Circulating tumor DNA (ctDNA) in early-stage NSCLC: a new frontier in precision oncology

Analytical platforms for ctDNA detection

Detection of ctDNA relies on a range of molecular technologies that differ in analytical sensitivity, genomic breadth, and scalability. These approaches can be broadly grouped into polymerase chain reaction (PCR)-based and next-generation sequencing (NGS)-based platforms.

PCR-based methods, such as digital PCR (dPCR) and its main variants droplet digital PCR (ddPCR), and BEAMing (Beads, Emulsion, Amplification, and Magnetics), offer high analytical sensitivity and relatively fast turnaround times^[21-24]. These techniques are particularly suited for detecting known mutations with high precision, often reaching limits of detection below 0.01% MAF^[25-27]. However, their major limitation lies in their restricted multiplexing capacity, which makes them less suitable for MRD applications where multiple independent variants must be tracked simultaneously to minimize false negatives.

NGS technologies revolutionized ctDNA detection, with the development of several innovative approaches. Amplicon-based strategies amplify selected genomic regions to enhance sensitivity; examples include Tagged-amplicon deep sequencing (TAm-Seq), which uses a two-step amplification process to accurately detect low-frequency variants^[28], and Safe-SeqS, which employs unique molecular identifiers (UMIs) to correct PCR errors, thereby lowering background noise^[29]. Hybrid capture-based approaches, exemplified by CAncer Personalized Profiling by deep sequencing (CAPP-seq), provide broader genomic coverage while retaining fragment length information^[30]. More recently, phased variant enrichment and detection sequencing (PhasED-seq) has been introduced as a capture-based assay that leverages clusters of mutations on the same DNA fragment to reduce both technical and biological errors, achieving detection limits in the parts-per-million^[31].

Although whole-exome sequencing (WES) and whole-genome sequencing (WGS) theoretically enable broad detection of tumor genomic alterations, these untargeted approaches have been mainly explored in tissue-based analyses and remain difficult to apply in liquid biopsy due to high costs, limited input material, and complex bioinformatic requirements. Feasibility is particularly challenging in early-stage disease, where ctDNA levels are extremely low. Emerging data indicate that WGS and WES may expand ctDNA detection capabilities, although their use remains largely experimental^[32,33].

Assay design: tumor-informed versus tumor-agnostic approaches

Beyond the analytical platform, ctDNA assays also differ in their design strategy, which can be either tumor-informed or tumor-agnostic^[12] [Figure 1]. Tumor-informed assays involve initial sequencing of a patient’s tumor tissue to create a customized panel of somatic mutations for plasma tracking^[34-38]. This strategy enhances analytical accuracy, achieving very low detection limits while reducing interference from clonal hematopoiesis of indeterminate potential (CHIP). Nonetheless, its implementation in routine clinical practice is constrained by practical issues: adequate tumor tissue must be available for sequencing, which is not always the case in early-stage disease or when only small biopsies are obtained before neoadjuvant therapy. The design of a patient-specific panel typically requires about 3-4 weeks, potentially delaying treatment decisions, and patients achieving a pathological complete response (pCR) may lack residual tumor tissue for assay generation^[35]. Moreover, the number of variants incorporated in the panel directly affects assay sensitivity and the limit of detection^[11,34,39]. Personalized multiplex PCR assays, such as Signatera (Natera) and RaDaR (Inivata), use tumor sequencing to identify patient-specific mutations that are subsequently tracked in plasma, reaching very high sensitivity^[40,41].

Figure 1. Tumor-informed vs. tumor-agnostic ctDNA assays: key features and limitations. CHIP: Clonal hematopoiesis of indeterminate potential; ctDNA: circulating tumor DNA; MRD: minimal residual disease. Created with BioRender.com (Created in BioRender. Cammarota, A. (2026) https://BioRender.com/txdc4kj).

