White Kidney Bean Extract

Also sold as: Phaseolus vulgaris, Starch Blocker, Phase 2

A bean seed protein that genuinely inhibits pancreatic alpha-amylase. The open question is not whether the enzyme is inhibited — it is — but how much starch that actually keeps out, and where the rest of it goes.

How White Kidney Bean Extract actually works

The active is a seed glycoprotein, not a phytochemical: alpha-amylase inhibitor, roughly 50 kDa, which binds pancreatic alpha-amylase through a protein-protein interaction and occludes the active-site cleft. It is a relative of phytohemagglutinin from the same gene family and the same seed, which is why the extraction method is a safety specification and not just a manufacturing detail. Blocking amylase is the first step of starch digestion and not the last — brush-border maltase-glucoamylase and sucrase-isomaltase still work on whatever gets past — so the practical effect is a narrowed doorway rather than a wall, and starch that escapes is fermented in the colon.

Where to get White Kidney Bean Extract

The evidence for White Kidney Bean Extract

Graded by what exists behind each claim.

✅ Clinically validated

• Meta-analysis of 8 randomized trials (n=543): −1.62 kg body weight, −1.17 kg fat mass, −1.58 cm waist, no serious adverse events — and no significant change in fasting glucose, triglycerides or insulin.
• A 12-week randomized trial of a proprietary extract standardized by amylase-inhibiting activity, at 700 or 1,000 mg three times daily before meals, reported dose-dependent weight and fat loss. Every arm was also on a calorie-restricted diet.
• A 4-month randomized trial at 1.5 g/day with carbohydrate intake deliberately maintained lowered HbA1c by 0.72% and shifted the gut microbiota toward short-chain-fatty-acid producers.

📊 Correlative data

• The only study that ever measured whether a starch blocker blocks starch CALORIES used one-day fecal calorie balance on a 100 g starch meal in 1982: 80 kcal in the stool against 78 on placebo, where 400 kcal was predicted.

🧪 Theoretical / extrapolated benefits

• Inhibiting amylase may shift starch digestion distally rather than prevent it — colonic fermentation, gas, and short-chain fatty acids rather than blocked calories. That is a different claim from the one on the label and it has some human support.

How to read these tiers: they say how much human evidence exists, not how well something works — and ✗ flags harm, never a disappointing trial. How the evidence tiers work →

What White Kidney Bean Extract actually does

Unusually for a botanical, the active ingredient here is a protein, and knowing that answers most of the questions the page raises. Phaseolus vulgaris seeds carry alpha-amylase inhibitor, a glycoprotein of roughly 50 kDa. It is not a polyphenol, not an alkaloid and not a fiber. It belongs to the same seed protein family as phytohemagglutinin and arcelin -- the same gene cluster, the same tissue, three different jobs -- and that shared ancestry is the reason the toxicity question and the efficacy question on this page cannot be separated.

The inhibition is a protein-protein interaction, not a substrate-analog block. The inhibitor binds human pancreatic alpha-amylase and occludes the active-site cleft, which is why it is effective at low molar ratios and why the reviews describe it specifically as reducing enzyme activity through protein-protein interaction rather than through competitive inhibition Peddio 2022. Two isoforms exist and they are not interchangeable: alpha-AI1 inhibits mammalian amylases, alpha-AI2 inhibits certain insect amylases and does essentially nothing to ours. A bean can carry either, both or neither.

Which enzyme is being blocked matters, because it is the first step and not the last. Alpha-amylase hydrolyzes internal alpha-1,4 bonds in starch and produces maltose, maltotriose and alpha-limit dextrins. Those fragments are then finished at the brush border by maltase-glucoamylase and sucrase-isomaltase into free glucose, which SGLT1 carries across. Inhibiting amylase therefore slows the entry to the pathway; it does not disable the pathway. Anything that escapes into the more distal small intestine still meets a full complement of brush-border enzymes and a long absorptive surface. This is the structural reason a starch blocker behaves less like a wall and more like a narrowed doorway.

