3
$\begingroup$

Does anyone have experience on conducting a network-meta analysis based on observational studies?

Although network meta-analysis were proposed as a method to produce indirect comparisons between two or more interventions based on randomized controlled trials, some reports have outlined a potential role for conducting this type of meta-analysis also with the inclusion of non-randomized studies (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4634799/).

What are the possible pitfalls? Is this a valid methodological approach?

$\endgroup$
1
  • $\begingroup$ So there are a lot of pitfalls that have to do with the inherent biases in the study designs themselves. As such, most would/ should argue that RCTs should not be combined with non-RCTs in a meta-analysis (pairwise or NMA). If you ignore study quality and bias then observational data is data and can be pooled just like data arising from RCTs (not recommended). $\endgroup$
    – abousetta
    Jan 14 at 3:43
1
$\begingroup$

The primary issues are the same as in observational meta-analyses, the exchangeability of subjects within studies between the treatment groups as well as differences between studies in what they controlled for in order to deconfound the treatment and outcome relationship.

In general the calculation of standard errors in network meta-analyses is still an area of active research which would carry over to the observational

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.