A few years in, a new picture for liquid biopsy​

An approach that fundamentally changes the sensitivity of liquid biopsy is starting with tumor tissue to inform the analysis and then “diving into liquid to find low-level signal with greater sensitivity,” Dr. Oxnard says. Foundation Medicine is exploring this approach for measurable residual disease monitoring with its FoundationOne Tracker ctDNA monitoring assay. “It takes baseline tissue NGS, designs bespoke probes, and then searches with high sensitivity for that specific signal, to monitor that signal at low levels, and to look for residual disease after curative care.” The research version of the assay was released in June.

Ignatiadis, et al., in an article published in May, said “the next frontier” for the clinical use of liquid biopsy is likely to be “the systemic treatment of patients with ‘ctDNA relapse,’” which is their term for ctDNA detection prior to imaging-detected relapse (Ignatiadis M, et al. Nat Rev Clin Oncol. 2021;18[5]:297–312). But blood testing on its own, Dr. Oxnard says, isn’t likely to supplant current methods for diagnosing recurrent cancer.

He recalls a patient case of recurrent lung cancer with a positive bone scan: “I sent a liquid biopsy and it showed the patient’s EGFR mutation. And I asked my colleagues, ‘Can I treat based on this blood test, or do I need to get a biopsy to confirm advanced recurrent lung cancer?’” His colleagues advised him to order the biopsy, arguing the patient was owed pathologic confirmation. “That was many years ago. Today, if we can sacrifice the biopsy, we do. But I do think there’s something true about pathologic confirmation, especially with regard to incurable cancer.” Using blood testing to instigate clinical workup and accelerate detection of recurrence is more likely. “We know there’s something true about biopsy confirmation, and I don’t see us losing that in the near term.” Rather, he says, “liquid biopsy will make us more nimble, more patient centered, and more efficient in how we undergo diagnostic workup and get patients the therapies they need.”

Population-level screening for multiple malignancies using genomewide analysis of cfDNA is another application. The Circulating Cell-Free Genome Atlas (CCGA) study (NCT02889978), supported by Grail, is an ongoing effort to determine if genomewide cfDNA sequencing in combination with machine learning can detect and localize multiple cancer types at sufficiently high sensitivity and specificity for general population screening.

This multicenter, longitudinal, case-controlled observational trial is split into three pre-planned substudies. Biospecimens were prospectively collected from participants with newly diagnosed untreated cancer and from healthy controls. CCGA1 focused on assay development. Three prototype sequencing assays were performed: paired cfDNA and white blood cell targeted sequencing of 507 genes for single nucleotide variants/indels; paired cfDNA and WBC whole genome sequencing for copy number variation; and cfDNA whole genome bisulfite sequencing (WGBS) for methylation. WGBS had the highest sensitivity and was taken forward for further assay and clinical development (Klein EA, et al. J Clin Oncol. 2018;36[15 suppl]:12021; doi:10.1200/JCO.2018.36.15_suppl.12021).

“The focus on methylation avoids the need for WBC sequencing to avoid the variable of clonal hematopoiesis of indeterminate potential that confounds targeted sequencing,” says Minetta C. Liu, MD, CCGA co-investigator (with Dr. Oxnard and others) and research chair, Department of Medical Oncology, Mayo Clinic. Methylation signals are also organ specific, allowing for tissue localization of the cancer signal. “This is critical in guiding the clinical evaluation to diagnostic resolution,” Dr. Liu says.

CCGA2 followed, she says, to develop, train, and validate a targeted methylation assay to classify cancer versus non-cancer and identify the tissue signal origin for multi-cancer detection across all stages. A total of 6,689 participants (2,482 with cancers of more than 50 types and 4,207 without cancer) were divided into training and validation sets. Plasma cfDNA was collected and subjected to bisulfite sequencing, targeting a panel of more than 100,000 informative methylation regions. A classifier was developed using machine learning algorithms. The classifier’s performance was consistent across the training and validation sets: specificity was 99.3 percent in the validation set (CI: 98.3 percent to 99.8 percent; 0.7 percent false-positive rate). And stage I to III sensitivity was 67.3 percent in a prespecified set of 12 cancer types and 43.9 percent across all cancer types. Sensitivity of detection increased with increasing stage. In the prespecified types (anus, bladder, colon/rectum, esophagus, head and neck, liver/bile duct, lung, lymphoma, ovary, pancreas, plasma cell neoplasm, stomach), sensitivity was 39 percent in stage I, 69 percent in stage II, 83 percent in stage III, and 92 percent in stage IV. Detection increased similarly across all 50 cancer types. Tissue of origin was predicted in 96 percent of samples with a cancer-like signal, and of those, tissue of origin localization was accurate in 93 percent (Liu MC, et al. Ann Oncol. 2020; 31[6]:​745–759). CCGA3 was designed for further refinement and validation of the classifier in a large population (Klein EA, et al. Ann Oncol. Online ahead of print June 23, 2021. doi:10.​1016/​j.annonc.​2021.05.806).