Tumor-agnostic (tumor-naïve) assays, by contrast, rely on predefined genomic panels that can be applied directly to plasma without prior tumor profiling^[42,43]. Their main advantages are shorter turnaround time and independence from tissue availability. However, this approach is generally less sensitive and more susceptible to background noise from CHIP-related variants. From a biological standpoint, analyzing circulating DNA derived from the entire tumor burden may also enable tumor-agnostic approaches to capture both spatial and temporal genomic heterogeneity, providing a broader view of tumor evolution and disease complexity.

Study characteristics and ctDNA detectability across cohorts

Over the past decade, multiple retrospective and prospective studies, with cohorts ranging from 33 to over 330 patients, have evaluated perioperative ctDNA using various platforms, most commonly NGS-based assays^[11,51-68] [Table 1]. These studies generally included unselected stage I-III NSCLC patients, with the exception of the study by Jung et al., which focused on EGFR-mutant cases^[61]. The study by Chaudhuri et al. also enrolled a subset of small cell lung cancer patients, and some cohorts included a minority of stage IV cases^[51,53,66]. Adjuvant therapy was administered according to standard practice, while neoadjuvant chemotherapy was limited to a minority of cohorts^[61,62,66]. In selected cases, definitive chemoradiotherapy or radiotherapy alone was delivered as curative-intent treatment^[51,58,67].

Across studies, presurgical ctDNA was detectable in approximately 20%-60% of patients, with higher rates in stage II-III compared with stage I disease, and consistently lower sensitivity in adenocarcinoma compared with squamous histology. In TRACERx, detection reached 82% (78/85) in non-adenocarcinoma vs. 42% (39/93) in adenocarcinoma and was also associated with a heavier smoking history and a higher risk of occult nodal disease^[11].

Within the TRACERx study, Abbosh et al. conducted a phylogenetic ctDNA analysis in 96 patients with resected early-stage NSCLC, using patient-specific multiplex PCR panels informed by multiregion exome sequencing of the primary tumor^[34]. Preoperative ctDNA was detected in 48% of patients and was independently associated with non-adenocarcinoma histology, lymphovascular invasion, and high Ki-67 expression. In ctDNA-positive cases, clonal variants were consistently identified at high frequency, while subclonal variants were detected in about two-thirds of patients, reflecting the heterogeneity of tumor evolution. ctDNA detection preceded radiographic relapse by a median of 70 days, underscoring its ability to anticipate recurrence and capture clonal dynamics beyond imaging.

Updated TRACERx results, with a median follow-up of 4.6 years in event-free patients, further refined the prognostic role of ctDNA^[11]. Among 187 evaluable patients, preoperative ctDNA detection was higher in non-adenocarcinoma (92%) than in adenocarcinoma (42%) and, in the latter, was associated with occult mediastinal nodal involvement (P = 0.001). Postoperative ctDNA was detected in 84% of patients with recurrence or incomplete resection, while false negatives were mainly observed in those lacking sampling near relapse or with isolated brain metastases, where ctDNA shedding is limited. At a postoperative landmark analysis (within 120 days after surgery), ctDNA was detectable in 25% of patients and was strongly predictive of recurrence (positive predictive value 93%, negative predictive value 68%, sensitivity 49%), with positivity increasing by stage (12% in stage I, 23% in stage II, 44% in stage III; P = 0.008). Landmark ctDNA positivity was associated with a higher risk of death (hazard ratio [HR] 5.3, 95% confidence interval [CI]: 2.9-9.7; P = 1 × 10^-9) and recurrence (HR 6.8, 95%CI: 3.7-12.3; P = 6 × 10^-13). During serial surveillance, 20% of landmark-negative patients later became ctDNA-positive before or at relapse, with a median lead time of approximately one year and detectable ctDNA levels as low as 0.02%.

Prognostic role of ctDNA and implications for postoperative risk stratification

The presence of detectable ctDNA before treatment consistently demonstrated prognostic value across studies, showing a significant association with poorer outcomes. Pre-treatment ctDNA positivity was associated with inferior recurrence-free survival (RFS) (HRs 2.2-4.2) and overall survival (OS) (HRs 2.0-5.5). In studies incorporating multivariable analyses, baseline ctDNA remained an independent prognostic factor for RFS and OS in two cohorts^[52,54], while in another the association did not reach statistical significance^[67], underscoring the heterogeneity of results across studies.

