Original URL: https://www.theregister.com/2014/08/24/cutting_cancer_rates_data_models_and_a_happy_ending/
Cutting cancer rates: Data, models and a happy ending?
How surgery might be making cancer prognoses worse
Posted in Networks, 24th August 2014 08:00 GMT
In theory, the definitive treatment for early stage cancer patients should be surgical resection. Just get in there and cut the tumour out. If the lump is small and well-constrained, then surgical removal is all that should be needed. Unfortunately, this isn’t usually what happens.
Even when the tumour is nice and small and easily got at, patients mostly have to go through the rigours of chemotherapy and/or radiotherapy and all the traumas that these treatments entail. Why? Because for many types of cancer, surgery is often followed by relapse – the disease pops up again, usually somewhere else in the body, even though all traces of it appear to have been removed.
In breast cancer, for example, the pattern of relapse is very clear. When looking at patient data we can see there’s a peak of relapse in the first two to four years, then there’s a lull and then a smaller wave of relapses much later on.
This bimodal relapse pattern occurs not just in breast cancer, but prostate, lung, pancreatic, melanoma and osteosarcoma. And in case any of you data nerds were wondering, this bimodal pattern seems to be pretty robust across different data sets, and appears to be independent of disease stage or other prognostic factors.
All of which begs the question of why cancers should behave this way. The standard model of cancer has assumed that there’s an exponential growth pattern when tumours are small, but then this rate of growth slows as they get bigger – technically, it’s a damped exponential.
If we can treat cancers by cutting them out, why do we need chemotherapy?
So, for a long time the working hypothesis is that even when tumours are small there are already microscopic pockets of disease hiding in the body, and when the primary tumour is cut out these microscopic metastases carry on growing. As they are very small the model says they should be growing very fast, so then you need to whack them with chemotherapy – which is basically a poison targeted at rapidly growing cells. Hence the need to follow surgery with chemo…
However, the problem is that this model suggests that relapse should be related to disease staging. The bigger your tumour, the longer it’s been around, the longer the micrometastases have been growing; hence the faster the relapse. But that’s not what the data says. The data says that it’s the surgical intervention that’s acting as the trigger mechanism.
A number of researchers and doctors (including Michael Retsky, Romano Demicheli, Michael Baum, William Hrushesky, Isaac Gukas and others) across the globe puzzled over this discrepancy when it first became apparent in the 1990s. They worked together to figure out what was going on and to try and tease out a better model of cancer growth. A model which could be used not just to explain the data but, more importantly, to see what could be done in terms of changing therapies.
And this is the key point, this isn’t just an arcane data science or computer modelling issue, it’s not about understanding the molecular biology of cancer, it’s about making a significant impact on treatment. Because it’s not the primary tumour that kills most patients, it’s metastatic disease spreading through the body.
Surgery might not be the answer in every single case
In time a new model has been developed, and rather than tumour dynamics being described by a damped exponential equation, it turned out to be intermittent and dynamically non-random. And in the case of breast cancer, most relapses are the result of a sudden spurt of metastatic growth that is triggered by the surgery itself. In other words, the treatment that should be most curative is kicking off a process that leads to cancer relapse within four years.
Now, funnily enough, this isn’t necessarily a new finding. It turns out that historically doctors in ancient Greece and Rome had also noticed that surgery to remove breast cancers could make things much worse. It also fits in with folk wisdom in certain parts of the world – for example in Africa many women presenting with cancer do not go to conventional oncologists because they fear that surgery would “provoke” the disease.
So, why should surgery have this effect? It turns out that there are probably multiple effects at work. Firstly, surgery itself causes a massive release of chemicals into the bloodstream that signals to the body that there’s been tissue damage and that it needs to repair itself.
This repair includes the sprouting of new blood vessels (a process called angiogenesis) at the site of the wound. Unfortunately, these same factors in the blood also help tiny pockets of cancer cells sprout the blood supply they need to get the nutrients and oxygen to grow.
