If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Current imaging-based cancer screening approaches provide useful but limited prognostic information. Complementary to existing screening tests, cell-free DNA–based multicancer early detection (MCED) tests account for cancer biology [manifested through circulating tumor allele fraction (cTAF)], which could inform prognosis and help assess the cancer's clinical significance. This review discusses the factors affecting circulating tumor DNA (ctDNA) levels and cTAF, and their correlation with the cancer's clinical significance. Furthermore, it discusses the influence of cTAF on MCED test performance, which could help inform prognosis. Clinically significant cancers show higher ctDNA levels quantified by cTAF than indolent phenotype cancers within each stage because of more frequent mitosis and cell death combined with increased trafficking of cell-free DNA into circulation because of greater vascularization and depth of tumor invasion. cTAF has been correlated with biomarkers for cancer aggressiveness and overall survival; cancers with lower cTAF had better survival when compared with cancers with higher cTAF and with the Surveillance, Epidemiology, and End Results–based survival for that cancer type at each stage. MCED-detected cancers in case-control studies had comparable survival to Surveillance, Epidemiology, and End Results–based survival at each stage. Because many MCED tests use ctDNA as an analyte, cTAF could provide a common metric to compare performance. The prognostic value of cTAF may allow MCED tests to preferentially detect clinically significant cancers at early stages when outcomes are favorable and avoid overdiagnosis.
Current approaches to cancer screening include imaging-based tests, such as mammography for breast cancer or low-dose computed tomography for lung cancer, and tissue visualization-based tests, such as colonoscopy for colon cancer. These approaches provide useful but limited information on the prognosis of the disease, and they potentially contribute to overdiagnosis of cancer.
Recently, molecular markers have been utilized for prognosis, therapy predictions, and treatment recommendations for cancer patients through tissue-based gene expression tests in melanoma, prostate, lung, colon, and breast cancer [eg, OncotypeDX (Genomic Health, Redwood City, CA) has been used for breast and colon cancer prognosis prediction and staging guidelines].
Just as use of molecular markers has revolutionized cancer treatment decisions, cancer screening technologies could also benefit from the use of molecular biomarkers to inform disease prognosis. As novel biomarkers emerge to select treatment and determine the prognosis of individuals already diagnosed with cancer, the question remains of whether there are biomarkers with prognostic capabilities for early detection tests and how these biomarkers can aid in avoiding overdiagnosis.
A screening or early detection test should preferably detect cancers in need of a timely treatment decision while avoiding burdening the health care system with the workup of indolent neoplasms. A new paradigm through which this may be achieved is multicancer early detection (MCED) tests, which evaluate analytes, such as circulating tumor DNA (ctDNA), a direct marker for the presence of an invasive malignant process in the body. This review aims to assess whether ctDNA as a biomarker in the screening setting possesses capabilities to characterize cancer similar to biomarkers employed for patient management after disease confirmation, and thus could enhance the utility of cancer screening as we know it. Furthermore, circulating tumor allele fraction (cTAF) is proposed as a metric to compare MCED tests because of its superior prognostic potential than cancer type and/or clinical stage.
Cancer Biology of ctDNA Release
Focusing on the biological signals that tumors emit in the body can help design better tests for screening or early detection of cancer. ctDNA is cell-free DNA (cfDNA) that is shed (ie, released) from tumors during cellular apoptosis and necrosis
and reaches the circulation. This genetic material in the circulation is packaged into short fragments of on average 165 nucleotides. Not accounting for DNA fragmentation has led to vast underestimation of the ctDNA in cfDNA samples that is available for detection by MCED tests in some studies.
and has been found to vary by histologic or molecular tumor subtype, which may be partially explained by the mitotic activities of different subtypes. In addition, trafficking of cfDNA into the circulation depends on the anatomic location of the tumor.
and due to potential shedding into systems other than lymph or blood vessels, like airways, intestinal lumen, urinary system, or cerebrospinal fluid. Notably, ctDNA levels do not depend only on the release rate of cfDNA from tumor cells into the tumor microenvironment (TME), but also on presence or absence of transport pathways from the TME into the circulation. Thus, trafficking of cfDNA into the circulation is a distinct process that affects circulating tumor fraction and the detectability of ctDNA in a screening test.
