Abstract

Both the European Neuroendocrine Tumor Society (ENETS) and the International Union for Cancer Control/American Joint Cancer Committee/World Health Organization (UICC/AJCC/WHO) have proposed TNM staging systems for pancreatic neuroendocrine neoplasms. This study aims to identify the most accurate and useful TNM system for pancreatic neuroendocrine neoplasms. The study included 1072 patients who had undergone previous surgery for their cancer and for which at least 2 years of follow-up from 1990 to 2007 was available. Data on 28 variables were collected, and the performance of the two TNM staging systems was compared by Cox regression analysis and multivariable analyses. All statistical tests were two-sided. Differences in distribution of sex and age were observed for the ENETS TNM staging system. At Cox regression analysis, only the ENETS TNM staging system perfectly allocated patients into four statistically significantly different and equally populated risk groups (with stage I as the reference; stage II hazard ratio [HR] of death = 16.23, 95% confidence interval [CI] = 2.14 to 123, P = .007; stage III HR of death = 51.81, 95% CI = 7.11 to 377, P < .001; and stage IV HR of death = 160, 95% CI = 22.30 to 1143, P < .001). However, the UICC/AJCC/WHO 2010 TNM staging system compressed the disease into three differently populated classes, with most patients in stage I, and with the patients being equally distributed into stages II–III (statistically similar) and IV (with stage I as the reference; stage II HR of death = 9.57, 95% CI = 4.62 to 19.88, P < .001; stage III HR of death = 9.32, 95% CI = 3.69 to 23.53, P = .94; and stage IV HR of death = 30.84, 95% CI = 15.62 to 60.87, P < .001). Multivariable modeling indicated curative surgery, TNM staging, and grading were effective predictors of death, and grading was the second most effective independent predictor of survival in the absence of staging information. Though both TNM staging systems were independent predictors of survival, the UICC/AJCC/WHO 2010 TNM stages showed very large 95% confidence intervals for each stage, indicating an inaccurate predictive ability. Our data suggest the ENETS TNM staging system is superior to the UICC/AJCC/WHO 2010 TNM staging system and supports its use in clinical practice.