Does Food Combining Work? One Real Trial in Ninety Years Says No

The rule is that you should never eat protein and starch in the same meal. No steak with potatoes. No eggs with toast. No cheese sandwich. Do it wrong, the theory says, and the food sits in your stomach and rots.

Here is the number that should end the conversation: in the ninety-odd years this idea has been sold, it has been put through a real trial — a control group, calories matched, every meal served in hospital rather than reported — once. Fifty-four people, for six weeks, in the year 2000 [1]. Two much smaller studies since, a dozen people each, pointed the same way; they are below.

We’ll get to what that trial found. But the more interesting question is how a rule about what you may put on a plate went seven decades before anyone ran that trial, when its central claim about the stomach had been tested, and had failed, within seven years of the book.

It began with one man who felt better

In 1905 a New York surgeon named William Howard Hay collapsed running for a train. He was diagnosed with kidney disease and an enlarged heart, and he was frightened enough to overhaul his life. He changed how he ate, and he also quit coffee and quit tobacco. Over about three months he went from 225 pounds to 175, and he felt transformed.

Something clearly worked. But look at what changed at once: his diet, his caffeine, his smoking, and his weight. There is no way to know which of those did the work — and Hay did not try to find out. He concluded it was the food pairing, and he spent the rest of his life saying so.

That is the entire original evidence base. One man, no control, four variables moving together.

His reasoning, in his own words

Hay laid the theory out in 1929. He believed starch digestion needed alkaline conditions and protein digestion needed acid, and that the stomach could not do both at once. His phrase was that no stomach could be both, “just as no room could be at once both light and dark” [5].

If digestion stalled, he wrote, the starch would ferment and the meat would putrefy. That is where “rotting in your stomach” comes from — it is a 1929 prediction, not an observation.

It is a tidy argument. It is also wrong in a way that is now easy to check.

Why the mechanism fails

Three things sink it, and none of them are controversial.

Your pancreas does not take orders. The enzymes that digest starch, protein and fat are made in the same cells, stored together, and released together into the small intestine [2]. There is no switch that sends one and holds back the others based on what you ate.

The conflict Hay described happens in two different organs. Protein digestion starts in the stomach, in acid. Starch digestion in the stomach is only what the saliva enzyme manages while the meal is still buffering the acid — the stomach makes no starch enzyme of its own [2]. The real work on starch happens further down, in the small intestine, where the environment is close to neutral [3]. Adding protein cannot interrupt a job the stomach was never doing.

And the rule cannot actually be followed. Wheat flour is about 10 grams of protein and 76 grams of carbohydrate per hundred grams. Rice is roughly 7 and 80. A potato is 2 and 17 [4]. Every food on the approved “starch” list is also a protein source. If mixing the two caused rot, a slice of plain bread would do it.

The one time anyone tested it properly

In 2000, researchers in Geneva took 54 people with obesity into hospital — so intake was controlled rather than reported — and put them on 1,100 calories a day for six weeks. Half ate separated meals, half ate normal mixed ones. Calories and macronutrients were closely matched, which is exactly the right way to test whether the separation does anything by itself.

The separated group lost 6.2 kg. The mixed group lost 7.5 kg. No significant difference — and the numbers, if anything, ran the other way. Body fat, waist measurements, cholesterol, blood sugar and insulin all improved about equally in both [1].

That trial has never been repeated at that standard. Two smaller studies have looked since, and both landed in the same place. In 2001 a group at the University of Rostock in Germany put twelve healthy adults on a ten-day Hay diet matched to their usual food for calories and protein, and reported no reduction in body fat, lean mass or weight: about 159 lb before and 157 lb after (72.2 vs 71.4 kg) [11]. The Hay diet did raise measured fat oxidation — how much fat the body was burning for fuel — over those ten days. That is the one result a proponent could cite, and it came with no change in body fat to show for it. Then in 2020 researchers in Jakarta ran a randomised crossover in twelve overweight adults aged 18 to 21, ten of them women: three weeks of food combining and three weeks of a plain 1,200-to-1,500-calorie diet, in random order, and no significant difference in weight change between the two [12].

The stomach claim was tested in the 1930s, and failed

The other half of the promise is about digestion: mix protein and starch, the theory goes, and the stomach cannot cope. That was put to the test in people at least twice within seven years of Hay’s book, both times in Philadelphia.

In 1934 a professor of medicine there reported on fifty hospital patients with heart, kidney, lung and other chronic illnesses. One morning each ate about 3½ ounces (100 g) of hamburger steak; the next, the same steak with the same weight of mashed potato. From two hours after the meal, samples were drawn from their stomachs until it had emptied. Steak alone took an average of three hours and fifty-one minutes to leave. Steak and potatoes took three hours and fifty-four, and the acid readings after the two meals were, the author wrote, “strikingly similar” [13]. Steak with potatoes is the first thing this rule forbids. Fifty sick people ate it, and their stomachs took three minutes longer.

In 1936 three more Philadelphia researchers went after the chemistry. Five people free of gut disease, one of whom made no stomach acid at all, swallowed test drinks again and again: plain water, 9 grams of egg-white protein, 15 grams of starch, and the protein and starch together. The team drew more than 900 samples from their stomachs and ran more than 6,000 measurements on them. Their summary: mixtures of carbohydrate and protein “do not, in any way, interfere with gastric secretion”, the acid and enzymes the stomach puts out. The protein was broken down as far in the mixture as alone. And the starch was broken down further with protein in the drink, probably, they wrote, because protein soaks up acid and lets the saliva enzyme go on working [14]. That is the mechanism described two sections up, measured.

