CancerSEEK - Hope for cancer screening?

CancerSEEK - Hope for cancer screening?
Screening tests for the various cancer types are generally invasive and unpleasant, often taking place at more advanced stages of disease. CancerSEEK is effectively a blood test which has proven to be capable of identifying 8 types of cancer known to have the highest mortality rates, five of which currently do not have an accurate early detection screening test available.

Which types of cancer can CancerSEEK identify?

It took more than 3 decades worth of cancer genetics research, but the team at Johns Hopkins Kimmel Cancer Center have determined that non-invasive screening is possible for the following types of cancer:

  • Ovarian
  • Liver
  • Stomach
  • Pancreas
  • Oesophagus
  • Lung
  • Breast
  • Colorectal

Currently, cancer affecting the liver, stomach, ovaries, pancreas and oesophagus do not have effective screening tests for medical professionals to use when making a diagnosis. These cancers typically only present symptoms at more advanced stages too, making treatment that much more difficult, and resulting in higher mortality rates.

How does CancerSEEK work?

The first genetic blueprints of cancer were developed by the Kimmel Cancer Center’s Ludwig Center Laboratory (Baltimore, USA). To date, this is where a large percentage of research has taken place. Through studies, researchers have managed to develop what they describe as a ‘non-invasive multianalyte test’ (i.e. a simultaneous detection test) capable of detecting levels of cancer proteins (by testing for 8 protein biomarkers) as well as identifying cancer gene mutations (in 16 different genes) within the bloodstream.

Previous research conducted by the Johns Hopkins Kimmel Cancer Center identified the relevance of using genes, and analysis of DNA in the blood, for early stage cancer detection. (1) Protein markers have also been studied for detection purposes. CancerSEEK allows for immunoassay testing which can measure the concentration of macromolecules (molecules containing a large number of atoms, like a protein molecule) in the blood.

Researchers started their work exploring hundreds of genes and 40 protein markers known to potentially help identify cancer before it spreads from the initial site of disease in the body (metastasis). finally identifying 8 key proteins and 16 DNA (gene) segments. (2) The final 8 protein biomarkers were sufficient enough to enable the team to identify a positive or negative result for their group of cancer types (i.e. in cancer patient participants versus non-cancer participants).

Determining the levels of the final 8 protein biomarkers is important as many early-stage cancer tumours do not normally secrete high enough amounts of ctDNA (circulating tumour DNA) to be detected in the blood on their own. The team felt the need to develop this multianalyte approach so as to factor in sensitivity when screening for cancer markers. When combining protein marker information with ctDNA analysis, the chances of more accurate results are theoretically higher.

The team took care to develop a small, yet potent mutation panel in order to detect and identify cancer gene mutations. This was identified as being important as it minimises the amount of false-positive results which may lead to unwarranted invasive treatments and ensures that the test is affordable (currently estimated to cost under $500).

Through the test screening, research teams were able to determine the location of a tumour using the multianalyte data received. This result in itself sets the test apart from current testing processes known as ‘liquid biopsies’, which also look for cancer markers in the blood but sometimes with unreliable accuracy. Currently, a liquid biopsy cannot accurately determine the origin of a tumour either. These tests are effectively geared towards analysing large numbers of genes in order to determine the best course of therapy. In contrast, CancerSEEK is aimed at effective screening.

Why is detection high on the agenda?

Early detection of cancer markers is largely regarded as critical when it comes to reducing high mortality rates. CancerSEEK offers some light at the end of the tunnel in this regard. By using a combination of selected biomarkers and genes, early and accurate detection may soon be possible and could change the way medical professionals go about screening for various cancers in future.

Currently, treatment practice uses therapeutic combinations as a way to battle the disease. If cancer could be identified early enough with CancerSEEK, the need to use so many combination therapies may not be all that necessary. It is even being speculated that surgery alone could potentially cure a person’s condition if the cancer present is detected early enough using CancerSEEK.

“Early detection provides many ways to improve outcomes for patients,” says Anne Marie Lennon, MD, PhD, professor of medicine, surgery and radiology; clinical director of gastroenterology, and one of the research corresponding authors.

How were gene mutations and protein biomarkers analysed during research?

Research first needed to identify mutations in ctDNA which could be analysed from a sample of blood taken from one of the study’s participants.

