Why most peptides cannot be a pill: oral bioavailability explained

The short answer

Most peptides cannot be swallowed as a pill because the digestive system breaks peptides apart, and whatever survives is too large and water-loving to cross the gut wall. Typical oral bioavailability is under 1 percent. The few that made it into a tablet, desmopressin, semaglutide and octreotide, did so with far larger doses, strict dosing rules, or both.

A peptide is a short chain of amino acids, the same building blocks your body assembles into proteins. That is also the whole problem. Your stomach and intestine are a purpose-built machine for taking chains of amino acids and cutting them into pieces small enough to absorb as food. A peptide drug that goes in by mouth is, from the gut's point of view, a snack.

The result is that peptide and protein medicines have historically required parenteral administration (an injection or infusion, any route that bypasses the digestive tract), which is a known deterrent to people taking their medicine [1]. A 2026 review of the field puts the general figure plainly: the barriers of the gut leave most peptides with an inherent oral bioavailability below 1 percent [3].

That figure is the honest summary, but it hides the interesting part. A semaglutide tablet exists and produces real weight loss [8]. An insulin tablet matched injected insulin in a trial and was cancelled anyway [9]. And one small peptide, desmopressin, has been given to the same volunteers by six different routes [4], which is the closest thing we have to a controlled experiment on the question in this article's title. The rest of this post walks through those cases with the actual numbers, because the numbers are where the answer lives.

What bioavailability actually measures

Bioavailability is the fraction of a dose that reaches the bloodstream intact, measured against an injection that puts all of it there. A 1 percent figure means 99 percent of every tablet was destroyed or never absorbed. It is read off blood concentration curves over time, not off whether a person feels an effect.

Bioavailability is the share of a dose that arrives in the bloodstream in its original form. To measure it, researchers give the drug by the route under test, take blood samples for hours, and compare the resulting concentration curve against the curve from an injection, usually intravenous or subcutaneous (under the skin), where the whole dose is assumed to reach the blood. The comparison is made on the area under the curve (the total exposure over time, abbreviated AUC), so a route that delivers, dose for dose, a hundredth of the injection's AUC has 1 percent bioavailability.

Two things make a molecule bad at this. The first is size and chemistry. The widely used rule of five predicts that poor absorption or permeation is more likely when a molecule has more than five hydrogen-bond donors, more than ten hydrogen-bond acceptors, a molecular weight over 500, or a calculated log P above 5 (log P is a measure of how much a substance prefers oil over water) [2]. Peptides fail this on several counts at once: they are large, highly polar, charged, and unstable in the presence of the enzymes that cut proteins [3].

The second is the environment. The stomach sits at a pH of roughly 1 to 2, and the enzymes pepsin, trypsin and chymotrypsin exist specifically to cut peptide bonds, so a swallowed peptide is degraded extensively and quickly [3]. Anything that survives then meets a dense, heterogeneous layer of mucus lining the intestine before it reaches the cells that could absorb it [3]. The strategies developed to get around all this, including permeation enhancers, inhibitors of gut enzymes, and mucus-penetrating and cell-penetrating peptides, are the subject of an entire research field [1]. Which is another way of saying the default outcome, with no help, is that the peptide does not get through.

One molecule, six routes: the desmopressin experiment

Desmopressin, a synthetic version of the hormone vasopressin, was given to eight healthy volunteers by six routes. Measured against the subcutaneous injection, intranasal delivery reached 3.4 percent and a 200-microgram tablet reached 0.1 percent. After sublingual and rectal dosing no drug was detectable in blood. Same molecule, a thousandfold spread.

The cleanest answer to "does the route matter" comes from a 1993 study of desmopressin (1-deamino-8-D-arginine vasopressin, a synthetic relative of the hormone that tells your kidneys to hold on to water). Eight healthy volunteers each received the drug by six routes on separate occasions at least a week apart, intravenous, subcutaneous, intranasal, oral, sublingual and rectal, and were followed for eight hours each time with blood and urine sampling [4].

