, according to findings from the TRACERx study.
The findings pave the way for clinical trials of MRD-driven treatment escalation, Chris Abbosh, MD, of University College London, reported during a presentation at the AACR virtual meeting I. Data in the presentation were updated from the abstract.
Dr. Abbosh and colleagues used phylogenetic circulating tumor DNA (ctDNA) profiling to assess MRD and predict relapse in patients from the TRACERx study who underwent surgery for stage I-III NSCLC.
“The approach we take is technically termed a ‘tumor-informed, personalized cell-free DNA-enrichment approach,’” Dr. Abbosh explained. “We take out the primary tumor from the patient, we multiregion sample that tumor, and submit each region for deep whole-exome sequencing.”
The researchers prioritize variants for MRD tracking based on clonality/subclonality, high copy number status, and low background sequencing noise. The researchers then construct an anchored-multiplex PCR panel against the positions of interest, which is applied to cell-free DNA in the pre- and postoperative setting.
“We’ve developed an MRD caller to go alongside this chemistry,” Dr. Abbosh said. “The main premise behind the MRD caller is that it can calculate intralibrary error rates to inform the MRD pool.”
Sensitivity and specificity
To validate their approach, Dr. Abbosh and colleagues tested the assay with low DNA input (5 ng, 10 ng) and high DNA input (30 ng, 60 ng). They found the assay to be more sensitive with higher DNA input, and variant fractions were detected down to 0.003%.
The researchers also assessed how sensitivity and specificity scale with an increasing number of variants – 50, 100, or 200 variants. When tracking 200 variants, the assay was powered to detect lower ctDNA fractions than when tracking 50 variants. On the other hand, specificity was higher with 50 variants (99.8%) than with 200 variants (99.4%).
Next, Dr. Abbosh and colleagues analyzed postoperative cell-free DNA collected at 271 time points from 37 NSCLC patients who did not relapse. This included 11 patients who developed proven second primary malignancies.
Of the 271 time points when MRD negativity was expected, MRD was not detected at 269 time points, which translates to 99.3% specificity for the assay.
Shedding, relapse, and disease-free survival
Dr. Abbosh and colleagues also found that non-adenocarcinoma histology is associated with preoperative ctDNA shedding in NSCLC. The researchers analyzed 88 early-stage preoperative samples from NSCLC patients. ctDNA was detected preoperatively in 49% of lung adenocarcinomas and 100% of lung squamous cell carcinomas.
“This finding is important when it comes to interpreting our non–small cell lung cancer relapse data from 53 TRACERx patients,” Dr. Abbosh said.
Of the 53 patients who relapsed, 42 had ctDNA detected prior to surgery and were thus considered shedders, while 11 were nonshedders. ctDNA was detectable at or before relapse in 91% (38/42) of shedders and 64% (7/11) of nonshedders.
The median time from ctDNA detection to clinical relapse was 164 days in shedders and 22 days in nonshedders. The median disease-free survival was 362 days and 640 days, respectively.
“So what these data suggest is that preoperative ctDNA detection status will be a proxy of the potential utility of ctDNA as an MRD biomarker in a clinical setting,” Dr. Abbosh explained.
Standard-of-care imaging findings in the 53 patients who relapsed further demonstrated the utility of ctDNA in this setting, Dr. Abbosh said.
All scans were divided into three categories: those showing unequivocal relapse, those with a new equivocal change (relapse, inflammation, or a nonspecific finding), and those with no evidence of relapse. Each was further categorized by preimaging MRD status.
Relapse occurred in 9 of 10 patients who were MRD positive but had a scan showing no evidence of relapse. Relapse occurred in 15 of 16 patients who were MRD positive and had scans showing new equivocal changes.
Patients with unequivocal evidence of relapse who were MRD negative at or before the scan were more likely to have a second primary cancer than to have NSCLC relapse (52% vs. 48%), which is a reflection of the specificity of the MRD assay to the primary tumor, Dr. Abbosh said.