Post-treatment ctDNA assessment has been performed using either landmark analyses, which assess prognostic value at predefined postoperative timepoints to minimize immortal-time bias^[69], or serial sampling enabling longitudinal monitoring. At postoperative landmark analyses, ctDNA positivity was reported in approximately 5%-25% of patients across studies. Detection at this prespecified timepoint was consistently associated with poor outcomes, with HRs for RFS ranging from 4.0 to 53.0, and for OS from 3.0 to 27.0.

Longitudinal monitoring identified ctDNA positivity in up to ~11%-56% of patients during follow-up. Across studies, longitudinal ctDNA positivity was associated with a significantly increased risk of recurrence, with HRs for RFS between 2.7 and 16.1, and for OS ranging from 2.9 to 11.4. In studies incorporating multivariable analyses, postoperative ctDNA positivity consistently emerged as an independent prognostic factor for survival outcomes^{[54,55,59-61,64,67,68]}.

Across these studies, ctDNA was investigated as an early biomarker of recurrence risk. Reported median lead times between ctDNA detection and radiological or clinical relapse ranged from less than 1 month to about 12 months, reinforcing its value as a tool for the earlier identification of molecular relapse compared with conventional imaging.

Of note, a limited number of studies observed that some patients who were ctDNA-negative after surgery subsequently relapsed with brain-only disease. This observation may reflect the limited negative predictive value of ctDNA and suggests that intracranial recurrence may be less reliably captured by plasma ctDNA assays^{[11,57,64,70]}.

Several studies have explored the prognostic implications of ctDNA status in the context of adjuvant chemotherapy, examining both dynamic changes before and after treatment as well as its potential to refine risk stratification beyond conventional clinicopathological factors. Although based on limited cohorts, findings have been consistent in showing that adjuvant chemotherapy was associated with longer RFS in ctDNA-positive patients (HRs 0.2-0.34), whereas no clear benefit was observed in ctDNA-negative patients, with some analyses even suggesting worse outcomes in this subgroup^{[54-56,60,68,70]}.

Several caveats should be considered when interpreting these findings. First, most of the studies were conducted in relatively small cohorts, which limits the statistical power and generalizability of the results. Moreover, the enrolled populations were heterogeneous, with differences in stage distribution and histologic subtype, both of which are known to impact ctDNA shedding and detectability. Additional variability derived from the definition of landmark timepoints and the frequency of blood sampling, as well as from the use of diverse assay platforms with different sensitivities and analytical thresholds. Moreover, indirect comparisons are further limited by heterogeneous preanalytical conditions, including sample handling, processing intervals, and cfDNA extraction, all of which can markedly affect cfDNA yield and ctDNA detection^[71]. Taken together, these factors preclude direct comparison across studies, and the reported ranges of detection rates and hazard ratios should therefore be interpreted with caution.

Role of ctDNA in neoadjuvant chemoimmunotherapy trials

The phase III CheckMate 816 trial established neoadjuvant chemoimmunotherapy as a new standard of care in resectable NSCLC, demonstrating significant improvements in pCR and event-free survival (EFS) with the addition of nivolumab to platinum-based chemotherapy^[1]. An exploratory analysis within this study investigated the clinical relevance of ctDNA as a prognostic biomarker, using a tumor-informed panel for WES (ArcherDX Personalized Cancer Monitoring)^[35]. Among the 358 patients enrolled, plasma samples suitable for ctDNA analysis were available for 43 (24.0%) patients in each treatment arm.

In the nivolumab plus chemotherapy arm, ctDNA clearance after neoadjuvant treatment occurred in 56% (24/43) of patients with baseline ctDNA positivity, compared to 35% (14/43) in the chemotherapy group. A higher proportion of patients achieving ctDNA clearance also had a pCR in the nivolumab plus chemotherapy group (46%) compared with the chemotherapy arm (13%). Furthermore, ctDNA dynamics provided prognostic information beyond pathologic assessment. Patients in the chemoimmunotherapy arm who cleared ctDNA experienced markedly improved long-term outcomes, with a 5-year OS rate of 75%, compared to 52.6% in those with residual ctDNA (HR 0.38, 95%CI: 0.15-1.00). Similarly, the 5-year EFS rate was higher in patients with ctDNA clearance (51.6% vs. 40.1%; HR 0.45, 95%CI: 0.18-1.11), reinforcing the prognostic value of ctDNA kinetics in the neoadjuvant setting.

