Tuesday, May 14, 2013
Science is Hard
Yes it is. Or it certainly can be. Back in Flexner's time and right through mid-Century, obviously, even though we didn't have any high quality randomized trials going on, doctors were doing stuff. Some of it was probably helpful much of the time. For example, they knew to amputate severely injured limbs, especially if there were signs of putrescence. If there's an accessible tumor, cutting it out can be helpful. It it isn't malignant, it's curative. Digitalis was used for heart disease since the 18th Century, and it is indeed helpful. There were other so-called empirical remedies back then as well, by which we mean remedies that appear to work but we don't know why.
Digitalis has survived as a useful treatment, but a lot of what doctors have done routinely for many years has not. In the 1946 National Formulary of the American Pharmaceutical Association, pills containing mercurous chloride were listed as treatment for "biliousness," a condition thought to be caused by insufficient flow of bile and characterized by constipation, headache, and general malaise. Mercury was thought to stimulate the liver; it did definitely counteract constipation, to put it mildly. Of course it is actually poisonous and long-term use of this compound was deleterious indeed.
So why did doctors believe in ineffective or even dangerous remedies? (It wasn't long before this time that they had given up bloodletting.) There are a few reasons.
The most basic is that most conditions that cause discomfort or suffering either get better on their own in a while, or fluctuate in severity. People are most likely to consult doctors when they have symptoms. Whatever nostrums or mumbo jumbo the doctor provides will then likely get credit for the patient shortly feeling better. This is how superstitions generally get started.
Furthermore, similar symptoms may have multiple causes. Even if half the people don't get better after consuming mercury, the treatment will end up getting credit for those who do. It might even really help some people, but harm twice as many. Nevertheless, thanks to confirmation bias, those who believe in it will continue to use it and be persuaded by their observations that it is sometimes effective. (Those it helps + those who get better regardless all redound to its credit; it is presumed unconnected to the harms it causes, because we have no such expectation.)
Another reason is that people just tend to like it when doctors do something, anything. The so-called placebo effect is greatly misunderstood and over-hyped, so I'll steer clear of the term for now. Let's just say that confirmation bias, and perhaps other psychological mechanisms, mean that if people expect to feel better, they will say they feel better and perhaps, in some sense, will feel better. "Feeling better" is, after all , a purely subjective state. I could have exactly the same physical symptoms but be less troubled by them. And our experience of pain is very much affected by how much attention we pay to it. Whatever signals are coming from the peripheral nerves, we may have very different degrees of caring about them. A doctor's kindly ministrations and our presumption that we're going to feel better could be all it takes to make it so, for a while -- even if the cancer is still spreading.
All of the above, in addition to inflicting the practice of licensed, scientifically trained physicians, is of course the foundation of all forms of quackery.
In extreme cases, what we call anecdotal evidence can be quite valid. As a classic example, no-one says we need a randomized controlled trial of parachutes. Everybody knows what will happen, pretty much inevitably, if a person falls from a height of 2 miles. That people usually do it safely using a parachute is all we need to know. The curative power of insulin for people with Type 1 diabetes falls in this category, as does lemon juice for scurvy. Dr. Lind would not actually have needed his various active controls to prove the point. But these cases are rare.
Next time, a bit on the difference between the concepts of science-based medicine and evidence-based medicine.