Search

e13563 Background: Chronic Obstructive Pulmonary Disease (COPD) is a heterogeneous disease with multiple phenotypes and endotypes and associated with lung cancer development. In this study, we analyzed the clinical features of COPD including various phenotypic and endotypic features from the electronic health records (EHR) as the potential risk factors for lung cancer development in a large cohort of COPD patients. Methods: We identified a COPD cohort based on electronic Medical Records and Genomics (eMERGE) network COPD cohort identification algorithms with minor modifications from the Mount Sinai Data Warehouse (2000-2020) and followed the patients on the first diagnosis of lung cancer. The development of lung cancer in COPD patients was confirmed by manual chart review. We retrieved the clinical features from EHRs and conducted Kaplan Meier (KM) analysis and multivariable Cox-regression modeling for hazard ratio (HR) analysis. Results: We found that 3.8 % of COPD patients (824 out of 21,658) developed lung cancer. While COPD patients with emphysema and smoking history (former or current) showed an increased risk of lung cancer onset, patients with concurrent asthma and corticosteroid (ICS) inhalation history showed a reduced risk of lung cancer onset (adjusted HR in Table). Interestingly, COPD patients with higher eosinophil counts showed late onset of lung cancer (>300 cells/ul, cancer rate at 5y=2.4%, p=< 0.002; 150-300 cells/ul, cancer rate at 5y=2.9%, p=0.003) when compared to patients with low eosinophils count (300 cells/ul; HR:0.72, 95% CI: 0.57-0.89, p=0.003) when compared to those with low eosinophils count (

e19512 Background: Targeted therapy is an important treatment for chronic lymphocytic leukemia (CLL). However, optimal strategies for deploying small molecule inhibitors or antibody therapies in the real world are not well understood, largely due to a lack of outcomes data. We implemented a novel temporal phenotyping algorithm pipeline to derive lines of therapy (LOT) and disease progression in CLL patients. Here, the CLL treatment pattern and time to the next treatment (TTNT) were analyzed in real-world data (RWD) using patient electronic health records. Methods: We identified a CLL cohort with LOT from the Mount Sinai Data Warehouse (2003-2020). Each LOT consisted of either a single agent or combinations defined by NCCN CLL guidelines. We developed a natural language processing (NLP)-based temporal phenotyping approach to automatically identify the number of lines and therapeutic regimens. The sequence of treatment and time interval for each patient were derived from the systematic treatment data. Time to event analysis and multivariate (i.e., age, gender, race, other treatment patterns) Cox proportional hazard (CoxPH) models were used to analyze the patterns and predictors of TTNT. Results: Four hundred eleven CLL patients received 1 to 7 LOTs. Ibrutinib was the predominant 1st LOT (40.8% of patients) followed by anti-CD20-based antibody therapies and chemotherapy in 30.6 and 19.2% of patients, respectively, followed by Acalabrutinib, Venetoclax, and Idelalisib in 3.4, 2.7, and 0.7% of patients, respectively (Table 1). The 2nd to 5th LOT showed the same or similar trends. We next analyzed the TTNT in the 1st line of each therapeutic class. Acalabrutinib resulted in a longer median TTNT than Ibrutinib. Both Acalabrutinib and Ibrutinib showed longer TTNT compared to Venetoclax (median TTNTs were 742 and 598 vs. 373 days: HR = 0.23, p=0.015 and HR = 0.48, p=0.03, respectively). In addition, patients with age equal to or older than 65 showed longer TNNT (HR=0.16, p=0.016). Conclusions: Our result shows the potential of RWD usage in clinical decision making as real-world evidence reported here is consistent with results derived from clinical trial data. Linking this study to genetic data and other covariates affecting treatment outcomes may provide additional insights into the optimal sequences of the targeted therapies in CLL. Table 1: Therapeutic class and patient numbers (%) in each line.[Table: see text]