These results indicate that ctDNA clearance after neoadjuvant therapy may provide additional prognostic information in resectable NSCLC, complementing established markers such as pathological assessment.

Role of ctDNA in perioperative chemoimmunotherapy trials

The NADIM trial (ClinicalTrials.gov: NCT03081689) was a phase II, single-arm study that established perioperative chemoimmunotherapy as a highly promising approach in resectable stage IIIA NSCLC^[74]. Patients (N = 46) received three cycles of nivolumab plus chemotherapy followed by surgery and one year of adjuvant nivolumab. At 5-year follow-up, the trial continued to report remarkably favorable outcomes, with long-term survival significantly outperforming historical controls^[42]. A prespecified exploratory objective was to assess ctDNA as a potential biomarker of treatment efficacy and prognosis. At baseline, ctDNA was detectable in 30 of 43 (69.8%) patients with available samples, and levels were significantly correlated with tumor size (P = 0.021). Quantitative assessment of ctDNA burden—measured as the summatory mutant allele fraction (sumMAF)—demonstrated strong prognostic value. Patients with low baseline ctDNA levels (< 1% sumMAF) experienced superior 5-year progression-free survival (PFS) and OS, at 83.8% (95%CI: 65.3-92.9) and 86.2% (95%CI: 67.1-94.6), respectively, compared to 48.6% (95%CI: 19.2-73.0) and 56.3% (95%CI: 24.4-79.1) in those with sumMAF ≥ 1% (HR 4.26; 95%CI: 1.02-17.75, P = 0.0089 and HR 4.38; 95%CI: 0.91-21.10, P = 0.016 for PFS and OS, respectively). Of note, four of the five patients with disease progression despite low baseline ctDNA burden harbored unfavorable genomic alterations, including one EGFR and three kelch-like ECH-associated protein 1 (KEAP1) mutations, highlighting the potential additive prognostic relevance of mutational profile^[75]. Clearance of ctDNA following neoadjuvant therapy—defined as undetectable or < 0.1% sumMAF—was also significantly associated with improved long-term outcomes. Among the 40 patients with evaluable post-treatment samples, 27 (67.5%) achieved ctDNA negativity prior to surgery. Patients who cleared ctDNA showed better 5-year survival: PFS was 85.2% and OS reached 92.3%, vs. 60.6% and 59.2%, respectively, in patients with residual detectable ctDNA (HR 3.06, 95%CI: 0.73-12.90, P = 0.079 and HR 6.88, 95%CI: 1.29-36.78, P = 0.0070 for PFS and OS, respectively). Interestingly, ctDNA clearance did not correlate with pCR, nor did it retain prognostic value in patients who had achieved pCR. However, in patients with residual viable tumor, ctDNA clearance maintained a strong prognostic association, suggesting that molecular response could complement—but not replace—histopathological assessment. Additionally, post-hoc analyses revealed that the emergence of new mutations during treatment, particularly in Kirsten rat sarcoma virus (KRAS) and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), but not their presence at baseline, was associated with inferior PFS and OS. This finding supports the use of longitudinal ctDNA monitoring to capture clonal evolution and resistance dynamics during therapy. Notably, ctDNA dynamics demonstrated a stronger association with long-term outcomes compared to conventional response assessments. While the majority of patients achieved a partial radiological response (71.7%) and 63.4% achieved a pCR, neither was significantly correlated with progression-free or overall survival. By contrast, ctDNA clearance after neoadjuvant therapy was independently associated with improved survival outcomes and demonstrated a higher prognostic discriminative ability, as reflected by a Harrell’s concordance index (C-index) of 0.82 for OS, compared to 0.72 for Response Evaluation Criteria in Solid Tumors (RECIST)-based radiologic response. Furthermore, the prognostic relevance of ctDNA clearance persisted even in patients without a pCR, suggesting that molecular response assessment may provide complementary information to pathological evaluation and potentially identify MRD not captured by histology alone. The addition of ctDNA dynamics to radiologic response significantly enhanced the prediction of OS (P = 0.003), indicating complementary prognostic value^[42]. Neither tumor mutational burden (TMB) nor PD-L1 expression levels were significantly associated with PFS or OS in this study. ctDNA showed superior performance compared to these established biomarkers^[42]. Among patients achieving pCR, 92% (24/26) did not have increased ctDNA after treatment. In contrast, 33% of patients with incomplete pathological response showed an increase in ctDNA levels after neoadjuvant therapy^[42]. Altogether, the NADIM trial provides robust and mature evidence that both baseline ctDNA levels and dynamic changes during neoadjuvant treatment are independent predictors of long-term survival. At baseline, tumor protein 53 (TP53) was the most commonly mutated gene detected in ctDNA, whereas its prognostic or predictive role was not reported. Notably, APC mutations were associated with significantly worse PFS and OS, highlighting the potential of ctDNA to provide prognostic insights beyond clearance dynamics^[42].

