Ultrasensitive circulating tumor DNA (ctDNA) minimal residual disease (MRD) detection in early stage non-small cell lung cancer (NSCLC).

8078 Background: Translation of ctDNA MRD in NSCLC has been hampered by suboptimal sensitivity of 1st-generation assays. Here, we explore how improvements in analytical sensitivity can drive improved clinical sensitivity for MRD after surgery in NSCLC. Methods: To understand MRD kinetics, we analyzed longitudinal ctDNA in early stage NSCLC using data from the TRACERx study. Patient-specific mathematical models were generated to predict MRD levels at the postsurgical landmark and estimate the impact of an assay’s 95% limit of detection (LOD95) on clinical sensitivity. To test these predictions, tumor-informed ctDNA testing was performed using an SNV-based assay (CAPP-Seq) and a phased variant-based assay (PhasED-Seq, Foresight Diagnostics) on 269 samples from 46 patients. We also assessed MRD performance for predicting 1) patient outcomes at the landmark timepoint and 2) the effect of adjuvant treatment. Results: ctDNA MRD dynamics were assessed in 23 patients from TRACERx with ≥3 consecutive samples with detectable ctDNA without intervening therapy. MRD kinetics strongly correlated with exponential growth, with a median ctDNA doubling time of 51 days. Extrapolating MRD levels to the postsurgical landmark predicted that improving MRD assay LOD95 from 100 ppm (0.01%) to 1 ppm could increase clinical sensitivity by 2.1-fold. We then evaluated 46 NSCLC cases using two assays. The median LOD95 was 1 ppm for PhasED-Seq and 84 ppm for CAPP-Seq. Twelve cases were MRD+ by PhasED-Seq, all of whom recurred (100%), with MRD levels as low as 0.19 ppm. In contrast, 6 cases were MRD+ by the SNV-based approach, of whom 5 (83%) recurred. Accordingly, PhasED-Seq had a higher clinical sensitivity than the SNV-based method (12/18 67% vs. 5/18 28%, P = 0.022). Kaplan Meier analysis for freedom from recurrence revealed significantly worse outcomes for patients with detectable MRD regardless of the method used; however, outcomes were better stratified using PhasED-Seq (HR 3.1 vs. 11.4). Patients who were MRD- after surgery by both assays had similar outcomes regardless of adjuvant therapy (chemotherapy and/or radiotherapy). However, MRD+ patients by PhasED-Seq receiving adjuvant therapy had significantly better outcomes than those who did not (HR 8.2, P = 0.00035). A similar benefit for adjuvant therapy was not observed with the SNV-based assay. Using PhasED-Seq, 80% (4/5) of MRD+ patients receiving adjuvant therapy cleared their MRD, compared to 0% (0/3) without adjuvant treatment. Conclusions: Ultrasensitive ctDNA detection improved the clinical sensitivity of MRD at key landmarks in early stage NSCLC. PhasED-Seq detected MRD at levels below 1 ppm and was associated with significantly better outcomes, revealing potential benefits of adjuvant therapy in MRD+ patients. This suggests that ultrasensitive MRD detection is promising for use in risk-adapted trials in early stage NSCLC.