From the Journals

Diagnostic test helps clinicians identify IPF with nonsurgical biopsy

View on the News

Molecular classification could help identify less clear-cut IPF cases

Use of a molecular classifier could be most helpful in situations where patients have atypical radiology results or in cases where multidisciplinary teams disagree on the diagnosis, Simon Hart, PhD, wrote in a related editorial.

According to the 2018 international guidelines for idiopathic pulmonary fibrosis, usual interstitial pneumonia certainty is defined as honeycombing seen on high-resolution CT (HRCT), probable if there is presence of traction bronchiectasis but not honeycombing, and indeterminate if there is no presence of usual interstitial pneumonia or another diagnosis. As radiologists “often disagree on HRCT patterns,” IPF sometimes becomes a working diagnosis based on progression of disease, Dr. Hart wrote. In these cases, molecular classifier samples could help identify IPF in patients who have undergone less invasive transbronchial lung biopsy.

Among patients for whom diagnoses using identical clinical features have different results, HRCT and pathology data, particularly in cases of nonspecific interstitial pneumonia and chronic hypersensitivity pneumonitis that follow a similar disease course to idiopathic pulmonary fibrosis, molecular classifier testing could help identify patients with these diseases so treatments such as to avoid treating these patients with anti-inflammatory or immunosuppressive therapy.

“It seems conceivable that in future interstitial lung diseases could be classified by a simple dichotomy: primarily scarring diseases characterized by molecular usual interstitial pneumonia to be treated with antifibrotics versus immune-driven conditions without usual interstitial pneumonia that need an anti-inflammatory approach,” he wrote.

Dr. Hart is from the respiratory research group at Castle Hill Hospital in Cottingham, England. These comments summarize his editorial in response to Raghu et al. (Lancet Respir Med. 2019 Apr 1. doi 10.1016/S2213-2600[19]30058-X). He reported receiving grants and support to attend conferences, and consultancy fees from Boehringer Ingelheim.



Researchers used a machine learning algorithm to identify a molecular signature for usual interstitial pneumonia in patients with suspected idiopathic pulmonary fibrosis, according to recent research published in the Lancet Respiratory Medicine.

Dr. Ganesh Raghu

The results of the molecular test, called the Envisia Genomic Classifier (Veracyte; San Francisco), had a high positive predictive value of proven usual interstitial pneumonia, and could be used in place of surgical lung biopsy to confirm a diagnosis of idiopathic pulmonary fibrosis (IPF), wrote Ganesh Raghu, MD, director at the Center for Interstitial Lung Diseases and professor of medicine at the University of Washington, Seattle, and his colleagues.* The Envisia Genomic Classifier recently received final Medicare local coverage determination for IPF diagnosis, according to a recent press release by Veracyte.

“IPF is often challenging to distinguish from other [interstitial lung disease], but timely and accurate diagnosis is critical so that patients with IPF can access therapies that may slow progression of the disease, while avoiding potentially harmful treatments,” Dr. Raghu stated in a press release. “Our results with molecular classification through machine learning [the Envisia classifier] are promising and, along with clinical information and radiological features in high-resolution CT imaging, physicians through multidisciplinary discussions, may be able to utilize the molecular classification as a diagnostic tool to make a more informed and confident diagnoses.”

The researchers prospectively recruited 237 patients from 29 centers in the United States and Europe who were evaluated with the Bronchial Sample Collection for a Novel Genomic Test for suspected interstitial lung disease and who underwent surgical biopsy, transbronchial biopsy, or cryobiopsy for sample collection. They used histopathology and RNA sequence data from 90 patients to create a training data set of an unusual interstitial pneumonia pattern for the machine learning algorithm.

The classifier found usual interstitial pneumonia diagnoses in 49 patients; the test had a specificity of 88% (95% confidence interval, 70%-98%) and a sensitivity of 70% (95% CI, 47%-87%). Of 42 patients with inconsistent or possible usual interstitial pneumonia identified from high-resolution CT imaging, there was a positive predictive value of 81% (95% CI, 54%-96%). When multidisciplinary teams made diagnoses with the molecular classifier data, there was a clinical agreement of 86% (95% CI, 78%-92%) with diagnoses made using histopathology data. In 18 cases of IPF, there was an improvement in diagnostic confidence using the molecular classifier data, with 89% of diagnoses designated as high confidence, compared with 56% of cases based on histopathologic data (P = .0339). In 48 patients with nondiagnostic pathology or nonclassifiable fibrosis histopathology, 63% of diagnoses with the molecular classifier data were high confidence, compared with 42% using histopathologic data (P = .0412).

This study was funded by Veracyte, creator of the Envisia Genomic Classifier. Some authors reported relationships with Veracyte and other companies.

SOURCE: Raghu G et al. Lancet Respir Med. 2019 Apr 1. doi: 10.1016/S2213-8587(19)300.

Correction, 4/25/19: An earlier version of this article misstated how the Envisia Genomic Classifier could be used. The Envisia test is not intended to replace high-resolution chest CT (HRCT). It is used when HRCT is inconclusive to help prevent patients from having to undergo invasive diagnostic procedures.

Next Article: