Predicting PD-L1 Expression in Non-Small Cell Lung Cancer With Machine Learning Tool

According to results of a performance comparison, a machine learning tool based on clinical and radiological features can accurately predict PD-L1 expression prior to neoadjuvant treatment in c-stage 1/2 non-small cell lung cancer (NSCLC) when PD-L1 expression is indeterminable by biopsy. 

In this study, 117 patients with c-stage 1/2 disease who underwent resection were identified and 3 951 radiomic features were extracted through identifying the tumor (within tumor contour), rim (contour ± 3 mm), and exterior (contour + 10 mm) on preoperative contrast computed tomography. Feature selection was performed using the Boruta algorithm and 3 prediction models (radiomics, clinical, and combined models) of tumor PD-L1 expression were constructed using the Random Forest algorithm. Performance was evaluated using 5-fold cross-validation and areas under the curve (AUCs) were compared using the Delong test. Patients were categorized into 2 study groups based on those who underwent biopsy (n = 38) and those who did not undergo biopsy (n = 79) and predictive ability of biopsy was compared to each prediction model. 

Among all study participants, results indicated that 33 patients had a PD-L1 ≥1%, and the mean AUC for the validation set in radiomics, clinical, and combined models were 0.80, 0.80, and 0.83 (P = .32 vs. clinical model), respectively. In patients who underwent biopsy, the diagnosis of malignancy was made in 22 patients and PD-L1 was measurable in 19 of those patients. The diagnostic accuracy of PD-L1 ≥1% from 19 determinable biopsies and 38 of all attempted biopsies was 0.68 and 0.34, respectively. In terms of the 3 validation sets, these were outperformed by machine learning by 0.71, 0.71 and 0.74, respectively. 

Kohei Hashimoto, MD, PhD, Cancer Institute Hospital, Japanese Foundation for Cancer Research, Tokyo, Japan, and coauthors stated, “Our machine learning could be an adjunctive tool in estimating PD-L1 expression prior to neoadjuvant treatment, particularly when PD-L1 is indeterminable with biopsy.” 

Source:

Hashimoto K, Murakami Y, Omura K, et al. Prediction of tumor PD-L1 expression in resectable non-small cell lung cancer by machine learning models based on clinical and radiological features: Performance comparison with preoperative biopsy. Clin Lung Cancer. Published online: August 10, 2023. doi: 10.1016/j.cllc.2023.08.010