The pharmacokinetic question for a protein is a survival question, and it is the one the product is actually built around. The inhibitor has to cross the stomach -- pH 1.5 to 3.5, pepsin active -- and arrive folded in the duodenum, because an unfolded protein inhibits nothing. That is the entire reason the dosing instruction is 30 minutes before a meal Jäger 2024: the meal buffers gastric acid and shortens the exposure, and the inhibitor rides in with the food it is meant to act on. There is no plasma concentration to measure and no half-life in the ordinary sense, because systemic absorption is not the goal and would be a problem if it happened; the relevant exposure is intraluminal, it is finished when gastric emptying is finished, and it is gone with the next bowel movement.

Then the part that is usually left out: the starch does not disappear. Starch that escapes amylase in the small intestine reaches the colon and becomes, by circumstance rather than by structure, resistant starch. Colonic bacteria ferment it to short-chain fatty acids -- acetate, propionate, butyrate -- plus hydrogen, carbon dioxide and, in some people, methane. That is where the gas comes from, and it is also a real metabolic signal rather than a side effect: butyrate is the preferred fuel of the colonocyte and propionate reaches the liver through the portal vein. A four-month randomized trial that deliberately kept carbohydrate intake up while giving the extract found Bifidobacterium, Fecalibacterium and Anaerostipes higher in the treated group and attributed the metabolic improvement to the enrichment of short-chain-fatty-acid producers Feng 2022. So the honest mechanistic statement is not that the calories are blocked. It is that some of them are handed to the microbiome instead of to the small intestine, and the two are not energetically equivalent but they are not zero and 100 either.

Cell, rodent, human — and where it stops

The measurement that should have settled this was made in 1982, and it is a better study than most of what came after it. Subjects ate a high-starch meal of 100 g of starch -- spaghetti, tomato sauce and bread -- with either placebo or starch-blocker tablets, and fecal calorie excretion was measured by a one-day calorie balance technique Bo-Linn 1982. If the tablets had prevented starch digestion, fecal calories should have risen by 400 kcal. They did not move: 80 +/- 4 kcal on the blocker against 78 +/- 2 kcal on placebo. The conclusion was that starch-blocker tablets do not inhibit the digestion and absorption of starch calories in human beings.

The caveat is real and has to be stated, because it is the product's best defense. The 1982 tablets were a crude antiamylase preparation and the paper does not characterize their inhibitory activity. A modern aqueous extract standardized by measured alpha-amylase inhibitor activity is a different material. What has not happened in the 44 years since is anybody repeating that balance study with the modern material, which means the only direct measurement of whether a starch blocker blocks starch calories in a person is a negative one on a material nobody sells any more.

What the modern trials measure instead is body weight. A 12-week, double-blind, placebo-controlled randomized trial gave 81 completers a proprietary aqueous extract of whole dried white kidney beans, standardized by its alpha-amylase inhibitor activity, at 1,000 mg or 700 mg or a microcrystalline cellulose placebo, three times a day 30 minutes before meals, during a calorie-restricted diet Jäger 2024. It reported dose-dependent reductions in body weight, fat mass, body mass index, waist and hip circumference, with thigh circumference moving only in the high-dose group. Registered as NCT02930668.

Read the design before the result: every arm was on a calorie-restricted diet. So the trial measures what the extract adds on top of an energy deficit, which is the right clinical question and is not the same question as whether the extract blocks starch. A compound that improves adherence, or slows gastric emptying, or reduces the glucose swing that drives the next snack, would produce the same result without blocking a single calorie.

Pooled across the literature the effect is small, consistent, and pointed in an awkward direction. Eight randomized controlled trials and 543 participants give a weight difference of -1.62 kg (95% CI -1.99 to -1.25), body mass index -0.58 kg/m2, fat mass -1.17 kg, waist -1.58 cm and hip -0.99 cm, with no serious adverse events attributed to the extract Shi 2026. And no significant difference in fasting blood glucose, in triglycerides, or in random insulin.

That last sentence is the interesting one and it deserves its own paragraph. If the mechanism is inhibition of starch digestion, glycemic measures are the first place an effect should show and body composition is the last. The pooled data have it the other way round. Three readings are available and the page cannot choose between them honestly: the trials were powered for anthropometry and not for glycemia; fasting measures are the wrong endpoint for a meal-restricted mechanism and post-prandial curves would have shown it; or the weight change is not coming from starch blockade at all. The third is not the least likely of the three.

