Please enable JavaScript.

Coggle requires JavaScript to display documents.

Meta-Analysis (Synthesis & Heterogeneity (Heterogeneity (When evidence…

- - - - These differences may just be statistical noise but could also reflect genuine differences in study findings
- - - - See diagrams in Drummond et al.
    - - (i) options being compared which are directly related to the decision
      - (ii) trials that include these options may also include other options which may not be related to the decision, but do contribute evidence relating to those that are
    - - NMA is increasingly used but it is important to be clear about the underlying evidence + the implications for the validity and uncertainty of the estimates
      - Necessary to consider heterogeneity between studies due to characteristics of the patients who were randomised or the features of the study (e.g. design/location/age)
      - Key issue: whether any differences in trials are likely to impact the treatment effect estimates
- - - - These characteristics can be used as covariables in regression analysis, generating subgroup-specific estimates of particular parameters (meta-regression)
        
        Meta regression can be a powerful tool for explaining heterogeneity, but should be used carefully
        
        Effective sample size for such analyses relates to the number of studies, which is often small
        
        Risk of confounding with meta-regression, where an association may actually reflect correlation with another variable not included in the synthesis
        
        One way to use meta-regression: adjust the estimates of a parameter (e.g. treatment effect) for the quality/relevance of the study which may be an important source of heterogeneity