Main content

Medical detection dogs

Can dogs smell cancer? Dr Mark Porter investigates the evidence for whether canine super noses can be used to accurately detect cancer.

Can dogs smell cancer? Ever since Hippocrates the odour of disease has been used to aid diagnosis but has this simple technique been forgotten? Dr Mark Porter investigates the evidence for whether canine super noses can be used to accurately detect cancer. There have been plenty of anecdotes reported but what about hard science? Studies since 2004 from the Medical Detection Dogs Centre in Milton Keynes have shown convincing results and they've now teamed up with MIT in the US, specialists in 'e-noses'. Could devices the size of a mobile phone be used to sniff for disease?

Available now

28 minutes

Programme Transcript - Inside Health

Downloaded from www.bbc.co.uk/radio4

 

THE ATTACHED TRANSCRIPT WAS TYPED FROM A RECORDING AND NOT COPIED FROM AN ORIGINAL SCRIPT.  BECAUSE OF THE RISK OF MISHEARING AND THE DIFFICULTY IN SOME CASES OF IDENTIFYING INDIVIDUAL SPEAKERS, THE BBC CANNOT VOUCH FOR ITS COMPLETE ACCURACY.

 

 

INSIDE HEALTH – Programme 1.

 

TX:  02.01.18  2100–2130

 

PRESENTER:  MARK PORTER

 

PRODUCER:  ERIKA WRIGHT

 

 

Porter

Hello and welcome to a new series of Inside Health.   Over the next three months we’ll be turning our forensic eye to all things evidence based in healthcare.  From the return of rickets, the so called English disease, to how computers are improving prosthetic limbs.  But today, in keeping with the festive season, we start with a riddle.  What do you think links these three?

 

Music – Beethoven’s Fifth – An Endoscopy clip – Dogs barking

 

Some dogs, Beethoven’s Fifth and me having an endoscopy.  Well the link, oddly is, smell – specifically the smell of diseases like cancer.  And if it’s not clear why, don’t worry, all will be revealed over the next half hour.

 

And, like all good stories, this one starts way back in the mists of time.

 

Barr

The ancients and perhaps the most commonly known ancient would be Hippocrates discussed the use of smell in detecting disease.

 

Porter

Professor Hugh Barr is a surgeon with a special interest in how smells might aid diagnosis, something he shares with the father of modern medicine.

 

Barr

They cultivated an art of understanding what certain diseases smelt like and they were able, therefore, to say to people whether they had a certain type of infection or another type of infection.  And in fact he made a very good insight for tuberculosis because tuberculosis was a big problem in those days and there were other infections that gave you a cough and sputum.  And Hippocrates said that if you heated the sputum, in fact if you cast it on hot coals, and it liberated a particular smell then the patient would die and it was usually TB.  That insight has informed how we detect the volatile from TB in the 21st century.

 

Porter

By volatile you mean this is the bit Hippocrates was smelling?

 

Barr

Yeah, the smell from hot sputum, if it’s TB, is very distinctive.  Whereas if it’s something else like pneumonia caused by other bacteria it is less distinctive.

 

Porter

Now of course in Hippocrates’ day, we’re talking – this is before the advent of microbiology, we don’t have all the x-rays, all the different scanning techniques we have.  As technology in medicine has advanced have we forgotten smell?

 

Barr

You’re absolutely right.

 

Dogs barking

 

Porter

Visitors receive a warm welcome at the Medical Detection Dog Centre near Milton Keynes.  Researches here were the first to publish clear evidence that dogs can identify people with cancer.  In 2004 the British Medical Journal published the team’s study showing that trained dogs could pick out the urine samples of people with bladder cancer.  The dogs were not infallible, but, on average, they got the diagnosis right in four out of 10 cases – way higher than might occur by chance alone.  And in recent years the charity’s research has broadened with ongoing studies into other cancers, including prostate and breast.

 

Guest

Stig, come on.

 

Porter

Claire Guest is the charity’s Chief Executive.

