GLP-1 Bone Density and Fracture Risk: 15 Study Reports by Drug

By Kaden Coziar, Founder & Managing Editor, The RX Index Research · Updated September 2026

In a 52-week trial of 64 adults at increased fracture risk and without diabetes, semaglutide left hip bone density 2.6% lower relative to placebo (Hansen et al., 2024). That is not a 2.6% fracture risk. The RX Index’s September 2026 GLP-1 bone density map separates 15 study reports by drug, outcome and population.

Below: the drugs in this map, what six U.S. labels say about fractures, and the exact counts behind the older-adult findings.

Key GLP-1 bone density statistics

1. Copy statistic, source and page link Semaglutide left hip bone density 2.59% lower and spine bone density 2.05% lower relative to placebo after 52 weeks in 64 adults at increased fracture risk and without diabetes (Hansen et al., eClinicalMedicine, 2024, 1.0 mg study dose).
2. Copy statistic, source and page link One three-year health-record analysis found opposite fracture associations by diabetes status: HR 0.91 with type 2 diabetes and HR 1.13 without it. Both compared GLP-1 users with people who had never used a GLP-1—not DPP-4 inhibitor users (Hamad et al., JAMA Network Open, 2026).
3. Copy statistic, source and page link A 52-week liraglutide trial in 37 women after diet-induced weight loss reported less bone mineral content loss than the control group: an estimated difference of 27 g (95% CI 5–48). Bone mineral content is not bone density or fracture risk (Iepsen et al., Journal of Clinical Endocrinology & Metabolism, 2015).
4. Copy statistic, source and page link In Wegovy's heart trial (SELECT), 24.2 people per 1,000 at age 75 or older had a hip or pelvis fracture on semaglutide vs 6.0 per 1,000 on placebo over the trial, or about 18 more per 1,000—not per year (The RX Index calculation from participant counts in the Wegovy prescribing information, rev. 06/2026; overall median follow-up 41.8 months).
5. Copy statistic, source and page link In the same trial, 9.8 women per 1,000 had a hip or pelvis fracture on semaglutide vs 2.1 per 1,000 on placebo over the trial, about 8 more per 1,000 (The RX Index calculation from participant counts in the Wegovy label).
6. Copy statistic, source and page link Across all 17,604 people in SELECT, reported fracture adverse events in the broader coding group were 3.5% on semaglutide vs 3.6% on placebo over a mean of 39.8 months. The trial did not systematically collect every nonserious fracture (Kushner et al., Obesity, 2025).
7. Copy statistic, source and page link Only 1 of the 6 U.S. prescribing-information documents reviewed here contains a clinical fracture finding: Wegovy's. No such statement was found in the checked Ozempic, Mounjaro, Zepbound, Saxenda or Foundayo editions (The RX Index label review, September 29, 2026).
8. Copy statistic, source and page link In 133,606 adults with type 2 diabetes, starting a GLP-1 was linked to a 21% lower fragility-fracture hazard over 3 years than starting a DPP-4 inhibitor (HR 0.79) and a 30% lower hip or femur fracture hazard (HR 0.70) (Hamad et al., JAMA Network Open, 2026).
9. Copy statistic, source and page link In adults 65 and older with type 2 diabetes, femoral (thigh-bone) fractures in one year were 2.5 per 1,000 on semaglutide vs 5.2 per 1,000 on DPP-4 inhibitors, and 1.9 vs 4.4 per 1,000 on tirzepatide vs DPP-4 inhibitors (Chen et al., Osteoporosis International, 2026; per-1,000 math by The RX Index).
10. Copy statistic, source and page link In 46,177 adults 65 and older with type 2 diabetes in Israel, starting a GLP-1 was linked to an 11% higher fragility-fracture hazard than starting an SGLT-2 or DPP-4 inhibitor (HR 1.11) over a median of 34.7 months (Kasher Meron et al., Journal of Clinical Endocrinology & Metabolism, 2026).
11. Copy statistic, source and page link Starting tirzepatide was linked to a 44% higher 14-month hazard of new osteoporosis or fragility fracture, combined, than starting another GLP-1 (HR 1.44) in 66,329 matched pairs; that is not a fracture-only number (Hsu et al., Diabetes Research and Clinical Practice, 2025).
12. Copy statistic, source and page link A five-year conference study of 73,483 matched pairs with type 2 diabetes and obesity reported osteoporosis diagnoses in 4.1% vs 3.2%, about 41 vs 32 per 1,000. Its printed case counts do not reproduce those percentages using the full cohort sizes; the mismatch is shown below (AAOS 2026 abstract, Table 1; rounded-percentage conversion by The RX Index).
13. Copy statistic, source and page link As of September 29, 2026, The RX Index's targeted search found no randomized human trial reporting tirzepatide's effect on hip or spine bone density; a 2026 pooled analysis of seven tirzepatide trials says bone density wasn't assessed in them (Alfaris et al., Diabetes, Obesity and Metabolism, 2026).

