Citations

Abstract Current unbiased approaches to mine the large amounts of patient-level data on mutations, structural variations and gene expression result in an unwieldy amount of interactions and correlations, which cannot be parsed to identify disease drivers. Here we present an approach to encode mutational and structural variant data at a patient level in a semantic association space. This approach transforms the presence of a mutation (or other feature) in each patient into the quantitative semantic association score of the corresponding gene and the phenotype of interest, which we have trained on all publicly available literature using word-embedding neural networks. Using data from The Cancer Genome Atlas (TCGA), we encoded the mutation or structural variant status (incl. copy number, fusion and chromothripsis) of all patients in the Lung Adenocarcinoma and Mesothelioma cohorts into our semantic space. For each cancer, we first defined the set of genes that are most associated to it according to the literature. To project each patient into this semantic space, we next determined if each patient had a mutation in the genes representing the disease semantic vector (e.g. NSCLC). For TCGA data we only counted non-Silent mutations and represented them as a binary number for each gene, i.e. 0 if the patient had no mutations in that gene and 1 if the patient had a non-Silent mutation in the gene. Each patient was then encoded in a binary vector with each member corresponding to a gene from the disease semantic vector. For example, lung adenocarcinoma was associated to 1,367 genes in our semantic space. A lung adenocarcinoma patient’s vector would them be composed of 1,367 binary numbers dictating if the gene is mutated or not in that patient. We then multiply these binary vectors with the semantic disease vector to obtain the patient’s projection in the disease space, which in effect replaces the binary number with the Semantic Association Score between the gene and the disease. Contrary to clustering patient samples by their mutation or structural variant data alone, our projected patient vectors clustered patients together into 22 groups with high patient-to-patient similarity. These clusters recapitulate canonical knowledge about the disease, e.g. Lung Adenocarcinoma patients form clusters that include EGFR-driven and KRAS-driven cohorts. We also see novel groups of patients driven by genes such as MET, STK11 and MALAT1. These clusters can be further stratified by their survival status and other clinical features. We validated our approach with a non-TCGA Mesothelioma cohort, revealing similarities in patient stratification regardless of the data source. This approach represents a dramatic shift in patient segmentation, delivering real-time grouping of patients and biomarker identification, which can accelerate clinical trial design and therapeutic development strategy. Citation Format: Enrique Garcia-Rivera, Aaron S. Mansfield, Karthik Murugadoss, Murali Aravamudan. Patient segmentation using machine-learning based literature and genomic data synthesis uncovers novel cohorts of NSCLC and mesothelioma patients [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 2452.

PURPOSE This pilot study examined the ability to operationalize the collection of real-world data to explore the potential use of real-world end points extracted from data from diverse health care data organizations and to assess how these relate to similar end points in clinical trials for immunotherapy-treated advanced non–small-cell lung cancer. PATIENTS AND METHODS Researchers from six organizations followed a common protocol using data from administrative claims and electronic health records to assess real-world end points, including overall survival (rwOS), time to next treatment, time to treatment discontinuation (rwTTD), time to progression, and progression-free survival, among patients with advanced non–small-cell lung cancer treated with programmed death 1/programmed death-ligand 1 inhibitors in real-world settings. Data sets included from 269 to 6,924 patients who were treated between January 2011 and October 2017. Results from contributors were anonymized. RESULTS Correlations between real-world intermediate end points (rwTTD and time to next treatment) and rwOS were moderate to high (range, 0.6 to 0.9). rwTTD was the most consistent end points as treatment detail was available in all data sets. rwOS at 1 year post–programmed death-ligand 1 initiation ranged from 40% to 57%. In addition, rwOS as assessed via electronic health records and claims data fell within the range of median OS values observed in relevant clinical trials. Data sources had been used extensively for research with ongoing data curation to assure accuracy and practical completeness before the initiation of this research. CONCLUSION These findings demonstrate that real-world end points are generally consistent with each other and with outcomes observed in randomized clinical trials, which substantiates the potential validity of real-world data to support regulatory and payer decision making. Differences observed likely reflect true differences between real-world and protocol-driven practices.

