Chemotherapy for Well-Differentiated Pancreatic Neuroendocrine Tumours with a Ki-67 Index ≥10%: Is There a More Effective Antitumour Regimen? A Retrospective Multicentre Study of the French Group of Endocrine Tumours (GTE)

Background: The best chemotherapy regimen for well-differentiated pancreatic neuroendocrine tumours (pNETs) with a Ki-67 index ≥10% is still debated. We evaluated the antitumour efficacy of various first-line chemotherapy regimens (streptozocin based, platinum based, or dacarbazine/temozolomide based) in this situation. Methods: In this retrospective multicentre study of the French Group of Endocrine Tumours (GTE), we recruited consecutive patients with advanced well-differentiated pNETs and a Ki-67 index ≥10% receiving chemotherapy between 2000 and 2012. The primary endpoint was progression-free survival (PFS) according to RECIST. Results: Seventy-four patients (42 men, median age 55.5 years) were enrolled from 10 centres. Fifty-one patients (69%) had grade 2 NET and 61 (82%) were stage IV. Median overall survival was 36.3 months. Forty-four patients (59%) received streptozocin-based, 18 (24%) platinum-based, and 12 (16%) dacarbazine/temozolomide-based chemotherapy regimens. These 3 groups were similar regarding age, functioning tumours, grade, the number of metastatic sites, and surgery for primary tumours, but not regarding surgery for metastases and time since diagnosis. Grade 3 NET (HR 2.15, 95% CI: 1.18-3.92, p = 0.012) and age above 55 years (HR 1.84, 95% CI: 1.06-3.18, p = 0.030) were associated with shorter median PFS in the multivariate analyses. Compared to streptozocin-based chemotherapy, no difference was found in terms of PFS for the platinum-based or for the dacarbazine/temozolomide-based chemotherapy regimen: median PFS was 7.2, 7.5, and 7.2 months, respectively (p = 0.51). Conclusions: Patients with intermediate or highly proliferative well-differentiated pNETs may benefit from 1 of the 3 chemotherapy regimens. Increased age and grade 3 were associated with shorter median PFS. Randomised studies searching for response predictors and the best efficacy-tolerance ratio are required to personalise the strategy.