NEW YORK – Exact Sciences and Guardant Health, two major players aiming for a share of the anticipated blood-based cancer screening market, shared updates this week on their respective multi-cancer early detection (MCED) assays.
At the annual meeting of the American Association for Cancer Research, each company shared results of their tests from early retrospective studies in a handful of cancer types.
Although there is already an MCED available for clinical use — Grail's Galleri — crucial questions have yet to be answered about whether the value of pan-cancer screening outweighs its costs. The field is also grappling with significant unknowns regarding clinical best practices, especially for cancer types without established screening paradigms and known follow-up procedures.
Speaking at AACR on Monday, Guardant Health co-CEO AmirAli Talasaz said that the firm is taking these questions seriously. "For a product to have clinical utility in this field, it needs to exhibit performance metrics optimized for [the particular] cancer or panel of cancers that are getting screened for."
Defining those performance requirements means considering not just the current screening modalities or lack thereof but also patients' anticipated post-screening diagnostic journey, he said.
"For cancers with established screening paradigms and established post-screening diagnostic pathways, our aim, we believe, should be [a test] that increases the compliance to screening," Talasaz said. "For cancers without [these], our aim should be to reduce false positive rates while still ensuring the sensitivity is clinically meaningful."
Practically speaking, this means that as Guardant has been building out data for tumor types beyond its first target, colorectal cancer, the company is developing specific interpretation methods and cutoffs optimized for the cancer type in question.
In his presentation, Talasaz shared new feasibility data for the firm's methylation, fragmentation, and mutation-based next-generation sequencing assay in detecting colorectal, lung, pancreatic, and bladder cancers.
For colorectal and lung cancers, the US Preventive Services Taskforce already recommends screening with available tools that include colonoscopy and stool-based tests. According to Talasaz, about 130 million people in the US are eligible for colon cancer screening and about 15 million are eligible for lung cancer screening. However, "the compliance with these screening modalities have been stubbornly low, at 66 percent for CRC and just 14 percent in the space of lung cancer."
For these cancers, a blood test need only have performances "on par with current screening tests," to make a difference, he said, assuming increased compliance with a more patient-friendly procedure.
But for other cancer types, specificity needs to be high enough that the benefit of screening outweighs the associated risk and harm.
Guardant's AACR data came from a combined cohort of more than 3,500 subjects, including over 1,800 cancer-free cases, over 1,200 colorectal cancers, about 200 lung cancers, 42 pancreatic cancers, and 84 bladder cancers across stages.
In lung cancer, where existing imaging screening methods yield around 76 percent specificity and 80 percent sensitivity, the firm's assay showed overall sensitivity of 91 percent at 90 percent specificity. For early-stage cancers (stage I and II), the test maintained a sensitivity of 87 percent.
For colon cancer, keeping specificity at 90 percent, Guardant's assay showed a 92 percent sensitivity overall, with 90 percent sensitivity in early-stage cases.
For the other two cancer types, the company set its analysis at a higher specificity threshold of 95 percent. With that locked in, the test detected pancreatic cancer with 81 percent sensitivity overall and early-stage sensitivity of 73 percent. In bladder cancer, Guardant saw 76 percent sensitivity overall and 52 percent sensitivity for early-stage cases.
The firm's technology can also predict tumor type, or location, which means that if broadly implemented in asymptomatic individuals, a multi-cancer test wouldn't necessarily leave patients who get a positive test result with a completely unknown path forward to diagnosis and treatment.
In the data presented at AACR, tumor origin prediction was correct at least 86 percent of the time.
According to Talasaz, caveats to the current data include that the samples came from real-world cohorts, which included some cancer cases detected by screening but others from already-symptomatic patients. The control cohort was also self-reported as healthy as opposed to known cancer-negatives.
In a separate symposium on Sunday, Hatim Allawi, Exact Sciences' senior VP of research and technology development, shared results from a retrospective training and validation study using an assay combining the detection of methylated DNA and protein targets identified in previous research.
The 475-subject cohort included 160 cases, representing six cancer types — esophageal, lung, liver, ovarian, pancreatic, and stomach tumors — with 315 matched, purportedly healthy controls.
Working with collaborators at the Mayo Clinic, Exact had previously identified a set of 15 methylated DNA sites that could distinguish cancer samples from controls, plus five proteins with added discriminatory value. The firm's current assay platform uses a modified qPCR chemistry developed in house, called TELQAS (Target Enrichment Long-probe Quantitative Amplified Signal).
In the firm's validation study, investigators split the 475 subjects into a training set and a test set.
Using about two-thirds of the cohort as training samples, the team tuned their assay to yield a sensitivity of 89 percent at a specificity of 98 percent.
They then locked the algorithm and applied it to the remaining third. "As expected, there was a slight drop in sensitivity to 85 percent and specificity of 95 percent," Allawi said. When the locked algorithm was applied to the whole cohort, accuracy bumped back up to 88 percent sensitivity and 97 percent specificity, he added.
Performance for specific cancer types ranged from 73 percent sensitivity for pancreatic cancer to 97 percent for lung cancer. Breaking down the group by stage, Allawi reported that the overall stage I and stage II sensitivity was 76 percent, compared to 92 percent for stages III and IV.
Interestingly, the assay sensitivity appeared to be much less impacted by stage in some cancer types, like liver cancer, compared to others, like pancreatic cancer, he added.
Many lingering questions remain unanswered, though, by the two companies' retrospective validation and feasibility data. Even if tests like Guardant's, Exact's, or Grail's demonstrate sufficient sensitivity and specificity for the cancers covered that the benefits are expected to outweigh the harms, clinicians lack protocols, guidelines, and best practices on the diagnostic follow-up for currently unscreened cancers. All of that will have to be built from the ground up.
Talasaz said that the first step to answering those types of questions will come in the form of prospective interventional trial data. Both Guardant and Exact have such studies ongoing, as do several other competitors.
Guardant has said its ECLIPSE trial in colorectal cancer should read out in a few months, supporting the commercial launch of its test, now branded Guardant Shield. Another, similarly designed study is ongoing in lung cancer, while additional tumor types like pancreatic and bladder cancer are still being explored retrospectively.
Exact is also approaching readout for its BLUE-C trial, prospectively assessing both a second generation of its Cologuard stool test and its blood-based assay for CRC screening. The company said earlier this year that results for the blood test are likely to be released in the first half of 2023.
While Guardant seems to have solidified its test methodology, employing the same core sequencing platform as it builds outward from colorectal cancer, Allawi said Exact is still exploring the technical side as it eyes future multi-cancer implementation.
"As far as what we are doing next, we are definitely expanding to include more cancer types, and we are employing new platforms [like] next-generation sequencing … and newly discovered methylated DNA markers to improve sensitivity and performance," he said.