Skip to main content

Together we are beating cancer

Donate now
  • Science & Technology

Following CUPiD’s arrow: A new blood test to find cancer of unknown primary

Jacob Smith
by Jacob Smith | In depth

23 April 2024

3 comments 3 comments

A medical professional holding a tube containing a blood sample
Olena Yakobchuk - Shutterstock

One of the most nefarious things about cancer is its ability to spread. 

This is also called metastasis, and happens because cancer cells can break away from a tumour and travel in the bloodstream. From there they can settle in another part of the body, where they start to grow a new tumour, making the cancer more difficult to treat. 

But while it might sound confusing, cancer growing in a second part of your body doesn’t mean you then have two types of cancer. 

For example, if someone’s cancer starts in their breast and spreads to their lung, it doesn’t mean that they have lung cancer.  

Instead, it’s called a secondary breast cancer. It’s made of the same cancer cells that were growing in their breast, only now they’re growing in their lung. 

In most cases where a person is diagnosed with a secondary cancer, we can use special imaging tests to find where the cancer originates, known as the primary cancer. 

But in rare instances, doctors can’t tell from any tests where the primary cancer is. They can only see the location of the secondary cancer. 

In these cases, the patient is diagnosed with cancer of unknown primary (CUP). 

Sometimes, doctors are able to identify the primary cancer after diagnosing a person with CUP. But that involves a long road of diagnostic tests for the patient, and unfortunately is only successful in the minority of cases. 

A team of researchers we funded at the Cancer Research UK National Biomarker Centre, part of the University of Manchester, and the Christie NHS Foundation Trust, are working to change that. 

Into the unknown

There are a couple of theories on why the primary cancer is difficult, or even impossible, to find in cases of CUP, but there isn’t a clear answer. 

“Sometimes when you look at the cells under the microscope, you can clearly see it’s a breast cancer, for example. You just can’t find that breast cancer in the breast,” explains Dr Alicia-Marie Conway, an academic clinical lecturer at the University of Manchester. 

“Some people think that there was a primary tumour that has regressed, or the immune system found and destroyed it, but the metastatic sites continue to grow. And we do have evidence of that in tumour types like melanoma. 

“Another theory is that the primary tumour is just difficult to find; it’s remained really small.”  

But whatever the reason, not knowing where a person’s primary tumour is presents a unique set of challenges when it comes to treatment. 

As treatments have developed over the past decade, they’ve become more targeted, meaning they’ll only work on cancers with certain mutations. Often these mutations will be specific to certain types of cancer or will affect one tissue differently to another. 

For example, the drug vemurafenib targets a mutation called BRAF in certain types of melanomas, but this drug is ineffective in treating bowel cancer with the same BRAF mutation. 

And that means targeted treatments, which are likely to be more effective, can’t be used to treat cases of CUP where we can’t find the primary tumour as we can’t be confident that they’ll work. 

What we need is a way of easily identifying where cancer cells first started growing, so we can determine what treatments would work best for them. 

That’s exactly what Conway and her team have been investigating. 

They’ve been exploring how we can use something called DNA methylation to reveal where the primary cancer started growing. And all they need is 2 tablespoons of blood.  

A model representing the structure of DNA
A model of the structure of DNA

What’s DNA methylation?

Methylation is a way that our cells can ‘tag’ DNA.  

Believe it or not, every single one of our cells contains every gene we need to create a human body. And there are around 100,000 of them. 

That means that, in theory, every type of cell has the instructions it needs to become every other type of cell. A skin cell has all the genes it needs to be a liver cell, and vice versa.  

But each type of cell plays a very specific role in our bodies and is highly specialised to perform that role. Our bodies need a way of making sure that each type of cell ends up in the right places.  

It shouldn’t come as a surprise to learn that having skin cells growing in your liver would be a bad thing. 

So, our bodies use DNA methylation to stop that from happening. 

While each cell contains all our genes, which are made up of strings of DNA, they can switch certain ones on or off depending on the jobs they need to do in the body. There are a few different ways that cells can do this, but one of them is DNA methylation.  

To do it, cells stick small molecules called methyl groups onto specific parts of DNA to make a particular pattern.  

That way, when one of your cells comes to read the sequence of your DNA, it can see which genes are switched on and off and knows to ignore those that are switched off. 

And that pattern will be the same in all cells of the same type, like a unique code that can be used to identify them.  

