Automating Access to Real-World Evidence.

Introduction: Real-world evidence is important in regulatory and funding decisions. Manual data extraction from electronic health records (EHRs) is time-consuming and challenging to maintain. Automated extraction using natural language processing (NLP) and artificial intelligence may facilitate this process. Whereas NLP offers a faster solution than manual methods of extraction, the validity of extracted data remains in question. The current study compared manual and automated data extraction from the EHR of patients with advanced lung cancer. Methods: Previously, we extracted EHRs from 1209 patients diagnosed with advanced lung cancer (stage IIIB or IV) between January 2015 and December 2017 at Princess Margaret Cancer Centre (Toronto, Canada) using the commercially available artificial intelligence engine, DARWEN (Pentavere, Ontario, Canada). For comparison, 100 of 333 patients that received systemic therapy were randomly selected and clinical data manually extracted by two trained abstractors using the same accepted gold standard feature definitions, including patient, disease characteristics, and treatment data. All cases were re-reviewed by an expert adjudicator. Accuracy and concordance between automated and manual methods are reported. Results: Automated extraction required considerably less time (<1 day) than manual extraction (∼225 person-hr). The collection of demographic data (age, sex, diagnosis) was highly accurate and concordant with both methods (96%-100%). Accuracy (for either extraction approach) and concordance were lower for unstructured data elements in EHR, such as performance status, date of diagnosis, and smoking status (NLP accuracy: 88%-94%; Manual accuracy: 78%-94%; concordance: 71%-82%). Concurrent medications (86%-100%) and comorbid conditions (96%-100%), were reported with high accuracy and concordance. Treatment details were also accurately captured with both methods (84%-100%) and highly concordant (83%-99%). Detection of whether biomarker testing was performed was highly accurate and concordant (96%-98%), although detection of biomarker test results was more variable (accuracy 84%-100%, concordance 84%-99%). Features with syntactic or semantic variation requiring clinical interpretation were extracted with slightly lower accuracy by both NLP and manual review. For example, metastatic sites were more accurately identified through NLP extraction (NLP: 88%-99%; manual: 71%-100%; concordance: 70%-99%) with the exception of lung and lymph node metastases (NLP: 66%-71%; manual: 87%-92%; concordance: 58%) owing to analogous terms used in radiology reports not being included in the accepted gold standard definition. Conclusions: Automated data abstraction from EHR is highly accurate and faster than manual abstraction. Key challenges include poorly structured EHR and the use of analogous terms beyond the accepted gold standard definition. The application of NLP can facilitate real-world evidence studies at a greater scale than could be achieved with manual data extraction.

Effects of Ethnicity on Outcomes of Patients With EGFR Mutation-Positive NSCLC Treated With EGFR Tyrosine Kinase Inhibitors and Surgical Resection.

INTRODUCTION: In addition to the higher prevalence of EGFR mutations found among lung cancer cases in East Asian patients, it is unclear whether there are differences in treatment outcomes by ethnicity-that is, East Asian versus non-East Asian. METHODS: Patients diagnosed with EGFR-mutant lung cancer between January 2004 and October 2014 at a single center were reviewed. Data captured included demographics, tumor and treatment information, and survival. Survival of patients of East Asian and non-East Asian ancestry was compared, including in the subgroup that received EGFR tyrosine kinase inhibitor (TKI) for advanced disease and in those with early-stage disease that underwent surgical resection. RESULTS: A total of 348 patients with EGFR-mutant NSCLC were identified. There was a higher proportion of nonsmokers among those of East Asian ethnicity. No significant difference in survival was seen between patients of East Asian and non-East Asian ethnicity, median 6.7 years (95% confidence interval [CI]: 5.4-not applicable) and 5.4 years (95% CI: 4.1-7.2), respectively (p = 0.09). Among 196 patients that received treatment with EGFR TKI, the median survival from TKI initiation was also similar for those of East Asian and non-East Asian ethnicity, 3.0 years (95% CI: 2.1-3.5) and 2.7 years (95% CI: 2.2-3.5), respectively. Among the early-stage patients that underwent surgical resection (n = 163), those of East Asian ethnicity had similar median recurrence-free survival from surgery compared with non-East Asian patients, 5.3 years (95% CI: 3.5-not applicable) and 5.1 years (95% CI: 3.3-7.2), respectively. CONCLUSIONS: In a cohort of patients with EGFR-mutant lung cancer with access to uniform standards of care, East Asian ethnicity was not associated with improved survival after treatment with EGFR TKI or surgical resection.

