Adults who started a GLP-1 receptor agonist had a 21% lower three-year risk of fragility fracture than adults who started a DPP-4 inhibitor, according to a study of 133,606 matched patients published July 24, 2026 in JAMA Network Open. But the study also found the opposite signal in people taking these drugs without type 2 diabetes, where fracture risk was 13% higher.
Key takeaways
- In the full matched cohort, GLP-1 users had about a 21% lower relative risk of fragility fracture (hazard ratio 0.79). In absolute terms the difference is small: a 0.79 percentage point drop, which works out to roughly 8 fewer fractures per 1,000 people treated for three years, or one fracture prevented for every 126 people treated (number needed to treat, 126).
- The benefit was concentrated in people with type 2 diabetes, whose risk was about 9% lower (hazard ratio 0.91). Users without type 2 diabetes went the other way, with about a 13% higher risk (hazard ratio 1.13).
- This is observational data, not a randomized trial, so it shows association, not cause.
What the study found
The researchers used target trial emulation, a method that takes real-world records and analyzes them the way a randomized trial would be designed. The data came from the TriNetX Research Network. After propensity score matching, the analysis included 133,606 patients, split into 66,803 per group, with a mean age of about 63 years.
Over three years, GLP-1 initiation was linked to lower fragility fracture risk compared with DPP-4 inhibitor initiation: about 21% lower in relative terms (hazard ratio 0.79; 95% CI, 0.76 to 0.83, meaning the true reduction is very likely somewhere between 17% and 24%). The absolute difference was modest. Risk fell by 0.79 of a percentage point, which is about 8 fewer fractures per 1,000 people over three years, or one fracture prevented for every 126 people treated for three years (number needed to treat, 126).
The protection was strongest at the sites that matter most in my line of work. Vertebral fractures were about 32% lower (hazard ratio 0.68) and hip or femur fractures about 30% lower (hazard ratio 0.70). The authors reported the association held independent of changes in BMI and HbA1c.
Then comes the part almost nobody is leading with. In the subgroup analysis, the protective effect was seen in adults with type 2 diabetes, whose risk was about 9% lower (hazard ratio 0.91), while GLP-1 use was linked to increased fracture risk in users without type 2 diabetes, about 13% higher (hazard ratio 1.13). The interaction test was significant at P less than .001, meaning there is less than a 1 in 1,000 chance the split between the two groups is a coincidence. One honest limit here: the study reported an absolute risk reduction and number needed to treat for the overall cohort, but not broken out for the non-diabetic subgroup, so I can give you the relative 13% for that group but not a reliable “X per 1,000.”
Dr. Kumar’s take
The headline everyone will run is “GLP-1s protect your bones.” That headline is only true for the population the study was built around, people with type 2 diabetes. The number I would circle is 1.13, because the non-diabetic subgroup is exactly the group now filling clinics and telehealth queues for weight loss alone.
Here is why I think the reversal makes sense. Diabetes itself damages bone quality through poor glucose control, advanced glycation end products and higher fall risk. Fixing the metabolic problem plausibly fixes some of that bone problem, and the benefit is big enough to swamp anything else going on. Take that metabolic repair away and you are left with the raw skeletal effect of fast, large weight loss: less mechanical load through the spine and hips, and loss of lean mass along with fat. Muscle pulls on bone, bone responds to load, and removing both quickly makes the skeleton downshift.
I see the downstream version of this in spine practice. Thin, deconditioned patients with low muscle mass fracture vertebrae from things that should not break anything, a hard sneeze or a missed step off a curb. A 32% lower vertebral fracture risk (hazard ratio 0.68) in the overall cohort is real, but it does not transfer to a 58-year-old with a normal A1c who has dropped 40 pounds in seven months.
The caveats matter. This is a matched observational cohort, not randomization, and people without type 2 diabetes who get these drugs differ from those who do. The non-diabetic group is also smaller, so its estimate is less stable. But the effect went in the opposite direction with a significant interaction, which is harder to explain away than a null result.
What it means for you
If you have type 2 diabetes and are on a GLP-1, this is reassuring. Nothing here argues for stopping.
If you are taking one purely for weight loss, treat your skeleton as something you have to actively defend. Resistance training at least twice a week, loading the spine and hips, not just walking. Protein at roughly 1.2 to 1.6 grams per kilogram of body weight per day so you lose fat and not muscle. Adequate vitamin D and calcium, and a slower rate of loss if the weight is coming off very fast. Lifting weights also lowers your risk of type 2 diabetes, so it pays twice.
Ask about a baseline DEXA scan before you start if you are postmenopausal, over 65, already thin, on steroids or a medication known to affect bone, or have fractured before. Knowing your starting bone density is what turns this study from a scary headline into something you can act on.