Enter Note Done

Media

Decrease font size Increase font size

EVALUATING CLINICAL UTILITY OF A UIP GENOMIC CLASSIFIER IN SUBJECTS WITH AND WITHOUT A HRCT PATTERN OF UIP

View Session Add to Schedule

Print Abstract


Presenting Author(s)
Jonathan ChungCo-Author(s)
Daniel Pankratz, Steven Nathan, Lori Lofaro, Gerard Criner, Fernando Martinez, Kevin Flaherty, David Lynch, Thomas Colby, Jeffrey Myers, Steve Groshong, Mark Steele, Giulia Kennedy, Kevin Brown, Ganesh Raghu, Yoonha Choi, Hannah Neville, Jing Huang, Sadia Benzaquen, David Lederer, Brandon Larsen, Karel Calero, Lars Hagmeyer, Amy Case, Umair Gauhar, Navdeep Rai, Murali Ramaswamy, John Davis, Neil Barth, Patric Walsh
Purpose:
Diagnostic evaluation for idiopathic pulmonary fibrosis (IPF) includes radiology and multi-disciplinary review (MDD). The Envisia Genomic Classifier (EGC), a 190-gene machine learning classifier using transbronchial biopsies, provides a binary prediction of UIP or non-UIP. The EGC shows sensitivity of 70% and specificity of 88% against a histopathology reference standard. We evaluated the utility and confidence of IPF diagnosis informed by EGC results.

Methods:
BRAVE is a multicenter observational study of patients with suspected interstitial lung disease. 94 BRAVE subjects with HRCT scans, central radiology review, central histopathology diagnoses, and EGC results were evaluated. Two blinded MDDs including an ILD pulmonologist, radiologist, and pathologist were convened. Two subject files were prepared with patient history and central radiology results. One version included histopathology images (pathology-MDD) and central histopathology evaluations while the other version included the EGC result (EGC-MDD). Subject versions were randomly assigned to MDDs. Each MDD independently determined a categorical clinical diagnosis of IPF versus non-IPF, with associated confidence. Clinical diagnoses were compared within radiology UIP categories.

Results:
IPF diagnoses by EGC-MDD and pathology-MDD were concordant (>86% agreement). Central radiology diagnosed 27% (25/94) of patients with HRCT-UIP (definite, probable or possible UIP) and 73% (69/94) as inconsistent with UIP. 17 of 25 with HRCT-UIP (68%) were diagnosed with IPF by both EGC-MDD and pathology-MDD. 94% of EGC-MDD IPF diagnoses were confident, compared to 53% with histopathology (p=0.008). For 69 subjects with radiology inconsistent with UIP, EGC-MDD and histopathology-MDD agreement is 90%, with 61 diagnosed with non-IPF and 1 diagnosed with IPF in both.

Conclusions:
EGC results support confident IPF diagnoses in subjects with HRCT-UIP. When radiology is inconsistent with UIP, EGC-MDD diagnoses are concordant with histopathology-MDD diagnoses of IPF vs. non-IPF.

Clinical Implications:
The Envisia Genomic Classifier enhances diagnostic confidence in IPF, with high concordance to multidisciplinary diagnoses informed by pathology.