We thank Dr. Dasheiff for his interest in our article.1 We agree that based on observational registry-based data, it is not straightforward to claim causality and impossible to account for all, including unknown, confounders. However, a thorough analysis based on propensity score–matched (PSM) sets of patients is mimicking a randomized experiment, and because the data set was very large, we were able to account for many known potential confounders simultaneously. It is common in real-life databases that there are some observations missing. In our study, we tackled this with a complex approach, in which we performed imputation and PSM repeatedly in multiple iterative steps. The observation that international normalized ratio (INR) values were incomplete is explained by the fact that standard operating procedures for intravenous thrombolysis (IVT) differed across centers. Some centers do not determine INR before IVT in patients with neither anticoagulation nor liver or coagulation disorder. We agree that duration of diabetes, nonpharmacological and other cotreatments may matter. This limitation was mentioned in the study. By accounting for the age difference between metformin (MET)-/MET+ groups with PSM, we intended to address this aspect. Our data may serve as an argument for patients and GPs to continue MET in patients with diabetes and vascular risk factors and spur planning of further trials that may prove the suggested protective effect of MET.