Often, mutations to DNA are not consistent across a cancer type, so it’s difficult to tell what the primary cancer is from that information alone. Methylation pattern, however, is distinct to a tissue type, making it much more useful in CUP. 

That’s a good start, but there’s a catch. 

The problem of scale in cancer of unknown primary

The way we traditionally examine tumour cells, and in these cases, look at the DNA methylation patterns, is by taking a biopsy of the tumour. Biopsies can be invasive for the patient, and often can only collect a small sample of tissue. 

“One of our big challenges currently is that the biopsies that we get from people with CUP are often quite poor quality, and don’t have very much tissue,” Conway explains. 

“And that means we don’t have enough tissue to look for the methylation patterns very easily in these patients’ cells.” 

So, to overcome that problem, the team looked to find an alternative to invasive biopsies, a way to easily take sample from patients but still get the same data needed from the tumour’s DNA. 

And there’s a good candidate for that: liquid biopsies. Or, as they’re more commonly known, blood tests. 

Cancer blood tests 

Interest in how we could use blood tests in cancer research has been growing over the past few years.  

In September 2021, a new blood test that has the potential to detect over 50 kinds of cancer began trials in the NHS, and earlier this year, researchers working on our flagship lung cancer study TRACERx published results showing how we could use a blood test to track the evolution of lung cancer. 

But to recap, let’s go through how they work. 

As a tumour grows, cells die, and new cells grow in their place. These dead cells get broken down and their contents, including their DNA, gets released into the bloodstream.  

These chunks of tumour DNA in the blood are known as circulating tumour DNA (ctDNA). 

Therefore, we can analyse ctDNA that we get from a blood test to determine information about the tumour, like what mutations it carries and, if we’re tracking it over time, how it’s changed. 

But in this case, it’s the methylation patterns that we’re interested in. If we look at the patterns on the pieces of ctDNA we collect in the blood samples, we may be able to identify which tissue or cell type they came from.  

And that would lead us to the site of the primary cancer. What we need is a test that combines finding tumour DNA in a blood test with identifying patterns in methylation and matching them to specific cancer types, or methylation profiling. 

So, Conway, with the team of researchers at the Cancer Research UK National Biomarker Centre, directed by Prof Caroline Dive CBE, have developed one. 

It’s called CUPiD. 

Prostate Cancer cell image taken using a Scanning Electron Microscope

For more cancer news delivered straight to your inbox, subscribe to our newsletter


Classified information

At its core, CUPiD is a machine learning algorithm called a classifier. Classifiers do what they say on the tin, they analyse data and classify them into discrete categories – in this case, into a cancer type. 

That may sound straightforward: take a blood sample, extract the tumour DNA, analyse the methylation, identify the cancer type.  

Unfortunately, it’s a little more complicated than that, for two reasons.  

The first is that the amount of DNA present in a blood sample is extremely small, which doesn’t give researchers a lot to work with. 

The second is that tumour cells aren’t the only cells that release DNA into the bloodstream when they die. Healthy cells do that too. And that means a lot of the free DNA in the bloodstream of a person with cancer isn’t tumour DNA. 

So, we need a way of accurately sifting through a very small amount of DNA to find an even smaller amount of tumour DNA. 

Luckily, that’s exactly the kind of thing CUPiD has been trained for. 

Training camp for CUPiD

CUPiD is actually what’s called an ensemble classifier, which is lots of individual classifiers put together. That allows it to perform multiple analyses at once. But to perform these analyses, classifiers need to be trained on existing datasets.  

In CUPiD’s case, a big existing dataset.  

The computer scientists in the team trained CUPiD using information from The Cancer Genome Atlas, a large scale, freely accessible database containing the genetic information of over 20,000 samples representing 33 cancer types.  

That way, CUPiD learned to recognise pieces of DNA from each type of cancer, and it could then use that information to classify unknown samples it was given. But to be clinically useful, it needs to work on very small amount of DNA. 

So, how did it fare? 

The team tested CUPiD on 170 DNA samples from blood. That included 143 samples from people with cancer, representing 13 different tumour types, and 27 samples from people without cancer. 

And CUPiD correctly identified the tumour type in 121 of those samples, a success rate of 84.6%.  

Importantly, the test had a high accuracy. When CUPiD makes a tumour prediction it is correct 96.8% of the time.  

The researchers then applied CUPiD to blood samples banked from patients with CUP. 

CUPiD made a tumour prediction in 32/41 (78.0%) cases. For 88.5% of cases the primary tumour predicted by CUPiD was consistent with the clinically suspected cancer type or subsequent primary cancer found through other tests. 

