 |
 |
Paul, I absolutely agree. But its helpful to remember that there are 2 different purposes to evidence hierachies: (A) a searching guide and (B) grading of evidence. The first is used by EBMers to answer specific clinical questions arising in the care of patients, whereas the latter occurs in guideline writing. As a searching guide it works something like this: A question about treatment arises in caring for a patient, and I have (only) a couple of minutes to find the best current evidence. 1. I check if there is a systematic review (I'd use PubMed:Clinical Queries as it picks up the Cochrane and non-Cochrane reviews). If I find a systematic review, I still need to check the quality is OK. 2. If there is no systematic review, then I look for individual RCTs (and focus on the largest well-conducted study first). If an RCT exists, I then need to appraise the quality. If its no good (e.g, a small and flawed study) or there are no RCTs then I'd move on to the next level, 3. If there are no RCTs, then I'd check if there are any cohort studies. If a cohort study exists, .... etc The evidence "hierarchy" is used as a search heuristic, with alternation between finding evidence at a particular level and checking the quality of what is found. A guideline writing (who is hopefully also an EBMer, but that seems pretty rare) would hopefully work through the same process, but also give a grade to the best evidence found. As the guideline writer will have more time they might also check the next level of evidence down the hierarchy. Like all heuristics, it is murky but makes doing bedside EBM possible. And better search heuristics would be great. Suggestions anyone? Cheers Paul Glasziou At 09/07/2005, Dr Paul Montgomery wrote: > The problem I have with these sorts of evidence hierarchies is that once > you go below an RCT, how to prioritise? It seems to me that depending on > the nature of the question posed, it may be that a cohort study with > perhaps historical controls could actually be better than a sloppy > case-control study where the controls were contemporaneous. And as we > progress down the hierarchy, things seem to get murkier and > murkier............. > > Or am I missing something? > > Paul > > Dr Paul Montgomery > Departmental Lecturer in Evidence Based Intervention > University of Oxford > Barnett House > Wellington Square > Oxford > OX1 2ER Paul Glasziou Department of Primary Health Care & Director, Centre for Evidence-Based Practice, Oxford ph: 44-1865-227055 www.cebm.net