6592 Background: Clinical trial phenotyping is the process of extracting clinical features and patient characteristics from eligibility criteria. Phenotyping is a crucial step that precedes automated cohort identification from patient electronic health records (EHRs) against trial criteria. We establish a clinical trial phenotyping pipeline to transform clinical trial eligibility criteria into computable criteria and enable high throughput cohort selection in EHRs. Methods: Formalized clinical trial criteria attributes were acquired from a natural-language processing (NLP)-assisted approach. We implemented a clinical trial phenotyping pipeline that included three components: First, a rule-based knowledge engineering component was introduced to annotate the trial attributes into a computable and customizable granularity from EHRs. The second component involved normalizing annotated attributes using standard terminologies and pre-defined reference tables. Third, a knowledge base of computable criteria attributes was built to match patients to clinical trials. We evaluated the pipeline performance by independent manual review. The inter-rater agreement of the annotation was measured on a random sample of the knowledge base. The accuracy of the pipeline was evaluated on a subset of randomly selected matched patients for a subset of randomly selected attributes. Results: Our pipeline phenotyped 2954 clinical trials from five cancer types including Non-Small Cell Lung Cancer, Small Cell Lung Cancer, Prostate Cancer, Breast Cancer, and Multiple Myeloma. We built a knowledge base of 256 computable attributes that included comorbidities, comorbidity-related treatment, previous lines of therapy, laboratory tests, and performance such as ECOG and Karnofsky score. Among 256 attributes, 132 attributes were encoded using standard terminologies and 124 attributes were normalized to customized concepts. The inter-rater agreement of the annotation measured by Cohen’s Kappa coefficient was 0.83. We applied the knowledge base to our EHRs and efficiently identified 33258 potential subjects for cancer clinical trials. Our evaluation on the patient matching indicated the F1 score was 0.94. Conclusions: We established a clinical trial phenotyping pipeline and built a knowledge base of computable criteria attributes that enabled efficient screening of EHRs for patients meeting clinical trial eligibility criteria, providing an automated way to efficiently and accurately identify clinical trial cohorts. The application of this knowledge base to patient matching from EHR data across different institutes demonstrates its generalization capability. Taken together, this knowledge base will be particularly valuable in computer-assisted clinical trial subject selection and clinical trial protocol design in cancer studies based on real-world evidence.

e18747 Background: Accurate longitudinal cancer treatments are vital for establishing primary endpoints such as outcome as well as for the investigation of adverse events. However, many longitudinal therapeutic regimens are not well captured in structured electronic health records (EHRs). Thus, their recognition in unstructured data such as clinical notes is critical to gain an accurate description of the real-world patient treatment journey. Here, we demonstrate a scalable approach to extract high-quality longitudinal cancer treatments from lung cancer patients' clinical notes using a Bidirectional Long Short Term Memory (BiLSTM) and Conditional Random Fields (CRF) based natural language processing (NLP) pipeline. Methods: The lung cancer (LC) cohort of 4,698 patients was curated from the Mount Sinai Healthcare system (2003-2020). Two domain experts developed a structured framework of entities and semantics that captured treatment and its temporality. The framework included therapy type (chemotherapy, targeted therapy, immunotherapy, etc.), status (on, off, hold, planned, etc.) and temporal reasoning entities and relations (admin_date, duration, etc.) We pre-annotated 149 FDA-approved cancer drugs and longitudinal timelines of treatment on the training corpus. A NLP pipeline was implemented with BiLSTM-CRF-based deep learning models to train and then apply the resulting models to the clinical notes of LC cohort. A postprocessor was developed to subsequently post-coordinate and refine the output. We performed both cross-evaluation and independent evaluation to assess the pipeline performance. Results: We applied the NLP pipeline to the 853,755 clinical notes, and identified 1,155 distinct entities for 194 cancer generic drugs, including 74 chemotherapy drugs, 21 immunotherapy drugs, and 99 targeted therapy drugs. We identified chemotherapy, immunotherapy, or targeted therapy data for 3,509 patients in the LC cohort from the clinical notes. Compared to only 2,395 patients with cancer treatments in structured EHR, this pipeline identified cancer treatments from notes for additional 2,303 patients who did not have any available cancer treatment data in the structured EHR. Our evaluation schema indicates that the longitudinal cancer drug recognition pipeline delivers strong performance (named entity recognization for drugs and temporal: F1 = 95%; drug-temporal relation recognition: F1 = 90%). Conclusions: We developed a high-performance BiLSTM-CRF based NLP pipeline to recognize longitudinal cancer treatments. The pipeline recovers and encodes as twice as many patients with cancer treatments compared with structured EHR. Our study indicates deep NLP with temporal reasoning could substantially accelerate the extraction of treatment profiles at scale. The pipeline is adjustable and can be applied across different cancers.

