Model Could Help Screen NSCLC Patients Likely to Benefit From Anti–PD-L1 Immunotherapy

A model developed by researchers in China could help identify patients with non–small cell lung cancer (NSCLC) most likely to benefit from anti– programmed death-ligand 1 (PD-L1) immunotherapy.

A new prediction model based on positron emission tomography (PET)/CT images and clinicopathological characteristics could help clinicians screen patients with non–small cell lung cancer (NSCLC) who would benefit from anti–programmed death-ligand 1 (PD-L1) immunotherapy. Results of the univariate analysis were published in Frontiers in Oncology.

“Early diagnosis and treatment play a key role in improving the 5-year survival rate of lung cancer,” the researchers explained, while in recent years the development and advancement of immunotherapy has become an important area of research for the disease.

Previous studies have shown immunotherapy against PD-L1 yielded favorable results in patients with NSCLC, especially in those with high PD-L1 expression, they added. However, measuring expression levels of PD-L1 can pose challenges and new methodologies could be beneficial.

The combination of functional-metabolic and morphological imaging and F18-fluorodeoxyglucose-PET/CT (18F-FDG PET/CT) is the most advanced noninvasive form of imaging technology and can reflect—up to a certain extent—the glucose uptake level of tissues, the authors wrote.

Using data from a retrospective study of patients with NSCLC who underwent a combined imaging protocol of 18F-FDG PET/CT between January 2019 and March 2021 in China, the researchers set out to develop a prediction model incorporating the radiomic signature and clinicopathologic risk factors for individual prediction of PD-L1 expression in this population.

Of the 255 patients included in the analysis, 170 made up the training cohort and 85 made up the validation cohort; there were no statistically significant differences in clinical characteristics between the 2 cohorts. All patients underwent standard 18F-FDG PET/CT before biopsy or surgery, and those with other types of cancer or incomplete clinical datasets were excluded from the analysis.

The majority of patients had adenocarcinoma (n = 188; 73.7%), and the remainder had squamous cell carcinoma (n = 67; 26.3%).

Immunohistochemistry (IHC) assays were used to determine PD-L1 expression and cutoff values of 1% and 50% were employed. The researchers extracted 80 radiomic features in the training dataset.

They found:

  • In the univariate analysis, the expression of PD-L1 in lung tumors was significantly correlated with the radiomic signature, histologic type, Ki-67, maximum standardized uptake value, metabolic tumor volume, and total lesion glycolysis (P < .05 for all)
  • Expression of PD-L1 was not correlated with age, tumor/node/metastasis stage, and history of smoking (P > .05)
  • The prediction model for PD-L1 expression level over 1% and 50% that combined the radiomic signature and clinicopathologic features resulted in an area under the curve (AUC) of 0.762 and 0.814, respectively
  • For the prediction of PD-L1 expression over 1%, the AUC scores were 0.754 (95% CI, 0.696-0.805), 0.636 (95% CI, 0.574-0.695), and 0.757 (95% CI, 0.699-0.808) for features derived from radiomics, clinics, and the combined model, respectively
  • For the prediction of PD-L1 expression over 50%, the AUC scores were 0.762 (95% CI, 0.705-0.813), 0.672 (95% CI, 0.611-0.730), and 0.814 (95% CI, 0.761-0.860) for features derived from radiomics, clinics, and the combined model, respectively

“The combination of the radiomic signature and clinicopathologic risk factors presented a better diagnostic efficacy compared with the simple radiomic signature or clinical feature model,” the authors wrote. Currently, radiomics is rarely used to predict PD-L1 expression in NSCLC based on PET/CT images and clinicopathologic risk.

The researchers hypothesized CT-extracted features performed better than PET features in both the 1% and 50% groups because the density resolution of the PET image was worse than that of the CT image. This could affect the extracting and selection of meaningful radiomic features, underscoring the importance of a combined approach.

Future investigations are warranted to validate the findings, they noted, adding that the study’s single-center retrospective nature and small sample size mark limitations.

“In the present study, we established a prediction model based on PET/CT images and clinicopathological characteristics to predict the expression of PD-L1 in NSCLC patients and provided a novel strategy for clinicians to screen the patients who could benefit from anti-PD-L1 immunotherapy,” they concluded.

Reference:

Li J, Ge S, Sang S, Hu C, and Deng S. Evaluation of PD-L1 expression level in patients with non-small cell lung cancer by 18F-FDG PET/CT radiomics and clinicopathological characteristics. Front Oncol. Published online December 16, 2021. doi:10.3389/fonc.2021.789014