The post-prandial data support reading two, at least in principle. In 23 women with abdominal obesity, a plant formula containing kidney bean, white mulberry leaf and green coffee extracts cut the incremental glycemic area by 17.1 percent for instant noodle soup (p = 0.005) and 40.6 percent for white rice (p = 0.004), and did nothing at all for strawberry sorbet Lange 2022. The null result is the mechanistically satisfying one -- a sorbet is sucrose and fructose, and an amylase inhibitor has no substrate to work on. The limitation is equally plain: three actives in one capsule, so nothing in that trial can be attributed to the bean.

And one trial went at the distal-shift hypothesis directly. Ninety subjects, 4 months, 1.5 g of white common bean extract or 1.5 g of maltodextrin half an hour before a meal, with carbohydrate intake deliberately maintained rather than restricted Feng 2022. Glycated hemoglobin fell 0.721 +/- 0.742 percent in the extract group. Fewer treated patients scored 6 or above on the Toronto Clinical Scoring System, and sural sensory nerve conduction velocity drifted down in the control arm and up in the treated one. Bifidobacterium, Fecalibacterium and Anaerostipes were more abundant and Weissella, Klebsiella and Cronobacter less abundant at month 2, with Bifidobacterium still higher at month 4. This is the strongest human result in the file and it is worth being clear about what it is: an open question turned into a specific, checkable hypothesis, in one trial, on a population with type 2 diabetes, that has not been replicated.

White Kidney Bean Extract — which form, and does it matter

The unit on the label is the whole problem, and it is a unit most buyers have never heard of. Alpha-amylase inhibitor activity is expressed in alpha-amylase inhibiting units, and an inhibiting unit is defined by the assay that measured it -- the substrate, the pH, the temperature and the source of the amylase. There is no single legally fixed definition shared across manufacturers. Two tubs both reading 1,000 mg of white kidney bean extract can therefore differ by an order of magnitude in the only property that matters, and a milligram figure with no activity figure has told you the weight of a powder.

The review literature says this out loud. A 2022 update on common bean alpha-amylase inhibitors sets out the types of seed extract, the inhibitor proteins in each and their mechanism, and then closes on the urgent need for further studies to confirm the clinical efficacy of the commercial products Peddio 2022. That is the gap in one sentence: the protein is characterized, the commercial products are not.

The same failure has been documented one shelf over, for the sister protein. Lectin activity measured across commercially available edible plants was highest in the Fabaceae, and the paper's title is a call for harmonization of the analytical methods and for a risk assessment built on them Adamcová 2021. When an assay is not harmonized, numbers from different laboratories are not comparable, and a label number carries no cross-brand meaning.

Cultivar is a specification too, and it is invisible. Italian Phaseolus vulgaris cultivars screened for alpha-amylase and alpha-glucosidase inhibition and for the absence of phytohemagglutinin produced a genuinely awkward result: every cultivar had alpha-glucosidase inhibitor activity, but alpha-amylase inhibitor was missing altogether in two of them, and only one cultivar -- Nieddone, accession ACC177 -- showed no hemagglutination activity Peddio 2023. The active protein and the toxic one are cultivar-level properties, they are measured separately, and they do not travel together conveniently. A tub that names the species and not the cultivar has skipped the level at which the answer lives.

The extraction method is the safety specification, because both molecules are proteins. Aqueous extraction concentrates the water-soluble seed protein fraction -- which is where both the inhibitor and the hemagglutinin sit. Heat destroys both, which is exactly why properly boiled beans are safe food and why a heat step that guarantees safety also guarantees inactivity. The manufacturer is therefore running a purification that has to keep one protein folded while removing another, and no label describes how or reports the residual hemagglutinating activity of the finished powder.

The dose that has actually been studied is a schedule, not a number. The proprietary standardized extract was given at 700 or 1,000 mg three times a day, 30 minutes before meals -- 2.1 to 3.0 g a day, split across the three starch-containing meals Jäger 2024. The four-month diabetes trial used 1.5 g half an hour before a meal Feng 2022. A product taken once a day with no starch in front of it is not that protocol however many milligrams it contains, because an enzyme inhibitor with no enzyme working and no substrate present cannot do anything at all.