 

Guest

So the BMJ study showed for the first time in the world that dogs could be trained to find human cancer, so these aren’t dogs that are anecdotally finding it on their owners, these are dogs that were going out and being trained to find them on samples and urine samples.  And this was revolutionary, I mean for many, many centuries medics have talked about the odour of disease but it had never been really established and in fact there had been in very many cases it had been completely forgotten as part of the diagnostic process.  Dogs are biosensors, you know whether you like dogs or you don’t they’ve got fluffy coats, waggy tails but they have this fantastic biosensor – their nose – can detect down to parts per trillion, a teaspoon of sugar in two Olympic swimming pools.  If you think that our odour, our own unique odour, we know all humans have can be changed by a disease process, you can easily see how the dog, as a biosensor, could tell us something very, very new about disease.  And this first study it opened that door again, the first robust study that was done in the world for indicating that dogs could be trained to find human bladder cancer with a reliability far higher than chance.  In this particular study the dogs – they never come in contact with the patient, they sniff small urine samples, so a point five of a mil of urine.  Training them – using samples where actually knew the diagnosis at that point.  So some of the patients sadly had cancer, others were healthy, others had other complicated diseases but they were believed to be cancer free.  Making the assumption that some of the volatiles, the smelly compounds, would have got into the urine and that the dog would be able to detect them.

 

Basher, are you going to do some work?

 

Duggett

Hi, my name’s Mark Duggett, I’m the bio detection trainer coordinator at Medical Detection Dogs.  I’ve got Alexa here and we’re about to run a training session.  We’ve got some dog treats for the dog as a reward for when she’s working.

 

Guest

We have five stands and on each stand there’s a metal arm, at the end of the metal arm you’ll see that there’s a sample and these are the samples that the dog’s going to sniff.  And if the dog finds the disease she’s been trained to find she’ll stop at the stand.

 

Duggett

And then I’ll click the clicker and that just communicates to her that she’s done the right thing.

 

Indication?

 

Guest

Correct.

 

Duggett

Good girl.  So she’s gone along there to position four, so she sniffed one, two and three, decided they weren’t the correct ones, got to position four and stopped and indicated.  Some more treats for her.

 

Guest

After the BMJ study in terms of the work that we were doing with the dogs there was a sort of real silence, there was nothing but a wall of scepticism.  The cancer work could have stopped but one of my trained dogs, Daisy, she was working on a prostate cancer study, she lives with me, she’s a very, very accurate prostate cancer detector, started to behave a little bit differently round me, she looked a bit wary.  I took her out one day for a walk, she was in the back of my car, I lifted the boot up to let her out of the car and she wouldn’t jump out, she kept on jumping into my face with big brown eyes and then sort of nudging at me.  Daisy, what on earth’s the matter, just off you go and she kept doing it, kept jumping, jumping, jumping very intently.  But I sort of felt where she’d bumped into me and I thought, funny I can feel a bit of a deep bruise, felt it for a couple of days and then thought I could feel a lump.  To cut a long story short I was referred and after a number of tests I was diagnosed with a very early grain stage breast cancer. I had a lumpectomy, lymph node removal and radiotherapy and I’m here to tell the tale over five years later.  And that of course just then reignited me in the fact that my life’s been saved because of my dog.  The dogs have got information that will save thousands of lives in the future.

 

McCartney

I’ve heard lots of anecdotes about this over the years and there probably is something going on here.

 

Porter

Inside Health’s resident sceptic, Glasgow GP Dr Margaret McCartney

 

McCartney

I have a friend who was certain that her dog knew before she did that she was pregnant.  And there have been lots of doctors writing into medical journals, famously in 1989 some doctors wrote in and said that a dog, they thought, had detected that a patient had skin cancer and seemed to insist on the person doing something about it and was only satisfied when the person had that skin tumour excised.

 

Porter

That’s a bit like the anecdote we’ve just heard and the dog wouldn’t know whether there was a breast lump there…

 

McCartney

Completely.

 

Porter

But the dog may have sensed something.

 

McCartney

Yes, that’s absolutely right but what we really want to do is investigate this scientifically and that’s of course what people have gone on to try and do.

