[Skip to Content]
[Skip to Content Landing]
Citations 0
March 1, 2000

Evaluating Clinical Studies of Drug Efficacy

Author Affiliations

Phil B.FontanarosaMD, Deputy EditorIndividualAuthorStephen J.LurieMD, PhD, Fishbein FellowIndividualAuthor

JAMA. 2000;283(9):1139-1140. doi:10.1001/jama.283.9.1137

To the Editor: The article by Dr McAlister and colleagues1 raises some interesting issues in evaluating clinical trials of drug efficacy. Although we agree that the best evidence for efficacy comes from head-to-head randomized controlled trials (RCTs), we think that observational studies can provide valuable results in a comparison of drugs and that this class of study should not merit the lowest weight. McAlister et al may be unaware of new developments in design, such as case-crossover studies2 and the use of propensity scores,3 which overcome some of the threats to validity. In fact, the whole purpose of propensity scores is to put observational data into an RCT framework. Moreover, we think that direct comparisons of clinical outcomes should carry more weight than the use of surrogate measures. For example, the effect of statins in preventing cardiovascular events may be due to improving lipid profiles or lowering of blood pressure or both.4

If we agree that a hierarchy of research studies is warranted, it is not clear how it would be developed. Should the evidence be weighted according to the number of studies, the number in each study, the representativeness of the study population, or some other factor? How many observational studies are equal to an RCT? Does one take the quality of the sample size of the study into account?

Observational studies are necessary to monitor drug safety. Randomized controlled trials are rarely carried out with a prior hypothesis relating to an adverse drug event. In fact, the assessment of adverse drug events is almost entirely dependent on observational data of a large number of subjects followed up over a long period.5

Finally, we were concerned that the clinician opted for the cheapest statin in the absence of any evidence of efficacy, while the policy maker was more cautious (and more scientific, in our view) in opting for the statin with proven efficacy. If we were the patient, we would side with the policy maker.

McAlister  FALaupacis  AWells  GASackett  DLfor the Evidence-Based Medicine Working Group, Users' guides to the medical literature, XIX: applying clinical trial results, B: guidelines for determining whether a drug is exerting (more than) a class effect.  JAMA. 1999;282:1371-1377.Google Scholar
Maclure  M The case-crossover design: a method for studying transient effects on the risk of acute events.  Am J Epidemiol. 1991;133:144-153.Google Scholar
Rubin  DB Estimating causal effects from large data sets using propensity scores.  Ann Intern Med. 1997;127:757-763.Google Scholar
Pierdomenico  SDBucci  ACostantini  FCuccurullo  FMezetti  A Additional antihypertensive effect of fluvastatin in hypertensive and hypercholesterolemic patients treated with trandolapril.  Am J Hypertens. 1999;12:A126.Google Scholar
Walker  AMStampfer  MJ Observational studies of drug safety.  Lancet. 1996;348:489.Google Scholar