We have posted a preprint that addresses the six major misconceptions currently circulating with respect to meta-analysis models with an aim to dispel these and move evidence synthesis forwards. Your comments will be appreciated.
3 Likes
I have listed what we believe are the six entrenched misconceptions here for ease of access:
Misconception 1: The chosen parameter assumption determines if we can infer beyond the meta-analytical set of studies.
Misconception 2: The FE model is the only model under the common parameters assumption
Misconception 3: All models that address heterogeneity are RE models
Misconception 4: Diversity in characteristics across trials (heterogeneity) should guide model selection in meta-analysis
Misconception 5: The RE model is the best way to address overdispersion seen with the FE model
Misconception 6: Heterogeneity renders the common parameters assumption to be unrealistic
1 Like
Paper published here
1 Like
In my field systematic reviews and meta-analysis are so poorly done I dont think they have clinical utility. They combine a lot of poorly designed, analyzed studies answering different questions with sloppy methodology in an R or Stata program that gives them a meaningless number. I agree with Professor Harrell that IPD meta-analyses are the only ones of value.
2 Likes
I agree with your thoughts as this leads to research waste. There still remains a role for aggregate data meta-analysis but it needs boundaries to stop research waste and we have set this out in this paper: Angry scientists, angry analysts and angry novelists
1 Like