Search

PURPOSE To inform precision oncology, methods are needed to use electronic health records (EHRs) to identify patients with cancer who are experiencing clinical inflection points, consistent with worsening prognosis or a high propensity to change treatment, at specific time points. Such patients might benefit from real-time screening for clinical trials. METHODS Using serial unstructured imaging reports for patients with solid tumors or lymphoma participating in a single-institution precision medicine study, we trained a deep neural network natural language processing (NLP) model to dynamically predict patients' prognoses and propensity to start new palliative-intent systemic therapy within 30 days. Model performance was evaluated using Harrell's c-index (for prognosis) and the area under the receiver operating characteristic curve (AUC; for new treatment and new clinical trial enrollment). Associations between model outputs and manual annotations of cancer progression were also evaluated using the AUC. RESULTS A deep NLP model was trained and evaluated using 302,688 imaging reports for 16,780 patients. In a held-out test set of 34,770 reports for 1,952 additional patients, the model predicted survival with a c-index of 0.76 and initiation of new treatment with an AUC of 0.77. Model-generated prognostic scores were associated with annotation of cancer progression on the basis of manual EHR review (n = 1,488 reports for 110 patients with lung or colorectal cancer) with an AUC of 0.78, and predictions of new treatment were associated with annotation of cancer progression on the basis of manual EHR review with an AUC of 0.84. CONCLUSION Training a deep NLP model to identify clinical inflection points among patients with cancer is feasible. This approach could identify patients who may benefit from real-time targeted clinical trial screening interventions at health system scale.

Although papillary thyroid carcinoma (PTC) is the most frequent endocrine tumor with a generally excellent prognosis, a patient developed a clinically aggressive PTC eleven years after receiving platinum chemotherapy for ovarian endometrioid adenocarcinoma. Germline and somatic analyses of multi-temporal and multi-regional molecular profiles indicated that ovarian and thyroid tumors did not share common genetic alterations. PTC tumors had driver events associated with aggressive PTC behavior, an RBPMS-NTRK3 fusion and a TERT promoter mutation. Spatial and temporal genomic heterogeneity analysis indicated a close link between anatomical locations and molecular patterns of PTC. Mutational signature analyses demonstrated a molecular footprint of platinum exposure, and that aggressive molecular drivers of PTC were linked to prior platinum-associated mutagenesis. This case provides a direct association between platinum chemotherapy exposure and secondary solid tumor evolution, in specific aggressive thyroid carcinoma, and suggests that uniform clinical assessments for secondary PTC after platinum chemotherapy may warrant further evaluation.

PURPOSE Evidence-based somatic and germline sequencing has transformed cancer care and improves patient outcomes. However, patients’ low genetic literacy and misunderstanding of their own genomic results poses a threat to the realization of precision oncology. To optimize patient genomic comprehension, we developed a Web-based, patient-directed, genomic sequencing education and return-of-results tool, HOPE-Genomics. METHODS The HOPE-Genomics prototype included somatic and germline sequencing results, embedded multimedia genomic education, and interactive features (eg, request for genetic counseling). Between January and April 2018, we elicited feedback on tool usability and comprehensiveness through participant surveys, 4 focus groups of patients with cancer and their family members, and 3 provider focus groups (comprising 8 patients, 5 family members, and 19 providers). RESULTS We identified themes in patient/family tool-related responses, including the desire to view a patient-friendly report, a desire to receive multiple types of genomic information (eg, prognostic and uncertain), high acceptability of report content, and interest in tool-enabled access to genetic counseling. Major themes from the clinician focus groups included believing the tool could help patients formulate questions and facilitate patients’ communication of results to family members. However, there were diverse responses from all participants in terms of tool implementation (ie, timing and nature of report release). Some participants preferred report release before meeting with the provider, and others preferred it during the appointment. Additionally, some clinicians were concerned about providing prognostic and treatment information through the tool. CONCLUSION There was high acceptability and interest from patients, family members, and providers in a patient-directed genomics report. Future work will determine whether direct-to-patient reporting of genomic results improves patient knowledge, care engagement, and compliance with genomically guided interventions.

PURPOSE Smoldering multiple myeloma (SMM) is a precursor condition of multiple myeloma (MM) with a 10% annual risk of progression. Various prognostic models exist for risk stratification; however, those are based on solely clinical metrics. The discovery of genomic alterations that underlie disease progression to MM could improve current risk models. METHODS We used next-generation sequencing to study 214 patients with SMM. We performed whole-exome sequencing on 166 tumors, including 5 with serial samples, and deep targeted sequencing on 48 tumors. RESULTS We observed that most of the genetic alterations necessary for progression have already been acquired by the diagnosis of SMM. Particularly, we found that alterations of the mitogen-activated protein kinase pathway ( KRAS and NRAS single nucleotide variants [SNVs]), the DNA repair pathway (deletion 17p, TP53, and ATM SNVs), and MYC (translocations or copy number variations) were all independent risk factors of progression after accounting for clinical risk staging. We validated these findings in an external SMM cohort by showing that patients who have any of these three features have a higher risk of progressing to MM. Moreover, APOBEC associated mutations were enriched in patients who progressed and were associated with a shorter time to progression in our cohort. CONCLUSION SMM is a genetically mature entity whereby most driver genetic alterations have already occurred, which suggests the existence of a right-skewed model of genetic evolution from monoclonal gammopathy of undetermined significance to MM. We identified and externally validated genomic predictors of progression that could distinguish patients at high risk of progression to MM and, thus, improve on the precision of current clinical models.

The emergence of immune checkpoint blockade therapies over the last decade has transformed cancer treatment in a wide range of tumor types. Unprecedented and durable clinical responses in difficult-to-treat cancer histologies have been observed. However, despite these promising long-term responses, the majority of patients fail to respond to immune checkpoint blockade, demonstrating primary resistance. Additionally, many of those who initially respond to treatment eventually experience relapse secondary to acquired resistance. Both primary and acquired resistance are a result of complex and constantly evolving interactions between cancer cells and the immune system. Many mechanisms of resistance have been characterized to date, and more continue to be uncovered. By elucidating and targeting mechanisms of resistance, treatments can be tailored to improve clinical outcomes. This review will discuss the landscape of immune checkpoint blockade response data, different resistance mechanisms, and potential therapeutic strategies to overcome resistance.

Purpose FGFR3-TACC3 (fibroblast growth factor receptor 3–transforming acidic coiled coil-containing protein 3) fusions have recently been identified as driver mutations that lead to the activation of FGFR3 in bladder cancer and other tumor types and are associated with sensitivity to tyrosine kinase inhibitors. We examined the clinical and molecular characteristics of patients with FGFR3-TACC3 fusions and hypothesized that they are enriched in a subset of patients with bladder cancer. Materials and Methods We correlated somatic FGFR3-TACC3 fusions with clinical and molecular features in two cohorts of patients with bladder cancer. The first cohort consisted of the muscle-invasive bladder cancer (MIBC) data set (n = 412) from The Cancer Genome Atlas. The second cohort consisted of patients with MIBC or high-grade non-MIBC at the Dana-Farber Cancer Institute that had targeted capture sequencing of a selected panel of cancer genes (n = 356). All statistical tests were two sided. The clinical response of one patient with FGFR3-TACC3 bladder cancer to an FGFR3 inhibitor was investigated. Results Overall, 751 patients with high-grade bladder cancer without FGFR3-TACC3 fusions and 17 with FGFR3-TACC3 fusions were identified in the pooled analysis of the data sets from The Cancer Genome Atlas and the Dana-Farber Cancer Institute. FGFR3-TACC3 fusions were enriched in patients age ≤ 50 years versus age 51 to 65 years versus those older than 65 years (pooled, P = .002), and were observed in four (12%) of 33 patients age ≤ 50 years in the pooled analysis. Similarly, FGFR3-TACC3 fusions were significantly more common in Asians (13%) compared with African Americans (4%) and whites (2%; pooled, P < .001), as well as in never smokers (5.6%) compared with ever smokers (1.1%; pooled, P < .001). One patient with the fusion who was treated with an FGFR3 inhibitor achieved complete remission for 10 months. Conclusion Clinical testing to identify FGFR3 fusions should be prioritized for patients with bladder cancer who are younger, never smokers, and/or Asian.

