Being able to predict accurately whether a patient’s cancer would come back after treatment would help doctors stay one step ahead of the disease.

Those predictions could indicate what treatment a patient should be offered, or perhaps suggest when it’s time to switch treatment earlier than is possible today.

While these kinds of predictions aren’t yet possible, a new study by our scientists shows that we’re heading in this direction, at least for lung cancer.

Led by Professor Charles Swanton at the Francis Crick Institute in London, the team is analysing the DNA released by tumour cells into the blood of lung cancer patients over the course of their treatment. And their latest findings, published in the journal Nature, reveal in incredible detail how these cancers change as they develop, all thanks to a simple sample of blood.

Their data suggest that in certain patients, blood tests could help doctors detect early signs that the disease will return after treatment, months in advance of a tumour showing up on scans. For some patients on the study, these predictions could have been made a year in advance.

“This means that for the first time in lung cancer, we could open up clinical trials where patients with disease, that isn’t yet visible through medical imaging, can be treated with new drugs,” says Swanton.

“We could also use tumour DNA in the blood to monitor how well the treatments are working.”

For a cancer that’s hard to treat, this could make a big difference for patients. But these results are still just an early glimpse from this ongoing study, so there’s more to be done before blood tests like this could be tested in clinical trials.

Fishing for clues

This latest research forms part of Swanton’s pioneering study, TRACERx. The first of its kind, the study has set out to follow almost 850 patients with the most common type of lung cancer, non-small cell lung cancer, from diagnosis through treatment and beyond. So far, they’ve analysed the first 100 patients, the results of which were also published today.

“TRACERx is designed to look at how lung cancers evolve over time and, in particular, work out how genetic faults in these cancers are involved in causing the disease to return following surgery,” says Dr Chris Abbosh, one of the lead researchers on the latest study from University College London.

“Through in-depth analysis of a large number of tumours we hope to develop ‘evolutionary rulebooks’ that dictate how lung cancers develop and progress.” This, says Abbosh, will lead to a new way to understand lung cancer biology.

Previous studies that set out to understand lung cancer evolution offer a narrow view of this process because they usually analyse single tissue samples from one part of the tumour, for example a single diagnostic biopsy sample.

We hope to develop ‘evolutionary rulebooks’ that dictate how lung cancers develop and progress

– Dr Chris Abbosh

Now scientists have revealed how diverse and complex a tumour can be, new methods have been developed that attempt to capture this complexity in full.

That’s where blood samples come in.

Tumours contain a mixture of dividing and dying cells. The balance is more heavily weighted to the dividing cells, which is what results in a tumour. But as some cancer cells die they release a trace of the tumour into the blood in the form of fragments of DNA. These fragments exist in miniscule amounts in the bloodstream, and are easily drowned out by the normal contents of our blood. But extremely sensitive lab techniques have allowed scientists to fish them out and analyse them.

Because these DNA fragments have come from cells right across the tumour, scientists believe they could offer more information than single biopsies can.

In TRACERx, the team will be using this circulating tumour DNA (ctDNA) in 3 different ways.

“First, we’re looking at blood samples taken before the patients had surgery to remove their tumours,” explains Abbosh. “We want to see what features of the tumour are associated with the release of ctDNA into the circulation.

“Next, we want to see which parts of the tumour’s genetic make-up, or ‘evolutionary tree’, can be detected in the blood.

“Finally, we want to find out what this DNA could tell us after the patients have had surgery.”

Spot the difference

There are several different types of non-small cell lung cancer, but the most common types are squamous cell and adenocarcinoma. Comparing the two, Abbosh and the team found that almost all patients (94%) with squamous cell type had tumour DNA in their blood. This was true for only 1 in 7 patients (13%) with adenocarcinomas.

Looking for clues behind these differences, the researchers found that blood samples from patients with lots of dying cells in their tumours, described as ‘necrotic’, were more likely to carry tumour DNA. In line with this, the squamous cell tumours were found to be much more necrotic than the adenocarcinomas. But in patients with adenocarcinomas that did carry more dying cells, the team picked up tumour DNA in their blood.

“Many studies have suggested that tumour DNA might be released when cancer cells are breaking down and becoming necrotic,” says Abbosh. “Squamous cell tumours are known to be a necrotic type of non-small cell lung cancer, so this might explain why we could detect tumour DNA earlier in these patients.

“This knowledge could also help us in the future in terms of identifying groups of patients in whom you would expect to detect ctDNA.”

Another characteristic that appeared alongside tumour DNA in the blood was a marker of the level of cell growth in the tumour, revealed through PET scans – a type of scan patients often have before lung cancer surgery.

Patients whose tumours contained lots of growing cells were more likely to have tumour DNA in their blood. And this could give the team important information about how tumours evolve.

Swapping crystal balls for science

As a cancer grows and evolves, new genetic faults emerge, leaving distinct patches of cells across the tumour that all differ in their genetic makeup. Think of these patches as ‘branches’ growing from the tumour’s original ‘trunk’.

Cells from these branches can break away from the tumour and spread around the body, acting as a seed for the growth of new tumours. In some TRACERx patients, whose disease ultimately returned and spread, the team found tumour DNA from these ‘branch’ cells in blood samples that were taken after surgery. And when the patients’ cancers came back, the researchers found that the DNA in these returning tumours matched what they had found in those earlier blood samples.

“In other words, we can detect an individual branch in the blood that forecasts the disease returning, before being able to biopsy it,” says Swanton.

But the predictive powers of tumour DNA in the blood don’t appear to end there.

In almost all patients whose disease returned, the team could detect rising tumour DNA levels in the blood ahead of the tumour becoming detectable by doctors.

We could now have a personal biomarker of treatment response in each patient, which is what the field has been waiting for

– Professor Charles Swanton

This suggests that the blood test could indicate if a cancer will come back well before it could be picked up by doctors. In fact, the team could identify these early warning signs, or ‘biomarkers’, up to almost a whole year in advance of the returned disease showing up in scans.

It’s important to say that these results are based on collecting the samples, monitoring the patients, and then looking back to see what information the blood tests can reveal. This remains an experimental approach, but it could have significant implications.

“This means we could now have a personal biomarker of treatment response in each patient, which is what the field has been waiting for,” says Swanton.

“We could use these biomarkers to map which part of the tumour is responsible for the disease returning and, if confirmed, target these with drugs or immunotherapy,” adds Abbosh.

“We could personalise treatment to the relapse process, all before the tumour becomes apparent.”

These are exciting future prospects, and Swanton is already working with others on immunotherapies that could target what’s unearthed in these blood samples.

So although we can’t yet see the future, it seems that, for lung cancer at least, we might be on our way to predicting it.



Abbosh, C. et al. (2017). Phylogenetic ctDNA analysis depicts early stage lung cancer evolution. Nature. DOI: 10.1038/nature22364