Search

29 Background: Brain metastases (BM) are rare in colorectal cancer (CRC) with 1-3% of patients developing BM over their lifetime. There are no screening recommendations for BM in patients with CRC and most are diagnosed after a patient develops symptoms. Identifying the clinical characteristics for patients with CRC and BM may aid clinicians to consider BM early in symptom onset and potentially screen appropriate patients. Methods: We performed a retrospective chart review of patients >18 years diagnosed with CRC and BM at Yale New Haven Hospital from 1/2012-9/2021. Clinical characteristics were analyzed by descriptive statistics, time intervals were compared using the Mann-Whitney test, and survival analysis was done using the Kaplan-Meier method. Results: We identified 93 patients with CRC and BM. Median age at diagnosis was 56 years, 55% were male, 75% of patients had a left sided primary tumor and 47% were rectal primary. At initial diagnosis, 38% and 39% were stage III and IV respectively. RAS/RAF mutations were present in 45% of patients and all were mismatch repair proficient. The median time from CRC diagnosis to BM was 3.81 years (95% CI 3.09, 4.41) and median time from metastatic CRC to BM was 2.21 years (95% CI 1.48, 2.75). Median overall survival after developing BM was 0.49 years (95% CI 0.24, 0.70). Patients with RAS/RAF mutations had a decreased median time from CRC diagnosis to BM of 3.33 years (95% CI 2.69, 4.37) compared to RAS/RAF wildtype at 5.09 years (95% CI 4.44, 7.97) (p 0.0028). Median time to BM from CRC diagnosis was also decreased in females (p 0.01). The location of primary tumor did not significantly decrease time to BM. Prior to BM diagnosis, 76% of patients had lung metastasis. 84% of patients received treatment for BM while 14% had no BM treatment and 2% were unknown. For treatment 50% had Gamma Knife Radiosurgery, 22% underwent surgical excision +/- adjuvant radiation, and 12% underwent whole brain radiotherapy. After BM treatment, 43% of patients received additional chemotherapy. Conclusions: Although BM in CRC is uncommon, early diagnosis in those at risk is critical given the urgent nature of BM. Patients with CRC that develop BM frequently have tumors that originate in the left side of the colon and preexisting lung metastasis. In patients with CRC and BM, the majority develop BM > 3 years after initial CRC diagnosis, and BM onset is significantly shorter in females and patients with tumors harboring a RAS/RAF mutation. After BM diagnosis and treatment, less than half of patients are able to receive subsequent therapy, and median survival is < 6 months. [Table: see text]

115 Background: BI 905711 is a tetravalent bispecific antibody that cross-links TRAILR2 with CDH17. This cross-linking drives CDH17-dependent TRAILR2 oligomerization, leading to caspase activation and eventual apoptosis, inhibiting tumor growth in preclinical models of GI cancer. Methods: This Phase Ia/b study of BI 905711 in patients (pts) with advanced GI cancers (NCT04137289) aimed to determine the maximum tolerated dose (MTD) based on the proportion of pts with dose-limiting toxicities (DLT) and explore preliminary antitumor activity. In Phase Ia, pts received BI 905711 every 14 days. One pt with colorectal cancer (CRC) was enrolled at each of the 2 lowest dose levels (0.02/0.06 mg/kg) and 4 pts with CRC were enrolled at each subsequent level (0.2/0.6/1.2/2.4/3.6/4.8 mg/kg). Up to 4 pts with non-CRC GI cancers were included at the dose level below the CRC cohort. Dose escalation was guided by a Bayesian logistic regression model. Results: As of 01 August 2022, 48 pts (CRC: n = 26; non-CRC: n = 22, including 13 with pancreatic ductal adenocarcinoma [PDAC]) had received BI 905711 (dose range 0.02–4.8 mg/kg); pts had a median age of 61 years (range 27–78) and had received a median of 3 (range 1–11) prior lines of treatment. No DLTs were observed and the MTD was not reached. 41 pts had AEs (grade [G] 3–5: 14 pts; serious: 13 pts). 17 pts had treatment-related AEs (TRAEs); 4 TRAEs were G3: AST increased (2 pts), fatigue and ALT increased (1 pt each). 1 pt had serious TRAEs: G2 decreased appetite and G3 fatigue. 2 pts had G1/2 infusion-related reactions that resolved and did not prevent resumption of treatment. PD biomarker modulation on the level of plasma caspase-3/7 activity was most common at 0.6 mg/kg (4/7 pts) or 1.2 mg/kg (2/6 pts). 13 pts achieved stable disease (SD) and 8 pts were progression-free for ≥4 months (PFS4). In pts with CRC, 6 pts achieved SD and 3 had PFS4. Among non-CRC pts, 7 pts achieved SD (PDAC: n = 6) and 5 had PFS4 (PDAC: n = 4). Median duration of treatment was 30.5 days (range 15–246) overall and 71 days (range 15–211) in the 0.6 mg/kg group (n = 8, predicted therapeutic dose; of whom 3 pts had PFS4: CRC: n = 1/4 [all CDH17+]; non-CRC: n = 2/4). Conclusions: In heavily pretreated pts, BI 905711 was associated with a tolerable safety profile and early signs of disease control. BI 905711 will be further assessed in Phase Ib in 4 dose groups: 0.6/1.2/2.4 mg/kg every 14 days, and 0.6 mg/kg weekly. Clinical trial information: NCT04137289 .

