Simple Blood-based Cancer Tests Show Early Promise

Liam Davenport

November 07, 2019

GLASGOW — The promise of simple, blood-based tests to detect cancers earlier and hopefully improve outcomes took a step closer with two studies, in breast cancer patients and individuals suspected of having a brain tumour, presented at NCRI Cancer Conference 2019.

Daniyah Alfattani, a PhD student at the Breast Surgery Division, University of Nottingham, and colleagues studied breast cancer and control blood samples for tumour-associated antigens (TTAs) associated with breast cancer.

Preliminary but Encouraging

Combining several of the antigens into a panel, they were able to single out breast cancer cases with a sensitivity of nearly 40% and a specificity approaching 80%.

Describing the data as "preliminary", the team says the study "opens the possibility of a blood test for screening and detection of breast cancer".

Daniyah Altaffani said in a news release that, while further validation of the test is required, "these results are encouraging and indicate that it's possible to detect a signal for early breast cancer".

If successful, she believes that a "blood test for early breast cancer detection would be cost effective, which would be of particular value in low and middle income countries".

"It would also be an easier screening method to implement compared to current methods, such as mammography."

Dr Iain Frame, CEO of NCRI, commented that the results "are promising and build on this research group's expertise in other cancers, such as lung cancer.

"It's obviously early days but we look forward to seeing the results from the larger group of patients that are now being investigated."

Other experts suggested that the results were perhaps too preliminary to be promising at this stage.

Second Study

For the second study, Dr Paul Brennan, senior clinical lecturer and honorary consultant neurosurgeon at the University of Edinburgh, studied patients suspected of having a brain tumour who were referred by their general practitioner for imaging tests.

They performed infrared spectroscopy to examine the signature of whole blood samples, finding that, compared with imaging, the test was 81% accurate in picking out patients with brain tumours, rising to 92% in those with glioblastomas.

Dr Matthew Baker is reader in chemistry at the University of Strathclyde, and chief scientific officer at ClinSpec Diagnostics Ltd, which combined the spectroscopy technique with artificial intelligence to create the test.

He said in a news release: "These results are extremely promising because they suggest that our technique can accurately spot who is most likely to have a brain tumour and who probably does not."

Dr Baker added: "Because the technique requires just a small blood sample, it offers the potential to test a large number of people with suspicious symptoms and give the best indication of who needs an urgent brain scan.

"This could ultimately speed up diagnosis, reduce the anxiety of waiting for tests and get patients treated as quickly as possible."

Dr Sarah Jefferies, clinical director for cancer at Addenbrooke's Hospital, Cambridge, and a member of the NCRI's glioma subgroup, commented: "The number of people being diagnosed and dying from brain tumours is increasing and we urgently need better ways to spot and treat the disease.

"This type of testing offers a number of potential advantages. It's relatively straight-forward for the patient, who need only have a blood test.

"For the health service, it could combine with clinical assessment to make the process of referring patients for brain scans more efficient."

Early Detection Biomarkers

Daniyah Alfattani and colleagues' research came from the observation that autoantibodies against TTAs are not only tumour biomarkers but also can be detected up to 5 years before the tumour becomes clinically apparent.

To develop a blood test to exploit this phenomenon and hopefully aid in the early detection of breast cancer, the team studied blood samples taken at the time of diagnosis in 90 primary breast cancer patients and 90 controls.

They screened the samples for 67 TTAs previously associated with breast cancer pathology, of which 18 showed a degree of discrimination between breast cancer cases and controls.

While individual TTAs did not show high sensitivity for identifying cancer cases, combining the most promising candidates into panels improved the results.

A panel of nine TTAs achieved a sensitivity of 38.2% for detecting breast cancer versus controls, with a specificity of 78.7%.

Next, the team will screen 800 samples with the panel of nine antigens to increase the sensitivity and specificity and determine the cut-off points for each antigen, followed by a validation study.

More Research Needed

Commenting on the findings, Dr Jeremy Carlton, King's College London & The Francis Crick Institute, London, pointed out in a statement that the results indicate the test "failed to detect cancer in 60%–70% of patients that were known to have cancer.

"Importantly, this test also mis-reported cancer in around 20% of patients known to be cancer-free."

He added: "In short, this study is too preliminary to support the claims made…and it will be important to ensure that any future diagnostic test is as accurate, sensitive, and reliable as possible."

Prof Paul Pharoah, professor of cancer epidemiology, University of Cambridge, added: "The search for a blood test that is capable of detecting cancer very early, before it causes any symptoms is one of the major goals of research in early cancer detection."

However, he believes, like Dr Carlton, that the data are "clearly very preliminary" and called for "a lot more research…before any claim can be made that this is likely to represent a meaningful advance in the early detection of cancer".

Alternative Approach

Presenting their approach for detecting brain tumours, Dr Brennan said that, "often when you think about blood tests, you think about prostate specific antigen and CA125 for cancer, which are single biomarker assays, and…that doesn't give you a huge amount of information".

He went on to say that the issue with liquid biopsies of circulating tumour DNA is whether enough material has been obtained and whether it is adequately sequenced.

"No doubt that will all be overcome at some point," he said, adding: "But this is the here and now."

He and his colleagues therefore developed a rapid blood test that does not look at a single protein or molecule but rather provides a snapshot of all the molecules in the blood.

The test, known as attenuated total reflection (ATR)-Fourier transform infrared (FTIR) spectroscopy analysed the presence of more than 20,000 proteins in serum samples.

This was combined with a machine learning algorithm and trained on a retrospective cohort of samples taken from 237 unaffected controls and 487 tumours, categorised using the World Health Organisation classification.

The initial results, which were recently published, were sufficiently promising for the test to be trialed on a first group of 104 patients recruited for a prospective clinical validation study.

In Glasgow, Dr Brennan explained that the cohort has now been expanded out to 400 patients referred from their GP with a suspicion of brain tumour.

For this, the blood test is performed by the laboratory, which is blinded to the patient history, and the results compared with the results on imaging.

When looking at primary glioblastoma, the test was found to have a sensitivity for identifying patients with the disease of 92%, and a specificity of 80%.

Dr Brennan said that, on a practical level, if a symptom support algorithm flags up that a patient has got a risk of a brain tumour, if the blood test comes back as negative, "you're going to be right 92% of the time".

"It's not perfect, but nothing in life is perfect, and this is part of clinical decision making, it's not a diagnostic test."

For all brain tumours, the sensitivity of the test was lower, at 81%, while the specificity remained at 80%.

Dr Brennan explained that the most common tumour included in the training dataset was glioblastoma, "and what that tells us is, if we increase the number of non-glioblastoma tumours in the algorithm—so more metastases, more lymphomas, more of the things which are a problem in the clinic—then the performance of the test will go up".

The researchers say that next, they will examine the test in a further 600 patients referred for a brain scan via their GP or the hospital emergency department.

Poster 2966 was funded by the University of Nottingham. Brennan and Baker's research was funded by Scottish Enterprise.

Matthew Baker is an employee of ClinSpec Diagnostics Ltd. No other conflicts of interest declared.

NCRI Cancer Conference 2019: Poster 2966. Presented 3 November; Artificial intelligence – does it have promise for cancer prevention and early diagnosis? Presented 4th November.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: