic insufficiency (Whitcomb et al., 2023). And relieving the symptom is not correcting the deficit: under-dosed enzymes relieved diarrhoea in nearly all real-world studies and improved nutritional status in none (Kadaj-Lipka et al., 2025).

"Fibre" is not a class with a shared effect. Laxation requires water-holding capacity and resistance to fermentation; psyllium was 3.4 times more effective than insoluble wheat bran for stool output; finely ground bran decreased stool water content; and inulin, fructo-oligosaccharide and polydextrose have insufficient evidence as laxatives (McRorie and McKeown, 2017; McRorie et al., 2020; Rao and Brenner, 2021). Response above control appeared only above 10 g/day, and flatulence rose significantly (van der Schoot et al., 2022). Two green kiwifruit matched 12 g of psyllium on the primary endpoint with fewer adverse events and fewer dissatisfied patients (Chey et al., 2021).

The best-evidenced botanical is a calcium-channel-blocking smooth-muscle relaxant, and its own meta-analysts call the evidence very low quality. Peppermint oil ranked first among five first-line intervention classes for irritable bowel syndrome, with no significant difference from any other active treatment, and beat placebo with a number needed to treat of 4 whose confidence interval runs to 71 (Black et al., 2020; Ingrosso et al., 2022).

Barrier products are sold on a construct whose leading reviewer says no disease has been cured by normalising it. Ten trials found no overall effect of glutamine on intestinal permeability; where permeability did fall dose-dependently, symptoms did not change; and the striking symptom trial reported a 5.8% placebo response rate in a condition that ordinarily produces 40 to 57% (Camilleri, 2019; Abbasi et al., 2024; Pugh et al., 2017; Zhou et al., 2019).

Bloating products are frequently aimed at the wrong variable. Girth can rise without a change in gas volume through abdominophrenic dyssynergia; gastric emptying studies should not be ordered routinely for bloating; and probiotics should not be used to treat bloating and distension (Moshiree et al., 2023; Malagelada et al., 2017; Melchior et al., 2025).

27 What would settle the open questions

Each of the following is a specific, feasible experiment, and in every case the reason it has not been done is not that it is impossible.

Replicate the glutamine trial with a credible placebo arm. A multicentre randomised trial in post-infectious diarrhoea-predominant irritable bowel syndrome with a pre-specified responder definition and a documented placebo response rate would resolve a fourteen-fold effect that has stood alone since 2019 (Zhou et al., 2019). Its own authors asked for this.

Give zinc carnosine a symptom endpoint in an unprovoked population. The existing experiment prevented an indomethacin-induced permeability rise in ten healthy volunteers with no symptom to relieve (Mahmood et al., 2007). The clinically useful question is whether it does anything for people with symptoms and no NSAID exposure.

Scale alpha-galactosidase. Eight healthy volunteers, one bean meal, and clear reductions in both breath hydrogen and flatulence (Di Stefano et al., 2007), unreplicated at size in nearly two decades, in the one product class whose mechanism is unambiguous.

Test enzyme supplements for dissolution and publish it. The method exists, was applied once to nine products, and produced the finding that reclassified the drug versions (Case et al., 2005). Repeating it in the supplement aisle is a laboratory exercise, not a trial.

Power a peppermint oil trial as a first-line treatment. Its meta-analysts state this explicitly, and a number needed to treat between 2.5 and 71 is the reason (Ingrosso et al., 2022).

Separate bloating from distension as co-primary endpoints. The STW 5 gas-challenge study measured both and they diverged — perception improved, distension did not (Aguilar et al., 2025). Any bloating trial that reports one composite score is discarding the most informative contrast available.

Replicate anything other than peppermint oil. Thirty-three trials, 18 preparations, 17 of them on one oil, and only three preparations replicated at all (Hawrelak et al., 2020). The field's problem is not a shortage of positive results. It is a shortage of second ones.

28 Cross-references

Ginger · Peppermint · Dietary-Fibre · Digestive-Enzymes · L-Glutamine · Zinc · Marshmallow-Root · Turmeric · Aloe-Vera — the single-subject deep references for the agents synthesised here.

