Skip to content
This GLP
← Research
Evidence

Bone density rose. Whether fractures fell, the trials cannot say

Twenty-five trials found higher bone density at every site measured. The fracture interval runs from 0.47 to 1.36 — consistent with a halving and with a rise.

Ruth Alvarez6 min read
Fracture risk ratio, 25 trials1.00.471.36Bone density gain, grams per square centimeterLumbar spine+0.07Hip neck+0.05Total hip+0.06

Losing weight quickly takes bone with it, which is why the skeleton keeps coming up around this drug class and why a reader who intends to stay on one for a year has a fair reason to ask about it. A meta-analysis of twenty-five randomized trials in type 2 diabetes gathered what has actually been measured [1], and the answer separates into two parts that point in different directions — a separation worth keeping straight before any of it reaches a claim on a seller’s page.

Density went up

Bone mineral density was higher on a GLP-1 receptor agonist than in the control arms at every site measured: the lumbar spine by 0.07 g/cm² (95% CI 0.06 to 0.09, P < 0.00001), the hip neck by 0.05 g/cm² (95% CI 0.03 to 0.08, P = 0.0001) and the total hip by 0.06 g/cm² (95% CI 0.04 to 0.07, P < 0.00001). The turnover markers moved consistently with that: formation markers rose — P1NP at a standardized mean difference of 0.33 (95% CI 0.07 to 0.59), osteocalcin by 1.46 µg/L (95% CI 1.10 to 1.83), bone-specific alkaline phosphatase by 0.91 µg/L (95% CI 0.19 to 1.63) — while β-CTX, which tracks resorption, fell at a standardized mean difference of −0.34 (95% CI −0.54 to −0.14). Calcium, phosphate and total alkaline phosphatase did not move at all.

Fractures did not answer

The outcome that matters to a person rather than to a scan is whether bones break, and there the pooled risk ratio was 0.80 with a 95% confidence interval of 0.47 to 1.36 (P = 0.41). The authors summarize that as no significant association, which is accurate, but the interval is the part to read: it is compatible with fractures falling by more than half and with them rising by more than a third. That is not reassurance. It is a set of trials too small, and followed for too little time, to distinguish between those two worlds — the same problem we found where two meta-analyses of the same six trials reached opposite conclusions.

And the cohort is not this market

Every trial pooled here enrolled people with type 2 diabetes, whose bone behaves oddly: density often reads normal or high while fracture risk is raised anyway, so density is a worse proxy in that group than in anyone else. A buyer without diabetes — most of the people reading a telehealth seller’s page — is outside the population these numbers describe. We have written before about a fracture finding that stopped at the edge of the diabetes cohort, and this is the same boundary seen from the other side.

The practical version is dull and unchanged: resistance training and enough protein are what the evidence supports for protecting lean tissue and bone during weight loss, and neither is something a seller charges for. What a seller does control is whether a clinician is reachable to raise it with, which is the gap we count in questions they leave open.

Frequently asked

Do GLP-1 drugs weaken bone?
In these twenty-five trials bone density was higher on the drug, not lower, at the lumbar spine, hip neck and total hip. Whether that translates into fewer broken bones is not established.
What does the fracture result actually say?
A pooled risk ratio of 0.80 with a 95% confidence interval of 0.47 to 1.36. That spans a near-halving and a rise of more than a third, so it distinguishes nothing.
Does this apply to me if I do not have diabetes?
Not directly. Every trial pooled here enrolled people with type 2 diabetes, in whom bone density is a poorer guide to fracture risk than it is in anyone else.

Sources

  1. [1] Tan Y, et al. (2025). Effect of GLP-1 receptor agonists on bone mineral density, bone metabolism markers, and fracture risk in type 2 diabetes: a systematic review and meta-analysis Acta Diabetologica. PMID 39985672

Where to get it

Best GLP-1 injections

Every injectable seller we can verify, with the price each one publishes and an honest read of what the trials measured.

Compare providers →

More in Evidence