Dichotomizing continuous prognostic factors can cause paradoxical results in survival models.

摘要

In their interesting contribution, Murakami and colleagues1 investigated the importance of lymph node metas-tases for overall survival in patients with pancreatic ductal carcinomas. They compared the number of metastatic lymph nodes (Nmeta) and the ratio of metastatic nodes with total number of examined lymph nodes (LNR) after tumor resection and concluded that "The number of metastatic nodes, but not LNR, is one of the most powerful prognostic factors after resection of pancreatic carcinoma." LNR is considered to be an important and robust prognostic variable for patients with various kinds of cancer.2'3 When a study on prognostic factors shows surprising results, however, it should be checked for methodological pitfalls.2'5 We noticed that both Nmeta and LNR entered the calculations in a dichotomized form.