Does Exercise Really Help You Keep Weight Off? We Checked the Study Everyone Cited

The story starts with an explainer written by Rachel Woods, a senior lecturer in physiology at the University of Lincoln, and published by The Conversation under the title "The exercise paradox: why workouts aren't great for weight loss but useful for maintaining a healthy body weight." It is a clear, sober piece of science writing, and its dateline reads October 20, 2025. It is not new research, and it was not new in August. It simply re-circulated.

What carried it back into everyone's feed was a repost. The syndicated version ran on ScienceDaily on August 22, 2026, under a punchier headline, and from there it went everywhere — aggregators, newsletters, hospital social accounts, the works.

Inside that piece is one sentence doing all the heavy lifting. Verbatim: "In a study of over 1,100 people, physical activity was shown to have little effect on the amount of weight a person initially lost. However, doing higher levels of activity after losing weight was strongly linked to maintaining the weight loss." That is the entire evidentiary spine of the news cycle, and helpfully, the words "maintaining the weight loss" are hyperlinked. Follow that link and you land on a specific paper. So we did.

We have gone looking at popular health advice this way before, and the habits that don't survive a closer look are rarely the ones you would expect.

The link leads to a 2021 paper in Frontiers in Nutrition by Marleen van Baak and colleagues, examining physical activity, weight loss, and weight maintenance within a large European trial called DiOGenes. And the figure of 1,121 does appear in it — once, in the Subjects section, describing how many people were enrolled in the parent trial. It is not the number of people in any analysis the paper actually reports.

The paper is completely open about this. In the researchers' own words: "However, not all data were available at all time points, especially the data on PA were missing in many participants. Therefore, the number of subjects will differ in the different analyses. Complete body weight and PA data were available for 193 participants (76M/117F)." Nothing was hidden. The caveat was simply left behind somewhere between the paper and the feed.

Once you start counting, six different sample sizes appear across the paper's analyses, and the one that matters for the maintenance claim is not the largest.

The Figure That Matches Nothing

There is a second wrinkle, and it is the one we cannot resolve. The parent DiOGenes trial published its own participant flow in the New England Journal of Medicine in 2010, and that report states that 1,209 adults were screened, 938 entered the low-calorie-diet phase, 773 were randomly assigned to a maintenance diet, and 548 completed the study. None of those numbers is 1,121, and neither paper explains the gap. We are reporting it as an open discrepancy rather than guessing at a cause, because guessing is how this kind of thing starts.

None of this makes the paper a bad one. It makes it a different kind of paper than the coverage implied, and the difference matters in three specific ways.

Nobody Was Assigned to Exercise

DiOGenes was a diet trial. Participants were randomly assigned to different eating patterns, never to different amounts of movement. The activity analysis looked at how active people already happened to be and searched for patterns in the results. That is an observation living inside a randomized trial, which is a genuinely useful thing — but it cannot tell you whether the activity caused the outcome, or whether the kind of person who stays active is also the kind of person who holds onto a weight loss.

The Measurement Was a Questionnaire

Activity was self-reported, using a standard survey that asks about work, sport, and leisure and scores each on a five-point scale. Nobody wore an accelerometer. Nobody's energy expenditure was measured. What the study captured was how much activity people said they did, which is a reasonable proxy and a long way from a stopwatch.

There Is No Dose in It

This is the part most likely to disappoint. The maintenance result, available in full through Europe PMC, is expressed as kilograms of body weight per one point on that questionnaire score. It does not convert into minutes, steps, sessions, or calories. There is no version of this study that tells you how much to move, which is worth knowing before you go looking for a number in it.

Here is the part that reframed the whole piece for us. Buried in the abstract of the very paper being used as proof is a sentence the authors chose to include, and it is a caution against reading their result the way the news cycle read it.

They wrote, in the paper's abstract: "The positive effect of higher PA on weight loss maintenance seems in contrast to randomized controlled trials that have not been able to confirm a positive effect of exercise training programmes on weight loss maintenance."

That is the researchers telling you, in their own abstract, that their observational finding does not match what happens when you actually assign people to exercise. They were transparent about their missing data, transparent about their design, and transparent about the tension in their own result. Every one of those hedges was intact at the source. None of them survived the trip.

So does exercise help you keep weight off or not? Yes — and this is the genuinely interesting part, because the case is stronger than the study everyone cited, and newer.

In June 2026, a systematic review published in Scientific Reports pooled eleven randomized controlled trials covering 568 people, all of them in the maintenance phase after losing weight. The exercise groups regained 2.81 kg less than the control groups, with a confidence interval running from 5.12 kg down to 0.51 kg. More striking than the effect itself is its consistency: the heterogeneity between trials measured zero, meaning the eleven studies essentially agreed with each other. That is uncommon in this literature, and it is the single best piece of evidence for the headline. It was published two months before the news cycle and, as far as we can tell, appeared in none of the coverage.

Then there is the year-long trial. In Copenhagen, 195 adults who had lost about 12% of their body weight on a very low-calorie diet were randomly assigned to four groups and followed for 52 weeks. Over that year the group left to its own devices regained 6.1 kg. The group assigned to a supervised exercise program regained 2.0 kg — a difference of 4.1 kg in favor of exercise. Their cardiorespiratory fitness rose too, while the comparison group's drifted down. Two of the four arms in that trial involved a weight-loss medication as a comparison condition, which is a separate story and not this article's.