Additional evidence comes from the NADIM II trial, which enrolled patients with resectable stage IIIA and IIIB NSCLC, thereby extending the evaluation of perioperative chemoimmunotherapy to a more advanced disease stage^[43]. In this randomized phase II study, patients received perioperative nivolumab plus platinum-based chemotherapy or chemotherapy alone. Baseline ctDNA was detectable in 57 of 86 patients and correlated with tumor burden. Patients with low baseline ctDNA levels (MAF < 5%) experienced significantly better outcomes than those with high ctDNA levels (HR 0.38, 95%CI: 0.15-0.95 for PFS; HR 0.22, 95%CI: 0.07-0.64 for OS). Post-treatment plasma samples collected before surgery were available in a subset of patients. Following neoadjuvant therapy, ctDNA was undetectable in 67% (18/27) of patients in the chemoimmunotherapy arm, compared to 44% (4/9) in the chemotherapy arm.

Further insights on the prognostic utility of ctDNA in the perioperative setting comes from the phase III AEGEAN trial, which evaluated neoadjuvant platinum-based chemotherapy with or without durvalumab followed by surgery and one year of adjuvant durvalumab or placebo in patients with resectable stage II-IIIB NSCLC^[3]. A comprehensive exploratory analysis focused on ctDNA dynamics at multiple timepoints: prior to each neoadjuvant therapy cycle, at surgery, and during adjuvant therapy (cycles 1, 3/4, and 10/11). A MRD landmark assessment was conducted at adjuvant cycle 1 day 1, corresponding to a median of 6.9 weeks after resection. WES of pretreatment tumor biopsies was used to identify somatic variants for the development of patient-specific, tumor-informed assays applied to plasma samples. At baseline, ctDNA was detectable in the majority of patients (89.6%, N = 283, 142 patients in the durvalumab arm and 141 in placebo arm), and its clearance during neoadjuvant treatment was observed more frequently in patients achieving a pathological response^[76]. Among patients who were ctDNA-positive at baseline, all those who achieved pCR and over 93% of those with major pathologic response (MPR) had achieved ctDNA clearance by pre-surgical assessment. Conversely, lack of early clearance—particularly by cycle 2 day 1—had a high negative predictive value (≥ 89%) for pCR, suggesting a potential role for ctDNA in identifying patients unlikely to achieve pathological responses. Moreover, pre-surgical ctDNA clearance was associated with improved EFS regardless of pCR status^[36]. At the post-operative landmark timepoint, MRD was observed in approximately 10% of patients (17/168, 10 in durvalumab arm and 7 in placebo arm) and was associated with a markedly inferior 12-months disease-free survival (DFS) (14.3%, 95%CI: 2.4-36.3) compared with MRD-negative patients (89.3%, 95%CI: 82.6-93.5)^[36]. MRD positivity at the post-surgical landmark was more frequently observed among patients with stage III disease or N2 nodal involvement. In the durvalumab arm, none of MRD-positive patients achieved MPR or pCR. Exploratory genomic analyses suggested that postoperative MRD positivity in the durvalumab arm was associated with mutations in KEAP1 (30.0% vs. 6.4%) and lysine methyltransferase 2C (KMT2C) (30.0% vs. 3.8%) (P < 0.05). Although based on small numbers, these findings highlight the added biological insight that ctDNA profiling can provide beyond its role as a dynamic quantitative biomarker.

