Drug Information Journal - March 2009 - (Page 171) BIOSTATISTICS 171 Trends in Meta-analysis While meta-analysis is a well-accepted tool in evidence-based medicine, both the science and the utility of meta-analysis continue to evolve in response to the empirical demands of health care providers, researchers, payers, and policymakers. Three important developments in metaanalysis are summarized, with relevant examples provided for illustration. First, the indications for performing metaanalyses using aggregate data versus individual patient data are discussed. Second, the advantages of cumulating data in real time as new studies are finished, that is, cumulative metaanalysis, are reviewed. Third, we describe the use of meta-analyses to provide indirect comparisons of various interventions when no head-to-head trials exist. Network meta-analyses, in particular, are a valid way to rank order the efficacy or safety of multiple interventions simultaneously. Those who use clinical research syntheses should appreciate these highly relevant developments in the field of meta-analysis, developments that hold great promise for all who wish to use information better in health care. Susan D. Ross, MD, FRCPC SDRoss Consulting, Cohasset, Massachusetts Key Words Meta-analysis; Evidencebased medicine Correspondence Address Susan D. Ross, MD, SDRoss Consulting, 720 Jerusalem Rd., Cohasset, MA 02025 (email: Sdross720@gmail.com). INTRODUCTION Meta-analysis, a statistical method of combining data from multiple sources, has been used for many decades in the social sciences, environmental science, and agriculture. In the early 1990s, meta-analysis also started to gain traction in medicine as a valid and useful way to combine the results of clinical studies. Today it is a well-accepted and necessary tool to produce the best available evidence for evidence-based decisions in health care (1–4). The field continues to evolve in several fascinating ways in response to the empirical demands of health care providers, researchers, payers, and policymakers. This article from the cofounder of MetaWorks, an original AHRQ Evidence-Based Practice Center, describes a few of the more important developments in this regard. It is aimed at nonstatistician audiences who nevertheless are frequent consumers of these kinds of evidence syntheses. Meta-analysis is only one step near the end of the long journey of a systematic review of clinical studies. The systematic formulation of a research question, identification and retrieval of appropriate studies to help answer that question, critical appraisal of each of those studies, data extraction, and database development all precede any effort to synthesize data, whether that effort is qualitative or quantitative. Once the statistical syntheses are done, the results still need to be presented in an intelligible and defensible way, with exploration of strengths and weaknesses, and, most important, the clinical meaning of it all. To keep the meta-analysis activity in proper perspective, it is helpful to see data like those presented by Allen and Olkin (5), where the statistical analysis activities of several typical systematic reviews averaged approximately 10–15% of the total time spent on the review. Regardless of the amount of time spent on a meta-analysis, the usefulness of the information generated will vary widely, from negligible to huge, depending upon the particular question posed and the overall amount and grade of evidence available to answer that question. Several recent developments in meta-analysis promise to expand its influence across the board. These developments have been driven by the empirical needs of various decision makers, framed in question and answer format below. I N D I V I D U A L PAT I E N T D ATA V E R S U S A G G R E G AT E D ATA Q. Do we need to have individual patient data (IPD) from all studies, or are aggregate data good enough when we combine studies? A. For first-order effects, aggregate data are sufficient. We know this from studies like that of SteinSubmitted for publication: March 10, 2008 Accepted for publication: July 22, 2008 Drug Information Journal, Vol. 43, pp. 171–176, 2009 • 0092-8615/2009 Printed in the USA. All rights reserved. Copyright © 2009 Drug Information Association, Inc.
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.