Using multiplexed assays of oncogenic drivers in lung cancers to select targeted drugs.

IMPORTANCE: Targeting oncogenic drivers (genomic alterations critical to cancer development and maintenance) has transformed the care of patients with lung adenocarcinomas. The Lung Cancer Mutation Consortium was formed to perform multiplexed assays testing adenocarcinomas of the lung for drivers in 10 genes to enable clinicians to select targeted treatments and enroll patients into clinical trials. OBJECTIVES: To determine the frequency of oncogenic drivers in patients with lung adenocarcinomas and to use the data to select treatments targeting the identified driver(s) and measure survival. DESIGN, SETTING, AND PARTICIPANTS: From 2009 through 2012, 14 sites in the United States enrolled patients with metastatic lung adenocarcinomas and a performance status of 0 through 2 and tested their tumors for 10 drivers. Information was collected on patients, therapies, and survival. INTERVENTIONS: Tumors were tested for 10 oncogenic drivers, and results were used to select matched targeted therapies. MAIN OUTCOMES AND MEASURES: Determination of the frequency of oncogenic drivers, the proportion of patients treated with genotype-directed therapy, and survival. RESULTS: From 2009 through 2012, tumors from 1007 patients were tested for at least 1 gene and 733 for 10 genes (patients with full genotyping). An oncogenic driver was found in 466 of 733 patients (64%). Among these 733 tumors, 182 tumors (25%) had the KRAS driver; sensitizing EGFR, 122 (17%); ALK rearrangements, 57 (8%); other EGFR, 29 (4%); 2 or more genes, 24 (3%); ERBB2 (formerly HER2), 19 (3%); BRAF, 16 (2%); PIK3CA, 6 (<1%); MET amplification, 5 (<1%); NRAS, 5 (<1%); MEK1, 1 (<1%); AKT1, 0. Results were used to select a targeted therapy or trial in 275 of 1007 patients (28%). The median survival was 3.5 years (interquartile range [IQR], 1.96-7.70) for the 260 patients with an oncogenic driver and genotype-directed therapy compared with 2.4 years (IQR, 0.88-6.20) for the 318 patients with any oncogenic driver(s) who did not receive genotype-directed therapy (propensity score-adjusted hazard ratio, 0.69 [95% CI, 0.53-0.9], P = .006). CONCLUSIONS AND RELEVANCE: Actionable drivers were detected in 64% of lung adenocarcinomas. Multiplexed testing aided physicians in selecting therapies. Although individuals with drivers receiving a matched targeted agent lived longer, randomized trials are required to determine if targeting therapy based on oncogenic drivers improves survival. TRIAL REGISTRATION: clinicaltrials.gov Identifier: NCT01014286.

Deep Learning-Based Psoriasis Assessment: Harnessing Clinical Trial Imaging for Accurate Psoriasis Area Severity Index Prediction.

Introduction: Image-based machine learning holds great promise for facilitating clinical care; however, the datasets often used for model training differ from the interventional clinical trial-based findings frequently used to inform treatment guidelines. Here, we draw on longitudinal imaging of psoriasis patients undergoing treatment in the Ultima 2 clinical trial (NCT02684357), including 2,700 body images with psoriasis area severity index (PASI) annotations by uniformly trained dermatologists. Methods: An image-processing workflow integrating clinical photos of multiple body regions into one model pipeline was developed, which we refer to as the "One-Step PASI" framework due to its simultaneous body detection, lesion detection, and lesion severity classification. Group-stratified cross-validation was performed with 145 deep convolutional neural network models combined in an ensemble learning architecture. Results: The highest-performing model demonstrated a mean absolute error of 3.3, Lin's concordance correlation coefficient of 0.86, and Pearson correlation coefficient of 0.90 across a wide range of PASI scores comprising disease classifications of clear skin, mild, and moderate-to-severe disease. Within-person, time-series analysis of model performance demonstrated that PASI predictions closely tracked the trajectory of physician scores from severe to clear skin without systematically over- or underestimating PASI scores or percent changes from baseline. Conclusion: This study demonstrates the potential of image processing and deep learning to translate otherwise inaccessible clinical trial data into accurate, extensible machine learning models to assess therapeutic efficacy.