Circulating tumor fraction represents the fraction of cfDNA in the circulation that originates from a tumor (ie, ctDNA proportion of all cfDNA). ctDNA carries the genomic and epigenetic state of the primary tumor and metastases if present, including fragments that contain tumor-specific somatic small variant and/or abnormal methylation and whose count reflects copy number alterations. Herein, we use the term allele to refer to genetic and epi alleles. Because not all tumor-derived DNA at a mutant locus harbors the tumor allele because of heterozygosity and/or intratumor and intertumor heterogeneity, cTAF is used as a measure of ctDNA signal availability for many cancer detection tests. cTAF is a modeled value that estimates the expected fraction of cfDNA in the circulation that originates from a tumor and contains a tumor-specific allele. Observations of allele frequencies at multiple loci of the tumor genome are combined into a single cTAF estimate. cTAF is therefore an attractive metric to evaluate the performance of an MCED test as it estimates the expected abundance of tumor-specific alleles in a sample. Consequently, the number of tumor-specific features for MCED tests is calculated as the total cfDNA fragments multiplied by cTAF. For example, with an MCED test that scans 105 regions to a depth of 100× per region, a sample with a cTAF of 10−4 would have an expected yield of 1000 tumor-specific alleles. As ctDNA has a short half-life (from 16 minutes to 2.5 hours),
tests based on ctDNA capture the status and activity of a tumor close to the time of a blood draw. Figure 1 visualizes the underlying biology that drives ctDNA levels in an individual with cancer and how the resulting set of frequencies of alleles at variant loci drives the estimation of cTAF.
Several factors related to tumor biology and the associated TME provide insight into what influences ctDNA levels in blood. First, fast-growing, clinically significant tumors have more cell divisions and therefore generate new cancer DNA at a higher rate. These tumors have cell growth rates that often exceed resource limitations (ie, outgrow vascular supply and further support functions of tumor stroma), which, in turn, limits tumor growth rates.
which is a possible contributor to increased trafficking of cfDNA into the circulation. Although all progressing and growing lesions are main contributors of ctDNA, metastatic lesions might contribute even more because they are located where a circulating tumor cell exited the circulation to colonize, and therefore they often have close connection to the bloodstream.
However, in the context of early detection and screening, mechanisms that lead to detectable ctDNA before a cancer metastasizes are of more interest. Even in local or locoregional stages, cfDNA fragments from clinically significant tumors could potentially have a path from the TME into the circulation (eg, due to deeper invasion into surrounding tissues or increased blood supply).
supports that ctDNA shedding can increase with tumor blood flow, perfusion, and angiogenesis, and this provides another mechanistic explanation as to why clinically significant cancers have higher ctDNA levels. As a potential pathophysiological link, increased perfusion and increased vascular permeability increase the flow of interstitial fluid in the TME. Cellular debris from cell death, including cfDNA fragments, can be transported together with such interstitial fluid back into postcapillary venules, terminal ends of lymphatic capillaries, or openings in the tumor neovasculature under development.
All of these factors together suggest that higher ctDNA levels, and consequently higher cTAF, are associated with clinical significance. Prioritizing such cancers for detection and early treatment may have a positive impact on outcomes in a screened population.
Multicancer Early Detection Tests and Their Ability to Preferentially Detect Clinically Significant Cancers
MCED tests are being developed that use next-generation sequencing and machine-learning classifiers to assess genomic features, such as methylation patterns,
However, additional research is needed to fully understand the differences between different cancer types and to inform clinical use for both single-cancer screening and MCED tests. In breast cancer, there were different levels of ctDNA detected, depending on the molecular tumor subtype,
and patients with HER2+/estrogen-receptor–negative tumors had higher ctDNA levels than other patients (HER2–/estrogen receptor negative, HER2–/estrogen receptor positive, HER2+/estrogen receptor positive; P = 0.02).
In multivariate analyses for breast cancer, clinical stage (stage III versus I and IV versus I) and hormone-receptor status were the factors most significantly associated with ctDNA levels and detection of tumor-derived mutations in the blood (P < 0.001 each).
Tumor burden and ctDNA levels were correlated for both localized and metastasized pancreatic ductal adenocarcinoma; pretherapeutic ctDNA levels were associated with shorter disease-free survival in localized cancers, whereas in metastatic pancreatic ductal adenocarcinoma, ctDNA levels were associated with worse overall survival (OS) in contrast to disease-free survival.
cTAF has also been associated with OS and/or progression-free survival in various individual cancer types, such as resectable colon cancer, non–small-cell lung cancer, head and neck squamous cell carcinoma, and advanced biliary cancers.
Although the evidence for the relationship between cTAF and cancer aggressiveness in individual cancers is strong, data are mostly limited to specific cancer types. In the context of MCED tests, evidence has been collected to demonstrate how cTAF is related to cancer aggressiveness and OS across cancer types, as well as MCED test performance.
The Circulating Cell-Free Genome Atlas (CCGA) study, a prospective, case-control, observational study, was conducted to develop and validate an MCED test. CCGA was divided into three sub-studies, each with different objectives. Sub-study 1 identified DNA methylation as the most promising genomic feature for an MCED test.
Sub-study 2 focused on the development of a targeted methylation assay and training and independent validation of a machine-learning classifier to differentiate cancer/noncancer and predict tissue of origin of the cancer signal.
Because CCGA is a case-control study, enrolled participants had a known newly diagnosed cancer status or noncancer status confirmed in multiyear follow-up. However, it still allowed for comparison of cTAF between participants with and without various types of cancer and different clinical presentations and outcomes. Results from different analyses of CCGA are included below.
cTAF Increases with Aggressiveness and Varies by Orders of Magnitude within Each Cancer Type and Stage
Samples from CCGA were analyzed for cTAF across stages in multiple cancer types. The results showed that median cTAF increased with stage across cancer types. However, cTAF varied by orders of magnitude within a given cancer type and stage (Figure 2A) (unpublished data). Strong differences in cTAF were observed between cancer types. High-mortality cancers, including esophageal, gastric, hepatobiliary, lung, and pancreatic [bottom 10th percentile of 5-year cancer-specific survival in Surveillance, Epidemiology, and End Results (SEER)], were observed to have higher cTAF than low-mortality cancers, including breast, prostate, and thyroid (top 90th percentile of 5-year cancer-specific survival in SEER) within each stage (Figure 2B) (unpublished data). High/low mortality was defined on the basis of histologic information for lung and esophageal cancers so that cancer subtypes that did not meet the mortality criteria were omitted. Thus, developing an understanding of the range of cTAF in different clinical presentations and outcomes is helpful for establishing and characterizing adequately sensitive screening tests with reference to the underlying cancer biology.
cTAF Influences Performance of Blood-Based cfDNA MCED Tests
An objective of the CCGA study was to determine the factors that influence MCED test performance. Any classifier whose features are derived from ctDNA-specific attributes should have better performance if more ctDNA with cancer signal is available to analyze (ie, higher cTAF). Indeed, a positive association was observed between detection and ctDNA level among cancer cases. Notably, once cTAF was included as a predictor in a multivariate analysis, cancer type and stage were no longer significant influences on test performance (unpublished data). Because high-mortality cancers have higher median cTAF than low-mortality cancers within each stage, the MCED test developed in the CCGA study preferentially detected high-mortality cancers at each stage (Figure 2B) (unpublished data). This may provide an opportunity to detect high-mortality or clinically significant cancer earlier, when patients can still have favorable outcomes, while avoiding overdiagnosis of indolent disease. Given these characteristics, cTAF is an attractive metric to compare MCED performance between approaches and studies.
Prognostic Value of cTAF
Participants from the second CCGA sub-study were followed up for up to 3 years to assess OS and explore how a cancer signal from cfDNA analysis was associated with cancer prognosis across cancer types.
Results from this study corroborated findings from unpublished data that higher cTAF was observed in later stages with higher tumor burden, and cancer detection rate increased with cTAF (Figure 3A). Cancers with lower cTAF had better survival compared with cancers with higher cTAF (Figure 3B), which suggests that tumor DNA shedding is a biological factor that is strongly associated with prognosis. Therefore, cTAF could be a useful surrogate biomarker for aggressiveness or clinical significance of cancer across multiple cancer types in addition to clinical stage.
In agreement with this inference, cancers not detected by the MCED test (due to low cTAF) had better survival than cancers that were detected; in addition, these nondetected cancers had better survival than expected based on SEER at each stage (Figure 3, C–F). More important, cancers detected by the MCED test had OS comparable to that expected on the basis of SEER at the same stage (Figure 3, C–F), indicating that these cancers could potentially benefit from early detection and thus have favorable outcomes.
In a multivariate analysis that included covariates of cancer mortality group (high versus low), method of diagnosis (clinical presentation versus screening), clinical stage (III/IV versus I/II), and age, cancer detection by the MCED test remained a significant predictor of OS (P < 0.0001), suggesting its ability to inform prognosis while avoiding overdiagnosis of indolent cancers.
