Women over 50 who already had osteoporosis and took a GLP-1 drug were recorded with about half as many collapsed vertebrae as similar women who did not. In a database of 56,142 matched pairs, the vertebral fracture rate was about 1.9 percent in the GLP-1 group and 4.3 percent in the comparison group. Spine surgeries to shore up a broken vertebra, the cement-injection procedures called kyphoplasty and vertebroplasty, were down by a similar margin.
Those are the headline numbers from a retrospective study published online September 1 in Clinical Spine Surgery ↗ by a team at Temple University in Philadelphia. They land on a question the last round of GLP-1 bone research left open.
Why this population is the one that counts
The worry that trails this drug class is that fast weight loss strips bone along with fat, leaving a lighter frame that snaps more easily. A large records study in June ↗ pushed back on that fear, finding fewer fractures in GLP-1 users than in people on other weight-loss drugs. But it carried a catch its own authors named. The comparison was GLP-1 against other diet drugs, not GLP-1 against nothing, and it was run in overweight adults, not in people whose bones were already thin.
This new study closes both gaps. Everyone in it already carried an osteoporosis diagnosis, the group where a fragile vertebra is a real and near-term risk rather than a decade-away hypothetical. And the comparison is GLP-1 use against no GLP-1 use, the absolute question the earlier paper could not answer.
What the study compared
The team pulled deidentified records from TriNetX, a network that pools electronic health records across many hospitals, and selected female patients over 50 with osteoporosis and no fracture on the books, across a ten-year window ending January 2025. They split them by whether a GLP-1 receptor agonist was prescribed after the osteoporosis diagnosis, drugs like semaglutide ↗ (Ozempic and Wegovy), dulaglutide ↗ (Trulicity), and the dual GIP and GLP-1 ↗ agonist tirzepatide ↗ (Mounjaro and Zepbound), all of which act through the GLP-1 receptor ↗.
Then they used propensity matching, a statistical method that pairs people with similar traits so the two groups look alike on paper. Here the matching went past the usual age and weight to the variables that actually drive bone risk: body mass index, HbA1c, kidney function, existing osteoporosis medications, and blood levels of calcium, phosphate, and vitamin D. On vertebral fracture the GLP-1 group came out at a relative risk of 0.43, meaning a bit under half the rate of the matched controls. On the surgeries the figure was 0.45.
The catch, and one that cuts the other way
Read it the way the last bone paper asked to be read. This is observational, built from care that already happened, not a trial that randomly assigned anyone to a drug. A doctor who starts an older woman on a GLP-1 drug tends to be treating someone healthier and more engaged with her own care than the woman who never gets the prescription, and those traits protect the spine for reasons that have nothing to do with pharmacology. A relative risk of 0.43 is a large effect, large enough that some of it is probably the patients rather than the drug. The authors grade their own work Level III evidence, the observational tier.
One bias, though, runs against the finding rather than for it. Women who see doctors often enough to be started on a GLP-1 drug also get imaged more, and more imaging finds more of the silent vertebral fractures that never reach a clinic on their own. That would inflate the fracture count in the GLP-1 group, not shrink it. The protective signal survived a headwind that should have pushed it the other way.
There is a plausible mechanism underneath. GLP-1 signaling appears to act on bone turnover directly, and these drugs have shown gains in spine bone density in earlier work. Weight loss offloading the joints does not explain a vertebral result, since a collapsed vertebra is a density problem, not a load problem.
What would settle it is the study that still does not exist: a randomized trial that assigns people with osteoporosis to a GLP-1 drug or a placebo and counts fractures over years. Until then, the fragile-skeleton fear keeps failing to show up in exactly the population that had the most to fear from it.