On this page

• What do the studies show for each GLP-1 drug?
• Why do the results differ by diabetes status?
• Does Ozempic (semaglutide) lower bone density?
• How many fractures showed up in Wegovy's heart trial?
• Which GLP-1 labels mention fractures?
• What do tirzepatide (Zepbound, Mounjaro) studies show?
• Are older adults at higher risk?
• Does exercise protect your bones on a GLP-1?
• Where did the "30% higher osteoporosis risk" headline come from?
• What should you ask your doctor?
• Why this matters now
• How we built this
• What this data does and doesn't show
• How to cite this page
• Download the data
• Frequently asked questions
• Sources

What do the studies show for each GLP-1 drug?

Semaglutide has the most bone evidence in this map: one randomized trial found 2.59% lower hip bone density relative to placebo after a year, and its big heart trial reported more hip and pelvis fractures in women and people 75 and older. Tirzepatide has fracture data from health records but no randomized human hip or spine bone-density result we could find. Only Wegovy's label, among the six checked here, contains a clinical fracture finding. Sources are linked in Table 1.

Read each row with one question in mind: did the study measure bone density, an osteoporosis diagnosis, or actual broken bones? Those are three different things. Bone density can drop without a single extra fracture showing up, at least over the time a study runs.

Source: The RX Index GLP-1 Bone Evidence Map v1.1, built from Hansen 2024, Jensen 2024, Kushner 2025, Hamad 2026, Huang 2026, Chen 2026, Hsu 2025, Liu 2026, Cheng 2019, Alfaris 2026, Iepsen 2015 and six U.S. prescribing-information documents. Full rows with links are in the dataset.

What the words mean

Source: definitions written by The RX Index for this page.

Look up a drug and an outcome

Use the full evidence table to find a drug and the kind of result you care about. Every row keeps its population, comparison, time period and source link. Nothing is ranked, and nothing is hidden: the full table and the CSV stay available without a form.

Why do the results differ by diabetes status?

In one three-year analysis, GLP-1 use was linked to a lower fragility-fracture hazard in people with type 2 diabetes (HR 0.91) but a higher hazard in people without it (HR 1.13). Both groups were compared with GLP-1 noninitiators (Hamad et al., 2026). That is a specific finding—not proof that diabetes determines what every GLP-1 does to bone.

Here's why the headlines keep fighting each other. One week a study says GLP-1s protect bones. The next week one says they hurt them. Check who was studied, what was measured and what the comparison was. Even one trial can report lower hip density and higher whole-body density.

Source: the linked original reports, checked September 29, 2026. This table keeps the outcome, group and comparator attached to each finding. Hsu's mixed-population drug comparison and Alfaris's pooled trial results appear in the tirzepatide section.

The no-diabetes evidence is not uniformly negative. Iepsen measured BMC, while Jensen found different results at different bone sites. Counting each whole study as “better” or “worse” would hide those findings.

The split inside the same studies

A count across different studies can mislead. The people differ, the drugs differ, the years differ. Looking inside one study helps, but the comparisons and follow-up still need to match. Four reports here present separate results by diabetes status.

Source: the four reports linked in the table, checked September 29, 2026. Hamad reports an interaction P<.001: evidence that the association differed by diabetes status in that analysis. That does not show why it differed. Epic's outcome includes a diagnosis or an osteoporosis-medicine order, not just scan-confirmed osteoporosis.

Source: Huang et al., Tables 1 and 2. “Usual care” means metformin users without starting semaglutide or a comparator drug. The correction is the paper's Benjamini–Hochberg false-discovery-rate procedure; the sitagliptin result shows why a confidence interval alone is not the full significance test here.

Hamad's main diabetes comparison against DPP-4 inhibitors also changed over time: the cumulative three-year HR was 0.79, but the third-year interval alone was 0.94 (0.86–1.02), with no clear difference. Those answer different time questions, not conflicting versions of one number (Hamad et al., interval analysis).

Why diabetes might change the answer

These studies do not prove that diabetes itself explains the differences. They point at three things.

The comparison is often different. Most diabetes studies here compare GLP-1 users with people on other diabetes drugs, like DPP-4 inhibitors, SGLT-2 inhibitors or glipizide. Most no-diabetes studies compare them with a placebo, with people on no drug, or with people on another weight-loss medicine. "Better than another diabetes pill" and "better than nothing" are different questions.

Weight loss may be doing more of the work without diabetes. In the Weill Cornell study, more weight loss went with more hip bone loss, and the authors concluded that weight loss seemed to drive bone loss in people without diabetes (Liu et al., 2026).

Something beyond weight and blood sugar may be at play in diabetes. In the JAMA Network Open study, a mediation model accounting for changes in BMI and A1c still estimated a lower fracture hazard (direct-effect HR 0.81, 0.76–0.86). That model does not prove a bone-protective mechanism (Hamad et al., 2026).

Does Ozempic (semaglutide) lower bone density?

