We would like to thank Hung et al. (1) for their interest in our recent publication in Diabetes Care (2). Our analysis showed that in patients with type 2 diabetes (T2D) without established cardiovascular disease, glucagon-like peptide 1 receptor agonist (GLP-1RA) users with a >251-day supply had a significantly lower risk of ischemic stroke hospitalization than non–GLP-1RA users (adjusted hazard ratio [HR] 0.28 [95% CI 0.11–0.71]) (2).
We would like to further clarify a few points in response to the submitted letter.
First, the question of whether it is the GLP-1RA or the effect of the GLP-1RA on glucose that determines the occurrence of stroke was difficult to determine in our study. The current study database did not examine the ability to control glucose. Glycated hemoglobin A1c (A1C) data were not available in the database, and this is one of our study’s limitations. However, we were able to calculate duration of GLP-1RA use and cumulative dose. Higher cumulative dose and longer duration of use of GLP-1RA were associated with lower risk of hospitalization for ischemic stroke. The cost of GLP-1RA was reimbursed in patients with T2D who have received the maximum tolerated dose of metformin and still do not have adequate blood glucose control. Higher cumulative and longer duration of use of GLP-1RA are likely to result in better glycemic control (3). A recent meta-regression analysis demonstrating a relationship in the degree of glucose lowering with GLP-1RA and cardiovascular benefit provides some support for a direct glucose effect (4). The estimated effects were 10–41% for A1C on major adverse cardiovascular event outcomes (5).
To further clarify the methodology used, for the non–GLP-1RA users, the period from the study entry to the index date was matched to the time of initiation of GLP-1RA users, so the potential immortal time bias was less likely to happen.
We agree there are still important residual confounders that may limit the interpretation of the study results. As much as possible, we identified potential residual confounders by using ICD codes. As Hung et al. (1) suggested, residual confounders, such as the duration of diabetes, the severity of the kidney disease, and the patient’s income and education level, may limit the interpretation of the study results. We further identify the duration of diabetes and the patient’s income. The duration of diabetes at index date was 9.05 and 6.88 years for the GLP-1RA group and non–GLP-1RA group, respectively. Regarding average personal income per month, the majority were at the level of NTD 40,000 (USD 1,440) and below, (70.6% and 77.5% in the GLP-1RA group and non–GLP-1RA group, respectively). GLP-1RA users with a >251-day supply during the study period had a significantly lower risk of ischemic stroke hospitalization, and we determined adjusted HRs of ischemic stroke adjusted for sex, age, each comorbidity, each concomitant medication, duration of diabetes, and monthly income (adjusted HR 0.25 [95% CI 0.10–0.63], P = 0.0031). Despite controlling for a large number of potential confounders, we still could not exclude the possibility of residual confounding, such as other psychosocial/lifestyle factors, which is an inherent limitation of most observational studies.
See accompanying article, p. e181.
Article Information
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