3535 Background: Approximately 20% of colorectal cancer (CRC) has an activating mutation in the PIK3CA oncogene. PIK3CA codes for the catalytic subunit of phosphoinositide 3-kinase alpha (PI3Kα), which ultimately activates the AKT and mammalian target of rapamycin (mTOR) pathway. The PI3Kα inhibitor alpelisib has been approved for breast cancer where a single PIK3CA activating mutation is sufficient for response. However, two activating mutations (multi-hit) in the PIK3CA allele substantially increases PI3Kα signaling compared to single hot spot mutations and results in exceptional response to PI3Kα inhibition. We aimed to identify the prevalence of PIK3CA multi-hit mutations in CRC to identify patients potentially susceptible to PI3K inhibitors. Methods: Tissue-based comprehensive genomic profiling (CGP) was performed in a Clinical Laboratory Improvement Amendments (CLIA)-certified, CAP (College of American Pathologists)-accredited laboratory (Foundation Medicine Inc., Cambridge, MA, USA) on all-comers during the course of routine clinical care from 2013-2021. Approval was obtained from the Western Institutional Review Board (Protocol No. 20152817). Hybrid capture was carried out for at least 324 cancer-related genes, including PIK3CA. Results: We identified 48,836 patients with advanced CRC who underwent Foundation Medicine testing and 846 (1.7%) patients with multi-hit PIK3CA mutations. Additional clinical and molecular data was available for 41,154 of these patients of which 710 (1.7%) had multi-hit PIK3CA and 7627 (19%) had any deleterious PIK3CA mutation . The local colon tumor was used for sequencing in 70% of cases and a separate site in 30% of cases. Patients with PIK3CA multi-hit mutations were 53% male with a median age of 60 (interquartile range 50-70). The microsatellite status was available for 697 of 710 patients with multi-hit PIK3CA and 123/697 (18%) were microsatellite instability-high. The Table outlines the genes with cooccurring mutations of >10% prevalence for multi-hit PIK3CA CRC, including the clinically relevant mutations in KRAS (65%) and BRAF (13%). The four most common PIK3CA variants were H1047R (9.8%), E545K (9.2%), E542K (9.0%) and R88Q (7.1%). The most common variant pair was E542K with E545K in 4.7% of multi-hit cases. Conclusions: Double-hit mutations in PIK3CA are seen in 1.7% of advanced CRC patients and may represent a subset of patients that may have enhanced sensitivity to PI3K inhibitors. Given the high prevalence of CRC in the United States and worldwide this represents a clinically meaningful prevalence of multi-hit PIK3CA. Future investigation on the clinical utility of PI3K inhibitors may be warranted in multi-hit PIK3CA CRC.[Table: see text]

136 Background: O6-methylguanine DNA methyltransferase (MGMT) is the principal repair mechanism of DNA damage created by alkylating agents. Approximately 40% of colorectal cancers (CRC) exhibit MGMT promoter hypermethylation, expected to result in gene silencing. The role of MGMT as a biomarker is poorly understood, and quantitative measurement of MGMT protein has not been performed in CRC. We hypothesized that altered tumor MGMT would affect DNA damage response, modulate adaptive anti-tumor immune responses, and influence survival in CRC. Methods: We used multiplexed quantitative immunofluorescence (QIF) to study 4 clinically annotated retrospective patient cohorts treated from 2000 to 2017 at Yale, consisting of 400 formalin-fixed paraffin-embedded CRC cases represented in tissue microarray format. These included paired tumors and adjacent non-tumor colonic mucosa (n = 112 pairs in Cohort 1), paired primary CRC and lymph node metastases (n = 31 pairs in Cohort 2), and 2 cohorts with primary stage I-IV tumors (n = 250 in Cohort 3; n = 150 in Cohort 4). We established a QIF panel for simultaneous, localized measurement of DAPI (all cells), cytokeratin (CK; tumor epithelial cells), MGMT, γH2AX (DNA damage response), and CD8. The measurement of fluorescent signals in CK-positive tumor cells or CK-negative stromal cells was achieved using co-localization strategies and the AQUA QIF platform. Results: We identified lower MGMT protein levels in CRC cells than in non-tumor colonic epithelium (p < 0.0001), and in lymph node metastases compared to paired primary CRC tumors (p = 0.0411). Using the visual detection threshold, tumor selective MGMT protein downregulation was identified in 20% of cases in Cohorts 3 and 4. Microsatellite instability-high (MSI-H) cases showed decreased tumoral MGMT protein compared to microsatellite stable (MSS) cases (p = 0.002). In Cohorts 3 and 4, low tumoral MGMT was consistently associated with increased CD8+ tumor infiltrating lymphocytes (p = 0.0472; p = 0.0002). The association between MGMT and γH2AX was inconsistent. The association between MGMT and CD8 was significant only in MSS cases (p = 0.0001) but not in MSI-H cases (p = 0.0979), supporting that the effect is not driven by MSI-H status. Tumor-selective MGMT downregulation evidenced by a low tumor-to-stroma ratio was associated with improved progression-free survival and overall survival in both Cohorts 3 and 4, but statistical significance was only seen in Cohort 3. Conclusions: MGMT protein downregulation occurs in 20% of CRCs and is associated with increased adaptive anti-tumor immune responses, mismatch repair (MMR) deficiency and better prognosis. MGMT deficiency may alter DNA repair in tumor cells and mediate the accumulation of antigenic mutations or neopeptides, independent from MMR status. Our results support a biomarker role of MGMT protein and suggest a role for immunotherapy combinations in MGMT deficient tumors.

