NEW YORK – Blood-based minimal residual disease testing has demonstrated great prognostic potential in colon cancer, but its translation into successful randomized controlled trials continues to be a challenge, according to new data shared last week at the American Society of Clinical Oncology Gastrointestinal Cancers Symposium (ASCO GI).
Investigators from trials of two leading commercial tests illustrated the two sides of this coin during a Saturday discussion session at ASCO GI. Research presented at the symposium included new data from the ongoing GALAXY trial, in which Japanese researchers have been prospectively tracking circulating tumor (ctDNA) in colorectal cancer patients using Natera's Signatera tests, which use upfront tumor sequencing to create bespoke ctDNA panels for each patient. Meanwhile, investigators also presented the first official report of final Phase II data from the COBRA trial of Guardant Health's blood-only tumor-agnostic Reveal assay, which was closed last year after it failed an interim futility analysis.
Discussing both trials during the meeting, Aparna Parikh, an associate professor of medicine at Massachusetts General Hospital who was not involved in either trial, stressed that in general, MRD researchers continue to seek answers about both analytic/clinical validity and utility.
"We certainly have many answers, but I'll make the argument that we also have many questions left to answer," she said.
More and more, Parikh added, trials like GALAXY have begun to concretize the analytic and clinical validity of what has become a crowded field of commercial and pre-commercial assays.
Guardant's COBRA trial, meanwhile, had been poised to address the question of clinical utility — but its failure illustrates the remaining challenges, which she and others now hope to overcome in ongoing trials.
A disappointing outcome
COBRA, which was led by trial research group NRG Oncology, recruited more than 600 early-stage colorectal cancer patients over 3.5 years, randomizing these individuals, who would not have normally been recommended adjuvant chemotherapy, to a treatment arm where those with positive results would get chemo, and a surveillance arm, where ctDNA would still be banked and analyzed, but not acted upon.
In total, there were 16 ctDNA-positive patients at baseline — seven in the surveillance arm and nine who received chemotherapy. But in direct contrast to the aim of the trial, which was to demonstrate a benefit from offering chemo to ctDNA-positive patients, there were actually more cases who cleared their ctDNA after six months with no treatment than there were treated patients with the same response.
Based on evidence from observational studies, chemotherapy administration should have led to ctDNA clearance in most patients, but only one of the treated individuals cleared their ctDNA while three untreated control subjects did.
Investigators had no choice but to close the trial.
Van Morris, associate professor of GI medical oncology at the University of Texas MD Anderson Cancer, who presented the COBRA findings, said during the meeting that while it's not clear yet why this inverted result occurred, the researchers are in continued discussions with the National Cancer Institute and Guardant to try to get to the bottom of it.
"We are every bit as surprised and frankly disappointed by these results" as anyone, he said.
Discussing possible factors at play, MassGen's Parikh suggested that simple stochastic sampling issues could have contributed, with the clearances reflecting a biological self-clearing phenomenon that other trials, including GALAXY, have recorded. However, previous data on self-clearing have mostly shown much lower frequencies.
Assay specificity, she added, can certainly have a greater effect in low-risk or rare disease settings than in higher-risk settings.
For example, she posed the question of a hypothetical cohort of 1,000 patients tested with an assay with 50 percent sensitivity. If that cohort had low risk of recurrence — a 10 percent rate for example — dropping specificity from 98 percent to 95 percent would lead to a reduction in positive predictive value of more than 20 percent. The same drop in a population with a 50 percent risk of recurrence would only lower PPV by 10 percent.
"We have to think about assay performance in the patient population that we are going into," Parikh said.
When news broke last year that the COBRA trial was closing, Guardant Health was vocal in denying that the failure was due to its assay's performance or to an issue of false positives, as some outside parties had suggested.
However, NRG Oncology did cite a possibly greater-than-expected false-positive rate when it supplied letters to trial investigators to share with involved patients.
Guardant responded in a statement this week that it was aware of these letters and had raised concerns about certain statements and characterizations, but ultimately the decision fell to NRG.
"It cannot be said for certain whether any individual patient in COBRA received a false-positive result since data on recurrence-free survival are not yet available," the company argued this week.
At least one case report does exist noting a putative false positive returned by the Reveal assay in a patient with a non-cancer disease state that researchers hypothesized had led to a methylation signal mimicking cancer. Reported by a group led by Weill Cornell Medicine researcher Pashtoon Kasi in Case Reports in Oncology, the findings were made when both Reveal and Signatera tests were performed in a patient being treated for stage III colorectal cancer, and the results conflicted, with Guardant's test a positive, and Natera's a negative.
Both tests were then negative after the patient was treated for hepatitis and remained negative during a long period of follow-up with repeat testing. Based on this, the authors hypothesized that methylation changes due to viral infection had confounded the initial Reveal test.
Guardant Chief Medical Officer Craig Eagle reiterated in an email this week that since COBRA was initiated, the firm has made multiple updates to its assay and "remains confident about [the test's] promise to significantly improve patient care."
The first version of Reveal combined mutation detection and epigenomic analyses, but a newer version focuses on epigenetics with much greater sequencing coverage at approximately 15 Mb compared to 450 kb for the assay used in COBRA.
Investigators also presented interim data at ASCO GI from the COSMOS study, which evaluated the use of this newer version of Guardant Reveal to identify MRD and predict disease recurrence in 130 patients with stage II and III colon cancer. "The data suggest the test is both highly specific and predictive for recurrence, without dependence on a tissue sample," Eagle said.
The company did not comment on how many, if any, of its early clinical/commercial patients may have been tested using the original Reveal assay that featured in COBRA. The firm launched its test clinically in early 2021, a little more than a year after COBRA began its accrual.