Abstract The MERIT study was a single-arm, phase II clinical trial of nivolumab for the second- or third-line treatment of patients with malignant pleural mesothelioma in Japan. MERIT confirmed that PD-1 inhibition has activity in mesothelioma and led to the regulatory approval of nivolumab for the treatment of mesothelioma in Japan. See related article by Okada et al., p. 5485

Abstract Mesothelioma affects mostly older individuals who have been occupationally exposed to asbestos. The global mesothelioma incidence and mortality rates are unknown, because data are not available from developing countries that continue to use large amounts of asbestos. The incidence rate of mesothelioma has decreased in Australia, the United States, and Western Europe, where the use of asbestos was banned or strictly regulated in the 1970s and 1980s, demonstrating the value of these preventive measures. However, in these same countries, the overall number of deaths from mesothelioma has not decreased as the size of the population and the percentage of old people have increased. Moreover, hotspots of mesothelioma may occur when carcinogenic fibers that are present in the environment are disturbed as rural areas are being developed. Novel immunohistochemical and molecular markers have improved the accuracy of diagnosis; however, about 14% (high‐resource countries) to 50% (developing countries) of mesothelioma diagnoses are incorrect, resulting in inadequate treatment and complicating epidemiological studies. The discovery that germline BRCA1‐asssociated protein 1 ( BAP1 ) mutations cause mesothelioma and other cancers (BAP1 cancer syndrome) elucidated some of the key pathogenic mechanisms, and treatments targeting these molecular mechanisms and/or modulating the immune response are being tested. The role of surgery in pleural mesothelioma is controversial as it is difficult to predict who will benefit from aggressive management, even when local therapies are added to existing or novel systemic treatments. Treatment outcomes are improving, however, for peritoneal mesothelioma. Multidisciplinary international collaboration will be necessary to improve prevention, early detection, and treatment.

e20639 Background: Multiple randomized controlled trials (RCTs) have shown a robust benefit of immunotherapy with immune checkpoint inhibitors (ICI) in non-small cell lung cancer (NSCLC). We did a meta-analysis to examine the benefit of ICI in various subgroups. Methods: PubMed was searched up to Jan 15, 2019 for RCTs comparing overall survival (OS) between ICI and control (without ICI) arms. Pooled hazard ratio (HR) and 95% confidence interval (CI) was calculated for each subgroup. Interaction tests were done to compare relative benefit between opposed subgroups of interest (eg. men vs women; reported as Pheterogeneity). All analyses were performed with a random effects model in Comprehensive Meta Analysis (v2). Results: Twelve phase 2/3 RCTs involving 7244 patients were included. A significant OS benefit of ICI was found in both squamous and non-squamous histology. Current/former smokers, EGFR wild-type, and KRAS mutant patients had a significant OS benefit from ICI, but never smokers, EGFR mutant, and KRAS wild-type patients did not. An OS benefit of ICI was found in patients with or without baseline brain metastasis, PD-L1 < 1% or ≥1%, men or women, age < 65 or ≥65, and ECOG PS 0 or ≥1. No significant difference of relative benefit from ICI over control was found in patients with different PD-L1 expression, sex, age, or ECOG PS (Table). Conclusions: OS benefit of ICI in NSCLC was associated with a smoking history, wild-type EGFR or KRAS mutation. However, the OS benefit of ICI was seen regardless of histology, PD-L1 expression, sex, age, and ECOG PS. [Table: see text]