Yes, in the 64-person trial of adults at increased fracture risk and without diabetes. After 52 weeks, semaglutide 1.0 mg left hip bone density 2.59% lower and spine bone density 2.05% lower relative to placebo, while the semaglutide group lost about 9% of its weight. The trial was not designed to estimate fracture risk (Hansen et al., 2024).

The semaglutide group’s average hip density went from 0.785 to 0.764 g/cm², about 2.7% down. The placebo average stayed at 0.793 g/cm². The study’s adjusted relative difference was 2.59%. That is a group comparison—not a prediction for one person’s bones.

Source: Hansen et al., eClinicalMedicine 2024;72:102624, Table 2. 64 adults, 32 per group. PINP was the primary outcome; the other rows were secondary. The authors did not adjust these tests for multiple comparisons. The hip/spine DXA measures are mineral per area; the shin/wrist 3D measures are mineral per volume. They are not measurements of the same bone compartment.

The hip, spine and shin carry your body weight every time you stand. The wrist doesn't. The trial's authors pointed to that pattern: bones that suddenly carry less weight tend to rebuild less. It fits, though the femoral neck, also weight-bearing, didn't show a clear change.

Where the "2.6%" number comes from

That hip number gets repeated a lot. Here's who was actually in the trial.

• The dose was 1.0 mg. That's an Ozempic dose, not the 2.4 mg Wegovy dose.
• 64 people, 32 per group. Mean age 63. 86% were women past menopause.
• They started at increased fracture risk. Everyone had low bone density or a recent fragility fracture.
• Average BMI was 27.7. That is in the overweight range, not the obesity range.
• It's a gap between groups. Hip density in the semaglutide group went from 0.785 to 0.764 g/cm², about 2.7% down. The placebo group stayed at 0.793.

So "2.6% bone loss on GLP-1s" stretches one small trial across millions of very different people.

Is 54.8% a bone-loss number?

No. It's CTX, a blood marker that rises when the body breaks down more bone. Its relative treatment difference was +54.8%; the hip BMD difference was −2.59%. These are different measurements. Dividing one percentage by the other does not tell you how much worse the bone loss was.

Is that more bone loss than dieting would cause?

The cited diet trial found hip BMD fell 2.2% over one year in its calorie-restriction group, whose weight fell 10.7%; hip BMD rose 1.2% in the control group (Villareal et al., 2006). Those are within-group changes from a different trial. Hansen’s 2.59% is a relative treatment difference, so placing them side by side does not show that dieting and semaglutide cause the same bone loss. These studies did not match equal amounts of weight loss head to head.

Bone density and muscle are separate questions. For what GLP-1s do to lean tissue, see our GLP-1 lean mass loss by drug research.

How many fractures showed up in Wegovy's heart trial?

Across all 17,604 people in SELECT, reported fracture adverse events in the broader coding group were 3.5% on semaglutide 2.4 mg vs 3.6% on placebo over a mean of 39.8 months. Hip and pelvis fracture proportions were higher in two groups: 24.2 vs 6.0 per 1,000 at age 75+, and 9.8 vs 2.1 per 1,000 in women. Those are trial-period participant proportions, not yearly rates (SELECT safety paper, Wegovy label).

The subgroup counts come straight from the Wegovy label. Here they are in plain per-1,000 terms.

Source: reported-fracture-adverse-event row from Kushner et al., Obesity 2025; subgroup counts from the Wegovy prescribing information, rev. 06/2026; per-1,000 math by The RX Index. The women and 75+ groups overlap. Don't add them together. The label gives an overall median follow-up of 41.8 months; the safety paper gives a mean of 39.8 months. Both are correct. SELECT systematically collected serious adverse events, events leading to stopping treatment, and selected events of special interest—not every nonserious fracture.

Scaled to 1,000 participants aged 75 or older, the observed counts equal about 24 people with a hip or pelvis fracture on semaglutide and about 6 on placebo. That's roughly 18 more over the trial—not a prediction that 18 extra fractures would occur in every new group of 1,000.

What the label says, and what it doesn't

The label reports counts. It doesn't say the drug caused them. The SELECT safety paper described no overall imbalance in the fracture adverse events it collected and called the hip and pelvis gaps "numerical" imbalances in small groups (Kushner et al., 2025). The numbers are small: 24 women with a hip or pelvis fracture vs 5, and 17 participants vs 4 at age 75 and older.

Numbers that small can swing a lot, so one trial can't settle it. That's why the older-adult studies and the diabetes split elsewhere on this page matter.

The liver (MASH) trial is a separate count

The Wegovy label also reports a trial in adults with MASH, a fatty-liver disease. There, 4.4% of people on Wegovy had a fracture vs 3.3% on placebo. Separately, the label reports rates of 2.6 vs 2.0 per 100 patient-years (Wegovy prescribing information). Different people, different yardstick. Keep it apart from the SELECT numbers.

Which GLP-1 labels mention fractures?