TPS375 Background: In normal conditions, Claudin (CLDN)18.2 is a tight junction protein with expression strictly confined to differentiated epithelial cells in gastric mucosa. CLDN18.2 has been found to be upregulated, and involved in tumor development and progression in a variety of tumor types such as gastric, pancreatic, and bile duct cancer (BTC). The biological characteristics of CLDN18.2 suggest it is an ideal therapeutic target for cancer drug development. IMAB362 is the first anti-CLDN18.2 monoclonal antibody (mAb) of high potency to have been tested in humans and it revealed clinical efficacy in gastric cancer in a phase II study. TST001, a humanized IgG1 mAb, binds to a distinct epitope of CLDN18.2 with higher affinity and mediates CLDN18.2 expressing cancer cell death through antibody-dependent cellular cytotoxicity (ADCC) in comparison with IMAB362; Furthermore, TST001 is produced using an optimized glycoengineering process to increase affinity to FcR. The enhanced binding to CLDN18.2 on tumor cells and FcR on NK cells results in more efficient engagement of the tumor cells with NK cells and antibody mediated cellular cytotoxicity. In preclinical xenograft studies, TST001 displayed potent anti-tumor activities in the tumor models with medium to high level of CLDN18.2 expression and synergy anti-cancer effect with checkpoint inhibitor. A mAb specific for CLDN18.2 was also developed as an IHC based biomarker for patient enrollment in the clinical trials. Methods: This is an open-label, multi-center, phase I clinical trial to evaluate the safety, maximum tolerated dose (MTD), pharmacokinetics (PK) profile and preliminary anti-cancer effect of TST001 in patients with locally advanced or metastatic solid tumors. (NCT04396821) The study consists of two parts: Part A is a 3+3 dose escalation design with sequential dose cohorts of 1, 3, 6, 10mg/kg in Q2W and Q3W schedules. Based on the emerging safety data, higher doses may be proposed for testing. About 27-54 patients will be enrolled. Dose expansion (Part B) will utilize doses of TST001 based on the emerging data from Part A. In Part B, up to 20 patients with CLDN18.2 overexpression per tumor specific cohort will be enrolled to 3 cohorts: A: TST001 single agent in gastric/gastroesophageal junction (G/GEJ) cancer; B: TST001 + nivolumab in G/GEJ cancer; C: TST001 single agent in pancreatic cancer or BTC. All patients in Part B will be selected by CLDN18.2 expression by central lab testing. The safety, anti-tumor activity, and PK will be further assessed in Part B. Enrolment began in July 2020 in the USA and is ongoing in multiple sites. As of 20 September, 2021, 23 subjects were dosed in Part A and the dose of 10mg/kg is being tested. Another phase I study of TST001 single agent and in combination with chemotherapy in patients with metastatic solid tumor is also ongoing in China (NCT04495296).

TPS222 Background: Activation of the tumor necrosis factor-related apoptosis-inducing ligand receptor 2 (TRAILR2) induces apoptosis via caspase activation. Targeting TRAILR2 is an attractive therapeutic strategy, but some early TRAILR2 agonists were associated with severe hepatotoxicity. Cadherin 17 (CDH17), a membrane protein highly expressed in GI cancers, is not expressed in normal hepatocytes so using CDH17 as a liver-sparing anchor may avoid hepatotoxicity. BI 905711 is a tetravalent bispecific antibody that cross-links TRAILR2 with CDH17 to induce CDH17-dependent TRAILR2 oligomerisation. In preclinical assays, BI 905711 demonstrated a potency shift of ̃1000 fold versus the 1st-generation TRAILR2 agonist lexatumumab. BI 905711 induced apoptosis in CDH17-positive tumor cells in vitro, impaired tumor growth in pt-derived colorectal cancer (CRC) xenografts, and no hepatotoxicity was observed. Methods: This phase Ia/Ib study (NCT04137289) aims to determine the safety, maximum tolerated dose (MTD), pharmacokinetics (PK), pharmacodynamics, and preliminary efficacy of BI 905711 in pts with advanced, refractory GI cancers. Up to 140 adult pts with histologically confirmed, advanced unresectable/metastatic colorectal, gastric, esophageal or pancreatic adenocarcinoma, cholangiocarcinoma, or gallbladder or small intestine carcinoma, who have progressed on standard-of-care therapies, will be enrolled. In phase Ia, pts will receive intravenous BI 905711 at escalating doses (range 0.02–4.8 mg/kg) every 14 days, until disease progression or unacceptable toxicities. Dose escalation will be guided by a Bayesian logistic regression model with overdose control based on dose-limiting toxicities (DLTs) in the first 28 days. In phase Ia, a minimum number of CRC pts will be enrolled to each cohort: ≥1 CRC pt at each of the 2 lowest dose levels (0.02/0.06 mg/kg) and 4 pts at each subsequent dose level (0.2/0.6/1.2/2.4/3.6/4.8 mg/kg). In parallel to dose escalation in CRC pts, up to 4 pts with non-CRC GI cancers will be included at the dose level below that of the CRC cohort. If an objective response (OR) per RECIST v1.1 is observed in CRC or non-CRC GI pts at a safe dose level, up to 10 additional pts with the same tumor type will be recruited at that dose level. In phase Ib, CRC pts will be randomized into up to 4 dose cohorts (as determined in phase Ia; n=20 each) to define the recommended phase II dose. The primary endpoints are determination of the MTD based on the proportion of pts with DLTs (phase Ia) and OR rate based on RECIST v1.1 (phase Ib). Secondary endpoints include PK parameters and OR in pts with measurable disease (phase Ia), and disease control, tumor shrinkage, duration of response, and progression-free survival (phase Ib). Trial enrollment is ongoing, with 33 pts enrolled to date. Clinical trial information: NCT04137289.