e14062 Background: Real world evidence generated from electronic health records (EHRs) is playing an increasing role in health care decisions. It has been recognized as an essential element to assess cancer outcomes in real-world settings. Automatically abstracting outcomes from notes is becoming a fundamental challenge in medical informatics. In this study, we aim to develop a system to automatically abstract outcomes (Progression, Response, Stable Disease) from notes in lung cancer. Methods: A lung cancer cohort (n = 5,003) was obtained from the Mount Sinai Data Warehouse. The progress, pathology and radiology notes of patients were used. We integrated various techniques of Natural Language Processing (NLP) and Artificial Intelligence (AI) and developed a system to automatically abstract outcomes. The corresponding images, biopsies and lines of treatments (LOTs) were abstracted as attributes of outcomes. This system includes four information models: 1. Customized NLP annotator model: preprocessor, section detector, sentence splitter, named entity recognition, relation detector; CRF and LSTM methods were applied to recognize entities and relations. 2. Clinical Outcome container model: biopsy evidence extractor, lines of treatment detector, image evidence extractor, clinical outcome event recognizer, date detector, and temporal reasoning; Domain-specific rules were crafted to automatically infer outcomes. 3. Document Summarizer; 4. Longitudinal Outcome Summarizer. Results: To evaluate the outcomes abstracted, we curated a subset (n = 792) from patient cohort for which LOTs were available. About 61% of the outcomes identified were supported by radiologic images (time window = ±14 days) or biopsy pathology results (time window = ±100 days). In 91% (720/792) of patients, Progression was abstracted within a time window of 90 days prior to first-line treatment. Also, 72% of the Progression events identified were accompanied by a downstream event (e.g., treatment change or death). We randomly selected 250 outcomes for manual curation, and 197 outcomes were assessed to be correct (precision = 79%). Moreover, our automated abstraction system improved human abstractor efficiency to curate outcomes, reducing curation time per patient by 90%. Conclusions: We have demonstrated the feasibility and effectiveness of NLP and AI approaches to abstract outcomes from lung cancer EHR data. It promises to automatically abstract outcomes and other clinical entities from notes across all cancers.

