A multimodal approach combining blood-based biomarkers and imaging data can predict which patients with advanced non-small cell lung cancer (NSCLC) will experience durable clinical benefit (DCB) from immune checkpoint inhibitors (ICIs), according to a new study published in Nature Communications. The research highlights the potential of early circulating tumor DNA (ctDNA) dynamics and normalized blood-based tumor mutational burden (bTMB) to identify responders and non-responders to immunotherapy, potentially improving treatment decisions and patient outcomes.
Identifying Predictive Biomarkers
The study, a multicenter prospective observational trial, screened 328 advanced NSCLC patients, ultimately including 83 patients without actionable EGFR or ALK mutations who underwent first-line immunotherapy. Researchers analyzed pretreatment ctDNA, bTMB, and peripheral blood mononuclear cell (PBMC) transcriptomes, along with early on-treatment ctDNA dynamics, to identify factors associated with DCB, defined as progression-free survival (PFS) of at least 6 months.
Pretreatment ctDNA was detected in 94% of patients, with TP53, KRAS, LRP1B, KEAP1, and MLL2 as the most frequently mutated genes. The DCB cohort exhibited numerically lower ctDNA concentration and significantly higher bTMB. Integrating ctDNA concentration and bTMB as ctDNA-normalized bTMB revealed that patients with higher normalized bTMB (cutoff at 0.0975) had significantly better clinical outcomes (median PFS: 11.7 vs 8.8 months, P = 0.03, hazard ratio [HR]: 0.47, 95% CI: 0.26–0.85).
Early ctDNA Dynamics as a Key Predictor
Early on-treatment ctDNA dynamics proved to be a strong predictor of ICI response. The change in ctDNA (ΔctDNA) was significantly lower in the DCB group. Patients with persistently undetected or decreased ctDNA had significantly longer PFS (median PFS: 25.7 vs 11.2 vs 4.0 months, P = 0.001). Defining a ctDNA molecular response as an 80% decrease in ctDNA concentration from pretreatment, 83.9% of patients achieving DCB were classified as ctDNA molecular responders, while 70.4% of non-DCB patients were ctDNA non-responders (P < 0.001). Early ctDNA dynamics outperformed all individual pretreatment factors, with patients achieving ctDNA molecular responses showing significantly better clinical outcomes (median PFS: 11.5 months vs 4.0 months, HR: 0.32, 95% CI: 1.27–5.40, P < 0.001).
Multimodal Model for Enhanced Prediction
Combining normalized bTMB with early on-treatment ctDNA dynamics in a multiparameter predictor significantly improved the accuracy of predicting ICI benefit. A logistic regression model, constructed using data from the study cohort and validated with data from the DIREct-On study, predicted DCB with an area under the curve (AUC) of 0.854, achieving 75.0% sensitivity and 88.9% specificity. The model demonstrated similar performance in the validation cohort, with an AUC of 0.798, sensitivity of 80.5%, and specificity of 67.6%.
Adding RECIST (Response Evaluation Criteria in Solid Tumors) response to the model further improved performance, increasing the AUC to 0.878, sensitivity to 79.2%, and specificity to 86.4% in the training set. The external validation set also showed improved performance, with an AUC of 0.887, sensitivity of 94.7%, and specificity of 85.3%. Patients with higher predicted scores had substantially longer PFS than those with lower scores in both cohorts.
Clinical Implications
These findings suggest that a non-invasive approach integrating blood-based biomarkers and imaging data can effectively predict durable clinical benefit from ICIs in NSCLC patients. Early assessment of ctDNA dynamics, combined with normalized bTMB and RECIST response, offers a promising strategy for identifying patients likely to benefit from immunotherapy and guiding treatment decisions. This approach could potentially improve patient outcomes by allowing for timely adjustments to treatment strategies based on early response assessment.
