Here is a riddle for you.
Q . What is the difference between a physician who is reluctant to use placebos and Guernsey McPearson?
A. The physician doesn’t fool sufferers gladly.
Actually, my worst problems are usually with physicians who are all too ready to fool patients… and themselves.
Last time I wrote on the subject of dichotomania. Well, unfortunately, dichotomania is becoming something of a monomania with me: an obsessive-compulsive disorder, no less. No sooner had I finished with Dr Angina Cutter and hypertension then Panacea re-assigned me. I don’t know what is wrong with these people. Clearly they place a very low priority on the quality of my life. (There is some justice in this, since I place a very low priority on ‘Quality of Life’. Indeed, I always say that whenever I have anything to do with it mine suffers.) Can’t they just allow me to draw my salary while sitting quietly in a corner re-reading the complete works of RA Fisher? Apparently not. I was immediately assigned to work on our latest ‘blockbuster’ in the field of depression, Perkupiton â. That was bad enough but imagine my horror on learning that the medic concerned was Dr Violet Shrink. Now, most medics I am assigned to work with I get on with fairly well. We establish pretty early on where the brains in the partnership lie and take it from there. I remember fondly, for example, my collaboration with Dr Durchfall on the Stromboliteâ project, with Dr Richard Rising on Fullagroâ and with Dr Harvey Puffer on Zefferâ. (A tricky project that. Zeffer was designed to improve airflow in the anatomical north but had the side-effect of increasing it in the anatomical south. It helped asthma patients get out more but reduced the number of people who wanted to meet them.) I didn’t mind giving advice on meta-analysis to Dr Percy Vere. Even working with Dr Cutter on Tenseoffâ had a certain je ne sais quoi but I don’t know what and I have to admit that she is very charming. However Dr Shrink is not only illogical but batty and has no sense of humour. Imagine my chagrin at our project initiation meeting when she got in first with the obvious remark.
‘So
‘Roses are red,’ I replied, making the obvious retort.
‘Hm,’ she said, 'a curious comment. Have you heard of KMT?’
‘KMT, an acronym for Knight’s Move Thinking (or Thought) a symptom of schizophrenia, often related to so-called, word-salad and the construction of neologisms’, I replied, ‘no I haven’t. Tell me about it.’
‘Well KMT, or Knight’s Move Thinking, to give it its full name, is a symptom of schizophrenia. The patient goes from one thought to another seemingly unconnected thought, although in the patient’s mind there will be some fantastic link that seems perfectly logical to him.’
‘Or her’, I replied. ‘Have you every heard of repeated measures?’
‘Isn’t that a topic in statistics and if so what is its relation to the current conversation?’ She hauled out her notebook and started scribbling in it furiously.
‘Enough of this banter,’ she resumed after a while. ‘You know how much Sir Lancelot* has said is riding on this project. I am determined to optimize our chances of success and would like to hear your proposals for dealing with placebo responders.’
‘I propose to have a placebo group or at least a control group,’ I replied.
‘I don’t think you understand,’ she replied. ‘ I mean what is your proposal for dealing with the fact that a number of trials fail due to the high response rate in the placebo-group?’
‘I am reminded of DeFinetti’s advice to oil companies who didn’t like uncertain forecasts but preferred hindsightism: don’t drill dry wells.’ Dr Shrink pulled out her notebook again. ‘By all means let us have trials in which the response rate in the placebo group is very low,’ I continued, ‘on average in such trials the treatment effect of Perkupiton compared to placebo will indeed be better than in trials in which the response rate to placebo is very high. In fact, I can confidently predict that in trials in which the response rate to placebo is 100%, Perkupiton will be shown to be no better than placebo.’
‘My point precisely, she replied’ ‘So what do you propose to do about it? Surely this is an issue where statistics can help?’
‘Perhaps. Suppose that we organise a game of dice: medical advisers versus statisticians.’ A knowing gleam came into Dr Shrink’s eye as I said this and she hauled out her notebook. ‘Whichever team gets the highest average score wins’, I said, ‘However, whenever a medical adviser gets a six he or she is disqualified. I think that this will increase the chances of the statisticians winning.’
‘But that’s unfair,’ she said, ‘ no-one would agree to such a procedure.’
‘Quite right,’ I said, ‘unless they were medics engaged in a responder analysis or perhaps comparing trials in which the treatment did well to those in which it did badly’.
‘Are you attempting to make a crude analogy? However, surely we could test all patients in a run-in period. Those that responded to placebo could be excluded. That would be fair’
‘We could,’ I replied. ‘I see three problems. First, we have to assume that the patients are dozy, second we have to assume that the regulatory authorities are asleep, and finally we have to get lucky.’
‘Do you have to speak in riddles?,’ she replied. ‘A bit rich’, I thought, ‘from a member of the profession that gave us the ink-blot test.’
I proceeded to explain. ‘The first problem is to do with what we tell the patients. If they are all given a placebo in the run-in, we have to make them believe that they are not. This is stretching the ethics and we also have to assume that they are too stupid to guess what we are up to. The second problem is that if the regulatory authorities are awake, they might insist that we have to find some way of excluding placebo-responders from being prescribed Perkupiton if and when it gets to market. This is not going to make our marketing men very happy. The third problem is that we have to assume that being a placebo responder, or perhaps more to the point a non-responder, is a condition with high reproducibility. If not, we will have lots of false inclusions and have wasted time and resources to no purpose.’
‘But that’s preposterous,’ she replied. ‘It is a well-known fact that there are a lot of placebo-responders in depression.’
‘It is a well-known fact that some depressed patients given placebo are observed to get better,' I replied, 'but that’s another matter.'
‘Well how do you explain it then?’
I looked out of the window, where the rain was bucketing down. ‘Let me demonstrate,’ I said, ‘the power of incantation. Rain, rain go away. Come again another day. I predict that by some pluvioincantory process at some time within the next ten days my spell will work and the weather will be dryer than it is now only for it to become wetter again at a further interval.’ More furious scribbling in the notebook.
‘Well,’ she continued,’ let me put it in a way that a statistician like you might understand. We plan all our trials with at least 80% and often 90% power don’t we?’
‘Yes.’
‘But in depression only 30% of the trials have a significant treatment effect and that is even having ignored the failed projects. How do you explain that?’
‘Oh that’s easy. The drugs aren’t very good. They have an effect but it is less than the clinically relevant difference.’
‘Well that’s preposterous. Of all the treatments we have for depression, including counselling, transactional analysis, holistic medicine and psychoanalysis, pharmacotherapy is the most successful.’
‘In the country of the blind, the one-eyed man is king.’
‘
‘In that case’, I replied, ‘there is only one thing I am qualified to be.’
‘And that is?’
‘Chief Executive Officer’.
Back to Guernsey McPearson’s Prose Selection