Abstract :
It might seem desirable (especially to trialists) that clinical practice should be determined by the evidence for efficacy of a particular intervention. But we know that real life is more complex and that there are many influences on clinical practice. The relative importance of these is different for different doctors, depending on such factors as age, time from medical qualification, etc.
There have been few formal studies of the influence of trials on practice and most have been related to trials of pharmaceuticals. Until the last decade most drug trials have been too small to yield clear results and so, for a given agent, some studies have yielded positive results, some negative and some equivocal. This causes maximal confusion for the profession and great opportunities for selective quotations in advertisements. Perhaps more importantly it gives the opportunity for selective quotation in promotional meetings organized by pharmaceutical companies. These companies can apparently show clear effects of such promotion on future prescription patterns, but this data is generally unpublished.
Lessons from acute MI
When large trails such as the GISSI and ISIS studies do show clear, concordant, and highly statistically significant results, then large changes in practice do occur, and relatively rapidly. This has been shown clearly for fibrinolytic treatment in acute myocardial infraction (1). But even here Ketley and Woods (2) showed a two-fold difference in the rate of prescriptions of thrombolytic treatment between neighboring hospitals (ranging from 30-60% of acute MI patients). These differences were not explained by the obvious clinical features of the patients, but were closely related to whether a particular hospital had had prior experience of thrombolytic drugs by taking part in a clinical trial. So it is clear that trials may influence not only by their results, but also by the promotional and educational aspects of taking part in a study.
Conversely Lamas et al (3) have shown that adverse trial results cause a reduction in prescriptions of a particular drug. It had been clearly shown that the use of calcium channel blockers in acute MI (particularly nifedipine and diltiazem) was harmful in patients with impaired LV function. In a subsequent trial of an ACE inhibitor in patients with acute MI and impaired LV function (SAVE study), there was a significant reduction in concomitant prescription of calcium blockers following publication of the MDPIT study of Diltiazem post MI. So prescribing can be rational.
On the other hand the use of intravenous betablockers in acute MI is more variable and generally lower than one might expect from the modest but clear reduction in mortality shown in the ISIS-1 (4) and MIAMI (5) trials. This may be because these two drugs are out of patent and not promoted for this purpose by advertising by the manufacturers.
Trials of drugs of hypertension
Hypertension presents particular difficulties for clinical trials. One crucial difficulty is that event rates are low and so the trials demand several years of follow-up as well as very large numbers of subjects. Another difficulty is that after the early studies of diuretics or betablockers versus placebo had shown the benefit of lowering blood pressure, it became unethical to test the newer agents against placebo except in very mild hypertensives as in the TOMHS trial. As a result we have a conflict between mortality and surrogate end-point studies.
Future directions
Providers of health care, whether HMO based or some variation of national health service, now demand hard evidence on which to base changes in practice from older to newer (and generally very much more expensive) remedies. How are they to get such evidence? I believe that there is no substitute for properly conducted randomized clinical trials. If a small proportion of a drugs promotional budget were spent on this, it would be hugely beneficial. Observational data from hospital or other databases is so hugely biased as to be worse than useless.
Secondly, all new trial evidence should be presented in the context of the previous evidence. Systematic overviews can rapidly influence practice and consensus among clinicians. This process is loosely called “evidence-based medicine.” We have recently put out feelers for future overviews in the cardiovascular field-the so-called Cochrane Cardiovascular Collaboration; the initial response has been very enthusiastic. Such regularly updated overviews will be available freely on INTERNET.