e21514 Background: Although some advanced non-small cell lung cancer (aNSCLC) patients respond favorably to immune checkpoint inhibitors, a larger portion either do not respond or experience negative side effects. At present, there is no consensus on how to best predict response to immunotherapy (IO) agents. Complete blood count (CBC) data have shown correlation with both progression and survival for patients treated with IO. However, previous studies have been relatively small, or did not focus on aNSCLC. Given CBCs are measured routinely for every patient throughout their course of therapy, we used real-world data (RWD) to assess the biomarker potential of CBCs in predicting response at the start and through the course of therapy. Methods: We performed a retrospective cohort study of 11,138 lung cancer patients treated at hospitals in the Mount Sinai Health System of whom 249 aNSCLC and were treated with anti-PD1/PD-L1 therapy. We extracted and harmonized medical record data using an automated abstraction engine and evaluated the impact of changes in CBCs, including eosinophils, lymphocytes, monocytes, neutrophils, red blood cells and platelets after treatment initiation. Time to treatment discontinuation (TTD) and overall survival (OS) were our clinical endpoints. Results: Abnormal baseline levels of monocytes, neutrophils, and red bloods cells were associated with OS; counts taken between 2-8 weeks were modestly to significantly associated with poorer OS, with hazard ratios ranging from 1.7-4.7. Most strikingly, neutrophil levels above 7 (HR = 4.3, p = 3.2*10−10, 95% CI = 2.7-6.7) and a neutrophil-to-lymphocyte ratio > 5 (HR = 3.4, p = 8.0*10−8, 95% CI = 2.2-5.4) were associated with poorer OS. For those with normal neutrophil levels at baseline, progressing to high levels at 2-8 weeks resulted in a hazard for OS of 5.0 (p = 4.4*10−8, 95% CI = 2.8-9.0). Those with progression are more likely to have high neutrophil levels at baseline, or an increase in neutrophil levels from normal to high levels between baseline and 2-8 weeks or 8-14 weeks. Our results remained consistent even after controlling for factors that may affect neutrophil levels including concurrent chemotherapy, administration of neutropenia medications, or concurrent infections. Repeating the analysis with TTD as the primary endpoint showed similar but less significant trends. Conclusions: RWD showed changes in blood cell counts after anti-PD1/PD-L1 therapy were associated with OS in aNSCLC and show promise as a predictive biomarker.

e14054 Background: Matching clinical attributes (i.e. indications, lab tests, treatment regimens) of clinical trial eligibility criteria with real world patient data is extremely challenging. Attribute phenotyping is one of the key components of Trial2Patient, a customized system developed by Sema4 to find patients for clinical trials. Transforming treatment regimens to a standard ontology and encoding drugs with standard nomenclatures will facilitate the semantic retrieval of treatments mentioned in clinical trial criteria. This will also enable the interoperation between different data sources that is often required for fast-learning and scalable healthcare information system. Methods: Free text containing treatment regimen/medication terms were extracted and preprocessed from three sources: 1) clinical trials listed in a commercial database citeline.com, 2) clinical trials listed in clinicaltrials.gov, and 3) National Comprehensive Cancer Network (NCCN) Guidelines. The regimen terms such as neoadjuvant therapy for non-small cell lung cancer, checkpoint inhibitor, EGFR inhibitor, androgen deprivation therapy (ADT), among many others, were profiled by AI methods (i.e. pattern reorganization and rule-based) and knowledge engineering via Sema4’s in-house knowledge base (CAV), Pharmaprojects in citline.com and NCCN Guidelines. The drugs related to each regimen were identified and mapped to RxCUI via RxNorm ontology. Results: We identified 76 regimen terms for non-small cell lung cancer (NSCLC), small cell lung cancer (SCLC) and prostate cancer (e.g. PD-L1 ≥1% nonsquamous NSCLC, bone antiresorptive therapy for M1 castration resistant prostate cancer), and 14,476 drug-category pair (e.g. pembrolizumab is a PD-1 inhibitor, pembrolizumab is used as the third line and beyond systemic therapy for M1 CRPC). All drugs identified were mapped to RxCUI for real world patient matching. Conclusions: This approach systematically extracted and normalized regimens and medications mentioned in clinical trials in NSCLC, SCLC and prostate cancer to standard codes. These standardized data can be used in mapping treatment histories of a patient to the eligibility criteria for clinical studies or for identifying studies relevant to a patient. The outcome of profiling cancer treatment regimens through standard ontology RxNorm may be particularly valuable in cancer studies based on real-world evidence.

