In simple terms no, because comparisons made are not necessarily randomised. However, I don't think you can make a direct comparison for the simple reason that network meta-analyses allows you to make comparisons that can't be made from the traditional analysis of RCTs - but more caution needs to be applied in interpreting conclusions.
No, network meta-analysis assumes that all interventions included in the "network" are equally applicable to all populations and contexts of the studies included. It is still an evolving method with much more complex methodological approach. It requires specialist statistical expertise and software. Network meta-analysis combines indirect (triangular approach) or direct comparisons (pairwise approach).
Hi Murali,
I'm conducting both, what I call traditional/standard meta-analyses and network meta-analyses. I think you cannot just say one or the other have a higher level of evidence. I agree very much with Peter and Oyekula in that network meta-analyses are still a fairly recent approach, but on the other hand the argument about comparisons not necessarily being randomised is as well true for standard meta-analysis if we're honest. Side note: The only actually randomised approach in this matter would be a registered replication report.
I would answer you question in the following fashion: If you want to know which of two interventions is favourable, then your level of evidence is better if you look at it from a traditional way, but if you want to assess which of multiple interventions scores the best, then the only way looking at it is a network MA.
As a final note I want to say that both approaches heavily depend on the quality of studies you use as the source for your analyses, and the scrutiny of the execution of all necessary steps and analyses.
Hope this helped.
Cheers,
Christoph
Murali Srinivasan Peter Donald Griffiths Oyekola Oloyede
Our simple formula shows the gap between the evidence expected and evidence observed: https://f1000research.com/articles/8-38/v1 also look at the Venn diagram to see the difference between the comparisons in trials, in meta-analysis and in network meta-analysis.
For our network meta-analysis, there was no usable data for 43% of pairwise direct comparisons in the network of intervention so network meta-analysis created 43% of evidence 'statistically'! That's why we have considered them high risk of bias.
Yes. NMA pools data from both direct and indirect comparisons where both exist, so in this particular scenario it has more power than pairwise MA. It also preserves randomisation. Sofia Dias
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis