Evaluating Circulating Tumor DNA Clearance as an Early Intermediate End Point Predicting Clinical Benefit in Metastatic Nonsquamous NSCLC After First-Line Chemoimmunotherapy

Introduction In the first-line (1L) NSCLC setting, overall survival (OS) as a trial end point is increasingly challenging due to longer survival and confounding from novel later-line therapies. This study aims to evaluate whether circulating tumor DNA (ctDNA) clearance (alone or with radiographic response) is a reliable early surrogate end point for progression-free survival (PFS) and OS in 1L NSCLC studies. Methods Data from a phase III trial (IMpower150) of patients with untreated metastatic nonsquamous NSCLC were used. ctDNA clearance, unconfirmed radiographic response, and confirmed radiographic response at approximately 6 months after treatment were evaluated as surrogates for PFS and OS. The meta-analytic approaches were used. The data were randomly split into four cohorts, treating each as an independent "trial" to simulate a multi-trial meta-analysis. This process was repeated 50 times for statistical robustness. Results ctDNA decline correlated with longer PFS and OS, with greater declines associated with better outcomes. Patients achieving ctDNA clearance (CL) by week 21 had the most significant PFS and OS improvements ( p < 0.001). Radiographic response (Resp) within 6 months posttreatment was also associated with improved survival. Combined ctDNA_CL and Resp further enhanced prognostic discrimination. When comparing atezolizumab plus bevacizumab plus carboplatin plus paclitaxel versus bevacizumab plus carboplatin plus paclitaxel (BCP), the individual-level association between 6-month ctDNA CL plus Resp and survival yielded a global odds ratio of 2.06 (95% CI: 2.02–2.11) for PFS and 6.08 (95% CI: 5.92–6.23) for OS. Similar global odds ratios were observed for atezolizumab plus carboplatin plus paclitaxel versus BCP. For the atezolizumab plus bevacizumab plus carboplatin plus paclitaxel versus BCP comparison, cohort-level associations between 6-month ctDNA CL plus Resp and PFS were R 2 WLS = 0.41 (95% CI: 0.33–0.50) and R 2 copula = 0.38 (95% CI: 0.30–0.45), whereas associations with OS were R 2 WLS = 0.48 (95% CI: 0.40–0.56) and R 2 copula = 0.51 (95% CI: 0.43–0.60) at 6 months. Slightly lower cohort-level associations were observed for atezolizumab plus carboplatin plus paclitaxel versus BCP (R 2 = 0.34–0.41). Conclusion ctDNA decline is associated with improved survival outcomes in a clear dose–response manner. ctDNA clearance and radiologic response provide complementary prognostic information, and combining these end points (ctDNA_CL + Resp) can further improve prediction of PFS and OS. This combination also enhanced the correlation between treatment effects on the early end points and PFS/OS at the trial level. However, their overall ability to predict treatment effects at the trial level remained modest for both PFS and OS, likely due to the limited data from a single study. Further validation in larger, multi-trial data sets is needed to establish ctDNA clearance as a reliable early surrogate end point in first-line NSCLC.