Do GLP-1 Drugs Weaken Bones? What the New Randomized-Evidence Review Actually Found

The best current answer is that we do not know whether GLP-1 receptor agonists weaken bones or increase fractures. A 2026 systematic review found a few agent-specific bone mineral density signals, including modest reductions at selected sites in one semaglutide trial and one liraglutide trial. But the randomized evidence consisted of only eight small trials with 493 participants, and certainty was rated very low for every drug studied.

That is neither a clean safety signal nor a clean bill of health. It is an evidence gap.

Bone scans, turnover markers, and fracture outcomes answer different questions. The 2026 review found too little randomized evidence to make a class-wide judgment.

GLP-1 bone density: the short answer

The review does not show that GLP-1 medicines cause osteoporosis or fractures. It shows that the randomized evidence on GLP-1 bone density is too small, too varied, and too unevenly distributed across drugs to settle the question.

The review by Lai and colleagues, published October 23, 2026, included:

• 8 randomized controlled trials with 493 participants in total
• 4 different agents: semaglutide, liraglutide, exenatide, and dulaglutide
• 5 trials in adults with type 2 diabetes and 3 in adults without diabetes
• 3 trials at low risk of bias and 5 with some concerns
• no more than one low-risk-of-bias trial for any individual agent
• very-low-certainty evidence for every agent

The authors did not calculate one pooled class-wide effect. That was a strength, not an omission: averaging clinically different drugs, doses, populations, comparators, weight changes, and skeletal measurements could have produced a precise-looking number with a misleading meaning.

What did the semaglutide trial find?

The single semaglutide trial found modest reductions at selected BMD sites and increased bone resorption, but one 64-person trial cannot establish a general semaglutide effect.

This was the Hansen 2024 randomized trial in 64 adults with increased fracture risk. The review rated it at low risk of bias. Compared with placebo, semaglutide produced the largest weight difference among the included trials: an estimated 6.8 kg greater weight loss.

The trial reported:

• lumbar-spine BMD lower by 0.018 g/cm² relative to placebo
• whole-body BMD lower by 0.02 g/cm²
• C-terminal telopeptide, or CTX, higher by 166.5 pg/mL

CTX is a marker of bone resorption. A higher value suggests more breakdown of bone tissue during remodeling, but it is not a fracture and does not by itself diagnose osteoporosis. Likewise, a difference in BMD is a surrogate skeletal measurement. Its clinical meaning depends on site, magnitude, duration, baseline risk, measurement precision, and whether it predicts actual fractures.

The trial is important because it was randomized and comparatively well conducted. It is limited because it was small, involved one agent in a selected population, and was not powered to determine fracture risk.

Why did the two liraglutide trials look different?

The liraglutide findings differed by trial, dose, population, and amount of weight loss, which is exactly why class-wide conclusions are premature.

In the Jensen 2024 trial, liraglutide at the studied 3.0 mg weight-management dose produced an estimated 6.7 kg greater weight loss than placebo. The trial reported lower BMD at two sites:

• lumbar spine: 0.019 g/cm² lower
• total hip: 0.013 g/cm² lower

By contrast, the Hygum 2020 trial studied liraglutide at 1.8 mg in adults with type 2 diabetes and reported about 3.8 kg of weight loss. Lumbar-spine BMD did not change materially: the mean difference was 0.00 g/cm², with a 95% confidence interval from −0.01 to 0.01.

These trials do not prove that a particular dose protects or harms bone. They differed in more than dose, including population and study context. Their divergence instead shows why one result should not be copied onto every use of liraglutide—or onto every GLP-1 receptor agonist.

What did the exenatide and dulaglutide evidence show?

Exenatide results were inconsistent, while dulaglutide was represented by only one trial. Neither evidence base supports a confident verdict.

Five of the eight included trials evaluated exenatide, but their findings did not line up. The review highlighted lumbar-spine estimates ranging from 0.03 g/cm² higher in one trial to 0.04 g/cm² lower in another; the latter estimate had a wide confidence interval spanning substantial benefit and harm.

Dulaglutide appeared in one trial. A single small study cannot define its skeletal profile, especially when the broader review found no agent supported by more than one low-risk-of-bias trial.

The correct synthesis is not that the drugs “cancel each other out.” It is that drug-specific evidence remains sparse and inconsistent.

Why didn’t the authors pool the eight trials?

A meta-analysis is only useful when combining studies answers a coherent question. Here, one pooled number could have hidden the differences that matter most.

The studies varied across several dimensions:

| Source of variation | Why it matters | |---|---| | Agent | Semaglutide, liraglutide, exenatide, and dulaglutide differ pharmacologically and in weight-lowering potency. | | Population | Most trials involved type 2 diabetes; only one focused on obesity without diabetes. | | Comparator | Some trials used placebo, while others used active diabetes treatments, including insulin. | | Weight change | The amount of weight loss varied substantially and may itself affect skeletal measurements. | | Skeletal endpoint | Trials measured different BMD sites and bone-turnover markers, not a uniform clinical outcome. | | Study quality | Five trials raised bias concerns, and no agent had more than one low-risk-of-bias trial. |

Readers sometimes assume a systematic review must produce a forest plot and one bottom-line estimate. It does not. As explained in PeptideBase’s guide to reading a peptide meta-analysis, refusing to pool incompatible studies can be more informative than manufacturing an average.

BMD, bone-turnover markers, and fractures are not interchangeable

BMD and bone-turnover markers can identify skeletal signals, but fractures are the clinical outcome needed to determine whether those signals translate into harm.

