CC-0
CC-0

EXPERT REACTION: AI-based blood test shows promise in detecting 50 cancers and pinpointing their location in the body

Embargoed until: Publicly released:
A new blood test aided by artificial intelligence (AI) can detect 50 different types of cancer, and tell doctors where the tumour is located in the patient's body, according to US and UK scientists. However, across all 50 cancers, the test is much more effective at picking up later stage disease. At stage I, it detected just one in five cancers, while at stage IV, it detected more than nine out of ten cancers. The researchers also gauged its performance in 12 types of cancer that are often the most deadly, and the test performed better, picking up two in five cancers at stage I, and more than nine in ten at stage IV. In more than nine out of ten cases, the test accurately predicted the whereabouts of the cancer in the body, and the rate of false positives was low at less than one in 100. The test is based on detecting genetic material shed into the blood by tumours, and the AI looks for known cancer signatures in the DNA.

Journal/conference: Annals of Oncology

Link to research (DOI): 10.1016/j.annonc.2020.02.011

Organisation/s: Mayo Clinic, USA

Funder: This work was supported by GRAIL, Inc. (Menlo Park, CA; no grant number). This publication was also partially supported by Princess Margaret Cancer Centre’s McCain GU BioBank in the Department of Surgical Oncology (grant number REB # 08-0124); its contents are solely the responsibility of the authors and do not necessarily represent the official views of the University Health Network. CS is supported by the Francis Crick Institute that receives its core funding from Cancer Research UK [grant numbers FC001169, FC001202]; the UK Medical Research Council [grant numbers FC001169, FC001202]; and the Wellcome Trust [grant numbers FC001169, FC001202]. CS is funded by Cancer Research UK (TRACERx; PEACE; and CRUK Cancer Immunotherapy Catalyst Network, no grant number), the CRUK Lung Cancer Centre of Excellence (no grant number); the Rosetrees Trust (no grant number); and the Breast Cancer Research Foundation (BCRF, no grant number).

Media Release

From: Elsevier

Blood test accurately detects over 50 types of cancer, often before symptoms show

Test also identifies where in the body the cancer has originated

Researchers have developed the first blood test that can accurately detect more than 50 types of cancer and identify in which tissue the cancer originated, often before there are any clinical signs or symptoms of the disease.

In a paper published in the leading cancer journal Annals of Oncology [1] today (Tuesday) the researchers show that the test, which could eventually be used in national cancer screening programmes, has a 0.7% false positive rate for cancer detection, meaning that less than 1% of people would be wrongly identified as having cancer. As a comparison, about 10% of women are wrongly identified as having cancer in national breast cancer screening programmes, although this rate can be higher or lower depending on the number and frequency of screenings and the type of mammogram performed.

The test was able to predict the tissue in which the cancer originated in 96% of samples, and it was accurate in 93%.

Tumours shed DNA into the blood, and this contributes to what is known as cell-free DNA (cfDNA). However, as the cfDNA can come from other types of cells as well, it can be difficult to pinpoint cfDNA that comes from tumours. The blood test reported in this study analyses chemical changes to the DNA called "methylation" that usually control gene expression. Abnormal methylation patterns and the resulting changes in gene expression can contribute to tumour growth, so these signals in cfDNA have the potential to detect and localise cancer.

The blood test targets approximately one million of the 30 million methylation sites in the human genome. A machine learning classifier (an algorithm) was used to predict the presence of cancer and the type of cancer based on the patterns of methylation in the cfDNA shed by tumours. The classifier was trained using a methylation database of cancer and non-cancer signals in cfDNA. The database is believed to be the largest in the world and is owned by the company involved in this research, GRAIL, Inc. (California, USA).

Senior author of the paper, Dr Michael Seiden (MD, PhD), President of US Oncology (Texas, USA), said: “Our earlier research showed that the methylation approach outperformed both whole genome and targeted sequencing in the detection of multiple deadly cancer types across all clinical stages, and in identifying the tissue of origin. It also allowed us to identify the most informative regions of the genome, which are now targeted by the refined methylation test that is reported in this paper.”

In the part of the Circulating Cell-free Genome Atlas (CCGA) study reported today, blood samples from 6,689 participants with previously untreated cancer (2482 patients) and without cancer (4207 patients) from North America were divided into a training set and a validation set. Of these, results from 4316 participants were available for analysis: 3052 in the training set (1531 with cancer, 1521 without cancer) and 1264 in the validation set (654 with cancer and 610 without cancer). Over 50 types of cancer were included.

The machine learning classifier analysed blood samples from the participants to identify methylation changes and to classify the samples as cancer or non-cancer, and to identify the tissue of origin.

The researchers found that the classifier’s performance was consistent in both the training and validation sets, with a false positive rate of 0.7% in the validation set.