Microbiome-Nutrition — probiotics, prebiotics, fermentable fibre, synbiotics, postbiotics, short-chain fatty acids, and the fermentation question this article hands over at section 24.

Immune-Resilience-and-Nutritional-Immunology — the mucosal immune compartment, layer three of Figure 3.

Stress-Anxiety-and-Adaptogenic-Nutrition — the gut–brain axis, visceral hypersensitivity, and the central neuromodulators named in the bloating literature.

Metabolic-Nutrition and Glycemic-Control-and-Metabolic-Supplements — viscous fibre's glycaemic and lipid effects, which arise from the same physics as its laxative effects but answer a different question.

Liver-Health-and-Hepatoprotective-Nutraceuticals — turmeric's liver signal at population scale, greater celandine, and the herbal hepatotoxicity file that section 25 opens and does not attempt to close.

Hydration-Electrolytes-and-Fluid-Balance — oral rehydration and glucose-assisted sodium absorption, which is the one place in this organ where diarrhoea has a decisive nutritional treatment, and the point at which this article stops rather than summarising it badly.

Where a cross-reference appears, the single-subject article is the deeper authority on chemistry, preparation, standardisation and dose-finding, and this article is the authority on how that agent performed against a digestive endpoint and against the comparator that endpoint deserves.

29 Standing constraint

30 References

31 Evidence handling

The numbered list above is generated at build from records retrieved from NCBI and from regulator, agency and chemical-database sources whose URLs were resolved and checked; nothing in it was typed from memory. Four candidate sources failed verification and were replaced rather than cited from recollection: two Food and Drug Administration pages, on pancreatic enzyme products and on oral sodium phosphate labelling, returned HTTP 404 and were replaced by the Drugs@FDA application record and the agency's dietary-supplement questions-and-answers page; a European Food Safety Authority opinion and its publisher mirror returned HTTP 403 and were replaced by the European Medicines Agency herbal article entries, which carry the preparation and dose information the argument needed; and one guideline that resolved correctly was dropped on scope rather than availability, being paediatric reflux.so that a later editor does not retry them.

Study types are named in the sentence that uses them. A hydrolysis rate in a buffer, a relaxation curve in a tissue bath, a patch-clamp recording, a wounded-monolayer migration assay, a rodent ulcer model, a rat interconversion pharmacokinetic study, a single-meal crossover in volunteers, a five-day provocation crossover with a permeability probe, a collected-stool trial with an active comparator, a four-week partially randomised food-versus-supplement comparison, a twelve-week primary-care trial, a network meta-analysis, a pooled reanalysis of raw trial data, a systematic review that declined to pool, a national dietetic guideline, a society clinical practice update, a laboratory assay of marketed products, a case report of liver injury, and a nationally representative exposure survey are not interchangeable, and the article does not let one stand in for another.

Where results conflict, both appear. Aloe's three-trial positive meta-analysis is reported in full and then superseded in the same section by the larger active-comparator trial that found no difference. Glutamine's fourteen-fold symptom result is reported beside the ten-trial meta-analysis that found no overall effect on the marker that is supposed to be its mechanism. Ginger's accelerated gastric emptying is reported together with the unchanged symptoms in the same eleven patients, and its pregnancy nausea benefit beside a pilot trial in irritable bowel syndrome that placebo won at both doses.

Where a synthesis reported its own certainty, that assessment is carried rather than paraphrased away. Peppermint oil is described as very low quality evidence because its meta-analysts describe it that way. The dietetic constipation guideline is reported as 8 moderate, 39 low and 12 very low quality statements out of 59. The non-steroidal anti-inflammatory drug barrier review is reported as hypothesis-generating because its authors say so. Where a confidence interval makes a point estimate uninformative — a number needed to treat of 4 with an interval running to 71 — the interval is given rather than the point.

Populations are named every time, because the central failure mode of this literature is transferring a result from a person with a measured deficit, a positive breath test, an induced barrier injury or a Rome-defined disorder to a reader who has none of those. Every reference resolves to a record retrieved from NCBI or to an agency or database page that returned HTTP 200 and the expected document title at build. Project 06 was consulted read-only as a general-science corpus and is not the evidence spine for this title; no write was made to it or to project06_catalog.sqlite. Marketing copy is not treated as scientific evidence anywhere in this document.

