GLP-1s, bone density, and fracture risk: what the evidence shows

Most of the conversation about GLP-1 medications and body composition is about muscle. Bone gets less attention, partly because you cannot see it change in a mirror and partly because the consequences show up years later rather than in the next few months. But the same physiology that makes rapid weight loss costly for lean tissue applies to the skeleton, and unlike some of the connective-tissue questions we covered in our article on tendon and joint risk, this one has real measurements behind it.

This article separates three things that often get blurred together: what has been directly measured in people taking GLP-1 medications, what is reasonably extrapolated from other weight-loss research, and what nobody has established yet. Where the evidence conflicts, we say so rather than picking the more dramatic side.

Why losing weight costs bone at all

Bone is living tissue that continuously remodels itself in response to the mechanical load it carries. Carry less weight, and the skeleton has less reason to maintain the density it built to support the heavier body. Energy restriction also changes the nutritional environment in which remodeling happens.

This is not theoretical. A systematic review and meta-analysis of 41 clinical trials of diet-induced weight loss in adults with overweight or obesity found statistically significant decreases in total hip bone mineral density at 6, 12, and 24 months, on the order of 0.010 to 0.015 g/cm². Lumbar spine and whole-body density showed no consistent significant change over 3 to 24 months, with one exception at 6 months. Early in the interventions, at 2 to 3 months, markers of bone breakdown rose. The authors described the hip losses as small relative to the metabolic benefits of losing excess weight.¹

That is the baseline against which any GLP-1-specific finding should be read. Some bone loss during substantial weight loss is expected, medication or not. The question is whether GLP-1 therapy adds to it.

What has actually been measured in GLP-1 users

The most direct evidence comes from a randomized, double-blinded, two-centre phase 2 trial published in 2024. Sixty-four adults, 86 percent of them postmenopausal women with a mean age of 63, all with a T-score below minus 1.0 at the hip or spine or a prior low-energy fracture, received weekly semaglutide 1.0 mg or placebo for 52 weeks. Semaglutide produced 6.8 kg more weight loss than placebo. It also produced significantly greater declines in lumbar spine areal bone density (a difference of minus 0.018 g/cm², p = 0.007) and total hip density (minus 0.020 g/cm², p = 0.001). Femoral neck density did not differ significantly. Tibial cortical thickness fell by a relative 1.8 percent in the semaglutide group. The trial's primary outcome, the bone formation marker P1NP, showed no difference, but the resorption marker CTX rose significantly in the semaglutide group.²

That is a measured finding, and it is the strongest one available. It is also a small trial in a deliberately high-risk population, and it did not measure fractures.²

Two other GLP-1-specific datasets point in a gentler direction. A post hoc analysis of a 20-week pilot trial randomized 20 older adults, mean age 72.7, with overweight or obesity and prediabetes or type 2 diabetes, to semaglutide 1.0 mg plus lifestyle counseling or lifestyle counseling alone. The semaglutide group lost significantly more weight, minus 5.3 percent versus minus 0.9 percent, but there were no significant between-group differences in whole-body bone density (p = 0.77), regional density at the leg, pelvis, or lumbar spine, or in CTX and P1NP. Weight change did correlate directly with pelvis bone density change (r = 0.62, p = 0.01), which is consistent with the loss being driven by the weight rather than by the drug.³ Twenty participants over 20 weeks is very small and very short, and the absence of a difference here is not evidence that no difference exists.

A larger retrospective study followed 255 patients on semaglutide or tirzepatide, 92 percent female with a mean age of 64, matched to 255 non-users by age, sex, BMI, and diabetes status, with DXA scans before treatment and at least six months after, over a median of 17 months. Both groups lost bone density at broadly comparable rates. Among participants without diabetes, hip bone loss was somewhat greater in the medication group, minus 1 percent versus minus 0.6 percent. Among those with diabetes, hip loss was similar between groups. Weight loss correlated with bone loss.⁴

Finally, a focused meta-analysis of placebo-controlled trials reporting DXA outcomes with GLP-1 receptor agonist therapy screened 2,618 articles and found nine eligible trials. Seven reported lean mass, and only two reported bone density. The authors concluded that roughly 30 percent of the body weight lost on these medications is lean mass, and that more data are needed on bone density outcomes.⁸ That last sentence is the honest summary of the field: bone has simply not been measured often enough.

