LAUSR

To the Editor: We read with great interest the article entitled ‘‘Prognostic Factors of Survival After Neoadjuvant Treatment and Resection for Initially Unresectable Pancreatic Cancer.’’ We would like to congratulate the authors for their insights into the current practice patterns. This is the largest to date cohort and further reinforces the clear advantages and benefits of neoadjuvant therapy in the treatment of initially unresectable pancreatic ductal adenocarcinoma. However, we would like to make a few comments on the methodology and the interpretation of the findings of this study. First, although the authors were careful about commenting on the fact that resection margins (R) status had a large P-value in the multivariable analysis, they still concluded that ‘‘In patients undergoing resection after neoadjuvant therapy for initially unresectable pancreatic ductal adenocarcinoma, preoperative Carbohydrate antigen 19-9 (CA19-9) levels, lymph node involvement, metastasis category, and vascular involvement, but not the R status, were independent prognostic factors of overall survival.’’ This statement could easily be interpreted that R status does not matter as long as neoadjuvant treatment is used, or that R status does not predict prognosis in the neoadjuvant setting. In fact, throughout the manuscript and their sensitivity analysis, the estimated hazard ratio (HR) of R status in multivariable analysis was never provided. In univariable analysis, the HR was 1.48 for R1 1 mm and 1.70 R1 direct, respectively, which is quite high. Even though P was>0.05 after multivariable analysis, it would be useful to know the estimated HR. Specifically, could it be even higher with a wider confidence interval that includes the null? The epidemiology literature suggests that ‘‘0.05’’ is not some magical cutoff and that odds ratio, HR, and confidence interval should be presented together with the P-value. Another major concern is the inclusion of M1 patients in the analysis of the significance of R status. One might anticipate that when metastasis has occurred before treatment, local negative margin might not matter as much. Although the author did a sensitivity analysis excluding the M1 patients, the HR of R status was not shown. The inclusion of only univariate variables with a significance of <0.05 in the multivarate may not be appropriate. Factors with clinical significance suggested by previous studies might also need to be included, regardless of their P-value. For example, in this paper, P 1⁄4 0.057 for age, suggesting it may have an impact on prognosis, but it was not included in the multivariable analysis. Furthermore, we noticed that ‘‘arterial/venous infiltration’’ seems to be associated with R status. In R0 patients, 23.2% had infiltration, whereas the incidence was 47.5% and 52.5% in the R1 1 mm and the R1 direct group, respectively. Another example is the N stage. We might anticipate that in patients with vascular infiltration and more positive lymph nodes, it might be much harder to achieve R0 resection. Therefore, it could be very possible that both ‘‘infiltration’’ and ‘‘N stage’’ predicted R status, which in turn predicts prognosis. Intraoperatively, the surgeons could hardly alter ‘‘arterial/venous infiltration’’ or ‘‘N stage,’’ but an optimal R status is pursuable. The paper is clearly a prediction analysis rather than a causal inference, and a randomized trial to evaluate R status in any setting would be highly unethical. Retrospective data is all we have. Again, we would like to congratulate the authors for a nicely done study.