For cancers that have existing screening options (eg, breast, lung, and prostate), the MCED test performance was consistent with that observed for all cancers: increased detection in aggressive cancer subtypes, such as hormone-receptor–negative breast cancer and small-cell lung cancer, as well as prostate cancers with high Gleason scores compared with less aggressive cancers with low Gleason scores (P < 0.0001).
Thus, an MCED test whose performance is influenced by underlying cancer biology could potentially overcome the issue of overdiagnosis by not detecting cancers with indolent phenotypes.
Correlation of cTAF with Clinical Biomarkers for Cancer Aggressiveness
A separate analysis of clinical features of tumor biology in addition to clinical stage (eg, tumor volume and mitotic or metabolic activity and accessibility of tumor DNA to the circulation) was conducted to further understand how cTAF varies and affects cfDNA-based cancer detection.
Clinical correlates of circulating tumor fraction [tumor size, mitotic activity, as reported by %Ki-67–positive (breast cancer), metabolic activity, as reported by positron emission tomography FDG standardized uptake value (lung), and depth of microinvasion (colorectal)], which are clinically established indicators of aggressive tumors, were associated with more frequent detection by the MCED test in the CCGA study.
Consequently, the MCED test was more sensitive for tumors that are associated with higher mortality. The consistent results across lung, breast, and colon cancers may reflect an underlying tumor biology that applies to cancers without current available screening paradigms.
As such, across cancer types and stages, cancers with cTAF above this cLOD will be detected at least half of the time.
The target cLOD for an MCED test that best balances overdiagnosis versus early cancer detection is not known, and may vary between cancer types. If the cLOD of a test is too low, there exists a risk of overdiagnosis or detecting conditions with low cTAF (ie, indolent neoplasms). Conversely, if the cLOD of a test is not low enough, there exists a risk of underdiagnosis or not detecting cancers that would have benefitted from appropriate treatment and management. Defining the optimal level of detection that is relevant to clinicians and patients without overdiagnosis (hypercancer detection) and underdiagnosis (missing cancer detection) will be useful in preferentially detecting clinically significant cancers that require treatment. In the CCGA study, it was observed that cancers not detected by the MCED test had a better prognosis at each stage than the prognosis expected on the basis of SEER, suggesting that such cfDNA-based MCED tests could potentially overcome the issue of overdiagnosis.
It does not mean that all cancers that were detected by the MCED test have high cancer-specific mortality. In fact, on the basis of modeling studies, early detection of cancers using this same MCED test has predicted stage shift,
which can subsequently lead to reduced mortality. Notably, the cancers that were detected followed the mortality rates observed in SEER, whereas the cancers that were not detected had an exceptionally good prognosis.
The distribution of cTAF in the intended-use population for screening is not yet known and will require assessment in future clinical studies. cTAF may be lower in individuals who are being screened for cancer than in individuals who are symptomatic for cancer or have a known diagnosis of cancer (even if that diagnosis is recent), such as participants of the case-control CCGA study. In the first CCGA sub-study, the cLODs of several different approaches that analyzed whole genome methylation, single-nucleotide variants, or somatic copy number alteration were compared, with whole genome methylation having the lowest cLOD (1.2 × 10−3) in comparison to whole genome somatic copy number alterations, fragmentation profiles, and targeted ultradeep sequencing of short somatic variants (unpublished data).
In the second CCGA sub-study, the whole genome methylation assay was refined to a targeted methylation assay targeting only the most cancer-informative regions of the genome. The associated targeted methylation classifier had an even lower cLOD (1.3 × 10−4) than the whole genome methylation classifier. It is expected that cTAF is lower in the intended-use population (screening population) based on differences reported between participants already diagnosed with cancer and participants prediagnosis in the CCGA study (unpublished data). However, the cLOD of the targeted methylation–based MCED test was about an order of magnitude lower than the whole genome methylation approach, which could allow for cancer detection in the intended-use population. The ability to detect cancer before clinical presentation using this targeted methylation-based MCED test has been demonstrated in the interventional PATHFINDER study.
In addition, in the exploratory prospective, interventional study, Detecting Cancers Earlier through Elective Mutation-Based Blood Collection and Testing, 26 preclinical cancers in ≈10,000 women, aged 65 to 75 years, with no personal history of cancer and with high adherence to standard-of-care screening were detected using a cfDNA-based MCED test and protein biomarkers.
Notably, the ability to detect low cTAF cancer cases overcomes issues associated with low ctDNA yield in blood samples for early-stage cancers. Future development of systemized methods for measuring cTAF in different settings for many cancer types will also help standardize comparisons of patient cases and improve biological understanding to reduce overdiagnosis.
