Furuium Clin Phurmucol zyxwvutsrqpo 1997; zyxwvutsrq 1 1 zyxwvutsrqponm :48 zyxwvutsr 1493 zyxwvutsrq 0 zyxwvutsrq Elsevier, Paris Review Standardization of terminology in meta-analysis: a proposal for working definitions P Nony, M Cucherat, MC Haugh, JP Boissel Clinicul Pharmacology zyxwvuts Unit, Curdiovascular Hospital. 69003 Lyon, France (Received 2 April 1997; revised 26 May 1997; accepted 10 September 1997) Summary - In this paper we present a brief overview of the growing concern to standardize definitions and terminology in meta-analysis. This tool ha? become inescapable in both drug research and therapeutic evaluation. The performed and published meta-analyses are increasing. as well as the variation in the meaning of the terms used in meta-analysis. In the second part of the paper we propose glossary of the most common terms used in reports of meta-analyses. The glossary has been written by only one group of scientists, the definitions are therefore p ~ > p o ~ e d to the scientific community as zyxwvutsrqp working definirions, to be subject to reactions from leaders in meta-analysis. meta-analysis /glossary / clinical trials OVERVIEW We believe there is a need to establish a common understanding of the terminology which is the most frequently used in meta-analysis. A basis for such a glossary is proposed in this review. However, one can wonder whether such a task can be achieved by only one group, or should be performed through a co-operative work, expressing the opinion of differ- ent investigators. This is the reason why the defini- tions given below are only proposed to the scientific community as working definitions, to be submitted to reactions from leaders in the field of meta-analy- sis. Some evidence argues for a growing concern to harmonize definitions and terms used in meta-anal- ysis, as discussed in the following sub-headings. Meta-analysis has become an inescapable tool in drug research and therapeutic evaluation Meta-analysis corresponds to a set of methods for analysing and interpreting the data collected from several unrelated clinical trials (D’ Agostino and Weintraub, 1995). The resulting information has a level of precision that could not have been possible using data from just one trial. Unlike general reviews, which zyxwvutsr are purely narrative, a meta-analysis can gen- erate quantified results and verify their coherence. The principle of a meta-analysis can be summar- ized as follows: each trial provides a certain amount of data that can be divided into a common core, reflecting the treatment effect, and a specific part to each trial representing the trial specificity. In a meta- analysis the core data are pooled using an appropri- ate technique to give a mean estimation of the treat- ment effect, and the specific data are separated. The mean estimation provides two kinds of information: first, a qualitative result of the expected treatment effect, and second, a quantitative result on the mag- nitude of this effect. The heterogeneity will be deter- mined from the differences in the specific data. Meta-analysis is now considered as the modern form of the general review. In specific diseases or health states with well-established classes of ther- apy and defined end-points, meta-analysis can be performed at different levels: same drug in the same dosage, same drug in different dosages, same basic effect (eg, blockade of platelet cyclo-oxyge- nase), same mechanism of action (eg, anti-aggrega- tory effect), various interventions for the same pur- pose (eg, lipid lowering by diet, drugs or surgery). Regulatory agencies are just beginning to consider the licensing of therapies on the basis of meta-analy- sis, and it can be anticipated that they will be more and more confronted with this question. Another important use of meta-analysis the near future will