Objective Liquid biopsies are being rapidly used in adult cancers as new biomarkers of disease. Circulating tumor DNA (ctDNA) levels have been reported to be proportional to disease burden, correlate with treatment response, and predict relapse. However, little is known about how frequently ctDNA is detectable in pediatric patients with solid tumors. Therefore, we developed a next-generation sequencing approach to detect and quantify ctDNA in the blood of patients with the most common pediatric solid tumors. Methods Detection of ctDNA requires assays sensitive to somatic events typically observed in the cancer type being studied. In pediatric solid tumors, structural variants are more common than recurrent point mutations. We adapted an ultralow passage whole-genome sequencing approach to capture copy number variants and a hybrid capture sequencing assay to detect translocations in liquid biopsy samples from pediatric patients. Results Copy number changes seen by ultralow passage whole-genome sequencing enabled detection of ctDNA in patients with osteosarcoma, neuroblastoma, alveolar rhabdomyosarcoma, and Wilms tumor. In Ewing sarcoma, detection of the EWSR1 translocation was a more sensitive approach. For patients with samples collected at multiple time points, changes in ctDNA levels corresponded to treatment response. We also found that disease-specific genomic biomarkers of prognosis were detectable in ctDNA. Conclusion This study demonstrates that liquid biopsy approaches that detect somatic structural variants are well suited to pediatric solid tumors. We show that children with the most common solid tumor malignancies have detectable levels of ctDNA, which may be used to track disease response and identify genomic subclassifiers of disease. Efforts to profile larger collections of clinically annotated specimens are under way to validate the clinical use of these assays.

Purpose Patients with castration-resistant prostate cancer (CRPC) receive second-generation androgen-deprivation therapy, but frequently experience relapse or do not respond. Understanding the genetic mechanisms of resistance will help to identify strategies and biomarkers that are essential for the next line of therapy. Patients and Methods We analyzed whole exomes of patient-matched pre- and post-treatment tumors from patients with CRPC. These patients had received the secondary androgen-deprivation therapy agent, abiraterone, which suppresses androgens to below castration levels, or enzalutamide, which competitively inhibits the key androgen signaling effector, androgen receptor. Results We observed that abiraterone-resistant tumors harbored alterations in AR and MYC, whereas enzalutamide-resistant tumors gained alterations in cell-cycle pathway genes, such as mutation in cyclin-dependent kinase N2A ( CDKN2A) or amplification of CDK6. Experimentally, overexpressing cell-cycle kinases promoted enzalutamide resistance in androgen-sensitive LnCAP cells that was mitigated via CDK4/6 blockade—palbociclib and ribociclib. Conclusion CDK4/6-mediated resistance observed in preclinical experiments suggests that CDK4/6 amplifications may sufficiently promote enzalutamide resistance in CRPC, and that these patients may respond to palbociclib or ribociclib. The overall observations suggest that, in genomically selected advanced CRPC, clinical strategies against abiraterone- or enzalutamide-resistant tumors may require treatment strategies that are tailored to the resistance mechanisms that are specific to those patient subpopulations.

3014 Background: Precision medicine has revolutionized oncologic care in the United States (US) in the past two decades. While the US cancer population is rapidly diversifying, enrollment of a diverse patient population into clinical trials lags behind. In particular, it is unclear whether minority patients are adequately represented in precision oncology trials. Herein, we report racial/ethnic representation in precision oncology studies spanning four common cancer types (breast, lung, prostate, colorectal cancers). Methods: Completed US clinical studies incorporating precision medicine objectives based on a set of 12 precision oncology search terms (including tumor biomarker, whole exome sequencing, tumor mutation testing, gene expression signatures, tumor microarray, tumor genomics, et cetera) were identified from Clinicaltrials.gov. Studies were reviewed for reporting race/ethnicity for inclusion in the analysis. The Surveillance, Epidemiology, and End Results (SEER) database was used to determine incidence of race/ethnicity in the US cancer population, correlated with disease site and median year of enrollment for each trial. The difference in incidence (D-I) was defined as the median absolute difference in study racial enrollment and SEER incidence, with a negative value corresponding to underrepresentation. Wilcoxon signed-rank test was used to compare median D-I to a value of 0 by racial/ethnic subgroups. Results: Overall, 156 studies were identified; 40.3 and 27.5% studies enrolling from 2000 through 2020 met the inclusion criteria for racial and ethnic subgroups reporting, respectively. Of 4,418 total enrollees, 82.5% were White, 10.5% Black, 3.8% Asian, and 0.4% American Indian/Alaskan Native (AIAN). Ethnically, 6.4% were Hispanic. The D-I was +2.2% for Whites (interquartile range (IQR) = -43.7% to 25.4%; P < 0.013), -0.74% AIAN (IQR = -0.8% to +5.9%; P < 0.001), -2.5% Asians (IQR = -4.1% to 30.4%; P < 0.152), -4.6% Blacks (IQR = -20.1% to +45.0%; P < 0.001), and -8.1% Hispanics (IQR = -14.8% to + 29.6%; P < 0.001). By disease site, Blacks were significantly underrepresented proportional to their cancer incidence among prostate (D-I of -11.8%, p = 0.009) and lung studies (D-I of -5.9%, p = 0.013), while prostate studies significantly overrepresented Whites (D-I +14.0%, p = 0.005). Lung studies overrepresented Asians (D-I +0.49%) consistent with the prominent role of targetable oncogene drivers in this population. Conclusions: Results demonstrate an underrepresentation of minority racial groups and an overrepresentation of Whites in precision oncology studies. Increased emphasis on equitable enrollment onto these studies is critical, as resulting precision Omic conclusions are used to stratify populations and personalize treatments. A continued lack of diversity among enrollees may further leave behind vulnerable minority populations in the era of precision oncology.

