Search

Genomic testing is an important part of personalized oncology. This study explored testing patterns and results, as well as patient (pt) characteristics, in the BC setting. Data from the nationwide, de-identified Flatiron Health (FH)-Foundation Medicine, Inc. (FMI) BC clinico-genomic database were included; pts had to have been treated in a FH network oncology practice and tested with a FMI CGP test. Pts were >18 years old and had a primary or metastatic BC (mBC) diagnosis between 1/1/2011 and 3/30/2020. Only pts with solid-tumor tissue of origin and solid-tumor assay tests were considered; pts whose tests occurred >120 days after their last visit (based on report date) were excluded. Pt cohorts were classified by timing of test result receipt: cohort 1, between primary diagnosis and 120 days post-diagnosis if no surgery recorded; cohort 3a, within 60 days prior to/after mBC diagnosis; and cohort 3b, >60 days after mBC diagnosis. Pt characteristics and actionable mutations according to the FMI report only (associated with sensitivity or resistance to ≥1 therapy) were described by cohort and BC subtype, defined at primary diagnosis for cohorts 1/2 and at mBC diagnosis for cohorts 3a/3b. Pt characteristics were compared with those of the FH cohort (not tested with next-generation sequencing or received other next-generation sequencing tests). Overall, 4183 pts were included: 285 in cohort 1, 412 in cohort 2, and 3486 in cohorts 3a/3b. There was an overrepresentation of triple-negative BC (TNBC) among pts tested in the early setting (cohorts 1/2) vs the underlying population (33%/32% vs 10%), and an underrepresentation of hormone receptor (HR)-positive/HER2-negative BC (45%/48% vs 71%). TNBC remained overrepresented among pts tested at mBC diagnosis (cohort 3a; 30% vs 11%), but not during mBC follow-up (cohort 3b; 14% vs 11%), which followed the underlying FH distribution. The majority of patients were tested using samples collected within their respective test windows: 100% in cohort 1, 63% in cohort 2, and 78% and 46% in cohorts 3a/3b. For pts tested in the mBC setting, 22% and 15% in cohorts 3a and 3b, respectively, were tested using tissue collected >60 days prior to mBC diagnosis. Percentages of pts with at least one FMI-report-actionable alteration ranged from 94% (n = 601) in HER2-positive samples to 75% (n = 668) in pts with TNBC. Among this BC cohort, FMI-CGP predominantly occurred in the mBC setting, and use of archival tissue for CGP occurred in later lines of treatment. Testing, in particular prior to mBC diagnosis, reflected a disproportionately higher prevalence of pts with TNBC, potentially due to there being fewer satisfactory treatment options in this setting. The distribution of actionable alterations varied by HER2 and HR status. Citation Format: Achim Rody, Nayan Chaudhary, Christopher Craggs, Marcio Debiasi, Gilles Erb, Virginia Fisher, Margaret McCusker, Tamara Snow, Maria Vidal Losada, Patricia Luhn. Comprehensive genomic profiling (CGP) in breast cancer (BC): Patterns and results from a clinico-genomic database [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 437.

Real-world data (RWD) are potentially a highly valuable resource to accelerate drug discovery and improve clinical practice. In particular, datasets combining clinical outcomes and comprehensive genomic profiling (CGP) could uncover new mechanisms of resistance and identify new patient populations for targeted therapies. Here, we used a de-identified clinico-genomic database (CGDB) from routine clinical care based on a clinical EHR-derived data from Flatiron Health linked to tumor sequencing results from Foundation Medicine to assess how prevalence of genomic alterations changes following progression on targeted therapies in the metastatic setting. One of the main challenges we faced is the lack of systematic and longitudinal genomic testing: among the patients receiving a specific treatment, some had a CGP performed prior to treatment, others only around the end of the line of therapy, but only a handful had multiple tests. We therefore established a methodology to assess changes in alteration prevalence and tested how our results are sensitive to cohort building parameters such as timing of testing, duration of therapy, or prior lines of treatment. Our approach successfully recovered three established cases of therapeutic resistance: (1) EGFR T790M in non-small cell lung cancer patients treated with erlotinib or gefitinib; (2) AR alterations following hormonal therapies in prostate cancers; and (3) ESR1 mutations in hormone-receptor positive breast cancer patients progressing on endocrine therapies other than fulvestrant. We delved into the latter case and identified differences in the type of ESR1 mutation that occurred following endocrine therapies if CDK4/6 inhibitors were co-administered or not. On one hand, RWD lack randomization and can thus be subject to confounding variables but on the other hand, RWD have the advantage of including a larger number of patients in a real-world setting compared to clinical trials. Our analysis shows that real-world CGDB can recapitulate the expected emergence of mutations following specific lines of treatment. Our study is a proof of concept of usage of RWD to assess how prevalence of genomic alterations changes following treatment. Our method allowed us to assess differences between patient cohorts and generate novel hypotheses on treatment-related changes in genomic profiles, which can ultimately lead to new drug indications. Citation Format: Nayan Chaudhary, Patricia Luhn, Gunther Jansen, Ciara Metcalfe, Marc Hafner. Identification of genomic alterations related to treatment progression in RWD [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 6323.

Stroke is the fifth leading cause of death in the United States and a major cause of severe disability worldwide. Yet, recognizing the signs of stroke in an acute setting is still challenging and leads to loss of opportunity to intervene, given the narrow therapeutic window. A decision support system using artificial intelligence (AI) and clinical data from electronic health records combined with patients' presenting symptoms can be designed to support emergency department providers in stroke diagnosis and subsequently reduce the treatment delay. In this article, we present a practical framework to develop a decision support system using AI by reflecting on the various stages, which could eventually improve patient care and outcome. We also discuss the technical, operational, and ethical challenges of the process. Geisinger Health Plan Quality Fund; National Institute of HealthUnited States Department of Health & Human ServicesNational Institutes of Health (NIH) - USA [R56HL116832] This work was sponsored in part by funds from the Geisinger Health Plan Quality Fund and National Institute of Health R56HL116832 (subaward) to VA and RZ. The funders had no role in study design, data collection, and interpretation, or the decision to submit the work for publication.