Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Observational Studies Require More Than Just High Standards!
- Guy Lloyd (23 October 2007)
|
|
|||
|
Guy Lloyd, Consultant Cardiologist Eastbourne DGH. BN21 2 UD
Send response to journal:
|
The BMJ has published a welcome set of standard criteria to judge prospective observational studies (1). What however has not been included is the government health warning which should accompany all such publications. While the accompanying leader article highlights the undoubted usefulness of this form of study to examine rare diseases, and in particularly has stressed the importance of good methodology (2); unfortunately it is not in rare and unusual diseases where such studies are usually performed. Rather, they are most often used to examine common illnesses such as cardiovascular disease and cancer for which end points are frequent and the prevalence high enough to make the randomised controlled trial overwhelmingly the more reliable form of assessment. While the quoted examples appear to be proof enough of the importance of the cohort, a list from cardiology alone demonstrates that the results of observational studies have often been seriously flawed. The cardioprotective effects of female sex hormones; the usefulness of antioxidants or homocystenine lowering strategies; rhythm control for atrial fibrillation or the advantages of dual over single chamber pacing are all examples of where the observational studies have suggested a clear treatment effect and greatly influenced practice. The subsequent randomised trials have however refuted each hypothesis. The problem in nearly all such cases is the ‘fellow traveller’ effects of many adverse (or beneficial) factors that tend interact and for which only a proportion can be statically accounted. Fat people tend on average to do less exercise, have a high saturated fat intake (of which a higher proportion will be trans-fats) while smoking and not attending to their insulin therapy or taking their blood pressure tablets; all of this is because of their poor education, as a result of their lower socio- economic status resulting in a high level of lifetime stress because of low feelings of social control. Fair enough this is an extreme example but does illustrate just how interdependent many of these factors are, and if cohorts are considered in isolation, they can result in findings that are basically wrong. This would not be a problem if it was generally accepted that cohorts studies were hypothesis generating and therefore should not be acted upon until a suitable randomised trial has been conducted, if feasible (clearly an RCT of smoking would have some ethical difficulties). Glasziou et al in an excellent recent article suggested that a rates ratio of at least ten and a p<0.01, combined, should be required before the ‘signal’ from a true effect can be accepted as being clearly different from background population ‘noise’(3). Very few of our current favourite targets for example mild obesity, salt intake or dare I say even passive smoking would pass this test. The findings of these studies should usually start rather than close the debate. Experts are too hasty to present a hypothesis, dressed in the clothing of a proven fact, it makes for good copy, and the medical profession is far too willing to accept such findings uncritically. 1. von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP; STROBE Initiative.Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007 Oct 20;335(7624):806-8. 2. Rothwell PM, Bhatia M.Reporting of observational studies. BMJ. 2007 Oct 20;335(7624):783-4 3. Glasziou P, Chalmers I, Rawlins M, McCulloch P When are randomised trials unnecessary? Picking signal from noise. BMJ 2007;334:349-351 Competing interests: None declared |
|||