The phrase “gluten free” on a packet does not mean the food contains no gluten. It means the gluten present is below twenty parts per million, which is twenty milligrams in every kilogram of food. That figure appears in food regulations in many countries and is one of the more consistent numbers in international food labelling.
Two questions follow immediately. Why twenty, rather than zero or fifty? And how does a laboratory measure something at that concentration in a food matrix that may be a biscuit, a sauce, a beer or a soy protein isolate?
The answers are connected. The threshold sits where it does because of research into how much gluten people with coeliac disease can tolerate without measurable intestinal damage, tempered by what analytical methods can reliably detect and by what food manufacturing can realistically achieve. It is a public health compromise built on biology and constrained by chemistry, and both halves are worth understanding before deciding how much confidence to place in a label.
Key takeaways
- Gluten free is a defined maximum of twenty parts per million, not an absence of gluten.
- The threshold derives from dose-response work showing that very small daily amounts do not cause detectable intestinal damage in most people with coeliac disease.
- Most testing uses a sandwich ELISA against a gluten peptide sequence, which requires the protein to be reasonably intact.
- Hydrolysed and fermented products such as beer and soy sauce break the assay’s assumptions and need different methods.
- Uneven distribution of contamination means sampling design matters as much as the assay.
What Gluten Free Legally Means
Gluten is not a single protein. It is the collective name for the storage proteins of wheat and its relatives, principally the gliadins and glutenins of wheat, the secalins of rye and the hordeins of barley. These proteins share sequences rich in proline and glutamine that resist complete digestion, leaving peptide fragments long enough to trigger an immune response in genetically susceptible people.
Food regulations in most major jurisdictions define gluten free as containing no more than twenty milligrams of gluten per kilogram of food as sold. The wording matters: the limit applies to the food as consumed or as sold, not to an ingredient before dilution, and the claim is a maximum rather than a target.
Terminology used voluntarily by manufacturers is a different matter. Statements such as “made in a facility that also handles wheat” or “may contain traces of gluten” are precautionary allergen labelling. In most places these are not defined by regulation, are not backed by any specific numerical limit, and are applied inconsistently. A product carrying a precautionary statement may in practice contain no detectable gluten; another with no statement may contain more. The regulated claim carries a defined meaning; the precautionary one does not.
The Research Behind the Threshold
The threshold is not arbitrary, and its derivation is the most useful thing to understand about it.
Coeliac disease is a dose-related immune reaction rather than an all-or-nothing one. Intestinal damage depends on how much of the triggering protein reaches the small intestine over time, and the relationship is graded: larger amounts produce more damage, smaller amounts produce less, and at sufficiently small amounts the effect becomes undetectable by the methods available to measure it.
Establishing where that point sits required feeding studies in which people with treated coeliac disease consumed known small daily amounts of gluten over a period of weeks to months, with intestinal biopsies before and after to look for changes in the structure of the small intestinal lining. The key measurement was mucosal architecture, since that is the objective marker of damage rather than symptoms, which are unreliable in both directions.
The broad finding was that daily intakes in the region of a few tens of milligrams could produce measurable changes in some participants, while daily intakes in the low single-digit milligrams did not produce detectable damage in the groups studied. Regulators then worked backwards. If the tolerable daily amount is around a few milligrams, and a person eating a gluten free diet consumes a certain quantity of substitute foods each day, what concentration in those foods keeps total daily intake below the tolerable amount? For a plausible daily intake of specially manufactured foods, a limit of twenty milligrams per kilogram keeps the total within that range.
Two features of this reasoning deserve emphasis. First, the limit protects against cumulative daily intake, not against a single exposure, which is why the concentration matters alongside how much of the food is eaten. Someone whose diet consists almost entirely of gluten free substitute products at exactly twenty parts per million is closer to the tolerable intake than someone eating mostly naturally gluten free whole foods with an occasional labelled product.
Second, individual sensitivity varies. The threshold was set to protect the large majority, and a minority of people appear to react at lower exposures. Those individuals may need to be stricter than the label allows, choosing certified products and avoiding foods produced on shared lines even when they carry a compliant claim.
How ELISA Detects Gluten Fragments
Extraction comes first and is a significant part of the method. Gluten proteins are poorly soluble in water, so the sample is ground and extracted with a solvent system, commonly involving alcohol, sometimes with reducing and disaggregating agents added to break up protein aggregates. This cocktail matters because heat treatment during cooking causes gluten proteins to aggregate and become harder to extract, and an extraction that leaves protein behind under-reports the result.
The sandwich assay then works as follows. Antibodies specific to a gluten sequence are immobilised on the wells of a microplate. The extract is added, and any gluten present binds. After washing, a second antibody carrying an enzyme label is added, which binds to a different site on the same captured molecule. After a further wash, a substrate is added, the enzyme converts it into a coloured product, and a plate reader measures the intensity. Colour is proportional to gluten concentration, read against a calibration curve made from standards.
