Dear Editor

We read Tanaka’s study [1] with great interest. In this study, the authors evaluated the preoperative computed tomography (CT) findings of thymoma patients to identify the CT features associated with tumor invasion, since preoperative investigations to detect invasion to neighboring organs are important for deciding on the most appropriate surgical approach. Based on their logical regression analysis, a lobulated tumor contour was significantly associated with both lung and pericardial invasion. However, we raise some concerns about the statistical method used in Tanaka’s study.

First, we noted that there were five variables in each of Tables 5 and 6 for the multivariate analyses performed to analyze the relationship between pathological invasion of the lung or pericardium and the clinicopathologic features, and that 18 patients with lung invasion and 11 patients with pericardial invasion as outcomes were recorded in these two Tables. This statistical method breaks a basic statistical rule that one variable at most analyzes 5–10 outcome events in multivariate logical regression analyses [2]. In other words, analyzing the risk factors for lung invasion or pericardial invasion needs 25–50 patients, instead of 18 or 11 patients. Five variables in 18 or 11 outcomes would result in overfitting the multivariate analysis model in Tanaka’s study. Usually, a comparison between patients who died and those who survived before the multivariate analysis would reduce the sample size of the variables. Then, the multivariate analysis could be analyzed after variable sizes were significantly reduced, also obeying the basic statistical rule that one variable at most analyzes 5–10 outcomes.

Despite these comments, we acknowledge Tanaka’s outstanding study, from which we learnt much.