Cell-Free DNA Detects Adjuvant MRD Positivity and Disease Relapse in NSCLC

Circulating tumor DNA may be a biomarker for the detection of postsurgical minimal residual disease and for determining the clonality of relapsing disease.

Circulating tumor DNA (ctDNA) may be a biomarker for the detection of postsurgical minimal residual disease (MRD) and for determining the clonality of relapsing disease. Results from the Lung TRACERx study, in which ctDNA was detected in various histological subtypes, have the potential to reshape clinical trial designs for patients with non–small cell lung cancer (NSCLC) who become MRD-positive after surgery.

Findings from the TRACERx (NCT01888601) lung study were presented by Chris Abbosh, MD, of University College London in the United Kingdom, during the 2020 American Association for Cancer Research Virtual Annual Meeting I.1 The ongoing observational study, which was launched in 2014, is designed to determine the genetic evolution of NSCLCs from diagnosis to relapse and to analyze the association of these changes with therapeutic outcomes.

Investigating MRD is important for advancing treatment for patients with high-risk early-stage disease who have undergone potentially curative resection of their lung cancer, Abbosh said. Patients with low-burden, residual cancer after surgery have a greater potential for a survival benefit from adjuvant chemoradiation therapy than do those in the metastatic setting. “If we want to improve outcomes further in non–small cell lung cancer, we really need to focus on innovation in the early-stage space,” he said.

The use of MRD as a biomarker would enable clinical investigators to zero in on the population most likely to benefit from further therapy, Abbosh noted. “The promise of an MRD-driven adjuvant trial is that you can differentiate between these populations and you can conduct adjuvant trials in smaller, more relevant populations when you are only escalating treatment in patients who are destined to relapse from their disease,” he said.

To measure MRD, investigators developed a 50-variant assay based on anchored-multiplex polymerase chain reaction (AMP) enrichment panels generated from cell-free DNA in plasma samples from patients who had undergone surgery for stage I to III NSCLC. Abbosh said the algorithm detected variant fractions down to 0.003 with high DNA input and 0.008 with low DNA input, with a specificity of 99.8% in analytical data.

To confirm the sensitivity of the assay, investigators applied the test to a real-world postoperative population of participants in the TRACERx study who either did not suffer a clinical relapse or who developed a second primary cancer.

Overall, 26 patients with a median of 8 postoperative samples per patient (range, 3-13) were included in the nonrecurrence cohort and followed up for a median of 1173 days (3.2 years). The second primary cancer cohort consisted of 11 patients with a median of 7 postoperative samples per patient (range, 1-10). This cohort had a median follow-up of 840 days of disease-free survival (DFS). The second primaries identified were 5 second primary lung cancers and 6 non-lung cancers. Of the 271 postoperative timepoints analyzed in 37 patients, 269 were MRD-negative, leading to a specificity of 99.3% for the assay in this data cohort.

“[In] both of these categories we would not expect to see a postoperative MRD signal,” said Abbosh. “An important concept here is in patients developing a secondary primary cancer, we do not expect to see MRD using a personalized assay since the assay is looking for a tumor signature associated with a primary tumor that was removed at surgery.”

Investigators went on to analyze whether preoperative ctDNA shedding is associated with histological subtype in patients with early-stage NSCLC. Of 88 early-stage preoperative samples, ctDNA was detected in 49% of lung adenocarcinomas (21 of 43), 100% of lung squamous cell carcinomas (33 of 33), and 75% of other NSCLCs (9 of 12). The other NSCLCs were large cell carcinomas (4 of 5), large cell neuroendocrine carcinomas (2 of 2), pleomorphic carcinomas (3 of 4), and adenosquamous disease (0 of 1).

These findings validate the results of a 2017 analysis of 96 early-stage preoperative samples that showed preoperative detection of ctDNA varies by histological subtype, Abbosh said. That study showed that ctDNA shedding was detected in 19% of lung adenocarcinomas (11 of 58), 97% of lung squamous cell carcinomas (30 of 31), and 71% of other NSCLCs (5 of 7).2

To determine the potential clinical impact of preoperative ctDNA detection, Abbosh and colleagues analyzed postoperative samples at different timepoints from 53 patients who relapsed. Of these, 42 had detectable ctDNA preoperatively (ctDNA shredders) and 11 lacked detectable ctDNA preoperatively (ctDNA nonshredders). They found that ctDNA was detectable at or before clinical relapse in 91% of the shredders (38 of 42 patients) compared with 64% of the nonshredders (7 of 11).

In patients who were shredders with detectable ctDNA at relapse, the median lead time from first ctNDA detection postoperatively to clinical relapse was 164 days (range, 6-1022) and the median DFS was 362 days (range, 41-1143). For non-shredders with detectable ctDNA at relapse, the median lead time was 22 days (range, 0-634) and the median DFS was 640 days (range, 404-1242).

“What these data suggest is that preoperative ctDNA detection status would be a proxy of the potential utility of ctDNA as an MRD biomarker in a clinical setting,” Abbosh said.

Additionally, Abbosh said, a comparison of MRD surveillance and standard-of-care imaging for the same 53 patients showed that MRD can be detected before evidence of relapse is observed on imaging. Nine of 10 patients who were MRD positive on a blood test but showed “no evidence of relapse” on surveillance scans subsequently relapsed from their disease. Additionally, 15 of 16 patients who were MRD positive but reported as “new equivocal changes” on scans eventually relapsed. Meanwhile, the rate of second primary cancers was higher than the relapse rate in patients who had “unequivocal new changes” on a scan and no MRD detection (52% vs 48%, respectively.)

In summary, Abbosh said, the findings show that the AMP methodology “can detect low-frequency variant DNA at low assay DNA inputs consistent with an MRD setting” and that preoperative shedding of ctDNA affects MRD outcomes. Based on the study results, the field of NSCLC is ready to move away from conventional adjuvant trials toward MRD-driven studies, he noted.

Conventional trials have heterogenous patient populations with a low relapse-event rate, he noted. Further, adjuvant studies must recruit large numbers of patients, take more than 10 years to complete, and create the risk of toxicities from the escalation of standard-of-care therapy in patients who are cured by surgery.

“The traditional adjuvant trials are challenging because you need hundreds, if not thousands, of patients to adequately power these studies, and you need to follow up patients for many years. These challenges largely arise from the fact that you’re dealing with a heterogeneous patient population, which includes some patients who are cured by surgery and some patients who have residual or metastatic disease, and you can’t differentiate between these populations,” said Abbosh.

Characteristics in MRD-driven trials that could overcome the challenges of conventional adjuvant trials include rapid readout of study data, therapy escalation only in patients who are destined to relapse, and an opportunity to establish DFS surrogates predicated on MRD clearance, Abbosh said.


1. Abbosh C, Frankell A, Garnett A, et al. Phylogenetic tracking and minimal residual disease detection using ctDNA in early-stage NSCLC: a lung TRACERx study. Presented at: 2020 American Association for Cancer Research Virtual Annual Meeting I; April 27-28, 2020. Abstract CT023. bit.ly/3idjVZC

2. Abbosh C, Birkbak NJ, Wilson GA, et al. TRACERx consortium; PEACE consortium; Swanton C. Phylogenetic ctDNA analysis depicts early-stage lung cancer evolution. Nature. 2017;545(7655):446-451. doi:10.1038/nature22364