Recurrence prediction using circulating tumor DNA in patients with early-stage non-small cel lung cancer after treatment with curative inten A retrospective validation study

Background Despite treatment with curative intent, many patients with localized non-small cell lung cancer (NSCLC) develop recurrence. The current challenge is to identify high-risk patients to guide adjuvant treatment. Identification of residual disease by detection of circulating tumor DNA (ctDNA) may allow more accurate clinical decision-making, but its reliability in NSCLC is not established. We aimed to build on previous data to validate a tissue-informed personalized ctDNA assay, to predict recurrence in patients with early-stage disease. Methods and findings Tumor tissue and plasma was collected from patients with stage 0–III NSCLC enrolled to LEMA (Lung cancer Early Molecular Assessment trial, NCT02894853). Serial plasma was collected before and after definitive treatment, with the latter including key timeframes of interest (1–3 days post-treatment, between 14 and 122 days after treatment end, and≥14 days after treatment end). Somatic mutations identified by tumor exome sequencing were used to design patient-specific assays, to analyze ctDNA. Results were compared and combined with an independent dataset (LUCID; LUng Cancer CIrculating Tumour Dna study, NCT04153526). In LEMA, 130 patients (57% male; median age 66 years (range 44–82); 69% adenocarcinoma, 22% squamous cell carcinoma (SCC); 3%/49%/19%/29% with stage 0/I/II/ III) were treated with curative intent. Tumor tissue originated from surgical resection or diagnostic biopsy in 118 and 12 patients respectively. LUCID included 88 patients (51% male; median age 72 years (range 44–88); 63% adenocarcinoma, 31% SCC; 49%/28%/23% with stage I/II/III). Before treatment, ctDNA was detected in 48% LEMA and 51% LUCID patients. Sensitivity, specificity, positive and negative predictive value of ctDNA detection post-treatment (≥1 positive sample ≥14 days after treatment end) to predict recurrence were 61%, 97%, 92% and 84% for LEMA and 64%, 96%, 90% and 83% for LUCID. In the combined cohort, ctDNA detection after treatment was associated with shorter recurrence-free survival (hazard ratio (HR) 11.4 (95% confidence interval (CI) [7.0,18.7]; p<0.001)) and overall survival (HR 8.1 (95% CI [4.6,14.2]; p<0.001)), accounting for guarantee-time bias. Of note, a key limitation of this work was the irregular sample collection schedules, during routine follow-up visits, of both studies. Conclusions ctDNA detection predicted recurrence in independent retrospective cohorts with notable reproducibility, including near-identical detection rates and predictive values, confirming its ability to differentiate patients at high- versus low risk of recurrence. Our results support the potential of tissue-informed ctDNA analysis as a decision-support tool for adjuvant therapy in NSCLC.