e18584 Background: Cancer progression rates following diagnosis are readily measured. However, the progression rate of cancer during the preclinical sojourn time is generally unobserved. Understanding the duration of preclinical stages (“dwell time”) would allow clinicians to better identify appropriate screening intervals for cancer. We therefore elicited estimates of progression rate during the preclinical sojourn time for a wide variety of malignancies from a panel of clinical experts. Methods: We used a validated consensus methodology (RAND/UCLA modified Delphi panel method) to elicit per-stage dwell time estimates for 20 solid cancers and lymphoma from experts. Eleven experienced oncologists (general and subspecialists) from community and academic centers reviewed literature on the natural history of disease and estimated in number of years (

2624 Background: KY1044, is a fully human IgG1 anti ICOS antibody designed to stimulate Teffs and to deplete ICOS high Tregs in the tumor microenvironment. Methods: Patients with advanced/metastatic malignancies received escalating doses of KY1044 as a single agent and in combination with atezolizumab 1200 mg by IV infusion every 3 weeks until disease progression or unacceptable toxicity. Dose escalation was guided by a modified toxicity probability interval design. The primary objective was to determine safety, tolerability, and maximum tolerated dose. Cohorts that were tolerated were later enriched with more subjects. AEs were classified according to CTCAE v5 and efficacy measures performed according to RECIST v1.1 every 8 weeks for the first 16 weeks and then every 12 weeks. Results: As of 16-Dec-2020, a total of 103 patients have been enrolled in the study (38 patients as monotherapy in 6 cohorts at doses ranging from 0.8 to 240 mg and 65 in combination with atezolizumab in 5 cohorts at doses 0.8 – 80 mg). 63% and 55% of patients received ≥4 prior anti-cancer therapies in the single agent and combination cohorts, respectively. All cohorts were completed without DLTs during the first 21 days of treatment. In the KY1044 single agent cohorts, 47.4% of patients experienced treatment-related AEs (TRAEs), all were grades 1 or 2. In the combination cohorts, TRAEs were observed in 58% of patients. Most of the TRAEs were grade 1 or 2 apart from 8 TRAEs that were ≥grade 3 occurring in 43 weeks (range 45 to 66 weeks). Conclusions: KY1044 is well tolerated as single agent and in combination with atezolizumab. Objective responses have been observed in this phase 1 part of the study. The phase 1 expansion and phase 2 part of the study is ongoing. Clinical trial information: NCT03829501.

TPS150 Background: ATP release from dying cancer cells in response to platinum-based chemotherapy increases extracellular immunosuppressive adenosine, which binds and activates the A2a and A2b receptors on immune cells. Adenosine-mediated signaling impairs activation, proliferation, and cytotoxic activity of effector T cells, resulting in inhibition of antitumor activity. Concomitant adenosine receptor blockade may therefore enhance the therapeutic efficacy of chemo/immunotherapy regimens. Etrumadenant (AB928), the first clinical-stage, small-molecule, dual adenosine receptor antagonist, is highly potent, pharmacodynamically active, and has been well tolerated in dose escalation studies as monotherapy or combined with chemo/immunotherapy. Recently, results were reported for the ARC-3 phase 1/1b study of etrumadenant + modified 5-fluorouracil + oxaliplatin (mFOLFOX-6) in patients (pts) with mCRC. In this study, etrumadenant + mFOLFOX-6 was well tolerated without significant additive toxicity. Disease control was observed in patients with RAS/BRAF mutated mCRC, as well as 3L+ disease previously treated with FOLFOX and/or FOLFIRI (PR and/or SD >4 mo). Due to deep responses in 1L-3L+ pts, 6 pts had the opportunity to pursue surgery and radiotherapy with curative intent. The encouraging results from ARC-3 warrant further evaluation of etrumadenant-based combination therapy for mCRC. Methods: ARC-9 is a phase 1b/2, multicohort, open-label, randomized platform study designed to evaluate safety and clinical activity of etrumadenant (150 mg orally once daily [QD]) in combination with standard-of-care (SOC) regimens or novel therapeutics in pts with mCRC (Table). Cohort eligibility is based on prior anticancer treatment history. Pts enrolled in Cohorts A and B will be randomized (2:1) into the experimental vs SOC arms. Cohort C consists of a single arm to allow inclusion of novel agents as they become available with built in early stopping rules for futility. Pts who progress on the SOC arm of Cohort A can enroll in Cohort B; pts who progress on the SOC arm of Cohort B can crossover to the experimental arm. Primary endpoints across cohorts are shown in Table. Safety monitoring will occur throughout the trial, disease assessments will occur every 8 weeks, and correlative study pre- and on-treatment biopsies will be performed. [Table: see text]

