There is a thread at the moment entitled I hate EBM stemming from one student's frustration at the varied evidence base for the management of stroke and stroke risk.
Perhaps I am just scraping the bottom of the barrel for material, but here is my contribution to this thread:
As I have written previously (somewhere in a comment on someone's blog - can't find it now), consultants and senior registrars are often very knowledgeable and up-to-date, but their teaching skills can be limited which means that medical students don't quite get the concept that the consultant is trying to convey.
There is a reason that medicine is sometimes considered more an art than a science - EBM is great for telling you what is the statistically best treatment for a large cohort of patients (e.g. male 60-75 year olds with SBP between 140-180), but it is very bad at telling you what to do with the patient that is sitting in front of you right now (e.g. female 76 year old with SBP of 138, major renal impairment, dilated cardiomyopathy and bedbound in a nursing home).
Look carefully at the inclusion and exclusion criteria of most of the RCTs that you review: these make or break a study and can define whether they obtain a statistically significant result or not. You may well find that the patient sitting in front of you does not strictly fit into the inclusion criteria of the trial, or on subgroup analysis failed to reach statistical significance in that study. Perhaps they have qualities that bring them in line with two conflicting studies.
Just because you are armed with one paper to support your point of view does not mean you are right, or that you have treated the patient's problem - all you have done is answered the question of whether to prescribe drugoxidil or pharmacoxib, or whether to perform a serum rhubarb level vs a radio-labelled hairy cell scan.
Two of my favourite quotes:
- Metanalyses are to Analyses as Metaphysics is to Physics.
- Doctors treat patients, not statistics.