In addition, because cTAF has been shown to correlate with clinical biomarkers of aggressiveness for the most common cancers,
it may be possible to use cTAF to both complement and substitute (when unavailable) the diverse, heterogeneous biomarkers in standard use to inform therapy decisions.
Competing models exist that differ in their characterization of which cancers are detectable by ctDNA and the impact of MCED on a screening population. One author suggests that detection by cfDNA-based MCED tests depends on the presence of at least micrometastases (unpublished data). However, although the presence of additional tumor cells in a metastasis contributes more ctDNA fragments, a local or regional primary lesion can also be detected before any metastases appear. In addition, a metastasis that is too small to be manifested on imaging is limited in the amount of ctDNA that it can shed. Mitotic activity, lymphovascular invasion, and the histologic type of lung cancer are well known as significant contributors to tumor fraction.
When modeling cTAF in lung cancers, only the size and metabolic activity of the primary cancer, but not the presence of tumor-involved lymph nodes, were identified as significant correlates in multivariable regression.
as these cannot sufficiently reflect the heterogeneity, varying detectability, and varying clinical significance of neoplasms of the same volume. Such models do not agree with empirical evidence, including prospective interventional studies. Large-scale interventional studies, such as NHS Galleri (N ≈ 140,000; https://www.nhs-galleri.org, last accessed July 13, 2022), are ongoing to better understand the clinical utility of an MCED test and its relationship to cTAF in large screening populations.
More research is needed on the relationship of cTAF and tumor burden (number of tumor cells), tumor growth and metabolism, metastatic potential, and tumor type, as there are likely cancer types where ctDNA levels depend on anatomic site, histologic type, and further tumor characteristics. In addition, much of the current cancer biology presented herein is representative of solid tumors; the role of cTAF in liquid cancers, like circulating lymphomas and leukemias, may differ because of the predominant hematopoietic origin of cfDNA as well as the access and DNA fragment trafficking conditions from lymph nodes and bone marrow to the circulation.
Use of cTAF in Other Settings by cfDNA-Based Tests
cTAF is also the common factor that allows the use of cfDNA tests in the minimal residual disease setting, where detection of putative ctDNA can indicate residual tumor cells after surgery or other interventions, which may lead to cancer recurrence and thus poor outcomes.
Caution should be exercised in extrapolating post-treatment minimal residual disease cTAF prognostication to screening populations; ctDNA positivity in a minimal residual disease setting indicates failure of treatment that is not generally appropriate for a screened population, where the individuals do not have a known prior cancer diagnosis and might subsequently have lower cTAF. Nevertheless, there are likely parallels between MCED tests preferentially detecting more clinically significant cancers with higher cTAF and minimal residual disease surveillance tests, which have increasingly higher sensitivity months after surgery or radiation that is consistent with progressive increase in the amount of residual disease over time.
In this review, we have shown that cTAF is a clinically meaningful biomarker that drives MCED test performance and informs prognosis of cancer. Because detection of cancer by MCED tests is influenced by cTAF, there is the potential for MCED tests to not only detect cancers earlier, but to inform prognosis, reduce overdiagnosis, and inform treatment decisions. Furthermore, cLOD computed against cTAF is an attractive biologically motivated metric to assess and compare MCED test performance.
We thank Mia DeFino, MS, ELS (DeFino Consulting, LLC, Chicago, IL), and Ruhi Ubale, PhD, CMPP (employee of GRAIL, LLC), for providing medical writing support; Kristi Whitfield (PosterDocs, Oakland, CA) and John Beausang, PhD (employee of GRAIL, LLC), for providing support for figure development; and Erin Spohr (ENGAGE Labs, LLC, Oak Ridge, NJ) for providing copyediting, all paid for by GRAIL, LLC, a subsidiary of Illumina, Inc.
de Hingh I.H.
The strengths and limitations of routine staging before treatment with abdominal CT in colorectal cancer.
Supported by GRAIL, LLC, a subsidiary of Illumina , Inc., which paid for medical writing support.
J.B. and O.V. are co–first authors.
Disclosures: All authors are employees of GRAIL, LLC, a subsidiary of Illumina, Inc., and hold equity in Illumina, Inc., at the time of this work. GRAIL, LLC, a subsidiary of Illumina, Inc., is currently held separate from Illumina, Inc., under the terms of the Interim Measures Order of the European Commission, dated October 29, 2021.