82 Background: The MORPHEUS platform consists of multiple, global, open-label, randomized Phase Ib/II trials designed to identify early efficacy signals and safety of treatment (tx) combinations across tumor types. Isatuximab (isa; anti-CD38) targets CD38 receptors expressed on immunosuppressive cells in the tumor microenvironment. We hypothesized atezolizumab (atezo; anti–PD-L1) + isa would induce an anti-tumor response beyond that of regorafenib (rego), a multi-kinase inhibitor, in patients (pts) with tx-refractory metastatic colorectal cancer (mCRC). Methods: This randomized Phase Ib/II trial (NCT03555149) enrolled pts with microsatellite stable/mismatched repair proficient mCRC who had received ≤ 2 prior tx lines (fluoropyrimidine-, oxaliplatin- or irinotecan-containing chemotherapy plus a biologic agent). Pts received atezo (1200 mg intravenously [IV] every 3 weeks [q3w]) + isa (10 mg/kg IV q3w) or control tx with rego (160 mg orally days 1–21; dose escalation to 160 mg during Cycle 1 allowed per institutional guidelines). The primary endpoint was objective response rate (ORR; investigator-assessed RECIST 1.1); secondary endpoints included progression-free survival (PFS), overall survival (OS), disease control rate (DCR) and safety. Results: Data cutoff date was March 3, 2020. Fifteen pts received atezo + isa and 13 pts received rego. Fourteen atezo + isa pts (93.3%) and 11 control arm pts (84.6%) had received 2 prior lines of tx; 9 atezo + isa pts (60.0%) and 9 control pts (69.2%) had liver metastases at enrollment. No responses were seen in either arm; 3 pts receiving atezo + isa (20.0%) and 8 control pts (61.5%) had stable disease as their best response. DCR (response and/or stable disease ≥ 12 weeks) was 6.7% with atezo + isa and 15.4% with control. One pt treated with atezo + isa beyond progression had prolonged disease stabilization. Median PFS was 1.4 mo (95% CI: 1.4, 1.8) with atezo + isa and 2.8 mo (95% CI: 1.6, 3.1) in the control arm; median OS was 5.1 mo (95% CI: 3.1, 7.8) with atezo + isa and 10.2 mo (95% CI: 4.8, not reached) with control. Tx-related adverse events (AEs, Grade 1-4) occurred in 13 atezo + isa pts (86.7%), and 12 control pts (92.3%). The most common tx-related AEs with atezo + isa were infusion-related reaction (73.3%), nausea (26.7%) and fatigue (20.0%). No Grade 5 AEs occurred in the atezo + isa arm, 1 (7.7%) was reported in the control arm (sepsis, considered unrelated to study tx). No atezo + isa pts and 1 control-arm pt (7.7%) withdrew from treatment due to a tx-related AE. Biomarker analyses did not identify any significant trends related to efficacy. Conclusions: In this trial, superior efficacy of atezo + isa vs rego was not shown. However, the atezo + isa combination was well tolerated, with a manageable safety profile. Clinical trial information: NCT03555149.

462 Background: To evaluate epidemiology, treatment and outcomes of stage 4 gastrointestinal stromal tumors (GIST). Methods: The 2010-2016 United States National Cancer Database was queried for adult patients diagnosed with invasive GIST (ICD code 8936) at AJCC stage 4, without prior malignant disease. Overall survival (OS) was evaluated using Cox proportional hazards regression, accounting for potential confounders in multivariable models. Results: A total of 1,578 stage 4 GIST were included (13.3% of all GIST) with a male:female ratio of 1.38:1. The most common cancer site was the stomach (55.4%) and small intestine (40% of stage 4 GIST). At diagnosis, median age was 62 years and median tumor diameter 10 cm. Distant organ metastases were reported in 58.7%, the most frequent being hepatic (n = 801, 50.8% of stage 4 GIST). The majority of stage 4 GIST patients received systemic therapy (78.6%), either alone (35.4%) or combined surgery+systemic therapy (43.3%). Systemic therapy was administered neoadjuvant in 6.9%, adjuvant in 32%, and neoadjuvant+adjuvant in 5.1% of surgically resected patients. Another n = 204 patients (12.9% of all stage 4 GIST) were treated with surgical resection alone. GIST overall survival rates were 88%, 77%, 67% and 51% at 1, 2, 3, and 5 years, respectively. On multivariable Cox proportional hazard models, primary GIST treatment independently affected OS: compared to combined surgery+systemic therapy, OS was shorter for patients receiving systemic therapy alone (HR = 2.77 (95% CI: 2.12-3.61, p < 0.001)) and no treatment (HR = 4.2 (95% CI: 2.75-6.43, p < 0.001)). No significant OS difference was evident comparing surgery+systemic therapy and surgery alone (HR = 1.23 (95% CI: 0.88-1.72, p = 0.227)). In subgroup analyses of patients undergoing surgical resection, no statistically significant OS difference was evident comparing neoadjuvant, adjuvant and combined neoadjuvant+adjuvant systemic therapy on multivariable analyses. Treatment at non-academic vs. academic centers was associated with shorter OS (multivariable HR = 1.36 (95% CI: 1.1-1.69, p = 0.005)). Further independent OS predictors were male sex (vs. female HR = 1.28 (95% CI: 1.03-1.59, p = 0.023)), older age, higher comorbidities, higher cancer grade, and larger cancer diameter. Conclusions: Stage 4 GIST is a rare gastrointestinal malignancy that most commonly manifests in the stomach and small intestine in male patients, with frequent hepatic metastases and excellent 5-year OS rates of up to 51%. Combined surgery+systemic therapy demonstrates best outcomes, although surgical resection alone might yield comparable results in selected patients.