The same authors had an explanation for the people who swore by the diet. It is the one this page gives in the next section, ninety years late. Swap concentrated starches for vegetables and fruit, they wrote, and “a weight loss will follow the smaller caloric intake” [14].

Be clear about what these two studies were. They measured stomach contents after test meals, in fifty patients and in five volunteers, and neither was randomised. They were not trials of how anyone feels: neither paper reports asking about bloating, and neither reports anything beyond the stomach. For that promise, that separating foods eases bloating or indigestion, we looked on 30 September 2026 for a trial that assigned people to separated or mixed meals and counted the symptoms, in PubMed, Europe PMC, the ClinicalTrials.gov registry and, for the decades PubMed does not reach, Crossref, and found none. The nearest later test we found was in pigs: a 1996 German experiment in eight sows, in which feeding the protein apart from the rest of the ration made no difference to how much of the feed’s energy they digested [15].

So the digestion half splits in two. The mechanism, that a mixed meal stalls the stomach, was tested and failed ninety years ago. The feeling, that people are less bloated on separated meals, has not been put to a trial that we could find.

But people do lose weight on it

They do, and it is worth being straight about why, because dismissing their experience would be both rude and inaccurate.

A rule that forbids common combinations removes a lot of food from your day. No sandwiches. No pasta with meat. No burger. You are not reorganising your enzymes; you are eating less, because you have made most normal meals against the rules.

The diet works. The explanation for why it works is the part that isn’t true. And that distinction matters, because you can get the same result without memorising a compatibility chart.

The twist: order genuinely matters

Here is the part that makes this more than a simple debunk.

There is real, repeated evidence that the sequence of a meal changes your blood sugar response. Eat the protein and vegetables first and the carbohydrate last, and the glucose spike afterwards drops sharply — in one crossover study of 11 adults with type 2 diabetes, the post-meal glucose level was about 29% lower at 30 minutes, 37% lower at an hour and 17% lower at two hours, and the total rise above the starting level, added up across those two hours, was 73% lower [8]. The same group then repeated it in 15 people with prediabetes, where the peak rise in glucose came down by more than 40% [9].

Whether the order does anything over months is another matter, and the longer trials do not agree. In the longest, 101 Japanese adults with type 2 diabetes were randomised to advice to “eat vegetables before carbohydrate” or to a standard meal plan, and followed for two years. HbA1c, the three-month blood-sugar average, fell from 8.3% to 6.8% with the vegetables-first advice and from 8.2% to 7.3% with the standard plan. But the advice was a package: more green vegetables, fewer fruits, foods that raise blood sugar slowly, chewing every bite more than 20 times. And diabetes medicine was changed along the way, raised in half of the standard-plan group and in 7% of the other. The order of eating is not the only thing that differed [16]. At the other end, sixteen weeks of carbohydrate-last coaching for 45 American adults with prediabetes, on top of standard diet counselling, saw the food-order group lose about 3.6 lb and the counselling-only group about 2.6 lb, with no difference between the groups in weight, HbA1c or glucose tolerance [10].

Pooling does not rescue it. A 2022 review of trials in type 2 diabetes found three that measured HbA1c, 147 people in all, 101 of them from that one Japanese trial, and called the result “slight to no difference” [17]. A March 2026 review pooled four, 211 people, and got a difference of 0.16 percentage points [18]; the trial carrying 70.6% of that result tested a drink of whey protein and guar, a thickening fibre, taken before meals, not the order of a meal [19]. The newest review we found, from September 2026, pooled 18 randomised trials of eating vegetables first and rated every one at moderate or high risk of bias. Its pooled results are for blood sugar and insulin in the two hours after a meal, all at very low certainty; on HbA1c its abstract quotes that same Japanese trial’s figures and says such results are hard to interpret because of the trials’ limitations [20].

These are small studies, and those described above were all in people with type 2 diabetes or prediabetes; they measure blood sugar rather than fat loss. For people whose blood sugar is already normal, a 2026 review found six studies, 107 healthy adults in all, lasting three to seven days; most saw a smaller rise in blood sugar after the meal when the vegetables, fruit or protein came first, and the review asks for longer trials [21]. But notice what they require: protein and carbohydrate in the same meal, minutes apart.

So the genuine finding is the opposite of the folk rule. Keep them on the same plate. Just eat them in a sensible order.

Where the idea went after Hay

The rules were formalised into a system in 1951 and reprinted for decades. In 1981 the Beverly Hills Diet added the claim that fruit enzymes could undo the fattening effect of other foods; it drew a formal medical rebuttal documenting real harms, including severe diarrhoea and dangerously low potassium [6].

Then in 1985 Fit for Life took it mainstream — reportedly 1.8 million copies in print within a year [7], built on the claim that wrongly combined food rots inside you. Its authors’ nutrition qualifications came from an unaccredited correspondence school.

Today it arrives as a 21-day protocol on your phone. The rules are Hay’s, unchanged. The mechanism is Hay’s, unchanged. The evidence arrived in 1934, in 1936 and in 2000, and each time it went the other way.