 

Guest

The evidence has been building around the world.  We published again in 2011 a more detailed study looking at bladder cancer and the reliability of dogs over a bigger sample size.  And there have been a number of published studies with very, very promising results looking at prostate cancer, colorectal cancers, lung cancers.  A lot of these studies are very small, small number of dogs, relatively small number of samples and what we knew we needed to do next was a much, much, much bigger study over a much larger sample size in order that we can really look at the evidence base for a number of dogs working over a number of years.  And that’s currently where we are at the moment, we’re working with Milton Keynes University Hospital looking at a much bigger group, we’ve currently already recruited a thousand patients, hope to recruit up to 3,000.  Looking at the dogs’ ability to detect the cancer but also looking at longitudinally how the dogs are able to pick it up over a number of years.  So is the dog able to pick it up before the clinician.  And we have about three years remaining.  Early results are looking promising, we’ve got a long way to go but certainly by the end of this we’ll have established the reliability of dogs over a much bigger group.

 

McCartney

So an ideal medical test would be one that was really accurate, could pick up a problem that could be treated easily before it develops into a big harder to treat problem and something that’s non-invasive and cheap to do, so that’s your kind of ideal characteristics.  There have been several studies done trying to get dogs to detect people with known breast cancer and lung cancer, compared with normal control people, people who don’t have any known cancer.  And some studies have shown that the dogs are quite good at this.  No study has shown that the dogs are as good as current medical practice.

 

Porter

The other issue of course with dogs, as much as I love dogs, it’s the practicality of how you would use a dog in a clinical setting.

 

McCartney

Well indeed, you know you could certainly take samples, which is the way these studies have been done, you take some kind of samples either tissue samples or breath samples that people with suspected lung cancer have exhaled and you use a kind of room separate from your patients to get the dogs to try and determine whether or not there’s an abnormality in them.  But the problem is they’re not as good as current tests.  And we know that all medical testing comes with a false positive and a false negative rate.  So for example one study in Germany that was done looking at whether or not dogs could detect lung cancer found a 7% false positive rate, so telling people they had a problem 7% of the time when they didn’t.  But more worryingly were missing around a third of the people who really did have lung cancer.  So that simply isn’t good enough to use.  But what they did say was that some dogs were better than others, so it could be that the dogs are sensing some kind of chemical that could be done in a different way without the variability that different animals might be subjected to.

 

Porter

And that’s the exciting bit about this, isn’t it, that it opens up possibly a new horizon for identifications that we can apply technology to.

 

McCartney

Yes, so if this works at all it’s not magic, there’s something scientific going on, there’s some chemical that the dogs are detecting that humans are not able to and that in turn would make you think well what is that chemical, could it be detected in a more reliable way, is there a machine that could do that for you.

 

Mershin

My name is Andreas Mershin, I’m a research scientist at the MIT Center for Bits and Atoms.  And one of the things that we do is we make machines, bio-electronic noses, that mimic a human or a dog or another animal’s sense of scent.

 

Porter

Researchers at Medical Detection Dogs have teamed up with Massachusetts Institute of Technology – and, yes, you heard right – Andreas Mershin’s lab really is in something called the Center for Bits and Atoms and they’re hoping to use their canine experience to  help develop an electronic nose.

 

Mershin

The dogs, in some cases, were able to generalise to other cancers that they weren’t trained for and as far as we can tell none of the molecules that are volatile coming from let’s say a skin cancer tumour versus a lung cancer tumour, none of the molecules that we could identify were the same yet the dogs were able to generalise and say both those things were cancers somehow.  We’re still not quite sure how they do it but it looks more and more like it’s something that rides on top of the molecules, something like a cancery melody, a cancery theme that they’re detecting.  If you think of a melody that you can recognise, let’s say Beethoven’s Fifth Symphony – ba, ba, bum etc.  I didn’t even sing it right, I didn’t have the right tempo or the right pitch but I kind of primed you and if you know Beethoven you will recognise this, even though none of the frequencies, none of the pitches, none of the tempos would be the same.  This is what seems to be operating in the olfactory detection of cancers also.

