GLP-1 drugs linked to lower fracture risk

Type 2 diabetes changes bone quality, and every glucose-lowering drug class gets scrutinised for whether it adds to or subtracts from fracture risk. GLP-1 receptor agonists are now used very widely, and because they also produce weight loss there has been a reasonable worry that bone could suffer. The published data had pointed in both directions, with individual trials too small to settle the question on their own. Pooling across trials, and comparing agents against each other, is one way to see whether a class-wide signal exists at all.

The authors ran a Bayesian network meta-analysis, registered on PROSPERO, searching electronic databases for published randomised controlled trials and ClinicalTrials.gov for unpublished trial data. Fifty-four randomised controlled trials met the criteria, covering 49,602 participants with type 2 diabetes, of whom 28,353 were treated with a GLP-1 receptor agonist. Comparators included placebo and other anti-hyperglycaemic drugs. Risk ratios with 95% confidence intervals were pooled for fracture outcomes, and the analysis also ranked individual agents by their probability of being the safest option for fracture.

As a class, GLP-1 receptor agonists were associated with a decreased fracture risk relative to placebo or other anti-hyperglycaemic drugs. Among individual agents, exenatide carried the lowest fracture risk versus placebo, with a risk ratio of 0.17 and a 95% confidence interval from 0.03 to 0.67. The ranking placed exenatide first, followed by dulaglutide, liraglutide, albiglutide, lixisenatide and semaglutide. The authors also report that a statistically significant inconsistency was observed in some of the comparisons within the network, which limits how firmly the ordering of agents can be read.

This addresses a literature the authors describe as conflicted, and it points away from the concern that this drug class harms bone in type 2 diabetes. Because fracture was not the primary endpoint of most contributing trials, the events being pooled are relatively sparse, which is the likeliest explanation for the very wide confidence interval around the best-ranked agent. Network inconsistency means the head-to-head ordering should be treated as exploratory rather than as a prescribing hierarchy. Dedicated trials with bone endpoints, and longer follow-up in people losing substantial weight, remain the open questions.

For anyone taking or considering a GLP-1 receptor agonist for type 2 diabetes, this is a reason to think the class is not associated with more fractures, and may be associated with fewer. It is not evidence that one specific agent protects bone better than another, because the comparisons between drugs were inconsistent and imprecise. The finding applies to people with type 2 diabetes enrolled in trials, not to healthy adults using these drugs for weight loss. Questions about which medication suits an individual belong with the prescriber who knows their history.