The muscle loss figure nobody checked

The figure travels everywhere: roughly 40 percent of the weight lost on semaglutide is lean mass. It is quoted in newspapers, on clinic websites and by people selling you protein. Almost nobody quoting it has asked the obvious follow-up, which is what those people were eating.

Nobody asked because, in most of the trials, nobody measured.

The hole at the centre of the evidence

A systematic review published in January went through 41 randomised controlled trials of liraglutide, semaglutide and tirzepatide, 50,690 participants between them, looking for one thing: did the trial assess what people actually ate?

Two of the 41 did. About five percent.

A separate scoping review went wider, across 129 trials of injectable weight-loss drugs. Fifty-seven described lifestyle modification as part of the intervention. Thirty-six recorded something about diet. Ten reported the outcome, and half of those ten used a single sitting, a buffet or an ad libitum lunch, measured once.

Sit with that, because it changes what the famous number means.

The 40 percent is not a measurement of what the drug does to muscle. It is a measurement of what happens when tens of thousands of people take an appetite suppressant and nobody checks their protein. Those are different claims, and only the second one is supported by the design.

What the number actually is

It comes from a dedicated body composition substudy inside STEP 1, 140 participants scanned by DEXA. Fat mass fell 8.36 kg on average. Lean mass fell 5.26 kg. That puts the lean share at about 39 percent.

Two things get lost every time it is repeated.

Lean mass on a DEXA scan is everything that is neither fat nor bone: skeletal muscle, yes, but also organs, connective tissue and the water held in all of it. A body 20 kg lighter needs less plumbing.

And the share of the loss is not the composition of what remains. Far more fat came off than lean tissue, so the proportion of the body made of lean tissue went up, by around three percentage points. Both statements are true at once. Only the alarming one gets printed.

Then somebody measured properly

A group at Rouen ran the study the trials did not. SEMALEAN followed 115 patients on semaglutide 2.4 mg through a year of ordinary clinical care, with DEXA, handgrip strength and resting energy expenditure at baseline, seven months and twelve. Mean BMI at the start was 46.3. A hundred and six finished.

Weight fell 10 percent by month seven, 13 percent by month twelve. Fat mass fell 14 percent, then 18.

Lean mass fell 3 kg by month seven. Then it stopped.

From seven months to twelve the fat kept going and the lean tissue held. Whatever was happening to muscle was happening in the first half of the year, and then the body stopped settling the bill from that account.

The finding that should be better known

Handgrip strength went up. By 3.7 kg at month seven, and 4.5 kg by month twelve.

Not held. Not declining slowly. Up, by a margin nobody would call trivial, in a cohort that had shed 13 percent of bodyweight and 3 kg of lean tissue getting there.

And the figure I keep returning to: the proportion of that cohort meeting the criteria for sarcopenic obesity fell from 49 percent to 33 percent. Half of them began the year with the condition everybody is afraid this drug causes. A third of them ended it that way.

It is not an isolated result. A systematic review of GLP-1 therapy in fatty liver disease, published in Liver International this year, pooled twelve studies covering 810 people. Where strength was measured, in three studies covering 477 of them, there was no deterioration in grip or in sit-to-stand performance despite meaningful weight loss. A real-world cohort from the obesity unit at Vall d'Hebron in Barcelona found absolute grip preserved and relative strength improved.

Something is going on that the lean mass number does not describe.

The explanation: the body spends the worst muscle first

This is the part worth understanding properly, because it reframes everything above.

Muscle in a person carrying a great deal of excess weight is not uniformly muscle. Fat infiltrates it, sitting between and inside the fibres, and the technical name for that is myosteatosis. Infiltrated muscle contracts worse, takes up glucose worse and is perfused worse. It weighs the same on a scan as healthy tissue and does considerably less work.

Imaging studies keep finding the same thing. In the post-hoc MRI analysis of SURPASS-3, published in Lancet Diabetes and Endocrinology, tirzepatide reduced muscle fat infiltration across every dose while muscle volume fell only modestly and in proportion to the weight lost.

The detail worth pausing on: the drop in fat infiltration was larger than weight loss alone would predict. Something happened to the quality of the tissue beyond simply having less of the person.

A separate MRI study of the thigh on liraglutide found the same direction, with muscle fat infiltration down and the proportion of adverse muscle composition reduced, even as the limb itself got smaller.

The working hypothesis, and it is a hypothesis rather than a settled fact, is that the tissue carrying the most intramuscular fat, which is the least useful tissue you own, is the tissue preferentially broken down first.

If that is right, then a DEXA scan showing lean mass down and a dynamometer showing strength up are not in conflict at all. You lost the bad muscle. What is left is a smaller, cleaner, better-perfused engine, and it grips harder than the larger dirty one did.

Quality is not the same as quantity, and for function it is the more important of the two.

The trial that proves the two can be separated

Then there is BELIEVE, published in Nature Medicine this year, and it is the most interesting thing in this field.

Bimagrumab is an antibody that blocks activin type II receptors. Activin signalling tells muscle not to grow; block it and the brake comes off. It has nothing to do with appetite and does not touch food intake at all. It works directly on fat and muscle tissue.

The phase 2b trial ran 72 weeks and compared semaglutide alone, bimagrumab alone, and the two together.

• Semaglutide 2.4 mg on its own: lean mass down 7.4 percent.
• Bimagrumab on its own: lean mass up 2.5 percent.
• The two combined: lean mass down only 2.9 percent, and body weight down 22.1 percent.

Of all the weight lost on the combination, 92.8 percent of it was fat. Visceral fat fell 58.2 percent against 35.8 on semaglutide alone. The inflammatory marker hsCRP fell by up to 84 percent.

And the detail that carries the argument is the one I nearly got wrong. The extra fat loss on the combination did not come from eating less. By week 24 the calorie reduction was near identical in both arms: a median of 482 kcal on semaglutide alone against 487 on the combination.

Same deficit. Same hunger. Wildly different bodies at the end of it.

Which settles a question that had been open. Fat loss and lean loss are not welded together. They can be uncoupled, and somebody has now done it in humans.

There is something quietly absurd about the pharmaceutical answer to muscle loss on a drug being a second drug, delivered intravenously, with muscle cramps and acne as its signature side effects and up to a fifth of participants stopping it. Bimagrumab is not available and is not your answer this year. But the principle it demonstrates is the reason to bother with protein and loading, because those are the versions of the same lever that you can actually pull.

The one place this might go wrong

Here is something I have not seen mentioned in a single article written for patients, and it is the part that concerns me most given what I do for a living.

A preprint posted this year took the question nobody had asked: what happens in muscle that is actively repairing? The researchers injured muscle in mice, then treated them with semaglutide. In the injured muscle, intramuscular fat formation increased markedly and the growth of regenerating fibres was inhibited. In uninjured muscle from the same animals, nothing.

Every reassuring imaging study above was done on intact muscle. Nobody has looked at the repairing kind.

Now the caveats, and they are large enough that you should weigh them before the finding. Mice, not people. A preprint that has not been through peer review. An artificial chemical injury rather than a torn muscle. This is about as preliminary as evidence gets, and none of it transfers to a hamstring in a human on Wegovy. I am not going to pretend otherwise, and if it fails review I will take this section down.

But if you are on one of these drugs and you are recovering from surgery or a significant injury, this is a question worth putting to your doctor rather than to the internet. It is the one scenario where the general reassurance may not hold, and it is the exact scenario a nutritionist who works in sport ends up seeing.

What actually has evidence behind it

The uncomfortable part, which you will not read on the sites selling the solution.

There is no completed randomised trial of protein supplementation during GLP-1 therapy. The single most-repeated piece of advice in this field has never been tested in the population it is aimed at. Several trials are registered, LEAN-PREP among them, and the results are still ahead of us.

What does exist is adjacent and good. Verreijen and colleagues, 2015, put 80 obese older adults through 13 weeks of hypocaloric dieting with resistance training and gave half of them whey protein with leucine and vitamin D against an isocaloric control. The supplemented group finished up 0.4 kg of appendicular lean mass. The control group finished down 0.5 kg. Fat loss was identical.

Nearly a kilogram of difference in the arms and legs, from the supplement alone, in thirteen weeks.

What is not in doubt is the shortfall. A cross-sectional study of 60 people on these drugs, using three-day food records, found 43 percent reaching 1.2 g of protein per kg per day, 10 percent reaching 1.6, and 5 percent reaching 2.0. The protective dose is being missed by nine people in ten.

A separate survey of 69 current users found the same pattern in the micronutrients: calcium, iron, magnesium, potassium and vitamins A, C, D, E and K all below reference intake, alongside fibre, while calories from fat, saturated fat and sodium sat above. Sixty-nine people is a small sample and I would not quote it as settled, but it matches the food logs in front of me.

The practical half of this, including why the daily total is the wrong number to chase, is on its own page.

So why does the gym feel harder?

Because feeling weaker and being weaker are separate problems, and after everything above, the second one is not the common one.

A body running on half its usual food is short of glycogen, short of sodium and short of what a hard set costs. The contractile tissue is largely still there. What is missing is what it burns. That presents as heavier weights, fewer reps and a session that ends early, and it is a fuelling problem wearing a strength problem's clothes.

The two have opposite fixes. One is solved by eating around training. The other is not solved by anything quick.

The exception is age. In older people, or anyone who started near the sarcopenia threshold, functional loss is more likely to be real rather than felt, and the reassurance in this article thins out considerably.

What I do with all of this

• The window is early. Lean mass loss is front-loaded into the first six or seven months, which is exactly the period when people are most delighted with the scale and least interested in hearing about protein.
• Mass and function are different measurements. A scan showing lean tissue down and a grip test showing strength up are not contradicting each other, and reading only the first one produces panic that the second one does not support.
• Weekly beats monthly. A four-week gap hides the direction of travel inside normal fluctuation. Weekly, with fat and lean read as separate lines, shows which account the weight is leaving while there is still time to change it. It also decides how hard the weight will be to hold later.

The sentence to keep

The most-quoted number in this field came out of trials that never asked what anybody ate. When people finally measured, the lean tissue loss stopped at seven months, grip strength rose, sarcopenic obesity halved, and the muscle that remained was cleaner than the muscle that went.

That is not permission to ignore it. The protein still has to be eaten and the muscle still has to be loaded, and neither happens by accident in somebody who is not hungry. But the panic is pointed at the wrong target, and the real work is dull, early, and almost entirely about what goes on the plate in the first six months.