The most widely used antibody targets a specific repeated peptide motif found in the gluten proteins of wheat, rye and barley, which is also the region relevant to the immune response in coeliac disease. This is elegant, because the assay measures something close to what actually matters biologically rather than total protein.
Three constraints follow from the sandwich design.
The molecule must be large enough to carry two antibody binding sites simultaneously. A fragment with only one site can bind the capture antibody but produces no signal, and it may also block sites, suppressing the reading. This is the origin of the hydrolysed food problem discussed next.
The result must be converted from measured protein to reported gluten. Assays are typically calibrated against a prolamin fraction, and because prolamins make up roughly half of total gluten, the measured value is conventionally multiplied by two. That convention assumes a composition that varies between cereals and between varieties, so it introduces uncertainty.
Antibody performance differs between wheat, rye, barley and oats, since the target sequences are similar but not identical. A method validated on wheat may respond differently to barley contamination, which matters when the likely contaminant is barley malt.
| Method | What it detects | Best suited to | Main limitation |
|---|---|---|---|
| Sandwich ELISA | Intact gluten peptides with two binding sites | Most solid and liquid foods | Underestimates hydrolysed gluten |
| Competitive ELISA | Single fragments, including short peptides | Beer, soy sauce, fermented and hydrolysed foods | Different calibration, harder to relate to intact gluten |
| Lateral flow strip | Presence above a set threshold | Rapid screening on production lines | Qualitative or semi-quantitative only |
| Mass spectrometry | Specific peptide sequences directly | Confirming disputed or complex results | Costly, specialist, no universal standard method |
| Polymerase chain reaction | Cereal DNA, not gluten protein | Detecting the presence of the grain | DNA and protein do not track each other after processing |
Hydrolysed and Fermented Product Problems
Fermentation and hydrolysis break proteins into smaller peptides. If those fragments are too short to bind two antibodies at once, a sandwich ELISA reports a low result even when substantial amounts of gluten-derived peptide remain. The danger is that the fragments may still be long enough to be immunologically active, since the peptides that trigger coeliac disease are themselves short. A product can therefore test below twenty parts per million by sandwich ELISA and still contain immunologically relevant material.
The partial solution is a competitive ELISA. In this format, gluten in the sample competes with a labelled reference for a limited number of antibody sites, so a single binding site is sufficient to generate a signal, and the signal falls as gluten rises. This detects fragments that the sandwich format misses. Its drawback is that the relationship between a competitive result and a sandwich result is not fixed, so the two are not directly comparable, and expressing the outcome in the same parts-per-million units as an intact-protein measurement involves assumptions.
For a consumer, the practical guidance is straightforward. Prefer fermented and hydrolysed products made from naturally gluten free ingredients over those made from gluten-containing grains and subsequently treated, because the first requires no analytical inference and the second does.
Sampling and Uneven Distribution
The reason is physical. Contamination usually arrives as discrete particles: a few grains of wheat in a bulk delivery of oats, flour dust settling on one part of a line, residue in a section of pipework released as a slug into one part of a run. Those particles do not distribute themselves evenly. A single wheat grain contains a large amount of gluten relative to a twenty parts per million limit, so a kilogram of otherwise clean flour containing one stray grain can exceed the limit, while the adjacent kilogram contains nothing.
Proper practice addresses it in two ways. First, take many small increments from across the batch, over time in a continuous process or from many points in a static one, and combine them into a composite sample. Many increments capture the distribution far better than one large grab from a single point. Second, grind and homogenise thoroughly before subsampling, because the analytical portion taken forward for extraction is small and any remaining particulate heterogeneity translates directly into variable results.
Oats, Cross Contact and Purity Protocols
Oat protein, avenin, is related to the prolamins of wheat, rye and barley but differs enough that the majority of people with coeliac disease tolerate pure oats without evidence of intestinal damage. A minority appear to react to avenin itself, which is why national coeliac organisations advise introducing oats cautiously and under guidance, and why some are more conservative than others.
The larger practical issue is contamination rather than avenin. Oats are commonly grown in rotation with wheat and barley, harvested with equipment used for those crops, transported in shared vehicles, and milled in facilities that handle them. Ordinary commodity oats therefore frequently contain wheat or barley at levels well above twenty parts per million, and the variation between samples of the same product can be wide.
Purity protocols address this along the entire chain: seed stock verified free of other cereals, fields with documented rotation history and inspection for volunteer wheat plants, dedicated or thoroughly cleaned harvesting equipment, segregated transport and storage, milling on dedicated lines, and testing at multiple points rather than only at the end. Oats produced this way are sold as gluten free oats and carry the claim legitimately; ordinary oats do not, regardless of whether the packet mentions wheat as an ingredient.