And Then Everyone Stopped

The most quietly persuasive result came a year later. The same Copenhagen participants were reassessed twelve months after all treatment ended, with the researchers deliberately staying out of touch to see what real life did. The people who had been in the exercise program regained 3.6 kg over that unsupervised year. The people who had been on medication alone regained 9.6 kg, according to the follow-up analysis. At that final visit, the former exercise group was still logging a median of 240 minutes of moderate-to-vigorous activity a week. The habit outlived the study. It is worth noting that this follow-up was observational rather than a fresh randomization, and that the people who showed up for it had responded better during treatment than those who didn't, so the comparison flatters the result somewhat.

Movement seems to keep paying out in ways the scale never captures, which is also the through-line of what mixed exercise does for longevity.

Five distinct claims circulated in this news cycle, and they are not equally supported. We graded each one against the best available evidence, using a fixed four-point scale that we will carry across future analyses like this one.

Two rows deserve a note. The first grade rests on an editorial in the International Journal of Obesity by Eric Robinson and David Stensel, who reviewed the umbrella evidence and concluded that activity produces real if modest weight loss, working best alongside dietary change rather than instead of it. Modest is not nothing.

The muscle row comes from a 2025 meta-analysis in BMJ Open Sport & Exercise Medicine covering 25 randomized trials and 1,608 people. Resistance training protected fat-free mass and increased fat loss without changing body weight at all — but the authors are explicit that the muscle-sparing effect showed up only in interventions running under five months, which they suspect reflects how hard the programs were to stick with. That caveat almost never travels with the claim.

If you want the fuller picture of what regular resistance work actually changes, we've covered what happens to your body when you lift weights regularly in more depth.

There is a tidy physiological explanation for the whole paradox, and it arrived this year. A review in Current Biology by Herman Pontzer and Eric Trexler re-analyzed fourteen human exercise interventions across 21 cohorts and 450 participants, and found something counterintuitive: total daily energy expenditure rose by only about 30% of what simple arithmetic predicted. In eleven of nineteen cohorts, daily energy expenditure did not measurably change at all.

The body, in other words, quietly reclaims most of what a workout is supposed to spend. And the compensation grows in step with the exercise dose, and grows further when exercise is paired with eating less — which is precisely the situation someone trying to lose weight is in. As the authors put it, with calorie restriction the compensation "may be greater, reducing or eliminating the impact of exercise on energy balance and weight loss." Take away the deficit, and there is far less for the body to defend against. That asymmetry is a plausible reason the same behavior underperforms during a diet and earns its keep afterward.

It also explains why the returns on exercise so rarely scale the way people assume, a pattern we ran into when looking at how much a short, hard workout can actually deliver.

The other number attached to this story deserves the same treatment. It comes from the National Weight Control Registry, a long-running database of people who have lost significant weight and kept it off, whose facts page reports that "90% exercise, on average, about 1 hour per day" and that "94% increased their physical activity." Those two figures describe different things — the first is about maintenance, the second about how members lost the weight in the first place — and they get merged constantly.

The peer-reviewed version is more careful. Writing in the American Journal of Clinical Nutrition in 2005, Rena Wing and Suzanne Phelan reported that women in the registry logged about 2,545 kcal a week in activity and men about 3,293, levels they characterized as roughly an hour a day of moderate activity such as brisk walking. Walking was what most people reported doing. But the registry is self-selected, self-reported, and has no comparison group — it describes people who succeeded, which is not the same as showing what made them succeed.

What Measured Activity Shows

The strongest objective check on that number came in 2019, when researchers used doubly labeled water — the gold standard for measuring energy expenditure — on 80 adults. People holding an average 26 kg weight loss for roughly nine years expended 812 calories a day in physical activity compared with about 620 to 640 in two control groups, and averaged 12,107 steps a day against roughly 8,900 and 6,600. The gap is real and it is measured rather than remembered. It is also cross-sectional, so it still cannot tell you which way the arrow points.

Round numbers in health tend to have stranger origins than their confidence suggests, as we found when tracing the actual history of the 10,000-step target.

None of this required special access. The study was free to read, the abstract was three paragraphs long, and the sample sizes were printed in plain text. What it took was following the link.

If you want a way to test the next health claim that lands in your feed, these are the four questions that would have caught this one:

• What is the actual number of people behind this specific finding? Not the trial, not the enrollment — the analysis being described.
• Was the thing being credited assigned, or just observed? Assigning people to a behavior and watching what happens is a different order of evidence from noticing who already does it.
• How was it measured? A questionnaire, a wearable, and a laboratory measurement of energy expenditure are three different levels of confidence wearing the same word.
• Do the researchers themselves agree with the headline? Abstracts frequently contain a sentence of caution that never makes it into the coverage. It is usually near the end.

The reassuring thing is that the answer here was not a debunking. Exercise really does appear to help people hold onto a weight loss, and the trials that test it directly say so more convincingly than the study that got the attention. The headline was fine. It just deserved better homework — a pattern that shows up again and again in the nutrition myths that refuse to die.