Large GLP-1 Drug Review Takes Closer Look at Risks, Rewards

GLP-1 medications showed a range of potential benefits and safety concerns in an umbrella review of non-cardiometabolic outcomes, though the data quality for many remained limited.

Across 60 meta-analyses that assessed 116 non-cardiometabolic outcomes, some of the most consistent signals were for adverse gastrointestinal events based on moderate- or high-quality evidence. This included higher odds of nausea (OR 2.47, 95% CI 1.84-3.34), vomiting (OR 2.78, 95% CI 1.91-4.06), and diarrhea (OR 1.94, 95% CI 1.52-2.49), Yongze Li, MD, PhD, of the First Hospital of China Medical University in Shenyang, and colleagues reported in JAMA Network Open.

High-quality, convincing evidence also suggested possible protective associations against infection-related outcomes, especially for serious infections (OR 0.89, 95% CI 0.87-0.92).

Other signals also emerged across fracture, respiratory, neurologic, psychiatric, and endocrine outcomes. Despite reaching significance, most pooled outcomes didn’t reach prespecified high-credibility thresholds, rendering the associations exploratory and hypothesis-generating.

“Clinically, gastrointestinal adverse events with GLP-1 receptor agonists are well recognized, and our results confirmed these as the most consistent and robust signals,” Li told MedPage Today.

“What was more notable, however, was that for many other non-cardiometabolic outcomes, especially those that have recently attracted attention — such as infections, respiratory diseases, and cancer-related outcomes — the evidence was often suggestive based on limited credibility,” he continued. “This reinforces the gap between emerging signals and high-certainty evidence.”

Some exploratory associations included lower odds for fracture, incident respiratory disease, and all-cause dementia. Other signals linked GLP-1 drug use to higher odds for resolution of metabolic dysfunction-associated steatohepatitis, gastroesophageal reflux disease, gallbladder or biliary diseases, and thyroid disease.

Not all statistically significant findings should be interpreted as clinically actionable, Li warned.

“For GLP-1 receptor agonists, gastrointestinal adverse events are consistent and expected. However, for many other outcomes … the current evidence should be interpreted cautiously,” he explained. “Clinicians should avoid overinterpreting early signals and continue to rely primarily on well-established cardiometabolic benefits when making treatment decisions.”

“Specifically, any identified safety signals should prompt nuanced, individualized risk-benefit discussions and targeted monitoring — particularly for symptomatic patients or those with high-risk clinical profiles — rather than routine alterations to GLP-1 receptor agonist prescribing patterns or surveillance protocols,” the team added. “GLP-1 receptor agonists continue to be a cornerstone of antidiabetic therapy but should not be misconstrued as a panacea.”

Li called for large-scale, long-term randomized trials and well-designed real-world studies to clarify whether these suggested associations translate into meaningful clinical benefits or risks.

The umbrella review included 1,751 randomized clinical trials with over 3.5 million participants. Studies were identified by a systematic search of databases through January 2026; the majority were published within the past decade.

Most study populations (71.7%) involved people with type 2 diabetes and a third involved those with obesity. Follow-up durations ranged from 3 months to 5.4 years or longer.

Li and co-authors categorized evidence certainty by the GRADE (Grading of Recommendations Assessment, Development and Evaluation) framework. Credibility of evidence was classified as convincing (class I), highly suggestive (class II), suggestive (class III), weak (class IV), or nonsignificant based on a handful of factors like number of trial participants and extent of statistical significance.

“This allowed us to distinguish between findings that are statistically significant and those that are truly robust and clinically reliable,” Li explained. “In this sense, our work helps move the field from signal detection to evidence appraisal.”

Evidence regarding cancer risks was inconclusive. While preclinical data have raised concerns about thyroid cancer, the researchers noted that the relevance of these findings to humans “remains unconfirmed.”

Since many non-cardiometabolic outcomes were captured in studies as adverse events rather than prespecified endpoints, data may have been susceptible to misclassification and reporting bias. Other limitations included a lack of stratification by GLP-1 drug dose or treatment duration.