Losing weight costs bone however the weight comes off (diet, surgery or drug), typically 1 to 3% at the hip and spine over a year or so. The question these ten papers circle is whether GLP-1 drugs do anything beyond that. A 2015 randomized trial suggested they actively protect bone, which launched the hypothesis. A larger 2024 trial complicated it: the drug alone lost more density than exercise alone at the same weight loss, and only the combination held bone steady. Meanwhile three meta-analyses of randomized trials find no fracture excess, and the two largest, with about 40,000 and 48,000 participants, found fewer fractures on the drugs. That reassurance has a specific limit. Those trials ran mostly under two years, in people with type 2 diabetes, and counted fractures as incidental adverse events rather than as an endpoint anyone was looking for. Nobody has yet followed a non-diabetic obesity population, the people filling Wegovy and Zepbound prescriptions, for long enough to answer the question that matters.
Ranked papers
#1S-LITE (bone substudy)
Jensen SBK, Sørensen V, Sandsdal RM, et al. · JAMA Netw Open · 2024
What it measured: Bone density at the hip, spine and forearm after a year
The best randomized bone data available, and more nuanced than the headline suggests. After an eight-week low-calorie diet, 195 adults with obesity were sorted into four year-long arms (liraglutide 3.0 mg, supervised exercise, the two together, or neither), with density read at hip, spine and forearm. Two results stand out. The combination held bone density unchanged against placebo, and liraglutide alone lost more density than exercise alone at the hip and spine despite comparable weight loss. That second comparison complicates the drug-protects-bone story and points squarely at training as the countermeasure.
PMID 38916894 ↗DOI 10.1001/jamanetworkopen.2024.16775 ↗
#2
Iepsen EW, Lundgren JR, Hartmann B, et al. · J Clin Endocrinol Metab · 2015
What it measured: Bone content over a year of maintaining a diet-induced weight loss
The trial that started the protective hypothesis. 37 women with obesity who had already lost 12% on a low-calorie diet were randomized to 52 weeks of weight maintenance on either liraglutide 1.2 mg daily or no drug. Bone mineral content dropped significantly in controls, while the drop on liraglutide was not significant (a four-fold difference), while the bone-formation marker P1NP rose 16% on the drug against a 2% fall in controls. The limits are real: 37 people, bone content rather than density, and a maintenance phase rather than active weight loss. It generated a hypothesis; it did not settle one.
PMID 26043228 ↗DOI 10.1210/jc.2015-1176 ↗
#3
Zhang Y, Chen G, Wang W, et al. · Bone · 2025
What it measured: Fracture risk pooled across 44 randomized trials in type 2 diabetes
The current randomized benchmark: 44 trials and 47,823 participants with type 2 diabetes. Fractures were less common on GLP-1 drugs than on placebo or other diabetes drugs (relative risk 0.77, 95% CI 0.61 to 0.96), but in subgroups the benefit appeared only with more than 78 weeks of treatment and, among individual drugs, only with liraglutide (relative risk 0.42). Every trial enrolled people with type 2 diabetes, and the authors call the evidence too limited for a clinical consensus. That leaves the long-term question for older Wegovy and Zepbound users exactly where it started.
PMID 39603373 ↗DOI 10.1016/j.bone.2024.117338 ↗
#4
Cheng L, Hu Y, Li YY, et al. · Diabetes Metab Res Rev · 2019
What it measured: Fracture risk pooled across 38 randomized trials
The previous generation of the same analysis: 38 trials and 39,795 participants with type 2 diabetes, with a pooled odds ratio of 0.71 (95% CI 0.56 to 0.91), fewer fractures on the drugs. Only liraglutide (0.56) and lixisenatide (0.55) reached significance on their own, and only with treatment longer than 52 weeks. It shares the structural flaw of every meta-analysis here: fractures were picked up as adverse-event reports rather than counted as a pre-specified outcome, which is a weaker way to find them.
PMID 30974033 ↗DOI 10.1002/dmrr.3168 ↗
#5
Mabilleau G, Mieczkowska A, Chappard D · J Diabetes · 2014
What it measured: Fracture risk pooled across the earliest randomized trials
The first attempt at the question. Of 28 eligible trials, only seven recorded any fracture, 19 in all. The odds ratio came out at 0.75, pointing toward fewer fractures, but the confidence interval ran from 0.28 to 2.02. In other words, almost no fractures occurred and the estimate means little. Its contribution was setting a default assumption of no harm that the two larger analyses above have since failed to overturn.