98 Background: Treatment (tx) options for patients (pts) with metastatic castration-resistant prostate cancer (mCRPC) to bone are limited. Radium-223 (R223) has demonstrated overall survival (OS) benefit, but objective clinical responses to R223 or the anti-PD1 checkpoint inhibitor (CPI) pembrolizumab (pem) are infrequent. As R223 may increase immunogenicity of mCRPC to bone and increase activity of CPI, we undertook a Phase 2 study to assess safety of the combination and differences in immune cell infiltrate in bone biopsies (bx) and preliminary clinical activity of R223 + pem vs. R223 alone. Methods: Eligibility required mCRPC to bone with no visceral metastases (mets) or lymph nodes > 2 cm, ECOG PS 0 or 1, Hgb ≥ 9 g/dL, and no prior R223 or CPI. Pts underwent bone bx at screening and at 8 wks. Pts were stratified by alkaline phosphatase ≥220 vs. < 220 U/L and high vs. low volume bony mets (CHAARTED criteria) and randomized 2:1 to receive R223 55 kBq/kg q4wks + pem 200 mg q3wks (Arm A) or R223 55 kBq/kg q4wks alone (Arm B). If restaging after 3 doses R223 showed at least stable disease, pts in Arm A continued pem alone until progressive disease (PD). Upon PD, R223 was resumed if no new visceral mets. Pts continued tx until clinical/radiologic PD, unacceptable toxicity or completion of 6 R223 doses. The primary endpoint was difference in CD4+ and CD8+ T-cell infiltrate in 8 wk vs. baseline bx; secondary endpoints were safety/tolerability, radiographic progression-free survival (rPFS) and OS. Exploratory endpoints included PSA response and rate of symptomatic skeletal events (SSEs). Results: Of 45 pts enrolled, 42 received study tx (29 Arm A, 13 Arm B) and were eligible for analysis. 21 pts in Arm A and 5 in Arm B had evaluable paired bone bx. Median fold-change of proportion of CD4+ T-cells/total cell count from baseline to 8 wks was 0.90 (range 0.0-26.6) in Arm A and 0.40 (0.0-13.0) in Arm B (P = 0.87); for CD8+ cells, median 0.67 (0.0-40.4) in Arm A and 0.40 (0.1-28.8) in Arm B (P = 0.77). Grade 3 treatment-related non-hematologic adverse events (AEs) occurred in 3 pts (10%) in Arm A (pneumonitis, diarrhea, AST increased); none in Arm B. Median rPFS was 6.7 mo (95% CI 2.7-11.0 mo) in Arm A and 5.7 mo (2.6-NR) in Arm B. Median OS was 16.9 mo (12.7-NR) in Arm A and 16.0 mo (9.0-NR) in Arm B. 3 pts (10%) in Arm A and 0 in Arm B had PSA reduction of ≥ 50%. SSE rate was 38% in Arm A and 54% in Arm B, with pathologic fractures in 0% of pts in Arm A and 23% in Arm B. Conclusions: In the 62% of treated pts with evaluable paired bx at baseline and after 8 wks, there was no evidence of increased CD4+ or CD8+ T-cell infiltration with R223 + pem. Additional biomarker analyses will be presented. This study revealed that R233 + pem did not result in unexpected AEs, but did not lead to prolonged rPFS or OS compared to R223 alone to support this two-drug combination in a biomarker-unselected population in this setting. Clinical trial information: NCT03093428.

9549 Background: Long-term response – the plateau of the survival curve – is the transcendent benefit from PD-1 blockade. However, only a subset of responses achieve substantial durability. The frequency, characteristics, and predictors of long-term responders (LTR) to PD-1 blockade are not well known and may differ from short-term responders (STR). Methods: Patients with advanced NSCLC treated with anti-PD-1/PD-L1 therapy from two institutions (MSK and DFCI) were examined. Responses were assessed by RECIST. LTR was defined as PR/CR lasting ≥ 24 months. STR was defined as PR/CR lasting < 12 months. Comparisons were also made to patients with progressive disease (PD). PD-L1 expression was assessed by IHC. TMB was assessed by targeted NGS; high TMB was defined as ≥ median of the cohort. A subset had detailed molecular profiling by MSK-IMPACT. Fisher’s exact and Mann-Whitney U tests were used to compare features, and the log-rank test was used to compare survival. Results: Of 2318 patients (MSK n = 1536, DFCI n = 782), 126 (5.4%, 95% CI 4.6-6.4%) achieved LTR, with similar rates in both cohorts. STR occurred in 139 (6%). Overall survival was longer in LTR compared to STR (median NR vs 19.6 months, HR 0.07, p < 0.001). LTR had deeper responses compared to STR (median best overall response -69% vs -46%, p < 0.001). Patients with LTR were younger ( < 65 years old) and had increased TMB (≥ median mut/Mb) compared to both STR and PD (p = 0.006, p = 0.03; p < 0.001, p < 0.001). The rate of LTR was enriched among patients with both high TMB/high PD-L1 compared to those with low TMB/low PD-L1 (9% vs 1%, OR 9.2, p < 0.001), while STR was similar in both groups (7% vs 6%). 2% of patients with sensitizing EGFR mutations (n = 243) achieved LTR. Loss of function variants in ARID1A (14% vs 2%), PTEN (8% vs 0%), and KEAP1 (12% vs 2%) were enriched in LTR compared to STR (p < 0.05 for each). Among patients with KRAS mutations, the rate of LTR was higher in those with co-mutation with TP53 compared to STK11 (11% vs 2%, p = 0.01). Conclusions: Long-term response (LTR, ongoing response ≥ 24 months) to PD-1 blockade is an uncommon but profound clinical outcome in metastatic lung cancers. Younger age and high TMB correlate with LTR; the combination of high TMB/high PD-L1 enriches for LTR but not STR. Features predicting long term response may be distinct from those predicting initial response.

1046 Background: Preclinical evidence suggests treatment (tx) with T-DM1 plus an anti-PD1 antibody triggers antitumor immunity. We conducted a phase 1 trial to determine the safety and explore the efficacy of T-DM1 plus pembro. Methods: Eligible patients (pts) had MBC previously treated with trastuzumab (H) and taxane (T), were T-DM1-naïve, and received >1 prior line of tx for MBC or developed recurrence within 6 months (mo) of adjuvant tx. A dose de-escalation (esc) design was used with 6 pts in the dose-finding cohort, followed by an expansion (exp) cohort at the recommended phase 2 dose (RP2D), with mandatory baseline biopsies (bx). The primary endpoint was safety and tolerability. Secondary endpoints included objective response rate (ORR), progression-free survival (PFS), and clinical benefit rate (CBR: complete response + partial response + stable disease >24 weeks). Associations between immune biomarkers and tx response were explored. Results: 20 pts started protocol tx (6 in dose de-esc cohort; 14 in exp cohort). Median follow-up was 23.5 mo. Pts had median age 54 yrs and median 1 line of prior MBC tx (range 0-2); 100% had received prior T, H, and pertuzumab. There were no dose-limiting toxicities in the dose de-esc cohort; thus full doses of T-DM1 (3.6 mg/kg q21 days) and pembro (200 mg q21 days) were the RP2D. 85% of pts experienced tx-related adverse events (AEs) > grade (gr) 1; 20% of pts experienced gr3 AEs. There were no gr>4 AEs. Gr3 AEs were fatigue; AST increase; ALT increase; pneumonia; pneumonitis; oral mucositis; and vomiting, each in 1 pt. 17 pts had baseline bx; 6 pts had repeat bx after 1 tx cycle. Efficacy results, overall and by PD-L1 Combined Positive Score (CPS; 22C3 staining) and tumor-infiltrating lymphocyte (TIL) status, are shown in the table. Tumors’ antigen presentation will be explored through HLA/dendritic cell marker staining and immune signatures by RNA sequencing. Conclusions: T-DM1 plus pembro was safe and tolerable. The regimen demonstrated clinical activity. Further exploration of immune-related predictive biomarkers is warranted. Clinical trial information: NCT03032107 . [Table: see text]