Only 1 of the 6 U.S. prescribing-information documents checked here contains a clinical fracture finding: Wegovy's. Ozempic contains the same drug, semaglutide, but its reviewed label has no such statement. Neither do the checked Mounjaro, Zepbound, Saxenda or Foundayo labels. This is a review of these six editions, not every GLP-1 product.

Source: official prescribing information for Wegovy, Ozempic, Mounjaro, Zepbound, Saxenda and Foundayo; each searched for “fracture,” “bone mineral,” “osteopor,” “hip” and “pelvis,” with the clinical context read. Mounjaro’s August 2026 edition was checked in DailyMed. Check done by The RX Index.

A silent label isn't a clean bill of health. It means we found no explicit clinical fracture or BMD statement in that reviewed edition. For how GLP-1 labels have changed over time, see our GLP-1 FDA label changes tracker.

What do tirzepatide (Zepbound, Mounjaro) studies show?

We found no randomized human hip or spine bone-density result for tirzepatide, and a 2026 pooled analysis says its phase 3 trials didn't assess bone density. In adults 65 and older with type 2 diabetes, it was linked to fewer femoral fractures than DPP-4 inhibitors (1.9 vs 4.4 per 1,000 in one year). Compared with other GLP-1s, it was linked to a 44% higher hazard of a combined “osteoporosis or fracture” outcome (Chen, Hsu, Alfaris).

That sounds like a contradiction. The studies compared tirzepatide with different things and counted different outcomes. Their differences explain why the numbers cannot be treated as the same question; they do not settle which finding best predicts future fractures.

Source: the studies linked in the table, checked September 29, 2026; per-1,000 math by The RX Index.

Why "44% more fractures" is the wrong summary

The 1.44 hazard ratio counts two things together: a new osteoporosis diagnosis or a fragility fracture. The abstract doesn't split them. So “44% more fractures” does not describe the reported composite. The same study found tirzepatide users were also more likely to start an osteoporosis medicine (HR 1.61, 1.22–2.12). That points to more diagnoses and treatment, but the abstract doesn't say how many of the events were actual breaks.

What the trials in people 65 and older counted

Eli Lilly pooled the side-effect reports from seven tirzepatide trials for people 65 and older. The authors described fracture and fall rates as broadly comparable; the small pooled counts do not establish equal risk. They also wrote that this is hard to read, because none of the trials measured bone density and the follow-up may have been too short.

Source: Alfaris et al., Diabetes, Obesity and Metabolism 2026, Table 1 (Lilly-funded post hoc analysis); per-1,000 math by The RX Index. Very few events, pooled from trials with different designs, and one person can appear in more than one row.

The missing bone-density data

Tirzepatide’s phase 3 obesity trials produced substantial weight loss. Neither reviewed label has a clinical fracture or BMD statement. The authors of the 2026 pooled analysis say bone density wasn't assessed in the phase 3 trials they pooled. As of September 29, 2026, we found no randomized trial reporting what it does to hip or spine bone density. That's the biggest hole in this whole map.

Are older adults at higher risk?

Some of the strongest warning signs come from older adults: 24.2 vs 6.0 hip or pelvis fractures per 1,000 at age 75+ in Wegovy's heart trial, and an 11% higher fragility-fracture hazard in people 65+ with type 2 diabetes in Israel. But another study of people 65+ with diabetes found about half as many femoral fractures on semaglutide or tirzepatide as on DPP-4 inhibitors.

Source: studies linked in the table, checked September 29, 2026; per-1,000 math by The RX Index.

A source mismatch in the Constantine report: the abstract prints 9.28% vs 5.10% for ages 78–88 and an odds ratio of 4.99. Those percentages give a crude odds ratio of about 1.90, not 4.99. For the full cohort, 3.05% ÷ 2.61% is about 1.17, not the printed risk ratio of 1.09. The BMI 40+ subgroup also prints 3.15% vs 1.91% with an OR of 1.26; the crude OR is about 1.67. We could verify what the original abstract reports, but not the full-text model details that might explain these differences. Do not treat the printed ratios as calculations from the percentages shown here.

Chen’s cohort also excluded people with recent falls, fractures or recorded frailty and required later prescriptions consistent with continued treatment. Its lower fracture association should not be applied unchanged to frail older adults (Chen et al., methods).

Age raises the starting line. Even without a GLP-1, 5.10% of people aged 78 to 88 in the orthopedic study broke a bone within a year, about twice the 2.61% rate for that study's whole comparison group. So even a small change in fracture risk matters more here. If you or a parent is over 65 and weighing these drugs, our guide to GLP-1 options for adults over 65 walks through the choices.

Does exercise protect your bones on a GLP-1?

It seems to help, but it isn't proven. In a 195-person trial, liraglutide alone left hip bone density 0.013 g/cm² lower than placebo, while liraglutide plus supervised exercise showed no clear difference from placebo. The exercise-plus-drug group also wasn't clearly better than the drug alone.

Here's how the trial ran. Everyone did an 8-week, 800-calorie diet first. Then they were split into four groups for a year: placebo, exercise, liraglutide 3.0 mg, or both. Bone density changes were measured from before the diet to the end, so each number covers about 60 weeks.

