About 2 years ago, Grail was formed as a spin-off company of the sequencing giant Illumina, with the goal of developing a blood test for early cancer detection . With close to $1 billion financing, Grail tops the list of the highest grossing deals for private biotech companies in 2017 . The firm’s testing relies on analyzing circulating tumor DNA (ctDNA); minuscule amounts of fragmented DNA originating from tumor cells . As the release of DNA in the circulation is not specific to cancer cells, there is also a larger proportion of circulating DNA (termed circulating free DNA, cfDNA) from normal cells .
As we explained in previous publications , , the ratio of tumor DNA to normal DNA (generally expressed as a percentage) is an important parameter for this test. For example, a 0.1% mutant allele fraction means that one out of every 1000 DNA molecules in the circulation, originates from the tumor and the other 999 DNA molecules originate from healthy cells.
Grail’s technology is based on the analysis of paired cfDNA and white blood cell (WBC) DNA at high depth (e.g. 60,000×). This involves the targeted sequencing of 507 genes previously associated with cancer, to identify single nucleotide variants or insertions/deletions (indels). An alternative analytical technology uses paired analyses of cfDNA and WBC DNA by whole genome sequencing (30× depth) to identify copy number variations. A third version of the assay employs cfDNA whole genome bisulfite sequencing at 30× depth to identify methylation differences. As Grail found that all three tests provide approximately similar results , , we will not comment on the clinical performance of each test separately.
To create their database of cancerous mutations, Grail initiated the Circulating Cancer Genome Atlas (CCGA) study, which aims to enroll 15,000 participants consisting of 70% cancer patients and 30% healthy controls . While this is a work in progress, preliminary data from two studies presented at the 2018 American Society of Clinical Oncology (ASCO) annual meeting (June 1–5, Chicago, IL, USA) will be commented upon here.
Grail is not the only group interested in early cancer diagnosis using ctDNA. For example, two recent papers in leading journals have addressed the same issue , . Phallen et al. evaluated 200 patients with colorectal, breast, lung and ovarian cancer and reported sensitivities between 60% and 70% in patients with stage I–II disease . Cohen et al. combined ctDNA screening with traditional serum biomarkers and reported sensitivities between 70% and 98% for the detection of ovarian, liver, stomach, pancreatic and esophageal cancers. Their investigation claimed specificity greater than 99% .
We have previously correlated ctDNA abundance with cancer volume based on published empirical data. The minimum percent mutant allele fraction of circulating DNA needs to be around 0.01% in order for any method to detect ctDNA fragments with a 10 mL blood draw , . Below this fraction, there will be less than one tumor DNA genome equivalent in a 10 mL blood draw. Consequently, diagnosis will be impossible due to sampling error (no retrieval of tumor DNA), regardless of the sensitivity or reliability of the analytical assay. A summary of our calculations regarding tumor diameter, tumor weight, tumor volume, percent fraction of circulating tumor DNA, number of genomes per 10 mL of blood and likelihood of detection of ctDNA are shown in Figure 1. When the tumor diameter is less than 6 mm, diagnosis with this principle becomes impossible due to absence of ctDNA in the collected sample. For more details see Refs.  and . Currently, imaging technologies can readily detect 6 mm diameter tumors .
Grail’s clinical sensitivity
For their study  Grail prospectively collected 1627 samples from 749 controls (no cancer) and 878 patients with newly diagnosed and untreated cancer (20 tumor types of all stages). The overall sensitivity of their blood test ranged between 50% and 90% (stages I–III) but for some cancers (low Gleason grade prostate, thyroid, uterine, melanoma and renal) the assay had less than 10% sensitivity. All these sensitivities were calculated at a fixed specificity of 95%. Grail concluded that the ctDNA-based blood test detected multiple cancers at various stages with good sensitivity and high specificity, showing promise as a multi-cancer screening test.
In their separate breast cancer study , Grail included 358 patients with invasive breast cancer (mostly stage I–II) and 452 controls. They also reported their sensitivities at 95% specificity. For symptomatically diagnosed breast cancer patients the average sensitivity was 58%, 40% and 15% in the three breast cancer subtypes, respectively (triple negative, HER2-positive/hormone receptor-positive, HER2-negative). However, when patients were classified according to the mode of diagnosis (symptomatic vs. screen-detected/no symptoms) the sensitivities were 44% for symptomatic patients and only 10% for screened-detected breast cancers.