The CCGA study recruited participants with a known cancer diagnosis in order to develop and validate the multi-cancer early detection test, now known as Galleri. “In parallel with CCGA, Grail has been conducting other prospective, longitudinal studies in asymptomatic intended use populations, where biospecimens and clinical data are collected to support population-based cancer screening,” Dr. Liu says. These large-scale efforts have established the necessary clinical validation cohorts post-assay development, she says. These studies include STRIVE (NCT03085888), which enrolled approximately 100,000 female participants at the time of screening mammography from 35 U.S. clinical sites (including five Mayo Clinic sites), as well as SUMMIT (NCT03934866), which enrolled approximately 25,000 smokers and former smokers in the United Kingdom at high risk of lung cancer.

Longitudinal follow-up is a critical component of these trials, Dr. Liu says, especially for participants who have a cancer signal-detected test result but a subsequent unremarkable diagnostic workup. Those participants could develop cancer later, reflecting a lead-time bias. Participants without a detected signal also may develop cancer later. “These observations will provide critical insight into overall test performance,” she says.

PATHFINDER (NCT04241796) is a prospective pilot implementation study of about 6,200 participants without a known diagnosis of malignancy, in which results of Grail’s Galleri test are returned to participants and their clinicians. Participant-reported outcomes and perceptions of the test are ascertained, and diagnostic pathways are recorded toward resolution of a signal-detected test result. The first results were presented at the ASCO annual meeting in June, simultaneously with Grail’s June 4 announcement that Galleri is available in the U.S. by prescription.

Of the 6,629 individuals age 50 or older enrolled in PATHFINDER, Grail’s test accurately detected 29 cancers across 13 types: breast, colon and rectum, head and neck, liver and bile duct, lung, lymphoid leukemia, lymphoma, ovary, pancreas, plasma cell neoplasm, prostate, small intestine, and Waldenstrom macroglobulinemia. Of the new cancers detected, nearly 40 percent (9/23) were localized (stage I–II), and 13 of 23 were detected before distant metastases (stage I–III). The positive predictive value was 44.6 percent (95 percent CI: 33.2–56.7 percent), which is consistent with findings from the CCGA study. Final results are expected in the first half of 2022.

Multi-cancer early detection screening detects malignancies that have no current screening paradigm, such as pancreatic and ovarian cancer, Dr. Liu says. “This is critical, as over 70 percent of cancer-related deaths between ages 50 and 79 are attributed to malignancies without recommended standard screening options.” Galleri and other multi-cancer early detection tests under development have demonstrated the ability to detect multiple cancer types across all stages of disease, she says. “It is not surprising that sensitivity of detection increases with increasing stage because circulating tumor DNA is a function of tumor burden.” But not every tumor sheds DNA into the circulation at the same rate, Dr. Liu notes, “so there is a function of tumor type and underlying tumor biology that plays into sensitivity of detection.” In CCGA2, cancers of every stage were detected, “but stage distribution is different across different cancers. For example, we remarkably detected a fair number of pancreatic tumors at an early stage, when intervention is more likely to reduce cancer-related mortality.” In the validation set, sensitivity of detection for pancreatic cancer was 63 percent in stage I and 83 percent in stage II.

Dr. Liu emphasizes that blood-based multi-cancer early detection tests should be used in conjunction with standard-of-care screening recommendations. “The goal is to enhance—not replace—single cancer screening paradigms,” she says. “We also have to keep in mind that ctDNA has a lower limit of detection. We are only as good as the assays are, and if we don’t detect a cancer signal, it doesn’t mean that cancer is not there.”

Practical issues are also under consideration, specifically with respect to incorporating multi-cancer early detection blood tests into physician practices. “We need to develop operational workflows related to ordering, interpreting, and managing the test results,” Dr. Liu says. “We assume primary care providers will order the blood test, as they already order mammograms, low-dose chest CT scans, and the like. But who will direct the diagnostic workup for a cancer signal-detected result, or follow patients who have a cancer signal-detected result but no identified malignancy? We are in the midst of a paradigm shift in cancer screening that requires collaboration across medical specialties to incorporate multi-cancer early detection into our general practices.”

“And those discussions are gratefully taking place.” 

Charna Albert is CAP TODAY associate contributing editor.