The doses tell you what the researchers already expected. The intravenous and subcutaneous doses were 2 micrograms. The intranasal and sublingual doses were 20 micrograms, the rectal dose 50 micrograms, and the oral route used a 200-microgram tablet, a hundred times the injected amount [4]. Using the subcutaneous curve as the reference, bioavailability came out at 3.4 percent after intranasal and 0.1 percent after oral administration [4]. After the sublingual and rectal doses, no desmopressin was detectable in blood at all; only low amounts turned up in the 24-hour urine collection [4]. The authors concluded that bioavailability by the nasal and oral routes seemed lower than had previously been reported [4].

Read that against the tablet dose. For every microgram swallowed, about a thousandth of what the same microgram delivers under the skin reached the blood [4]. That is not a rounding error, it is the mechanism this article is about. Desmopressin tablets are themselves a marketed medicine, and the DDAVP tablet label puts their bioavailability at about 5 percent compared with the nasal spray and about 0.16 percent compared with intravenous desmopressin [12], in line with the volunteer study. Two cautions travel with the result. This is one small peptide in eight people, and other peptides will land at different points on the ladder. And the sublingual finding matters because "sublingual" is now a marketing word for research peptides: in the one controlled comparison here, holding a peptide under the tongue produced nothing measurable in the blood [4].

How semaglutide got into a tablet

Rybelsus pairs semaglutide with SNAC, an absorption enhancer that buffers the stomach lining locally and lets the peptide cross stomach cells right beside the tablet. Under the label's rules bioavailability is about 0.8 percent, rising with a longer fast and falling with more water. The 30-minute wait is the mechanism, not a formality.

The landmark case is oral semaglutide, approved for type 2 diabetes as Rybelsus [1]. The 2018 paper that explained how it works found something unusual. Absorption of semaglutide from the tablet takes place in the stomach, not the intestine where small-molecule drugs are normally absorbed; it is confined to an area close to the tablet's surface; and it requires co-formulation with SNAC (sodium N-[8-(2-hydroxybenzoyl) aminocaprylate], an absorption enhancer) [5]. SNAC protects the peptide from enzymatic degradation through a local buffering action and enhances absorption only transiently, by a route that goes through the stomach cells themselves rather than between them [5]. The authors also reported that the mechanism is compound specific [5], which is worth remembering the next time a product page promises that adding SNAC to any peptide makes it oral.

What that engineering buys is spelled out on the FDA label. Semaglutide's estimated absolute bioavailability was approximately 0.4 to 1 percent after oral doses of 3, 7 and 14 mg of Rybelsus, and 1 to 2 percent after 1.5, 4 and 9 mg of the newer Ozempic tablet formulation, with peak concentration about one hour after dosing [6]. The two tablet products are not substitutable milligram for milligram [6]. Hence the ritual: take one tablet once daily on an empty stomach in the morning with up to 4 ounces of water, no other liquids, swallow it whole, and wait at least 30 minutes before eating, drinking or taking other oral medicines [6]. In the label's own dosing-conditions study, absorption was higher after a longer post-dose fast, and higher with 50 mL of water than with 240 mL [6].

A pharmacokinetic model built from six clinical pharmacology trials put a single number on it: bioavailability was 0.8 percent when the tablet was taken under the recommended conditions (a 30-minute post-dose fast, no more than 120 mL of water), increasing with a longer fast and decreasing with a larger water volume [7]. Within-subject variability in bioavailability was 137 percent, which, because the drug is dosed daily and has a long half-life, works out to 33 percent variability in steady-state exposure [7]. So the 30-minute rule is not bureaucratic caution. It is the difference between the drug working and the tablet being an expensive way to feed your stomach lining.

And it does work. In OASIS 4, a 71-week double-blind trial at 22 sites in four countries, 205 adults with overweight or obesity and no diabetes were randomised to a 25 mg oral semaglutide tablet once daily and 102 to placebo, both alongside lifestyle intervention [8]. The estimated mean change in body weight at week 64 was minus 13.6 percent against minus 2.2 percent on placebo, an estimated difference of 11.4 percentage points [8]. For comparison, the injectable semaglutide dose the trial names as the alternative is 2.4 mg [8], and that injection is given once weekly [13], against a tablet taken every day. That ratio is the price of the pill.

