Author response: Clinical outcomes associated with prior glucagon-like peptide-1 receptor agonist exposure in burned patients

Correspondence to: Saif Badran (e-mail: badran@wustl.edu)

Division of Plastic Surgery

Department of Surgery

WashU Medicine

Suite 1150

Northwest Tower

660 South Euclid Avenue

St Louis

Missouri

63110

USA

BJS, https://doi.org/10.1093/bjs/znag048, published 20 April 2026

_____

We thank Li and colleagues for their close reading of our study¹ and for the opportunity to respond.

On calendar time, we accept the substance of the point. Index year was not among our matching variables, and exposure prevalence rose from 1.3% in 2018 to 10.1% in 2025, so exposed patients are weighted toward later index years. Adjustment on the index year is the right approach in confirmatory work. We would distinguish this from time zero alignment, however: exposure was recorded entirely before the index burn, so exposure assignment, time zero, and the start of follow-up coincide—consistent with the TARGET framework's requirement that eligibility, treatment assignment, and follow-up initiation are synchronized to avoid design-related biases². The concern here is that burn care itself changed over these years, and those changes would have to be very large to explain a halving of the odds of death. Furthermore, the consistency of the mortality effect across all three exposure windows (OR 0.46–0.48), which span different calendar-time distributions, provides indirect evidence against secular trend as the sole explanation.

On competing risks, the mechanism is real, though its likely size here is small. In the primary cohort, there were 177 fewer deaths and 19 more recorded hypertrophic scarring events in the exposed arm, so differential survival could account for the scarring difference only if more than one in ten of those additional survivors had been recorded with hypertrophic scarring. This proportion substantially exceeds the overall hypertrophic scarring rate observed in the cohort, making a purely survivorship-driven explanation unlikely. Mortality reduction was of similar magnitude across all windows (OR 0.46 to 0.48), so the same holds at 3 years. We have also previously reported an association in the same direction in GLP-1 receptor agonist users undergoing elective abdominoplasty, over a fixed 90-day postoperative window in which death is not a meaningful competing event³. Preclinical data further support biological plausibility: GLP-1 receptor activation modulates fibroblast function, matrix remodeling, and inflammatory pathways involved in scar formation⁴. providing a mechanistic basis independent of survival bias. Our paper called for a prospective study of the scarring association rather than treating it as established, and we hold to that.

On multiplicity, the principal associations—mortality, intensive care unit admission, intubation, sepsis, and pneumonia—were each significant at P < 0.001 and would survive Bonferroni correction across the outcomes of the primary analysis, which is a stricter standard than the false-discovery-rate correction they proposed. The 3-year scarring association, at P = 0.031, would not survive that correction—a point we acknowledged by framing it as hypothesis-generating. We would also note that the exposure windows were sensitivity analyses on largely the same patients, so they were not three independent sets of tests.

On the active-comparator new-user design, we agree this is the stronger framework and that residual healthy-user and adherer effects cannot be excluded here. After matching, 80% of control patients carried a diagnosis of diabetes mellitus, and mean haemoglobin A1c was equivalent between cohorts, so for most of the cohort, the comparison was not with metabolically healthy controls. We have recently applied an active comparator, though not a new-user design, to the scarring outcome in the abdominoplasty cohort, using dipeptidyl peptidase-4 inhibitor users—a class that also raises endogenous GLP-1—and the association persisted.³ An active comparator drawn from other glucose-lowering drugs would also restrict the study to patients with diabetes, dropping the roughly one-fifth of our exposed patients with no recorded diabetes diagnosis; this trade-off between reduced confounding and reduced generalizability should be weighed in future study design. We note that a recent independent TriNetX study by Dao et al., using a different design (GLP-1 receptor agonist initiation within 30 days after burn), similarly found reduced mortality, infection, and opioid prescriptions in GLP-1 receptor agonist users5, providing convergent evidence from a complementary analytic approach.