Authors’ reply

We express our gratitude to Dziewierz et al¹ for their thorough and insightful analysis of our research letter published in the recent issue of Polish Archives of Internal Medicine.² Due to the constraints of the research letter format, only essential findings from our large‑scale data analysis were included in the final manuscript and supplement. We appreciate and largely agree with the comments provided, and are pleased to offer further clarification.

However, we respectfully disagree with the suggestion that the evolving landscape of glucagon‑like peptide‑1 receptor agonists (GLP‑1 RAs) should limit our analysis to newer molecules, such as semaglutide and tirzepatide, which indeed have recently dominated the market. Liraglutide still remains a relevant and frequently prescribed treatment for patients with type 2 diabetes and obesity in Poland.

Regarding the identification of dulaglutide as an independent predictor of vision loss in our study, we acknowledge that this finding warrants further investigation. The retrospective nature of our analysis, coupled with limited baseline demographic and clinical data, restricts our ability to provide deeper insights. Unaccounted variables (time from diagnosis and control of diabetes, other comorbidities, patients adherence or compliance, etc.) may have introduced selection bias, which could explain this association. Nevertheless, we agree that this observation may serve as a hypothesis‑generating finding for future research by other groups with access to larger, more comprehensive datasets.

Furthermore, the median time to onset of first adverse events was calculated based only on data from the patients that experienced the event; hence, it ignores censoring and should not be compared directly with the Kaplan–Meier analysis provided in the Supplementary material. However, it can be observed that the Kaplan–Meier curve declines slightly more rapidly in the GLP‑1 RA group, which can reflect shorter time‑to‑event. We were able to perform a comparable Kaplan–Meier analysis that ignored censored patients altogether, and the resulting P value was 0.01 for the log‑rank test—similar to the P value for comparison of the median time‑to‑event. However, this analysis was not reported in the paper.

In our study, in the patients with diabetes, the use of GLP‑1 RAs as an independent predictor of primary outcome occurrence yielded a P value of 0.06; however, this finding did not meet the threshold for statistical significance. Current evidence regarding the association between GLP‑1 RA use and nonarteritic anterior ischemic optic neuropathy or vision loss remains conflicting. Our analysis did not provide conclusive evidence supporting such an association. Nonetheless, we strongly recommend continued vigilance and extended follow‑up observations, particularly for newer agents, such as tirzepatide, in patients receiving GLP‑1 RAs for the management of type 2 diabetes or weight management (both on- and off‑label).³