3627 Background: Differential gene expression (DGE) methods, initially developed for analyzing bulk RNA changes in pure tumor cell lines under experimental settings, are commonly used to identify biomarkers in and infer biological differences between patient tumor samples, which are admixtures of tumor and non-tumor components. Methods to sensitively and accurately detect cell type-specific expression differences in admixed patient samples are not well characterized but may greatly affect emerging targeted and immunotherapy biomarker strategies. To address this issue, we developed a simulation framework to benchmark our ability to detect changes in tumor-intrinsic gene expression. Methods: Pseudobulk RNAseq melanoma cohorts were simulated by sampling from melanoma single cell RNAseq data. Simulation parameters were optimized to maximize concordance of gene expression means and variances (Spearman r = 0.81, 0.68, respectively) between the TCGA SKCM cohort (n = 462) and matched simulated cohort, and then validated in two independent melanoma cohorts (n = 42, 129; means Spearman r = 0.80, 0.78; variances Spearman r = 0.68, 0.63). Using this simulation framework, we benchmarked the effect of sample size, magnitude of differential expression, and differences in cell type proportions on the sensitivity and positive predictive value (PPV) of detecting true differentially expressed genes in the tumor-intrinsic compartment. Results: Reference cohorts of 50 total tumors (n = 10) were simulated to contain a 2 standard deviation tumor-intrinsic expression change in 50 randomly selected genes and a 11% difference in mean purity between two equally sized 25-tumor subgroups. DGE analysis using DESeq2 with an FDR q-value threshold of 0.1 yielded a sensitivity of 0.37 and PPV of 0.29. DGE analysis of the same simulated cohorts using a non-parametric Mann-Whitney U test with an FDR q-value threshold of 0.1 yielded a sensitivity of 0.13 and PPV of 0.76. Conclusions: Commonly used DGE methods for existing expression-based biomarker strategies have poor sensitivity and PPV in admixed tumor samples, limiting our ability to find meaningful transcriptional biomarkers in clinical cohorts. We are currently developing methods to more accurately detect true differentially expressed genes in admixed bulk RNAseq samples and applying these approaches for biomarker discovery in immunotherapy-treated patient cohorts and other clinical tumor cohorts.

1518 Background: Germline genetic analysis is an essential tool for implementing precision cancer prevention and treatment. However, only a small fraction of cancer patients, even those with features suggestive of a cancer-predisposition syndrome, have detectable pathogenic germline events, which may in part reflect incomplete pathogenic variant detection by current gold-standard methods. Here, we leveraged deep learning approaches to expand the diagnostic utility of genetic analysis in cancer patients. Methods: Systematic analysis of the detection rate of pathogenic cancer-predisposition variants using the standard clinical variant detection method and a deep learning approach in germline whole-exome sequencing data of 2367 cancer patients (n = 1072 prostate cancer, 1295 melanoma). Results: Of 1072 prostate cancer patients, deep learning variant detection identified 16 additional prostate cancer patients with clinically actionable pathogenic cancer-predisposition variants that went undetected by the gold-standard method (198 vs. 182), yielding higher sensitivity (94.7% vs. 87.1%), specificity (64.0% vs. 36.0%), positive predictive value (95.7% vs. 91.9%), and negative predictive value (59.3% vs. 25.0%). Similarly, germline genetic analysis of 1295 melanoma patients showed that, compared with the standard method, deep learning detected 19 additional patients with validated pathogenic variants (93 vs. 74) with fewer false-positive calls (78 vs. 135) leading to a higher diagnostic yield. Collectively, deep learning identified one additional patient with a pathogenic cancer-risk variant, that went undetected by the standard method, for every 52 to 67 cancer patients undergoing germline analysis. Superior performance of deep learning, for detecting putative loss-of-function variants, was also seen across 5197 clinically relevant Mendelian genes in these cohorts. Conclusions: The gold-standard germline variant detection method, universally used in clinical and research settings, has significant limitations for identifying clinically relevant pathogenic disease-causing variants. We determined that deep learning approaches have a clinically significant increase in the diagnostic yield across commonly examined Mendelian gene sets.

5010 Background: Immune checkpoint inhibitors targeting the PD-1 pathway have transformed the management of many advanced malignancies, including clear cell renal cell carcinoma (ccRCC), but the drivers and resistors of PD-1 response remain incompletely elucidated. Further, the common paradigm in solid tumor immunology that pre-existing CD8+ T cell infiltration, in combination with high numbers of nonsynonymous mutations (which, in the context of diverse HLA class I alleles, may be presented as neoantigens) drives response to PD-1 blockade, has not been thoroughly explored in ccRCC. Methods: We analyzed 592 tumors collected from advanced ccRCC patients enrolled in prospective clinical trials (CheckMate 009, CheckMate 010, CheckMate 025) of treatment with PD-1 blockade (n = 362) or mTOR inhibition (as control arm; n = 230) by whole-exome (n = 454) and RNA-sequencing (n = 311), integrated with CD8 immunofluorescence analysis (n = 219), to uncover the immunogenomic determinants of therapeutic response and survival. Wilcoxon rank-sum test was used to compare somatic alteration burden between clinical benefit (CB) v.s no CB (NCB); Fisher’s exact test was used to compare mutations and copy number alteration by infiltration state; and hazard ratio (HR) was calculated from Cox PH model for progression-free (PFS) and overall survival (OS) endpoints. All tests were at a significance level of p < 0.05. Results: Conventional genomic markers (tumor mutation burden, p = 0.81; neoantigen load, p = 0.47 for CB vs. NCB) and degree of CD8+ T cell infiltration (p = 0.88 for PFS; p = 0.65 for OS) were not associated with clinical response or altered survival with PD-1 blockade. These advanced ccRCC tumors were highly CD8+ T cell infiltrated, with only 22% having an immune desert phenotype and 5% with an immune excluded phenotype. Our analysis revealed that CD8+ T cell infiltrated tumors are depleted of clinically favorable PBRM1 mutations (p = 0.013) and enriched for unfavorable chromosomal losses of 9p21.3 (p < 0.001) when compared to non-infiltrated tumors. When found within infiltrated tumors, del(9p21.3) was associated with worse CB rate (36% (9/25) for del(9p21.3) vs. 88% (7/8) for wildtype at that locus, p = 0.017) and worse survival (HR = 2.38, p = 0.01 for PFS; HR = 2.44, p = 0.01 for OS) with PD-1 blockade. Conclusions: These data demonstrate how the potential interplay of immunophenotypes with somatic mutations and chromosomal alterations impacts therapeutic efficacy in advanced ccRCC.

