Network analysis of clinical trials data: merits and problems

Network meta-analysis (NMA) of randomized controlled trials has been introduced as an extension of traditional pairwise meta-analysis. This method is very valuable to health care providers, payers and for assessing benefit-risk balance where multiple options are available. This method has the advantage of providing treatment effect comparisons of interventions that have not been studied in head-to-head studies. Due to time and costs constraints, it is usually impossible to compare all relevant treatments in one large study. NMA provides data that would otherwise not be available. Before, to perform any analysis, a clear objective in terms of interventions and populations, a systematic literature review following the standard guidelines and a thoughtful feasibility assessment analysis are necessary. In the large context of diabetes, the issues encountered during NMA are usually different for type 1, where few treatment options are available, than for type 2, where several treatments and pathways are possible. When scarce evidence is available, we might favor population adjustment indirect treatment comparisons method to obtain a closer match of two pivotal studies, while when a large amount of studies are available, we might prefer to perform various sensitivity analyses or meta-regressions to explore the robustness of the results. NMA is a broad term and appropriate tools and process should be used to provide unbiased evaluation of treatment alternatives.