• Bone mineral density estimates the mineral content of bone at measured sites, usually by dual-energy X-ray absorptiometry, or DXA.
• Bone-turnover markers such as CTX reflect aspects of ongoing bone remodeling.
• Fractures are patient outcomes influenced by bone strength, falls, age, muscle function, prior fractures, medications, and other factors.

A drug could alter a biomarker without changing fracture incidence. A modest BMD change could matter over time—or remain clinically small. And a fracture association could arise even without a directly measured BMD pathway.

This is the same endpoint problem covered in PeptideBase’s guide to biomarkers versus clinical benefit: an intermediate measure can be important without being a substitute for the outcome people actually care about.

The Lai review was specifically built around randomized trials reporting DXA-derived BMD. It concluded that adequately powered, agent-specific trials with fracture endpoints are still needed.

Why observational fracture studies do not settle the question

Observational studies can examine far more fractures than small BMD trials, but their results depend heavily on who received each drug and what comparison group was chosen.

Two 2026 studies illustrate the problem:

• A retrospective cohort of adults aged 65 or older with type 2 diabetes found GLP-1 receptor agonist initiation associated with an 11% higher fragility-fracture rate than the combined comparator group. The adjusted hazard ratio was 1.11.
• A separate target-trial emulation in adults with type 2 diabetes found GLP-1 receptor agonist initiation associated with a 21% lower three-year fragility-fracture rate than initiation of a DPP-4 inhibitor. The hazard ratio was 0.79. In one sensitivity analysis, the association differed by diabetes status.

Those estimates point in different directions. Neither randomly assigned treatment, so neither can fully remove residual confounding, prescribing differences, selection effects, or comparator effects. They also do not evaluate identical populations or questions.

Observational evidence is useful for detecting patterns in large real-world populations. It does not erase the randomized evidence gap identified by Lai and colleagues, and the small randomized BMD trials do not invalidate the observational fracture studies. The two evidence streams answer different questions with different vulnerabilities.

Does weight loss explain the BMD findings?

The review raises weight loss as an important possible context, but it does not prove that weight loss caused the observed BMD changes.

The two trials with the largest relative weight loss—the semaglutide trial and the liraglutide 3.0 mg trial—both reported reductions at selected BMD sites. That convergence deserves further study.

But it is not a causal test. Participants were randomized to a drug, not to a specific amount of weight loss. The trials cannot cleanly separate a drug-specific skeletal effect from the consequences of weight change, differences in body composition, altered mechanical loading, or other pathways. A pattern can generate a hypothesis without proving the mechanism.

What can we conclude without overreaching?

The review supports caution in interpretation, not a claim that GLP-1 drugs cause osteoporosis and not a claim that the class is proven bone-safe.

The defensible conclusions are:

1. Randomized GLP-1 bone-density evidence is small: eight trials and 493 participants.
2. The trials are clinically heterogeneous, so the decision not to pool them was appropriate.
3. Semaglutide and liraglutide 3.0 mg produced modest reductions at selected BMD sites in individual trials alongside substantial weight loss.
4. Lower-dose liraglutide did not show the same pattern, and exenatide findings were inconsistent.
5. BMD and CTX are not fracture outcomes.
6. Observational fracture estimates can move in opposite directions and cannot establish causality.
7. Conclusions should remain agent-specific until larger randomized trials report clinically meaningful skeletal outcomes.

For broader context on the peptide-drug category, see GLP-1 peptides versus “research peptides” and the PeptideBase database pages for GLP-1, semaglutide, liraglutide, exenatide, and dulaglutide.

Frequently asked questions

Do GLP-1 drugs cause osteoporosis?

The available evidence does not establish that GLP-1 drugs cause osteoporosis. The 2026 systematic review found very-low-certainty evidence for every agent and called for larger, agent-specific trials with fracture outcomes.

Did semaglutide lower bone density?

One randomized 64-person semaglutide trial reported modest reductions in lumbar-spine and whole-body BMD relative to placebo. That signal is worth studying, but one trial cannot establish a general effect or determine fracture risk.

Is lower BMD the same as a higher fracture risk?

No. BMD is an important surrogate measure, while fractures are clinical outcomes influenced by many factors. A BMD difference cannot automatically be translated into a fracture estimate.

Why can fracture studies disagree?

Observational fracture studies may use different populations, comparators, follow-up periods, and statistical adjustments. Treatment is not randomly assigned, so residual confounding can also move estimates in either direction.

Did the review prove a class effect?

No. The authors reported results separately for semaglutide, liraglutide, exenatide, and dulaglutide because the agents and study contexts were too different for a credible class-wide estimate.

Sources

1. Lai CH, Li YT, Wu TS, et al. Skeletal effects of GLP-1 receptor agonist therapy in adults with overweight or obesity: a systematic review. Archives of Osteoporosis. 2026;21:148.
2. Hansen MS, Wölfel EM, Jeromdesella S, et al. Once-weekly semaglutide versus placebo in adults with increased fracture risk. EClinicalMedicine. 2024;72:102624.
3. Jensen SBK, Sørensen V, Sandsdal RM, et al. Bone health after exercise alone, GLP-1 receptor agonist treatment, or combination treatment. JAMA Network Open. 2024;7:e2416775.
4. Kasher Meron M, Hornik-Lurie T, Twig G, et al. GLP-1 receptor agonists and the risk of fragility fractures in older adults with type 2 diabetes. Journal of Clinical Endocrinology & Metabolism. 2026;111:1949–1958.
5. Hamad CD, Wiener J, Golzar A, et al. Glucagon-Like Peptide-1 Receptor Agonists and Fragility Fracture Risk in Type 2 Diabetes. JAMA Network Open. 2026;9:e2625141.

This article is educational and explains research evidence. It does not provide medical advice or recommend starting, stopping, or changing any treatment.