Asymptotic Philosophers

A philosophy of science blog

On failure to find a cure and the disease model of mental illness

Posted on Author KinoPosted in Philosophy of Science, Scattered Plot1 Reply
I read a popular article (linked below) by the philosopher of science/statistics, Jacob Stegenga, on the lack of evidence on the effectiveness of antidepressants in treating depression. He presents and explains a number of major problems plaguing this kind of research: industry-funding, publication bias, low study quality, placebo effect combined with blind-breaking, lack of validity in outcome measurement, the misleading use of odds ratio as a measure of effect size in meta-analyses. Many of these problems have gained a lot of philosophical attention in recent years, usually under the banner of philosophy of medicine or philosophy of psychiatry. There are also two problems, pointed out towards the end of this article, that are more or less specific to antidepressants: 1) there does exist a few high-quality meta-analyses, which have found antidepressants ineffective; 2) the popular theory that SSRIs treat depression because depression is often caused by serotonin imbalance and that seretonin is important in depression because SSRIs are effective is clearly circular, and very probably ungrounded.
Insofar as this article is a critical analysis of scientific methods and the nature of evidence, it’s pretty solid. It’s also very accessible, which is a skill I try to train myself through this blog. In any case, all this is to say that I agree with most of what he says concerning “scientific evidence” proper. However, I’d like to comment on a few points concerning the limitations of “scientific evidence”.
Here’s the article:
Depression is a very complex disorder and we simply have no good evidence that antidepressants help sufferers to improve …

Towards the end of the article, Stegenga affirms the reality of individual experiences, but cautions us against taking experience, however real and legitimate, as evidence. He mentions misattributing natural fluctuation to effect of drug, placebo effect, and confirmation bias as the three main reasons. They are good reasons, and there are probably more.

However, if you have been personally affected by this, you might not find these reasons persuasive. This is exactly the sentiment expressed by Marcus Arvan, who is (among other things) the owner of the Philosophers’ Cocoon blog.

Anti-depressants do work: reply to Stegenga
In a piece at Aeon linked to at Daily Nous, philosopher of science Jacob Stegenga (University of Cambridge) contends that “we simply have no good evidence that antidepressants help sufferers to improve.” I don’t think the evidence available supports Stegenga’s argument …

One of Arvan’s major point of contention is that randomized control trials (RCT) are not the only sensible outcome measure. Importantly, they overlook individual level improvements that would, by all other standards, be considered causally meaningful, but are nevertheless too scarce to show up above background noise in such trials.

I want to explore this tension a little bit further. Suppose a drug is found ineffective: say, 20 people in the placebo group improved and 20 in the drug group improved, out of, say, 10,000 in each group. (Of course, this is unlikely to happen, but let’s suppose that the drug is clearly ineffective by RCT standards.) How do we make sense of the 20 people in the drug group who improved?

Statistically speaking, we cannot tell them apart from placebo improvements or even measurement error. However, for us to claim that “the drug couldn’t possibly have caused the improvements because we have just found the drug to be ineffective” seems both arrogant and circular. But it is true that “we have just found the drug to be ineffective”. How else do you find drugs effective?

To put this puzzle in another way, suppose the same drug works for you but not for me, there seems to be a tension. Why? Why should the drug’s effect on you be the same as on me?

This is where the disease model comes in. (A quick caveat: I’m not a philosopher of medicine, so my use of the term “the disease model” may differ from many other uses.) Very roughly, what I mean by the disease model is the idea that a disease is a specific dysfunction in a specific biological pathway or mechanism, and that an intervention “cures” the disease by causally affecting this dysfunction. Stegenga has used examples of curing scurvy with vitamin C and treating Type 1 diabetes with insulin in his article. He argues that the pessimistic state of the effects of SSRIs suggests that serotonin is not the relevant mechanism of depression.

Under the disease model, it makes perfect sense why we should expect the same drug to have the same effect on different people. If you and I have the same disease with the same dysfunction, then a drug either is a causal agent on that dysfunction or is not. If it is, we should both find it beneficial. If it is not, then the person who improved must have benefitted from something else unrelated to the drug.

My skepticism is over whether we should hold on to this framework at all. In earlier days, the disease model was invoked primarily to argue against stigmatization. Mental health related stigmatization and discrimination may still occur today, unfortunately, and the disease model may still be good for that purpose. However, the DSM-V’s grand vision of moving away from the symptomatic model of diagnosis to a causal model has failed, suggesting that we still don’t have good evidence that the disease model is the right way of understanding mental illness.

Of course, all this isn’t to say that mental illness isn’t biological or doesn’t have biological components. Understanding depression as a “condition” or “syndrome” does not preclude the possibility that some, or many, people will have the same biological manifestation of this condition that can be “cured” by the same intervention. What it does not demand is that if two people share symptoms, they also share a cause and they also share a cure. The process of [cause -> symptoms -> cure], where a cause produces a set of symptoms, which allows us to find a cure that removes the cause and consequently the symptoms, is especially unreliable in mental illnesses. It was this unreliability that motivated the DSM to move to (and stay at) the prototypical model, where illnesses are classified by the clustering of symptoms. While the current DSM is, by all means, not the final picture, I haven’t encountered any obvious reason why this model is wrong-headed. Of course, this doesn’t mean that I have good reasons to believe this model to be the right one.

The problem is this. There is a certain level of generality and “well-behavedness” that is necessary for something to be studied scientifically (at least with current scientific methods). This may be a limitation of science, or a misconception of what it is we are trying to study, or an irreconcilable fact of nature. We don’t know. However, until this tension is concretely reconciled, we must remember that, while science measures experience and guides it accordingly, experience should always take precedence over science

Kino
Latest posts by Kino (see all)

1 comment

  1. If we allow that similar symptom profiles can be caused by different dysfunctions, then I think the disease model can make sense of the idea that a person with (say) ‘depression’ improve after taking a particular drug while another person with ‘depression’ doesn’t improve. Really, these people have ‘depression_1’ and ‘depression_2’, which should be classified based on their causal etiology, not their symptomatic realizations. On this line of thought, it’s an empirical question how many different varieties of mental illness there are, but our successful interventions are a reliable guide for helping to answer this question.

Comments are closed.