Noninvasive Image Texture Analysis May Provide Tumor Heterogeneity Information With Prognostic Potential in NSCLC
Noninvasive quantitative computed tomography–based texture analysis can accurately differentiate the presence of a KRAS mutation from pan–wild-type non–small cell lung cancer (NSCLC), based on the results of a study in PLOS ONE. Weiss et al reported that this tumor heterogeneity information may be associated with patient survival, thus possibly helping to identify appropriate molecularly defined targeted therapy for patients with this type of lung cancer.
Molecular characterization of tumors may aid in the identification of a potential vulnerability for targeted therapy. Although most modalities to assess tumors require collection of tissue by invasive means, a new noninvasive imaging approach centering on quantitative computed tomography–based texture analysis may help to evaluate tumor characteristics and shed light on the tumor’s molecular genotype. Weiss and colleagues examined the potential of this analysis to provide tumor heterogeneity as well as prognostic information in NSCLC.
Host of Oncogenes Under Study
Formalin-fixed, paraffin-embedded lung tumor tissues were obtained from patients with early-stage non–small cell lung cancer. The tumor DNA was analyzed on the LungCarta Panel, with 249 multiplexed assays interrogating 213 somatic mutations in 26 oncogenes and/or tumor-suppressor genes. The genes under study were AKT1, ALK, BRAF, DDR2, EGFR, EPHA3, EPHA5, ERBB2, FGFR4, JAK2, KRAS, MAP2K1, STK11, MET, NOTCH1, NRAS, NRF2, NTRK1, NTRK2, NTRK3, PIK3CA, PTCH1, PTEN, PTPN11, PTPRD, and TP53.
Of the 48 total cases selected for imaging analysis, 27 were KRAS-mutant and 21 were pan–wild-type for all 26 oncogenes and tumor-suppressor genes. The mean age of the patients was 70.4 years (range, 45.1–85.1 years), and 29 were men. Nearly all of the patients (46) had a smoking history. Of the non–small cell lung tumors, 33 were adenocarcinomas and 10 were squamous cell carcinomas.
Results and Clinical Implications
Positive skewness (a measure of asymmetry of the histogram; P = .031) with fine-texture analysis and lower kurtosis (a measure of peakedness and tailedness; P = .009) with coarse-texture analysis were significantly associated with the presence of a KRAS mutation. The investigators suggested that these features may indicate more focal fibrosis, as a result of the hostile microenvironment within a KRAS-driven tumor. In previous studies, this fibrosis has been linked to inferior outcomes in both squamous cell carcinoma and adenocarcinoma of the lungs.
From the Monte Carol analysis, values for the five-node decision tree for sensitivity, specificity, and accuracy were 96.3% (95% confidence interval [CI] = 78.1–100), 81.0% (95% CI = 50.5–97.4), and 89.6% (95% CI = 72.9–97.0), respectively.
There appeared to be no major differences in overall survival and disease-free survival between patients with lung tumors with a KRAS mutation and those that were pan–wild-type. The investigators noted that this finding was in line with other published data in early-stage NSCLC.
Standard deviation, skewness, and kurtosis were found to be significantly associated with overall survival. A lower kurtosis value was linked with poorer survival than a higher kurtosis value (P = .048).
Mean (a measure of average brightness) and kurtosis were significantly associated with disease-free survival. A lower mean was related to shorter disease-free survival (P = .009). In addition, lower kurtosis values were related to shorter disease-free survival compared with higher kurtosis values (P = .049).
Turning to the prognostic effect of this noninvasive imaging approach, the investigators demonstrated that standard deviation, kurtosis, skewness, and mean were significantly associated with overall survival. With fine-texture analysis, in those with a KRAS-mutant tumor, higher standard deviation was associated with shorter overall survival, whereas there were no major differences in overall survival with standard deviation in those who had pan–wild-type tumors. With course-texture analysis, in those with KRAS-mutant tumors, lower kurtosis was significantly associated with shorter overall survival (P = .044), whereas there were no major differences in overall survival with kurtosis in those who had pan–wild-type tumors. Without filtration in those with a KRAS mutation, a lower mean was significantly associated with a shorter disease-free survival (P = .015), but no such difference was seen in those with a pan–wild-type tumor.
Closing Thoughts
To characterize non–small cell lung tumors noninvasively would be of great assistance to clinical oncologists in making treatment decisions, noted the investigators. However, validation of these findings is the next requirement.
“With further refinement of treatment decisions utilizing molecular analysis, rapid identification of molecular targets associated with resistance or responsiveness to molecularly based therapies in a noninvasive manner for advanced [NSCLC] can improve treatment efficiencies by providing go-no go decisions faster than or complementary with traditional and/or evolving laboratory assay techniques,” concluded the investigators.
Glen J. Weiss, MD, MBA, of the Translational Genomics Research Institute, Phoenix, is the corresponding author of the article in PLOS ONE.
This study was supported by TGen Foundation, SHC Foundation, Flinn Foundation, and Sequenom, Inc. Funding was also provided by the UK Department of Health’s NIHR Biomedical Research Centre.
Dr. Weiss is on the speaker’s bureau for Genentech, Pfizer, Celgene, and Eli Lilly and has received honoraria from Quintiles and Medscape. Ronald L. Korn, MD, is an employee and stockholder in Imaging Endpoints. Balaji Ganeshan, PhD, and Kenneth A Miles, MD, FRCR, are shareholders in TexRAD Ltd, a company developing and marketing the tumor textural analysis software described here.
The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.