The study focussed on 1 005 cancer patients who had been diagnosed with stage one, two or three non-metastatic cancers (i.e. cancers that have not spread from the primary site). The research team also included 812 healthy participants. These participants had no cancer history and could thus be used in the study as a ‘control population’ group.

DNA samples were then purified from plasma (the yellowish liquid component of blood) and amplified using a multiplex-polymerase chain reaction (i.e. a robust 61-amplicon panel) to help identify rare mutations. The amplified products (i.e. 61 primer pairs) could then be augmented further to identify between 66 and 80 base-pair segments for comparison with 16 genes of interest. Each amplified sample was then labelled with a unique DNA barcode and could be matched against the dataset in the Catalogue of Somatic Mutations in Cancer.

Rack of blood sample tubes with barcode labels.

What was the study able to determine?

Participating cancer patients all had one of the 8 types of cancer being analysed and had not yet undergone any surgical treatment. The research team found that the majority of enrolled cancer patient participants had a stage two diagnosis (49%), followed by stage three (31%) and stage one (20%). The average age of diagnosis amongst patient participants was 64. The average age of the control population group was 55 years.

The final group of participants were carefully screened beforehand. The study’s parameters dictated that participants should not have had a prior neoadjuvant therapy (i.e. the administration of a therapeutic agent before a primary treatment measure), a diagnosed stage four cancer diagnosis (following the removal of all or part of an organ, tissue or other structure) or a record of blood having been drawn while under anaesthesia.

For those cancer patient participants who required a surgical resection procedure, whereby removal of either all or part of a specific tissue structure or organ was necessary around the time of study commencement, blood was drawn beforehand.

Researchers used CancerSEEK to analyse all participant blood samples against gene mutations (2001 genomic positions within 16 genes) and levels of numerous different protein markers which included the final 8 proteins.

The team applied machine-learning tools and statistical analysis to identify an average sensitivity percentage (i.e. the ability of the test to accurately detect cancer). This is important for being able to detect early stage cancer.

A positive result was determined if the mutation frequency in a cancer participant’s sample showed in at least 1 of the 16 genes. A positive result was also determined if significantly elevated levels in at least 1 of the 8 proteins were identified (in comparison to the control population group).

The research team found that their developed CancerSEEK blood test boasted:

  • A 78% sensitivity rate when testing for stage three cancers
  • A 73% sensitivity rate for stage two cancers
  • A 43% sensitivity in detecting stage one cancers

It was observed that CancerSEEK could identify an overall average sensitivity of 70% for all 8 cancers, with individual ranges for each of the separate types (e.g. an average sensitivity of 98% for ovarian cancer and 33% for breast cancer). Higher sensitivity percentages could be identified with stage 1 liver cancer (100%) and the lowest were found in detecting cancer of the oesophagus (20%).

At the moment, less than 50% of localised cancer is detectable at during the early stages and can be picked up in the blood using current screening tools. CancerSEEK thus shows promising capabilities when it comes to increasing the o of more accurate detection.

Significantly, cancer sensitivity percentages for those cancer types for which there is currently no screening test available (i.e. liver, stomach, ovaries, pancreas and oesophagus) ranged between 69% and 98%.

It was also determined that 7 of the 812 control population group (approximately 1%) tested positive for sensitivity, meaning that it is possible that these individuals have a predisposition for developing cancer (or potentially have the disease in the very early stages) but are currently asymptomatic. Their results were thus labelled as false-positive as they are not actually diagnosed cancer patients.

Nevertheless, the test findings thus show that there is a less than 1% margin of error when it comes to detecting disease in undiagnosed individuals. Follow-up screening on these individuals using other available measures may be required to either confirm or rule out a cancer diagnosis based on the results of the test.

Once a positive result had been determined, the research team went on to see if they could identify tissues of origin (i.e. where the malignancy started in the body). To do this, an artificial intelligence algorithm was developed factoring in ctDNA, protein biomarker levels, 31 other known proteins involved and a patient’s gender.

The team grouped together cancer types according to the most likely follow-up screening procedure. For instance, oesophageal and gastric cancer participants are likely to have a recommended endoscopy screening, and so were appropriately grouped.