e18321 Background: Despite the availability of targeted therapy for ROS1 NSCLC since 2016, there is limited information on real-world ROS1 testing practices to identify patients potentially eligible for targeted treatment among community practices. The objective of this study was to characterize ROS1 testing rates and identify potential barriers to ROS1 testing. Methods: The Flatiron Health EHR-derived database was used to identify patients diagnosed with advanced NSCLC between 7/1/2016 - 7/30/2018 and received systemic treatment in a community practice setting. Descriptive statistics were used to summarize the population, biomarker testing rates and timing of testing. Logistic and multinomial regressions including patient demographics and clinical characteristics were used to identify factors associated with the following outcomes: ROS1 testing, testing for biomarkers other than ROS1 (PDL1, ALK, EGFR) , longer times from advanced diagnosis (adv. dx) to ROS1 test results and initiation of therapy prior to ROS1 testing. Results: Of 9,275 eligible patients, ROS1 testing rates in squamous (SQ) and non-squamous histologies increased over the study period (30% to 52%, 62% to 75% respectively). Older age, males, poorer performance status, SQ histology, history of smoking and recurrent disease (RD) patients were significantly less likely to be tested. Among patients not tested for ROS1, 62% were tested for other biomarkers including > 50% of non-squamous patients tested for PDL1, EGFR or ALK and 51% of squamous patients tested for PDL1. The median time from adv. dx to test result was 24 days, with patient characteristics (i.e. Hispanic ethnicity, history of smoking) as well as test type predicting longer times to test results ( > 41 days vs. < 24 days after adv. dx). Patients who had a delay in test results were more likely to receive other therapy prior to availability of their ROS1 test results (25+ days vs. < 24 days OR [95% CI]: 9.18 [7.87, 10.70]). Conclusions: In real-world practice, patient characteristics were associated with ROS1 testing and testing delays which may result in some patient subgroups being less likely to be identified for potential targeted therapy.

e18240 Background: The time between a patient’s positive biomarker test result and initiation of targeted therapy may vary due to a number of factors, including the use of chemotherapy prior to or after the biomarker test result. The objective of this study was two-fold: first, to investigate the impact of delayed ALK inhibitor (ALKi) therapy on overall survival (OS) and second, to examine the association between the use of chemotherapy prior to or after biomarker testing and OS. Methods: The Flatiron Health EHR-derived database was used to identify patients with ALK-positive (ALK+) advanced NSCLC diagnosed between 1/1/2011 - 9/30/2018. The median time from ALK+ test result to ALKi start was used to separate patients into early versus delayed treatment cohorts. To account for potential immortal time bias, times from ALK+ test result to ALKi start from the delayed cohort were sampled with replacement and used to create modified index dates among the early cohort. Cox proportional hazards models adjusting for baseline characteristics (i.e. ECOG) were used to assess the association between delayed ALKi start and OS and the use of chemotherapy with OS. Results: 422 patients with ALK+ aNSCLC were included in this analysis with a median time from test result to ALKi start of 3 weeks. 88 patients (20.9%) received chemotherapy prior to starting their ALKi. Delayed ALKi use was associated with a 2.3 fold increase in risk of mortality (HR [95% CI]: 2.30 [1.28, 4.15], p < 0.01). There was no difference in survival observed between those who had received chemotherapy prior to initiating their ALKi and those who did not receive chemotherapy (HR [95% CI]: 0.99 [0.62, 1.58]). Among those who initiated chemotherapy prior to their ALK+ test result, the continued use of chemotherapy prior to initiating an ALKi did not result in differences in OS compared to those who switched to an ALKi without continuing chemotherapy (HR [95% CI]: 1.03 [0.44, 2.41]). Conclusions: Delayed initiation of ALKi may result in poor outcomes in patients with ALK+ NSCLC. Receipt of chemotherapy prior to ALKi or the duration of chemotherapy did not impact survival. Future strategies to improve the time to therapy initiation may be useful in improving patient outcomes.

6581 Background: While AC is recommended for all patients with stage IIA, IIB, and IIIA NSCLC, its use and benefit among the elderly population is unclear. Methods: We identified patients with stage IIA, IIB, or IIIA NSCLC using the National Cancer Database from 2006-2014. Patients were divided into age groups ≤65, 66-70, 71-75, 76-80, > 80 years. Trends in AC use, factors influencing AC administration, and outcomes associated with AC were studied. Results: Out of 27368 patients, 13464 received AC and 13904 did not. 11% had stage IIA disease, 50% had stage IIB disease and 39% had stage IIIA disease. AC use was lower with increasing age (49% age ≤65 vs. 3% age > 80 (P < 0.0001)). Temporal use of chemotherapy for each age group was unchanged from 2006-2014. In multivariate logistic regression analysis, factors predictive of lower AC use were older age, Medicare or Medicaid insurance, academic center, and higher Charlson Comorbidity Index (CCI). Stage IIIA (OR 1.4, 95% CI 1.3-1.6) and Stage IIB (OR 1.3, 95% CI 1.2-1.4) had a higher chance of receiving AC than stage IIA. In a Cox proportional hazard model, younger age, female sex, academic center, private insurance, higher income, lower CCI, West coast center and AC were associated with better outcomes. When stratified by age and stage, AC was still associated with better survival (Table). Conclusions: AC utilization in stages IIA, IIB and IIIA NSCLC remains low, despite its association with improved survival in the elderly. Future clinical trials may be needed focused on elderly patients to establish the best regimen to optimize outcomes. [Table: see text]

