Plenty of protein, and still frail: in 5,679 older Japanese adults, the ones whose plates leaned hardest towards carbs over fat were frail more often. The frailty was measured first, the diet six months later

Frailty is the word doctors use for the slow loss of reserve in old age: the point where a fall, an infection or a week in bed is no longer something a person bounces back from. Eating enough protein is the standard dietary advice for holding it off. The researchers behind this study, from Japan's National Institutes of Biomedical Innovation, Health and Nutrition and two universities, had already noticed something awkward in their own data. Among older people in the top quarter for protein, roughly 25 to 30% were frail anyway. Protein, they point out, usually supplies only 10 to 20% of the energy in a human diet. Carbohydrate and fat share the other 80 to 85%, and how that share is split varies enormously from one person to the next. Their paper on what that split looks like in frail people came out in the European Journal of Nutrition on 7 October 2026.

The data come from the Kyoto-Kameoka Study, which follows people aged 65 and over in Kameoka, a city in Kyoto Prefecture. Residents filled in a 25-question frailty checklist by post, and later a food questionnaire asking how often they had eaten 46 foods and drinks over the past year. The analysis covers the 5,679 people, 2,957 of them women, who answered everything needed; they were in their early seventies on average. From the food answers the team worked out one number per person: energy from carbohydrate divided by energy from fat. A ratio of 2 means twice as many calories from carbohydrate as from fat. A quarter of the group came in below 1.69, and a quarter above 3.02.

Three in ten of them, 1,705 people, scored as frail on the checklist. In the quarter with the lowest ratio it was 26.1%; in the quarter with the highest, 36.1%. After the team allowed for age, sex, body weight, total calories, drinking, smoking, long-term illness, schooling, money worries, living alone and how people rated their own health, the top quarter still had 1.92 times the odds of being frail (the plausible range runs from 1.55 to 2.38). The line rose steadily, with no sign of a safe zone or a tipping point. And it rose the same way in the low, middle and high thirds for protein, which is the paper's main claim: eating a lot of protein did not cancel it out.

It was also not just one corner of frailty doing the work. The checklist covers seven areas, and a higher ratio went with more trouble in six of them: everyday tasks such as shopping and handling money, physical function, chewing and swallowing, going out, memory and low mood. The strongest link was with rarely leaving the house. The one area that showed no clear link was nutritional status itself, which on this checklist means weight loss and a low body mass index.

What the authors do with this is unusual. They write that the results 'do not support recommendations for older adults to lower their CFr, restrict carbohydrate intake, or increase fat intake', and that current advice should not change on the strength of this study. Their reasons are sound. Everything was measured at about the same moment, so nobody can say which came first. A ratio can go up because someone eats more carbohydrate or because they eat less fat, and this one says nothing about whether the carbohydrate was white rice or beans, or the fat was fish or pastry. And a frail person may eat differently because they are frail: poor appetite, sore teeth, trouble getting to the shops or standing at a stove all push towards soft, simple, starchy food.

The rest is our reading. On that last point the dates in the paper matter more than the authors make of them. The frailty checklist went out in July 2011. The food questionnaire followed about six months later, in February 2012. So the diet in this study is what people were eating after their frailty had already been recorded, which is the right order for frailty shaping the plate and the wrong one for the plate shaping frailty. Two other things stand out. Nearly two-thirds of the group (64.6%) said money was tight, and starch is the cheapest way to fill a plate; the analysis allows for that with a single yes-or-no question. And the list of things the model adjusted for has no measure of physical activity in it, although activity is one of the strongest known influences on frailty. The answers are also 14 years old, and from a country where rice sits at the centre of most meals, so the numbers won't carry over neatly to other kitchens. What does carry over is the question: once the protein is on the plate, the rest of the plate may not be neutral. The study was funded by Japanese government research grants and the authors declare no conflicts of interest.