Search

PURPOSE Key oncology end points are not routinely encoded into electronic medical records (EMRs). We assessed whether natural language processing (NLP) can abstract treatment discontinuation rationale from unstructured EMR notes to estimate toxicity incidence and progression-free survival (PFS). METHODS We constructed a retrospective cohort of 6,115 patients with early-stage and 701 patients with metastatic breast cancer initiating care at Memorial Sloan Kettering Cancer Center from 2008 to 2019. Each cohort was divided into training (70%), validation (15%), and test (15%) subsets. Human abstractors identified the clinical rationale associated with treatment discontinuation events. Concatenated EMR notes were used to train high-dimensional logistic regression and convolutional neural network models. Kaplan-Meier analyses were used to compare toxicity incidence and PFS estimated by our NLP models to estimates generated by manual labeling and time-to-treatment discontinuation (TTD). RESULTS Our best high-dimensional logistic regression models identified toxicity events in early-stage patients with an area under the curve of the receiver-operator characteristic of 0.857 ± 0.014 (standard deviation) and progression events in metastatic patients with an area under the curve of 0.752 ± 0.027 (standard deviation). NLP-extracted toxicity incidence and PFS curves were not significantly different from manually extracted curves ( P = .95 and P = .67, respectively). By contrast, TTD overestimated toxicity in early-stage patients ( P < .001) and underestimated PFS in metastatic patients ( P < .001). Additionally, we tested an extrapolation approach in which 20% of the metastatic cohort were labeled manually, and NLP algorithms were used to abstract the remaining 80%. This extrapolated outcomes approach resolved PFS differences between receptor subtypes ( P < .001 for hormone receptor+/human epidermal growth factor receptor 2− v human epidermal growth factor receptor 2+ v triple-negative) that could not be resolved with TTD. CONCLUSION NLP models are capable of abstracting treatment discontinuation rationale with minimal manual labeling.

PURPOSE Biliary tract cancers (BTCs), which include intrahepatic cholangiocarcinoma (ICC), extrahepatic cholangiocarcinoma (EHC), and gallbladder cancer (GBC), have limited treatment options. We sought to comprehensively examine the clinical and molecular characteristics of BTCs with amplification or mutation of ERBB2. METHODS Demographic, outcome, and treatment response data were collected for patients with ERBB2-altered BTC identified by next-generation sequencing with Memorial Sloan Kettering-Integrated Mutation Profiling of Actionable Cancer Targets from 2014 to 2018. RESULTS A total of 517 patients with BTC underwent next-generation sequencing (ICC, n = 313; EHC, n = 93; GBC, n = 111). Twenty-eight patients (5.4%) had ERBB2 alterations, including 2.7% with ERBB2 gene amplification, 2.3% with ERBB2 mutation, and 0.4% with concurrent amplification and mutation. The prevalence of ERBB2 gene alterations was significantly higher in GBC (12.6%) than in ICC (2.2%) and EHC (7.5%; P < .001). In ERBB2-amplified tumors, the median fold change was 6.4 (range, 2.1 to 19.7), while in ERBB2-mutant tumors, the most frequent mutated domain was the extracellular domain (32%), with all mutations in this region involving the S310 codon. Frequent co-altered genes in this cohort were TP53 (54%), PIK3CA (21%), and CDKN2A (18%); KRAS amplification/mutation was found in 7% of patients. One patient with ERBB2-amplified EHC who enrolled in a basket trial (ClinicalTrials.gov identifier: NCT02675829 ) had a partial response to the human epidermal growth factor receptor 2–targeted antibody-drug conjugate ado-trastuzumab emtansine. CONCLUSION ERBB2 alterations are present in 5.4% of BTCs. When present, the degree of ERBB2 gene amplification is often high, and S310 codon mutations are the most common hotspot. These features, along with the presented case, support further development of human epidermal growth factor receptor 2–targeted therapy in ERBB2-mutant and/or -amplified BTC.

12100 Background: Rash associated with increased peripheral eosinophils develops in approximately 50% of metastatic breast cancer patients receiving alpelisib. Antihistamines and corticosteroids have limited benefit. Refractory rash may lead to decreased dose intensity and affect clinical outcome. Benralizumab is an anti-IL-5Rα chimeric monoclonal antibody that depletes peripheral eosinophils and has demonstrated benefit in eosinophilic asthma and hypereosinophilic syndrome. We investigate the efficacy and safety of benralizumab for the treatment of alpelisib rash. Methods: We performed a single-center, single-arm, prospective phase 2 study to evaluate the efficacy and safety of benralizumab in cancer patients who developed CTCAE grade 2/3 skin events resulting from immunotherapy or targeted therapies with absolute blood eosinophil counts of ≥300/mcl. While remaining on culprit drugs, patients were treated with benralizumab 30mg once every 4 weeks for the first 3 doses followed by once every 8 weeks for 3 additional doses (approved dosing for eosinophilic asthma). Primary endpoint was clinical response measured as reduction in CTCAE grade 2/3 skin event to grade ≤1 by week 4. Secondary endpoints were patient quality of life (QoL) measured by skindex16, safety data, need for supportive oral corticosteroids, and changes in cytokines and eosinophil biomarkers. This interim analysis focuses on patients with PIK3CA-mutant metastatic breast cancer receiving alpelisib. Results: Between September 16th 2020 and January 1st 2022, we enrolled 10 metastatic breast cancer patients with grade 2/3 rash attributed to alpelisib (5 pts with G3). All patients had a reduction of rash to grade ≤1 (n = 10, p < 0.0001), and a decrease in peripheral absolute eosinophils (mean 500/mcl to 0, p < 0.0001). Of these, 6 patients had been on prophylactic oral antihistamines and 2 had oral steroid coadministration. QoL significantly improved (Skindex16 mean score 58 to 16, p = 0.0001) and eosinophils in skin histology decreased per HPF (mean 6.25 to 0.25, n = 8, p = 0.2) by week 4. An increase in IL-5 > 600% and reduction IL-6 and TNF-α > 50% were reported by week 4 and 8. Grade 1/2 mucositis in 4 patients were reported as adverse events. Conclusions: Our findings suggest that benralizumab is safe and effective for the treatment of grade 2/3 rash with eosinophilia related to alpelisib in patients with breast cancer. A reduction in rash severity was evidenced in all patients, along with improved QoL. Larger controlled studies are in development to evaluate the efficacy of benralizumab for the prevention of alpelisib rash. Clinical trial information: NCT04552288.

