BACKGROUND:To date, single-agent immune checkpoint inhibitor (CPI) therapy has proven to be ineffective against biomarkerunselected extrapulmonary poorly differentiated neuroendocrine carcinomas (EP-PDNECs). The efficacy of CPI in combination with chemotherapy remains under investigation. METHODS: Patients with advanced, progressive EP-PDNECs were enrolled in a two-part study of pembrolizumab-based therapy. In Part A, patients received pembrolizumab alone. In Part B, patients received pembrolizumab plus chemotherapy. Primary endpoint: objective response rate (ORR). Secondary endpoints: safety, progression-free survival (PFS) and overall survival (OS). Tumours were profiled for programmed death-ligand 1 expression, microsatellite-high/mismatch repair deficient status, mutational burden (TMB), genomic correlates. Tumour growth rate was evaluated. RESULTS: Part A (N = 14): ORR (pembrolizumab alone) 7% (95% CI, 0.2-33.9%), median PFS 1.8 months (95% CI, 1.7-21.4), median OS 7.8 months (95% CI, 3.1-not reached); 14% of patients (N = 2) had grade 3/4 treatment-related adverse events (TRAEs). Part B (N = 22): ORR (pembrolizumab plus chemotherapy) 5% (95% CI, 0-22.8%), median PFS 2.0 months (95% CI, 1.9-3.4), median OS 4.8 months (95% CI, 4.1-8.2); 45% of patients (N = 10) had grade 3/4 TRAEs. The two patients with objective response had high-TMB tumours. DISCUSSION: Treatment with pembrolizumab alone and pembrolizumab plus chemotherapy was ineffective in advanced, progressive EP-PDNECs. CLINICAL TRIAL REGISTRATION: ClinicalTrials.gov NCT03136055.

PURPOSE: Somatostatin analogs octreotide long-acting release (octLAR) and lanreotide are equally acceptable in National Comprehensive Cancer Network guidelines for neuroendocrine tumors (NETs). Lanreotide is more expensive and given by deep subcutaneous injection, whereas octLAR is given intramuscularly. We evaluated patient preference between these agents in terms of injection site pain. MATERIALS AND METHODS: Randomized, single-blinded study. Patients with NETs received injections every 4 weeks. Arm 1: octLAR × 3, then lanreotide × 3; arm 2: reverse order. Self-reported injection site pain scores (range, 0-10) were obtained after each of the first three injections. Primary end point was comparison of mean pain scores over the first three injections. Secondary end points included patient-reported preference. RESULTS: Fifty-one patients enrolled (26 in arm 1 and 25 arm 2), all evaluable for primary end point. No significant difference was identified in the mean pain score over the first three injections (2.4 ± 1.9 v 1.9 ± 1.5, P = .5). Thirty-four of 51 (67%) patients (15 in arm 1 and 19 in arm 2) completed post-therapy questionnaires and were evaluable for secondary end points. Seven patients (47%) in arm 1 and eight patients (42%) in arm 2 indicated no drug preference at the end of treatment. In the other 19 patients, more patients indicated mild or strong preference for octLAR over lanreotide. CONCLUSION: We found minimal pain with octLAR and lanreotide and no significant pain score differences between the two. Patients indicating a drug preference trended toward favoring octLAR.

No abstract

653 Background: PanNENs represent 1-2% of all pancreatic neoplasms. The genomic landscape derived from PanNEN tumor tissue has been described previously. There are little data detailing the frequency of genetic alterations identified in cfDNA in an advanced PanNEN population, the plasma-tissue concordance of detected alterations, and the clinical utility of cfDNA. Methods: Patients (pts) with metastatic PanNENs underwent collection of cfDNA for NGS using the MSK-IMPACT 505 gene assay between March 2017 and April 2020. Matched tissue based NGS with the FDA authorized MSK-IMPACT gene assay was completed when tumor tissue was available. For some pts, plasma and tumor tissue were sequenced at multiple time points. Clinical actionability of sequence variants was annotated by OncoKB. Clinicopathologic characteristics were extracted, and data are herein reported. Results: 25 unique pts with metastatic PanNENs had 32 plasma samples analyzed. The majority had well differentiated (22/25; 88%), intermediate grade disease (13/25; 52%). 6 (24%) pts had well differentiated high grade disease and 3 (12%) had poorly differentiated neuroendocrine carcinomas. After extraction, median cfDNA yield per sample was 23.98ng (range: 3.2 to 500.1). Mutations were detected in 21(66%) of 32 samples (10 pre systemic therapy, 10 at progression, 12 post response to therapy or while stable on therapy). The most frequently mutated genes occurring in >10% of patients were DAXX (28%), TSC2 (24%), MEN1 (24%), ARID1B (20%), ARID1A (12%) and ATRX (12%). 23 (92%) pts underwent tumor tissue sequencing with MSK-IMPACT with a median time of 6.9 (range: 0.5-33.4) months between tissue collection and time of plasma analysis. NGS of cfDNA identified the most common mutations observed in tumor tissue for: DAXX (5/6; 83%), TSC2 (3/6; 50%), MEN1 (5/12; 42%), ARID1A (3/5; 60%) and ATRX (3/6; 50%). In 21/23 (91%) paired samples, additional mutations not seen in tissue were detected in plasma and included TSC2, TP53, EGFR, VHL, and BRCA2. Potentially actionable mutations were identified in sequenced cfDNA in 8/25 (32%) patients including 4 TSC2 mutations (level 3b), 1 ATM mutation (level 3b), 2 ARID1A mutations (level 4) and 1 KRAS mutation (level 4). One patient who was treated with larotrectinib for an ETV6:NTRK3 fusion detected on tumor sequencing ultimately developed resistance with a NRTK3 G623R alteration identified through sequencing of cfDNA at radiographic disease progression. Conclusions: NGS of cfDNA in metastatic PanNENs, across the spectrum of WHO-defined tumor grade/differentiation, revealed tumor-associated genetic alterations in 66% of plasma samples. Clonal evolution, actionable alterations, and resistance mechanisms can be detected through circulating cfDNA genotyping and may serve as a powerful tool to better understand disease biology of a disease that often changes over time and through therapy.