What can go wrong with network meta-analyses?
You often have multiple endpoints but these are not collected consistently across all the studies included in the NMA. How do you deal with this?
Even with the same variables assessed over studies, they might be collected at different time points. This often has an effect on both efficacy and safety results. What do you do?
Your placebo treatment might change over time and consequently your placebo response. Often placebo plays a crucial part in your network, but can you really pool all your placebo arms? The patient population differs between studies and hence between treatments. How do you control this bias?
The pool of relevant treatments differs across countries when using network meta-analyses for HTA submissions. Should you adapt your NMA for each country’s submission?
You’re not submitting all your HTA dossiers at once and thus the literature search needs to be updated repeatedly hence you input data into the NMA. How big are the differences and how long will it take to update everything (and what will it cost?)?
Some of the input data is missing for some studies. How do you deal with these missings – just exclude them?