The classifier’s ability to correctly identify when cancer was present (the true positive rate) was also consistent between the two sets. In 12 types of cancer that are often the most deadly (anal, bladder, bowel, oesophageal, stomach, head and neck, liver and bile duct, lung, ovarian and pancreatic cancers, lymphoma, and cancers of white blood cells such as multiple myeloma), the true positive rate was 67.3% across clinical stages I, II and III. These 12 cancers account for about 63% of cancer deaths each year in the USA and, at present, there is no way of screening for the majority of them before symptoms show. The true positive rate was 43.9% for all cancer types in the study across the three clinical stages.

Detection improved with each cancer stage. In the 12 pre-specified cancers, the true positive rate was 39% in stage I, 69% in stage II, 83% in stage III and 92% in stage IV. In all of more than 50 cancer types, the corresponding rates were 18%, 43%, 81% and 93%, respectively.

The test was also consistent between the training and validation sets in its ability to identify the tissue where cancer had originated, with an accuracy of 93% in the validation set.

Dr Seiden said: “These data support the ability of this targeted methylation test to meet what we believe are the fundamental requirements for a multi-cancer early detection blood test that could be used for population-level screening: the ability to detect multiple deadly cancer types with a single test that has a very low false positive rate, and the ability to identify where in the body the cancer is located with high accuracy to help healthcare providers to direct next steps for diagnosis and care.

“Considering the burden of cancer in our society, it is important that we continue to explore the possibility that this test might intercept cancers at an earlier stage and, by extension, potentially reduce deaths from cancers for which screening is either not available or has poor adherence. To our knowledge, this is the largest clinical genomics study, in participants with and without cancer, to develop and validate a blood test for early detection of multiple cancers.”

The study is funded by GRAIL, the maker of the blood test. Researchers are continuing to validate the test in large, prospective studies in the USA (STRIVE and PATHFINDER studies) and the UK (SUMMIT study), and to examine its feasibility for screening populations [2].

A strength of the CCGA study is that it includes more than 15,000 participants from 142 clinics in North America, ensuring results are generalisable to a diverse population. The ongoing studies are assessing the test’s performance in even broader populations. Limitations include: all the participants with cancer had already been diagnosed with cancer (e.g. via screening or patients presenting with symptoms); the study was not designed to establish the test’s impact on death from cancer or other causes; at the time of this analysis, not all patients had been followed for a year, which is needed to ensure their non-cancer status was accurate; and some inaccuracy occurred in the detection of the tissue of origin for cancers that are driven by the human papilloma virus (HPV), such as cancers of the cervix, anus, and head and neck – this information is being used to improve the test’s performance.

Editor-in-chief of Annals of Oncology, Professor Fabrice André, Director of Research at the Institut Gustave Roussy, Villejuif, France, said: “This is a landmark study and a first step toward the development of easy-to-perform screening tools. Earlier detection of more than 50% of cancers could save millions of lives every year worldwide and could dramatically reduce morbidity induced by aggressive treatments.

“While numbers are still small, the performance of this new technology is particularly intriguing in pancreatic cancer, for which mortality rates are very high because it is usually diagnosed when it’s at an advanced stage.”

Notes:

[1] “Sensitive and specific multi-cancer detection and localization using methylation signatures in cell-free DNA”, by M.C. Liu et al. Annals of Oncology. doi: https://doi.org/10.1016/j.annonc.2020.02.011

[2] ClinicalTrials.gov numbers: NCT03085888, NCT03934866, NCT04241796.

 

The study was funded by GRAIL Inc. Publication was partially supported by Princess Margaret Cancer Centre’s McCain GU Biobank, Ontario, Canada.

 

N.B. Journalists: please provide the following permanent link to the published paper on the Annals of Oncology website in your stories:

https://www.annalsofoncology.org/article/S0923-7534(20)36058-0/fulltext

 

Annals of Oncology is a monthly journal published on behalf of the European Society for Medical Oncology (ESMO) by Elsevier.

Please acknowledge Annals of Oncology as a source in any reports.

 

Annals of Oncology website: https://www.annalsofoncology.org

ESMO website: http://www.esmo.org/

Attachments:

Note: Not all attachments are visible to the general public

  •  Research
    Elsevier
    Web page
    The URL will go live after the embargo ends

Expert Reaction

These comments have been collated by the Science Media Centre to provide a variety of expert perspectives on this issue. Feel free to use these quotes in your stories. Views expressed are the personal opinions of the experts named. They do not represent the views of the SMC or any other organisation unless specifically stated.

Dr Kristina Warton is a Postdoctoral Fellow at the Ovarian Cancer Early Diagnosis Project and Gynaecological Cancer Research Group in the Lowy Cancer Research Centre at the University of New South Wales

This is exciting work bringing together cutting edge laboratory techniques with AI. It highlights the potential of a test for cancer DNA in blood.

One of the strengths of the study is the large number of samples from healthy people that were included, because every cancer test has to do two things. Firstly, it has to correctly identify people who have cancer. But it also has to correctly identify people who don’t have cancer, so false positive results aren’t given out. You need lots of samples from people without cancer to show that the test doesn’t give false positives, and this study had several thousand.