e16052 Background: To assess microsatellite instability (MSI) and KRAS mutation prevalence, discrepancies and their impact on outcome in AJCC stage IV colorectal adenocarcinoma (colorectal cancer; CRC). Methods: The 2010-2016 United States National Cancer Database was queried for stage IV CRC patients > 18yo and information on MSI and KRAS mutation. Microsatellite status was stratified as stable microsatellites (MSS, including low instability) and microsatellite instability high (MSI-H). KRAS status was stratified as mutation and wildtype. Prevalence and discrepancies were evaluated according to patient demographics, US geography and CRC factors. Overall survival (OS) was assessed using Cox proportional hazards models adjusting for age, gender, race, comorbidities, metastatic burden, CRC treatment, treatment center type, and year of CRC diagnosis. A priori, a statistical interaction test between microsatellite status, KRAS status and primary CRC site was planned. Results: Microsatellite/KRAS status was available for n = 10,844/n = 25,712 patients, respectively. OS was assessed in n = 5,904 patients with data on both microsatellite and KRAS status, and follow-up. Overall prevalence of MSI-H was 3.1% and KRAS mutation 42.4%. Microsatellite and KRAS status varied according to primary CR site, as presented in Table. Further variation was evident according to US-geography, with MSI-H rates ranging from 1.6% to 4.1%, and KRAS mutation rates ranging from 41.1% to 44%. On multivariable analyses, longer OS was observed in patients with KRAS wildtype versus mutation (HR = 0.91, 95% CI: 0.85-0.97 p = 0.004), MSS versus MSI-H (HR = 0.75, 95% CI: 0.62-0.9, p = 0.003), and left-sided versus right-sided CRC (HR = 0.65, 95% CI: 0.6-0.7, p < 0.001). The effect of KRAS mutation further varied with CRC site and microsatellite status (interaction p = 0.002). Conclusions: Depending on its primary site and US geography, stage IV CRC shows distinct mutational behavior. KRAS mutation, MSI-H, MSI-H and primary CRC sidedness independently affect overall survival and interact with distinct prognostic profiles. Generically classifying adenocarcinomas of different site as “CRC” might deprecate this diversity. [Table: see text]

770 Background: Patients (pts) receiving treatment for metastatic pancreatic cancer (MPC) experience significant symptoms, treatment related side effects and psychosocial burdens. The ASCO guidelines recommend early palliative care consultation (PCC) to improve quality of life and survival. However, limited real-world data is available regarding the timing of PCC, hospice enrollment and location of death (LOD) in pts with MPC. Methods: We conducted a retrospective observational analysis of pts treated with chemotherapy for MPC at the Yale Smilow Cancer Hospital and affiliated community care centers (CCC) from January 2011 to April 2019. Patient demographics, treatment dates, initial PCC, enrollment of hospice at the time of death and LOD were manually abstracted from the electronic medical record. Univariate and multivariable logistic regression analyses were conducted to predict for PCC and death outside the hospital. Results: Of 363 pts identified with MPC who received chemotherapy, 38% (138) had a PCC. 67% (93) of patients’ initial PCC was in the hospital versus 33% (45) in the outpatient setting. The median time from the start of first-line chemotherapy to the first PCC was 5.2 months (interquartile range [IQR] 1.2 – 12.9). The median time from the first PCC to death was 1.5 months (IQR 0.5 – 4.42). At the time of our analysis, 300 pts had died and of those 76% (229) were enrolled on hospice at the time of death while 24% (71) were not. With respect to LOD, 47% (139) of pts died at home with hospice, 31% (94) at an inpatient hospice facility and 22% (67) died in the hospital. Female gender was associated with an increased likelihood of a PCC (HR 1.78, 95% CI 1.07-2.94, P = 0.026). Pts treated at a CCC were less likely to have a PCC (HR 0.21, 95% CI 0.12-0.36, P < 0.001). A PCC was not associated with a higher likelihood of death outside the hospital (HR 1.31, 95% CI 0.75-2.29, P = 0.346). Conclusions: Although most pts with MPC enroll in hospice, PCC is generally underutilized. In fact, many pts receive PCC near the end of life and in the hospital. Further studies are warranted to determine how best to incorporate early PCC to maximize supportive care for pts with MPC.

609 Background: While carcinoid and neuroendocrine tumors (NET) can manifest throughout the human body, approximately 2/3 of these tumors arise in the gastrointestinal (GI) tract. Methods: The 2019 version of the National Cancer Database was searched for adult patients with carcinoid or neuroendocrine tumors of the GI tract. Tumor incidence, patient demographics, treatment and cancer variables of GI carcinoid/NETs were compared to other GI cancers. Cox proportional hazards models were used to evaluate overall survival (OS). Results: A total of 101,744 patients were included. Carcinoid/NET incidence varied with primary location, with highest proportion in the small intestine (55.7% of all small intestinal cancers). Compared to other histologies, carcinoid/NETs were more common in younger patients (median age 60 vs. 68years, p < 0.001) and African Americans (18 vs 12%, p < 0.001). Median carcinoid/NET diameter 15mm vs. 40mm for other cancers (p < 0.001). Carcinoid/NETs were diagnosed at lower stages than other histologies (AJCC stage 3/4, 30% vs. 47%, p < 0.001), with most metastases in the liver (8.9%). Stage 1/2 carcinoid/NETs were most often treated by surgical resection with or without concurrent chemotherapy (65%; 22%), whereas stage 3/4 patients received chemotherapy alone or in combination with surgical resection more commonly (39%; 25%). After multivariable adjustment, primary cancer site emerged as an independent prognosticator: longest OS was observed in patients with carcinoid/NET of the small intestines (vs. colonic frame HR = 0.40, 95% CI: 0.37-0.44, p < 0.001). Comparable results were evident in patients with hepatic metastases (small intestines vs. colonic frame carcinoid/NETs; HR = 0.20, 95% CI: 0.17-0.24, p < 0.001). Other independent OS prognosticators were patient age, gender, race and comorbidities, as well as tumor size, stage and treatment. Conclusions: Compared to other GI-associated tumors, carcinoid/NETs are more common in younger patients, African Americans, diagnosed with smaller diameter at low stage. Carcinoid/NETs of the small intestines are associated with improved overall survival compared to other primary disease sites.