5023 Background: We previously showed that levels of CD8+ tumor infiltrating cells (TIC) expressing PD-1 but not TIM-3 and LAG-3 (CD8+ PD1+TIM3−LAG3−) were associated with response to nivolumab (nivo) in pretreated mccRCC pts (Pignon et al, 2019). Here, we sought to validate these findings in a randomized Phase III trial of nivo versus everolimus (evero) (CM-025) and explore the association of the biomarker with transcriptomic profiles. Methods: Tumor tissues from the CM-025 trial were analyzed (nivo arm: n = 116, evero arm: n = 107). Density/percentage of CD8+ PD1+TIM3−LAG3− TIC was evaluated by immunofluorescence (IF) and PD-L1 expression on tumor cells (TC) was evaluated by IHC. Linear association with outcomes was assessed using binary logistic (ORR, clinical benefit (CB) defined as CR/PR and PFS≥12 months) and Cox PH (PFS, OS) regression models (1-sided p-values shown). Bulk RNA-seq was performed in a subset of samples (n = 71) and data analyzed using ssGSEA and Gene Signature Scores (GSS). Results: In the nivo arm, density of CD8+ PD1+TIM3−LAG3− TIC (IF biomarker) was associated with ORR (OR = 1.43, p = 0.03) and CB (OR = 1.54, p = 0.02) while a trend was observed with PFS (HR = 0.87, p = 0.06). At an optimized cutoff, nivo treated pts with high IF biomarker (24/116, 20.7%) had higher ORR (45.8% vs 19.6%, p = 0.01) and CB (33.3% vs 14.1%, p = 0.03) and longer median PFS (9.6 vs 3.7 months, p = 0.03) than pts with low IF biomarker. A significant interaction between the IF biomarker and treatment was seen for both PFS and OS (2-sided p = 0.02 and 2-sided p = 0.08, respectively; significance determined as p < 0.15). By bulk RNA-seq, several inflammatory pathways (FDR q < 0.1) and inflammatory GSS (FDR q < 0.05) were enriched in the high IF biomarker group. When combined with the IF biomarker, TC PD-L1 expression (≥1%) further separated clinical outcomes (ORR, CB and PFS) in the nivo arm. In the evero arm, the IF biomarker was neither prognostic nor predictive of any clinical outcome. Conclusions: High levels of CD8+ PD1+TIM3−LAG3− TIC predicted response to nivo (but not to control evero) in mccRCC pts and were associated with activation of inflammatory response. Combination with TC PD-L1 further improved its predictive value, confirming our previous findings (Pignon et al, 2019). Further validation in the setting of first-line anti-PD-1 therapy is ongoing.

162 Background: Through gene expression profiling (GEP), the PAM50 classifier demonstrates prognostic value in localized prostate cancer (PCa). Pre-clinical drug response models predict increased taxane sensitivity in luminal subtypes compared to basal subtype. Men with mHSPC and high-risk features have greatest benefit from androgen deprivation therapy (ADT) plus docetaxel (D) vs ADT alone. We therefore sought to test the prognostic and predictive value of PAM50 in pre-ADT specimens from E3805 CHAARTED. Methods: Whole transcriptomic profiling of formalin-fixed, paraffin-embedded primary PCa biopsies from pts enrolled in the E3805 CHAARTED trial of ADT vs ADT+D was performed using the Human Exon 1.0 ST microarray platform (Decipher Biosciences). Normalized gene expression was used to classify subjects as luminal A, luminal B or basal subtype. Multivariable analyses (MVA) adjusted for ECOG status, de novo metastasis vs prior local therapy and volume of disease. The primary endpoint was overall survival (OS). Secondary endpoint was time to castration resistant PCa (TTCRPC). Results: Successful GEP was completed in 160 of 198 pts with available specimens. Eighty (50%), 77 (48%) and 3 (2%) pts were classified as luminal B, basal and luminal A, respectively. High volume disease was similarly present in luminal B (79%) and basal (78%) subtypes. In the ADT arm, luminal B subtype was associated with shorter OS vs basal (HR 1.75, p=0.05); consistent in MVA. Pts with luminal B subtype treated with ADT+D showed significant improvement in TTCRPC and OS (Table). By contrast, basal subtype showed no OS benefit from ADT+D even in pts with high volume disease. Conclusions: We demonstrate that GEP identifies tumor subtypes associated with differential benefit from chemohormonal therapy for mHSPC. Luminal B subtype is associated with poorer OS with ADT alone and benefits from addition of D. Basal subtype shows a lack of OS benefit from upfront ADT+D. We plan to validate these findings in independent trial cohorts.[Table: see text]

507 Background: There is no established neoadjuvant therapy (NAT) for patients (pt) with muscle invasive bladder cancer (MIBC) ineligible for cisplatin-based chemotherapy preceding radical cystectomy. Encouraging prospective data indicate PD-1/PD-L1 inhibitors, including pembrolizumab and atezolizumab, are safe and active as NAT for MIBC. Durvalumab (D), a PD-L1 inhibitor, is FDA approved for treating locally advanced or metastatic urothelial carcinoma following platinum-based chemotherapy. The safety and activity of D as NAT in MIBC have not been reported. Methods: We are conducting a single-center sequential multicohort trial (NCT03773666) of D alone (Cohort 1, N=10) and D plus the CD73 inhibitor oleclumab (Cohort 2, N=10) in cT2-T4aN0M0 MIBC pts who are RC candidates and are ineligible for or declined cisplatin-based chemotherapy. The primary endpoint is feasibility, defined as ≥7 of 10 pts receiving at least 1 dose of D followed by radical cystectomy without dose limiting toxicity (DLT) up to 12 wks post-RC. In Cohort 1, D is administered at 750mg IV Q2W for 3 cycles followed by RC 2-4 weeks after the last dose. Baseline and RC tissue and baseline and on-study blood are collected for correlative studies, including immunohistochemistry, genomics, transcriptomics, and metabolomics. Results: Cohort 1 has completed enrollment; ten pts were enrolled between Feb 2019 to Sept 2019. Median age was 67 (Range: 53-85) and 8 (80%) were men. All 10 pts completed 3 durvalumab doses. Eight pts completed planned RC with at least 12wk follow-up post-op to date. No DLTs were observed. One Grade 3 treatment-related adverse event (trAE) was reported (anemia), with no Grade 4 or higher trAE. Pathologic response (

111 Background: Cediranib, a vascular endothelial growth factor receptor tyrosine kinase inhibitor, suppresses expression of BRCA1, BRCA2, and RAD51 and increases sensitivity of tumors to poly (ADP-ribose) polymerase (PARP) inhibitors in vitro. Olaparib, a PARP inhibitor, demonstrates clinical efficacy in men with DNA repair deficient, mCRPC. We therefore performed a randomized phase 2 trial comparing olaparib with or without cediranib in men with mCRPC. Methods: Men with a minimum of one prior line of systemic therapy for mCRPC were randomized 1:1 to receive cediranib 30mg po daily plus olaparib 200mg po BID (Arm A) or olaparib 300mg BID alone (Arm B). At radiographic progression, patients (pts) in Arm B could crossover to Arm A. The primary endpoint was radiographic progression-free survival (rPFS). Secondary endpoints were objective response rate (ORR) and PSA50 decline rate (PSA50). Tumor biopsy specimens were obtained for biomarker analyses pre- and on-treatment. Results: Baseline characteristics of the 90 pts enrolled are summarized below. The median rPFS was 11.1 versus 4.0 months in Arm A and Arm B, respectively (Hazard Ratio 0.54, 95% CI 0.317, 0.928, p=0.026). Trends toward a higher ORR (19% and 12%), Disease Control Rate (Stable Disease + Partial Response) (77% and 64%,) and PSA50 (29% and 17%) were observed in Arm A compared to Arm B, respectively. Thirteen pts in Arm B crossed over to Arm A. One pt had a PR after crossover. Grade 3/4 adverse events (G3/4 AEs), irrespective of attribution, occurred in 77% and 58% of Arm A and Arm B pts, respectively. G3/4 AEs occurring in >10% of pts were hypertension (32%), fatigue (23%) and diarrhea (11%) in Arm A, and anemia (16%) and lymphopenia (11%) in Arm B. Conclusions: The cediranib/olaparib combination significantly improves rPFS in unselected, mCRPC pts. AEs were manageable. Analyses of mutation status in homologous recombination DNA repair genes are pending and will be key in interpreting the data. Clinical trial information: NCT02893917. [Table: see text]

