About seven years ago, researchers at the US DNA sequencing company Illumina started to notice something odd. A new blood test it ran on 125,000 expectant mothers looking for genetic abnormalities such as Down’s syndrome in their foetuses returned some extremely unexpected signals in 10 cases. Chillingly, it dawned on them that the abnormal DNA they were seeing wasn’t from the foetuses but was, rather, undiagnosed cancer in the mothers. Cancers of different types were later confirmed in all 10. “This was not a test developed for cancer screening,” says Alex Aravanis, then Illumina’s senior R&D director. “But it was evidence that it might be possible.”
In 2016, Illumina created Silicon Valley-based spin-off company Grail, with Aravanis as chief scientific officer. Backed by more than $1.5bn in funding, including money from Microsoft cofounder Bill Gates and Amazon founder Jeff Bezos, Grail is on a quest to detect multiple types of cancer before symptoms, via a single, simple blood test. The test looks at cell-free plasma to find fragments of so-called circulating tumour DNA (ctDNA) sloughed off by cancer cells.
Detecting cancer sooner – before symptoms – means you can intervene earlier and people are less likely to die. While doctors can screen for breast, colon and lung cancer, most varieties of the disease can only be detected after symptoms appear. And though it is far from the only approach (see box), the beauty of blood is that it is minimally invasive to collect. “A relatively simple blood-based test that can screen for evidence of cancer… might improve or even replace some screening programmes over time,” says Jacqui Shaw, professor of translational cancer genetics at the University of Leicester, who studies ctDNA.
Looking for ctDNA has become a viable proposition in recent years because of improvements in DNA sequencing technologies that make it possible to scan fragments and find those few with alterations that may indicate cancer. While other blood-based biomarkers are being investigated, the advantage of ctDNA is that, because it has a direct link to the tumour, it can be very specific at identifying cancer. For that reason, ctDNA is also showing promise as a way to profile and monitor advanced stage cancers, a “liquid biopsy”.
Early detection is a harder problem. Early on, when the tumour is small, there is not as much ctDNA to detect. The women Illumina identified as having cancer were all late, not early stage.
To date, there is one company offering a blood test based on ctDNA for early cancer detection: Epigenomics began offering its test for colon cancer in 2016 based on detecting biochemical modification of a single gene.
But the dream being imagined by Grail and others is an inexpensive test, perhaps no more than $500, which could conceivably be given annually to those over a certain age, with a high chance of detecting many cancers at once with high accuracy (Grail hasn’t announced a final number but thinks it will be in the region of 10). It’s a test that all of us, if it works, could one day get. “The big studies are still to be done,” says Nitzan Rosenfeld, a researcher who studies ctDNA at the Cancer Research UK Cambridge Institute and a cofounder of the UK-based liquid biopsy company Inivata, “but there has been considerable progress.”
Cancer essentially begins when a normal cell’s DNA gets mutated or altered. From that point, the cell multiplies too often and a mass or tumour of abnormal cells forms. A proportion of the cells invariably die and “shed” genetic material into the bloodstream, mixing with larger amounts of DNA fragments coming from the death of normal cells.
It was first reported that fragments of DNA carrying cancer-causing mutations could be found floating freely in the blood of cancer patients in the mid-1990s. The findings caught the attention of Dennis Lo, now a professor of medicine and chemical pathology at the Chinese University of Hong Kong, who thought that “a baby living in a mother is a little bit like a cancer growing in a patient”. Based on that insight, he went on to discover foetal DNA fragments in maternal blood and to pioneer non-invasive prenatal testing (NIPT) (he licensed his patents to Illumina and other companies). He also began, along with others, to apply those insights to how ctDNA fragments might be used in the monitoring and detection of cancer. In 2017, Grail merged with Lo’s company, Cirina, also aimed at early detection.
Grail’s competitors include Guardant Health, a liquid biopsy company valued at $3.5bn that was founded in 2011 and is also based in Silicon Valley. It recently branched out to work on an early-detection test for four common cancers: lung, breast, colorectal and ovarian.
There are also multiple academic efforts with designs on commercialisation. Last year, researchers at Johns Hopkins University school of medicine published details of a potential blood test called CancerSEEK that covers eight cancers. And in December last year, University of Queensland researchers made headlines with a “10-minute test” they have called a methylscape and which they say could potentially give a yes or no answer to the presence of cancer in the body, though it wouldn’t identify its location.
Each group or company has or is developing a way of detecting ctDNA. Grail and Guardant’s tests are based on sequencing the cell-free DNA. This can look for mutations, increases in the number of chromosomes or genes or unusual biochemical changes known as epigenetic changes, all of which can occur in the DNA of cancer cells. Grail has been experimenting with all three but hasn’t announced which method its final test will use. Guardant’s method takes them all in. CancerSEEK, meanwhile, looks for a small number of mutations as well as protein markers known to increase in particular cancers. The small mutation panel keeps costs down says Nickolas Papadopoulos, a professor of oncology who is co-leading work on the latter method. Instead of sequencing, methylscapes use gold nanoparticles to detect epigenetic alterations.
