Background: Surgical resection for asymptomatic, localized, well-differentiated pancreatic neuroendocrine tumors (PNETs) is common yet controversial. Studies using overall survival as an endpoint have shown a benefit for resection; however, these results may be due to treatment selection bias. We assessed the impact of surgery on both overall (OS) and cancer-specific survival (CSS) for asymptomatic patients with stage I PNETs (AJCC 8th edition).
Methods: Using SEER data, we identified 709 patients from 2007 to 2015 with well- and moderately differentiated stage I PNETs. We performed Kaplan-Meier survival estimates and adjusted Cox regression for OS and CSS.
Results: Among 709 patients, 628 (88.6%) underwent surgery. There were 37 overall deaths and 11 cancer-specific deaths. All cancer-specific deaths occurred within 3 years of diagnosis. Five-year OS and CSS rates were 89% and 98%, respectively, for the population. Five-year OS rates were 56% in the non-surgical cohort versus 92% in the surgical cohort (log rank, p < 0.001). However, the 5-year CSS rates were similar; 94% in the non-surgical group and 98% in the surgical group (log rank, p = 0.207). On multivariable analysis, surgery predicted improved OS but not CSS.
Conclusion: Although OS is superior in surgically treated stage I PNETs, CSS is not improved, implying treatment selection bias towards surgery being performed in healthier patients. These data suggest that overall survival is a problematic endpoint for the study of asymptomatic, stage I PNETs. Surgery should be individualized for this cohort of patients as the primary cause of death is non-cancer related.
Keywords: Neuroendocrine; Pancreas; Surgery; Survival; Tumor.