What would have to be true, and how you would know it was not

1. The balance study nobody has repeated, and the single experiment that would settle the category. Bo-Linn's design with the modern standardized material: a 100 g starch meal, one-day fecal calorie balance, placebo against 1,000 mg of extract Bo-Linn 1982 Jäger 2024. Prediction: fecal calorie excretion rises, and rises by well under 100 kcal rather than the 400 kcal that complete blockade of a 100 g starch load would give. If it rises by 400 kcal the 1982 conclusion does not transfer to the modern extract, and this page is wrong in the most useful possible way.

2. Breath hydrogen, which is the cheap version of the same question. If starch is being shifted distally rather than blocked, unabsorbed carbohydrate reaching the colon is fermented and some of the hydrogen produced is exhaled. Prediction: breath hydrogen after a starch meal taken with the extract is higher than after the same meal without it, peaking somewhere in the 2 to 4 hour window. If breath hydrogen does not move, either nothing is escaping the small intestine or nothing is fermenting it, and the distal-shift explanation this page leans on is wrong.

3. HbA1c at 16 weeks, on the protocol that produced the result. Prediction: on roughly 1.5 g a day taken before meals with carbohydrate intake maintained rather than cut, glycated hemoglobin falls, on the order of the 0.721 +/- 0.742 percent reported over four months Feng 2022. Retest at 16 weeks and not sooner, because HbA1c reflects roughly three months of red cell exposure and cannot be moved by one well-behaved fortnight. That slowness is the reason it is the marker worth trusting.

4. The prediction that cuts against the product. Fasting insulin and fasting glucose in a metabolically healthy adult, baseline and 12 weeks. Prediction: no change, because the pooled randomized data across eight trials found no significant difference in fasting glucose or in insulin Shi 2026. If a healthy person's fasting insulin drops meaningfully on this, then whatever is happening is not the mechanism described on this page, and the mechanism should be rewritten rather than the finding explained away.

5. The test that separates a standardized extract from bean flour. Two products matched on stated milligrams, one declaring alpha-amylase inhibiting units and one not, same person, same 100 g starch meal, breath hydrogen as the read-out, alternated over six pairs. Prediction: the undeclared product produces no rise, because activity is not a function of powder weight Peddio 2022 and some bean material carries no alpha-amylase inhibitor at all Peddio 2023. If the two perform identically, the standardization argument on this page is wrong and the unit on the label does not matter.

What nobody has tested yet

Nobody has measured how much intact, folded, active inhibitor reaches the human duodenum. This is a protein swallowed into an acid compartment full of pepsin, and its survival is the entire pharmacokinetic question. There is no published intubation study, no recovery figure, and therefore no way to know whether the 30-minute pre-meal instruction is optimal, arbitrary, or the only thing keeping the product working at all Jäger 2024.

No trial has reported the residual hemagglutinating activity of the extract it used. Given that the inhibitor and phytohemagglutinin come from the same seed protein family and are concentrated by the same aqueous extraction Peddio 2023, and that the analytical methods for lectin activity are not harmonized Adamcová 2021, this is the safety number the literature is missing rather than a number that has been measured and found reassuring.

Nobody has run a dose-response in alpha-amylase inhibiting units. The best trial compared 700 mg with 1,000 mg of one proprietary material Jäger 2024, which is a two-point milligram comparison and not a dose-response in the unit that defines the product. Nobody can say what the minimum effective activity is, so nobody can say which shelf products are under it.

Nobody has measured short-chain fatty acids in a white kidney bean trial. The best mechanistic result in the human literature infers short-chain-fatty-acid production from 16S abundances of the bacteria that make them Feng 2022. Fecal or plasma acetate, propionate and butyrate were not reported. Until somebody measures them, the most interesting hypothesis about this supplement rests on the taxonomy of the organisms rather than on the metabolites they are supposed to be producing.

And nobody has tested the extract without a diet attached. The 12-week trial ran every arm on calorie restriction Jäger 2024. A trial of the extract alone, on a weight-stable eucaloric diet, would say whether the compound does anything by itself or only adds to a deficit -- which is the question a person deciding whether to buy it is actually asking.

White Kidney Bean Extract — its own safety story, not its category's

The risk that belongs to this product rather than to its category is the seed's own lectin, and it is a real toxin rather than a theoretical one. Phytohemagglutinin in raw or undercooked kidney beans causes acute vomiting and diarrhea within a few hours; boiling destroys it and soaking alone does not. A standardized extract is not raw bean, and there is no signal of that syndrome in the trial literature -- eight randomized trials in 543 participants reported no serious adverse events attributed to the extract Shi 2026. What is missing is the confirmatory number: no manufacturer publishes the residual hemagglutinating activity of the finished powder, and the assays used to measure lectin activity are not harmonized between laboratories Adamcová 2021.

