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.