Source: Jensen et al., JAMA Network Open 2024. 195 adults with obesity and no diabetes, ages 18 to 65. Table 2 gives the spine interval −0.034 to −0.004. The abstract/results sentence prints a conflicting +0.002 upper limit; the table value is used here.

No clear difference from placebo does not mean no bone was lost. In the combination group, hip BMD fell 0.019 g/cm² from before the diet to the end of treatment (95% CI −0.026 to −0.011). The same study reported higher whole-body BMD with liraglutide and with the combination than with placebo. Hip, spine and whole-body findings must stay separate (Jensen et al., Table 2 and BMD results).

Read the last row closely. Adding exercise to liraglutide didn't produce a clear difference compared with liraglutide alone. So the honest takeaway is "exercise looks helpful," not "exercise fixes it." No fragility fractures were reported in any group; this small trial was not powered to establish fracture prevention. For practical steps on holding on to muscle, see how to limit muscle loss on a GLP-1.

Where did the "30% higher osteoporosis risk" headline come from?

It came from a conference study of 73,483 matched pairs of adults with type 2 diabetes and obesity. The authors reported osteoporosis diagnoses in 4.1% of GLP-1 users vs 3.2% of non-users over five years, with a relative risk of 1.29 (95% CI 1.22–1.36). Those rounded percentages equal about 41 vs 32 per 1,000, or about 9 more—not 30% of bone lost (original AAOS abstract).

“About 30% higher” is the reported relative association. “About 9 more per 1,000” translates the rounded reported percentages. But there is a mismatch in the source’s counts that a writer needs to see.

Source: AAOS 2026 conference abstract and Table 1, also discussed by Lawand and Abboud, AAOS Now, September 18, 2026. Per-1,000 math by The RX Index from rounded percentages.

The denominator check: the same abstract prints 2,955 osteoporosis cases among GLP-1 users and 2,284 among non-users, with 73,483 people in each matched cohort. Dividing those counts by the full cohort sizes gives 4.02% vs 3.11%—40.2 vs 31.1 per 1,000—not 4.1% vs 3.2%. The abstract does not supply a different outcome-at-risk denominator that resolves this. We preserve the reported percentages and the count check separately; neither is silently substituted for the other.

The authors themselves say the study shows a link, not a cause. They note that a prescription doesn't prove someone took the drug, and that no bone scans confirmed the diagnoses. They also reported osteomalacia, a softening of the bones, at 0.2% vs 0.1% (RR 2.55, 95% CI 1.83–3.55). Its 125 vs 49 case counts equal about 1.7 vs 0.7 per 1,000 using the full cohort sizes; those crude values are kept apart from the rounded reported percentages. This is conference-level evidence with an investigator write-up, not a full peer-reviewed paper verified in this audit.

What should you ask your doctor?

These studies can sharpen the questions you bring to your prescriber. They can't tell you which drug is best for your bones or whether to start or stop one. That call depends on your health history and on the benefits you'd get from treatment.

Questions worth asking, based on where the evidence points:

• Do I already have low bone density or a past fracture? The 2024 semaglutide trial was done in exactly that group.
• Would a bone density (DXA) scan before or during treatment make sense for me?
• I don't have diabetes. Does that change how you'd watch my bones?
• I'm over 65 (or 75). How do my fall and fracture risks factor in?
• I'm eating a lot less now. Am I getting enough protein, calcium and vitamin D?
• Is strength training safe for me, and where would I start?

Our list of questions to bring to your GLP-1 prescriber covers the rest of that conversation. If you're in perimenopause, see GLP-1 choices during perimenopause. And if you're still comparing options, Find My GLP-1 Path sorts the choices by your goals, budget and insurance in about two minutes. It doesn't estimate fracture risk.

Why this matters now

This September 2026 edition brings newer fracture reports alongside earlier bone-density trials. It includes 15 selected GLP-1 study reports, and the checked Wegovy prescribing information is revised June 2026. The underlying observations come from different years; a new paper or label edition does not make its patient data new.

Source: publication or online dates from each source in the Sources list. These are publication dates, not the years the data were collected.

How we built this

We collected published results on GLP-1 drugs and bone density, osteoporosis and fractures, then recorded who was studied, what they were compared with, what was measured and for how long, before copying any number. The retained claims were checked on September 29, 2026, at the source depths listed below.

