Overall survival (OS) is considered as the standard measure of outcome in oncology. However, considering that resectable pancreatic neuroendocrine neoplasms (Pan-NENs) usually have a long OS, the feasibility of prospective studies is questionable due to a long follow-up period needed. The primary endpoint was to validate the use of disease-free survival (DFS) as a surrogate measure of OS. The secondary endpoint was to calculate the gain in sample size using DFS instead of OS in hypothetical prospective studies with two parallel groups. A systematic review of studies reporting both OS and DFS in resected Pan-NENs was carried out. Multivariate linear regression analysis was used to evaluate if DFS predicts the OS in patients undergoing radical resection. Monte Carlo simulation was performed to estimate the gain in sample size, supposing the use of DFS instead of OS, to evaluate a hypothetical adjuvant treatment after surgery in a randomized trial. Six studies reporting data about seven cohorts of resected Pan-NENs were included, for a total of 1088 patients. The median OS and DFS were 144 (27–134) and 122 (50–267) months, respectively. There was a significant correlation between DFS and OS (R ² = 0.988; P = 0.035). Monte Carlo simulations showed that the number of patients needed to demonstrate a significant reduction of probability of a ‘target event’ in a hypothetical two-arm group exploring the hypothetical role of adjuvant therapy was reduced using DFS instead OS. This finding supports the legitimacy of using DFS as an acceptable surrogate for OS in surgical clinical trials.

Spartalizumab, a humanized anti-programmed death protein 1 (PD-1) MAB, was evaluated in patients with well-differentiated metastatic grade 1/2 neuroendocrine tumors (NET) and poorly differentiated gastroenteropancreatic neuroendocrine carcinomas (GEP-NEC). In this phase II, multicenter, single-arm study, patients received spartalizumab 400 mg every 4 weeks until confirmed disease progression or unacceptable toxicity. The primary endpoint was confirmed overall response rate (ORR) according to blinded independent review committee using response evaluation criteria in solid tumors 1.1. The study enrolled 95 patients in the NET group (30, 32 and 33 in the thoracic, gastrointestinal, and pancreatic cohorts, respectively), and 21 patients in the GEP-NEC group. The ORR was 7.4% (95% CI: 3.0, 14.6) in the NET group (thoracic, 16.7%; gastrointestinal, 3.1%; pancreatic, 3.0%), which was below the predefined success criterion of ≥10%, and 4.8% (95% CI: 0.1, 23.8) in the GEP-NEC group. In the NET and GEP-NEC groups, the 12-month progression-free survival was 19.5 and 0%, respectively, and the 12-month overall survival was 73.5 and 19.1%, respectively. The ORR was higher in patients with ≥1% PD-L1 expression in immune/tumor cells or ≥1% CD8+ cells at baseline. The most common adverse events considered as spartalizumab-related included fatigue (29.5%) and nausea (10.5%) in the NET group, and increased aspartate and alanine aminotransferases (each 14.3%) in the GEP-NEC group. The efficacy of spartalizumab was limited in this heterogeneous and heavily pre-treated population; however, the results in the thoracic cohort are encouraging and warrants further investigation. Adverse events were manageable and consistent with previous experience.

No validated prognostic tool is available for predicting overall survival (OS) of patients with well-differentiated neuroendocrine tumors (WDNETs). This study, conducted in three independent cohorts of patients from five different European countries, aimed to develop and validate a classification prognostic score for OS in patients with stage IV WDNETs. We retrospectively collected data on 1387 patients: (i) patients treated at the Istituto Nazionale Tumori (Milan, Italy; n = 515); (ii) European cohort of rare NET patients included in the European RARECAREnet database (n = 457); (iii) Italian multicentric cohort of pancreatic NET (pNETs) patients treated at 24 Italian institutions (n = 415). The score was developed using data from patients included in cohort (i) (training set); external validation was performed by applying the score to the data of the two independent cohorts (ii) and (iii) evaluating both calibration and discriminative ability (Harrell C statistic). We used data on age, primary tumor site, metastasis (synchronous vs metachronous), Ki-67, functional status and primary surgery to build the score, which was developed for classifying patients into three groups with differential 10-year OS: (I) favorable risk group: 10-year OS ≥70%; (II) intermediate risk group: 30% ≤ 10-year OS < 70%; (III) poor risk group: 10-year OS <30%. The Harrell C statistic was 0.661 in the training set, and 0.626 and 0.601 in the RARECAREnet and Italian multicentric validation sets, respectively. In conclusion, based on the analysis of three ‘field-practice’ cohorts collected in different settings, we defined and validated a prognostic score to classify patients into three groups with different long-term prognoses.