We read with interest the letter by De Giorgio and associates1 on cancer with or without diabetes mellitus (DM), published in the ARCHIVES. Among 974 patients with lung cancer admitted to our division during a 25-year period, 90 patients were diagnosed as having DM. Using the generalized Wilcoxon test, a statistically significant difference in survival was not observed between patients with DM and patients without DM (P = .24). In multivariate regression analysis with the Cox proportional hazards model,2 sex (P<.001), stage (P<.001), and performance status (P<.001) were significantly related to survival; however, coexisting DM was not a significant prognostic factor for survival (P = .94). Therefore, we disagree with the conclusion by De Giorgio and associates that "diabetes is associated with longer survival rates in patients with malignant tumors," including lung cancer. We would appreciate hearing about the diagnostic criteria of DM in patients with lung cancer. We believe that multivariate regression analysis is one of the suitable methods for determining the prognostic factor. We also question the best statistical method for determination of the prognostic factor.
Satoh H, Ishikawa H, Kurishima K, Ohtsuka M, Sekizawa K. Diabetes Is Not Associated With Longer Survival in Patients With Lung Cancer. Arch Intern Med. 2001;161(3):485. doi: