"Blood test could use DNA to spot eight of the most common cancers, study shows," The Guardian reports.
The test, called the CancerSEEK blood test, was developed to spot cases of cancer early on, by looking at markers in the blood such as proteins and fragments of DNA from tumours.
Researchers carried out the test on a group of people who had already been diagnosed with cancer, as well as some who had no previous disease. They focused on 8 specific types of cancer which had not yet spread.
Overall, 70% of the people who had cancer were correctly identified by the test. The proportion detected was much higher in some types of cancer (such as ovarian or liver cancer), but quite low in other types – only 33% of the people with breast cancer were correctly identified. It was also less accurate for people with early stage disease.
Although this is a promising new development, the test failed to spot cancer in 30% of the people included in the study who did have cancer. This means the test would need to be improved before it could be used on its own as a screening tool; especially in early stage disease. Ideally you would want an accuracy level at least similar to that of breast cancer screening (which is estimated to be around 87%).
Where did the story come from?
The study was carried out by researchers from a number of US institutions, including the Johns Hopkins University in Baltimore and the Memorial Sloan-Kettering Cancer Centre in New York. Other international centres in Melbourne and Italy also collaborated in this research.
The research was published in the peer-reviewed journal Science and received funding from a number of charitable foundations, clinics and public-sector sources.
The UK media`s coverage of the study was good, correctly reporting the accuracy of the test overall, while highlighting some of the limitations of the current findings. The fact that this test is not ready for use in medical settings was made clear in most of the stories.
What kind of research was this?
This study involved testing the accuracy of a new blood test designed to detect specific types of cancer. A test like this has the potential to be used in population screening programmes where people who aren`t yet showing symptoms can be tested to see whether they have signs of disease.
The study was well designed to evaluate whether the blood test could detect particular types of cancer. However, it was carried out mostly in people already known to have cancer where rates of detection could be expected to be higher. It does not tell us whether the test is better than existing ways of detecting these cancers in people without symptoms, or whether there is any overall benefit or harm from screening large numbers of people rather than limiting testing to people who actually have symptoms.
What did the research involve?
Researchers developed a blood test called CancerSEEK to find cases of cancer in the ovaries, liver, stomach, pancreas, oesophagus, colorectum (bowel), breasts, and lungs.
The test analyses small fragments of DNA circulating in the blood which have come from tumours. These DNA fragments are expected to be different from a person`s normal DNA, as cancer cells have undergone genetic mutations. The researchers also wanted to know if the test would show where any detected tumours might be located.
The CancerSEEK test also detects protein biomarkers as another way of identifying cancer, because some early tumours don`t release enough DNA into the blood to be picked up. In total, the test looks at 8 protein biomarkers and mutations at more than 2,000 locations in the DNA.
The researchers selected 1,005 people who already had one of the specified cancers which had not spread to other sites in their bodies. People participated in the research before they had received any chemotherapy. Researchers also recruited a group of 812 healthy people without any known history of cancer to act as a control group.
Both the people with cancer and the healthy control group were tested using CancerSEEK. The researchers then calculated various standard measures of how accurate the test was. This included the sensitivity (how many people who had cancer were correctly detected by the test) and the specificity (how many people who received a positive test result actually had cancer).
What were the basic results?
The performance of the test as shown by the percentage of people who were known to have cancer and received a positive test result varied across the cancer types. Ovarian and liver cancers were correctly picked up in nearly all cases. Around 70% of stomach and pancreatic cancers were also identified by the tests. Only 33% of the known cases of breast cancer were detected by CancerSEEK.
Overall the test had 70% sensitivity across all the cancer types – in other words 70% of the people who had cancer were correctly identified as having cancer by a positive test result.
The specificity of the test was 99% or more – in other words 99% of the people who had a positive test result did have cancer. This means that very few people would receive an incorrect positive test result (known as a false-positive) if the test was used in practice.
How did the researchers interpret the results?
The researchers acknowledged there were reasons why the test might not perform as well in practice compared to a test setting (for example, if people without cancer had other diseases which might cause false positive results).
They noted that for some of the cancer types in their study, there isn`t currently any kind of screening test for people who aren`t already considered to be high risk.
This is a promising piece of research on a new test which could potentially be used alongside other tests for detecting cancer. However, it has some limitations which means it cannot yet be used in clinical settings:
- People in the healthy control group who got a positive result from the test were considered to be "false positives". However, it is possible that some of them may actually have cancer which had not previously been detected, and this wasn`t investigated further in the research.
- While the test correctly identified most of the people with ovarian and liver cancer, it performed a lot less well on some types of cancer. Overall, 30% of people with known cancers were not identified as having cancer at all. This would suggest that the test or the cut-off points for abnormality may need to be changed or improved.
- This study was not designed to see whether this test was as good as or better than other types of cancer screening (for example, imaging or biopsies). It would need to go through this kind of evaluation before it could be used more widely.