In the phase III CheckMate 77T trial, patients with resectable stage IIA-IIIB (N2) NSCLC were randomized to receive neoadjuvant nivolumab plus platinum-based chemotherapy followed by surgery and adjuvant nivolumab or placebo^[4]. Among biomarker-evaluable patients (N = 190 across both arms), baseline ctDNA was detectable in 76 (33%) patients in the nivolumab arm and 64 (28%) patients in the placebo arm^[72]. Pre-surgical ctDNA clearance occurred more frequently with nivolumab (66% vs. 38%) and was associated with longer EFS across treatment groups, with the greatest benefit observed in the nivolumab arm (HR 0.41, 95%CI: 0.20-0.86). Importantly, the association between ctDNA clearance and improved EFS was evident regardless of pCR status. The EFS benefit with nivolumab was consistent across key genomic subgroups, including patients harboring KRAS, serine/threonine kinase 11 (STK11), or KEAP1 mutations^[72].

Role of ctDNA in adjuvant immunotherapy trials

In the phase III IMpower010 trial, patients with completely resected stage IB-IIIA NSCLC received 1-4 cycles of adjuvant platinum-based chemotherapy before randomization to atezolizumab or best supportive care^[5]. A prespecified exploratory analysis evaluated ctDNA at two key timepoints: post-surgery (before adjuvant chemotherapy) and post-chemotherapy (cycle 1 of adjuvant treatment in post-surgery ctDNA-positive patients), using a tumor-informed Signatera (Natera) assay^[38]. Among 600 ctDNA-evaluable patients with stage II-IIIA disease, 112 of 534 (21%) were ctDNA-positive after surgery. Detection of ctDNA after surgery or completion of adjuvant chemotherapy was associated with inferior DFS. The addition of adjuvant atezolizumab after chemotherapy improved DFS and OS in PD-L1-positive population, irrespective of ctDNA status, and also showed benefit in longitudinal ctDNA monitoring. Among patients who were ctDNA-negative after adjuvant chemotherapy, the median time to conversion to ctDNA positivity was longer with atezolizumab (HR 0.60, 95%CI: 0.31-1.17), and DFS was also prolonged in this subgroup (31.3 vs. 13.3 months; HR 0.70, 95%CI: 0.37-1.34)^[73].

In the phase III BR.31 trial, adjuvant durvalumab did not improve survival outcomes in resected EGFR/ALK-negative NSCLC^[77]. A retrospective tumor-informed MRD analysis, performed in 1,131 patients (80% of the trial population), showed that postoperative MRD positivity was identified in approximately 10% of cases (116/1131) and was strongly associated with inferior OS in the placebo arm. A survival benefit with durvalumab was observed in MRD-positive patients within both PD-L1 ≥ 25% (P = 0.011) and ≥ 1% (P = 0.024) subgroups, suggesting a potential predictive role of MRD status.

Role of ctDNA in targeted therapy approaches for early-stage NSCLC

Based on the results of the phase III ADAURA trial, three years of adjuvant osimertinib are now recommended for patients with completely resected stage IB-IIIA EGFR-mutant NSCLC^[6]. A retrospective, post hoc, tumor-informed MRD analysis was performed in a subset of 220 patients^[37]. At randomization, MRD was detectable in 8% of patients (5/112, 4% with osimertinib; 13/108, 12% with placebo), all with stage II-IIIA disease. Baseline MRD positivity appeared to be associated with poor prognosis, with 17 of 18 patients experiencing a DFS event, although the subgroup size was small. Four of five MRD-positive patients receiving osimertinib achieved ctDNA clearance, compared with none in the placebo arm. During follow-up, MRD events occurred in 31% (68/220) of patients. Among those MRD-negative at randomization, MRD-free status was maintained in 91% of patients in the osimertinib arm versus 58% in the placebo arm. In the osimertinib group, most MRD or DFS events (68%) occurred after treatment discontinuation, often within 12 months. Across both arms, MRD positivity preceded clinical recurrence by a median of 4.7 months. At 36 months, the DFS and MRD event-free rate was 86% vs. 36% for osimertinib and placebo, respectively (HR 0.23, 95%CI: 0.15-0.36).

