Going lactose-free eased the belly pain of most people with a positive breath test. The proof: nine studies going back to 1979, and not one randomised trial made the cut

Plenty of people with irritable bowel syndrome try dropping milk, and plenty of them have already been sent for a lactose breath test. Gastroenterologists and paediatric specialists in Brussels and Rouen set out to see what the evidence says about whether it helps. Their systematic review, in Nutrients on 31 October, covers adults with IBS and children with recurrent abdominal pain, the older label for what would now largely be called IBS in a child.

They searched four databases up to 15 August 2026 with no limits on language or date, and went through 973 records. Studies had to cut lactose for at least four weeks. Nine qualified: three in adults, all in Europe, and six in children, from Europe, the USA and India. Between them the study groups come to 455 people, 224 adults and 231 children (our addition). No randomised controlled trial met the criteria. Shorter trials were excluded by design, so some may exist, but what's reviewed here rests on cohorts. The studies date from 1979 to 2017, and seven of the nine came out in 2006 or earlier (our count).

On the face of it the results are striking. Among children with a positive test, pain improved in 59% to 100%, depending on the study. In children who tested negative, the three studies reporting it found 0% to 32%. In one Italian study of 18 test-positive children, pain episodes fell from an average of 16.9 a month to 1.6, and that's the only study in the review to put a number on how often pain happened. In the largest adult study, also Italian, 44% of 110 test-positive patients on the diet reported complete remission of symptoms and 39% partial. Among 22 test-positive patients who stayed off the diet, the review gives 18% with partial remission. It doesn't say how those 22 came to skip it.

Now look at how those figures were made. The authors rated the risk of bias as high in every study, and critical in several categories. Four of the nine had no control group. Where there was one, patients and investigators knew the breath test result, and the authors warn that this could have biased how the diet's effect was seen. Improvement was often just that word, undefined, or the judgement of the treating doctor. None of the studies used a validated symptom questionnaire, and adherence to the diet wasn't checked. Some of the comparison groups were also tiny: the 0% comes from 6 children and another test-negative figure from 4.

Expectation isn't a small matter in this condition. The review points to work in which IBS patients who believed they were sensitive to gluten reacted similarly to gluten, wheat and a sham, and to a low-fructose diet that helped regardless of breath test result. Either finding, if it applied here, would muddy the idea that a positive lactose test picks out the people who'll benefit.

The long-term picture is thin too. A Dutch study saw improvement fall from 17 of 18 children at 5 months to 11 of 18 at 15 months, while a German one held at 15 of 17 at 6 months. The adult results are odder. In a Spanish study, 1 of 50 test-positive adults on the diet for six months reached complete symptom resolution, and which symptoms that covered wasn't specified. And IBS was defined by criteria now well out of date: Manning or Rome II in the adult studies, with only one of the children's studies using any Rome definition, Rome III. Rome V is current.

Nobody measured quality of life, healthcare visits, days of work lost or how hard the diet was to live with. Those are what a person weighing up a lasting change would want to know. The authors still conclude that pain improves on the diet when malabsorption is present, and in the same breath that the studies are at high risk of bias and hard to generalise. To us, that reads as a signal and not yet a settled result. They note that guidelines advise against routine testing for carbohydrate malabsorption in IBS, and say a lactose breath test should be kept for patients whose history strongly suggests a link. What they want next is a trial with current diagnostic criteria, validated measures and patients and staff kept blind to the test result.

The paper reports no external funding. Three of the nine authors declare speaker fees or consultancy with companies, among them Nestlé HealthScience, Biocodex and Norgine. The review also has loose ends, listed in the data tab, that we couldn't resolve from the published text.