4541 Background: Anti-VEGFR2 therapy (ramucirumab/paclitaxel [RAM/TAX]) and ICI are approved as 2nd- and 3rd-line therapy (Tx), respectively, for pts with mGEA. We unexpectedly saw durable responses in 2 pts on RAM/TAX after progression on an ICI trial (KN-059; PMID 29674442). We performed a pilot to examine the clinical activity of ICI followed by RAM/TAX. Then we retrospectively compared the outcomes of pts who received this serial Tx to pts who received RAM/TAX without prior ICI. Methods: All pts with mGEA at Mayo Clinic who received RAM/TAX (2014-19) were included (N = 87). Outcomes were best objective response rate (ORR: complete [CR] or partial response) per RECIST1.1, progression-free survival (PFS), duration of response (DOR), and overall survival (OS). Chi square and multivariate (MV) logistic and Cox regression were used. Results: 15 consecutive pts with measurable mGEA received ICI immediately followed by RAM/TAX after irRECIST progression. Most pts (95%) did not respond to ICI. Yet on RAM/TAX, 100% (15/15) had tumor reduction (range -8% to -100%) with an ORR of 73% (11/15), including 3 CRs. In these pts (who received ICI followed by RAMTAX), PFS on RAMTAX was longer than on last chemotherapy before ICI (12.3 vs 3.0 m, P < .001). Outcomes on RAM/TAX in these pts were significantly better than in pts who received RAM/TAX alone (see Table). Associations were strengthened after adjusting for total lines of Tx, line of Tx of RAM/TAX, age, and ECOG PS. Exploratory analysis of paired tumor biopsies collected pre-ICI and on RAM/TAX in a small subset revealed that the frequency of intratumoral immunosuppressive FOXP3+ Tregs decreased on RAM/TAX, whereas the frequency of antitumor CD8+ T cells was preserved. Conclusions: RAM/TAX immediately preceded by ICI was associated with significantly higher OS, ORR, and DOR than RAM/TAX alone, suggesting ICI may enhance efficacy of subsequent anti-VEGFR/taxane therapy. This novel sequence of therapy will be tested prospectively in a new randomized phase 2 trial (NCT04069273). [Table: see text]

512 Background: The role of immune checkpoint inhibitors (ICIs) in the treatment of extrapulmonary neuroendocrine carcinoma (EP-NEC) has yet to be established. While objective responses have been observed, it is still unknown which patients are likely to derive benefit. We investigated the genomic profiles of patients who did and did not benefit from ICIs. Methods: Previously we reported the objective responses to ICI in a retrospective series of patients with EP-NEC. RECIST 1.1 criteria were used to categorize patients as achieving disease control (DC = CR, PR, or SD) vs progressive disease (PD). The EMR was reviewed to identify patients who had genomic panels performed, and results were extracted for analysis. Results: Of 31 patients eligible for RECIST assessment, 19 had genomic panels available (9 with DC: 4 SD, 5 PR vs 10 with PD). Of those with NEC histology specified, 9 were small cell, 1 combined large and small cell, 3 were large cell. All tumors were microsatellite-stable. All but one (with TMB = 25) of 16 tumors with TMB status available were < 10 m/MB. Of those with disease control, 67% had both TP53 and RB1 alterations, compared with only 10% in those with progressive disease; this was statistically significant ( p = 0.0198, Fisher exact). Of 7 tumors with TP53 + RB1 alterations, 4 were specified as small cell carcinoma. Half of those with PD showed alterations in β-catenin pathway genes CTNNB1 or APC, compared to only one of the DC group, but this did not reach significance ( p = 0.1409, Fisher exact). In an analysis of all Mayo patients (not just those treated with ICI) with NGS data available (Tempus and FoundationOne), concurrent TP53 + RB1 alteration was significantly more common in SCLC than in EP-NEC (Table). Conclusions: In this small series of patients with EP-NEC treated with ICIs, the SCLC-like genomic signature of concurrent TP53 + RB1 alterations was significantly more common in those with disease control than in those with progressive disease. An analysis of all patients with NGS data (not just on ICI) showed that the dual alteration was more common in SCLC, and SCLC also had more TMB-high tumors. This may explain why ICIs are more effective in SCLC than in EP-NEC. Further study is warranted to determine whether TP53 + RB1 mutations predict response to ICI in NEC.[Table: see text]

92 Background: Since the publication of the MOSAIC trial, stage III CC has been treated with a six-month (mo) regimen of FOLFOX (fluorouracil, leucovorin, and oxaliplatin) or CAPOX (capecitabine and oxaliplatin). Recently, the IDEA collaboration challenged this practice by demonstrating that the 3-year rate of disease-free survival (DFS) was non-inferior to 6mo of treatment (Rx) when given for low risk CC (83.1 vs. 83.3%) and resulted in significantly lower rates of grade 2 and higher neuropathy. In high risk (T4, N2) patients (pts) the DFS of 3mo of CAPOX was equivocal to 6mo (64.1 vs. 64.0%), while 3mo of FOLFOX was inferior to 6mo (61.5 vs. 64.7%). We hypothesized that trends in prescribing would favor shorter courses of Rx with a preference towards CAPOX given its efficacy across both high and low risk CC. Methods: We performed a retrospective analysis of stage III CC pts from 4 institutions. We evaluated prescribing patterns of 3mo or 6mo of Rx and CAPOX vs. FOLFOX over a period of 5 years from Jan 2016 to Jan 2021, a time period that traverses before and after the release of IDEA. Logistic and multinomial logistic regression models, with a linear time trend, were used to estimate the percentage of pts receiving CAPOX vs. FOLFOX and the combination of Rx and duration, respectively, while adjusting for baseline characteristics. The prescribing patterns in important subgroups were examined by incorporating the interaction term in the models. Results: A total of 366 pts met inclusion criteria. From 2016-2021, there was a significant increase per quarter in patients treated with CAPOX when compared to FOLFOX (OR 1.16 95% CI 1.11 – 1.21, p <.001). Prior to IDEA, 78.3% of pts received 6mo FOLFOX and 7.4% received 3mo CAPOX. Two years after IDEA, only 17.3% of pts were on 6mo FOLFOX compared to 67.5% of pts on 3mo CAPOX (Table). At present, high risk pts are more likely to receive 6mo FOLFOX (47.8%) than 3mo of FOLFOX (3.9%), 3mo CAPOX (25.8%), or 6mo CAPOX (22.4%). Low risk pts are more likely to receive 3mo of CAPOX (67.9%) than other Rx. Conclusions: Our findings suggest that since IDEA, physician practice has significantly changed in favor of CAPOX and shorter courses of Rx. The use of CAPOX has significantly increased overall, presumably due to its efficacy across all risk groups and relatively reduced toxicity.[Table: see text]

