1. DiGiulio, Sarah

Article Content

In non-small cell lung cancer (NSCLC), some patients' tumors shrink after immunotherapy regimens-and the patients have very good outcomes. Some patients' tumors don't respond at all. And some patients' tumors get worse. Patients in this third group are sometimes referred to as "hyperprogressors."

NSCLC. NSCLC... - Click to enlarge in new windowNSCLC. NSCLC

"There is relatively little known about what predisposes some patients to become hyperprogressors," noted Anant Madabhushi, PhD, Donnell Institute Professor of Biomedical Engineering and Director of the Center for Computational Imaging & Personalized Diagnostics at Case Western Reserve University. "Regardless, it is clear that these patients should be identified early and potentially avoid immunotherapy."


Previous work from Madabhushi's team had suggested that specific radiomic features and vascularity features of NSCLC tumors may predict immunotherapy response, so the researchers compared what those biomarkers looked like across CT scans of NSCLC tumors from patients who had good response to immunotherapy, no response to immunotherapy, and worse response after immunotherapy. The data was published online ahead of print in the Journal for ImmunoTherapy of Cancer (2020; doi: 10.1136/jitc-2020-001343).


The researchers did find specific patterns that allowed them to distinguish between responders, non-responders, and hyperprogressors just by looking at the pre-treatment tumor CT scans, explained Madabhushi. "This study showed that radiomic image features from CT scans of lung cancer patients prior to receiving immunotherapy could allow for prediction of treatment response in the future."


Radiomics refers to computer-extracted features or patterns from radiologic images, such as CT or MRI scans, Madabhushi noted. Examples of radiomic features of tumors, more specifically, are texture patterns that capture the heterogeneity of the tumor.


If these findings are replicated in larger multi-site studies, this approach could be used clinically to look for these biomarkers on patient CT scans after diagnosis and determine if a patient should receive immunotherapy or might better be served by conventional therapies, like chemotherapy or radiation, Madabhushi said.


Patterns in Tumor Nodules & Vascularity

For this study, the researchers analyzed CT scans from 109 patients with advanced NSCLC who had been given programmed cell death protein-1 (PD-1)/programmed death ligand-1 (PD-L1) inhibitors. Within this group of patients, 50 responded to treatment and 59 did not. With the group of non-responders, 19 were identified as hyperprogressors.


The researchers used AI technology to compare the radiomic and vascularity features of the patients' tumors and look for patterns among the various groups.


AI is used in many different contexts, Madabhushi explained, "The centerpiece of any AI system is a decision-making mechanism or a classifier. The classifier accepts features (patterns) extracted from the imaging data and employs the patterns to make a prediction."


In this study, the researchers looked at patterns in the tumor nodules (or texture) and the tumor vessels (tortuousness) across the CT scans to define the machine classifier, which in turn would allow the researchers to predict which category (responder, non-responder, or hyperprogressor) a patient fell into.


To put the findings to use, the researchers would "teach" computers to identify these patterns in CT scans taken from other NSCLC tumors, and thus predict immunotherapy treatment response.


Before these biomarkers can be used clinically to potentially identify which patients would do best with immunotherapy and who would not, the signature needs to be validated in a multi-institutional dataset and in a larger number of cases, Madabhushi noted. "Ideally, we are looking to validate this signature in retrospective completed clinical trials first and then start to move this signature into prospective clinical trials."


Additionally, the researchers want to explore whether this radiomic signature is able to predict mutational status of hyperprogressors, such as KRAS and STK11 mutations, Madabhushi noted.


Important Findings

"This is a very important study given that immunotherapy has been a preferable choice for patients who have not received immunotherapy in the first-line," stated Nagla Abdel Karim, MD, Professor of Medicine and Medical Director of the Georgia Cancer Center Clinical Trials Program at Augusta University. But first, identifying which patients become hyperprogressors-and not giving them immunotherapies-is important.


Having a biomarker like this one could allow patients who will do worse on immunotherapy to be excluded from clinical trials or given drugs they would ultimately do worse on, she explained.


Even though this biomarker isn't ready for clinical use yet, oncologists should be aware that there are NSCLC patients who will progress faster and do worse with some immunotherapies, Karim stated.


In addition to helping oncologists better identify these patients before they start therapy, the approach might also be helpful to monitor responses in patients undergoing immunotherapy (and make sure patients are progressing as expected), Madabhushi concluded.


Sarah DiGiulio is a contributing writer.