Oral insulin worked in the trial and was cancelled anyway

In 2019 Novo Nordisk reported an oral insulin tablet that matched injected insulin glargine on fasting glucose over eight weeks in 50 patients with type 2 diabetes. The project was still discontinued. The doses required were so high that producing enough for wide public use was judged not commercially viable.

Insulin is the peptide that oral-delivery efforts have chased longest [1], and the most instructive result is one that succeeded [9]. I338 was a long-acting basal insulin analogue formulated in a tablet with the absorption enhancer sodium caprate [9]. In a phase 2, eight-week, randomised, double-blind, double-dummy trial at two research institutes in Germany, 50 insulin-naive patients with type 2 diabetes were assigned to once-daily I338 plus a placebo injection or once-daily insulin glargine plus a placebo tablet, 25 in each group [9].

At eight weeks, least-squares mean fasting plasma glucose was 7.1 mmol/L on the tablet and 6.8 mmol/L on the injection, a difference of 0.3 mmol/L that was not statistically significant (p equals 0.46) [9]. Adverse events were reported in 60 percent of the I338 group and 68 percent of the glargine group, diarrhoea in 12 percent of each, with no severe adverse events and low hypoglycaemia in both [9]. The authors' interpretation was that I338 could safely improve glycaemic control with no evidence of a difference from a widely used injectable basal insulin [9].

Then the sentence that matters. Further development was discontinued because I338 doses were high, and producing the required quantities for wide public use was therefore deemed not commercially viable [9]. The recommended starting dose in the trial was 2700 nmol of I338 against 10 units of glargine, with a ceiling of 16,200 nmol against 60 units [9]. An oral peptide can clear the scientific bar and still fail the arithmetic, because low bioavailability means most of what you manufacture is destroyed on purpose. Even insulin could not afford that waste [9]. The trial was funded by Novo Nordisk [9], the company that also made the semaglutide tablet, which tells you the decision was about the molecule, not the ambition.

The 200-to-1 tablet: oral octreotide

Octreotide became a capsule by pairing the peptide with an absorption enhancer and releasing it in the intestine. In 75 healthy volunteers, a 20-milligram oral dose produced blood levels equivalent to a 0.1-milligram injection. That is a 200-fold dose to reach the same exposure, which is roughly what an oral peptide costs.

Another peptide that reached the market as a capsule is octreotide, a drug for acromegaly (a condition of excess growth hormone). The human pharmacokinetic work behind it consisted of four single-dose studies in 75 healthy volunteers comparing oral doses of 3, 10 and 20 mg with a single subcutaneous injection of 100 micrograms [10]. Oral octreotide appeared in the circulation within an hour, rose with dose, and decayed at a rate similar to the injection [10]. The headline comparison is that 20 mg by mouth and 0.1 mg under the skin produced equivalent pharmacokinetics: mean peak concentration 3.77 against 3.97 ng/mL, and mean area under the curve 16.2 against 12.1 hours times ng/mL [10].

Twenty milligrams by mouth to match a tenth of a milligram by injection [10] is a 200-fold dose ratio. The 2026 review describes the marketed product, Mycapssa, as containing sodium caprylate and stabilisers in an enteric capsule (a coating that survives the stomach and dissolves in the intestine) and puts its oral bioavailability at around 0.7 percent in humans [3]. Both numbers say the same thing from different directions.

The same review names the pattern. Formulation strategies have relied on a small set of well-characterised excipients applied across many peptides and rarely produce bioavailability beyond the low single-digit range in large animals and humans, a persistence it calls an apparent glass ceiling [3]. What has changed is not the ceiling but the peptides: structural modifications that raise potency, metabolic stability and half-life have made therapeutic efficacy possible even at low single-digit bioavailability, with oral semaglutide, Merck's macrocyclic PCSK9 inhibitor and an oral IL-23 receptor antagonist given as the examples [3]. In plain terms, a peptide gets to be a pill when it is potent enough, and cheap enough per milligram, that throwing away 99 percent of every dose is affordable.

