31 March 2020
A new paper provides an intriguing update on recent efforts to develop blood tests to detect cancer early. These non-invasive tests analyse circulating tumour (ct) DNA – small fragments of tumour DNA found in the circulation – with the goal of finding out whether or not a patient has cancer. Whilst reliably detecting early stage cancers is still a challenge, the research has made progress in developing a test that can help to identify cancer tissue of origin, that is, where a cancer is growing in the body – which can aid diagnosis and treatment planning.
The international consortium of researchers from the UK and the US, comprising input from the Circulating Cell-Free Genome Atlas Study, the STRIVE Study and GRAIL, Inc., used a genome sequencing method called bisulphite sequencing to analyse chemical changes to DNA that do not change the underlying sequence. The pattern of these epigenomic changes – called methylation marks – differs between tissues in the body, and also between healthy and disease states. In principle, analysing these marks should provide information not only on whether a patient has a disease, but in the case of cancer, where the disease is in the body.
The team analysed methylation patterns across the genome found in the cell free DNA (DNA released naturally into the circulation by healthy cells in the body) and ctDNA from 1,531 patients with cancer and 1,521 patients without cancer. They then trained a machine learning model to identify which patterns reflected different cancer types, and then tested their model using samples from an additional group of patients with (654) and without (610) cancer.
They found that across more than 50 cancer types seen in study participants they could detect around 18% of the earliest and least well developed stage I cancers, rising to 93% of the most advanced stage IV cancers. Of those cancers they detected, the test could determine the type of cancer (tissue of origin) in just over 9 in 10 of these cancers. The general pattern was that the test was able to detect more cancers the more advanced they were, however there was a great deal of variation between different cancer types. Across the whole study, fewer than 1% of participants were inaccurately classified as having cancer when they did not.
These results come from a much larger programme of work planned to investigate the use of these kinds of tests in clinical trials. As we have written before, whilst detecting cancer at earlier stages I or II (when treatment is more likely to be successful) is an admirable aim, it is not without its challenges. This study, like many others before it, detected a low proportion of cancers at these stages. There are a number of potential reasons for this, including:
Concerns over using very sensitive technologies for early cancer detection include how we know whether the result indicates a genuine cancer or a pre-malignant condition (i.e., one that without treatment might, or might not, progress to cancer), and what the next clinical steps should be. While tissue of origin information can help to minimise further investigative testing, it is still likely to be needed. With concerns around over-diagnosis and over-treatment already an issue of concern with current screening programmes, particularly for breast cancer, health systems will have to consider the relative risks and benefits of using these types of tests for early detection and/or screening of cancer with particular care.
There is still much work to be done. The main aim of this test is to detect cancers early, yet for early detection to be effective in improving patient outcome most cancers have to be at an early stage of development. Currently this approach is not able to detect many of these early stage cancers.
In addition, the results so far have been based on analysis of patients known to have cancer – we don’t yet know if this type of testing works in asymptomatic patients. This will be important to determine and more trials are planned to investigate this question. The amount of data available for different cancers also needs to increase, which will take time to collect.
Finally, we need more information on how to determine how often the cancers that are being detected early using these methods would progress to more advanced cancers, or would not develop any further. Work is ongoing to answer these questions, but it will be some time before tests of this type make their way into clinical practice – we’re not there yet.