769 Background: FOLFIRINOX (FFX) and Gemcitabine plus nab-paclitaxel (GN) are established first line (1L) therapies for metastatic pancreatic cancer (MPC) but real-world data on their comparative effectiveness is limited. Methods: All cases of MPC treated with 1L FFX or GN at Yale Smilow Cancer Hospital and the affiliated community care centers (CCC) from January 2011 – April 2019 were reviewed. Patient (pt) demographics, prior therapy, initial and subsequent dose reductions (DR), time to treatment discontinuation (TTD), overall survival (OS), and second line (2L) treatment data were manually abstracted from the electronic medical record. Categorical and continuous variables were compared between 1L FFX and GN cohorts via the Chi-squared and Wilcoxon rank-sum tests. Median OS was calculated by the Kaplan-Meier method. Results: We identified 363 MPC pts treated with 1L FFX or GN; 269 (74%) pts were treated with FFX and 94 (26%) with GN as 1L therapy. 204 (56%) pts were treated at the main campus and 159 (44%) at a CCC. Demographics and baseline characteristics (FFX/GN) were as follows: gender (male) 55% / 41%; race (white) 82% / 77%; age < 76 90% / 71% ( P < 0.001). 332 (91%) of pts received no prior therapy; 21 (6%) had prior surgery plus adjuvant gemcitabine and 10 (3%) had surgery alone. 98% of FFX-treated pts were treated with upfront DR, compared to 78% of GN-treated pts ( P= 0.003). 78% and 53% of FFX-treated and GN-treated pts, respectively, had subsequent DR ( P< 0.001). Median TTD was 4.8 months with FFX and 3.4 months with GN ( P= 0.0029) and the median OS was 11.3 months with FFX versus 7.2 months with GN ( P < 0.0001). After 1L, 33% and 61% of FFX- and GN-treated pts, respectively, received no further chemotherapy ( P< 0.001). Conclusions: In the largest manually abstracted retrospective analysis to date, MPC pts treated with 1L FFX were younger, more likely to receive 2L therapy, and had increased survival compared to pts treated with GN. The OS of pts treated with FFX was similar to the OS reported by Conroy et al despite upfront dose attenuations in 98% of pts. A randomized trial is needed to confirm optimal sequencing of chemotherapy in MPC.

272 Background: Expert knowledge is often shared among academic oncologists at tumor boards (TBs) at National Cancer Institute Designated Cancer Centers (NCI-CCs), but not documented or made accessible to community oncologists. Using an oncologist-only question and answer (Q&A) website, we sought to disseminate expert insights from TBs at NCI-CCs to provide educational benefit to the oncology community. Methods: A process was designed with faculty at 11 NCI-CCs to document and share discussions from TBs focused on areas of clinical complexity and practice variation on theMednet.org, an interactive Q&A website of over 8,700 US oncologists. One faculty member from each TB was selected as a site leader. She or he distilled discussions about patient management from the TB into a question that addressed the clinical situation being discussed. After the question was posted, faculty at the participating NCI-CCs were asked to answer the question on theMednet. Answers were peer reviewed, indexed, stored and disseminated via email newsletters to registered oncologists. Community engagement was measured by Q&A page views, upvotes of Q&A, and poll participation. Results: A total of 15 Breast, Thoracic, and Gastrointestinal programs from 11 NCI-CCs participated. Between 12/2016 and 5/2019, faculty highlighted 146 questions from their TBs. Q&A were viewed 43,291 times by 3,585 oncologists including 2,264 community oncologists. One hundred and eighty-four answers are posted by 56 academic physicians and peer reviewed by 76 academic physicians. One hundred and eighty-five publications were cited. Community oncologists upvoted Q&A 808 times and voted in 45 polls related to the questions 1,667 times. Viewership of NCI-CC Q&A increased by 419% over time. Q&A were repeatedly searched and viewed, with 90% of all TB Q&A viewed every month. Conclusions: Via the online Q&A theMednet platform, NCI-CC providers effectively made expert knowledge easily accessible to community oncologists across the US. Timely access to evidence based recommendations from expert faculty can inform future practice choices in the community. [Table: see text]