e20642 Background: While optimal sequencing of systemic therapy in aNSCLC is critical to achieve maximal clinical benefit, it is practically challenging to study tx sequencing through clinical trials. RWD allow retrospective, observational studies to examine tx patterns and associated clinical outcomes. Methods: 1,609 aNSCLC patients who received systemic therapies at Mount Sinai hospitals were analyzed for the number of line of therapy (LOT), therapeutic modalities (chemotherapy, targeted therapy and immunotherapy), and the sequence in which treatments were given when LOT > 1. Time to tx discontinuation (TTD) was used as a surrogate clinical endpoint for outcomes. Results: 578 of the 1,609 (36%) patients received more than one LOT. 356 (22%) received tyrosine kinase inhibitors (TKIs), and 297 (16%) received immune checkpoint inhibitors (CPIs). Kaplan-Meier analysis revealed that among 297 patients who received CPIs, median TTD was longer in the 1st line setting (295 days, 95% CI 169 to 523; n=132) than when LOT > 1 (169 days, 95% CI 113 to 269; n=165), although the difference was not statistically significant (P=0.092, log-rank test). No difference of TTD on TKIs was observed between LOT = 1 and LOT > 1 (P=0.51). With respect to tx sequencing, when patients (n=94) received TKIs as the 1st LOT, 60%, 35%, and 5% of them received another TKI, chemotherapy, or a CPI-containing regimen, respectively, as the 2nd LOT. Among patients (n=370) who progressed on 1st line platinum-based chemotherapy, 52%, 32%, and 16% received another chemo regimen, a CPI-containing regimen, or a targeted therapy, respectively, as the 2nd LOT; these percentages shifted significantly toward more CPIs (24%, 66%, 10% for chemo, CPI, targeted, respectively) when only 2016-2018 data were examined. In the 2nd line setting after platinum therapy, TTD was significantly longer in the CPI group (332 days, 95% CI 169-484) compared to the chemo group (88 days, 95% CI 65-100; P

e18141 Background: Clinical trials remain the most reliable means to evaluate drug efficacy and safety for evidence-based cancer care. However, there is a significant racial disparity among participants in clinical trials. For example, only 1-3% of the participants in registration trials of immune checkpoint inhibitors (CPIs) in advanced non-small cell lung cancers (aNSCLCs) were Black. RWD data enable assessment of whether clinical trial results can be generalized to broader populations. Methods: We analyzed clinical records of > 145,000 cancer patients treated at Mount Sinai hospitals. Therapeutic agents approved by FDA in recent years were assessed for responses across various race groups. Time to treatment discontinuation (TTD) was used as a surrogate clinical endpoint for outcomes. Results: Overall we did not observe significant differences in TTD among different race groups for the following drugs and indications we examined: palbociclib in metastatic breast cancer (mBC), EGFR tyrosine kinase inhibitors in aNSCLC, EGFR antibody cetuximab and panitumumab in metastatic colorectal cancer with wild type KRAS, abiraterone in metastatic castration-resistant prostate cancer (mCRPC), enzalutamide in mCRPC, sorafenib in unresectable hepatocellular carcinoma (HCC), CPIs in metastatic melanoma. For example, the median TTD of palbociclib in combination with fulvestrant or letrozole in post-menopausal women with HR+HER2- mBC was 181, 261, and 160 days in White (n = 114), Black (n = 55), other (n = 48) race groups, respectively (P = 0.61, log-rank test). Among patients with unresectable HCC treated with sorafenib, the median TTD was 64, 49, and 67 days in the White (n = 201), Black (n = 127) and other (n = 243) race groups, respectively (P = 0.70). However, when CPIs in aNSCLC were examined, we observed a significantly longer TTD in the Black group (median TTD not reached; n = 77) compare to the non-Black group (169 days, 95% CI 133-331; n = 211), P = 0.0049, median follow up 194 days. Conclusions: RWD showed there are no apparent differences of treatment response in various race groups for most new therapeutic agents. Preliminary results of CPIs in aNSCLC suggested a favorable response in the Black than the non-Black population.