 

Guest

We know now that the technology is now starting to learn, it’s learning like a dog’s nose would learn, we don’t think it’s as simple as hearing a few notes and a bar of music and saying that those notes would be identical for each person.  One person’s cancer may smell slightly different to another person’s cancer and that’s where the dog comes in because he has this incredible ability to match patterns, he’s not looking for absolutes like the machines do at the moment, he can match the patterns and say that pattern to me is the same as that pattern, I can make an indication.

 

Barr

Cancer is molecularly very different from ordinary cells and it does liberate volatile organic compounds.  You will perhaps know the example of the dogs who can smell, black spots on people’s skin, to detect melanoma.  Now although it sounds surprising in fact that black area does liberate a different smell from the surrounding skin and it has a specific signature.  So we can train systems to try and interrogate that.

 

Mershin

We were asked by the Department of Defence research arm, called DARPA, to create a machine that will beat the dogs’ limit of detection for several sets.  And they insisted, we were forced to do this, and then we found out that our machines were working way better than what we expected and we’re still trying to figure out how exactly they’re working, although we know they’re working really, really great.  And that’s where the collaboration with dogs comes into play.  The machines that we’ve built they don’t get programmed, so much as they have to get trained, the same way that dogs get trained.

 

Porter

But how do you put olfactory receptors into a machine?

 

Mershin

That’s also very interesting.  What happens is we have sequenced the DNA of various animals that have a nose, including the human, the rat, the mouse and that means we have a series of letters – A, D, G, Cs – they stand for the nucleic acids.  It’s important to note that we don’t ever take any real living dogs or any animals.  And the series of letters codes for protein that we can then design your gene on the computer, it’s all zeroes and ones now, you send it off to a company that sends you back synthetic DNA that has never seen anything living and then that synthetic DNA gets processed with commercially available kits into making a properly folded protein that then you immobilise on top of an electronic circuit and the electronic circuit’s job is to recognise when that protein is being activated by an odorant, an odorant is a fancy name for a molecule that carries scent.

 

Porter

So you are actually creating an electronic nose, a true electronic nose that can smell rather than just detect the odd signature.

 

Mershin

Yeah the project was called Real Nose for that precise reason.  Just like when you sniff coffee for instance you don’t count molecules, you don’t say – mmm I have 800 of this kind of molecule and 44,000 of that kind of molecule – you just say – mmm coffee.  There’s different characters of coffee, there’s different intensities of coffee but you’re not counting molecules, any more than you would be counting pixels if you were looking at painting or counting frequencies if you’re listening to a song.

 

Music

 

If the dogs have this capacity to generalise and look at this emergent property that, for lack of a better word, we can call cancery, and that the humans have the capacity to recognise something Beethoveny versus something Guns and Roseny – okay?  So since you’re not looking at any specific frequency of sound when you’re trying to tell the difference between Beethoven and Guns and Roses but you’re looking at an overall feel of it, well we expect the enoses to be capable of doing the same thing.

 

Porter

That is uncanny, I’ve just written down here Beethoven versus Guns and Roses.

 

Music

 

Great minds think alike, obviously, at least when it comes to analogies.

 

Now, fair warning, you might want to turn the volume down for this next bit.  No.  No more Guns N Roses – worse than that, it is Professor Hugh Barr examining my stomach a few years ago with an endoscope looking for damage to my gullet.

 

Barr

Right it’s in now, that’s looking fine.  We’re in the oesophagus, the food pipe now, moving down, it looks perfect, it looks perfect, it looks perfect, absolutely fine, just breathe.  Are you alright Mark?

 

Porter

I tell you what I’m never going to be a sword swallower.

 

But these days he doesn’t just look for trouble, he sniffs for it too using a sniffing endoscope.

 

Barr

It’s exactly what animals and dogs do.  What do they do when they meet a pal?  They go up and sniff, it may be in an unsavoury area.  But if we go down with the endoscope and we are inspecting somebody’s intestine we push air down so that we can inflate the organ to look at its lining.  But we suck the air out also.  The device, the enose, is outside the patient, it’s a little box attached to a computer, it can be attached to a mobile phone, you can put sensors inline to detect the smell and say well wait a minute, there’s something fishy here.  And it highlights the electronic nose and says well actually maybe there is a bacterial infection here and we can get a rapid diagnosis and put them on very specific antibiotic therapy.

