Comment on “Impact of tumor site on the prognosis of small bowel adenocarcinoma”

We were interested to read the article “Impact of tumor site on the prognosis of small bowel adenocarcinoma” by Falcone and colleagues1 in Tumori Journal. The authors aimed to evaluate clinical outcome of patients diagnosed with small bowel adenocarcinoma. They drew the conclusion that disease stage and primary tumor site were independent predictors of overall survival. However, some statistical problems need to be resolved before drawing that conclusion. The author mentioned in the article that multivariate Cox regression was used to test independence of significant factors in univariate analysis. All the variables (including nonsignificance factors) in the univariate analysis were imported into the multivariate regression model, resulting in an inevitable problem with overparameterization.2,3 As we know, age is a factor closely related to tumor incidence and progression. In this multivariate Cox regression, age was presented as nonsignificant, maybe caused by overparameterization. Therefore, we suggest excluding the sex factor from multivariate Cox regression. Another important question, especially in studies with small sample size, is the validity of the prediction model. This seems to be the major limitation of the Rosa Falcone et al. study, and it can be solved with statistical methods such as cross-validation and bootstrapping.4 Otherwise, the authors could draw an incorrect conclusion.