Bone is the outcome that gets raised in comment threads and never in a product page. It has an obvious mechanism behind the worry — rapid weight loss of any cause is associated with bone loss — and it now has two large syntheses behind it, which reach the same conclusion about the outcome that matters and different conclusions about the one that is easier to measure. It is a useful contrast with the reporting-database picture in what the adverse event database says, where volume of reports and weight of evidence pull in opposite directions.
The outcome that matters: fractures
A 2025 systematic review searched nine databases and pooled 25 studies of GLP-1 receptor agonists in type 2 diabetes [1]. Fracture incidence was not significantly raised: a relative risk of 0.80 with a 95% confidence interval from 0.47 to 1.36 (p=0.41).
That interval is worth reading rather than skipping. It runs from a halving of risk to a third more risk. The honest translation is not that the drug protects bone; it is that the pooled trials were not large enough to detect a difference in either direction, and no difference emerged.
The 2026 review reached the same place from a much larger base: 60 articles covering 46 randomized trials, 13 real-world studies and one pharmacovigilance study, across 1,250,717 individuals [2]. No effect on fractures at any site.
The outcome they disagree about: density
The 2025 review found bone mineral density better on treatment than control at every site measured — lumbar spine by a mean difference of 0.07 g/cm² (95% CI 0.06 to 0.09), hip neck by 0.05 (0.03 to 0.08), total hip by 0.06 (0.04 to 0.07). Bone turnover markers moved in the direction that usually accompanies that, with β-CTX, a marker of resorption, falling (SMD −0.34, 95% CI −0.54 to −0.14).
The 2026 review found none of it. Once its models used the most adjusted effect estimate available from each study, no effect on bone mineral density remained at any site.
The finding underneath both
The 2026 review’s clearest result was not about bone at all. Across 28 comparisons it found lean body mass falling consistently on these drugs (standardized mean difference −0.52, 95% CI −0.8 to −0.23), driven mainly by liraglutide and semaglutide, and robust to every sensitivity analysis it ran. Joint symptoms did not change.
That is the musculoskeletal effect with evidence behind it, and it has a price attached that nobody quotes — resistance training, protein, and the time both take. We put numbers on that in the cost nobody prices: keeping muscle, rather than repeating the argument here.
What a buyer does with this
Very little, which is the useful conclusion. Bone is not a reason to avoid these drugs on current evidence, and it is not a reason to buy one either. It belongs on the list of things worth mentioning to a clinician if you already have osteoporosis or a fracture history — and that presumes a clinician you can reach, which is its own variable in this market.
What is worth noticing is the asymmetry in what gets published. A seller will tell you about weight loss because it sells, and will not tell you about lean mass, bone, or what happens when you stop, because none of those do. The shape of that silence is the subject of what nobody says about a higher dose and what stopping costs, and the underlying question of whether the whole purchase is worth it is in is a GLP-1 worth what it costs.