8540 Background: The standard of care for patients with inoperable early stage or locally advanced NSCLC is definitive stereotactic body radiotherapy (SBRT) or conventional radiation therapy (RT) with systemic therapy. Circulating tumor DNA (ctDNA) testing can be used for the assessment of MRD and predict risk of recurrence. Few studies have prospectively evaluated MRD detection and ctDNA dynamics specifically among patients with early or locally advanced NSCLC receiving definitive RT. Methods: In a prospective clinical cohort of patients with stage I-III NSCLC (n = 17), serial plasma samples (n = 70) were collected before and after SBRT as well as before, during, and after conventional RT with or without concurrent systemic therapy and adjuvant durvalumab. Patients were followed-up for a median of 29 months (range: 4 to 54 months) with the last serial plasma collected at a median of 5 months from completion of RT (range: 1 – 26 months). A personalized, tumor-informed multiplex PCR assay (Signatera™ bespoke mPCR NGS assay) was used for the detection and quantification of ctDNA and tracked 16 tumor variants among 16 patients and 15 tumor variants in one patient. This study evaluated the prognostic value of ctDNA, correlating MRD status with clinical outcomes, in addition to ctDNA clearance kinetics during RT. Results: Among 17 patients with early-stage and locally advanced NSCLC, baseline ctDNA was detected in 82% of patients (14/17). Clinical progression was confirmed radiographically for 53% (9/17). All events of clinical progression were detectable by ctDNA (sensitivity 100%, 0.63 – 1.0), with a median lead-time of 5.5 months for MRD detection compared to radiographic disease progression. Durable ctDNA clearance was observed in 29% (5/17) of patients, all of whom then remained recurrence-free until the end of follow-up (median 12 months; specificity 100%, 95% CI 0.6 – 1.0). Transient ctDNA clearance was observed in 3 patients, and recurrent ctDNA was detected before or at the time of disease progression in all 3. ctDNA status after treatment at a single time point and longitudinally were highly predictive of disease recurrence (p < 0.0001). Conclusions: ctDNA detection is feasible for patients with stage I-III NSCLC undergoing definitive chemoradiation. and can serve as a powerful predictive biomarker for disease recurrence. High baseline detection rate is essential for feasibility of a ctDNA-based MRD assay. Residual detectable ctDNA represents a powerful predictive tool to identify patients who might benefit from intensification of adjuvant therapy following definitive RT.

1010 Background: We recently reported inferior outcomes to CDK4/6 inhibitors and endocrine therapy (CDK4/6i-ET) associated with germline BRCA2 (g BRCA2) in a cohort of estrogen receptor (ER) positive breast cancers. Co-occurrence of gBRCA2 with loss of heterozygosity (LOH) of neighboring RB1 was found to portend particularly poor outcomes. Here, we sought to define the effects of pre-treatment RB1 allelic copy number status on outcomes of CDK4/6i-ET and the likelihood of developing RB1 loss-of-function (LOF) mutations on CDK4/6i through the analysis of an expanded cohort of metastatic ER+ breast cancer patients. Methods: Patients who underwent sequencing on MSK-IMPACT from April 2014 to May 2021 were included. For every sample preceding CDK4/6i-ET, we performed FACETS to infer RB1 allele specific copy number, ploidy, tumor purity and fraction genome altered (FGA). Patients were categorized based on RB1 allelic status: HetLoss (total of one allelic copy), copy neutral LOH (CNLOH), other allelic imbalance including all other aneuploidy states, and diploid. Progression free survival (PFS) was assessed using univariate and multivariate Cox proportional hazard models adjusted for ET partner and FGA. Firth penalized logistic regression was used to study association of pre-treatment RB1 status with acquired RB1 LOF variants in paired post-CDK4/6i samples. Results: Of 2,630 potentially eligible patients, 279 patients had genomic sequencing performed prior to 1st line CDK4/6i-ET. Of these, 75 (26.8%) exhibited RB1 HetLoss, 39 (14.0%) had CNLOH of RB1, 111 (39.7%) exhibited diploid RB1 state, while 54 (19.4%) had other patterns of RB1 allelic imbalance. All non-diploid RB1 states were associated with significantly shortened PFS relative to diploid (univariate HetLoss HR: 2.05, 95% CI: 1.42, 2.97; CNLOH HR: 2.08, 95% CI: 1.32, 3.25; other imbalance HR: 1.70, 95% CI: 1.11, 2.58). Only HetLoss remained significant when adjusted for FGA (HR 1.61, 95% CI: 1.09, 2.38, p = 0.017). RB1 LOF was rare in pre-CDK4/6i tumors (< 1%); excluding these cases did not change our results. Of the 176 patients with paired pre- and post-CDK4/6i samples, only RB1 HetLoss in pre-CDK4/6i sample was significantly associated with development of RB1 LOF mutations in post-CDK4/6i sample (18.4%) as compared to diploid (4.2%, OR 4.25, 95% CI 1.02, 17.7, p = 0.047). These results indicate that tumors with one functional copy of RB1 are more likely to acquire RB1 LOF on CDK4/6i to achieve biallelic RB1 loss as a mechanism of CDK4/6i resistance. Conclusions: We demonstrate that LOH and allelic imbalance of RB1 are associated with shorter PFS on CDK4/6-ET. We postulate this may occur partly as a result of more frequent acquired RB1 LOF mutations under selective pressure of CDK4/6i. These data supports the implementation of more refined allele-specific copy number methods and identifies a high-risk population for escalated monitoring and treatment approaches.

1011 Background: The approvals of CDK4/6 inhibitors (CDK4/6i) and alpelisib (a PI3Ka inhibitor, PI3Ki) have overhauled early treatment of hormone positive aBC. While some clinical trials have investigated mechanisms of resistance to these drugs, their impact on tumor evolution requires further exploration. Here we use cfDNA to examine molecular changes pre- and post-CDK4/6i or PI3Ki treatment and use RWD to assess the impact of putative resistance alterations on response to treatment. Methods: Patients (pts) with aBC were identified via the Guardant INFORM database and included if they had a cfDNA test within 90 days prior to therapy initiation and/or 90 days after therapy discontinuation with CDK4/6i or PI3Ki. Pts with RB1 loss of function (LOF) alterations (alts) who received CDK4/6i and pts with PTEN LOF alts who received PI3Ki were separately identified, and these cohorts were matched 1:3 with a RB1/PTEN negative population respectively, by age (+/- 5 years), sex, year of cfDNA test, and line of therapy. Log-rank tests were used to assess differences in time to discontinuation (TTD) and time to next treatment (TTNT). Results: Differences in the frequencies of certain alts detected in pts pre- and post-CDK4/6i or PI3Ki treatment are shown (Table). Pts with RB1 LOF alts prior to the start of CDK4/6i had significantly worse TTD and numerically worse TTNT versus controls (TTD = 3 mos vs 4.7 mos, p=0.018; TTNT = 7.3 mos vs 8 mos, p=0.082). Pts with PTEN LOF alts prior to start of PI3Ki had no significant difference in TTD or TTNT versus controls (TTD = 4.1 mos vs 4.1 mos, p=0.92; TTNT = 7.4 mos vs 7 mos, p=0.32). Notably, 54% of pts receiving CDK4/6i and 84% of pts receiving PI3Ki were on their third or later line of therapy. Conclusions: Using cfDNA, we were able to further characterize the resistance landscape of both CDK4/6i and PI3Ki, and identified specific ESR1, RB1 and PTEN alterations that appear likely to occur under the pressure of therapy. Our real-world analysis examining RB1 LOF alts added further evidence to suggest it may be both a primary and acquired resistance mechanism to CDK4/6i. As a non-invasive alternative to tissue biopsies, this data further illustrates that cfDNA can provide unique insight into tumor evolution and disease progression in the aBC setting. [Table: see text]

3145 Background: Clonal hematopoiesis (CH) has well established associations with adverse clinical outcomes including all-cause mortality, cardiovascular disease, and progression to hematologic malignancy. The presence of CH has also been demonstrated to adversely impact survival from non-hematologic cancers, however whether CH may modulate response to radiation therapy (RT) in solid tumors is not known. Here we investigate the potential impact of CH mutations on radiation outcomes. Methods: We analyzed data from two previously well annotated cohorts of patients with tumors harboring somatic ATM mutations (n = 358) and FAT1 mutations (n = 365) who received RT and underwent prospective tumor and matched WBC sequencing utilizing the MSK-IMPACT assay. CH variants were detected in the blood samples utilizing a well-validated variant detection and filtration pipeline. Given that pathogenic mutations in ATM have been shown to be strongly associated with improved response to RT, these patients were excluded to avoid confounding. Additionally, patients with blood sampling for CH assessment that occurred more than 6 months after RT were excluded to address the possibility of therapy-related CH. We compared outcomes including irradiated tumor progression in patients with and without CH. Results: The final analysis consisted of 412 patients who underwent 811 total courses of radiation. A wide spectrum of solid tumor types were represented, most commonly non-small cell lung cancer (32.5%) and breast cancer (11.9%). A total of 161 patients (39.0%) had CH, with the most commonly mutated genes being DNMT3A (25.6%), PPM1D (6.2%), TET2 (5.8%), and TP53 (5.0%), consistent with prior studies of CH. Patients with CH were older at blood sample collection (67.6 vs 60.2 years, p < 0.001), reflecting an expected increase in CH burden with age. Fine Gray competing risks analysis, with death treated as a competing event and with clustering around patient identifier, showed no difference in irradiated tumor progression between patients with and without CH (HR 1.03, 95% CI 0.69 – 1.53, p = 0.896). Similarly, subanalyses by CH variant allele frequency and putative CH-driver mutations did not reveal an association between CH and response to RT. A hypothesis generating subgroup analysis by common cancer types, however, suggested that CH was associated with increased risk of progression post-radiation in prostate (HR 4.68, 1.14 – 19.1) and thyroid (HR 3.13, 1.55 – 6.34) cancer cohorts, warranting further investigation. Conclusions: We found no difference in irradiated tumor progression among patients who did and did not have CH. There may be an association between CH and poor radiation outcomes in certain cancer types, and further studies are needed to clarify the specific clinical and genomic factors that may influence radiation response.

9067 Background: Small cell lung cancer (SCLC) is an aggressive disease, characterized by inevitable chemotherapy resistance and rapid progression. We hypothesized that circulating tumor DNA (ctDNA) analysis can rapidly identify sensitivity to platinum-based therapy. Methods: Patients with SCLC at Memorial Sloan Kettering Cancer Center underwent serial plasma collections, including prior to the start of treatment and prior to Cycle 2 Day 1 of therapy (C2D1). Tumor mutations were identified from pre-treatment biopsies by MSK-IMPACT and/or pre-treatment plasma by CAncer Personalized Profiling by deep Sequencing (CAPP-Seq). Median variant allele fraction (VAF) of all mutations was monitored on subsequent blood draws using CAPP-Seq. Progression free survival (PFS) was measured from the time of first pre-treatment blood draw. Results: Plasma was collected from 19 patients treated with carboplatin and etoposide, including three who received concurrent atezolizumab. Seven were female, and mean age was 64.5 years. ctDNA was detected in 17 patients (89%), including in the two patients in our series with limited stage disease. The most common mutations were in TP53 and RB1 in 14 and 6 patients, respectively. Fourteen patients had available plasma at C2D1. At baseline prior to treatment, median VAF did not differ significantly between radiologic responders and non-responders (9.4% versus 30.3%, p = 0.35). After one cycle of chemotherapy, the VAF percent decrease was significantly more in responders versus non-responders (-96.9% versus -10.3%, p < 0.001). Median VAF was therefore significantly lower by C2D1 in patients who responded compared to non-responders (0.51% versus 27.2%, p = 0.02). Those who ultimately responded to therapy all had a > 2 fold decrease in VAF by C2D1. With a median follow-up of 180 days, PFS was significantly longer in patients with > 2 fold decrease in VAF by C2D1 (6.4 versus 1.9 months, log rank p < 0.001). Conclusions: A 2-fold decrease in plasma VAF by C2D1 predicted platinum-sensitivity in SCLC and was associated with longer PFS. ctDNA may permit early assessment of benefit and expedite alternative treatment options for those without significant decrease in median VAF after one cycle of therapy.

3529 Background: Circulating cell-free DNA (cfDNA) next-generation sequencing (NGS) is a promising strategy for non-invasive molecular profiling of cancers. MSK-ACCESS (Analysis of Circulating cfDNA to Evaluate Somatic Status) is a hybridization-capture targeted NGS assay that detects somatic variants in select exons of 129 genes with matched white blood cell sequencing. We present the initial clinical experience with MSK-ACCESS among patients with advanced non-small cell lung cancer (NSCLC). Methods: Patients with stage IV NSCLC underwent prospective MSK-ACCESS testing at initial diagnosis or progression of disease on targeted therapy between June 2019 and January 2020. A subset of patients had matched tissue-based NGS testing with the MSK-IMPACT 468 gene assay. We assessed oncogenic driver detection, turnaround time, plasma-tissue concordance, and matching to therapy. National Comprehensive-Cancer Network designated driver alterations were included in evaluation of tissue-plasma concordance (EGFR, ALK, KRAS, MET, RET, BRAF, HER2, ROS1, NTRK). Turnaround time was compared by a two-sided Wilcoxon signed-rank test. Results: A total of 201 patients with NSCLC had MSK-ACCESS testing at initial diagnosis (n = 79) or following progression of disease (n = 122). The median turn-around-time from plasma collection to MSK-ACCESS report was 16 days (range: 9 – 36 days) compared to 19 days from lab receipt of tissue to report (range: 12 – 57) for MSK-IMPACT (p < 0.001). Among patients with a driver detected on MSK-ACCESS, 100% (92/92) had an identical driver detected on MSK-IMPACT. Among patients with a driver detected on MSK-IMPACT, 75% (92/123) had an identical driver detected on MSK-ACCESS. This rate was similar among patients who were treatment-naive (74%; 64/86) and had disease progression (76%, 28/37) at the time of MSK-ACCESS. MSK-ACCESS identified driver alterations that directly guided first-line targeted therapy (n = 18) with response in all patients with available radiographic follow-up (n = 10), including a patient without confirmatory tissue testing. MSK-ACCESS identified resistance alterations among patients with disease progression including EGFR T790M, EGFR C797S, ROS1 G2032R, as well as a BRAF fusion. Conclusions: MSK-ACCESS successfully identified driver alterations with high concordance to tissue-based testing, directly guided patients to therapy with clinical responses, and detected known and novel resistance mechanisms. This assay warrants further clinical development to guide and facilitate precision oncology.

1063 Background: Rash develops in approximately 50% of breast cancer patients receiving alpelisib, often requiring dose modifications. Herein, we describe the characteristics of alpelisib-related dermatologic adverse events (dAEs). Methods: A single center retrospective analysis was conducted via review of electronic medical records. We collected clinical, laboratory and management data relevant to patients treated with alpelisib for advanced breast cancer under four different randomized clinical trials or post approval by regulatory agencies from 6/1/2013 to 7/31/2019. Type and severity of dAEs was recorded using the Common Terminology Criteria for Adverse Events version 5.0 (CTCAE v5.0). Results: A total of 102 patients (mean age 56 years, range 27-83) receiving alpelisib from 200 to 350 mg daily, most frequently in combination with endocrine therapy (79, 77.5%) were included. We identified 41 (40.2%) patients with all-grade rash [CTCAE grade 1/2 = 22 (21.6%); CTCAE grade 3 = 19, (18.6%)] distributed primarily along the trunk (18, 78%) and developing, on average, within 12.8 +/- 1.5 days of treatment initiation (n = 38). Mean duration of rash was 7.1 +/- 3.8 days; and no grade 4 dAEs were observed. Of 29 patients with documented morphology of alpelisib-related dAEs, the majority (26, 89.7%) had maculopapular rash. Thirteen (68%) of 19 patients with any-grade rash and report of any associated symptoms had pruritus (7, 36%) or burning pain (6, 32%). All-grade dAEs correlated with an increase in serum eosinophils from 2.7% to 4.4% (p < 0.05), and prophylaxis with non-sedating antihistamines (n = 43) was correlated with a reduction of grade 1/2 rash onset (OR 0.39, p = 0.09). Sixteen (84.2%) of 19 patients with grade 3 dAEs had interruption of alpelisib, followed by management with antihistamines, topical and/or systemic corticosteroids. We did not observe rash recurrence in 12 (75%) of these 16 patients who re-initiated therapy; and the majority (9, 56.3%) were re-challenged without a dose reduction. Conclusions: Pruritus and increased blood eosinophils occur with maculopapular rash within the first two weeks of initiating alpelisib and persists for approximately seven days. To reduce onset of grade 1/2 rash, non-sedating antihistamines (i.e. cetirizine) are recommended during the first eight weeks. While grade 3 rash leads to interruption of alpelisib, dermatologic improvement is evident with systemic corticosteroids; and most patients can resume therapy at a maintained or reduced dose upon re-challenge.

1535 Background: Clonal hematopoiesis (CH) is most commonly associated with mutations in genes with a known or putative role in leukemia and with an increased risk for hematologic malignancies. Nearly 50% of primary breast cancers contain leukocytes with CH mutations and CH mutations can be found in breast cancer (BC) tumor infiltrating leukocytes years prior to the development of secondary leukemia. It is currently not known whether CH is more prevalent among BC patients with both germline pathogenic variants and likely pathogenic variants (collectively referred to as PV) in cancer predisposition genes. Here we evaluated the relationship of CH to PV in 546 BC patients (pts). Methods: 546 BC pts underwent targeted capture sequencing of tumor and peripheral blood (PB) samples using MSK-IMPACT. Sequencing results from PB were used for identifying PV affecting up to 89 cancer predisposition genes and somatic mutations in 15 genes commonly associated with CH using previously validated methods. All pts consented via an IRB-approved protocol. Results: The majority (82.3%) of pts had non-metastatic disease. 59.7% of the 546 pts received chemotherapy and 42.7% received radiation. Of the 546 pts, 90 patients had germline PV in a cancer predisposition gene for a total of 98 germline PV identified (8 patients had 2 germline PV). Mutations in DNMT3A followed by TET2 were the most common CH mutations identified. CHEK2 PV were statistically significantly associated with CH in a multivariate analysis after controlling for age, prior chemotherapy and radiation therapy (OR: 3.94, CI: 1.51-10.26, p-value ≤0.005). Age by decade was significantly associated with the presence of CH in that model (OR: 1.93, CI: 1.53-2.42, p value ≤0.001). Of the 8 patients with CHEK2 germline PV and CH, the most common mutated CH genes were DNMT3A (4 pts), followed by TET2 (2 pts), although our sample size is too small to test for significant enrichment for any specific CH gene, type of CH mutation, or trinucleotide context. Further, CH was not associated with germline PV in other cancer predisposition genes. Conclusions: Our findings provide evidence of an association between CHEK2 germline PV with CH in BC pts, which will be further tested in an expanded cohort. If this association is confirmed, it might have theoretical and practical implications, including cancer prevention strategies.

