The Problem with Evidence-Based Medicine...

Today I have had to put yet another patient with mild osteoarthritis, a Mrs Smith (of course), onto an unnecessary prescription for non-steroidal anti-inflammatories. It’s not the first time. I’ve made a list of about 60 people who will be coming in at some point to have this same consultation. I’m only seen about 8 of them so far. Ah well, only 52 to go.

The reason these patients (and I am proud still to have patients, rather than clients, customers or prescribing units) are having to surf the wave of renal failure and gastric ulceration is that the Worcestershire Area Prescribing Committee don’t believe them when they say their pain is better when they take glucosamine. The PCT pharmaceutical advisors (who, I hope it’s not too creepy to say, usually do a great job) have looked at “the evidence”, have weighed it in the balance and found it wanting. Apparently glucosamine is no better than placebo at dealing with pain from osteoarthritis. We know that. The scientists have told us. So there.

The trouble is, I think they’re wrong.

Now before anyone tries to burn me at the stake for heresy, I would backtrack a little and say that I don’t think they are completely wrong. In fact, I think they have spoken the truth and nothing but the truth. But it isn’t the whole truth. I believe that when you look at whole populations of hundreds or thousands or people with mild osteoarthritis and give, fully blinded, half of them placebo and half of them glucosamine, there is not an appreciable difference in the outcome measures for each group. I believe that there will be a significant effect even in the placebo group, and that the mechanism of this effect is not fully understood. None of that is up for debate. I just don’t think that much of it is relevant to Mrs Smith.

The thing is, Mrs Smith isn’t a population of hundreds or thousands of people. Mrs Smith is a patient (there is that wonderful word again) with mild to moderate osteoarthritis. There are several important differences between individual patients and populations, and I don’t think the Area Prescribing Committee fully appreciate them.

Firstly, Mrs Smith’s choice isn’t: take glucosamine or take placebo. She has to choose between glucosamine and nothing. Now there has been a lot written about this argument, and I don’t intend to rehearse it here, except to say that glucosamine has few side effects, and even fewer serious ones. So by withdrawing what may be a placebo effect, we are forcing her to have pain or medication which is much more dangerous in the long run, both for her and for the purse string holders of the NHS.

Secondly (and again this is a common argument that has no need of review here) is that the placebo effect is a real effect. It has value. It has special value when the alternatives are much more dangerous. Furthermore, we can take advantage of it without duping Mrs Smith. We can be upfront about the fact that the effect she is having may be the placebo effect, relieving her pain through a complex pathway based in the biochemistry of the psyche rather than the spinal pain pathways. We can explain we don’t understand it fully. In short, we can explain that the label "psychology" is acceptable shorthand for "biochemistry we don't yet understand." And we can believe her if she tells us it’s working.

However, the most important difference between Mrs Smith and a thousand Mrs Smiths is the realisation that one is a model and the other is reality. There have been many occasions through my career when I have seen people mix up these two things, and Evidence Based Medicine has become the cause celebre of the mistake in recent years. Until EBM took the crown, the most obvious example was in teaching The Consultation through the famous models well known to any GP Training Scheme graduate. There we would be, on the video screen in front of our peers, sweat on our brow and palms dripping trying to make every consultation fit so-and-so’s model, in so doing completely missing the fact that the patient has already had a very satisfactory consultation thank-you-very-much, which has covered all ground it needs to and is complete without revisiting the patient’s dog’s birth history so we can tick the “holistic approach” box. We were making the classic mistake. In our minds the model had become the goal and if we didn't twist reality to fit the model, we had in some way failed.

It wasn’t until I thought about the linguistics of this that I finally understood it. The term “model” is not an accident. What is a model? It is a simplification of something complex, whose purpose is to aid the understanding of that complex entity by focussing the attention on just that part which is being studied. There is a great example of a model in the crypt of St Paul’s Cathedral. It is a wonderfully constructed replica of the entire cathedral made in dark wood, which has not been further coloured. It is usually surrounded by tourists who often find that looking at it helps them understand the proportions of the building (for instance, the way the dome relates to the lower sections) more easily than they could if they stood in the street and looked upwards. On the other hand it is a fair guess that it has never been surrounded by architecture students who wished to know the effect of the white light reflected from the marble façade on the brilliance of contrast of the decoration on the dome. It is, after all, made of wood and kept in a very dark room. So this model, this simplification, helps us understand one aspect of the reality at the expense of complete ignorance of another. All models do the same.

The reality of what goes on between a doctor and a patient is many times more complex than St Paul’s Cathedral, and yet both doctors and patients want to understand it. It is only by understanding it that it feels like a controllable and useful tool. And to aid our understanding or it, it has been modelled. A large part of the success of an interaction between a doctor and a patient comes from the medical knowledge of the doctor. That knowledge is gained from textbooks, journals, colleagues etc, but ultimately, where possible, from randomised controlled trials of treatments. These gold standard trials are spoken of in hushed tones when they are well designed and show an important causation, and there is a slight flinch of mistrust when one discovers that a particular piece of information came from anything less rigorous. They are so well regarded, in fact, that we seem to have forgotten what they were invented for.

And this is my heresy: that doctors, in our arrogance, believe that with enough study everything can be predicted and understood. We have forgotten that the randomised controlled trial is just a model of a much more important reality, and that reality is our individual patients, their illnesses and how we interact with them to seek improvement. I am not trying to take merit away from the RCT. It is still the best tool we have for trying to predict what is going to happen to a particular individual given a particular treatment. But like a craftsman working with tools which are wonderful constructions in themselves, we need to add the final tier of quality control, which is a critical inspection of the reality produced by those tools. It is that reality, after all, which will generate a response in the world.

And we need to do this quickly. We are living in a time when politicians and healthcare managers believe that medicine is a discipline which can be squeezed into protocols and delivered by… well, sometimes it feels as though they think it can be delivered by anyone who can read. At least part of the blame for this stems from the lie encouraged by EBM, that you can measure the outcome of a consultation by its performance against predictions from a statistical study. Until these priorities are returned to their original positions, then the balance between scientia and caritas will continue to be elusive, and Mrs Smith will continue to have to buy her glucosamine over the counter, or risk a stomach ulcer.