• What we included: 15 selected GLP-1 study reports (randomized trials and secondary analyses, a meta-analysis, health-record studies, a research brief and a conference study), the 6 reviewed U.S. prescribing-information documents, and 1 separate diet-trial report for context. The dataset has 80 records, including 3 dated evidence-gap records. A record is one result, label observation or gap—not one independent study. Reports may use overlapping participants.
• How deep we checked: full text and relevant results/tables for Hansen, Jensen, SELECT safety, Hamad, Huang, Chen and Alfaris; original published abstracts for Kasher Meron, Hsu, Liu, Constantine, Cheng, Iepsen and the Villareal diet trial; the full Epic Research brief, definitions and PDF text; and the original AAOS conference abstract/table plus the September investigator report. Some publisher pages were inaccessible, so original abstracts were read in PubMed or the author institution’s repository. Epic’s graphical confidence bounds were not numerically extracted. Each record states its access depth.
• Labels: we searched each official prescribing-information document for “fracture,” “bone mineral,” “osteopor,” “hip” and “pelvis,” and read the matches in context. We checked the PDF revision pages. For Mounjaro, conflicting cached PDF revisions led us to the consistent August 2026 DailyMed edition. “No statement found” applies only to the six reviewed editions.
• Population comparisons: we kept diabetes status, bone site, outcome and comparator on every record. We did not count whole studies as “better” or “worse”: Jensen’s hip and whole-body findings differ, and some reports change comparators or follow-up across diabetes groups. Hamad’s within-report sensitivity analysis is shown with its actual never-user comparator.
• The math: per 1,000 = people with the event ÷ people in the group × 1,000. Example: 17 ÷ 703 × 1,000 = 24.2. Differences use unrounded inputs, then round once to one decimal. Rounded reported percentages are multiplied by 10 and marked “about.” Modelled HRs, odds ratios and confidence intervals are copied from the primary report, not recreated from patient data. They need not equal crude proportions.
• Source conflicts: Liu’s original abstract gives −0.6% for the control group’s annualized hip BMD change. Jensen’s Table 2, rather than its conflicting abstract sentence, supplies the spine interval. Constantine’s printed ratios cannot be reconstructed from its displayed percentages without model details we could not access. AAOS’s printed case counts and full cohort sizes do not reproduce its printed percentages. Those unresolved relationships stay visible.
• What we didn't do: we didn't pool studies into one effect estimate, reanalyze patient data, or rate any drug as safest. This is a targeted evidence map, not a systematic review. Search-gap records mean no qualifying result was located in this review, not that no such study can exist.

To reproduce an original conversion, open its record in the CSV, follow the source link, and redo the math in the calculation column. To check a reported study estimate, use the listed source table or passage. Recreating a clinical model would require its data and methods, which this map does not provide.

What this data does and doesn't show

The map shows what the 15 selected GLP-1 study reports and 6 reviewed labels reported as of September 29, 2026, with one diet trial kept as context. It shows lower hip and spine BMD in selected randomized comparisons, favorable findings for other bone measures, and differing fracture associations. It does not establish that GLP-1s cause fractures, and it cannot tell you your own risk.

What it does show:

• GLP-1 treatment lowered hip and spine bone density a little in two randomized trials of people without diabetes.
• In SELECT, reported fracture adverse-event proportions were similar overall, with higher hip/pelvis proportions in women and adults 75+. Nonserious fractures were not all systematically collected.
• Some reports find different associations by diabetes status. Different endpoints, comparators, follow-up and mixed bone-site findings prevent a simple whole-study tally.
• Only Wegovy’s label among the six reviewed editions contains a clinical fracture finding.

What it doesn't show:

• That any GLP-1 causes fractures. Most fracture results come from health records, where the people who get a drug can differ from those who don't in ways the researchers can't fully measure.
• A ranking of drugs. Semaglutide and tirzepatide were never compared head to head for bones in these studies.
• What happens to human BMD after stopping treatment. No qualifying human recovery result was located in this targeted search. Animal withdrawal findings do not answer that question.
• Anything about doses or products with no data, including a randomized estimate of tirzepatide’s hip/spine BMD effect or a qualifying human bone-outcome report for orforglipron.
• Your personal risk. Trial averages don't account for your age, bone density, falls history, diet or activity.

Several studies were funded by drugmakers (Alfaris by Eli Lilly; SELECT by Novo Nordisk; Novo Nordisk supplied the drug for Hansen). Health-record studies depend on diagnosis codes, which can be wrong or missing.

How to cite this page

Cite the original study for its own findings. Cite this page when you use The RX Index's evidence map, label review or per-1,000 calculations.

Coziar, Kaden. "GLP-1 Bone Density and Fracture Risk: 15 Study Reports by Drug." The RX Index Research. September 2026; evidence checked September 29, 2026. https://therxindex.com/research/glp-1-bone-density/

Reuse: You’re welcome to reuse The RX Index’s original text, table layout, chart designs and calculations with credit to The RX Index. Keep the underlying study and label citations; their text, figures and other protected material retain their original rights and terms. No link is required.

Download the data

Download the GLP-1 Bone Evidence Map CSV.

The file contains 80 records, not 80 studies: study results, six label observations, a diet-trial context record and three dated evidence-gap records. Every record has a source or search-log reference, check date, outcome definition and verification status. Empty numeric cells mean unavailable or not extracted, not zero.