The importance of specificity in population screening
We outlined earlier that the specificity of the ctDNA test could theoretically be 100% . However, there are many recent examples whereby mutations in circulating DNA from apparently normal individuals are found relatively frequently , , , , , .
In the two Grail studies, the specificity was fixed at 95% , . However, in cancer screening programs, disease prevalence needs to also be carefully considered. The test’s predictive value, positive or negative, indicates the possibility of somebody having or not having the disease if the test is positive or negative, respectively. For example, a positive predictive value of 80% indicates the chance of somebody from the screened population having cancer, if the test is positive. Let us consider a hypothetical example whereby the disease prevalence is one patient in every 4000 individuals in the screened population (such is the case with ovarian, pancreatic and other serious, but relatively rare cancers). A test with 99% specificity, if positive, even with 100% sensitivity, will lead to a positive predictive value of only 2%. Even if the test has 99.9% specificity and 100% sensitivity, the positive predictive value will only be 20% (Figure 2). These examples emphasize that when a test is used for cancer screening (such as testing asymptomatic individuals), high specificities (e.g. >99%) are necessary in order for the positive predictive value to be clinically useful. In the data reported at ASCO by Grail, the finding of 10% sensitivity for detecting breast cancer in asymptomatic individuals, at 95% specificity is very problematic.
In studies investigating ctDNA for early diagnosis, including the two aforementioned high-profile recent papers ,  and the Grail investigations, it is very common to study patients who have already been clinically diagnosed. In such cases, the results may look more favorable than they actually are, because tumor size, biology and amount of ctDNA (% mutant allele fraction) do not usually represent accurately asymptomatic patients.
The true clinical validity of a test for early cancer diagnosis needs to be established with asymptomatic individuals, in a prospective fashion, against a predicate method, to determine the true sensitivity, specificity and positive and negative predictive value of the test. Diagnostic effectiveness aside, it will also be necessary to address other important issues related to cancer screening, including over-diagnosis and over-treatment .
Based on our previous analyses ,  and the latest results from Grail , , we conclude that ctDNA as an early diagnostic and screening cancer biomarker test is still far from clinical implementation. At present, the necessary sensitivity is problematic due to the miniscule amounts, or even absence, of ctDNA from the circulation in early, asymptomatic stages. It seems that new, non-invasive methods for extraction of ctDNA from larger volumes of blood, or even from the whole circulation, may be necessary to address the sensitivity issue. The issue of specificity will also remain under close consideration, as we better understand the nature of mutations and their functional significance in apparently normal cells . Finally, despite the difficulties with early detection, it is important to mention that ctDNA testing has important and immediate applications in other areas of cancer management such as prognosis, monitoring and prediction of therapy , .
Author contributions: All the authors have accepted responsibility for the entire content of this submitted manuscript and approved submission.
Research funding: None declared.
Employment or leadership: Dr. Eleftherios P. Diamandis discloses that he holds a consultant/advisory role with Abbott Diagnostics. Miss Clare Fiala has nothing to disclose.
Honorarium: None declared.
