Sutton AJ, Abrams KR, Jones DR; International Society of Technology Assessment in Health Care. Meeting.
Annu Meet Int Soc Technol Assess Health Care Int Soc Technol Assess Health Care Meet. 1999; 15: 55.
Department of Epidemiology and Public Health, University of Leicester, England.
A recent systematic review of meta-analysis methods [1] commissioned by the NHS Health Technology Assessment Programme forms the starting point for this work. This review identified a considerable body of work, produced over the past two decades, which developed the methodology for carrying out meta-analysis. The review highlighted areas where methodology was missing or where it could benefit from being developed further. One such topic is the inclusion of more information in meta-analysis than is routinely done at present. This issue is given further consideration in this paper. Taking the standard random effects model as the starting point, each of the following extensions are considered: 1) including (multiple) study level covariates; 2) including information from multiple study types (including randomized and observational); 3) considering multiple outcome measures simultaneously which includes the benefit of allowing studies which report different outcomes to be incorporated in a single meta-analysis model; and 4) including extra information via prior distributions in a Bayesian analysis. Each of these ideas is illustrated using data from a variety of sources on serum cholesterol and its effect on coronary and mortality outcomes. The relative merits of each extension is discussed and unique results derived using each method are highlighted. Several of these extensions are then applied simultaneously in an attempt to produce a "best" model. This model goes some way to forming a response profile of treatment effects, and assists in the assessment of optimum treatment regimens for specific patient groups. A brief discussion of further opportunities to include extra information in a synthesis is given. Approaches which are currently being developed include the addition of individual patient data to a meta-analysis of summary estimates and the incorporation of information from qualitative studies. REFERENCE: 1. Sutton AJ, Abrams KR, Jones DR, Sheldon TA, Song F. Systematic reviews of trials and other studies. Health Technology Assessment 1999; 2(19).
Publication Types:
Keywords:
- Bayes Theorem
- Biomedical Research
- Biomedical Technology
- Clinical Trials as Topic
- Humans
- Meta-Analysis
- Models, Biological
- Qualitative Research
- Research Design
- Technology Assessment, Biomedical
- methods
- hsrmtgs
Other ID:
UI: 102193805
From Meeting Abstracts