Frequently asked questions

Does Ozempic affect bone density? Yes, in one trial of people at increased fracture risk. Semaglutide 1.0 mg, a dose used by Ozempic, left hip BMD 2.59% lower and spine BMD 2.05% lower relative to placebo after 52 weeks in 64 adults without diabetes (Hansen et al., 2024). The trial was not designed to estimate fracture risk; the reviewed Ozempic label has no clinical fracture statement.

Can bone loss from Ozempic be reversed? No qualifying human study was located in this targeted review that shows whether BMD returns after stopping a GLP-1 (checked September 29, 2026). The 2024 semaglutide trial’s authors said their design could not test recovery (Hansen et al.). Ask your prescriber whether a bone density scan makes sense for you.

Does Wegovy increase fracture risk? SELECT reported similar fracture adverse-event proportions overall: 3.5% on semaglutide vs 3.6% on placebo. But not all nonserious fractures were systematically collected, and hip/pelvis proportions were higher at age 75+ (24.2 vs 6.0 per 1,000) and in women (9.8 vs 2.1 per 1,000) over the trial. The Wegovy label lists those counts; they are not annual rates or proof of a drug-caused increase.

Does tirzepatide (Zepbound or Mounjaro) cause bone loss? This map does not establish a drug-specific BMD effect. We found no randomized human trial reporting tirzepatide’s effect on hip or spine bone density, and a 2026 pooled analysis says its phase 3 trials didn't assess it (Alfaris et al., 2026). Health-record studies point both ways depending on the comparison.

Is tirzepatide worse for bones than semaglutide? The studies in this map don't show that. One linked tirzepatide to a 44% higher hazard of a combined “osteoporosis or fracture” outcome than other GLP-1s (Hsu et al., 2025), while another found both drugs linked to fewer femoral fractures than DPP-4 inhibitors in people 65+ with diabetes. Neither is a head-to-head trial.

Is it the drug or the weight loss? Weight loss may explain part of it, but the studies here do not isolate one cause. Liu linked greater weight loss with greater hip BMD decline; the diet trial also found BMD loss, but it was not matched head to head with semaglutide. Hamad’s diabetes analysis retained a lower fracture-hazard estimate after modelling weight and blood-sugar changes (Liu, Villareal, Hamad).

Does exercise stop GLP-1 bone loss? It has not been shown to stop every loss of bone density. In a 195-person trial, liraglutide plus exercise had no clear hip/spine BMD difference from placebo, but hip BMD still fell within the combination group. No fragility fractures were reported in any group; the trial was not powered to establish fracture prevention (Jensen et al., 2024).

Should I stop my GLP-1 because of my bones? These studies can't answer that for you. The right choice depends on your bone health, your other conditions and what the drug is doing for you. Bring the numbers on this page to your prescriber.

Sources

All checked September 29, 2026.

