Weight for rhythm: is the evidence for atrial fibrillation risk with semaglutide still premature?: reply

We sincerely thank Dr Shahmohamadi and colleagues1 for their thoughtful and constructive comments regarding our recent meta-analysis on semaglutide and the incidence of new-onset atrial fibrillation (AF). We fully agree that AF was not a prespecified primary endpoint in most of the included trials. As acknowledged in the Limitations section of our manuscript, this represents an important methodological constraint.2 We have clearly stated that AF events were mostly collected as adverse events, without systematic electrocardiogram (ECG) monitoring or adjudication, and that this may have led to underreporting or detection bias. Nevertheless, we maintain that meta-analyses based on adverse event reporting are an accepted and widely used strategy in the literature, particularly when transparently described and interpreted with appropriate caution, as in our work. While AF was not a primary endpoint, its consistent association with adverse outcomes makes its prevention an important secondary objective in cardiometabolic care. Moreover, unfortunately, individual participant data regarding baseline AF history are not consistently reported across trials, limiting the feasibility of adjustment in meta-analytic synthesis based on aggregate data. We also appreciate the authors’ effort in replicating our findings using a Bayesian approach. As they correctly point out, Bayesian hierarchical models may offer additional insight in the context of rare events. However, we believe this approach should be seen as complementary, not a replacement for conventional random-effects meta-analysis, which remains the standard in evidence synthesis. Importantly, our analysis showed no heterogeneity (I² = 0%) and was supported by consistent results across multiple sensitivity analyses. To further address the constructive concern raised by the authors regarding the impact of sparse event data, we performed an additional post hoc analysis excluding all trials with zero events in either treatment arm, as recommended in rare-event methodology. This analysis included 12 trials (19 557 semaglutide-treated and 17 651 controls) and yielded an odds ratio of 0.86 (95% confidence interval: 0.72–1.03; P = 0.10), with no statistical heterogeneity (I² = 0%) (Figure 1). While this result did not reach statistical significance, it remains directionally consistent with our primary analysis, supporting the robustness of the observed trend, albeit with reduced precision due to the smaller number of included studies. Effect of semaglutide therapy on new onset of atrial fibrillation. Random effects, 95% confidence intervals and I2 statistics. As the authors rightly note, the hypothesis of semaglutide providing protection against AF is biologically plausible, due to its effects on weight loss, glycaemic control, and inflammation. In our view, this provides further justification for considering the observed association as a signal that deserves attention—hypothesis generating, as already reflected in the tone of our conclusions. Future meta-analyses using individual patient-level data could allow for better control of baseline characteristics, including AF history and ECG surveillance methods. Indeed, in a previous Letter to the Editor in this journal, we had already emphasized the need for prospective trials specifically designed to assess semaglutide’s potential antiarrhythmic effects through rigorous and standardized AF monitoring.3 We are grateful to the authors for their contributions to the scientific dialogue on this important topic, and we believe that such discussions help refine future research priorities. We therefore appreciate the authors’ cautionary perspective and fully endorse their call for future trials specifically designed to evaluate AF as a primary endpoint, using continuous or protocolized monitoring strategies. Until such data become available, our findings offer a comprehensive synthesis of current evidence on this topic, and we believe they contribute meaningfully to the ongoing discussion around the cardiovascular pleiotropy of glucagon-like peptide 1 receptor agonists.