Tobacco exposure and immunotherapy response in PD-L1 positive lung cancer patients.

BACKGROUND: Tobacco exposure contributes to over 80 % of lung cancer cases. Smoking is associated with programmed death-ligand 1 (PD-L1) tumor expression and better outcomes from anti-programmed cell death protein 1 (anti-PD-1) therapy in patients with advanced non-small cell lung cancer (NSCLC). PD-L1 tumor expression is now routinely used to predict benefit from anti-PD-1 therapy in patients with advanced NSCLC. In this study, we explored the impact of smoking status on patient outcomes with anti-PD-1 therapy in addition to PD-L1 tumor expression. METHODS: A prospective real-world cohort of 268 patients with advanced NSCLC treated with anti-PD-1 monotherapy at the Princess Margaret Cancer Centre (PMCC) was used for this analysis. Logistic regression was performed to test factors associated with treatment response (RECIST v1.1), including PD-L1 tumour proportion score (TPS) and smoking status. RESULTS: Overall response rates (ORR) to immunotherapy were significantly higher in current and former smokers than never smokers (36 % vs 26 % vs 14 %; p = 0.02). In patients with PD-L1 tumour proportion score (TPS) ≥50 %, current smokers continued to experience better ORR to anti-PD-1 therapy than never smokers (58 % vs 19 %; p = 0.03). Current smoking was associated with higher response even after adjusting for level of PD-L1 TPS expression (adjusted odds ratio 5.9, 95 % CI 1.6-25.0, p = 0.03). Exploratory analysis demonstrated higher 1-year survival rates in smokers compared to never smokers (p = 0.003). CONCLUSIONS: Smoking remains an important factor associated with response to anti-PD-1 monotherapy. Advanced NSCLC patients with positive PD-L1 expression are more likely to respond to anti-PD-1 monotherapy if they are current smokers compared to never smokers.

Correlation of Neutrophil to Lymphocyte Ratio and Absolute Neutrophil Count With Outcomes With PD-1 Axis Inhibitors in Patients With Advanced Non-Small-Cell Lung Cancer.

INTRODUCTION: Programmed death-1 (PD-1) axis inhibitors have become standard therapy in advanced non-small-cell lung cancer (NSCLC). Response might be delayed and pseudo-progression occasionally occurs in patients who eventually benefit from treatment. Additional markers beyond programmed death ligand 1 (PD-L1) expression are needed to assist in patient selection, response evaluation, and treatment decisions. MATERIALS AND METHODS: The relationship between prospectively collected clinical outcomes (response, disease control rate [DCR], treatment duration, overall survival) and hematologic parameters (neutrophil to lymphocyte ratio [NLR], absolute neutrophil count [ANC], and platelet to lymphocyte ratio [PLR]) was explored retrospectively in advanced NSCLC patients treated with PD-1 axis inhibitors at a major cancer center from May 2013 to August 2016. Hematologic parameters at baseline and during treatment (week 2 or 3 and week 8) were included. RESULTS: Of 88 patients treated with PD-1 axis inhibitors, 22 (25%) experienced partial response. Baseline NLR ≤4 was associated with superior DCR (74% vs. 50%; P = .025), treatment duration (P = .037), time to progression (P = .053), and overall survival (P = .019), with no differential association according to PD-L1 tumor expression. Lower NLR and ANC during treatment were also associated with response to treatment (P = .025 and P = .017, respectively), and treatment duration (P = .036 and P = .008). No association was found between baseline PLR and DCR, response, treatment duration, nor overall survival. CONCLUSION: Baseline NLR ≤4 and lower NLR and ANC during treatment might correlate with disease control and treatment response and should be explored further as potential predictors of treatment benefit in larger studies.

Translation of Knowledge to Practice-Improving Awareness in NSCLC Molecular Testing.