235 Background: Tumor archival tissue (TT) MA of PBM is challenging due to insufficient specimen acquired from fine needle aspirations. This is a barrier to analyzing specimens for targeted therapy (clinical trial enrollment). The correlation of cell free ctDNA and TMA has not been studied extensively. We hypothesize that somatic TMA correlates with ctDNA and will be a surrogate for enrollment to targeted trials. Methods: We retrospectively identified PBM patients evaluated in the Phase I Program and analyzed ctDNA and TT. ctDNA was extracted from plasma, genomic alterations were analyzed by parallel sequencing of amplified target genes (73 genes) using Illumina Hi Seq (Guardant360). Direct sequence analysis was performed on genomic DNA isolated from FFPE TT using the Illumina MiSeq platform, 592-whole gene targets (Caris MI/X). The primary objective was to correlate the mutation concordance rate (mCR, common mutations/all mutations) between ctDNA and TT. We utilized Stata 15 for descriptive analysis, Spearman correlation coefficient, and Wilcoxon Sign Rank test for statistical analysis. Results: Between 03/2016-08/2017, 28 patients (M:F 1:1) were identified, 89.2% had pancreatic adenocarcinoma, 54% were treatment naïve and age ranged from 52-83 years old. The mCR for ctDNA was 31% (21/67) and the mCR for TT was 38% (21/55). These mCRs were highly correlative with a Spearman’s Rho = 0.89, (p < 0.0001). The mCR in ctDNA was compared to mCR in TT using Wilcoxon Sign Rank Test and was found to not be statistically different (p = 0.69). Adequate TT for MA occurred in 64 % (18/28) of patients. Percentage of DNA alterations found was 89.3% (25/28) in ctDNA and 94.4% (17/18) in TT. Moreover, 61% (11/18) of patients had at least 1 shared mutation. Percentages of driver mutations identified were 92% (23/25) in ctDNA and 100% (17/17) in TT. The most common driver mutations in ctDNA were TP53, KRAS, SMAD4, and NF1. The most common driver mutations in TT were KRAS, TP53, CDKN2A, and SMAD4. Conclusions: There was a statistically significant correlation between mCR of ctDNA and TMs in PBM (p < 0.0001).

e16752 Background: Undifferentiated osteoclast-like giant cell carcinoma of the pancreas is an aggressive malignancy only described by a few case reports in the literature. In this study, we sought to better characterize this entity by examining patients seen at Mayo Clinic. Methods: This study identified patients with osteoclast-like giant cell carcinoma of the pancreas using Mayo Clinic databases (MN, AZ, FL) from the year 2000 to present. Patient demographics, genetic data, and treatment modalities were reviewed. Kaplan-Meier analysis was used to evaluate median overall survival (mOS) for the cohort as well as mOS and mPFS for treatment subgroups. Results: 15 patients were identified (9 female, 6 male). Median age at diagnosis was 59 years and mOS for all patients was 11.0 mos (95% CI 6.2-15.7 mos). 3 patients (20%) had metastatic disease at diagnosis with the liver being the most common site (n = 3). Metastatic disease was associated with significantly shorter OS (3.5 vs. 14.1 mos, p = 0.005; HR 7.98 [95% CI 1.43-44.4]) compared to locoregional disease (LRD). 4 patients underwent genetic testing. The most common mutation was CDKN2A (n = 3), followed by TP53 (n = 2) and KRAS (n = 2). 13/15 patients had detailed follow-up information. 6/7 patients undergoing chemotherapy received a gemcitabine-based regimen as first line: with capecitabine (n = 2), with nab-paclitaxel (n = 2), or as monotherapy (n = 2). 1 patient received FOLFIRINOX. In patients with LRD and adequate follow-up (n = 11), 8/11 underwent surgical resection and had longer OS compared to those without resection (17.0 vs. 8.4 mos, p = 0.09). No surgical patients received neoadjuvant chemotherapy. 5/8 received adjuvant chemotherapy, 2 did not undergo chemotherapy, and 1 was lost to follow-up after surgery. PFS from time of surgery was 15.3mos (95% CI 5.0-25.6mos). 6 patients (40%) were alive at time of analysis; all 6 underwent surgical resection. Conclusions: Osteoclast-like giant cell carcinoma of the pancreas is a rare malignancy with a poor prognosis, even when diagnosed at an early stage. OS for patients of all stages in this study was less than 1 year suggesting prognosis may be even worse than that of pancreatic adenocarcinoma. The optimal therapy remains unknown but most patients received similar chemotherapy as in adenocarcinoma. Patients with LRD amenable to surgical resection experienced a PFS over 1 year from surgery, possibly associated with OS benefit, however overall prognosis is poor. Further study is warranted on a larger scale to better understand disease course and treatment options.