Every minute counts

The results the team saw with CUPiD, published last week in Nature Communications, were exciting but we’re still a little way from rolling it out in the clinic. 

This research was what’s called a proof-of-concept test.  

The team have shown that CUPiD can work in the lab using known samples, but we need to make sure that it could identify a wider range of cancer types using samples from real patients.  

The team are now recruiting for a trial to validate their results in a much bigger cohort of real patients with known cancer types as well as testing in a larger group of patients with CUP. 

But while it’s in early stages, a liquid biopsy like CUPiD has the potential to completely change the diagnostic journey of someone with CUP. 

“The beauty of the test in my eyes is that you could introduce it right at the beginning of the diagnostic journey for patients with metastatic cancer,” says Conway. 

“Patients often get a scan that shows multiple lesions that look like cancer, but they don’t see an oncologist until it’s been confirmed where that cancer has come from.”  

“And that sometimes isn’t straightforward. Sometimes patients have multiple biopsies, maybe they’ll need more scans, maybe they’ll have an endoscopy, and still, nobody is quite sure where this cancer came from.  

“So, even if all of those investigations found a primary tumour, the blood test could have done it a lot quicker.” 

Every minute counts when it comes to starting cancer treatment. The sooner a person with cancer can start their treatment, the more likely it is to be successful. 

That’s why developing tests like CUPiD, which could greatly reduce the time it takes to get a primary diagnosis, is so crucial for people with CUP. 

People like Lee. 

Living with cancer of unknown primary: Lee’s story 

Lee was diagnosed with CUP in January 2022, more than 3 months after he first found two lumps and made a GP appointment. After an initial ultrasound, he was referred to his local hospital for investigative tests. 

“I had an MRI scan, a CT scan, chest x-rays, and blood tests and everything was coming back inconclusive,” he says. “So, the last stop was a biopsy on these lumps. 

“The biopsy suggested I had cancer, so I went for a PET scan to confirm that. 

“The doctor said, ‘it’s cancer of unknown primary’. I didn’t understand what that meant, I thought cancer was just cancer. He said it’s incurable, and because of where it is it’s inoperable.” 

After his diagnosis, he was referred to the Christie NHS Foundation Trust in Manchester, where Conway and her team are based, in February 2022. 

“I had more blood tests, and they offered me a clinical trial. They told me to go away and think about it, but I didn’t have to. I wanted to get straight on it, because even if it didn’t help me, it could help people in the future.  

“We wouldn’t be where we are today if people hadn’t gone on trials years ago.” 

Once it was confirmed he was eligible for the trial, Lee started treatment with a new chemotherapy drug in March of that year. After the first 3 sessions, his tumour had shrunk by almost half. When he’d completed the 6 rounds of chemotherapy, he was moved onto a different trial using a less invasive immunotherapy treatment. 

Lee, now 57, knows just how revolutionary a test like CUPiD could be for people with CUP. 

“I’ve had loads of tests, loads of tissues taken and sent everywhere, blood tests, everything,” he says. “And they still can’t find the primary cancer.” 

“There’s no doubt a test like this would’ve made things better for me, but I’m so fortunate. And I’m so grateful for the team at the Christie. If it weren’t for them, I don’t think I would be here today.  

“The prognosis for cancer of unknown primary isn’t very good. I got told I had 9 months, but it’s been 2 years now. 

“Every day is a bonus. You wake up every morning with a smile, because you’ve got another day.” 



  • Hannah Bethany Cooksley
    30 April 2024

    Very interesting. Simplified to make it an easy read. And exciting to see what’s happening in the world of medicine! I am an aspiring medical student, and this has given me something to go away and research in more depth. Thank you so much for all the information! I look forward to reading more articles!

  • David Archer
    23 April 2024

    A very understandable explanation of methylation. Thank you

  • Tracey Featherstone
    23 April 2024

    I have liver cancer they say its treatable

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.


  • Hannah Bethany Cooksley
    30 April 2024

    Very interesting. Simplified to make it an easy read. And exciting to see what’s happening in the world of medicine! I am an aspiring medical student, and this has given me something to go away and research in more depth. Thank you so much for all the information! I look forward to reading more articles!

  • David Archer
    23 April 2024

    A very understandable explanation of methylation. Thank you

  • Tracey Featherstone
    23 April 2024

    I have liver cancer they say its treatable

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.