To the Editor:We very much appreciated the letter submitted by Ansari et al 1 in response to our recently published work regarding 30day outcomes after carotid endarterectomy (CEA) and carotid artery stenting (CAS) using the American College of Surgeons National Surgical Quality Improvement Program database. 2 In brief, motivated by the observation that real-world outcomes do not necessarily parallel findings from prospective, randomized trials, we conducted a retrospective analysis of both procedures in a large, national cohort using a propensity matching method. 3 We found that, in this data set, CAS was associated with increased odds of stroke relative to CEA. We did not find significant differences between rates of myocardial infarction and death between the 2 procedures.Propensity matching is a statistical method that was developed to reduce the bias inherent in retrospective analyses by balancing potentially confounding factors and, therefore, imperfectly approximating the design of a randomized trial. All statistical methods, no matter how well considered they may be, have significant limitations. At a basic level, the ability of propensity matching to balance confounding factors is limited by the number of relevant factors that is available in a given data set. 4,5 The use of propensity matching may also result in a significantly diminished available sample size or poor covariate balance despite optimization. 5 Furthermore, binary classifications of covariates (such as presence or absence of heart failure) fail to account for differences within subtypes of disease severity or duration, both of which likely have large and likely incompletely measured effects on clinical outcomes. 6 As is astutely pointed out by Ansari et al, conclusions drawn from any retrospective analyses, regardless of the used statistical method, should be considered in the context of their limitations and of previously available data and generated hypotheses.Ansari et al 1 used propensity matching to compare a large sample of patients undergoing CAS and CEA in the vascular procedure-targeted National Surgical Quality Improvement Program database, which contains more specific information regarding patient-level anatomic risk factors and surgical history. In their primary analysis, the authors do not find a statistically significant difference in rates of stroke or transient ischemic attack between the 2 procedures. In a secondary analysis of patients with high-risk features for complication, the authors also found that patients undergoing CAS had significantly reduced the odds of myocardial infarction relative to those undergoing CEA.