Network meta-analysis is a coveted skill in the systematic review world and can be associated with complexity and confusion. Luckily for me, Evidence Synthesis Ireland was on hand to host a 3-day workshop in Galway to explore the theory and methodology involved in conducting a network meta-analysis (NMA) and lead us through conducting an NMA on an example data set. Experts were called in from Université Paris Cité, with Dr Anna Chaimani and Dr Theodoros Evrenoglou leading the lectures and practical sessions over the course of the workshop, and they couldn’t have been more helpful and accommodating of everyone’s questions and queries.
Day 1 focussed on discussing in what context NMA is the appropriate analysis method, and what assumptions must be made when planning an NMA. Some key messages from Day 1 were:
- · NMA is used to compare two or more interventions, and it can enable inference for comparisons of interventions that have never been made in individual studies.
· The combined estimates from both direct and indirect comparisons provides a higher precision estimate than pairwise meta-analysis, which is known as the mixed estimate.
This day became slightly more technical, starting with looking at different approaches for evaluating incoherence in your data. Pastries and caffeine were the saviours of the morning!
We then looked at approaches for ranking interventions, delving in depth into a model for this, known as SUCRA. A key piece of advice was to be extremely cautious with ranking – it doesn’t take everything into consideration, so it’s not a good idea to recommend one intervention above all others based on the ranking results alone.
We also got familiar with NMAstudio on this day, which is an amazing piece of software for conducting the NMA and visualising the results. One surprising fact I learned was that most network graphs on NMA reviews have been conducted by hand using PowerPoint, which sounds like arduous work. Fortunately, NMAstudio is now able to make these look good for you, so I’d recommend trying it out.
By Day 3 we were up to speed with when an NMA is the appropriate strategy, and how to plan and conduct the analysis, so it was time to put it into practise using R studio. Using an example dataset, we performed our own NMA. Now that we were confident in the standard NMA technique, we finally looked at how to assess rare events and complex interventions with multiple components by using alternative models.
By the end of the workshop, I felt assured that I could recognise when an NMA should be implemented, and how to conduct it, ensuring that the correct assumptions are made, and complexities within the interventions have been accounted for.
My main take home from the workshop was the importance of having a knowledgeable and diverse author team. Expertise in both the interventions and the methods is crucial to allow the team to build a comprehensive protocol that anticipates the complexities that may influence the assumptions made in the analysis. Here at CIDG and READ-It we are fortunate to be connected to a diversity of individuals and organisations that can contribute this level of expertise to our review topics, so I’m now looking forward to connecting with my co-authors to contribute my new knowledge to an upcoming NMA review on digital devices for tuberculosis treatment adherence – watch this space!
Tilly Fox and participants at the Evidence Synthesis Ireland Network Meta-analysis Workshop (April 2023)
Tilly Fox is supported by the Research, Evidence and Development Initiative (READ-It) project. READ-It (project number 300342-104) is funded by UK aid from the UK government; however, the views expressed do not necessarily reflect the UK government’s official policies.