• English
  • Español

Introduction to meta-analysis 2: Effect measures for dichotomous outcomes

Topic category Statistical methods and meta-analysis
Date and Location
Date: 
Friday 21 October 2011 - 16:00 - 17:30
Location: 
Methods Group
Methods Group: 
Statistical Methods Group
Contact person
Contact person: 
Georgia Salanti (Contact this person)
Facilitators
First nameLast nameAffiliation and Country
First name: 
Dimitris
Last name: 
Mavridis
Affiliation and Country: 
University of Ioannina School of Medicine, Ioannina, Greece
First name: 
Areti Angeliki
Last name: 
Veroniki
Affiliation and Country: 
University of Ioannina School of Medicine, Ioannina, Greece
Other contributors
First nameLast nameAffiliation and Country
First name: 
Georgia
Last name: 
Salanti
Affiliation and Country: 
University of Ioannina School of Medicine, Ioannina, Greece
Target audience
Target audience: 
review authors
Is your workshop restricted to a specific audience or open to all Colloquium participants?: 
Open
Level of knowledge required: 
Basic
Type of workshop
Type of workshop: 
Training
Abstract text
Abstract: 
Objective:
The Cochrane Statistics Methods Group has developed a series of workshops addressing statistical guidelines as formulated in the Cochrane Handbook for Systematic Reviews of Interventions. This workshop will provide review authors with the knowledge to perform meta-analysis with dichotomous outcome data, and to understand the implications of choosing different effect measures.

Description:
Dichotomous data are perhaps the most common type of outcome in studies included in Cochrane reviews. However, there are several ways in which a simple 2x2 table from an individual clinical trial can be analysed, and even more ways to combine results across studies in a meta-analysis. This workshop will review the notions of risk, odds, risk ratio, odds ratio, risk difference and number needed to treat. It will then discuss the relative merits of the different effect measures, including the implications of swapping interest from ‘events’ to ‘non-events’. A simple overview of the different methods for combining results in meta-analysis will be provided (Peto method, Mantel-Haenszel methods and inverse-variance method). The last of these is newly implemented in RevMan 5, and the rationale for including it will be discussed. Issues will be illustrated by examples. Time will be allowed for discussion of the issues raised.