COIOSDeep risk insights through multi-disciplinary research across the globe
A bright stationery store interior with shelves of greeting cards and office supplies
Item contradicts Peer-reviewed Added

GLP-1 receptor agonist cancer benefits may reflect immortal time bias in study design

Simulated study designs with no true effect gave hazard ratios of 0.39 to 0.85, close to published GLP-1 cancer estimates.

The authors examined two observational studies citing target trial emulation that linked GLP-1 receptor agonists with lower cancer risk. Both counted patients as exposed after a second prescription but started follow-up at the first. Monte Carlo simulation reproduced these designs, with 500 replications of 5,000 people and a true hazard ratio of 1.0. Varying the gap between prescriptions gave hazard ratios of 0.72 to 0.39 over longer follow-up, 0.85 to 0.60 over shorter. Aligning eligibility, treatment assignment and time zero removed the bias and recovered hazard ratios centred on 1.0. The work uses simulated data only and does not re-analyse the original patient records.

Why it is interesting: Recent observational studies reported large protective associations between GLP-1 receptor agonists and cancer. This simulation shows similar estimates arise when exposure is defined after time zero, with no true effect.

Source
Journal of Clinical Epidemiology, 20 November 2026
Paper
Laundered Survival: When Target Trial Emulation Doesn't Emulate-Implications for GLP-1 Receptor Agonists and Cancer Outcomes
Authors
Brophy JM
DOI
10.1016/j.jclinepi.2026.112595
Type
Peer-reviewed article
Design
Conceptual appraisal plus Monte Carlo simulation (500 replications, n=5,000 each) reproducing two published GLP-1/cancer target trial emulation designs under a null hazard ratio
Verdict
Contradicts prior evidence
Driver
Obesity, diabetes and GLP-1

Other drivers in Health today