6001 Background: SGCs are rare tumors with no approved therapy for metastatic disease. HER2 amplification occurs in 8% among all SGC histologies, and 25-33% of the aggressive salivary duct carcinoma (SDC) histologic subtype. We hypothesized that ado-trastuzumab emtansine, a HER2 targeted antibody drug conjugate, may be clinically active in these patients. Methods: A cohort of patients with HER2 amplified SGCs were enrolled into a multi-histology basket trial of ado-trastuzumab emtansine, treated at 3.6mg/kg IV every 3 weeks. The primary endpoint was overall response rate (ORR) by RECIST v1.1 or PERCIST. A Simon two-stage optimal design was applied with type I error rate under 2.7%, power of 89%, H0 10%, H1 40%; H0 will be rejected if 6 or more responses are observed in 24 patients. Other endpoints include duration of response (DOR), progression-free survival (PFS), and toxicity. HER2 amplification was identified by next generation sequencing (NGS), and tumors were subsequently tested by fluorescence in situ hybridization (FISH) and immunohistochemistry (IHC). Fluorescence lifetime imaging microscopy - Förster resonance energy transfer (FLIM-FRET) assessed the propensity for HER2-HER3 heterodimerization, which leads to receptor internalization. Results: 10 patients with HER2 amplified SGCs were treated. The median age was 65 (range 36-90 years), 90% were male. The median lines of prior systemic therapy was 2 (range 0-3). ORR was 90% (9/10, 95% CI 56-100%) including 5 complete responses after prior trastuzumab, pertuzumab and anti-androgen therapy. After a median follow up period of 12 months (range 4-20 months), median DOR (range 2-19+) and median PFS (95% CI 4–22+ months) were not reached. Toxicities included grade 1 or 2 infusion reaction, thrombocytopenia and transaminitis; there were no treatment related deaths. HER2 amplification by NGS (fold change 2.8 to 22.8) correlated with HER2/CEP17≥2 by FISH (8/8 tested) or IHC3+ (10/10 tested). FLIM-FRET tested positive in 3/3. Conclusions: Ado-trastuzumab emtansine is highly efficacious in patients with HER2 amplified SGCs as identified by NGS. This study has met its primary endpoint, and cohort expansion is warranted to confirm these results. Clinical trial information: NCT02675829.

1053 Background: Sequential retreatment with endocrine therapy (ET) has been the clinical paradigm for ER+ MBC due to persistent dependence on hormone signaling. Recently CDK4/6i + ET have improved PFS and are routinely utilized in the first/second- line setting. Whether this paradigm of sequential retreatment holds for CDK 4/6i is unknown. To evaluate the potential benefit of CDK4/6i re-treatment we conducted this retrospective analysis. Methods: We identified ER+/HER2- MBC pts treated with ≥ 2 lines of CDK4/6i at our institution between 2015-2018. We categorized pts based on reason for discontinuation of their first CDK4/6i: cohort 1 – switch to alternate CDK4/6i due to toxicity; cohort 2 – retreatment with same CDK4/6i beyond progression with change of ET and cohort 3- switch to alternate CDK4/6i as monotherapy or with same or another ET. We analyzed pt demographics, imaging reports and time to subsequent therapy (TTST) for every CDK4/6i line for each cohort. If a pt received > 2 lines of CDK4/6i, then that pt was evaluated for every CDK4/6i exposure. Results: 135 pts received ≥ 2 lines of CDK4/6i treatment (Tx). Cohorts 1, 2 and 3 had 23, 43 pts and 84 pts respectively. In Cohort 2, 95% of pts received 2 subsequent CDK 4/6i + ET Tx; 56% had the second CDK4/6i in second-line met setting. TTST1 (1st CDK4/6i Tx) was 9.6m (95% CI 4.9 - 11 m), TTST2 (second CDK4/6i Tx) was 4.5m (95% CI 3.3 – 7.6 m) and 35% had TTST2 ≥ 24 weeks. For Cohort 3, 48% were retreated with a different CDK 4/6i in ≥ fifth-line. 51% received 2 subsequent CDK 4/6i Tx with 18% in second-line met setting. TTST1 was 9.6 m (95% CI 5.9 – 12 m), TTST2 was 4.4 m (95% CI 3.8 – 5.9 m) and 29% had TTST2 ≥ 24 weeks. In cohort 3, 29% (n=24) pts had PD as best response at the time of first CDK4/6i exposure but 29% (7/24) achieved a radiologic response to their second CDK4/6i Tx. Pts had tumor sequencing using MSK-IMPACT which will be correlated with TTST. Conclusions: This large single institution retrospective analysis suggests that retreatment with a CDK4/6i regimen should be evaluated in prospective trials. Additionally, despite PD as best response with the first CDK4/6i (palbociclib/ribociclib) regimen, a subset of pts had radiologic response to a subsequent abemaciclib-containing regimen, which is an hypothesis generating observation.

1009 Background: Genomic profiling of ER+ metastatic breast cancer (MBC) has revealed highly prevalent genomic alterations (e.g. ESR1, NF1, ERBB2) associated with exposure to antiestrogen therapy and endocrine resistance. It is not known whether any such acquired genomic alterations are observed after exposure to now standard CDK4/6 inhibitors (CDK4/6i). Methods: To identify genomic alterations associated with acquired resistance to CDK4/6i + antiestrogen combinations, we prospectively performed tumor and matched normal sequencing on 1059 ER+ breast cancers from 845 MBC patients collected prior to (n = 838) or post-treatment with CDK4/6i (n = 221), including 110 pre- and post-treatment pairs. We performed gene enrichment analyses to identify the oncogenic mutations and copy number alterations that were more frequent in post-CDK4/6i samples compared to CDK4/6i-naïve samples and further compared these results to those of post-hormone alone therapy datasets. Results: The post-CDK4/6i samples were collected following exposure to CDK4/6i plus aromatase inhibitors (51%), plus fulvestrant (28%), or multiple/other (21%). Along with alterations previously associated with resistance to hormonal therapy alone, our analysis identified multiple genes to be specifically enriched in the post-CDK4/6i tumors. Among these, loss-of-function alterations in RB1 were significantly enriched in the post-CDK4/6i compared to CDK4/6i-naïve samples (7.9% vs. 2.7%, p-value = 1.5e-5). The majority of the RB1 mutations in the post-CDK4/6i tumors had loss of heterozygosity (LOH), while LOH was uncommon in the CDK4/6i-naive tumors. Additionally, multiple alterations in effectors of PI3K/AKT signaling (excluding PIK3CA), cell cycle (such as CDKN2A loss) and in Hippo signaling were more frequent in the post-CDK4/6i tumors. Conclusions: This large clinico-genomic analysis of the post- and pre-CDK4/6i MBCs reveals multiple genomic lesions to be enriched after exposure to CDK4/6i, highlighting therapy-related genomic evolution as a recurrent phenomenon and identifying unique genomic subsets that have strong potential to alter the benefit of subsequent lines of therapies.