PMID 24164867 ↗DOI 10.1111/1753-0407.12102 ↗
#6
Ko HY, Bea S, Jeong HE, et al. · JAMA Netw Open · 2023
What it measured: Fractures among new users of two different diabetes drug classes
The largest real-world comparison, and the one closest to a practical prescribing question. A nationwide South Korean claims cohort of postmenopausal individuals with type 2 diabetes compared new SGLT2 inhibitor users against new users of incretin drugs, using an active-comparator new-user design that avoids the classic trap of comparing treated patients against untreated ones. Against GLP-1 drugs (111,835 SGLT2 inhibitor users and 8,177 GLP-1 users), fracture rates came out similar (hazard ratio 0.92, 95% CI 0.68 to 1.24). Against DPP-4 inhibitors, the other incretin class, SGLT2 inhibitor users had fewer fractures (hazard ratio 0.78). On the GLP-1 question, neither class looks bone-safer than the other.
PMID 37751205 ↗DOI 10.1001/jamanetworkopen.2023.35797 ↗
#7
Karam L, Mabilleau G, Paccou J · Osteoporos Int · 2025
What it measured: Narrative review of GLP-1 obesity-trial bone data
A review restricted to obesity rather than diabetes populations, which makes it useful. It pulls together the DEXA substudies from the STEP and SURMOUNT programs alongside the two randomized bone trials above. It concludes that the density loss that shows up is what weight loss produces, that a drug-specific protective effect remains plausible but unproven outside diabetes, and that the only mitigation with evidence behind it is resistance training plus enough protein.
PMID 40920189 ↗DOI 10.1007/s00198-025-07664-1 ↗
#8
Anastasilakis AD, Paccou J, Palermo A, et al. · Diabetes Obes Metab · 2025
What it measured: Bone effects compared across every major weight-loss drug
A critical appraisal covering bone effects across every major weight drug (orlistat, phentermine-topiramate, naltrexone-bupropion, liraglutide, semaglutide and tirzepatide), which makes it the best single place to see how this class compares with the alternatives. For GLP-1 drugs, clinical studies show a neutral or negative effect on bone density and turnover that is not clinically significant, and fracture risk does not seem to rise at clinical doses, but there are no long-term population data at all. The authors name the gap directly, and it is the right one: millions of healthy adults are now expected to take these drugs for decades.
PMID 40555693 ↗DOI 10.1111/dom.16541 ↗
#9
Xiao Y, Zhou M, Xiao W, et al. · Acta Diabetol · 2025
What it measured: Disproportionality analysis of fracture reports in FAERS
A disproportionality analysis of the FDA's adverse-event database from 2004 to 2022. GLP-1 drugs had the lowest fracture reporting of any diabetes drug class (adjusted reporting odds ratio 0.44), which the authors read as a protective effect. Spontaneous reports have no denominator: nobody knows how many people took the drug and did not report anything, so this cannot produce a rate or establish cause. It earns its place because it is the largest such dataset that exists and because it agrees with the randomized evidence.
PMID 39556224 ↗DOI 10.1007/s00592-024-02415-w ↗
#10
Al Refaie A, Baldassini L, Mondillo C, et al. · Calcif Tissue Int · 2024
What it measured: Bone density plus a texture score that captures bone quality, over a year
A prospective single-center study following 65 adults with type 2 diabetes for a year after starting semaglutide or dulaglutide, measuring density and trabecular bone score, a texture measure that captures the bone quality density alone misses. Of the 54 who finished (30 on dulaglutide, 24 on semaglutide), density by DXA fell significantly at the lumbar spine and modestly at the hip, while the texture score rose marginally and bone turnover markers rose. The authors read that as preserved bone quality. There was no comparison group, and the study was small and single-center, so it cannot separate the drug from the weight loss.
PMID 38864922 ↗DOI 10.1007/s00223-024-01240-1 ↗
About this list
Obesity medicine throws up the same handful of questions over and over, and we answer them with ranked lists anchored to the papers themselves. On 2026-10-30 every citation below was put back through PubMed. Where a trial has a registry entry, the card carries that link too, alongside the PubMed one.
Browse our full index of research lists or our long-form research articles.