Portal de Periódicos da CAPES

Clinical Evidence Epistemology

2007; Elsevier BV; Volume: 127; Issue: 11 Linguagem: Inglês

10.1038/sj.jid.5700894

1523-1747

Robert P. Dellavalle, Scott Freeman, Hywel C Williams,

Clinical practice guidelines implementation

randomized controlled trials TO THE EDITOR We write to respond to points raised by Dr Rees in the recent editorial “The Nature of Clinical Evidence: Floating Currencies Rather than Gold Standards” (Rees, 2007Rees J. The nature of clinical evidence: floating currencies rather than gold standards.J Invest Dermatol. 2007; 127: 499-500Google Scholar). We agree with Dr Rees that change is the only constant (Heraclites, c. 475 BCHeraclites (c. 475 BC) Wikiquote. Available at: http://en.wikiquote.org/wiki/Heraclitus. Accessed 23 March 2007Google Scholar) in clinical evidence, that clinical decisions should be based on the best available evidence (Collier et al., 2006Collier A.P. Johnson K.R. Dellavalle R.P. Williams H.C. The Cochrane Skin Group: promoting the best evidence.J Cut Med and Surg. 2006; 9: 324-331Google Scholar), and that randomized controlled trials (RCTs) are not always necessary (Glasziou et al., 2007Glasziou P. Chalmers I. Rawlins M. McCulloch P. When are randomised trials unnecessary? Picking signal from noise.BMJ. 2007; 334: 349-351Google Scholar). Furthermore, the results of single RCTs, even when published in highly prestigious journals, can be hazardous when read in isolation (Ioannides, 2005Ioannides J.P.A. Contradicted and initially stronger effects in highly cited clinical research.JAMA. 2005; 294: 218-228Google Scholar), highlighting the need for systematic reviews of all available RCTs. Rees points out that RCTs should not be blindly regarded as a “fixed” gold standard. We agree. While RCTs offer the strongest design to minimize bias in human intervention studies, their quality (in terms of design, conduct, analysis, reporting, and relevance) is highly variable, furthering the argument for systematic reviews to draw attention to such multidimensional components (Bigby, 2003Bigby M. The hierarchy of evidence.in: Williams H. Bigby M. Diepgen T. Herxheimer A. Naldi L. Rzany B. Evidence-based dermatology. BMJ Publishing Group, London2003: 44-48Google Scholar; Williams, 2003Williams H.C. How to critically appraise a study reporting effectiveness of an intervention.in: Williams H.C. Bigby M. Diepgen T. Herxheimer A. Naldi L. Rzany B. Evidence-based dermatology. BMJ Publishing Group, London2003: 56-63Google Scholar). Because systematic reviews are the most cited form of clinical literature (Patsopoulos et al., 2005Patsopoulos N.A. Analatos A.A. Ioannidis J.P.A. Relative citation impact of various study designs in the health sciences.JAMA. 2005; 293: 2362-2366Google Scholar), if one accepts citation impact as a measure of importance, we disagree that the importance of systematic reviews has “never been subject to experimental study” (Rees, 2007Rees J. The nature of clinical evidence: floating currencies rather than gold standards.J Invest Dermatol. 2007; 127: 499-500Google Scholar). Systematic reviews of all available evidence are also the main source of evidence used by the UK National Institute of Health and Clinical Excellence (http://www.nice.org.uk/) and the US Agency for Healthcare Research and Quality (http://www.ahrq.gov/) to inform health policy. We agree with Rees that there is increasing awareness that “failure to report studies may lead to error” (Rees, 2007Rees J. The nature of clinical evidence: floating currencies rather than gold standards.J Invest Dermatol. 2007; 127: 499-500Google Scholar). For example, the concealment of serious cardiovascular adverse events for rofecoxib made a significant difference to the positive light thrown on the drug from other trials at that time (Krumholz et al., 2007Krumholz H.M. Ross J.S. Presler A.H. Egilman D.S. What have we learned from Vioxx?.BMJ. 2007; 334: 120-123Google Scholar). When the truth of the real risks emerged, the drug was withdrawn. Systematic reviews serve to foster such enlightenment. We agree with Rees' description of P-values as indicating false positive error rates when used in the context of elementary frequentist statistical theory. We still maintain, however, that one of the advantages of systematic reviews is that they can improve the precision of estimates derived from smaller similar trials, thus making lots of additional trials unnecessary when an acceptable level of false positive error has been reached. To date, 14 placebo-controlled trials of topical pimecrolimus for atopic eczema have been conducted – do we really need any more to convince the clinical community that pimecrolimus is effective when compared to vehicle? A more serious example is the use of intravenous streptokinase as thrombotic therapy for myocardial infarction. By performing a cumulative meta-analysis of similar RCTs, Lau et al., 1992Lau J. Antman E.M. Jimenez-Silva J. Kupelnick B. Mostellar F. Chalmers T.C. Cumulative meta-analysis of therapeutic trials for myocardial infarction.N Eng J Med. 1992; 327: 248-254Google Scholar showed clear and consistent benefit for streptokinase after just eight RCTs involving 2,432 patients had been performed in 1973. Yet by 1988, a further 25 RCTs involving another 35,542 patients had been performed in isolation addressing the same question without altering the risk estimate, but making the overall P-value very small indeed (P<0.0001). Such an observation illustrates the human cost of failing to observe the need to bring all evidence together within the body of a systematic reviews, as opposed to wasting valuable patient and doctor resources in continuing to perform one inconclusive trial after another. There comes a point when the likelihood of a false positive results becomes very remote, and even the most skeptical person would say “enough is enough”. Dr Dellavalle and Dr Williams are editor and the co-ordinating editor, respectively, for the Cochrane Collaboration Skin Group. RPD and SRF were supported by National Institutes of Health (Bethesda, MD) Grant nos. CA92550 and T32 AR07411, respectively, the University of Colorado Cancer Center (RPD), the Dermatoepidemiology Research Unit of the Department of Dermatology (RPD and SRF), and the UK NHS R&D programme (HCW).