Response to the letter titled “Analysis of semaglutide's effect on Alzheimer's disease risk”

Abstract

To the Editor,

We appreciate the opportunity to respond to the comments by Hsu et al. regarding our study.¹ In their comments, the authors discussed several limitations in our study. We used a target trial emulation framework to examine the associations of semaglutide with first-time diagnosis of Alzheimer's disease (AD) and related medication prescriptions in patients with type 2 diabetes (T2D) who had no prior diagnosis of AD. We acknowledged that our study has the limitations inherent in observational studies using patient electronic health records (EHRs) including uncontrolled and unmeasured confounding factors and bias that preclude causal inferences, which will need to be examined in randomized clinical trials. The following are our responses to the specific comments raised by the authors. (1) In this study, semaglutide was compared with other anti-diabetic medications. Individuals were assigned to this treatment strategy compatible with their first prescription and assumed randomization by propensity-score matching for baseline covariates. Therefore, the dosage and duration of use of semaglutide were not accounted as confounders. We acknowledge that future analyses in comparing different dose or durations of semaglutide use among patients prescribed semaglutide could provide additional insights into dose–response relationships and temporal effects of semaglutide on AD incidence. (2) The groups were matched for lifestyle factors including tobacco use, alcohol drinking, lack of physical exercise, inappropriate diet and eating habits, and other problems such as antisocial behavior and sleep deprivation. (3) A1c ≥8.5% was an inclusion criterion that was applied to both the semaglutide and comparison groups. Semaglutide is known to be better in A1c control than other anti-diabetic medications, which could be one of the mechanisms underlying our observed reduced AD incidence associated with semaglutide, which is not a confounding factor. (3) In our study, groups were matched for individual comorbidities and were well balanced. We believe that matching for individual comorbidities could be better than matching for one single aggregated comorbidity index. For example, if two groups were well balanced for each comorbidity, they would automatically be balanced based on an aggregated comorbidity index. In summary, we acknowledged many limitations inherent in EHR-based observational studies, emphasized that no causal can be drawn in our study, and called for future randomized clinical trials to assess the causal relationships between semaglutide and AD prevention.