Recently, preliminary findings from the ADMIN study evaluating ctDNA dynamics in patients with resected NSCLC harboring uncommon EGFR mutations treated with adjuvant osimertinib were presented^[78]. Treatment was given with or without prior adjuvant chemotherapy. ctDNA dynamics were assessed using a tumor-informed assay. Among 51 enrolled patients, 43 were evaluable for MRD analysis. ctDNA was detected in 23.3% (10/43), and half of these patients (5/10) achieved clearance after 12 weeks of osimertinib. At baseline, 18.6% (8/43) were plasma EGFR mutation-positive, with clearance observed in 62.5% (5/8).

In the phase III NeoADAURA trial, neoadjuvant osimertinib-containing regimens significantly improved MPR over chemotherapy alone in resectable, EGFR-mutated stage II-IIIB NSCLC^[79]. An exploratory analysis from NeoADAURA evaluated the associations between baseline and pre-surgical MRD status with EFS and MPR. MRD testing was performed using the Personalis NeXT Personal® assay, an ultra-sensitive tumor-informed approach, in 189 patients. Compared with single EGFR mutation testing by Cobas® (30% detection rate), tumor-informed MRD analysis improved baseline ctDNA detection (71%). Patients without detectable MRD generally had less extensive disease at baseline, in terms of stage (II vs. III), tumor size, and nodal involvement. Notably, MRD negativity was associated with longer EFS (HR 0.24, 95%CI: 0.07-0.80). Treatment with osimertinib-containing regimens resulted in higher rates of MRD clearance compared with the control arm. Moreover, both pre-surgical MRD clearance and undetectable MRD status were associated with MPR (P = 0.0378 and P = 0.0546, respectively)^[80]. The prognostic relevance of ctDNA was further supported by a phase II multi-institutional trial (NCT03433469) investigating neoadjuvant osimertinib in resectable stage I-IIIA EGFR-mutated NSCLC^[81]. Among 23 evaluable patients, ctDNA testing was performed with the Guardant Reveal assay. Baseline ctDNA detection prior to treatment (C1D1) was independently associated with inferior EFS, with a median of 8.5 months in ctDNA-positive patients compared with 45.6 months in ctDNA-negative patients (HR 3.78, 95%CI: 1.09-13.13; P = 0.036). Similarly, ctDNA detection after one cycle of osimertinib (C1D28) was associated with shorter EFS in univariate analysis (18.7 vs. 45.6 months; HR 5.90, 95%CI: 1.09-32.01; P = 0.038), although this did not remain significant in multivariable analysis^[82].

In a single-center pilot study (NORA trial) of 25 patients with stage IA-IIIA EGFR-mutant NSCLC, longitudinal tumor-informed ctDNA monitoring was integrated into a treatment strategy of neoadjuvant osimertinib (two 28-day cycles), surgery, and three years of adjuvant osimertinib^[83]. Personalized ctDNA assays were developed for 20 patients, with baseline ctDNA positivity in six (30%). All achieved clearance after the first neoadjuvant cycle, irrespective of stage. Postoperatively, 95% (18/19) of recurrence-free patients remained ctDNA-negative, showing strong concordance between MRD status and disease control. One patient became ctDNA-positive while on adjuvant osimertinib without radiological recurrence, and another relapsed 2.6 months after treatment discontinuation despite ctDNA negativity. In terms of efficacy, neoadjuvant osimertinib achieved an objective response rate of 44% and a major pathological response rate of 24%, with no complete pathological responses reported.