Seven runners, 21 days of plant-based meal kits, 3.4 kg less body fat. Nobody weighed what they left on the plate, and their usual diet always came first

Plenty of runners wonder whether they can drop meat without losing anything on the road. A team from two Catholic universities in Brazil, in Campo Grande and Brasilia, gave it a small test, published in Frontiers in Nutrition on 2 November. Of 21 people who answered an online form, 7 men qualified: healthy, aged 25 to 45, eating meat and running for more than two years, with a median of 35 km a week.

The design is simple. Each man was tested at the start, ate his normal diet for 21 days and was tested again, then spent 21 days on a plant-based diet and was tested a third time. A company delivered four meals a day, with fruit, vegetables, whole grains, legumes, seeds and soy foods such as fermented tofu, and no meat, fish, eggs or dairy. The kits were calculated to match the calories and macronutrients each man reported eating, and he was told to eat all of it and nothing else.

After the plant-based weeks, median body fat stood at 9.1% against 13.3% after the normal diet. Fat mass was 5.7 kg against 9.1 kg, and weight 64.3 kg against 68.2 kg. Fat-free mass barely changed, at 60.7 kg against 60.3 kg. Time to exhaustion on the treadmill was longer than after the normal diet, though not significantly longer than at the first test, and the text gives only p-values, no minutes. For anyone hoping to swap the steak without paying for it in muscle, that reads well.

Now the sums. A common rule of thumb puts a kilo of body fat at about 7,700 calories. On that basis, and treating a difference in medians as if it were one man's change, losing 3.4 kg in 21 days would take a shortfall of about 1,250 calories a day (our rough arithmetic). These were lean men, with a median BMI of 23.3, on food meant to replace what they normally ate, calorie for calorie. And the kits were built from a three-day food recall, so if the men's memories fell short, the kits did too. Add that the ultrasound reading may have moved more than the fat did, or that they may have run more, and there's no shortage of explanations.

The paper can't tell us which, because leftovers weren't weighed or photographed and no food diaries were kept during the kit weeks. Contact was by phone and weekly symptom logs, and the men reported sticking to it. The authors themselves raise one explanation: fibre rose by 64.5 g, which may have filled the men up enough that they didn't finish their meals. Their guts noticed. Scores for flatulence went from 4 to 7 and for bloating from 1 to 2. Body composition came from an ultrasound device, and the normal diet was never monitored, just assumed to be stable because recreational athletes tend to be disciplined.

There's also a strange number the authors don't discuss. Median weight fell from 74.9 kg at the first test to 68.2 kg after the normal-diet weeks, a bigger drop than anything in the plant-based phase, before any kit had arrived, though the table doesn't mark that drop as significant. Body fat over that stretch hardly moved, at 13.6% and 13.3%. Medians from a group of seven can jump around like that, because a single man can shift them, but it's a warning not to lean on any one of these gaps. Whatever was going on, it wasn't the plant food.

The order matters as much as the food. Every man did the normal diet first and the kit second, with no randomisation and no washout, so anything that changes with time looks like a diet effect. The treadmill is the plainest case: there was no practice session, and the plant-based test was always a man's third go at it, a point the authors raise themselves. Training was meant to stay steady, but mileage and intensity weren't recorded in either phase.

The blood markers point in no clear direction. TNF-alpha, an inflammation signal, didn't differ between the two diets. An antioxidant enzyme, superoxide dismutase, came out lower after the kits than after the normal diet, but no different from the first test. Nitric oxide and calprotectin both rose significantly with exercise only in the plant-based phase, but those comparisons weren't adjusted for multiple testing. The authors are straightforward about all this. They call the work exploratory and hypothesis-generating, set no primary outcome and did no sample-size calculation, and they ask for randomised crossover trials with washout periods. Their funding came from Brazilian public research bodies, and they declare no conflicts. Seven men got through three weeks of it, with more wind. That's about as far as it goes.