This article is more than 5 years old. Information may no longer be current.
Prototype assays may facilitate development of blood tests for early cancer detection
Click Here to Manage Email Alerts
CHICAGO — Prototype sequencing assays evaluated in the Circulating Cell-free Genome Atlas study may contribute to the development of a highly specific blood test for early cancer detection, according to training set findings presented at American Association for Cancer Research Annual Meeting.
The early results suggest it may be feasible to design a comprehensive blood-based test that can detect cancer among asymptomatic individuals, perhaps with a specificity greater than 99%, said Alexander Aravanis, MD, PhD, vice president of research and development at life sciences company GRAIL Inc.
“This is really the starting point, and there is an opportunity to get deeper sequencing of those regions that we discover that are most informative.” Aravanis said during a presentation. “We also observed improvement in classification based on our machine learning approach with increasing data sets.”
The prospective, longitudinal, multicenter, observational Circulating Cell-free Genome Atlas (CCGA) study — the largest study of plasma cell-free DNA-based early cancer detection — has a target enrollment of 15,000 individuals. So far, more than 11,000 people have been enrolled from 141 sites in 24 states. Enrollment is expected to conclude this year.
Investigators collected blood from participants with newly diagnosed, treatment-naive cancer and controls at the time of enrollment. Tumor tissue, when available, also is collected and sequenced.
Longitudinal follow-up will continue for up to 5 years. Participants with cancer will be followed for data on treatment, recurrence and outcomes. Controls will be followed for data on cancer events, treatment and outcomes.
Aravanis presented data from the first training set. The analysis included analyzable assay data from 1,399 individuals: 841 with cancer and 558 noncancer controls. The two groups were balanced with regard to race, age, sex and BMI; however, individuals with lung cancer tended to be older and were more likely to be ever-smokers.
The cancer cases represented 20 tumor types — the most common of which were breast, lung, prostate and colorectal cancers — and encompassed all clinical stages. Cancer diagnoses resulted form a mix of screen-detected malignancies and clinical presentation, Aravanis said.
“We designed the CCGA study to get a very representative cohort of individuals in the United States,” Aravanis said. “Additionally, we’re very careful not to exclude other pathologic conditions or diseases that we would ultimately encounter in a real-world setting.”
Investigators isolated cell-free DNA and — with a single blood draw — analyzed samples using three sequencing methods:
Targeted sequencing measured nonsynonymous single nucleotide variants and indels in particular stretches of the genome;
Whole-genome sequencing measured changes in copy number across the genome; and
Whole-genome bisulfite sequencing measured aberrant methylation patterns.
The majority of cell-free DNA variants were white blood cell-matched likely derived from clonal hematopoiesis (noncancer controls, 78%; cancer cases, 66%).
White blood cell-matched variants increased with age, and several were previously unreported noncanonical loss-of-function mutations.
Less than 1% of participants who enrolled with no clinical diagnosis of cancer exhibited a consistent “cancer-like” signal, according to researchers. This finding suggests the potential to develop a test with a specificity greater than 99%.
In addition, a positive result among those who had no clinical diagnosis of cancer could suggest undiagnosed cancer, researchers wrote. Two people in the noncancer group who exhibited cancer signals subsequently were diagnosed with cancer — one with stage III ovarian cancer and one with stage II endometrial cancer.
After white blood cell variant removal, canonical driver somatic variants appeared highly specific to cancer. For example, none of the control participants had variants in EGFR or PIK3CA, whereas 11 participants with cancer exhibited variants in EGFR and 30 exhibited variants in PIK3CA. Also, of the eight control participants who had somatic copy number alterations detected with whole-genome sequencing, four were derived from white blood cells.
The strength of cancer-like signals detected by all three assays increased with disease stage.
Mean levels of nonsynonymous single nucleotide variants and indels per megabase were 1.01 (standard deviation [SD], ± 0.86) for controls, 2.43 (SD, ± 3.98) for those with stage I to stage III cancer, and 6.45 (SD, ± 6.79) for those with stage IV cancer.
Whole-genome sequencing scores were 0 ± 0.08 for those with no cancer, 0.27 ± 0.98 for those with stage I to stage III cancer, and 1.95 ± 2.33 for those with stage IV cancer.
Methylation scores were 0 ± 0.5 for controls, 1.02 ± 1.77 for those with stage I to stage III cancer and 3.94 ± 1.7 for those with stage IV cancer.
Researchers also observed high biological signal in typically unscreened cancers. This result emerged from a subset analysis of 196 cases of cancer associated with 5-year cancer-specific mortality greater than 50%.
This subgroup included 118 lung cancer cases, 25 pancreatic cancer cases, 24 esophageal cancer cases, 16 ovarian cancer cases and 13 hepatobiliary cancer cases.
Aravanis reported 65% detection of stage I through stage III cancers (n = 117) with whole-genome bisulfite sequencing, and 95% detection of stage IV cancers (n = 79).
Researchers plan to present additional data at a future meeting that will break down these results by tumor type.
“The next steps are to analyze and report more from the test set, and then move into validation studies using the larger Circulating Cell-free Genome Atlas cohort with more early-stage participants,” Aravanis said. “Together, these findings lay the groundwork for the development of blood based tests for early cancer detection.” – by Mark Leiser
Reference:
Aravanis AA, et al. Abstract LB-343. Presented at: American Association for Cancer Research Annual Meeting; April 14-18, 2018; Chicago.
Disclosures: GRAIL sponsored the study. Aravanis reports employment with and shareholder interests in GRAIL. Please see the abstract for all other authors’ relevant financial disclosures.
Perspective
Back to Top
Andrew Wang, MD
We always want to have better cancer screening tests. Because we know cancer is a disease of mutations, the hope has always been that genomic technologies or sequencing could offer value in this area. From that standpoint, this is exciting. This is one of the larger studies to detail this effort.
However, it is the first step of many, many steps that need to be taken.
One caveat is, this doesn’t address all of the pitfalls of cancer screening in general.
When we think about tests, we tend to use terms like sensitivity or specificity. However, in the clinical realm — especially for cancer screening — those are not the terms we use. The clinically meaningful terminology is negative and positive predictive values. It’s a math problem because of how you calculate negative or positive predictive values. Specificity and sensitivity are factors in that equation, but one of the most important factors is the prevalence or the incidence in the population. Even when a test with very high sensitivity and specificity is used on a not-so-common cancer, the negative and positive predictive values become very low, and this becomes problematic for screening tests.
In terms of cancers we already screen for, this will be valuable and it will compete with existing technologies, and then it will be a matter of who wins. That’s helpful, especially if it becomes cheaper. But if you think about the lack of good screening tests for a lot of other cancers — such as ovarian cancer or pancreatic cancer — I don’t know if this is going to overcome that problem. That math problem is still the same math problem, and that’s the same thing that has dogged every other test. As prevalent as breast cancer and prostate cancer are, we still have controversial stances about breast imaging or PSA testing.
There definitely have been success stories in certain enriched populations. Think about how we do chest CT for lung cancer in an enriched population of high-risk individuals. Instead of CT, maybe we’d have a blood test. So, I can certainly see potential impact and value, but for people who are looking for a screening test for uncommon cancers, the data we have seen don’t tell us if this is going to work.
The liquid biopsy arena is an exciting space, but the bigger impact of these blood tests is on surveillance and monitoring. Down the road, I envision we are going to have a complementary set of tests rather than one. We are going to use them to look at different things and gain a lot more information. Then maybe we will be able to rely less on technologies such as imaging that are more expensive, more labor intensive and more taxing on patients.
Click Here to Manage Email Alerts