Cultivar makes that a live variable rather than a settled one. Screening Italian bean cultivars found alpha-amylase inhibitor absent in two of them and hemagglutination activity present in all but one Peddio 2023. So the raw material can vary in both the active and the toxin independently, and the buyer has no visibility into either.

The predictable adverse effects are the mechanism working. Flatulence, bloating, borborygmi and loose stool are what colonic fermentation of escaped starch feels like, and they should track the starch content of the meal rather than the dose of the capsule. A product that produces none of these on a large pasta meal is either inactive or being taken at the wrong time.

The interaction that actually matters is with glucose-lowering therapy, and it is arithmetic rather than pharmacology. An agent that flattens the post-meal glucose rise while the mealtime insulin dose or the sulfonylurea stays fixed is stacking two glucose-lowering actions on one meal, and the risk window is the first 1 to 3 hours after eating. A trial of the extract moved glycated hemoglobin by roughly 0.7 percent over four months Feng 2022, which is the size of a real medication change and is a conversation with the prescriber before it is a purchase.

Two more specifics and a limit. Anyone with a diagnosed legume allergy is being offered a concentrated legume seed protein and should treat it as such. Anyone with irritable bowel syndrome whose symptoms are driven by fermentation is being offered more substrate for exactly that, which is the mechanism and not a side effect. And the honest limit of the product is written into how it works: taken away from a starch-containing meal it has no substrate, no enzyme to inhibit and no way to do anything at all Lange 2022. Nothing here is medical advice, this page does not diagnose or treat diabetes, prediabetes or obesity, and none of these statements has been evaluated by the Food and Drug Administration.

Sources read for this page

• Bo-Linn GW. Starch blockers--their effect on calorie absorption from a high-starch meal. N Engl J Med 1982 · PMID 6182469
• Jäger R. Proprietary alpha-amylase inhibitor formulation from white kidney bean (Phaseolus vulgaris L.) promotes weight and fat loss: a 12-week, double-blind, placebo-controlled, randomized trial. Sci Rep 2024 · PMID 38830962
• Shi N. White kidney bean extract reduces body weight and adiposity with acceptable safety in adults with overweight and obese: a systematic review and meta-analysis. Nutr Res 2026 · PMID 42066439
• Peddio S. Common bean (Phaseolus vulgaris L.) alpha-amylase inhibitors as safe nutraceutical strategy against diabetes and obesity: An update review. Phytother Res 2022 · PMID 35485365
• Peddio S. Biochemical and Phylogenetic Analysis of Italian Phaseolus vulgaris Cultivars as Sources of alpha-Amylase and alpha-Glucosidase Inhibitors. Plants (Basel) 2023 · PMID 37631130
• Feng Y. White common bean extract remodels the gut microbiota and ameliorates type 2 diabetes and its complications: A randomized double-blinded placebo-controlled trial. Front Endocrinol (Lausanne) 2022 · PMID 36303868
• Lange E. Comparison of Glycemic Response to Carbohydrate Meals without or with a Plant-Based Formula of Kidney Bean Extract, White Mulberry Leaf Extract, and Green Coffee Extract in Individuals with Abdominal Obesity. Int J Environ Res Public Health 2022 · PMID 36231426
• Adamcová A. Lectin Activity in Commonly Consumed Plant-Based Foods: Calling for Method Harmonization and Risk Assessment. Foods 2021 · PMID 34829077

How you would know if it worked

White Kidney Bean Extract — safety & side effects

Bloodwork to run alongside White Kidney Bean Extract

Baseline first, then again at 8–12 weeks.

The Metabolic Health & Prediabetes panel covers these in one order — 8 markers, $81.45 with the discount applied.

Check results you already have → · All 103 markers A–Z

White Kidney Bean Extract — frequently asked questions

What White Kidney Bean Extract is used for

White Kidney Bean Extract appears under 1 goal in the goal router.

Related Metabolic & Weight supplements

← Browse the full Supplement Vault

↑ Back to on this page