Results have been published or presented for all these methods, demonstrating that cancer-related signals can be seen. They are based on small studies of 1,000 people or fewer with cancer at various stages. The key for the tests is achieving both a high likelihood of detection (a good sensitivity) and a low false positive rate to avoid needless anxiety and unnecessary follow-up. (In a test being designed for a general population, where cancer actually isn’t very prevalent, the latter is particularly important because even low false positive rates will result in a substantial number of incorrect diagnoses.)
Grail’s best detection rates, based on a prototype test for detecting epigenetic changes, ranged from 80% down to 47% for nine cancers (respectively ovarian, liver, lymphoma, multiple myeloma, pancreatic, colorectal, oesophageal, head and neck and lung). Breast cancer ranged from 56% to 11% depending on the type. The false positive rate was set at 2%, though says Aravanis, further work suggests it could hit less than 1%. “From a single test, we detected a large fraction of the highest mortality cancers in early stage with very high specificity,” he says.
Of Grail’s competitors, CancerSEEK’s sensitivity ranged from 98% for ovarian cancer to 33% for breast cancer with a false positive rate of less than 1%. Guardant showed it could detect lung cancer in 71% of cases and colorectal cancer in 67% of cases with a false positive rate of 2%. Methylscape meanwhile had a sensitivity of 90% but the false positive rate was higher at 10 to 15%.
Yet while the companies are bullish about what they are seeing, not everyone is convinced. Eleftherios Diamandis, professor of clinical biochemistry at the University of Toronto, has made a name for himself criticising grand designs for revolutionary blood tests. Well before the media questioned the validity of startup company Theranos’s blood-testing technology – the firm and its chief executive were charged last year with “massive fraud” by the US Securities and Exchange Commission – Diamandis had raised doubts in the scientific literature.
Last year, he and his associate Clare Fiala published a series of journal pieces questioning how useful ctDNA could really be in early cancer diagnosis. Diamandis’s calculations, based on experimental literature data, give a sense of the size of tumour a ctDNA-based blood test might be able to pick up. His work indicates that with the technology available for analysing ctDNA, tumours would need to be approximately 1cm in diameter or greater to be detected. “That is a fairly large tumour,” he says. There is simply unlikely to be a single fragment of DNA from tumours under that size in the 10ml of blood that is a standard sample. While more blood may help a little, it would make things much less pleasant.
As Diamandis sees things, the tests do seem to be performing well, but that is because they are being applied to people who have already been diagnosed with cancer. “If you go to a real population of asymptomatic individuals, their success rate will be, predictably in my opinion, much less,” he says. “For the [multi-cancer] blood test – as of today – I am relatively pessimistic.”
For their part, the groups and companies acknowledge that large studies in people without diagnosed cancer are what is needed next. CancerSEEK is looking to enrol 10,000 women without known cancer to take its blood test; it will then follow them over five years to see whether they develop cancer. And Grail has two large, long-running studies of those who are asymptomatic, one of 100,000 and another of 50,000, in the pipeline. “They are large, clinical validity studies that complement the evidence that we are generating,” says Aravanis.
They also point out that they are finding a large fraction of those solid tumours that are the most clinically aggressive and therefore have potentially the highest mortality. “Even if you find 10%, that is 10% more than zero,” notes Papadopoulos.
Aravanis questions whether finding lots of very early-stage, slow-growing, potentially inconsequential cancers may actually be that helpful. “Whether or not you can detect every single cancer at the earliest stages or pre-cancer is an interesting scientific question, but I don’t know that that is really the important clinical question,” says Aravanis. “[Overdiagnosis] is something that some existing screening programmes struggle with.”
Perhaps the answer, say others such as Rosenfeld and Shaw, is to go for a more limited approach – tackling high-risk populations or focusing on those cancers where the diagnosis is often late. “[Then] I think we might find places where we can improve the current clinical situation,” says Rosenfeld. And Diamandis doesn’t rule out success in fluids other than blood, such as urine and cervical fluid for bladder and ovarian cancer respectively, where ctDNA concentrations could be far higher because they are in contact with, or proximity to, the tumour.
The million-dollar question is, when can we expect a once-a-year blood test that checks for multiple cancers? The typical answer from experts is that it will be at least five years before there is sufficient data to show whether it can work and then it will need to hit the desks of regulators and health economists. Patience, it seems, is a virtue in the fight against cancer.
Breathomics and cancer detection
While blood is relatively easy to collect to look for a cancer signal, what about if you could detect it on breath – no needle required? UK-based company Owlstone Medical is among those looking to discover whether this is possible. It has developed a breathalyser based on the idea that there may be chemicals – volatile organic compounds – in breath indicative of early-stage cancer.
It is currently running a trial in partnership with the NHS to see whether there are differences in the chemicals that can be detected in breath between people with lung cancer and those without it. Lung cancer, says Max Allsworth, the company’s chief scientific officer, is a good place to start because the air we breathe directly moves through the lung and past any tumours.
The company has also recently started a separate, smaller study looking at whether six other cancer types may also be detected early in this way (in this case, the chemicals would find their way into the breath less directly, via blood, which exchanges volatile chemicals with air in the lungs). Depending on the trial results, it may end up being a more generic cancer detection test – not telling you specifically where the cancer is, but that a common signal has been found. In all cases, says Allsworth, if there are chemicals in breath that suggest cancer, they will be present very early on, before you are likely to have circulating tumour DNA (ctDNA).