TPS5597 Background: Alterations in DNA damage repair genes are common in metastatic castration-resistant prostate cancer (mCRPC), and are implicated in responses to carboplatin, PARP inhibitors and immunotherapeutics. The ATR kinase is involved in the DNA damage response, and ATR inhibitors have been demonstrated in preclinical models to have synergistic activity with platinum compounds due to induction of replication stress. Methods: This is a randomized open-label Phase 2 study of the ATR inhibitor M6620 + carboplatin vs. docetaxel + carboplatin in mCRPC. Patients (pts) previously treated with at least one secondary hormonal therapy and taxane-based chemotherapy undergo mandatory pre-treatment biopsy and are randomized 1:1 to receive Arm A (docetaxel 60 mg/m2 day 1 + carboplatin AUC 4 day 1) or Arm B (M6620 90 mg/m2 days 2,9 + carboplatin AUC 5 day 1) every 21 days. Pts randomized to Arm A who are not candidates for docetaxel receive carboplatin AUC 5 monotherapy. Stratification factors are 1) prior PARP inhibitor (yes vs. no) and 2) evaluable disease by RECIST 1.1 (yes vs. no). Pts on Arm A crossover to Arm B (M6620+carboplatin) at the earlier of PSA or radiographic progression. For the primary endpoint of overall response rate (ORR; PSA reduction by ≥ 50% or radiographic response by RECIST 1.1), with 65 pts on each arm (total N = 130), there will be 80% power to distinguish ORR of 40% vs. 20% using a chi-square test (one sided α = 0.05). 136 pts will be enrolled to account for 5% dropout. Secondary endpoints include time to PSA progression, radiographic PFS, PFS by PCWG3 criteria, safety and adverse events in each arm. Biomarker studies include whole exome sequencing, RAD51 focus formation, and ATM IHC from tumor specimens. Circulating cell-free DNA from pre-treatment and progression plasma specimens will undergo ultra-low pass whole genome sequencing and deep targeted sequencing. The goal of this study is to expand therapeutic options in mCRPC through a novel approach to targeting the DNA damage response, and to identify biomarkers associating with response and resistance to both standard and trial therapy. Enrollment began June 2019 (NCI/ETCTN #10191, NCT03517969). Clinical trial information: NCT03517969 .

715 Background: S/R RCC are highly aggressive tumors but recent pilot clinical data have suggested that these tumors respond well to ICI. Our aim was to perform integrative molecular characterization of S/R RCC tumors in order to characterize potential features that underlie their poor prognosis and responses to ICI. Methods: We compared genomic (1), transcriptomic (2) and immune microenvironment (3) data between S/R and non-S/R tumors. (1) S/R patients from 3 cohorts [N = 209]: The Cancer Genome Atlas [TCGA], CheckMate 010/025 & panel sequencing from Dana-Farber/Harvard Cancer Center [DF/HCC]. (2) RNA-seq on S/R from 2 cohorts [N = 98]: TCGA & CheckMate 010/025. (3) Immunofluorescence for CD8+ T cells [N = 17] & Immunohistochemistry for PD-L1 expression on tumor cells [N = 118] from CheckMate 010/025. Overall Response Rate (ORR), Progression Free Survival (PFS), and Overall Survival (OS) in S/R RCC was compared between ICI and non-ICI in clinical cohorts (Table). Results: S/R tumors were significantly enriched in mutations in BAP1, NF2, RELN, and MUTYH, deletions of CDKN2A/B & amplifications of EZH2 (q < 0.05) compared to non-S/R tumors. Gene Set Enrichment Analysis showed upregulation of epithelial-mesenchymal transition, immune pathways, and proliferation programs compared to non-S/R tumors in both RNA-seq cohorts independently (q < 0.25). S/R tumors exhibited greater infiltration by CD8+ T cells at the tumor margin (p = 0.048) and PD-L1 expression on tumor cells (43.2% vs 21.0%, p < 0.01) compared to non-S/R. S/R had improved ORR, PFS, and OS on ICI vs. non-ICI (Table). Conclusions: S/R RCC tumors have distinctive molecular features that may account for their association with poor prognosis and outcomes on ICI.[Table: see text]

740 Background: Metastatic vRCC are aggressive tumors with poor prognosis. Our phase 2 trial of AB in vRCC showed a response rate of 33%. We investigated on-therapy changes in circulating immune cells and cytokines and their association with outcomes. Methods: Blood was collected at baseline (C1D1) and on-therapy (C3D1). Peripheral blood mononuclear cells were analyzed for cell type, expression of immune checkpoints, markers of activation, proliferation and function using flow cytometry; circulating cytokines by multiplex immunoassay. Relationship with progression-free (PFS) and overall survival (OS) was assessed by cox regression models. Results: Baseline and on-therapy samples were collected from all 60 patients. High baseline levels of immunosuppressive cytokines IL1α, IL6, CCL4 and IL13, as well as high baseline levels of CD4+ lymphocytes expressing CD69, were associated with inferior PFS and OS (Table). However, a decline in these markers on-therapy was not predictive of outcomes. On-therapy assessments showed an increase in the IFN-γ inducible cytokine CXCL10 (p

746 Background: Higher levels of soluble PD-L1 (sPD-L1) are associated with poor prognosis in patients with solid tumors including in renal cell carcinoma (RCC). Here we have tested whether in patients with advanced RCC, sPD-L1 levels are associated with PD-L1 expression on tumor tissue or with clinical outcomes on PD-1 blockade. Methods: Serum from 91 patients with advanced clear-cell RCC on a biomarker study of nivolumab (NCT01358721) obtained at baseline (Day 1), Day 29 and Day 63, was tested by SiMoa™ for sPD-L1 (capture mAb 298.12B1, detection mAb 339.4C10; Freeman laboratory and Quanterix). Tumor PD-L1 (tPD-L1) was assessed on pretreatment biopsies (Dako). Association of sPDL1 and tPD-L1 with clinical outcomes was analyzed, including best overall response by RECIST (BOR), objective response of >20% (OR), progression free survival (PFS), and overall survival (OS). Results quote Wilcoxon Rank Sum test or paired t-test with significance at P < 0.05. Results: Median sPD-L1 was highest in patients with progressive disease (PD) at all timepoints (Table). Compared to baseline, sPD-L1 levels significantly increased in patients with PD on Day 29 and Day 63, while sPD-L1 levels significantly decreased in patients with CR/PR on Day 63. In addition, we found significantly higher baseline sPD-L1 in patients with prior therapy compared to those who were treatment-naïve. High tPD-L1 was weakly associated with favorable OR, but also weakly associated with high baseline sPD-L1. Conclusions: Unlike tPD-L1, sPD-L1 levels may show promise for association with clinical response to nivolumab in RCC. In this exploratory study, sPD-L1 increase on-treatment was significantly associated with lack of OR, and may be of utility as an early marker for PD worthy of future validation. Analysis of RNASeq from patients’ tumor specimen is underway to assess whether high sPD-L1 with PD is associated with immune suppressive signatures.[Table: see text]