Measured, extrapolated, or unknown

Measured in GLP-1 users: a decline in spine and hip bone density over 52 weeks of semaglutide in adults at elevated fracture risk, with a rise in bone resorption.² No significant short-term difference in a 20-person pilot.³ Bone loss broadly comparable to matched non-users in a 17-month retrospective DXA cohort.⁴

Extrapolated from other weight-loss research: that hip bone density falls during substantial diet-induced weight loss,¹ that resistance training blunts that loss,⁷ that higher protein intake modestly attenuates it,⁹ and that weight loss reduces calcium absorption.¹⁰ These were studied in dieting adults, not in GLP-1 users. The mechanisms are the same and the extrapolation is reasonable, but it is an extrapolation.

Unknown: whether GLP-1 therapy raises the long-term risk of an actual fracture, and whether the countermeasures below change bone outcomes specifically in people taking these medications. No trial has tested that.

Who is most at risk

Two groups stand out, and it is not a coincidence that they are the same groups the research has focused on. Postmenopausal women already lose bone at an accelerated rate, and the trial that found the clearest semaglutide effect was conducted almost entirely in postmenopausal women who had low bone density or a prior fracture to begin with.² If you are in that group, our article on muscle loss in women over 50 covers the parallel lean-tissue picture.

Older adults are the second group. Bone loss stacks on top of age-related decline, and the same population is losing muscle and balance at the same time, which is what turns low bone density into a fracture. The trial showing that exercise mode changes bone outcomes during weight loss was conducted specifically in older adults with obesity.⁷ Our guide to coming off a GLP-1 after 65 covers the related transition question.

The fracture evidence genuinely conflicts

Bone density is a surrogate. What actually matters is whether people break bones, and here the literature does not agree with itself.

On one side, a 2025 meta-analysis pooled 44 randomized controlled trials involving 47,823 patients with type 2 diabetes and 292 incident fractures. The pooled relative risk for fracture with GLP-1 receptor agonists versus placebo or other antidiabetic drugs was 0.77 (95 percent CI 0.61 to 0.96), a statistically significant reduction. The benefit appeared only in trials longer than 78 weeks, and in drug-specific analysis only liraglutide reached significance (RR 0.42, 95 percent CI 0.21 to 0.85). The authors noted that further study is needed before drawing a clinical consensus.⁵

On the other side, a large retrospective cohort presented at the 2026 Annual Meeting of the American Academy of Orthopaedic Surgeons matched 73,483 patients with type 2 diabetes and obesity treated with a GLP-1 receptor agonist to controls on age, sex, race, BMI, HbA1c, tobacco use, and comorbidities including baseline osteoporosis. At five years, GLP-1 users had a significantly higher rate of osteoporosis, 4.1 percent versus 3.2 percent, a risk ratio of 1.29, along with higher rates of gout and osteomalacia.⁶ That is conference-presented data that has not completed peer review.

These findings are not directly comparable. One pools randomized trials in which fracture was a rare adverse event the trials were not designed to detect; the other analyses diagnosis codes for osteoporosis, which is not the same outcome as a fracture. Both can be correct and still leave the practical question open. Fracture risk on GLP-1 therapy has not been established in either direction.

What protects bone

The interventions with actual randomized evidence behind them are the same ones that protect muscle, which is convenient but not accidental. Bone and muscle respond to the same signal.

Mechanical loading. In a randomized controlled trial, 160 older adults with obesity lost roughly 10 percent of body weight over six months while assigned to aerobic exercise, resistance training, or both. Hip bone density fell 2.6 percent in the aerobic group, 1.1 percent in the combined group, and 0.7 percent in the resistance group, a statistically significant difference between groups (p = 0.001). Resistance and combined training were also associated with a smaller weight-loss-induced rise in bone turnover.⁷ Separately, in the LIFTMOR trial, 101 postmenopausal women with low bone mass were randomized to eight months of twice-weekly supervised high-intensity resistance and impact training or a low-intensity home program. The training group gained 2.9 percent in lumbar spine bone density while controls lost 1.2 percent, and gained 0.3 percent at the femoral neck while controls lost 1.9 percent, with a single minor adverse event across the whole program.¹¹ If you have never lifted, our beginner's resistance-training plan is the place to start. Resistance training also prevents a substantial share of the muscle loss that accompanies caloric restriction in older adults.¹²

