October 10, 2023 | The feasibility and design of a randomized controlled trial (RCT) for evaluating multi-cancer detection (MCD) assays suited to population-based cancer screening is the aim of a pilot Vanguard Study being undertaken by the National Cancer Institute (NCI). The platform study, which will enroll 24,000 individuals in one of four arms, is scheduled to launch in the first half of 2024 to simultaneously test three MCD interventions against a common control group, according to Lori M. Minasian, M.D., the agency’s deputy director of cancer prevention who presented on the topic at last month’s Next Generation Dx Summit.
These assays could “potentially” improve cancer screening, screen for cancers at organ sites currently without a screening test, detect cancers hard to spot at an early stage (e.g., pancreatic and ovarian), and identify cancer from many different organ sites with only a single test, she says. Too little data exists for the clinical benefits and harms to be known.
“From the perspective of the clinician, it appears that there’s a tsunami of these assays coming forward and how do they begin to choose which ones to use for which scenario?” Minasian rhetorically asks. Physicians, patients, and payers all need evidence demonstrating that their use in cancer screening will improve health outcomes.
“The principles of population-based cancer screening are dependent on a large burden of disease; recognizable pre-clinical stage; the potential for effective treatment in early-stage disease; acceptability of the screening test to patients and physicians and payers; reasonable sensitivity, specificity, and positive predictive value of the screening test; and, ultimately, improvement in all-cause mortality ... from the cancer that you’re screening for,” she begins. Cancer screening exists for the most common cancers in a limited number of organ sites—breast, prostate, lung, and colorectal—but “more than half of the cancer deaths are from organ sites that do not have screening tests currently available.”
Screening with MCD assays comes with many unknowns, notes Minasian. For example, “If we screen a population of asymptomatic patients long-term... will there be a reduction in mortality from cancer?” It is also unknown how many diagnostic tests are needed to make a cancer diagnosis. “What happens if you have a positive MCD test, you go through the diagnostic workup, and then there’s no diagnosis? Do you get tested again in a month? Two months? Three months? A year? What’s that timeframe?”
It likewise remains a mystery how many people might be subjected to unnecessary MCD tests and if they would have complications from those procedures, she continues. Another outstanding concern is whether people who test negative would stop doing their cancer screening. “For some cancers, just because you detect it early doesn’t mean you can make a difference to the clinical outcome.”
To evaluate MCD technologies, the NCI has set up a formal, three-pronged program known as the Vanguard Study. The framework for the pilot feasibility study involves study planning work, development of a network of investigators to evaluate the assays, and an assay selection process. The goal is to “understand how to better evaluate these assays before we launch what is arguably a very large randomized controlled trial,” says Minasian.
Study planning began with a workshop two years ago to explore study design assumptions, she continues. The gathering included academic investigators who were involved in biomarker development, primary care physicians with significant cancer screening expertise, and clinical trials experts, and results of those discussions published earlier this year in the Journal of the National Cancer Institute (DOI: 10.1093/jnci/djac218).
Agreement was reached that the NCI needed to evaluate MCD assays for clinical benefit, and there was significant support expressed by everyone for the need to “vigorously capture and understand fully harms and benefits together,” says Minasian. The group also strongly supported the idea of conducting a pilot study to look at the feasibility of randomization, assess the clinical workflow for the diagnostic pathway, and identify any unexpected issues that might be encountered before launching a much larger trial.
A lot of time was spent talking about clinical trial endpoints and the design of a screening study for a MCD assay, which differs significantly from a trial for a single organ modality where the primary endpoint is mortality from that cancer type and the natural history of the targeted cancer type is well understood, she notes. To complicate matters further, they were trying to figure out how to simultaneously study multiple MCD assays targeting different groups of cancers with different natural histories where staging systems don’t always mean the same thing for every kind of cancer.
Then there’s the question of what next steps would follow a positive result from a MCD assay because there is no clear pathway as there would be for, say, a positive low-dose CT scan for lung cancer, says Minasian. Having a tissue of origin might point clinicians in the right direction, but if there is no tissue of origin then a “whole body image of one kind or another” would likely be done.
The presumption has often been that a blood-based MCD test is much easier than these imaging modalities, she adds, but that might not be the perspective of patients subject to both diagnostic methods. “Acceptability is more than just the test itself.”
In the NCI’s National Lung Screening Trial (NLST), “the workup for a positive finding was pretty much left to physician discretion because... the variability was relatively limited [across study radiologists],” says Minasian. In a clinical trial that will evaluate MCD assays, “we have to think about what it means to direct or not direct what the diagnostic pathway would be.”
The NLST enrolled about 50,000 smokers who were randomized to low-dose CT scan or chest X-ray, had three annual screening rounds, six to seven years of diagnostic follow-up after a positive screening, and the primary outcome powering the study was lung-cancer-specific mortality, she reports. “The challenge with the NLST is that the uptake has been pretty poor.”
A chief advantage of the NLST, in addition to the mortality endpoint, is having a tool to identify nodules of indeterminate significance and understand them, says Minasian. “If you look carefully at the study design [for the MCD assay RCT], there is a huge opportunity for us to detect a variety of cancers early and begin to understand potentially the natural history of early-stage disease that we’ve not really seen before.”
“The [current] paradigm in oncology drug development has focused on understanding the extent to which a drug can modify the natural history of active disease, so virtually every drug goes through early phase trials where we look at the response rate to late-stage disease and based on.... [that] then we move it up earlier,” she later adds. “The purpose of setting up something like this [platform RCT] is to in fact be able to capture and understand how we can start capturing earlier stage much more efficiently... and then potentially be able to develop agents that are uniquely targeted to early-stage disease.”
