Using Network Meta-Analysis Of Individual Patient Data (IPD) & Summary Aggregate Data (SAD) To Identify Which Combinations Of Interventions Work Best For Which Individuals

OBJECTIVES: In many settings interventions are comprised of a number of potential components, and are sometimes therefore termed "complex". Such a range of potential interventions means that not only do we need to consider which combination is best for a population overall, but also which combination is best for particular sub-populations. The use of Individual Patient Data (IPD) allows such a question to be answered whilst minimising the problem of ecological bias. METHODS: Using a recent Cochrane Collaboration systematic review and subsequent pairwise metaanalysis on the safe storage of medicines we undertook a Network Meta-Analysis (NMA) of both IPD and Summary Aggregate Data (SAD), adjusting for heterogeneity in study design, in order to identify which combination of interventions was the most appropriate for specific sub-populations defined by individual level covari-ates. RESULTS: Based on SAD from 13 Randomised Controlled Trials (RCTs) the use of any intervention led to a statistically significant increase in the safe storage of medicinal products [OR: 1.53,95% CI: 1.27 to 1.84). However, interventions could comprise up to 5 different separate components, and using a NMA approach, and including IPD from 9 of the 13 RCTs, we were able to explore the heterogeneity between both component combinations and their effect in specific sub-populations. CONCLUSIONS: NMA of IPD and SAD can allow identification of the optimal potential combination of individual components for specific sub-populations and when there is a high level of uncertainty be used to help identify and design appropriate further RCTs.