CT-Based Radiomics for Predicting PD-L1 Expression in Non-small Cell Lung Cancer: A Systematic Review and...

Area of Science:

• Radiomics
• Medical Imaging
• Oncology

Background:

• Immunotherapy efficacy in non-small cell lung cancer (NSCLC) correlates with PD-L1 expression.
• Current PD-L1 assessment via immunohistochemistry is invasive and may miss tumor heterogeneity.

Purpose of the Study:

• To evaluate CT-based radiomics models for predicting PD-L1 expression status in NSCLC.
• To assess the quality and identify limitations of existing radiomics models.

Main Methods:

• Systematic literature search across major databases (Scopus, PubMed, Web of Science, Embase, IEEE) up to February 14, 2025.
• Meta-analysis of diagnostic test accuracy for models predicting PD-L1 tumor proportion scores (TPS1 and TPS50).
• Quality assessment using METRICS and QUADAS-2 tools.

Main Results:

• Meta-analysis included 11 studies (997 subjects).
• Pooled AUC for TPS1 models was 0.85 (sensitivity 0.76, specificity 0.79).
• Pooled AUC for TPS50 models was 0.88 (sensitivity 0.72, specificity 0.86).
• Significant risk of bias identified in QUADAS-2; average METRICS score was 58.1%.

Conclusions:

• CT-based radiomics is a potentially non-invasive tool for predicting PD-L1 expression in NSCLC.
• Methodological improvements are crucial for clinical generalizability and reliability.