Reader Response: Association of Prestroke Metformin Use, Stroke Severity, and Thrombolysis Outcome

When I first read the article by Westphal et al.,¹ I was excited about the possible positive effects of metformin on stroke. I note that a one-point difference in NIHSS is not clinically significant, a one-point change in modified Rankin Score is somewhat significant, and 13% vs 18% death rate is more clinically significant. However, further critical analysis challenges this interpretation. First, the authors conclude that “[t]his suggests a protective effect of MET [metformin].” Neurology reviewers and editors should admonish authors about inferring causality from inappropriately designed studies. In addition, as the authors report, “The number of confounders in this study was 19; together with the treatment group (MET+/MET−), 20 independent variables resulted.” This can only partially be compensated by their propensity score matching and multiple imputation. I would point out several other problems. For example, missing values included international normalized ratio (INR) on admission (29.7%). How could this be if everyone got IV thrombolysis? Per the article by Scheitz et al. cited in reference 8, “Laboratory measures obtained on hospital admission include glucose (mmol/L), blood cell counts (platelets, hemoglobin, and leucocytes), international normalized ratio and partial thromboplastin time/ Not all centers have to provide data on all variables but have given a commitment to add missing variables retrospectively, if considered relevant to answer a specific research question.” So, another question is whether everyone in this study had an INR before IV thrombolysis? Finally, perhaps the most important factor is the severity of the subject's type 2 diabetes. Although blood glucose and HbA1c are surrogates for the patient's clinical state, there is no assessment of other diabetes comedications. Patients usually move along a spectrum of diet control, oral hypoglycemics, and insulin. The composition of the MET group could be weighted toward insulin dependence, which would make them, on average, treated longer for their diabetes and having a more severe illness. That alone might account for their worse clinical outcomes, and not be a protective effect of metformin.