2621 Background: TMB is an emerging predictor of survival with immunotherapy. TMB is determined by taking the difference between somatic and germline datasets when tumor-normal pairs are available. In the case of commonly utilized tumor-only sequencing, additional steps are needed to estimate the somatic alterations. Computational tools have been developed that determine germline contribution based on sample copy state, purity estimates and occurrence of the variant in population databases. Given the potential bias of population datasets, we hypothesized that tumor-only filtering approaches may overestimate the actual TMB. Methods: We assessed the TMB from 50 tumors in 10 diseases including all missense, indels, and frameshift variants with an allelic fraction (AF) ≥5% and Coverage ≥100X within the tumor. Tumor-only TMB was evaluated against the gold standard of matched germline subtracted TMB at three levels. Level 1 removed all the tumor-only variants with AF in the non-TCGA ExAC database ≥1%. Level 2 removed all variants observed in population databases simulating a naive approach of removing germline variation. Level 3 used an internal tumor-only pipeline for calculating TMB. Results: There were significantly higher estimates of TMB with Level 1, Level 2 and Level 3 tumor-only filtering approaches than that determined by germline subtraction, resulting in significant bias. Whereas there was no correlation between TMB estimates and tumor-germline TMB for Level 1 filtering, there were improvements in correlations for Level 2 and Level 3. Conclusions: The tumor-only approaches that filter variants in population databases overestimate TMB compared to that determined by germline subtraction. Despite improved correlations with more stringent filtering approaches, these falsely elevated estimates may result in the inappropriate categorization of tumor specimens and negatively impact clinical trial results and patient outcomes. [Table: see text]

Abstract Background Immune-related adverse events (irAEs) have emerged as a serious clinical issue in the use of immune checkpoint inhibitors (ICIs). Risk factors for irAEs remain controversial. Therefore, we studied sex differences in irAEs in patients treated with anti-programmed cell death protein 1 (PD-1) therapy. Materials and Methods All patients with metastatic melanoma and non-small cell lung cancer (NSCLC) treated with anti-PD-1 therapy at Mayo Clinic Rochester and Florida from 2015 to 2018 were reviewed. Kaplan-Meier method and log-rank test was used for time-to-event analysis. Results In 245 patients with metastatic melanoma, premenopausal women were more likely to experience irAEs (all grades) compared with postmenopausal women and men (67% vs. 60% vs. 46%), primarily because of an increase in endocrinopathies (33% vs. 12% vs. 10%, respectively). In patients with NSCLC (231 patients), women (all ages) were also more likely to develop irAEs of all grades (48% vs. 31%). Women with NSCLC were more likely to develop pneumonitis (11% vs. 4%) and endocrinopathies (14% vs. 5%). No differences in grade ≥3 toxicities were seen across sexes in both cohorts, but women were more likely to receive systemic steroids for the treatment of irAEs compared with men. Better progression-free-survival was observed in women with NSCLC and irAEs (10 months vs. 3.3 months) compared with women without irAEs. Conclusion Women with metastatic melanoma and NSCLC are more likely to experience irAEs compared with men. We also observed differences between sexes in the frequency of certain irAEs. Larger studies are needed to investigate the mechanisms underlying these associations. Implications for Practice The results of this study suggest that women may be at a higher risk for immune-related adverse events (irAEs) compared with men when treated with anti-programmed cell death protein 1 therapy. In addition, women were more likely to develop certain irAEs, including endocrinopathies and pneumonitis. Close follow-up of women undergoing treatment with immune checkpoint inhibitors will allow clinicians to diagnose these treatment-related complications early, potentially reducing their associated morbidity and mortality. In addition, a possible association between irAEs and response to therapy was observed.