Adversarial interrogation — the required adversarial challenge list — is disposed as follows.

Which claims are overstated? ACCEPTED as the organising question. The barrier-repair claim, the enzyme-need claim inferred from feeling better, the class claim implied by the word "fibre", the gas claim made for bloating products, and the notion that "digestive support" names a quantity are each stated and refused in the section that reports their evidence.

What contrary trials exist? ACCEPTED and given priority. Ahluwalia and colleagues' 160-patient null is placed directly against Hong and colleagues' positive meta-analysis and treated as governing. Abbasi and colleagues' null on permeability is placed against Zhou and colleagues' symptom result. Van Tilburg and colleagues' placebo win is reported at both ginger doses. Hu and colleagues' unchanged symptoms are reported in the same paragraph as their positive emptying result. Kozłowska-Jalowska and colleagues' null in infants is reported beside Ojetti and colleagues' positive result in adults.

Are effect sizes clinically meaningful? ACCEPTED and answered structurally in section 06. Helicobacter pylori eradication, with a number needed to treat between 4.5 and 14 and a stated cost in adverse events, and a low-FODMAP diet that outperformed symptom-matched drug therapy, are the comparators against which every supplement result is read. Nothing that moved a marker is described here as clinical benefit.

Is the evidence population relevant? ACCEPTED as the article's spine. Section 04 separates replacement from luminal physics from mucosal repair on exactly this ground, and the right-hand column of Table 2 exists to make the population limit visible for every agent. Barrier results obtained under indomethacin or heat stress are labelled as provoked-population findings wherever they appear.

Is there publication bias? ACCEPTED as a structural expectation and asserted as a detected finding nowhere, because none of the syntheses cited here reported a funnel-plot analysis in its abstract. Hawrelak and colleagues' finding that only three of eighteen botanical preparations have been replicated at all is reported as the related and better-evidenced problem, and 17 of 33 trials concerning a single oil is reported as a distribution rather than a bias.

Are commercial claims stronger than the evidence? ACCEPTED. Figure 1 states the gap as a ladder, and section 09 states it as a regulatory asymmetry: the same enzyme units on a panel, with and without a dissolution requirement.

Are mechanisms being mistaken for outcomes? ACCEPTED, and treated as the article's central error. Ginger's faster emptying with unchanged symptoms, glutamine's dose-dependent marker fall with unchanged symptoms, and zinc carnosine's abolished permeability rise with no symptom endpoint at all are the three reference cases, and each is reported in the sentence that reports the mechanism.

Are safety concerns underrepresented? ACCEPTED, and derived from the efficacy trials rather than appended as a caveat. Peppermint oil's adverse-event excess is the same pharmacology as its benefit; fibre's dose-limiting flatulence is the symptom other products in the aisle are sold to treat; and the nine-plant preparation with the best combination evidence here is also the one with a documented case of acute liver injury.

Are doses and formulations genuinely comparable? ACCEPTED and treated as decisive four times: guideline-concordant against under-dosed lipase, above and below 10 g/day of fibre, coarse against finely milled wheat bran, and enteric-coated peppermint capsules against oil. Marshmallow root's only controlled pharmacology used the flower rather than the mucilage-bearing root, and that is stated rather than glossed.

Does any paragraph imply certainty not supported by evidence? ACCEPTED as a drafting constraint. Every efficacy statement carries an in-place grade, "superseded" is used where a larger trial replaced a positive meta-analysis rather than defaulting to "mixed", and "not studied" is used where the population that buys a product was never enrolled.

Unresolved science remains substantial, and section 27 sets out the seven experiments that would resolve most of it. The largest gaps are a replication of the glutamine trial with a credible placebo arm; a symptom endpoint for zinc carnosine in people who have not been given indomethacin; a dissolution survey of marketed enzyme supplements, which is a laboratory exercise rather than a trial; an adequately powered first-line trial of peppermint oil; a bloating trial that reports sensation and girth as co-primary endpoints rather than as one score; and a second positive trial of almost any botanical in this field other than peppermint oil.