1. Copy source citation and page link S01 Hansen MS, Wölfel EM, Jeromdesella S, et al. Once-weekly semaglutide versus placebo in adults with increased fracture risk: a randomised, double-blinded, two-centre, phase 2 trial. eClinicalMedicine. 2024;72:102624. https://pmc.ncbi.nlm.nih.gov/articles/PMC11087719/
2. Copy source citation and page link S02 Jensen SBK, Sørensen V, Sandsdal RM, et al. Bone Health After Exercise Alone, GLP-1 Receptor Agonist Treatment, or Combination Treatment: A Secondary Analysis of a Randomized Clinical Trial. JAMA Network Open. 2024;7(6):e2416775. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2820308
3. Copy source citation and page link S03 Novo Nordisk. Wegovy (semaglutide) U.S. prescribing information, revised 06/2026. https://www.novo-pi.com/wegovy.pdf
4. Copy source citation and page link S04 Kushner RF, Ryan DH, Deanfield J, et al. Safety profile of semaglutide versus placebo in the SELECT study: a randomized controlled trial. Obesity. 2025;33(3):452–462. https://pmc.ncbi.nlm.nih.gov/articles/PMC11897845/
5. Copy source citation and page link S05 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(7):e2625141. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2852068
6. Copy source citation and page link S06 Huang YN, Tsou MY, Li PH, et al. Associations of Semaglutide With Skeletal Outcomes in People With Obesity, With and Without Type 2 Diabetes: A Target Trial Emulation. Diabetes, Obesity and Metabolism. 2026;28(7):5834–5847. https://dom-pubs.onlinelibrary.wiley.com/doi/full/10.1111/dom.70786
7. Copy source citation and page link S07 Chen HY, Wu JY, Chu YH, Chen TW, Huang CF. Lower fall and femoral fracture risks with semaglutide and tirzepatide compared with DPP-4 inhibitors in older adults with type 2 diabetes. Osteoporosis International. Published July 11, 2026. https://link.springer.com/article/10.1007/s00198-026-08134-y
8. Copy source citation and page link S08 Kasher Meron M, Hornik-Lurie T, Twig G, Rotman-Pikielny P. GLP-1 receptor agonists and the risk of fragility fractures in older adults with type 2 diabetes. Journal of Clinical Endocrinology & Metabolism. 2026;111(7):1949–1958. https://pubmed.ncbi.nlm.nih.gov/41665888/
9. Copy source citation and page link S09 Hsu YH, Liang YC, Chan KC, et al. Association of tirzepatide use with risk of osteoporosis compared with other GLP-1 receptor agonists: A retrospective cohort study using the TriNetX database. Diabetes Research and Clinical Practice. 2025;230:112995. https://www.sciencedirect.com/science/article/abs/pii/S0168822725010101
10. Copy source citation and page link S10 Liu Y, Walzer D, Schmitz S, et al. Skeletal effect of semaglutide and tirzepatide in patients with increased risk of fractures. Journal of Clinical Endocrinology & Metabolism. 2026;111(7):1959–1966. https://academic.oup.com/jcem/article-abstract/111/7/1959/8467122
11. Copy source citation and page link S11 Epic Research. GLP-1 Use Associated with Lower Osteoporosis Risk in Adults with Type 2 Diabetes but Higher Risk in Adults Without Diabetes. March 3, 2026. https://www.epicresearch.org/articles/glp-1-use-associated-with-lower-osteoporosis-risk-in-adults-with-type-2-diabetes-but-higher-risk-in-adults-without-diabetes/
12. Copy source citation and page link S12 Constantine E, Enthoven L, Kahan R, Pflug E, Lauder A. The Impact of Glucagon-Like Peptide-1 Receptor Agonists on Fracture Risk in Overweight or Obese, Nondiabetic Patients. Journal of the American Academy of Orthopaedic Surgeons. 2026;34(12):e1610–e1619. https://doi.org/10.5435/JAAOS-D-24-01505
13. Copy source citation and page link S13 Wajahath M, Lawand JJ, Hill BW, Khan A, Abboud JA, Horneff JG. AAOS 2026 conference abstract and Table 1, submission 008736. https://submissions.mirasmart.com/Verify/AAOS2026/Submission/out/AAOS2026-008736.PDF. Investigator report: Lawand J, Abboud J. New studies link GLP-1 receptor agonist medications to tendon rupture risk and changes in bone metabolism. AAOS Now. September 18, 2026. https://www.aaos.org/aaosnow/2026/sep/research/research01/
14. Copy source citation and page link S14 Cheng L, Hu Y, Li YY, et al. Glucagon-like peptide-1 receptor agonists and risk of bone fracture in patients with type 2 diabetes: A meta-analysis of randomized controlled trials. Diabetes/Metabolism Research and Reviews. 2019;35(7):e3168. https://doi.org/10.1002/dmrr.3168
15. Copy source citation and page link S15 Alfaris N, Kushner RF, Li J, et al. Tirzepatide for Obesity in Adults ≥65 Years: A Post Hoc Analysis of the SURMOUNT and SUMMIT Clinical Trials. Diabetes, Obesity and Metabolism. 2026;28(9):8072–8083. https://pmc.ncbi.nlm.nih.gov/articles/PMC13449079/
16. Copy source citation and page link S16 Novo Nordisk. Ozempic (semaglutide) U.S. prescribing information, revised 05/2026. https://www.novo-pi.com/ozempic.pdf
17. Copy source citation and page link S17 Eli Lilly. Mounjaro (tirzepatide) U.S. prescribing information, revised 08/2026. https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=d2d7da5d-ad07-4228-955f-cf7e355c8cc0
18. Copy source citation and page link S18 Eli Lilly. Zepbound (tirzepatide) U.S. prescribing information, revised 08/2026. https://pi.lilly.com/us/zepbound-uspi.pdf
19. Copy source citation and page link S19 Novo Nordisk. Saxenda (liraglutide) U.S. prescribing information, revised 02/2026. https://www.accessdata.fda.gov/drugsatfda_docs/label/2026/206321s025lbl.pdf
20. Copy source citation and page link S20 Eli Lilly. Foundayo (orforglipron) U.S. prescribing information, revised 04/2026. https://www.accessdata.fda.gov/drugsatfda_docs/label/2026/220934Orig1s000lbl.pdf

1. Copy source citation and page link S21 Iepsen EW, Lundgren JR, Hartmann B, et al. GLP-1 Receptor Agonist Treatment Increases Bone Formation and Prevents Bone Loss in Weight-Reduced Obese Women. Journal of Clinical Endocrinology & Metabolism. 2015;100(8):2909–2917. DOI: 10.1210/jc.2015-1176. Original abstract in the authors’ institutional repository: https://research.regionh.dk/en/publications/glp-1-receptor-agonist-treatment-increases-bone-formation-and-pre/
2. Copy source citation and page link S22 Villareal DT, Fontana L, Weiss EP, et al. Bone mineral density response to caloric restriction-induced weight loss or exercise-induced weight loss: a randomized controlled trial. Archives of Internal Medicine. 2006;166(22):2502–2510. https://pubmed.ncbi.nlm.nih.gov/17159017/. Diet context, not a GLP-1 study.

The RX Index Research is the research and reference section of therxindex.com.