Cross contact in the wider food supply follows the same logic. The risk points are shared production lines, shared bulk transport, airborne flour dust in bakeries, and shared equipment in food service. Manufacturing controls include physical segregation, dedicated equipment, production scheduling that runs gluten free products first after a full clean, validated cleaning with verification testing, and air handling to control dust.
In domestic kitchens the same principles scale down. Porous items such as wooden boards, worn plastic utensils and shared toasters retain crumbs. Shared condiment jars accumulate crumbs from knives. Flour dust from baking settles across surfaces and stays airborne for a surprisingly long time. These are the routine sources of accidental exposure for someone who otherwise reads every label carefully.
Reading a Gluten Test Report
Start with the method. The report should name the assay, the antibody or kit used, and the format, whether sandwich or competitive. For a fermented or hydrolysed product, a sandwich ELISA result alone is not sufficient evidence, and the report should say what else was done.
Find the limit of detection and the limit of quantification. These are different. The detection limit is the lowest concentration distinguishable from background; the quantification limit is the lowest that can be reported as a number with acceptable precision. A result reported as less than the quantification limit means the laboratory could not measure it reliably below that point, not that the food contained nothing. Where the quantification limit is close to twenty, the test has little headroom to demonstrate compliance convincingly.
Look for measurement uncertainty. A competent laboratory states it, and at these concentrations it is not small. A result of eighteen parts per million with an uncertainty of several parts per million does not demonstrate compliance with any confidence, and a result of twenty-two does not by itself prove non-compliance. Judgements near the limit require repeat testing rather than a single number.
Check what the sample was and how it was taken. The report should describe the material, the number of increments, and how it was homogenised. A certificate covering a single grab sample from one point in a batch says considerably less than one covering a properly composited sample.
Frequently asked questions
Does twenty parts per million mean the food is completely gluten free?
No. It means gluten was not detected above twenty milligrams per kilogram, which is a maximum permitted concentration rather than an absence. The figure was chosen so that a person eating a normal quantity of such foods keeps their total daily gluten intake below the amount shown not to cause detectable intestinal damage in most people with coeliac disease. Someone whose diet is heavily weighted toward manufactured substitute foods accumulates more than someone eating mostly naturally gluten free whole foods.
Why is beer so difficult to test?
Fermentation breaks gluten proteins into shorter fragments. The standard sandwich ELISA needs a fragment long enough to be gripped by two antibodies at once, so short peptides go undetected even though they may still be immunologically active. A beer brewed from barley can therefore return a low result while containing gluten-derived peptides. Competitive ELISA formats and mass spectrometry detect these fragments better. Beers brewed from naturally gluten free grains avoid the ambiguity entirely, whereas barley beers treated with a gluten-degrading enzyme are the hardest case to verify.
Are oats safe on a gluten free diet?
Most people with coeliac disease tolerate pure, uncontaminated oats, but two separate issues apply. Ordinary commodity oats are frequently contaminated with wheat or barley through shared fields, harvesting, transport and milling, often well above the twenty parts per million limit, so only oats produced under a purity protocol should be used. Separately, a minority of people appear to react to oat protein itself, which is why introduction is usually advised cautiously and with medical guidance rather than assumed to be safe.
Are home gluten test kits reliable?
Portable and consumer test devices generally use lateral flow strips with the same antibody chemistry as laboratory assays, so the underlying detection is real. The limitations are practical and severe. The tested portion is tiny, and because contamination is typically particulate and unevenly distributed, a small sample can easily miss it or hit it. Results are qualitative or semi-quantitative rather than precise, and matrix effects in complex or hydrolysed foods can mislead. A negative result on a fragment of a meal should not be treated as evidence that the meal is safe.
Why do two laboratories report different results for the same product?
Several sources of difference stack up. The sample they received may differ genuinely if contamination was uneven. Different kits use different antibodies, different extraction solvents and different calibration standards, and the conversion factor from measured prolamin to reported gluten involves an assumption about composition. Measurement uncertainty at these low concentrations is proportionally large. Results within a few parts per million of each other are broadly consistent rather than contradictory, which is why proficiency testing and accreditation matter when choosing a laboratory.
The number on the packet is best understood as the output of a control system rather than a measurement of a particular biscuit. It rests on dose-response research establishing what intake is tolerable, on an assay that works well for intact protein and less well for hydrolysed material, and on sampling and process controls that determine whether the tested portion said anything about the rest of the batch. For most people that system works, which is why the labelling framework has held up. For those who react at lower exposures, the reasonable response is not to distrust the threshold but to recognise what it was designed to do, and to lean on certified products, naturally gluten free foods and careful attention to cross contact in kitchens, where the largest accidental exposures usually originate.
This is education, not medical advice. Laboratory results only carry meaning alongside your symptoms, history and examination. Talk to a qualified clinician about your own results before changing anything about your care or supplements.