Collagen peptides are absorbed, and that is not a counterexample

Collagen supplements do reach the blood after swallowing, but as two- and three-amino-acid fragments such as Pro-Hyp, not as the original protein. In volunteers who drank 9.4 to 23 grams of gelatin hydrolysate, peptide-bound hydroxyproline peaked at 20 to 60 nanomoles per millilitre within two hours. Fragments surviving is different from a drug arriving intact.

At this point a reasonable reader objects that collagen peptides are sold by the tub, swallowed by the spoonful, and studied in trials, so surely peptides can be absorbed. They can, and the details show why it does not carry over. In a 2005 study, healthy volunteers drank 9.4 to 23 grams of gelatin hydrolysates (collagen that has already been chopped into small pieces) made from porcine skin, chicken feet and cartilage after a 12-hour fast [11]. Before ingestion, negligible amounts of hydroxyproline in peptide form were found in their blood. Afterwards, peptide-form hydroxyproline rose significantly, reached a maximum of 20 to 60 nanomoles per millilitre of plasma at one to two hours, and fell to half that level by four hours [11].

The identity of what arrived is the point. The major constituent found in serum and plasma was Pro-Hyp, a two-amino-acid fragment, with small but significant amounts of other dipeptides and tripeptides such as Ala-Hyp, Ala-Hyp-Gly, Pro-Hyp-Gly, Leu-Hyp, Ile-Hyp and Phe-Hyp [11]. Nothing resembling collagen crossed the gut. What crossed were two- and three-amino-acid fragments, the forms the study identified in blood [11].

A peptide drug cannot use that door. A peptide drug has to reach its receptor as the sequence that was designed for it; a fragment of semaglutide is not a weaker semaglutide, it is an amino acid snack. That is why "bioavailable" means different things on a collagen tub and on a drug label, and why the gram-scale doses of a collagen supplement tell you nothing about a milligram-scale peptide. It is also why the sublingual and "advanced coating" claims attached to research peptides deserve the same test the products in this article passed: a human study with blood samples, compared against an injection, with the dose ratio stated. The desmopressin data above are what happens when that test is actually run on a sublingual peptide [4]. For BPC-157 specifically, where the oral claims are loudest, we have gone through the animal pharmacokinetics in a separate post.

How to read an oral peptide claim

Ask three questions of any oral peptide claim: was bioavailability measured in humans, what route was it measured against, and what dose ratio did it take. The Rybelsus label answers all three. A product that answers none of them is asking you to trust a number no regulator has examined.

Everything above reduces to a short checklist you can apply to any capsule, troche or spray sold as an oral peptide. First, was it measured in people? The semaglutide mechanism paper used dogs alongside human trials [5], but the number that matters was measured in humans, across six clinical pharmacology trials [7]. Second, against what? A percentage is meaningless without the reference route; the desmopressin study says subcutaneous, in so many words [4]. Third, at what dose ratio? A hundred to one for desmopressin [4], two hundred to one for octreotide [10], and a 25 mg daily tablet [8] against a 2.4 mg once-weekly injection [13] for semaglutide. A product that claims high oral bioavailability at the same dose as the injection is claiming to have solved a problem that Novo Nordisk could not solve for insulin [9].

None of this means oral peptides are a dead end. The 2026 review's whole argument is that the field has moved from "is it feasible" to "how do we optimise it", with molecules designed to be potent enough to tolerate a 1 percent yield [3]. It means the pills that exist are engineering achievements with fine print, and the fine print is the mechanism. If you want the full picture of how semaglutide works in the body, injected or swallowed, our semaglutide mastery course covers the pharmacology end to end, and the free unit on how peptides are taken is the beginner's version of this article.