Further analysis showed that two anatomic origin locations could be effectively identified using CancerSEEK within an average of 83% of positive scoring cancer patients. The test could determine tissues of origin to a single organ at a 63% average too. Accuracy was determined as being highest for colorectal cancer patients, averaging at 84%, and lowest for those with lung cancer, averaging at 39%.

A high level of specificity was of the utmost importance for the research team. The nature of cancer as a disease is very specific in itself. To be able to develop a blood test that can ensure significantly low chances for false-positive results makes for better screening detection purposes.

The research team acknowledges that the majority of information providing localisation data was obtained from the protein biomarkers. Normally, main gene mutations are not typically tissue-specific.

The team views this recently developed test as a step in a new direction which shines a positive light on cancer detection research. Many developments are already in play and being thoroughly researched. Screening as well as treatment detection techniques, like the MasSpec Pen which has been shown to be able to pick up malignant tissues at a much faster rate during treatment procedures, are already showing promising results. But cancer is a tricky area of medicine with a vast array of specifics to factor in. And specifics matter.

CancerSEEK is not intended to replace other screening tests currently in use, including liquid biopsies, which already are a non-invasive means to detect cancer, either at a time of diagnosis or potential disease relapse. Liquid biopsies and CancerSEEK are both useful detection tools since DNA from malignant cells mutate, and these mutations typically present (at some level) in the bloodstream. Both detection tools work by sequencing DNA in the blood.

CancerSEEK does appear to have potential to provide additional information which would be highly valuable for individuals with a predisposition for malignant tumours (i.e. ideally detecting cancer before becoming symptomatic). That said, there is a still a long way to go before this screening blood test can be feasible for clinical use for the general public.

Limitations of the study

One of the most obvious observations regarding the study is that the majority of enrolled participants had one of 8 diagnosed cancer types. Although the research ensured that none had any clinical signs of metastatic conditions, all had been diagnosed within one of three stages of the disease.

The CancerSEEK test is intended to be used as a screening test. In this sense, the blood test would need to prove most useful in a true screening setting (i.e. be able to identify cancer before a diagnosis is actually made). The majority of individuals in the study were already in a more advanced stage than the average population candidate. This key observation was, however, acknowledged by the research team.

Some comments by those in medical circles have also highlighted that even the sensitivity percentages achieved with CancerSEEK testing on cancer patient participants is a little lower than they would like. This is a relevant concern, especially given that the test results showed a higher the percentage of accuracy in participants with more advanced conditions, and not as much so in those with earlier stage malignancies. Effectively, this could mean that the screening capability, as it is now, could miss a sizable proportion of cancers at earlier stages when doctors would prefer to make a diagnosis.

Another point to note relates to identified cancer-related proteins which could determine tissue damage. Such tissue damage could also potentially show up in individuals with inflammatory conditions, like arthritis. It is already known that many conditions of an inflammatory nature do tend to influence protein behaviour, causing (up and down) fluctuations in serum proteins. It is also an agreeable point that a fair percentage of the overall population are not all in impeccable health either. Effectively, this means that the original 1% false-positive result of non-cancer patient participants would be expected to be higher if tested in less healthy population individuals. Thus, this aspect of the CancerSEEK screening test requires further specific analysis.

Also identified as a limitation of the current research is the use of a single dataset to compare participant information. This dataset provided validation to the research findings, but it is entirely possible that a comparison with another dataset could show somewhat different results. A question mark lingers here for many an expert in the field, who appear curious to see if there are any distinctive differences when using different datasets for comparative validation.

The Johns Hopkins team does not seem to refute these points. Much is yet to be done in order to accomplish the vision of the test, and they do acknowledge this.

Nurse with syringe withdrawing blood sample for testing at the doctor's office.

What are some of the initial thoughts of CancerSEEK?

CancerSEEK is in no way yet ready for clinical use. There is certainly potential, but there is a great deal of work still to be done in order to make the kind of progress the general public, and those practicing in medical fields, can actually benefit from.

Cancer research, in general, is an enormous undertaking. Many research studies, like CancerSEEK, show numerous holes and drawbacks in their findings. With such high areas of sensitivity and information specifics needed to be taken into consideration, it’s important to ‘get it right’. And that is no easy feat.

For those in the medical field, specificity is critical. It can even come down to specific information regarding one tissue compared with another, which can make all the difference. It’s that intricate. CancerSEEK does suggest that this is a possibility, which offers scope for further, more extensive research.

