Body-composition scans report how much lean mass a patient loses on semaglutide or tirzepatide. They do not report whether the patient got weaker. That can be measured in about fifteen minutes.
Ask a patient on a GLP-1 drug how treatment is going and you will hear a number from the scale. Ask the clinic and you may hear a number from a body-composition scan. Neither number says whether the patient can still climb the stairs.
That gap matters more now than it ever has. GLP-1-based drugs produce large, sustained weight loss, and public debate has settled on one question: how much of the loss is muscle? Body-composition studies answer part of it. In the SURMOUNT-1 substudy, tirzepatide reduced lean mass by 10.9% and fat mass by 33.9% over 72 weeks, and about a quarter of the weight lost was lean mass.1 A 2026 systematic review of 35 randomized trials found a median of 28.3% of weight lost from muscle-related measures. Not one of those 35 trials measured objective physical function.2 Of 417 registered randomized trials of these drugs, fewer than 5% measured physical performance and fewer than 5% measured muscle strength.3
So we know, roughly, how much tissue goes. We mostly do not know whether strength goes with it.
Mass is not strength
The working assumption is that lean mass stands in for strength. In older adults it fails even without a drug. In the Health, Aging and Body Composition Study, leg strength fell by 2.6% to 4.1% a year, about three times faster than leg lean mass, and gaining lean mass did not keep strength.4 In the same cohort, strength, not muscle mass, was associated with mortality.5 Pooling cohort data, the Sarcopenia Definitions and Outcomes Consortium found that DXA lean mass was not associated with adverse outcomes, while weak grip combined with slow gait predicted falls, hip fracture, mobility limitation and death.6,7
DXA lean mass is not even muscle. It also counts water, organs and connective tissue, and it shifts with hydration and with the weight lost. A patient can keep a good lean-mass share on the scan and still need the rail on the stairs. After thirty-two years in rehabilitation medicine, I watch the stairs more than the scan.
The evidence runs both ways
Function on these drugs is not uniformly worse. Randomized trials mostly measured walking, and walking improved: semaglutide lengthened the six-minute walk in heart failure with preserved ejection fraction and in knee osteoarthritis.8,9 But walking distance in those patients depends on the heart, the knee and the weight being carried, so a longer walk does not show that strength was kept.
Studies that measured strength point both ways. In the SEMALEAN cohort, handgrip strength was about 4 kg higher at 12 months on semaglutide.10 In a cohort of older men with type 2 diabetes, grip strength, gait speed and Short Physical Performance Battery scores fell on semaglutide.11 Neither was randomized. In a 52-week randomized trial after diet-induced weight loss, liraglutide alone did not improve stair-climb time or fitness; exercise did.12
The honest summary is that the question is open. And what is missing is narrower than the debate suggests: I found no study reporting how often, within one patient, strength or performance falls while lean mass looks reassuring.
Measurement error is the missing rule
Recent recommendations already advise assessing strength and function during treatment.13,14 What I did not find in any of them is a rule for telling a real decline from noise. Every functional test has measurement error. For four-meter gait speed in older adults, the 95% minimum detectable change was 0.108 m/s in intermediate-speed walkers,15 while a small meaningful change is near 0.05 m/s and a substantial one near 0.10 m/s.16 A decline that matters can hide inside the noise of a single visit. Without a rule for error, a clinic will either chase random wobble or miss real loss.
A practical rule
In a review just published in Diabetes, Obesity and Metabolism,17 I propose a definition, functional discordance: a confirmed decline, within one patient, in strength or physical performance beyond the test’s measurement error, without an adverse lean-mass signal. The monitoring is simple.
-
Who. Patients at higher risk: chiefly older adults, and those with low strength, frailty or rapid weight loss.
-
What. Grip strength, five-times chair-stand time and usual gait speed, with a timed stair climb where a flight is available. About fifteen minutes with a dynamometer, a chair and a marked walkway.
-
When. Before treatment, after a practice session; at 3, 6 and 12 months; and around dose changes.
-
How to read it. Count a decline only when it exceeds the test’s error and a repeat session within two weeks confirms it. Record at every test whether the patient used a rail, a cane or a helping hand, because a score earned with more help is not the same score.
-
What then. Look for causes, restore load with progressive resistance and power training, check protein intake, and review the pace of dose escalation.
The schedule is pragmatic and not validated. Stating it plainly makes it testable. If lean-mass change reliably identifies the patients whose function declines, functional monitoring adds little. If discordance is common, the scan is a poor stand-in for the capacity patients actually live on.
For trialists and sponsors
The ask is modest. Add grip, chair stand and gait speed to the visits where DXA is already done, score them against measurement error, and report within-person discordance, not only group means. The cost is minutes per visit. The return is an answer to the question patients and prescribers are already asking.
The goal of treatment is a lighter patient who can still do what life asks. We should measure the second half of that sentence.
Author Bio