 

Porter

It’s looking for what?  Bacterial infection only or can you look for suspicious malignancy as well – cancers?

 

Barr

The aim is to look for cancers.  Bacterial infection, I know this may sound odd, is quite straightforward because that is exactly what they do.  The bacteria talk to each other by liberating smell and taste.  What we’ve done in some areas is we’ve had to get the volatile to be liberated, much as Hippocrates says – look if you want to smell TB chuck it on a hot coal.  So similarly we are using the acquisition of the smell is done by sniffing but we have to generate the smell in sufficient quantities that it overwhelms any other smell that would be there.  For example, in the upper stomach we can find people who are infected with a little organism that lives there and if we give people a little bit of a drink beforehand, just a small amount that feeds the organism, it goes hungry, talks to each other, liberates this volatile, it can go down and smell and say well that’s here.

 

Porter

And that’s presumably helicobacter pylori isn’t it?

 

Barr

Exactly, you’re quite right Mark.  Helicobacter pylori, yes we can certainly sniff it and that really matters because that is strongly associated with gastric stomach cancer and this is now recognised very strongly by the medical device industry as a way of near patient testing, giving the patient information straightaway and acting appropriately.

 

Porter

And as you say this is near patient, so you’re not waiting for the results to come back four, five days, a week or so later, you, as the clinician, looking at the lesion there can say right this is H pylori infection, this patient needs treatment.

 

Barr

Yes, that is the idea.  Test, treat, reassured, no waiting for the dread letter from the consultant saying I’m afraid I’ve found this.

 

McCartney

There’s a big leap between sniffing an infection that’s linked to cancer and sniffing a cancer itself.  Now helicobacter definitely is linked to gastric cancer but most people with helicobacter will not go on to develop gastric cancer.  So you have to be quite clear what you’re testing for and what the impact of having a positive test for that would be.

 

Porter

But even if the technology could advance so that you could detect something that was being released from abnormal tissue that was pre-cancerous or indeed turning cancerous it still wouldn’t be a one stop shop would it, I mean obviously when people have got an infection we can diagnose it, treat them and send them on their way as a day case but you’re not going to be doing that with cancer.

 

McCartney

No this opens the door into quite a lot of dilemmas and paradoxes.  So the question is, first of all, are you using this as a diagnostic test for someone who’s already got symptoms or are you using this as a screening test where someone doesn’t have symptoms but they want to pick up on a problem that they don’t yet know about or have any symptoms of themselves.  And the statistics for each are quite different and quite complicated.  And the bottom line is you can end up with an awful lot of false positive tests, where you think there’s a problem but there isn’t, and you may not have effect of intervention to offer at that point that will have a useful impact on that person’s quality or quantity of life.

 

Porter

What’s your impression of this technology so far as an overview?

 

McCartney

So there’s obviously something going on, the question is what is it?  Can it be used reliably, could it be as good as or better than what we’re currently using and that’s going to come through a process of science trying to refine what’s already known and trying to make it better.  We know that lots and lots of things happen in medicine through serendipitous discovery, people following their noses, and I think this was a good example of that.

 

Porter

Literally in this case.  And what they’re all hoping to find is a way of picking up cancers earlier than current techniques.

 

Guest

We all know that early detection of cancer is key.  The very, very interesting thing that we’ve discovered with the dogs is that it doesn’t seem to make any difference to them what stage the cancer is, they seem as able or perhaps even at times more able to find an early stage cancer than a late stage.  And we believe this is because by the time somebody becomes very unwell with cancer there is perhaps a smell of decay much, much later, their body is reacting in a whole range of other ways and it’s a bit like having to peer through a fog in order to see what’s causing it.  But when the cancer’s growing rapidly as a new cancer against a healthy background the dog is able to see it very clearly.  So the challenge is to find out what that pattern is so that the dog can assist in early diagnosis.  If the dog could only help in late diagnosis it would be helpful but it wouldn’t improve survival rates but the indication is that it’s the early cancers they can help us with and that’s going to be the key to early detection through a non-invasive means.