How to structure the primary endpoint in an MCD clinical trial is not a straightforward proposition, Minasian says. “Is it mortality for every kind of cancer? Is it mortality just from the cancer that assay detects? Do we include those cancers for which there is standard-of-care screening? Do we exclude the rarest cancers?”
The “current best thinking” coming out of the of the study planning workshop is to move toward a huge platform-style RCT where participants would be randomized to a different MCD assay or the control arm and everyone would be offered standard-of-care screening. The assays would be compared to the one, larger control arm independently and not compared to one another.
The control arm will need more participants than any of the intervention arms, since those are the cancer cases that get used more than once, Minasian says. “But a participant being randomized would still have a better than 50% chance of being [assigned] to an intervention.”
The primary endpoint for each intervention arm would be morality from the cancers being detected by that assay, relative to the same set of cancers in the control arm, she explains. “This then allows us to add new arms to this platform structure over time... that’s independent of the other intervention arms [and] dependent upon the control arm.” These could include technologies other than MCD assays, she notes.
The entry of a new arm to the platform study would trigger a reassessment of the control group to ensure “sufficient and appropriate controls,” says Minasian. “This becomes a highly efficient way to evaluate multiple different assays. We potentially could consider an arm with more than one assay in it, but then those two companies would have to be highly collaborative because they’d be able to see each other’s information about the performance characteristics for that arm.”
To answer a question every company has asked—how to consider that assays may well improve between the start and end of the study period—Minasian says that generally what’s changing over time is the software algorithm determining what the cut point is, not the biochemical measurement itself. Even with the Vanguard study, although the assay would be locked down at the beginning, the NCI will prospectively include an endpoint to reevaluate the data based upon the new algorithm at the end, she explains. Some sort of accommodation will still have to be made if the upgraded algorithm detects additional other cancers.
Statisticians estimate that the RCT timeline for an assay that detects lung cancer would be between five to 10 years, depending on the population that gets recruited, she reports. Although mortality rates among study participants will be relatively low even after a decade, it should be possible to get to the primary endpoint in a “timely fashion” since most of the cancers being detected are not routinely screened for anyway.
Given the long wait for results on mortality endpoints, some experts in the study planning workshop suggested that the study instead use stage shift, she shares. Others in the group countered with their concern about stage shift having too much bias.
“Different types of cancers may not have the same response to therapy, and to date there have been no completed randomized controlled trials using MCD assays for cancer screening, which would in fact validate it for use of advanced stage shift as a surrogate for mortality,” says Minasian. Over time, the envisioned RCT could evaluate potential surrogates “because everything would be anchored to mortality endpoints.”
Minasian’s prime example for not using stage shift is prostate cancer versus pancreatic cancer, with a five-year survival rate of 97.1% and 12.5%, respectively. In the case of ovarian cancer, the figure is 50.8% because most patients are diagnosed at a late stage of the disease.
As was demonstrated by the 20-year UK Collaborative Tral of Ovarian Cancer Screening study (The Lancet, DOI: 10.1016/S0140-6736(21)00731-5), earlier detection doesn’t necessarily translate to changes in mortality, she says. Lacking an effective therapy for early-stage disease, the U.S. Preventive Services Task Force in fact recommends against screening in asymptomatic women.
It was the same story with a 15-year-old intervention trial for high-grade serous ovarian cancer—the most-deadly variety—where one group of patients with stage II and III disease were immediately followed every three months with CA-125 blood tests. They had their cancer recurrence detected five months early and began therapy. “They had far more toxicity and side effects but no survival advantage whatsoever,” says Minasian.
To develop a network of investigators, the NCI released a funding opportunity announcement last November, Minasian says. The network will have three components, including accrual and enrollment sites, a coordinating and communication center, and a statistics and data management center.
The announcement received a “robust response,” she says, adding that the network should be established by early next year. The network will ultimately develop cancer screening trials to evaluate a variety of emerging technologies, although “the primary function is to assess the clinical utility of cancer screening technologies and programs for the purpose of improving health outcomes.” It will be applying precision medicine approaches, such as risk stratification, as well as “evaluating the effectiveness and feasibility of screening strategies, conducting surveillance of cancer screening, and developing cancer screening studies to evaluate the clinical workflow.”
Step one is to develop the Vanguard study, which will look “very similar” to the proposed RCT other than being shorter and smaller, says Minasian. The ratio of total patients receiving an active intervention versus a placebo will also not be 1:1. It will have specific objectives helping the NCI refine the design for launching the larger RCT, including how many patients will be needed in the control arm.
The initial study will involve two blood draws, at baseline and year one, to pilot the idea of multiple blood draws over time and will not have a mortality endpoint, she stresses. “This is simply a feasibility study looking at the willingness of participants to be randomized, to determine the adherence to the diagnostic follow-up, determine the feasibility of diagnostic workflows... determine the reliability and timeliness of the return of results, and identify any potential barriers or facilitators for recruiting a generalizable population.”
The 24,000-person sample size was driven by the fact that many MCD assays have established their specificity at 98-99%, says Minasian, so 1-2% positive screens are anticipated. “The whole purpose of this is to have sufficient numbers of positive screens to be able to work them up, understand them, and learn something from that process.”
A broadly announced virtual workshop in early May kicked off the assay selection process, with applications due in June. Those are now being reviewed, she says. The process includes verification of the performance characteristics of the MCD assays using an NCI-developed Alliance Reference Set or similar tools.
The NCI will be funding the investigator network for four years, says Minasian. “If all goes well, we anticipate that the Vanguard study itself will take two to three years.