Search

Rationale: Studies of bronchoscopy have reported diagnostic yield (DY) using different calculation methods, which has hindered comparisons across studies. Objectives: To quantify the effect of the variability of four methods on DY estimates of bronchoscopy. Methods: We performed a simulation-based analysis of patients undergoing bronchoscopy using variations around base case assumptions for cancer prevalence (60%), distribution of nonmalignant findings, and degree of follow-up information at a fixed sensitivity of bronchoscopy for malignancy (80%). We calculated DY, the rate of true positives and true negatives (TNs), using four methods. Method 1 considered malignant and specific benign findings at index bronchoscopy as true positives and TNs, respectively. Method 2 included nonspecific benign findings as TNs. Method 3 considered nonspecific benign findings cases as TNs only if follow-up confirmed benign disease. Method 4 counted all cases with a nonmalignant diagnosis as TNs if follow-up confirmed benign disease. A scenario analysis and probabilistic sensitivity analysis were conducted to demonstrate the effect of parameter estimates on DY. A change in DY of >10% was considered clinically meaningful. Results: Across all pairwise comparisons of the four methods, a DY difference of >10% was observed in 76.7% of cases (45,992 of 60,000 comparisons). Method 4 resulted in DY estimates that were >10% higher than estimates made with other methods in >90% of scenarios. Variation in cancer prevalence had a large effect on DY. Conclusions: Across a wide range of clinical scenarios, the categorization of nonmalignant findings at index bronchoscopy and cancer prevalence had the largest impact on DY. The large variability in DY estimates across the four methods limits the interpretation of bronchoscopy studies and warrants standardization.

Rationale: Stereotactic body radiation therapy (SBRT) is the standard of care for inoperable early stage non-small cell lung cancer (NSCLC). Use of image guided thermal ablation (IGTA; including microwave ablation [MWA] and radiofrequency ablation [RFA]) has increased in NSCLC, however there are no studies comparing all three., Objective: To compare the efficacy of IGTA (including MWA and RFA) and SBRT for the treatment of NSCLC., Methods: Published literature databases were systematically searched for studies assessing MWA, RFA, or SBRT. Local tumor progression (LTP), disease-free survival (DFS), and overall survival (OS) were assessed with single-arm pooled analyses and meta-regressions in NSCLC patients and a stage IA subgroup. Study quality was assessed with a modified methodological index for non-randomized studies (MINORS) tool., Results: Forty IGTA study-arms (2,691 patients) and 215 SBRT study-arms (54,789 patients) were identified. LTP was lowest after SBRT at one and two years in single-arm pooled analyses (4% and 9% vs. 11% and 18%) and at one year in meta-regressions when compared to IGTA (OR = 0.2, 95%CI = 0.07-0.63). MWA patients had the highest DFS of all treatments in single-arm pooled analyses. In meta-regressions at two and three-years, DFS was significantly lower for RFA compared to MWA (OR = 0.26, 95%CI = 0.12-0.58; OR = 0.33, 95%CI = 0.16-0.66, respectively). OS was similar across modalities, timepoints, and analyses. Older age, male patients, larger tumors, retrospective studies, and non-Asian study region were also predictors of worse clinical outcomes. In high-quality studies (MINORS score ≥ 7), MWA patients had better clinical outcomes than the overall analysis. Stage IA MWA patients had lower LTP, higher OS, and generally lower DFS, compared to the main analysis of all NSCLC patients., Conclusions: NSCLC patients had comparable outcomes after SBRT and MWA, which were better than those with RFA., Competing Interests: Declaration of Competing Interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: NF, AN, GW, YZ, and AL are employees of CRG-EVERSANA Canada, which was contracted by Johnson and Johnson Inc. to perform the SLR and analyses. IK, BL, SG, MZ, and PS are employees of Johnson and Johnson Inc. IK, BL, SG, and MZ hold stock in Johnson and Johnson Inc. PL, CN and MP are consultants for Johnson and Johnson Inc. PL is a consultant for HistoSonics. PL is a consultant for Elucent Medical. CN and MP are consultants for Medtronic. MP is a consultant for Intuitive. MP is a consultant for Bodyvision. MP is a consultant for Biodesix. MP is a consultant for Philips. MP is a consultant for Noah Medical. MP is a consultant for AstraZeneca. PL received a research grant from HistoSonics. PL received a research grant from Siemens Medical. CN and MP received research grants from Johnson and Johnson Inc. CN and MP received research grants from Medtronic. CN received a research grant from Siemens Healthineer. MP received a research grant from Intuitive. MP received a research grant from Bodyvision. MP received a research grant from Biodesix. MP received a research grant from Philips. CN and MP have been a speaker for Johnson and Johnson Inc. CN has been a speaker for Medtronic. CN has been a speaker for Siemens Healthineer. MP has been a speaker for Intuitive. MP has been a speaker for Bodyvision. MP has been a speaker for Biodesix. MP has been a speaker for AstraZeneca. PL and CN are advisory committee members for Johnson and Johnson Inc. MP is an advisory committee member for AstraZeneca. MP is an advisory committee member for Pfizer. PL holds stock in McGinley Orthopedic Innovations. PL holds stock in HistoSonics. PL holds stock in Elucent Medical., (Copyright © 2023 The Author(s). Published by Elsevier B.V. All rights reserved.)