At the same time the surgery induces a drop in the immune system – so where tiny cancers have been kept in check by the immune system, they are suddenly let off the leash as the body recovers from surgery. The surgery itself may also release tumour cells into the bloodstream, cells that can circulate and find a welcoming home somewhere else in the body.
The breast screening minefield
Our little team of researchers working on this problem suddenly found themselves with explanations for all kinds of anomalous results. For example, the breast cancer data showed that while the bimodal pattern exists for all women, the effect is twice as strong in pre-menopausal women compared to post-menopausal. Here was a possible explanation for the racial disparity in breast cancer outcomes.
Medical science: It's not as bad as Star Trek just yet
African-American women present with cancer at an average age of 46, while for European American women it’s 57. Surgical intervention in both groups would lead to relatively more African American women suffering from relapse simply because they are more likely to be pre-menopausal.
And similar thinking also lead our intrepid group of scientists into one of the most disputed and mendacious areas in medicine – breast screening. For those who think that medicine doesn’t suffer from the poisonous atmosphere that afflicts climate science, for example, think again.
The breast screening controversy is a minefield of disputed data, models and theories. Now the basic idea of screening is a simple one – catch the cancer early and you can intervene before it has spread and it’s too late to do anything. The core dispute revolves around the data.
Critics suggest that some women will be “over-diagnosed” – that is, they will have lumps that are not actually dangerous but, once detected will be treated as though they are invasive cancers. These “over-diagnosed” women will become cancer patients when really they should be no such thing. And both sides of the argument – both the entrenched “screeners” and the opposing “sceptics” – are arguing about the numbers. Will the number of women who benefit from screening exceed the women who will not?
Applying the model to the situation
Into this fray comes this new understanding of tumour dynamics and the bimodal pattern of relapse. It suggests that pre-menopausal women who are treated will have a greater rate of relapse than post-menopausal women. All of which suggests that the screening of 40-49 year old women will have a much lower overall benefit than the screening of 50+ year old women.
Unfortunately this was an unwelcome intervention as far as those entrenched “screeners” were concerned. They were campaigning (successfully, as it happens) for screening to be introduced for all women over the age of 40. Suddenly all research funding for this project dried up…
All was not lost, however
Now it just so happens that this is a story that has a potentially very happy ending. In 2010 a paper was published in an anaesthesiology journal by a group from Brussels. Patrice Forget and colleagues looked at the data for 327 consecutive mastectomy patients to see if there was a relationship between relapse and the pain-killing drugs used during surgery.
It has to be said that this is a pretty unusual thing to do. Mostly anaesthesiologists have a very focused attention span – their job is to keep the patient alive and pain-free during and immediately after surgery. Thankfully this enterprising group took a longer view and came up with a surprising result. Patients treated with the non-steroidal anti-inflammatory drug ketorolac (a cousin of diclofenac, which is an over-the-counter drug in most countries) had significantly fewer instances of relapse compared to other drugs, particularly opiate-based pain killers.
Was this a statistical fluke? Well, it looks like the effect is real. And it may extend to other cancer types, not just breast cancer. This is where we get our happy ending – potentially. It means that a simple change in procedure, switching to an existing and cheap drug like ketorolac, can make a huge impact on the risk of subsequent relapse and death. It almost abolishes that bimodal relapse pattern.
At a stroke the cut of a cancer surgeon’s knife becomes curative again, with the threat of rapid metastatic spread disappearing. If a new drug appeared on the market with that much potential we’d be marching in the streets demanding access to it for our loved ones with cancer.
What stands in the way? The effect needs to be confirmed in a clinical trial. Statistics are one thing, but actually having a randomised controlled trial is still the gold standard in medicine. Thankfully such a trial is going on at the moment. A Belgian charity, the Anticancer Fund, is working with Patrice Forget and colleagues on a trial with breast cancer patients.
If confirmed, then this is a story with a happy ending that saves thousands of lives across the world. ®