1089 Background: Microsatellite instability (MSI) and mismatch repair deficiency (MMRd) can occur sporadically in solid tumors across histologic subtype and predict clinical response to immune checkpoint blockade (ICB). MSIsensor uses next generation sequencing to report the percentage of unstable microsatellites in a given tumor sample as a score. While scores of ≥10 and < 3 have excellent sensitivity/specificity for MSI/MMRd, intermediate scores of 3-10 are of unknown clinical significance, and may include patients with MSI-like genomic signatures who would be overlooked by historical assays. Methods: MSIsensor testing was performed on primary or metastatic tumor specimens from 2,371 patients with breast cancer. Scores of 3-10 were considered MSI-Indeterminate (MSI-I). Concurrent somatic mutations were determined using the MSK-IMPACT assay. Results: Of the 2,371 patients tested, 147 were found to have MSI-I tumors. Ninety-nine (67.3%) were metastatic specimens, and 48 (32.7%) were primary specimens. Of the MSI-I patients, 64 (43.5%) were estrogen receptor positive, 57 (38.8%) were triple negative, and 24 (16.3%) were HER-2 amplified. Somatic TP53 mutations were detected in 109 (74.1%) MSI-I patients, including 54 (94.7%) triple negative patients. Fourteen MSI-I patients were found to have somatic mutations in MMR genes ( MLH1, MSH2, MSH6, PMS2 or EPCAM), but no germline mutations were found. Of the 99 patients with MSI-I metastatic disease, 15 received ICB as part of their clinical care, exhibiting a median progression-free-survival (PFS) of 3 months. Of the 4 patients who demonstrated a PFS of > 6 months on ICB, 2 were found to have somatic mutations in MMR genes. Conclusions: We introduce a subpopulation of breast cancer patients whose tumors exhibit genomic signs of microsatellite instability without meeting the classic criteria for MSI/MMRd as previously defined in colon and endometrial cancers. A subset of these patients may exhibit the properties of MSI/MMRd at the epigenetic and protein-expression levels, and thus potentially benefit from ICB.

e24181Background: Targeting HER2 alterations including mutation, amplification and protein overexpression has been an active area of new drug development. However, the relationships between these H...

LBA11516 Background: ctDNA assays can noninvasively assess tumor burden and biology by identifying tumor-derived somatic alterations. For broad applicability, including early cancer detection, an unprecedented high-intensity approach (ultra-deep sequencing of plasma cell-free DNA (cfDNA) with broad genomic coverage) is needed to address intra-patient and population-level heterogeneity. We present initial results with this approach in patients (pts) with metastatic breast (BC), non-small cell lung (NSCLC), and castration-resistant prostate cancer (CRPC). Methods: Blood and tissue were prospectively collected w/in 6 wks with no intervening therapy change from pts with de novo or progressive cancer. cfDNA and white blood cell (WBC) genomic DNA from each pt were sequenced with a 508-gene panel (2 Mb; >60,000X raw depth). cfDNA variant calling used molecular barcoding for error correction and filtering for WBC variants. Tissue was sequenced using the MSK-IMPACT assay (410 genes, 1.4 Mb, >500X depth) blinded to plasma/WBC sequencing. Variants were classified as clonal or subclonal based on tumor sequencing in BC and NSCLC. Results: Of 161 eligible pts, 124 (39 BC, 41 NSCLC, and 44 CRPC) were evaluable for concordance. In tissue, 864 variants were detected across the 3 tumor types, with 627 (73%) also detected in plasma: single nucleotide variants/indels - 75%, fusions - 67%, and copy number alterations - 58%. In 90% of pts, at least 1 of the variants detected in tumor tissue was also detected in plasma: BC - 97%, NSCLC - 85%, CRPC - 84%. Most actionable mutations detected in tissue were also detected in plasma (54/71, 76%; SNVs only: 28/31, 90%). A subset of driver mutations (eg. in ESR1, PIK3CA, ERBB2, EGFR) were observed in plasma but not tissue. Clonal variants in tissue were more likely to be detected in plasma than subclonal variants (p

