Hat tip goes to Epiphenom who wrote about this article on Friday. Dr David Gorski has also written a piece on this paper for Science Based Medicine here.
We’ve been talking a bit about evidence based medicine here at Beyond the Short Coat. In the Hard Conversation posts (1, 2, 3, 4, guidelines here), we’ve noticed a lot of dissenters bring up alternative treatments. Invariably, these treatments aren’t evidence based, and don’t work. They generally are “treatments” that seem to be “safe”, but often have no basic science basis. The classic example is homeopathic treatments – which are generally just sugar water. Sometimes people bring up more controversial, less benign things – chelation therapy comes to mind.
So why in the world would people use a treatment like that? Classically we talk about placebo effect: some people improve if you just tell them you’re giving them strong medicine. Occasionally I bring up the fact that complicated diseases have variable courses. That’s not the whole story though, and in this paper Tanaka, Kendal, and Laland, show us another side of the issue.
The authors created a simple mathematical model to simulate the spread of possible treatments in a population. This isn’t a model of doctor dispensed or recommended treatments. This model is specifically about “self medicating” – more similar to over the counter meds, and traditional therapies, things with a pretty low barrier to access. They start out with a population, everyone is either sick (“diseased state”), or healthy(“healthy state”). People who are healthy can become sick and people who are sick can become healthy. Then they introduce a “treatment”. This treatment can either be adaptive (it works), neutral (it doesn’t work), or maladaptive (it makes things worse).
They assume that sick people will expose others to this treatment, and that people adopt the treatment at a constant rate of the time they “see” the treatment. Notice, I said nothing about the treatment working or not. The authors assumed people had no way of telling if a treatment works or not. Don’t be insulted, it’s a pretty reasonable assumption. In fact that’s the reason we have to do evidence based medicine – people generally can’t tell if something actually worked, or if they just got better anyway. They also assume that the longer a treatment doesn’t work, the more likely a sick person is to abandon the treatment.
This model in hand, the authors can ask a lot of questions. They can test a variety of situations. What if the disease is short lived, and rapidly gets better on its own? What if it never gets better? How about if people can catch the disease multiple times? What if the treatment works really well? What if it makes things worse? What if people only expose others to the treatment when they’re sick? What if they expose others to the treatment forever? How fast do people abandon a treatment that doesn’t work?
Their results are interesting to say the least. By their model If people can only get the disease once, the disease is relatively short lived, and only show people the treatment to others when sick, then treatments that don’t work, or hurt you, spread better than effective treatments! Why? Bad treatments give longer “exposure times”, which lets more people pick up on the treatment.
Now if you flip it – if the disease is long lived, and if people spread treatments all the time instead of just when sick, this model tends to select effective treatments as more dominant.
If you can get sick more than once – that too trends towards effective treatments over the long haul.
Interestingly, their model shows that prophylactic treatments spread badly, because they don’t offer a lot of opportunity to get “converts.”
So how much stock can we put in this model? How closely does it resemble real life? Across the board, it’s limited. Not enough of medicine fits in with their initial assumptions. While they discuss CAM in the paper, this model seems to resemble a lower tech, more “traditional medicine” approach. So it may model the spread of modalities in the past better than in the age of the internet.
One situation that does seem to work is flu season – it’s an illness that resolves quickly, and generally the flu “cures” are only used when you have the flu. Admittedly, one can get the flu every year since there are different strains every year. The situation fits the model pretty well actually – there are all manner of non-efficacious flu treatments out there!
How about our topic of choice recently – vaccines? Vaccines are generally pretty easy to come by – yes technically a doctor should be involved, but outreach is pretty high, and the barrier to vaccination is low. It’s a prophylactic treatment, so once you’ve gotten vaccinated you aren’t showing the whole world the joy of vaccination while you’re sick – because you don’t get sick! Well that actually fits the model pretty well, we have to do all kinds of things to get people vaccinated, and it’s not easy!
Where does this model breakdown? The easy one is chronic disease. This model predicts that long duration of disease leans away from bad treatments. Yet those patients who have chronic diseases will often try a wide variety of non-efficacious treatments. I think the inherent assumptions don’t fit chronic disease well though. In chronic disease, rather than pick therapies based on time exposure, one is more likely to actively seek out therapies. Additionally I think there are psychological components that are a little too subtle for this model to take into account.
One obvious issue is that this model doesn’t take into account the ways doctors can spread good practices. Nor does it give us a model through which other “thought leaders” can spread practices. I’m thinking of course of the celebrities – Like She Who Shall Not Be Named of the vaccine denialists. These are powerful forces today.
I think this model is an interesting counterpoint to our regular ongoing discussion here. Regularly commenters bring up “treatments” that have “helped children recover”, in an effort to spread their particular brand of “medicine”. They never cite evidence, they never see the need for it. I find it comforting to think that a model assuming patients know nothing about treatments can lead to this behavior. In my mind that means that maybe, just maybe, science based medicine can make a dent in this situation through the appropriate education.
Tanaka, M., Kendal, J., & Laland, K. (2009). From Traditional Medicine to Witchcraft: Why Medical Treatments Are Not Always Efficacious PLoS ONE, 4 (4) DOI: 10.1371/journal.pone.0005192