Biomarkers and the price of innovation in rare diseases

Let’s talk about biomarkers. Biomarkers are molecules found in the body, usually identified in the bloodstream or urine, that are the telltale traces of internal processes unfolding deep in our organs and cells. Different levels of these usually harmless byproducts can be used to flag up occasions when a patients symptoms may be something worth investigating, or alternatively rule out a given condition. Classic examples are human chorionic gonadotrophin (HCG), which is the biomarker measured in pregnancy tests, or blood glucose, which can indicate diabetes.

A couple of years ago, an American teenager called Jack Andraka won a major youth science prize by developing a blood test for a pancreatic cancer biomarker he had identified, mesothelin. Although, to my knowledge, the story hasn’t really moved on since then, it has recently started cropping up all over my newsfeeds. While it is an exceptional piece of work for any teenager to be doing, there are a couple of reasons to be cautious in our enthusiasm.

In cancer, where specific cell types multiply and overactivate, we can see that some of the biomarkers associated with processes carried out by those particular cells are produced at much higher levels than normal. The problem, then, is that healthy people also produce these markers in varying levels. The presence of a biomarker doesn’t, by itself, indicate anything at all. To overcome this, tests only appear as positive if a certain, abnormal level is measured, indicated by the reference range, the range of levels that commonly occur in a healthy population.

Mesothelin, as any quick Google search will tell you, is a biomarker for a large number of cancers, including pancreatic. It can also be found at elevated levels in healthy people. So far, Jack appears to have done a proof of concept study - his technology is capable of detecting accurately the presence and absence of mesothelin in blood samples from mice, and it seems to extend as far as being able to detect the excess found in mice with pancreatic tumours. However, there are a couple of challenges, particularly in the case of pancreatic cancer, that seriously limit the likelihood of this technology ever being used for screening or early detection; the stated goal.

One of these is that the evidence just isn’t there yet. Forbes has a good article on what evidence is missing, and also breaks down some of the claims surrounding cost and time savings.

But say this test does turn out to be as sensitive as promised, even in real people? The problem is that sensitivity is only half the challenge. The greater challenge in rare diseases like pancreatic cancer is specificity - how well the test detects the absence of disease. The aim is for this test to detect people before they become symptomatic, but people could have elevated mesothelin levels in their urine for any number of reasons including reduced kidney function, which controls which molecules even make it into the urine. Previous studies evaluating mesothelin have found that it is elevated in pancreatic cancer, benign pancreatic tumours, ovarian cancer and mesothelioma - those are all bad, so maybe that’s ok. More concerningly, at least one study has found overlapping levels in serum between subjects who are healthy and those who have pancreatic cancer.

So it’s probably reasonable to assume that this test won’t be perfectly specific. Let’s assume that for every 100 healthy people, the test wrongly assumes two of them have cancer. In 2012, 8900 new pancreatic cancer cases were diagnosed, and 96% are in the over-50s. If we screen everyone, taking the crude rate of 13.8 cases per 100,000, and apply it to the adult population of the UK of around 51 million, then we will detect around 7000 cases of cancer, assuming sensitivity is perfect and we catch everyone. However, from the remaining 50,993,000, we also end up with 1,019,860 false positives - healthy people who have elevated mesothelin levels for any number of other reasons, the majority of which are highly likely to be entirely harmless.

Even if people would rather take the risk of a false positive, with the associated stress, imaging and potentially even surgery, which of course carries its own risks, then the pressure on hospitals and oncology units of trying to identify who is actually sick would be catastrophic. It isn’t much of a stretch to think that the 1 patient in every 145 who is actually sick might be missed in the follow up, eliminating the value of the exercise entirely.

So this, then, is the challenge with trying to screen for cancers using biomarkers, and why screening is in fact becoming less common, not more. Perhaps the most worrying aspect of this development is the fact that Jack is reportedly hoping to launch an over-the-counter test, meaning that people would be able to show up to their doctor’s office with what looks like a positive test result for pancreatic cancer. What does the doctor do in that situation? The best case scenario is they triage based on likelihood, maybe suggest monitoring, or further investigation if the symptoms match up, with all the associated cost and burden on the system, in order to identify that rare patient who is actually ill. The worst case is that the patient believes, probably incorrectly, that they have pancreatic cancer. How do you live with that news, and what is the cost to that person’s mental health of having to live with that diagnosis, even if it is wrong? How does a doctor attempt to reassure someone, and minimise the number of invasive procedures carried out, each of which has a real risk of doing harm?

Innovation is a brilliant thing, and Jack Andraka is clearly a highly intelligent, driven person of the type that is likely to do a lot of good in the world. I absolutely applaud him. But I find the idea of an over-the-counter test for pancreatic cancer using a single biomarker genuinely frightening.