1. Aravanis AM, Lee M, Klausner RD. Next-generation sequencing of circulating tumor DNA for early cancer detection. Cell 2017;168:571–4. Search in Google Scholar
2. Huggett B. Bringing up baby. Nat Biotechnol 2018;36:393–401. Search in Google Scholar
3. Leung F, Kulasingam V, Diamandis EP, Hoon DS, Kinzler K, Pantel K, et al. Circulating tumor DNA as a cancer biomarker: fact or fiction? Clin Chem 2016;62:1054–60. Search in Google Scholar
4. Fiala C, Kulasingam V, Diamandis EP. Circulating tumor DNA for early cancer detection. J Appl Lab Med 2018;3:300–13. Search in Google Scholar
5. Diamandis EP, Fiala C. Can circulating tumor DNA be used for direct and early stage cancer detection? F1000Res 2017;6:2129 Search in Google Scholar
6. Klein EA, Hubbel E, Maddala T, Aravanis A, Beausang JF, Filippova D, et al. Development of a comprehensive cell-free DNA (cfDNA) assay for early detection of multiple tumor types: The Circulating Cell-free Genome Atlas (CCGA) study. J Clin Oncol 2018;36(Suppl):abstr 12021. Search in Google Scholar
7. Liu MC, Maddala T, Aravanis A, Hubbell E, Beausang JF, Filippova D, et al. Breast cancer cell-free DNA (cfDNA) profiles reflect underlying tumor biology: The Circulating Cell-Free Genome Atlas (CCGA) study. J Clin Oncol 2018;36(Suppl):abstr 536. Search in Google Scholar
8. US National Library of Medicine. The Circulating Cell-free Genome Atlas Study (CCGA). https://www.clinicaltrials.gov/ct2/show/NCT02889978. Accessed: June 2018. Search in Google Scholar
9. Phallen J, Sausen M, Adleff V, Leal A, Hruban C, White J, et al. Direct detection of early-stage cancer using circulating tumor DNA. Sci Transl Med 2017;9:eaan2415. Search in Google Scholar
10. Cohen JD, Li L, Wang Y, Thoburn C, Afsari B, Danilova L, et al. Detection and localization of surgically resectable cancers with a multi-analyte blood test. Science 2018;359:926–30. Search in Google Scholar
11. Erdi YE. Limits of tumor detectability in nuclear medicine and PET. Mol Imaging Radionucl Ther 2012;21:23–8. Search in Google Scholar
12. Diamandis EP. Present and future of cancer biomarkers. Clin Chem Lab Med 2014;52:791–4. Search in Google Scholar
13. Genovese G, Kahler AK, Handsaker RE, Lindberg J, Rose SA, Bakhoum SF, et al. Clonal hematopoiesis and blood-cancer risk inferred from blood DNA sequence. N Engl J Med 2014;371:2477–87. Search in Google Scholar
14. Alexandrov L, Jones PH, Wedge DC, Sale JE, Campbell PJ, Nik-Zainal S, et al. Clocklike mutational processes in human somatic cells. Nat Genet 2015;47:1402–7. Search in Google Scholar
15. Schwaderle MC, Husain W, Fanta PT, Piccioni DE, Kesari S, Schwab RB, et al. Detection rate of actionable mutations in diverse cancers using a biopsy-free (blood) circulating tumor DNA assay. Oncotarget 2015;33:11004. Search in Google Scholar
16. Gormally E, Vineis P, Matullo G, Veglia F, Caboux E, Le Roux E, et al. TP53 and KRAS2 mutations in plasma DNA of healthy subjects and subsequent cancer occurrence: a prospective study. Cancer Res 2006;66:6871–6. Search in Google Scholar
17. Fernandez-Cuesta L, Perdomo S, Avogbe PH, Leblay N, Delhomme TM, Gaborieau V, et al. Identification of circulating tumor DNA for the early detection of small-cell lung cancer. eBioMedicine 2016;10:6–12. Search in Google Scholar
18. Newman AM, Lovejoy AF, Klass DM, Kurtz DM, Chabon JJ, Scherer F, et al. Integrated digital error suppression for improved detection of circulating tumor DNA. Nat Biotechnol 2016;34:547–55. Search in Google Scholar
19. Li M, Diamandis EP. Technology-driven diagnostics: from smart doctor to smartphone. Crit Rev Clin Lab Sci 2016;53:268–76. Search in Google Scholar
20. McConnell MJ, Moran JV, Abyzov A, Akbarian S, Bae T, Cortes-Ciriano I, et al. Intersection of diverse neuronal genomes and neuropsychiatric disease: the Brain Somatic Mosaicism Network. Science 2017;356:eaal1641. Search in Google Scholar
21. Wan JC, Massie C, Garcia-Corbacho J, Mouliere F, Brenton JD, Caldas C, et al. Liquid biopsies come of age: towards implementation of circulating tumour DNA. Nat Rev Cancer 2017;17:223–38. Search in Google Scholar
©2019 Walter de Gruyter GmbH, Berlin/Boston