Error-corrected plasma whole-genome sequencing for personalised and tumour-agnostic minimal residual disease detection in TRACERx

TRACERx

Patient / Population Intervention / Exposure Comparison Outcome

In patients with early-stage non-small cell lung cancer, landmark ctDNA detection by error-corrected plasma whole-genome sequencing was associated with shorter recurrence-free and overall survival.

N

73patients

Design

Translational analysis applying ppmSeq error-corrected plasma whole-genome sequencing to longitudinal plasma samples from 73 individuals with early-stage NSCLC in the TRACERx prospective cohort study (NCT01888601) for tumour-informed and tumour-agnostic MRD detection, validated in 50 individuals with stage II-III colorectal cancer, 25 with recurrence, from the prospective Danish IMPROVE cohort study (NCT03637686).

Endpoint

No formal primary endpoint is stated. The main measure was the association of landmark ctDNA detection (MRD) with recurrence-free survival and overall survival.

ResultIn the NSCLC landmark analysis (n=73), MRD detection was associated with shorter RFS (HR 9.81, P = 3.2x10-9) and OS (HR 5.04, P = 9.6x10-6). The association persisted after adjustment for pathological stage, histology and adjuvant chemotherapy (HR 8.53, P = 1.5x10-7). ctDNA was detected at fractions as low as 2.1 ppm, with a median fraction of 667 ppm among detected samples (IQR 13-5,781). 16% of detections (11/69) occurred below 10 ppm. In the CRC cohort (n=50), landmark ctDNA by ppmSeq was associated with shorter RFS (HR 5.45, P = 5.29×10-5), and ctDNA was detected in 72% (18/25) of individuals with recurrence. In patients who did not receive adjuvant chemotherapy, discrimination of recurrence was AUC 0.93 with ppmSeq vs 0.70 with ddPCR (P = 0.03).

Wan JC, Pellan-cheng A, Ward S, et al. Presented at ESMO Congress 2026, abstract LBA94, 25 October. Not yet published.

Discussion & critique

This translational study shows that error-corrected plasma whole-genome sequencing can detect ctDNA at very low fractions, and that landmark detection strongly predicts recurrence in two modest prospective cohorts. The abstract reports no confidence intervals, no follow-up duration, no lead time over clinical recurrence, and no sensitivity or specificity for NSCLC. The ddPCR comparison is limited to a subgroup of unstated size, and the tumour-agnostic signatures are described as quantified without any performance data. This is prognostic evidence only: it does not show that acting on MRD status improves outcomes, and the assay comes from Ultima Genomics, with which several authors disclose financial ties.