11526 Background: A large fraction of cfDNA fragments are derived from hematopoietic sources. Somatic alterations in cfDNA can be tumor-derived but also could represent somatic changes associated with clonal hematopoiesis. We performed deep sequencing of both plasma cfDNA and matched white blood cell (WBC) genomic DNA (gDNA) to determine the contribution of clonal hematopoiesis to the variants observed in cfDNA. Four cohorts were investigated: metastatic breast (BC), non-small cell lung (NSCLC), castration-resistant prostate cancer (CRPC), and non-cancer participants (pts). Methods: Metastatic cancer pts with de novo or progressive disease were prospectively enrolled. Non-cancer pts were blood bank donors. Plasma cfDNA and matched WBC gDNA were sequenced using a targeted 508-gene panel (2 Mb) to > 60,000X raw depth. Variant calling used a novel pipeline that employed molecular barcoding for error suppression followed by de novo assembly and graph-based variant calling. Results: Of 151 metastatic cancer pts (48 BC, 49 NSCLC, 54 CRPC), median age was 64 (30-87) with 53% female and 33% treatment naive. Of 47 non-cancer pts, median age was 61 (20-78) with 51% female. Analysis of cfDNA identified 1072 variants (AF > 0.1%, > 2 mutant reads, passing bioinformatic quality filters) which were also detected in WBC gDNA as non-germline ( < 35% allele frequency [AF]) non-synonymous variants. For these cfDNA variants, AF ranged from 0.1-14.4% and correlated with AF in WBC gDNA (r2 = 0.47, p < 0.001). Mutated genes were consistent with clonal hematopoiesis, with the most frequently mutated genes being DNMT3A, TET2, PPM1D, and TP53 (215, 77, 45, and 36 variants, respectively). For both cancer and non-cancer pts (age > 45), median number of overlapping variants was 5 per pt (range 0-22). The number of WBC gDNA and cfDNA variants per individual was positively associated with age (p < 0.001) in both cancer and non-cancer pts (interaction p = 0.08). Conclusions: Somatic cfDNA variants are frequently derived from clonal hematopoiesis and increase with age. Accurate assessment of somatic alterations in cfDNA should account for this phenomenon to distinguish between tumor-derived and WBC-derived variants.

e23237Background: Traditional methods for HER2 assessment such as IHC and FISH are semi-quantitative and often limited by reproducibility issues. This may be a problem in tumors with borderline HER...

11500Background: E17K mutation of AKT1 is a recurrent, oncogenic event in solid tumors where treatment (Tx) with AKT inhibition has shown promise. Little is known about the significance of genomic ...

604 Background: Approximately 15-20% of invasive breast carcinomas (BCA) show HER2/ERBB2 gene amplification which determines eligibility for treatment by targeted therapy. Traditionally, HER2 statu...

TO THE EDITOR: In their retrospective cohort study, den Exter et al found no significant differences in recurrent pulmonary emboli (PE), bleeding, or death when comparing patients with cancer who had incidental PE identified on routine staging computed tomography scans and patients with cancer who were diagnosed with PE on the basis of the presence of symptoms. The authors are the first, to our knowledge, to use as a comparator symptomatic patients with dedicated scans demonstrating PE to assess the clinical significance of incidental PE. However, we previously demonstrated that 75% of patients with incidental PE, which we referred to as unsuspected PE (UPE), were actually symptomatic. Variability in the treating oncologists’ index of suspicion for PE may significantly influence whether patients with cancer are referred for radiologic studies to rule out PE, even in the presence of symptoms. Thus, patients who were included in the incidental PE group in the study by den Exter et al may have, at a different institution or with a different physician, been referred for a dedicated study to rule out PE. In a follow-up study that included 70 patients with UPE and 137 ageand stage-matched patients with similar cancers but without PE, we found that UPE adversely affected survival. Notably, patients with symptomatic UPE had significantly shorter survival than patients with asymptomatic clots (unpublished analysis, Fig 1). In fact, respiratory symptoms and fatigue have been independently associated with shorter survival in patients with cancer. Symptoms should thus be considered a critical variable in any survival analysis of patients with cancer and suspected or unsuspected PE. As research and treatment recommendations arise from these and related data, we caution against the use of the words incidental and asymptomatic as interchangeable terms. We recommend that symptoms be clearly reported when possible and that the term unsuspected rather than asymptomatic or incidental be used because of the assumptions and potential bias that are associated with the latter terms.

1556 Background: Light at night as in shift work suppresses nocturnal secretion of melatonin, a pineal hormone with oncostatic properties. Several studies have associated night shift work with higher risk of cancer, leading WHO in 2007 to classify rotating night shift work as “probably carcinogenic”. We conducted one of the most comprehensive studies, to date, to evaluate the effects of light and night shift work on melatonin measurements in the field. Methods: Study participants were 130 active nurses (84 current rotating night shift workers and 46 day shift workers) participating in NHS2. Each nurse wore a head-mounted light- and accelerometer for a 3-day study period, during which each spontaneous urine was collected for repeated urinary 6-sulfatoxymelatonin (melatonin) measurements. In addition, nurses were asked to fill out paper questionnaires and diaries. We used mixed models to evaluate the influence of light, activity and night shift work on urinary melatonin level adjusting, for age, lifestyle, and occupational history. We log-transformed main variables and report geometric means (GM [standard deviation]). Results: Greater levels of light were associated with lower melatonin (P < 0.0001), independent of activity level. An increase in light intensity from 10 to 100 lux was associated with a 12% decrease in geometric mean of melatonin level; however, this inverse association was only significant at night (Ptrend = 0.01). At night, each hour increase in exposure to ≥ 20 lux light lowered melatonin level by 5.7% (Ptrend < 0.0001). A single night shift affected the circadian system by lowering melatonin peak by 22% (day shift: GM = 17.57 [2.73]; night shift: GM = 13.64 [2.54]) and induced a phase shift (PS) of 0.9 hours, -changes that reset to normal by the next day. Two consecutive night shifts had a similar effect as a single shift. However, the effect was worse after three consecutive night shifts (GM = 10.11 [2.77]; PS = 2.2 hours). Conclusions: We found significant inverse associations of intensity and duration of exposure to light at night with urinary melatonin, independent of activity level. Three consecutive night shifts affected the circadian system more strongly than two consecutive, or a single night shift.