Good science measures the outcome of an intervention on a subject and compares that with the outcome of that same intervention on a control. The less identical the subject and control, the less reliable will be the outcome.
An engineer can compare outcomes on identical subjects, which is why your mechanic can give you a money-back guarantee. If the engineer wants to know what happens if you put sawdust in a gas tank insteal of petrol, he can find a reliable result because he can use an identical control.
A doctor doesn’t have 10 copies of any patient (except in twin studies or in the lab) so the conclusions of his science are much less reliable.
When patients are involved, controls are created by randomisation. You take a population and randomise that population into one or more groups.
While this might create two similar groups, they are by no means identical. Nor are any two patients within any of the groups identical. The hope is that any variables that could influence the outcome are distributed evenly between the groups. While you can use design or statistics to control for factors which might influence the outcome, there are many unknowns that cannot be controlled for. Genetics, past trauma, adverse childhood experiences, diet and many others add to the noise which might contaminate the observed outcomes.
Medicine tries to compensate for that noise with a larger number of participants, but numbers onluy increase statistical power, which is only necessary when the clinical effects are small. Would you rather have a treatment if it provided a total cure 10 out of 10 times or 10 out of a 1000 times?
Thus, medicine measures the average outcome of the treatment group and compares this to the average outcome of the control group.
The trouble is that average results can never apply to any individual. Worse, in medical science, it is assumed that once a result reaches statistical significance, not only CAN it be applied to every individual, it SHOULD be applied to everyone. Thus we have the widespread drugs like statins and antihypertensives that benefit less than one percent of people who take them, meaning 99% of people take them for no benefit whatsoever.
This makes no sense. Nobody is “average.” The average individual has one breast and one testicle. In the treatment group, some got better and some got worse, the same in the control group. The average result has absolutely no relevance to any individual yet many doctors will confidently claim that they know the right treatment for you “based on the science.”
This is garbage.
in the same way that individual results may not be generalisable to the group, population-based results may not be specifically applied to the individual.
If we want to get closer to the truth, we have to pay attention to the quality of the control.
Only when the control and the treatment subject is identical in all respects, can a reliable result be achieved. If we want to be certain of the effects of a given treatment, is it better to give the treatment to 20 different people and measure the average result, or give it to 1 person 20 times?
If a patient can act as their own control, then we achieve the gold standard of evidence for that patient. If the results are reasonably consistent, then they would also be generalisable.
To demonstrate this principle, let me show you this patient who had a headache for 27 years and watch as we use the patient as their own control. We can do this as many times as we need, to be sure that the treatment we are proposing will have the desired effect – which it did. Her headache vanished and never returned when the tooth problem was sorted.
or this:
Eventually it becomes a pattern. Of course, we cannot make any assumptions about groups generally, but in these cases, if removing the offending foreign body results in complete resolution of symptoms then we have proven causation in these cases. To be absolutely sure, we would have to reinstate the offending crown, so far none of my patients has wanted to pay for that, and there may be a few ethical issues with that line of enquiry.
This is true patient-centred evidence-based medicine. Patient-centred must mean more than the patient agreeing to the treatment. It must mean that the treatment is tailored to the patient, and this can only be achieved by using the patient as their own control.