Abstract: We read with interest the article on clinical outcomes of thoracic aortic surgery in hemodialysis patients by Hibino and colleagues [1Hibino M. Oshima H. Narita Y. et al.Early and late outcomes of thoracic aortic surgery in hemodialysis patients.Ann Thorac Surg. 2016; 102: 1282-1288Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar]. However, we have concerns regarding the methodology used in this study. First, the propensity score is defined as the conditional probability of assignment to a particular treatment [2Rosenbaum P.R. Rubin D.B. The central role of the propensity score in observational studies for causal effects.Biometrika. 1983; 70: 41-55Crossref Scopus (17156) Google Scholar]. The authors used a propensity score-matching analysis to compare the outcomes after thoracic aortic surgery in patients with and without hemodialysis. Because preoperative hemodialysis is not a treatment choice but a preoperative characteristic that is fixed and unchangeable, it seems inappropriate to use a propensity score to compare patients with and without hemodialysis. For example, if the propensity score for a patient in the hemodialysis group is estimated to be 0.50, is it true that the patient will receive preoperative hemodialysis with a 50% probability? Propensity score analysis is not a magic bullet that allows for a comparison of “apples and oranges.” Second, their propensity score-matching analysis might have been limited by poor statistical power because of the considerable reduction in sample size after the propensity score matching. Although they reported that the difference in hospital mortality (14.3% in hemodialysis patients versus 0.0% in control patients) was not statistically significant in their propensity score-matched groups, we could not determine whether this statistical result was due to adequate risk adjustment or the large reduction in sample size. To evaluate the prognostic impact of preoperative hemodialysis in patients undergoing thoracic aortic surgery, it might be better to include preoperative hemodialysis as a risk factor in the multivariate logistic regression analysis. We agree with Hibino et al. [1Hibino M. Oshima H. Narita Y. et al.Early and late outcomes of thoracic aortic surgery in hemodialysis patients.Ann Thorac Surg. 2016; 102: 1282-1288Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar] regarding the efficacy of thoracic aortic surgery for hemodialysis patients. However, further investigation is needed to evaluate the short-term prognostic impact of preoperative hemodialysis. If the authors would present the odds ratio of preoperative hemodialysis for early mortality in a multivariate logistic regression analysis, the result would be informative and interesting to the readers of The Annals of Thoracic Surgery. Early and Late Outcomes of Thoracic Aortic Surgery in Hemodialysis PatientsThe Annals of Thoracic SurgeryVol. 102Issue 4PreviewThe number of cardiovascular surgeries among hemodialysis patients is increasing according to the growing population of hemodialysis patients; however, the clinical outcome has not yet been clarified, especially in thoracic aortic surgery. The purpose of this study was to assess the early and late outcomes of thoracic aortic surgery in hemodialysis patients. Full-Text PDF ReplyThe Annals of Thoracic SurgeryVol. 104Issue 3PreviewWe appreciate the comments from Ueki and colleagues [1] regarding our article [2] reporting on the clinical outcomes of thoracic aortic surgery in hemodialysis patients. The comment, which used the metaphor “apples and oranges”, suggested that propensity score matching is appropriate for adjusting only the probability for the treatment choice, not unchangeable patient characteristics. Propensity score matching was proposed by Rosenbaum and Rubin [3]. These days, this type of matching is widely used to adjust the covariates in observational studies. Full-Text PDF