ctDNA clearance adds a new layer of risk signal in resectable NSCLC
Resectable non-small cell lung cancer still carries a substantial relapse risk even after perioperative treatment, and NeoCOAST-2 was designed to test whether newer combinations could improve early tumor control before surgery. In this phase 2 randomized platform trial, patients with stage IIA–IIIB disease received standard chemotherapy plus durvalumab alongside one of several investigational partners. The first readout already showed encouraging pathological activity, with pCR rates of 20.3% for the oleclumab arm, 25.7% for the monalizumab arm and 35.2% for the datopotamab deruxtecan, or Dato-DXd, arm.
The new analysis shifts the focus from resection specimens alone to what is circulating in blood before surgery. Using a tumour-agnostic assay, investigators tracked circulating tumour DNA clearance at neoadjuvant Cycle 3 Day 1 and paired those data with event-free survival. The result is a sharper clinical picture: molecular response in blood lined up with better outcomes, and the association was strong enough to suggest a practical role for ctDNA in perioperative decision-making.
A blood test outperformed the eye test of imaging alone in this analysis.
Molecular response after neoadjuvant therapy separated prognostic groups
ctDNA clearance rates at neoadjuvant Cycle 3 Day 1 differed across the three arms, reaching 50.9% in the oleclumab arm, 54.5% in the monalizumab arm and 72.5% in the Dato-DXd arm. The numerical spread matters because it suggests the antibody-drug conjugate combination may be producing deeper early tumor kill than the immunotherapy-based alternatives in this specific setting. The trial does not prove superiority across all endpoints, but the directional pattern is difficult to ignore.
The clearest signal came from the relationship between clearance and survival. Patients who cleared ctDNA had significantly longer event-free survival than those who did not, with a pooled hazard ratio of 0.22 and a 95% confidence interval of 0.11–0.45. The reported P value of 0.0000443 makes this association hard to dismiss as noise. In practical terms, the blood-based readout behaved like an early surrogate of treatment sensitivity, capturing biologic response before long follow-up accrued.
That pattern also fits the biology of minimal residual disease. Surgery may remove all visible disease, yet microscopic tumor deposits can persist and later drive relapse. When ctDNA disappears from plasma during neoadjuvant therapy, it likely reflects a major reduction in residual tumor burden. When it persists, the tumor may still be biologically active despite a favorable scan or even a partial pathologic response.
pCR and ctDNA together refined the prognostic picture
The most informative result came when ctDNA dynamics were combined with pathology. Patients who achieved both ctDNA clearance and pathological complete response had substantially better event-free survival than those who achieved neither, with an EFS hazard ratio of 0.10 and a 95% confidence interval of 0.02–0.42. In other words, the best clinical outcomes clustered in patients whose disease was invisible both in the resected specimen and in plasma.
This combined endpoint has obvious appeal for thoracic oncology. pCR is a meaningful tissue-based endpoint, but it only becomes available after surgery and can miss early biologic resistance that blood testing may detect. ctDNA, by contrast, offers a dynamic measure during treatment. When both markers agree, confidence in the depth of response rises. When they diverge, clinicians may be looking at a patient whose risk remains higher than pathology alone would suggest.
The findings also help explain why the Dato-DXd arm drew attention. It produced the highest pCR rate in the earlier report and the highest ctDNA clearance rate in this analysis, alongside the numerically highest 24-month EFS rate at 77.9%. The other arms posted 24-month EFS rates of 70.5% and 75.7%. These are not definitive comparative claims, but they place the Dato-DXd combination in the most favorable position across the measured endpoints.
Why this matters for perioperative lung cancer care
Perioperative NSCLC trials have long struggled with how best to connect early response markers to long-term outcomes. Pathology remains essential, but it is a retrospective measure of what happened in the tumor bed. ctDNA offers a moving target, one that can be measured while treatment is still underway. That matters for a disease in which relapse often reflects occult residual disease rather than the primary mass that was removed.
The current data support a model in which blood-based minimal residual disease assessment could complement surgical pathology and help stratify postoperative risk. A patient with ctDNA clearance and pCR looks biologically different from a patient with residual ctDNA after induction therapy, even if both undergo resection. The first group may need less escalation; the second may be the group in which intensified postoperative strategies deserve the closest attention.
The study also strengthens the rationale for continued development of datopotamab deruxtecan in resectable NSCLC. Its numerical advantage across pCR, ctDNA clearance and 24-month EFS does not establish practice change on its own, yet it gives the regimen a coherent efficacy story across tissue and liquid endpoints. In a field where many perioperative strategies improve one metric without moving another, concordance across multiple measures is notable.
What remains unresolved in NeoCOAST-2
Several questions stay open. The analysis is based on a phase 2 randomized platform design, so the sample sizes in each arm are modest and the trial was not built to settle comparative efficacy definitively. The ctDNA assay was tumour-agnostic, which is clinically attractive, but broader implementation will depend on assay performance in everyday practice and on how easily those results can be operationalized around surgery.
The timing of ctDNA assessment also deserves attention. Cycle 3 Day 1 is early enough to capture treatment effect before surgery, yet the optimal time point for risk stratification may differ across regimens and disease stages. The study shows that a single on-treatment blood draw can separate patients with very different outcomes, but it does not answer whether an earlier or later sample would perform better, or whether serial sampling would add further prognostic precision.
For now, the clearest takeaway is straightforward: in resectable NSCLC, ctDNA clearance tracked with better event-free survival, and the combination of ctDNA clearance plus pCR identified the patients with the most favorable outcomes. The Dato-DXd arm produced the strongest numerical signals across the reported endpoints, and that pattern now deserves validation in larger studies before it is used to guide treatment choices.