The researchers were training the system to look for patterns of methylation, which are small chemical changes to our DNA that help control whether a gene is active or inactive. DNA from cancers tend to have very different patterns of DNA methylation than DNA from healthy cells.
The multi-centre trial, including the Dana-Farber Cancer Institute, the Mayo Clinic, Harvard Medical School, the Francis Crick Institute and University College London, used next-generation sequencing techniques to detect changes in methylation patterns of DNA, which has been released from tumours and normal cells into the bloodstream. The researchers first trained the system using data on methylation patterns in DNA from patients with and without cancer. The system was then able to sort these patterns into distinct groups and the researchers taught the system which patterns corresponded to which type of cancer.
Dr Geoffrey Oxnard of Boston's Dana-Farber Cancer Institute, part of Harvard Medical School, said: 'Our results show that this approach to testing cell-free DNA in blood can detect a broad range of cancer types at virtually any stage of the disease, with specificity and sensitivity approaching the level needed for population-level screening'.
The researchers used 6689 blood samples, of which 2482 came from patients with cancer and 4207 without. The samples included over 50 different types of cancer, including bowel, lung and pancreatic cancer. The test resulted in a specificity of 99.3 percent. It is vital for any screening test to make sure that people are not tested positive when they are actually healthy.
On the other hand, the same is true for sensitivity, which assesses the number of cancers that are missed in the test. The new blood test has a sensitivity of 67.3 percent in an assay that looked at 12 different cancers. The detection improved significantly with advanced stages of the cancer.
The overall sensitivity for all 50 cancer types combined for stage one to three was 43.9 percent, meaning that more than half of the cancers were still missed. However, when a cancer was detected, over 90 percent were attributed correctly to the tissue or organ of origin. This test could prove particularly exciting for cancers that are hard to detect, such as pancreatic cancer, where 63 percent of stage one early pancreatic cancers were correctly picked up by the test.
Cancer Research UK early detection head Dr David Crosby said: 'Detecting cancers at their earliest stages, when they are less aggressive and more treatable, has a huge potential to save lives and we sorely need tech innovations that can turn this potential into reality… Although this test is still at an early stage of development, the initial results are encouraging.'
Dr Crosby continued 'But more research is needed to improve the test's ability to catch early cancers and we still need to explore how it might work in a real cancer-screening scenario.'