We will never have enough data, so we must do the best with what we acquire. Systematic reviews and meta-analyses of clinical trials represent one approach to maximizing the usefulness of clinical data. The Cochrane Collaboration thrives on the analyses of the results of tens of thousands of controlled trials in its database. The Collaboration bears the name of the British epidemiologist Archibald L. Cochrane, “Archie” to his friends. Cochrane excelled as an epidemiologist but prevailed with his seminal 1971 book, Effectiveness and Efficacy: Random Reflections on Health Services. In this book he advocated the use of systematic, ongoing reviews of clinical trials as the best single guide to what works in medicine. His influence grew even after his death in 1988 and culminated in the founding of the Cochrane Collaboration in 1993. As he noted in his own obituary, “He was a man with severe porphyria who smoked too much and was without the consolation of a wife, a religious belief, or a merit award—but he didn’t do so badly.”1
One of the most active areas within the Cochrane Collaboration is the Stroke Group. Prof Graeme Hankey, Associate Editor of Stroke for the Pacific Rim, has been an active member of this group. Through his efforts and those of Prof Peter Sandercock from Edinburgh, Stroke has agreed to publish brief summaries of meta-analysis of the Cochrane Collaboration’s Stroke Group. This will provide the readership of Stroke with the main conclusions of relevant analyses and the Cochrane Collaboration with an additional forum. Prof Graeme Hankey will edit this Cochrane Corner. The first summary, by Counsell and Sandercock, is on “Low-Molecular-Weight Heparins or Heparinoids Versus Standard Unfractionated Heparin for Acute Ischemic Stroke” (found on pages 1925–1926 in this issue), a timely and unsettled topic.
Meta-analyses have several advantages: the comprehensive and systematic categorization of data, the increased power derived from pooling patient data from multiple similar trials, the possibility of drawing conclusions in aggregate that cannot be derived from individual studies, and the possibility of displaying large amounts of data in a graphic and easily comprehensible form. Meta-analyses also have disadvantages: bias toward large, English-language, positive trials, since smaller, non-English language, negative results are less likely to be published. The concordance with the results of large randomized clinical trials is only 65% to 90%,2,3⇓ and the conclusions represent the average of individual results, ranging from nil to high treatment responses. Clinical trials address the question “Does it work for most patients?” not “Does it work for this patient?”4
Fortunately, methodological standards for clinical trials are rising, negative results increasingly find their way into the literature, and genetic techniques are helping to characterize treatment responders and nonresponders and to predict and prevent adverse drug reactions.5 No analysis can be better than the information upon which it is based, but categorizing data in a standardized form allows for alternate analyses and interpretation as knowledge grows. Even when reading wrong conclusions, clarity about the facts and transparency of the analyses allow for subsequent correction and redress. As Francis Bacon noted almost half a millennium ago, “Truth is more likely to arise from error than from confusion.”