 

Mershin

We do know that the dogs currently can detect early cancers, some early cancers, earlier than any other hospital test.  And this is important because for some cancers, specifically some skin cancers, if you detect it and go to the dermatologist within the first six months of a mole turning into a melanoma, let’s say, then the dermatologist can remove that lesion, half hour procedure, you go home.  If you wait six months often times that’s a death sentence.

 

Barr

We need a very big database of different stages of cancer and we don’t want to smell advanced cancer because we want to get these things very, very early, so we don’t want to smell putrefaction when the cancer is decaying and ulcerating and it’s got bacterial infections on it.

 

Porter

That actually you’re detecting it, very clever of you, but it makes little or no difference to the patient and it’s about impact.

 

Barr

We want to screen patients, they can go along and say look actually I’ve got a family history of this, that and the other, I wonder if I’m not feeling quite so well, they can go and have a test and it’ll say look no actually you’re alright. 

 

Porter

If the technology goes to plan can you envisage a day when someone might be screened solely on their breath rather than you having to put a scope down?

 

Barr

Yes, that would be the aim.

 

McCartney

If we could get better tests that would be ideal, that’s what we want, we want better accuracy, we want to have less false positives, we want to have less false negatives, we want to have less invasive tests – things that are easier and cheaper to do.  All those things are good but you really have to have the data that shows that first.  And I think it’s very unlikely that we’re going to be moving far away from needing a biopsy, for example, to look at what you’ve got under the microscope, to do testing for example for hormone receptors on tissue that would guide your treatment.  I mean there’s unlimited possibilities for how much better that could get.  But right now I think this would only ever be a small part of a much bigger picture.

 

Mershin

In making it into an electronic machine you can actually do some things that the dogs cannot do.  For instance, in the case of a machine you can even have this work under water or let’s say inside your bloodstream so that it can detect things that are not necessarily in the same phase as an odorant that’s volatile and flying through the air.  So it’s generalizable in a different way and it can be more powerful because it can operate outside the normal range of living things.  Also the fact that the dogs aren’t connected to the cloud, they can’t transmit their data to other dogs, they can’t really learn from each other very easily, they have to be trained individually.

 

Porter

How do you assemble a database of what your new nose is going to look for?

 

Mershin

If we deploy these kinds of noses inside of smart phones, and the technology’s almost there, the prototypes that we have now are about the size of an iPhone and they keep shrinking, the idea is to deploy a large number of these that will smell generalised population and then we can kind of back out what happened after a large enough number of people have been carrying these smart phones equipped with noses around for a while and then statistically you’ll start seeing some people developing various diseases.  If we manage to collect all this data into the cloud and look back we’ll start being able to mine it and look for correlations.  And the dream is there that with even a few cases here and there if one nose detects something, another one detects something similar but somewhere else we can be able to generalise and teach all the noses that didn’t have the experience to learn from the ones that did. 

 

Guest

So for some diseases I believe the dog will remain the biosensor of choice.  MPs going into the Houses of Parliament know that it’s the spaniels that have been into the House of Commons looking for explosives, why do we use the dog in that situation, well because it’s quick, it’s reliable, it can screen an area very, very accurately.  I believe for some diseases that will remain the case, the dog will remain the biosensor of choice.  But for cancer I think the electronic nose would be the end result, the end goal, because the number of patients that would need to be screened by cancer and taking these dogs into medical environments would probably not be feasible.  But it could well be that the dog will lead us to the answer, lead us to the electronic system.  I suppose what frustrates me at the moment is that while there’s a huge interest in the electronic noses many people are missing the fact that the dogs have the answers currently and with the dogs help we can get to that solution much more cheaply, much more easily and much more quickly.

 

Sunny, what have you got?  Come here.  Good boy.  Very good.

 

Porter

Claire Guest from the Medical Detection Dogs Centre.  And, as always, you can find links to more background information on the Inside Health page of the Radio 4 website.  Where you’ll also find details of how to get in touch.  So do email us if you have a health issue that you would like clarified.  It is your chance to help set our agenda.

 

And talking of agendas, next week I will be investigating the confusing world of milk allergies in children and Margaret meets the robots working in one of the world’s most advanced pharmacies.

ENDS

Broadcasts

Podcast