Introduction to evidence synthesis for decision making
Sensitivity analyses should be used to explore robustness of conclusions to variations in assumptions, when the evidence is compatible with alternative interpretations or analyses. The document ends with a brief description of the other documents in the series
| Main Authors: | , , , |
|---|---|
| Corporate Author: | |
| Format: | eBook |
| Language: | English |
| Published: |
London
National Institute for Health and Clinical Excellence (NICE)
2012, 2012
|
| Series: | NICE DSU technical support document
|
| Subjects: | |
| Online Access: | |
| Collection: | National Center for Biotechnology Information - Collection details see MPG.ReNa |
| Summary: | Sensitivity analyses should be used to explore robustness of conclusions to variations in assumptions, when the evidence is compatible with alternative interpretations or analyses. The document ends with a brief description of the other documents in the series Attention should be drawn to known or potential effect modifiers, biases in the trial evidence, or markers associated with risk of bias, as consideration should be given to the potential role of methods for addressing heterogeneity and bias. Some guidance is suggested on approaches that can be taken to disconnected evidence networks. Section 4 sets out some "good practice" in presenting the evidence base for relative treatment effects, transparency and reproducibility of methods, presentation of results, and presentation of model critique and model selection. We encourage the use of network diagrams and tables to clarify exactly what comparative evidence is used in the relative efficacy synthesis, and the production of tables that show both the relative and the absolute effects of treatments that are taken forward for use in the cost-effectiveness analysis. Justification and explanation should be given for choice of statistical model, for example fixed or random effects synthesis. This document serves as a brief introduction and orientation to the series of Technical Support Documents on Evidence Synthesis. It describes (Section 2) the overall analytic approach to evidence synthesis, in terms of separate models for the "baseline", that represents the natural history in the specified target population under a standard comparator, and the model for the treatment effects relative to that standard. It then clarifies the impact of the decision making context on synthesis methodology. It concludes that, in order for the estimates of treatment effects to be appropriate for cost-effectiveness analysis and comparative effectiveness research alike (and indeed for any scientific enquiry into the relative efficacy of medical interventions), the decision maker should restrict the study inclusion criteria to a target patient population that has a relatively narrow definition. In the specific context of cost-effectiveness analysis, the document then refers to the need for synthesis methods that are compatible with the use of probabilistic methods. Criteria for inclusion of treatments are described (Section 3), distinguishing between the comparator set of treatments in the decision analysis, and the comparator set of treatments used in synthesis. Once a target patient population has been defined, a suggested trial inclusion rule that avoids potential ambiguity regarding the relevance of evidence is to include any trial that compares at least two treatments in the synthesis comparator set. We suggest that any exceptions to this should require special justification. Trials with small sample sizes or trials with sample size motivated for particular reasons (such as "noninferiority trials") should not be excluded unless specific reasons can be given. |
|---|---|
| Item Description: | "Last updated April 2012" |
| Physical Description: | 1 PDF file (24 pages) illustrations |