Texture Features on CT May Predict Overall Survival for NSCLC Treated with Immunotherapy

By Michael Vlessides, /alert Contributor
Save to PDF By

New findings from the University of Miami Health System have shown that texture features on pre-immunotherapy computed tomography (CT) imaging may help predict the duration of overall survival for patients with advanced non small-cell lung cancer (NSCLC) who are treated with nivolumab monotherapy. 

In a presentation at the 2019 the American Society of Clinical Oncology Annual Meeting (abstract e20720), the investigators explained that targeted therapies can prove ineffective in most NSCLC patients, while disease response rates are less than 20% for patients with advanced NSCLC on immuno-monotherapy. Nevertheless, predictive models that help distinguish patients who do and do not respond to immunotherapy may help guide clinical practice. 


MRI. Source: Getty

“Texture analysis,” they wrote, “is a data-mining tool used to identify intensity patterns in diagnostic imaging. We hypothesized that texture features on pre-immunotherapy CT imaging can be associated with clinical outcomes for patients with advanced NSCLC treated with nivolumab.”

To help test this theory, the researchers turned to an IRB-approved database containing 159 individuals with advanced NSCLC who had been treated with nivolumab monotherapy. From this group they chose the 20 patients with the longest overall survival and 20 with the shortest overall survival for retrospective analysis. 

As part of the analysis, the last pre-immunotherapy positron emission tomography-CT for each patient was transferred to a computer program for segmentation. The software helped delineate FDG-avid intrathoracic tumors per guidelines from the Radiation Therapy Oncology Group. 

Ninety-two texture features within each tumor were analyzed for association with the study’s primary endpoint, overall survival. For purposes of the analysis, overall survival time was dichotomized to either 1) less than one year or 2) more than one year. A univariate logistic regression model estimated the odds ratio, 95% confidence interval, and p value for each feature. 

The analysis revealed that 11 of the 92 texture features showed significant association with overall survival time, with p values from 0.009 to 0.044. Of these, seven texture features exhibited large effect, with odds ratios of either less than 0.5 or greater than 1.5. 

What’s more, an additional group of 15 texture features trended toward statistical significance, with p values ranging from 0.05 to 0.10. In total, 26 of the 92 texture features either showed significant associations or trended toward such associations with duration of overall survival. 

Although the researchers acknowledged that the findings are preliminary in nature, they seemed encouraged by the early findings. “We are in the process of validating a multivariate predictive model,” the authors wrote.

“Future directions,” they added, “include expansion of this study across the full database, survival analyses and correlation of texture features with tissue biology.”

 

© 2024 /alert® unless otherwise noted. All rights reserved.
Reproduction in whole or in part without permission is prohibited.
Privacy Policy | Terms of Use | Editorial Policy | Advertising Policy