437 Background: Induction FFX for PC deemed either BR or LA at diagnosis provides an opportunity to downstage pts with the aim of an R0 surgery. The addition of RT after induction FFX may further downstage. However, there is a paucity of data regarding long-term survival for BR and LA patients successfully downstaged and resected. We performed a retrospective review of BR and LA PC treated with induction FFX +/- RT followed by surgery or consolidative RT at the Yale Cancer Center (YCC) to assess survival in these two cohorts. Methods: Clinical data was abstracted for pts with BR or LA PC who had surgery or received consolidative RT without surgery after induction FFX +/- RT at the YCC from 2010-2018. Surgical pts were re-reviewed by a radiologist to assess vascular involvement (BR vs. LA) using NCCN criteria. PFS and OS for surgery and consolidative RT were analyzed by the Kaplan-Meier method. Survival was compared via the log rank test. Results: 102 pts met inclusion criteria (BR=47, LA=55), 41 pts had surgery [BR=29/47 (62%) LA=12/55 (22%)] and 61 pts had consolidative RT [(BR= 18/47 (38%), LA= 43/55 (78%)] after induction FFX. 18 surgery pts received RT prior to resection and all surgery pts had R0 resection. Median follow up was 25 mo (range 5 – 97). Median PFS with surgery was 22 mo (95% CI 15 – 59) vs 14 mo (95% CI 10.9 – 20.1) with consolidative RT (p

630 Background: TAS-102 is an oral combination of the anti-metabolite 5-trifluorothymidine (FTD) and a thymidine phosphorylase inhibitor (TPI), preventing the degradation of FTD. It is approved as mCRC monotherapy with improved survival. Oxaliplatin is often reintroduced in mCRC after progressive disease (PD) on maintenance 5-FU although response is poor. The decreased efficacy may be related to acquired 5-FU resistance. We therefore explored the safety and efficacy of oxaliplatin in combination with an alternative and non cross-resistant anti-metabolite, TAS-102. Methods: Phase 1 of TAS-OX is a 3+3 dose-escalating study at a starting dose of TAS-102 25 mg/m2 and oxaliplatin 85 mg/m2 with three dose levels (table). TAS-102 is administered days 1-5 and oxaliplatin day 1, every 2 weeks. Eligible patients previously received 5FU, oxaliplatin, irinotecan, appropriate biologics, had measurable disease, usual laboratory parameters, and ECOG PS 0-1. The primary objective was to determine the recommended phase II dose (RP2D). Results: Twelve patients were evaluable for dose limiting toxicity (DLT). No DLTs were observed. Treatment related grade ≥ 3 AEs were neutropenia (n = 4) and thrombocytopenia (n = 1). No AEs resulted in treatment discontinuation. Two patients (dose levels 2 and 3) required dose reductions for prolonged neutropenia. Median number of cycles for all treated patients was 6 ± 4. The disease control rate (DCR) at 8 weeks was 67%. Best response in all evaluable patients was 1 PR (8%) 7 (59%) SD and 4 (33%) PD. Conclusions: The RP2D of TAS-102 is 35 mg/m2 in combination with oxaliplatin 85 mg/m2. No DLTs were observed and no unexpected AEs were seen. The DCR in this heavily pretreated patient population is encouraging. Phase II is now enrolling at this dose (NCT 02848079). Clinical trial information: NCT02848079. [Table: see text]

TPS4137Background: Gastric cancer remains a significant health problem in the US and globally with more than 951,600 annual cases worldwide. Moreover, the incidence of GEJ-centered adenocarcinoma i...

61 Background: Early results from recent studies using immune checkpoint blockade targeting programmed death 1 (PD-1) have suggested that pleomorphic undifferentiated sarcomas may have response rates of over 40%. As of now predictive biomarkers for response and resistance have been incompletely characterized. Methods: A 58-year-old man who received radiation for a head and neck squamous cell cancer, developed a radiation-induced undifferentiated pleomorphic sarcoma (UPS) in his neck in 2014. His sarcoma was resected but recurred in 2015. He received adjuvant radiation after surgical debulking in September 2015 and was found on radiation simulation scan to have new widespread metastatic disease involving liver, lung, and bone. He started treatment on nivolumab in November 2015 and has had a sustained near complete response in all lesions. Results: All sites had a near complete response to nivolumab treatment. Pre treatment tissue analysis revealed no mutations or amplifications in 134 cancer-related genes, on the Oncomine Assay (Life Technologies, Inc.). Normal tissue was noted to be heterozygous for BRCA2 N372H, while the allelic fraction of the 372H variant in the tumor was found to be 85%. Programmed death ligand-1 (PDL-1) immunohistochemistry staining of the tumor tissue was negative. Tumor infiltrating lymphocytes were not noted in the tumor tissue specimen. Conclusions: This patient exhibited a near complete response to all sites of disease, with remaining PET avidity in a single hilar node. The role of the BRCA2 variant and radiation just prior to starting nivolumab is unknown. BRCA2 N372H is a common single nucleotide polymorphism (SNP) in the population with a minor allele frequency of 0.25. Although the effect of the 372H variant on BRCA2 protein structure is predicted to be minimal, population studies have suggested a slightly increased risk of breast and ovarian cancer in homozygotes. It is possible that the radiation treatment to the site of recurrence in the head and neck, resulted in an abscopal effect enhancing the therapeutic effect of nivolumab therapy. Additional testing on his tissue is being done to further investigate the response.

e15124Background: TAS-102 (tri-fluoro-thymidine and tipiracil hydrochloride; a novel combination oral nucleoside anti-tumour agent) was recently approved for patients with refractory metastatic col...

e18188Background: Multidisciplinary tumor board (MTB) meetings have become a standard for evidence based decision making at medical centers worldwide. We believe that the expertise at subspecialty ...