Background: Ground-glass opacities (GGOs) appearing in computed tomography (CT) scans may indicate potential lung malignancy. Proper management of GGOs based on their features can prevent the development of lung cancer. Electronic health records are rich sources of information on GGO nodules and their granular features, but most of the valuable information is embedded in unstructured clinical notes., Objective: We aimed to develop, test, and validate a deep learning-based natural language processing (NLP) tool that automatically extracts GGO features to inform the longitudinal trajectory of GGO status from large-scale radiology notes., Methods: We developed a bidirectional long short-term memory with a conditional random field-based deep-learning NLP pipeline to extract GGO and granular features of GGO retrospectively from radiology notes of 13,216 lung cancer patients. We evaluated the pipeline with quality assessments and analyzed cohort characterization of the distribution of nodule features longitudinally to assess changes in size and solidity over time., Results: Our NLP pipeline built on the GGO ontology we developed achieved between 95% and 100% precision, 89% and 100% recall, and 92% and 100% F 1 -scores on different GGO features. We deployed this GGO NLP model to extract and structure comprehensive characteristics of GGOs from 29,496 radiology notes of 4521 lung cancer patients. Longitudinal analysis revealed that size increased in 16.8% (240/1424) of patients, decreased in 14.6% (208/1424), and remained unchanged in 68.5% (976/1424) in their last note compared to the first note. Among 1127 patients who had longitudinal radiology notes of GGO status, 815 (72.3%) were reported to have stable status, and 259 (23%) had increased/progressed status in the subsequent notes., Conclusions: Our deep learning-based NLP pipeline can automatically extract granular GGO features at scale from electronic health records when this information is documented in radiology notes and help inform the natural history of GGO. This will open the way for a new paradigm in lung cancer prevention and early detection., (©Kyeryoung Lee, Zongzhi Liu, Urmila Chandran, Iftekhar Kalsekar, Balaji Laxmanan, Mitchell K Higashi, Tomi Jun, Meng Ma, Minghao Li, Yun Mai, Christopher Gilman, Tongyu Wang, Lei Ai, Parag Aggarwal, Qi Pan, William Oh, Gustavo Stolovitzky, Eric Schadt, Xiaoyan Wang. Originally published in JMIR AI (https://ai.jmir.org), 01.06.2023.)

Background: Robot-assisted bronchoscopy (RAB) is among the newest bronchoscopic technologies, allowing improved visualization and access for small and hard-to-reach nodules. RAB studies have primarily been conducted at academic centers, limiting the generalizability of results to the broader real-world setting, while variability in diagnostic yield definitions has impaired the validity of cross-study comparisons. The objective of this study was to determine the diagnostic yield and sensitivity for malignancy of RAB in patients with pulmonary lesions in a community setting and explore the impact of different definitions on diagnostic yield estimates., Methods: Data were collected retrospectively from medical records of patients ≥ 21 years who underwent bronchoscopy with the Monarch® Platform (Auris Health, Inc., Redwood City, CA) for biopsy of pulmonary lesions at three US community hospitals between January 2019 and March 2020. Diagnostic yield was calculated at the index RAB and using 12-month follow-up data. At index, all malignant and benign (specific and non-specific) diagnoses were considered diagnostic. After 12 months, benign non-specific cases were considered diagnostic only when follow-up data corroborated the benign result. An alternative definition at index classified benign non-specific results as non-diagnostic, while an alternative 12-month definition categorized index non-diagnostic cases as diagnostic if no malignancy was diagnosed during follow-up., Results: The study included 264 patients. Median lesion size was 19.3 mm, 58.9% were peripherally located, and 30.1% had a bronchus sign. Samples were obtained via Monarch in 99.6% of patients. Pathology led to a malignant diagnosis in 115 patients (43.6%), a benign diagnosis in 110 (41.7%), and 39 (14.8%) non-diagnostic cases. Index diagnostic yield was 85.2% (95% CI: [80.9%, 89.5%]) and the 12-month diagnostic yield was 79.4% (95% CI: [74.4%, 84.3%]). Alternative definitions resulted in diagnostic yield estimates of 58.7% (95% CI: [52.8%, 64.7%]) at index and 89.0% (95% CI: [85.1%, 92.8%]) at 12 months. Sensitivity for malignancy was 79.3% (95% CI: [72.7%, 85.9%]) and cancer prevalence was 58.0% after 12 months., Conclusions: RAB demonstrated a high diagnostic yield in the largest study to date, despite representing a real-world community population with a relatively low prevalence of cancer. Alternative definitions had a considerable impact on diagnostic yield estimates., (© 2023. The Author(s).)