Predictive value of spectral dual-detector computed tomography for PD-L1 expression in stage I lung adenocarcinoma: Development and validation of a nomogram

Abstract

Background Programmed death ligand 1 (PD-L1) expression is a predictive biomarker for the efficacy of immune checkpoint inhibitors in treating early-stage lung adenocarcinoma (LA). Given that studies on the relationship between PD-L1 expression and spectral computed tomography (SDCT) quantification, qualitative parameters, and clinical biomarkers are limited, we aimed to explore this aspect for stage I LA and developed a nomogram to aid in preoperative individualized identification. Methods We analyzed SDCT parameters and PD-L1 expression in patients with invasive non-mucinous LA diagnosed by postoperative pathology. Based on the expression of PD-L1 (threshold, 1%), the patients were divided into positive and negative groups. Results Overall, 356 participants were included in the analysis. After univariate analysis, 18 parameters were associated with PD-L1 expression. Multiple regression analysis was performed on the significant parameters with AUC > 0.6, further, CT.value, CT40kev.a, Zeff.a, ED.a, ED.v, TK1, and Dsolid were found to be independent risk factors for PD-L1 expression in stage I LA. The clinical decision-making curve of the SDCT parameter-clinical nomogram had the highest net benefit over a wide range of threshold probabilities, and the clinical impact curve indicated its clinical validity. Model comparisons revealed the higher discriminatory accuracy of the nomogram than that of any individual variable. Conclusion Quantitative parameters based on SDCT can predict the expression of PD-L1 in early-stage LA, with Zeff.a playing a prominent role; our nomogram showed excellent predictive performance and good calibration. This approach may promote better non-invasive prediction of PD-L1 expression.