4583 Background: NccRCC and ccRCCsd are aggressive tumors associated with poor prognosis and response to therapy. Combination strategies co-targeting VEGF signaling and inhibitory immune checkpoints are highly active in clear-cell renal cell carcinoma, but data is lacking in NccRCC and ccRCCsd. We conducted a multicenter, open-label, single arm phase II trial of atezolizumab plus bevacizumab in NccRCC and ccRCCsd. Methods: Patients with NccRCC and ccRCCsd ( > 20% sarcomatoid differentiation), and ECOG performance status of 0-2 were eligible. Prior systemic treatment was allowed with the exception of prior PD-1/PD-L1-directed therapy. Atezolizumab 1200mg and bevacizumab 15mg/kg were administered every 3 weeks until progression, unacceptable toxicity, or patient withdrawal. Primary endpoint was objective response rate (ORR) per RECIST 1.1. Exploratory biomarker analyses included PD-L1 expression on tumor (TC) and immune cells (IC), and spatial analysis of the immune infiltrate. Results: Sixty patients received at least 1 cycle of treatment, among whom 56 were evaluable for response (17 ccRCCsd and 39 NccRCC). ORR was 34% in the overall population, 53% in ccRCCsd and 26% in NccRCC. Median progression-free survival was 8.4 months (95%CI, 6.9-16.5). Baseline tumor tissue was available for 36 patients. TC PD-L1 expression ≥1% was associated with improved ORR (9/14, 64%) compared to patients with PD-L1 expression < 1% (4/20, 20%). Patients with TC PD-L1 expression ≥1% who experienced progressive disease as best response had shorter average distance between tumor cells and nearest neighboring immune cells at baseline. Further analysis of the immune tumor microenvironment on an expanded cohort, including IC PD-L1 expression and correlation with clinical outcomes, is ongoing and will be updated. Conclusions: The combination of atezolizumab plus bevacizumab is active in NccRCC and ccRCCsd. Candidate predictive biomarkers include PD-L1 expression in TC and topological analysis of the immune infiltrate. Clinical trial information: NCT02724878.

1528 Background: While small studies have supported a genetic cancer predisposition among subjects with urothelial carcinoma (UC), systematic germline evaluation of this population is lacking. Here, we report the prevalence of germline variants among subjects with UC from multiple centers completing panel-based testing at a large, commercial laboratory. Methods: 1149 UC subjects underwent germline testing of 1 to 126 genes using massively parallel sequencing with customized capture bait-sets to analyze exonic regions, flanking intronic sequences, and copy number alterations. Pathogenic (P) and likely pathogenic (LP) were confirmed using orthogonal technology in accordance with Invitae standard operating practices. Analysis was limited to 645 subjects who completed testing of a shared set of 42 genes. P/LP variants including single nucleotide variants/indels/ copy number variants are reported. De-identified personal and family cancer histories were evaluated. Fisher’s Exact test and the Mann-Whitney test were used to analyze categorical and continuous variables respectively. Results: Among the 645 UC subjects with 42-gene testing for any indication, median age at testing was 60 years (6-88) and 326 (51%) were female. P/LP variants were identified in 21 (50%) of the 42 genes in 98 (15%) of subjects, including Lynch syndrome genes (n = 26 [4%]), BRCA1/2 (n = 16 [2.5%]), CHEK2 (n = 15 [2.3%]), and heterozygous MUTYH (n = 12 [1.9%]). Among 18 DNA damage repair (DDR) genes assessed, 90 P/LP variants were detected in 88 subjects (12.2%). There was no significant association between presence of a DDR gene variant and age at diagnosis, gender or reported family history of UC in a first degree relative (n = 48). Among subjects with documented history of UC only without other cancers (n = 195), 24 (12.3%) had P/ LP variants, of which 23 (11.8%) were in a DDR gene. Conclusions: Germline P/LP variants were identified in 15% of UC subjects most of which (92%) were in DDR genes, including 27% in Lynch syndrome genes. PARP and T-cell checkpoint inhibitors may warrant evaluation in subjects with germline DDR mutations. Further validation in unselected UC pts is warranted to propose examining germline P/LP variants in all UC patients.

4514 Background: S/R mRCC are poorly characterized rapidly progressing tumors associated with poor prognosis. Although conventional therapies are less effective for these tumors, emerging data suggests that ICIs may be especially effective. Our aim was to characterize the genomic alterations (GA) in S/R mRCC tumors and evaluate their response to ICIs. Methods: We retrospectively compared the activity of first-line ICIs to non-ICI-based therapies for S/R mRCC patients (pts) treated at DFCI and analyzed sequencing data from an NGS panel (275-447 genes) on a subset of these patients (matched by histology to non-S/R mRCC). For S/R mRCC pts treated with ICI vs non-ICI therapies, overall survival (OS) and time to treatment failure (TTF) were compared by Cox regression and objective response rate (ORR) by logistic regression. GA frequencies were compared by Fisher’s test and tumor mutational burden (TMB) by Mann Whitney U between S/R and non-S/R mRCC. Results were considered statistically significant if p < 0.05 or q < 0.10. Results: 125 S/R mRCC pts were included (88 S, 23 R, 14 S&R) among which 103 were clear cell and 48 had sequencing data. GA in BAP1 were significantly more frequent in S/R vs non-S/R (25% vs 4.3%; q = 0.096) while other GA had similar frequencies and TMB (median [IQR]) was similar (7.2 [5.2-8.4] vs 6.8 [5.3-9.1] mut/Mb; p = 0.98). Median follow-up was 35.4 (95% CI = 24.9 – 46.0) months (m). On multivariable analysis, S/R mRCC pts treated with ICI had significantly better clinical outcomes (Table). Conclusions: Pts with S/R mRCC have a higher frequency of BAP1 GA and better outcomes on ICIs compared to non-ICI-based therapies. Future studies should determine the molecular mechanisms underlying the improved response to ICIs in S/R mRCC. [Table: see text]

