Bone is load-bearing tissue that adapts to the load it carries. Take a sixth of a person’s body mass away over a year and the skeleton recalculates, which is why every form of rapid weight loss ever studied — dieting, bariatric surgery, and now this drug class — has a bone question attached. It matters more here than the arithmetic suggests, because the people buying these prescriptions skew older and female, and that is the group already losing bone for other reasons. The lean-tissue half of the same problem is in the muscle article.
The one randomized trial that scanned bone in four arms
A predefined secondary analysis of a Danish randomized trial is the clearest evidence in this field. 195 adults aged 18 to 65 with a body-mass index of 32 to 43 and no diabetes completed an eight-week 800 kcal diet, then were randomized for 52 weeks to a moderate- to vigorous-intensity exercise program, liraglutide 3.0 mg daily, both combined, or placebo. Bone mineral density was measured by dual-energy X-ray absorptiometry at hip, lumbar spine and distal forearm, from before the diet to the end of treatment.[1]
Weight lost across the whole study ran 7.03 kg on placebo, 11.19 kg with exercise, 13.74 kg with liraglutide and 16.88 kg with both.[1] The bone result does not follow that order.
In the combination group — the arm that lost the most — density was unchanged against placebo at the hip (mean change, −0.006 g/cm²; 95% CI, −0.017 to 0.004; P = .24) and at the lumbar spine (−0.010 g/cm²; 95% CI, −0.025 to 0.005; P = .20). Against the exercise group, the liraglutide-alone arm finished lower at the hip (−0.013 g/cm²; 95% CI, −0.024 to −0.001; P = .03) and at the spine (−0.016 g/cm²; 95% CI, −0.032 to −0.001; P = .04).[1]
So the arm that lost 16.88 kg held its density and the arm that lost 13.74 kg did not. How the weight came off decided the bone result, not how much of it came off. That is the single most useful sentence available on this question, and it comes from one trial in 195 people with a mean age of 42.8.
A second trial, older people, and the opposite conclusion about training
The obvious inference — add resistance training and the problem is handled — has been tested directly and did not hold. A 2025 single-blind randomized trial enrolled 150 adults with obesity, mean age 66.4, of whom 112 (74.7%) were women, and assigned them for 12 months to caloric restriction alone, caloric restriction plus a weighted vest worn about seven hours a day, or caloric restriction plus supervised progressive resistance training three times a week.[2]
Weight loss was similar across arms, from 9.0% to 11.2%. Total hip trabecular volumetric density fell significantly in every group, by 1.2% to 1.9%, with no difference between the vest and control arms (+0.91 mg/cm³; 97.5% CI, −0.27 to 2.09; P = .13) and the vest non-inferior to resistance training. Areal density behaved the same way.[2] The authors conclude that exercise may be insufficient on its own to offset weight-loss-associated bone loss in older adults.
Two trials, two answers, and the difference between them is the population: 42-year-olds in the first, 66-year-olds in the second. No trial has yet combined a GLP-1 with resistance training in an older cohort, which is precisely the combination this market sells. The general case for training alongside treatment is made in the exercise article, and it is a case about muscle and function rather than a proven case about bone after 65.
Protein moved strength, and strength is not density
A 2026 randomized trial assigned 187 older adults with overweight or obesity, through six months of caloric restriction with aerobic exercise and a twelve-month maintenance phase, to protein at the recommended dietary allowance of 0.8 g/kg/day or to 1.2 g/kg/day for either the weight-loss period or the full 18 months. Hip strength was estimated by finite element modeling of a sideways fall.[3]
At six months, hip bone strength rose 3.8% in the 18-month higher-protein group against 0.5% at the recommended allowance (p = 0.02), despite similar weight loss across groups of −8.0%. Group differences at 18 months were not significant.[3]
The same paper contains the complication. Greater weight loss was associated with greater gains in hip bone strength at six months (p = 0.007) and with greater losses of trabecular density at 18 months (p = 0.011) and of areal density at both six and 18 months (p < 0.001).[3] Strength and density are different measurements of the same hip, and in these participants they pointed opposite ways. A page quoting only the density direction would describe a skeleton getting weaker when the modeled strength was rising; a page quoting only strength would do the reverse. Protein intake alongside treatment is covered more generally in the protein article.
What the pooled drug evidence says, and how sure it is
A 2026 systematic review and meta-analysis gathered 60 articles — 46 randomized trials, 13 real-world studies and one pharmacovigilance study, covering 1,250,717 individuals — on the musculoskeletal effects of this drug class.[4]
On bone it found no effect: neither on bone mineral density at any site nor on fractures at any site, once the models used the most adjusted effect estimates. On muscle it found a consistent decrease in lean body mass, a standardized mean difference of −0.52 (95% CI, −0.80 to −0.23) across 28 comparisons, with heterogeneity of 88% and robustness across every sensitivity analysis.[4] Its own certainty grading for the muscle finding was low, and it calls for models that account for confounding.
A null in a meta-analysis is not a clean bill of health. It is the average of studies that mostly were not designed to measure bone, in populations mostly younger than the ones at fracture risk, over durations mostly under two years. The randomized trial above found a significant hip and spine difference between two arms; the pooled model, drawing on different studies, did not. Both statements are true at once.
Fractures, and the group most likely to be reading this
Fracture is the outcome that matters, and the largest study of it is a 2026 target trial emulation in adults aged 50 to 90 with type 2 diabetes who newly started either a GLP-1 receptor agonist or a DPP-4 inhibitor. After matching, 133,606 patients were analyzed, and three-year fragility fracture risk was lower on the GLP-1: hazard ratio 0.79 (95% CI, 0.76 to 0.83).[5]
The stratification is what a cash-pay reader needs. Among patients with type 2 diabetes the reduction held at 0.91 (95% CI, 0.88 to 0.95). Among those without diabetes the association reversed, to a hazard ratio of 1.13 (95% CI, 1.04 to 1.23), with an interaction P < .001.[5] Almost nobody buying a compounded prescription online has type 2 diabetes. This is one retrospective database study whose authors call for prospective work, and it is the only fracture number in the literature that splits by the indication most of this market is actually using.
The other reason density readings should be handled carefully is that they are a poor proxy for fracture in exactly this population — the fracture analysis across the menopause transition is set out in the menopause article, and the age-specific reading of the fracture data is in the article on older adults.
What follows from all of it
The defensible summary is short. Losing weight on this drug class without exercise cost more hip and spine density than losing weight through exercise, in the one trial designed to compare them, and combining the two preserved density while producing the largest weight loss of any arm. In older adults, structured resistance training did not prevent hip bone loss during caloric restriction — and that trial did not involve a GLP-1 at all. Higher protein intake improved modeled hip strength over six months and not over eighteen. And nobody has run the trial that matters most: a GLP-1, with and without resistance training, in postmenopausal women.
A baseline bone assessment before treatment is therefore a prescriber conversation rather than a formality, particularly for anyone shopping the programs on the board marketed to women. Every figure above was produced with branded product at labeled doses, and no bone measurement has been published on a compounded preparation. Compounded drugs are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed, as the compounding article sets out. How a figure here is established before it is published is described in the methodology.