Protein. A meta-analysis of 10 randomized trials found that higher-protein weight-loss diets modestly attenuated bone loss compared with normal-protein diets: total body density 0.006 g/cm² better (p = 0.016) and lumbar spine 0.017 g/cm² better (p = 0.019). The authors were candid that a higher-protein diet does not prevent bone loss during weight reduction, only softens it, and that the clinical significance is questionable.⁹ Protein has a much stronger evidence base for muscle, where supplementation meaningfully improves resistance-training outcomes.¹³

Calcium and vitamin D status. Weight loss itself reduces how much calcium you absorb. In a randomized, double-blind trial in 82 postmenopausal women, weight loss decreased true fractional calcium absorption, while vitamin D3 supplementation increased it. The authors concluded that with a calcium intake of 1.2 g per day, either 10 or 63 micrograms of vitamin D per day was sufficient to maintain calcium balance.¹⁰ That was studied in dieting women, not GLP-1 users, but reduced food volume on a GLP-1 makes adequate intake harder for the same arithmetic reason it makes protein harder. Our GLP-1 micronutrient checklist covers vitamin D alongside B12 and iron. Supplement decisions belong with your clinician, not a website.

Not losing weight faster than you need to. Across the pilot trial and the retrospective DXA cohort, the amount of bone lost tracked with the amount of weight lost.³⁴ That association does not prove that slowing the rate of loss protects bone, because nobody has randomized people to fast versus slow loss and measured bone. It does mean the trade-off is worth discussing with a prescriber rather than assuming faster is always better.

What to ask a clinician, and when a DXA scan makes sense

Useful questions are specific ones. Whether your age, sex, menopausal status, fracture history, or existing bone density puts you in the group where the semaglutide trial found a measurable effect.² Whether a baseline DXA scan early in treatment is worth having, so that any later scan has something to compare against. Whether your vitamin D and calcium status has been checked, given that weight loss reduces calcium absorption.¹⁰

A baseline scan is most obviously reasonable if you are postmenopausal, over 65, have a prior low-energy fracture, or already carry an osteopenia or osteoporosis diagnosis, since that is the population in which the clearest bone density signal was measured.² Screening decisions are clinical decisions and depend on more than this article can cover. If you are already considering a scan for body composition reasons, our comparison of DXA versus InBody scans explains what each one does and does not tell you. Note that a body composition DXA and a bone density DXA are related but reported differently, so ask which one you are getting.

What the evidence does not say

Several limits are worth stating plainly. No trial has been designed and powered to measure fracture incidence in GLP-1 users; the fracture meta-analysis pooled trials in which fractures were incidental adverse events, only 292 across 47,823 patients.⁵ The osteoporosis signal from the large matched cohort is conference-presented and has not completed peer review.⁶ The trial that measured the clearest bone density decline enrolled 64 people already at elevated fracture risk, the population where an effect is easiest to detect and hardest to generalize from.² Only two of nine placebo-controlled GLP-1 trials with DXA outcomes reported bone density at all.⁸ And no study has tested whether resistance training, protein, or vitamin D changes bone outcomes in GLP-1 users specifically; that evidence comes from dieting adults.⁷⁹¹⁰ We do not know whether taking one of these medications makes a person more likely to break a bone.

The bottom line

Bone density falls during substantial weight loss whether or not a medication is involved, mostly at the hip, and that is well established.¹ In the one randomized trial that looked hardest, in people already at elevated fracture risk, semaglutide added to that decline over 52 weeks.² Smaller and shorter GLP-1 studies did not find a clear separation from matched or control groups.³⁴ Whether any of this translates into more broken bones is unresolved, with randomized-trial pooling pointing one way and a large unpublished cohort pointing the other.⁵⁶ What is not in dispute is that resistance training reduces weight-loss-related hip bone loss by a meaningful margin in older adults,⁷ that it can add bone in postmenopausal women with low bone mass,¹¹ and that adequate protein and adequate calcium and vitamin D make the arithmetic easier rather than harder. Those are the same habits this site recommends for muscle, which means the practical response to an unresolved question is one you were probably already making.

Frequently asked