NHS to evaluate a new blood test that detects more than 50 cancers

It is hoped that Grail’s Galleri test, which is based on identifying circulating tumour DNA, will provide early diagnosis and save many lives. On 27 November 2020 NHS Chief Executive Sir Simon Stevens said the NHS in England would offer the test annually to 165,000 people: 140,000 with no symptoms aged 50 to 79 and 25,000 patients with symptoms of cancer who have been referred to hospital for investigation. The trial should start in mid-2021. If results by 2023 prove promising numbers will be expanded to one million people over the next two years. Sir Harpal Kumar, President of Grail Europe and until 1918 Chief Executive of Cancer Research UK, said: “Grail is thrilled to partner with the NHS and UK government to support the NHS Long Term Plan for earlier cancer diagnosis, and we are eager to bring our technology to patients in the UK as quickly as we can. I’m delighted we have the opportunity to work together to see if we can save many thousands more lives from cancer in the UK every year.”

A report of the Galleri test was published by Dr M C Liu, Dr G R Oxnard, three co-authors and many colleagues in the Annals of Oncology in June 2020. They used the highest performing of three sequencing assays for the analysis of cell-free fragments of DNA (cfDNA) in the plasma of patients with cancer from centres in the USA and Canada, targeting methylated DNA regions, to identify a wide spectrum of cancers and their tissues of origin.

The authors reported a case-control study of cfDNA from 4316 participants, 4207 without cancer and 2482 patients with more than 50 types of cancer with known tissues of origin: patients’ specimens were taken before treatment. The 4316 participants were divided into two sets. A training set of 3052 (1531 cancer, 1521 non-cancer) was used to program a classifier by machine learning. A validation set of 1264 (654 cancer, 610 non-cancer) with a matched distribution of cancer types and of factors such as age and sex was used to determine the efficiency of the derived classifier. Non-cancer specimens were used in training to ensure that the classifier achieved greater than 99% specificity.

The validation results showed that overall specificity was 99.3%, with 0.7% of false positives. Overall sensitivity - the overall proportion of cancers identified - was lower at 43.9%. All cancer tumours had been classified clinically by stages (I small, II larger, III with local spread and IV with wide spread). As anticipated, detection rates were lower the earlier the stage. In a pre-specified set of 12 cancer types which account for about 63% of US annual cancer deaths, Stage I-III overall sensitivity was 67.3%. Two cancers that are difficult to diagnose early – head & neck and pancreas – had sensitivities in Stage I alone that were shown graphically to exceed 50%. (Data were neither shown nor commented on for ovarian cancer which is also difficult to diagnose early.) The tissues of origin were predicted accurately in 90% of tumours. The researchers concluded that, considering the potential value of early detection in deadly malignancies, further evaluation of the test was justified in prospective population-level studies.

Commenting on the proposed NHS evaluation of the Galleri test, Dr Mangesh Thorat, Deputy Director (Clinical) of the Cancer Prevention Trials Unit at King’s College London said “The NHS pilot appears to be this next research study” and added that those invited should be appropriately informed that they are being asked to participate in research. They will need to consider the potential harms associated with such investigations and the anxiety caused by a positive test result.

It seems likely that over the course of the NHS study the large amount of new analytical and clinical information that will be acquired will allow Grail to refine the Galleri test further. In particular, the large number of volunteers who prove in time to be cancer-free should allow the classifier to reduce the proportion of false positives, a critical factor in population screening. UK expert reactions to the announcement of the NHS pilot published by the Science Media Centre appeared generally favourable. The Centre reported that NHS England had confirmed that the project will be funded by Grail. (For those interested in the science, there is a good review of the large amount of published work on DNA methylation as a cancer biomarker in Frontiers in Genetics, November 2019.)