The test in the study was good at locating which organ of the body the cancer was in. Unlike imaging, a blood test may not readily identify where the cancer is, and the authors use the phrase ‘diagnostic odyssey’ to describe what might happen to a patient who is told they have cancer, perhaps a small, early stage one, but you can’t tell where. In this study, the location of the cancer was specifically addressed.

Finally, where the challenge is, for this screening test and for all cancer screening tests, is to identify small, early stage cancers. Advanced cancers are a lot easier to detect. I would say that detecting the small, early ones is still a work in progress.

Last updated: 30 Mar 2020 12:34pm
Declared conflicts of interest:
None declared.
Associate Professor Michelle Hill is Group Leader of Precision and Systems Biomedicine at QIMR Berghofer

This large study, together with earlier reports confirm the potential to detect late stage cancer by a blood test. Blood test could be useful for screening high risk population, but it needs to detect cancer at an early stage, ideally pre-malignant stage.

Biological signals for cancer progression from early stage (I and II) to late stage (III and IV) are likely different to development from benign pre-cancer conditions to early cancer. As reported in this paper, algorithm for late stage cancer detection does not perform as well for early stage cases.

To enable early detection, research should focus on the progression from known precursor lesions to early cancer (dysplasia).

Last updated: 30 Mar 2020 12:26pm
Declared conflicts of interest:
None declared.
Dr Sriganesh Srihari is an Advanced Queensland Industry Research Fellow at the QIMR-Berghofer Medical Research Institute

Every cell in our body can ultimately be traced back to the single-cell zygote, but despite carrying the same genome (DNA) the cells have differentiated into numerous cell types forming different organs and tissues. Such differentiation has been possible due to methylation of the DNA that turns on and off selected genes, thereby driving cells down different differentiation ‘paths’. Abnormal methylation, often arising from lifestyle, environment, or other reasons can also lead to cells being driven down the ‘cancer path’.

However, this also means methylation patterns provide one of the strongest signals to detect cancers, particularly the types and tissues of origin (TOO) of these cancers. The large consortium study presents a technology for cancer detection as a minimally invasive test (a blood test) by measuring methylation patterns from the DNA released by cancer cells into the blood stream (i.e., cell-free DNA).

The trained classifier was able to achieve >90% true-positive rates for detecting cancer type and tissues of origin specifically for stage III-IV cancers, an impressive detection rate because of the complications arising from metastasis to distant organs at these advanced stages.

Although it is hard to say at this occasion how immediately useful the test could be for population-wide screening or for catching cancers in their early stages, the technology and the test could be valuable tools to monitor treatment progress/response and to catch early signs of metastasis in advanced cancers via a simple blood test.

Last updated: 30 Mar 2020 12:24pm
Declared conflicts of interest:
Sriganesh has declared no conflicts of interest.
Professor Hans Zoellner is the Head of Oral Pathology at the University of Sydney

A paper to be published on 31st of March in Annals of Oncology, describes a powerful new blood test that could be used to screen for cancer. 

Cancer screening usually looks for one particular type of cancer at a time. However, the new test screens for many different cancers, and has remarkable specificity identifying where the cancer comes from. 

The test relies on DNA shed into the circulation by cancer. Although every cell has the same DNA, cells vary greatly in different organs; cells of the heart for example, are very different from those in the liver. This is because different stretches of DNA become inactivated, in different types of cells. One way DNA gets inactivated, is for it to be chemically decorated by ‘methylation’. From this, cells can be told apart by their DNA methylation patterns, and this is how the new test identifies the type of cancer. 

An advanced computer technique ‘machine learning’, was used to ‘train’ a computer to identify cancer DNA. This was done with blood from around 1,500 cancer patients, and a similar group without cancer. They then checked the trained computer with blood from around 650 different patients with cancer, and a similar number without. Use of machine learning, foreshadows a progressive shift from medicine’s dependence on humans, to what might become more reliable computer diagnosis. 

While the test is very good at telling where a cancer is, it is less good at detecting very early cancers, where only 39 per cent of cases were picked-up. The more advanced the cancer, the more reliable the test becomes. There were very few ‘false positive’ results, which is an advantage for any clinical test. 

There seems good reason to be optimistic, that the test can be further improved to become much more reliable at detecting very early cancers. If that happens, we might all be lining up for regular blood tests, to catch most cancers well before they get deadly."

Last updated: 30 Mar 2020 12:02pm
Declared conflicts of interest:
None declared.
Dr Ilaria Pagani is a biomedical researcher at the South Australian Health and Medical Research Institute (SAHMRI)

To my knowledge this is the first study of its kind which represents an important milestone for non-invasive cancer diagnosis. This study has the potential to lead to a unique predictive test for all forms of cancer eliminating the need for multiple screenings. I am confident this work will open new horizons toward the non-invasive early detection of cancer, through an increase of the sensitivity of the sequencing technology in the future.

Last updated: 30 Mar 2020 12:00pm
Declared conflicts of interest:
None declared.

News for:

International
NSW
QLD
SA

Media contact details for this story are only visible to registered journalists.