1508 Background: Chronic lymphocytic leukemia (CLL) is among the most heritable cancers, with 60% of disease risk genetically determined. However, most of the genetic heritability of CLL remains unexplained. Previously, we identified ATM as the first CLL risk gene. Here, we leverage a deep-learning-based germline variant calling algorithm to explore germline mutational enrichment in DNA repair and cell cycle genes in CLL. Methods: A two-stage case-control analysis was conducted using gene-based mutational enrichment analysis of 50 established cancer predisposition DNA repair and cell cycle genes. In the discovery phase, a total of 285 Spanish patients and 5,608 ancestry-matched controls were evaluated. In the validation stage, an independent cohort of 514 European patients and 27,173 ancestry-matched controls were analyzed. An FDR correction was applied to both datasets and genes with a q-value < 0.2 in both cohorts were considered significant. Results: Our joint analysis of 799 CLL patients from 2 genetically distinct cohorts and 32,781 ancestry-matched cancer-free controls identified ATM and CHEK2 as significantly enriched in both CLL datasets. First, our analysis recaptured the previously reported finding of ATM variant enrichment in CLL patients. Carriers of pathogenic ATM mutations in our cohorts (n = 9 patients, discovery: 1.05%, validation: 1.17%) were 2.8–3.7 times more likely to develop CLL compared to cancer-free individuals (discovery: OR = 2.8, 95%CI = 0.7–9.0, q-value = 0.181; validation: OR = 3.7, 95%CI = 1.6–8.3, q-value = 0.0454). In addition, our analysis identified 21 CLL patients carrying pathogenic CHEK2 alterations (discovery: 1.40%, validation: 3.31%), making CLL patients 4.4-8.0 times more likely to carry such alterations compared to controls (discovery: OR = 8.0, 95%CI = 2.3–27.0, q-value = 0.026; validation: OR = 4.4, 95%CI = 2.5–7.3, q-value < 0.001). Conclusions: Our analysis of genetically distinct CLL cohorts, using a high-sensitivity variant calling algorithm, supports CHEK2 as a potentially novel CLL predisposition gene that may explain a portion of the missing monogenic heritability of CLL. In addition, this study highlights the DNA repair and cell cycle regulation pathways as potential drivers of CLL susceptibility.

2515 Background: ICB has improved survival in melanoma. Patients with stable disease (SD) as best treatment response represent an intermediate response phenotype whose biology has been incompletely characterized. Methods: Whole exome and transcriptome sequencing from pre-treatment tumors in melanoma patients treated with ICB (anti-CTLA-4 and/or anti-PD-1) were assembled and uniformly analyzed (WES n = 293; WES+RNA-seq n = 159). RECIST (v1.1) was used to determine complete or partial response (CR/PR; n = 94), SD (n = 42), or progressive disease (PD; n = 157). Gene set enrichment analysis (GSEA) was performed on 50 “hallmark” gene sets to identify pathways differentially expressed in patients with SD. CIBERSORT was used to infer relative proportions of 22 immune cell types in each sample. Mutation antigenicity was determined by calculating patient-specific mutation affinity for MHC class I peptides. Results: GSEA identified enrichment of multiple immune-related gene sets in SD tumors, including TNF-α signaling and interferon-ɣ response (FDR q < 0.1, SD vs CR/PR and SD vs PD). SD tumors had higher HLA and antigen presentation pathway expression, and increased cytolytic T cell activity compared to CR/PR and PD. CIBERSORT analysis identified higher total immune infiltrate in SD patients compared to CR/PR and PD (Mann-Whitney U p = 0.03 and p < 0.001, respectively) but not in patients with CR/PR vs PD (p = 0.124). However, checkpoint expression, including PD-1, PD-L1, and LAG3, was also higher in SD patients. Mutation load did not differ between SD and CR/PR or PD patients (SD median 2.87 vs CR/PR median 7.98, Mann-Whitney U p = 0.104; PD median 3.42, p = 0.210). However, SD patients had more antigenic passenger mutations (SD vs CR/PR, p = 0.001; vs PD, p < 0.001); there was no difference in antigenicity of driver mutations. Conclusions: Pre-treatment melanomas from patients with SD contain more antigenic passenger mutations and demonstrate a global increase in immune signaling. This may describe a subset of patients with pre-existing dysfunctional immune response that is minimally responsive to ICB. Further characterization of the tumor-immune interaction in these patients may inform improved interventions.

103 Background: ICIs have shown benefit across several metastatic carcinomas, yet predictive biomarkers are still lacking. 20% of malignancies harbor alterations in ≥1gene that is part of PBAF complex. With recent data suggesting an association between PBRM1 mutations (mts) and outcomes in renal cell carcinoma (RCC) pts treated with ICIs (Miao, Science, 2018), we examined the association between PBAF mts and OS in ICI-treated patients across several solid cancer (ca) types. Methods: Of 6007 pts with different ca histologies and targeted exome sequencing (Oncopanel) at Dana Farber Cancer institute (DFCI), 138 pts had truncating mts in any PBAF gene (SMARCA4, PBRM1, and ARID2) or oncogenic missense mts in SMARCA4 and were treated with ICIs. 138 histology-matched DFCI pts had none. A publicly-available cohort (2:1 histology matched) from Memorial Sloan Kettering (MSKCC) (Samstein et al., Nature Genetics, 2019) of 621 ca pts (PBAF mutant [MT] = 207, PBAF wild type [WT] = 414) treated with ICIs was analyzed for association between PBAF mts and OS. OS was defined from time from ICI initiation. OS was compared by Cox regression between PBAF MT and PBAF WT. Hazard ratio (HR) was derived using univariable and multivariable analysis (MVA) adjusted for ICI regimen (single vs combination) and age. Results: Median (Md) follow-up for the combined cohort (n = 897) was 27 months (m). Major histologies were non-small cell lung ca (268; 29.9%), melanoma (220; 24.5%), RCC (181; 20.2%), and bladder ca (65; 7.2%). Results on univariable and MVA analyses from individual and combined cohorts are presented below. Conclusions: PBAF mts are associated with survival in ICI-treated ca pts. Work in progress with non-ICI treated pts will determine if this is prognostic or predictive of response. [Table: see text]

Off-label prescribing in oncology is common and unregulated. The aim of this study was to describe the off-label use of rituximab, a novel anti-CD20 antibody, among patients from a large proprietary insurance database to understand how frequently and appropriately off-label prescribing occurs for this medication.In this descriptive study, 11,232,642 patients were enrolled in the D2 Hawkeye commercial insurance database between 2001 and 2007, and 2,782 patients received rituximab. The main outcome measures were quantity and type of off-label usage, and expenditures for off-label usage.Seven hundred five (25.3%) patients received rituximab for off-label indications, and of those, 332 (47.1%) received rituximab for uncertain or inadequate evidence-based diagnoses. Expenditures for off-label indications were 17.1% of expenditures for rituximab usage.The frequent use of rituximab for off-label indications should lead to improved postapproval surveillance of biologics by the US Food and Drug Administration, so that use can be adequately studied. This will also facilitate improved regulatory mechanisms to ensure evidence-based use.

TPS1118Background: Triple negative breast cancer (TNBC) has an aggressive clinical course with higher relapse rates and shorter overall survival compared to patients with hormone receptor-positive ...

e13501Background: A challenge in cancer research is the availability of tumor samples linked to clinical information. To address this, we worked with patients (pts) to develop patient-driven studie...

3074Background: While immune checkpoint blockade (ICB) produces durable responses in some patients with melanoma, most patients derive no clinical benefit, and the molecular underpinnings of ICB re...

6509Background: Hospitalization or death within 30 days of palliative-intent chemotherapy for metastatic cancer represent undesirable outcomes. An automated framework for predicting the risk of hos...

9581Background: Immune checkpoint inhibitors (ICI) have revolutionized treatment in metastatic melanoma (MM), but progression occurs in the majority of patients (pts), and the evolution of resistan...

TPS4600Background: CheckMate 025 established NIVO as a standard of care (SOC) option for advanced clear cell RCC (ccRCC) after prior antiangiogenic therapy. The combination of IPI plus NIVO is anti...