Future studies will need to establish very specific data to show that detecting cancer is possible for all members of the public, and also prove that screening using this process is capable of improving survival rates, something experts would very much like to see. For medical professionals, to be able to screen for undiagnosed malignancies, and determine specific cancers at a particular point during the symptomatic stage where they are more likely to be curable is a favourable notion. If this can be achieved somewhere down the line, it could change so much when it comes to cancer survival rates.

In relation to the obvious limitation of establishing cancer detection in already diagnosed patient participants, a professor of cancer epidemiology at the University of Cambridge, Paul Pharoah, PhD further emphasises the point to ‘get specific’.

Protein biomarkers for cancer detection is not necessarily a new line of research. Pharoah points out that they have been on the radar for many years and demonstrated some success in diagnosing symptomatic malignancies. Key here, is that none have yet shown great success in detecting cancer in a screening test capacity. What the CancerSEEK team have done is combine the use of protein biomarkers with ctDNA analysis for detection purposes. Both areas of focus have been independently studied with regards to cancer detection. Now, the information is being analysed together in one study. This in itself is a new detection method and shows potential in obtaining more accurate detection results across a wider range of cancer types.

What some of the team’s findings certainly determine is that the screening test could show high percentage detection results in more advanced stage cancer conditions. Pharoah stresses the point that this does not necessarily make CancerSEEK as it is now able to effectively and accurately detect early stage symptomatic cancer (with an average in the 40% range), or even pre-symptomatic conditions. Here, research needs more refined work.

In general, initial responses to the published report do appear to be sceptical, with many questions being raised, but do not seem to regard this newly developed screening tool as a total write-off. There is certainly evidence in the research findings thus far that does show significant promise.

Could CancerSEEK be the next ‘big thing’?

It is an agreeable possibility that CancerSEEK could very well become the kind of screening tool its team envisions – a non-invasive, easy-to-use blood test which is simple enough for practitioners to perform much like they would any other (i.e. a routine test). If proved a worthy screening tool following more extensive research, CancerSEEK could become a highly valued addition to any physician’s toolbox.

Accurate positive results could point a physician in the appropriate direction for further screening and testing, potentially taking place far sooner than it can now, in order to make a diagnosis. In this way, for instance, a positive result for stomach cancer could be indicated far earlier, prompting a second screening (i.e. endoscopy) to confirm a diagnosis. A person who may not yet be showing signs and symptoms can then begin appropriate treatment sooner and thus have a higher chance of combatting the disease than many do now.

A large-scale ‘healthy’ population study, at the very least in the tens of thousands, in order to really confirm accurate sensitivity and specificity rates is now expected to be done. In simple terms, CancerSEEK is designed for the general public and thus must be able to be applied in a real-world setting as precisely as possible.

The team is not promising hope for a cure as a direct result of their newly developed test, but they have, in some respects, laid foundations for further testing which has the potential to demonstrate an effective means to save more lives.

CancerSEEK has shown a small step forward in the area of early detection and has much ground to cover before it can be reliable enough to take a successful giant leap forward, not only for 8 types of cancer with high mortality rates and fewer screening options, but for all known forms of the disease. In principle, CancerSEEK has sparked something which could be highly valuable for all cancer types. Further developments will certainly be including additional types of protein biomarkers for just this purpose.

The good news is that the team is already thinking ahead, and research aimed at demonstrating that CancerSEEK can actually be the life-saving screening tool it is designed to be has already begun.

Currently, a 5 year long study has begun whereby close to 50 000 female volunteers between the ages of 65 and 75 have been enrolled. Each participant has never been diagnosed with cancer. The study is estimated to be costing around the $50 million mark, which has been provided as private funding by The Marcus Foundation.

We, the world, wait in anticipation…

The research team published their initial report in Science on 18 January 2018, an online platform for medical news and developments. (3)


1. National Institutes of Health. 29 August 2017. Detecting early signs of cancer in the blood:  [Accessed 24.01.2018]

2. Johns Hopkins Medicine. 18 January 2018. Single Blood Test Screens for Eight Cancer Types:   [Accessed 25.01.2018]

3. Science Mag. 18 January 2018. Detection and localization of surgically resectable cancers with a multi-analyte blood test: [Accessed 24.01.2018]