BACKGROUND: Molecular testing in advanced lung cancer is standard in guiding treatment selection. However, population-wide implementation of testing remains a challenge. We developed a knowledge translation intervention to improve understanding among diagnostic specialists about molecular testing and appropriate diagnostic sampling in lung cancer. METHODS: Specialty-specific education programs were developed from existing literature and input from Canadian leaders in lung pathology, respirology, interventional radiology, thoracic surgery, radiation oncology, and medical oncology. The programs, including key messages, review of current data, existing guidelines, group discussion, and participant feedback, were administered at provincial and national specialty meetings. Participant knowledge was assessed before and after the intervention by using anonymous questionnaires. Molecular (EGFR) testing rates in Ontario were also evaluated before and after the intervention period. RESULTS: Ten programs were administered to diagnostic specialists, including respirologists, pathologists, thoracic surgeons, radiologists, radiation oncologists, and medical oncologists, with completion of 255 preintervention and 219 postintervention surveys. At baseline, 30% were unsure of tissue handling methods for molecular testing, 20% chose an incorrect technique, and half were unfamiliar with how to initiate testing. After intervention, specialist knowledge improved regarding tissue handling and appropriate fixation techniques and uncertainty decreased from 30% to 2% (p < 0.001). A 12% increase (relative increase 57%) in molecular (EGFR) testing requests in Ontario was observed over the intervention period (p = 0.0032). CONCLUSIONS: Significant knowledge gaps exist among diagnostic specialists regarding molecular testing and targeted therapy in lung cancer. This initiative significantly improved understanding of the importance and methods of successful molecular testing and correlated with increased testing rates.

Evolution of Symptom Burden of Advanced Lung Cancer Over a Decade.

INTRODUCTION: Lung cancer is associated with higher levels of symptom distress and unmet needs than other cancer types. We assessed changes in symptoms, function, understanding, and preferences of patients with advanced lung cancer over a 10-year period. MATERIALS AND METHODS: A 26-item self-administered questionnaire was used to assess symptom burden, functional impairment, knowledge of disease and treatment, and information preferences. The survey was administered to consecutive outpatients with advanced lung cancer first in 2002 and a second cohort in 2012. RESULTS: A total of 108 patients with advanced lung cancer were surveyed in 2002, and 100 in 2012. Rates of severe physical symptoms were similar over the 10-year period. The most common symptoms remained fatigue, cough, and dyspnea. One-third perceived major impairment of daily activities from lung cancer. Significant anxiety was reported by at least 20%; nearly a quarter reported being unable to meet family needs. More patients in 2012 received information on treatment benefits, side effects, and clinical trials. Only 40% reported having end-of-life wishes, and fewer than half had discussed these with their oncologist. Over time, more patients expressed a preference for treatment associated with increased survival even if it compromised quality of life. Half were interested in Internet-based resources, most in print media, and a growing number in telephone support. CONCLUSION: Patients with advanced lung cancer continue to experience significant symptom distress and unmet needs despite advances in treatment. Comprehensive assessment and symptom, psychological, financial, and information support remain key areas for improvement in the care of patients with advanced lung cancer.

Patients with Advanced Non-Small Cell Lung Cancer: Are Research Biopsies a Barrier to Participation in Clinical Trials?

OBJECTIVES: Clinical trials of therapies for non-small cell lung cancer (NSCLC) are increasingly requiring mandatory tumor samples or research biopsies, both of which are potential barriers to trial participation. We assessed the impact of performance of research biopsies on the enrollment of patients with advanced NSCLC in clinical trials. METHODS: The cases of patients with advanced NSCLC who had been evaluated for clinical trials of systemic therapy at the Princess Margaret Cancer Centre from January 2007 to March 2015 were reviewed. RESULTS: Of the 55 clinical trials identified, 38 required tumor samples for enrollment. Six mandated repeat biopsies, whereas 32 permitted use of archival samples. Trial participation was offered to 636 patients at 940 unique study encounters, with some patients enrolling in multiple trials. Of the patients in 549 encounters during which participation in a therapeutic trial was offered, 60% received study treatment. More patients received study treatment (83% versus 55%, p < 0.0001) and study treatment was started earlier (after 9 days versus after 16, p = 0.002) when the trial did not have a mandatory tissue sample requirement. A similar trend was noted for trials permitting use of archival tissue versus mandatory repeat biopsies. The most common barriers to trial enrollment included absence of a required biomarker (34%), withdrawal of consent (20%), deterioration or death (17%), other exclusion criteria (15%), and insufficient biopsy tissue (10%). CONCLUSION: A growing number of NSCLC trials are requiring tumor tissue for treatment eligibility, which appears to be a significant barrier to trial enrollment. Potential solutions include use of available diagnostic samples (e.g., cytology samples), development of peripheral blood assays for molecular markers, faster central laboratory testing turnaround time, and more resources for rapid biopsy.