MD Anderson's Morris said that despite the COBRA outcome, he is confident that randomized trials remain not just feasible but necessary in the MRD space. One major takeaway, he noted, is the necessity of banking samples in future efforts to account for continuous evolution and improvement of assays.
Parikh agreed, adding that extra sampling would also allow for inter-assay comparisons, "so we don't keep reinventing the wheel."
"As a community, we should absolutely encourage and demand that this happens," she added.
Companies have yet to jump on that particular bandwagon thus far, though more informal analyses in trials like TRACERx have found some evidence that there could be significant sensitivity improvements associated with MRD assays that include more targets compared to those with fewer.
Future still bodes well
Although the MRD field continues efforts to prove utility, a randomized trial failure is a tough pill to swallow when the clinical validity and prognostic power of genomic MRD tests have been so clearly demonstrated, both Morris and Parikh said.
"Time and time again, in study after study, ctDNA has emerged as the most powerful prognostic biomarker we have," Parikh explained.
The most recent update to Natera's GALAXY study — being conducted under the larger CIRCULATE-Japan effort — was highlighted in the same ASCO discussion session and offers what Parikh called the largest prospective cohort of its type to date, with what is now 24 months of follow-up.
Investigators confirmed in their newest data that serial testing picks up more patients with emerging ctDNA than tests performed at an initial landmark time point, and that patients treated with adjuvant therapy can clear their ctDNA, some for a sustained time period. Even just a decrease of at least 50 percent in ctDNA was associated with better disease-free survival.
With 24 months of tracking, GALAXY has also offered a more detailed window into the breakdown of outcomes in the adjuvant-treated population, which appears to separate into two main groups: those who have sustained ctDNA clearance and good outcomes versus a combined group that includes non-clearers plus those who clear their ctDNA but only transiently. Both groups end up with similar rates of recurrence by the two-year mark.
Whether this transient clearance is biological, or potentially reflective of the sensitivity limits of Signatera MRD tests, is up in the air, Parikh said.
For the clinic, Parikh cautioned that it's important to consider the meaningfulness of sensitivity at different time points for different tests. In GALAXY, the landmark sensitivity has been in line with what other studies of other technologies, including Guardant's, have demonstrated, she said. And the improvement in sensitivity with serial testing also reflects what has been seen in other trials.
The landmark, post-surgery time point is what is most actionable in the clinic right now, Parikh argued, because that is when clinicians and patients are actually making adjuvant chemotherapy decisions.
In her own clinical practice, she said that she already considers treatment escalation for ctDNA-positive patients given ctDNA's demonstrated prognostic power and early data suggesting improvement in short-term outcomes with adjuvant chemotherapy.
"[For] a patient that is stage II and has maybe one high-risk feature where you're not entirely sure about giving adjuvant chemotherapy, if they're positive … depending on the test you are using [you might feel] more comfortable proceeding," Parikh said at ASCO GI. "Or, sometimes in a stage III patient with marginal performance status … if it's positive, I might push a little bit harder to give the chemotherapy."
Early data from Natera's BESPOKE CRC trial was also presented at ASCO GI. This study is also observational rather than randomized, but according to Alexey Aleshin, Natera's chief medical officer and general manager of oncology and early cancer detection, the fact that its results match so closely with what has been reported in GALAXY, but in a completely new US cohort, has raised the firm's confidence in its platform.
For ctDNA-positive patients receiving adjuvant chemo in BESPOKE, there was significantly longer average disease-free survival compared to those undergoing observation. No benefit to adjuvant treatment was observed in ctDNA-negative patients.
A separate ASCO GI poster surveying patients on their experiences in the study concluded that more than 90 percent valued the information they received from their Signatera results, while 73 percent reported that Signatera results reduced their anxiety about cancer recurrence.
Based on the current GALAXY data and evidence for other commercial tests, Parikh said that she views sensitivity as still too low to be used to de-escalate treatment in patients for whom chemotherapy would be otherwise indicated.
And although she does not yet believe oncologists can use surveillance testing to guide treatment decisions, MRD testing for ongoing monitoring is something that is a shared decision with patients, she added.
She also noted that there are already ongoing clinical trials where lingering ctDNA positivity is being used to recruit and randomize patients for additional chemotherapy or molecularly targeted treatment after an initial standard adjuvant regimen. Other efforts, including one by Parikh and her colleagues at MGH, are also in the works to extend this approach to novel therapies such as antibody-drug conjugates.
Investigators at ASCO GI also shared a short abstract describing another randomized trial, SU2C ACT3, which is using Guardant's Reveal assay. Although that effort, in stage III patients, was described by the researchers as initially struggling to meet its recruitment goals, it could prove the clinical utility that COBRA could not.
A sub-analysis from MD Anderson's INTERCEPT program was also highlighted in a poster. In that trial, researchers are using Signatera in post-adjuvant patients, with ctDNA positivity upgrading individuals to extra treatment with TAS-102. Although the trial is not randomized, early data indicated that more than 50 percent of treated patients had cleared their ctDNA at three months, dropping to about 30 percent at six months.
"It's a small study size but 10 patients had a decrease in their ctDNA with this treatment," Natera's Aleshin said in an interview. "Five actually had clearance of detectable levels. And at six months, three patients were still cleared and these are patients that we typically would all expect all to recur."
"It's early signs that actually treating these ctDNA-positive patients can be beneficial," he added.
Natera's randomized treatment escalation trial ALTAIR, another sub-study under the CIRCULATE-Japan umbrella, completed its recruitment in mid-2023 and is expected to read out results in the first half of 2024.