I am working on a systematic review and meta-analysis of RCTs and I am wondering if there is a different approach for rating the quality of the evidence if the difference is statistically significant versus when the difference is not. Or should the quality-rating be restricted to only the significant outcomes? To me it does not seem logical to downgrade the quality of the evidence for an outcome (with a non-significant result) because there is a high risk of bias in the included studies.
In a word, no, the approach is exactly the same. Studies showing null effects are very important in meta-analysis. Imagine what would happen if you omitted them. You would seriously bias the results of the meta-analysis towards finding an effect. Much of the labour in meta-analyses goes precisely towards finding studies with less impressively positive results.
You should GRADE evidence quality for non-significant outcomes in just the same way you would do for significant ones. If imprecision is a factor, yes, you must take that into account. However, there are other reasons (e.g. clinical equivalence) for statistical non-significance in a given outcome.
You can find the entire GRADE series of articles in the Journal of Clinical Epidemiology, here:
http://www.jclinepi.com/content/jce-GRADE-Series
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis