Seminar: "Network meta-analysis in a nutshell. Current state and future challenges"

AUEB STATISTICS SEMINAR SERIES 2024-2025

Presenter: Dimitris Mavridis, Department of Primary Education, University of Ioannina

Room Τ102, Troias Building

ABSTRACT

Randomized clinical trials (RCTs) are the gold standard of clinical trials for assessing interventions' efficacy. For most healthcare problems, there is a plethora of RCTs and competing interventions. Network meta-analysis (NMA) is a powerful statistical method that allows us to estimate the relative efficacy/effectiveness between competing interventions addressing the similar research question. Suppose that we have interventions A, B and C. A key concept of NMA is that of indirect comparison; if we have data on AvsB and AvsC comparisons, then we can indirectly assess the relative efficacy between BvsC because both B and C are compared to A. The validity of such a comparison lies on the AvsB and AvsC populations being similar in terms of their distribution of effect modifiers. NMA lies in the top of the evidence based methods and several worldwide institutions (e.g., World Health Organization, European Medicine Agency, National Institute for Clinical Excellence in the UK) and medical societies use NMA to produce guidelines and form healthcare policy worldwide. The role of NMA in decision making is expected to further increase due to the new EU Health Technology Assessment (HTA) that harmonizes the HTA processes across the EU, placing much emphasis on evidence synthesis methods. Despite the vast improvement in NMA methodology during its first two decades, there are still  continuing controversies remaining and, additionally, we have progressed to a degree that we are faced with more important and complex questions. In this presentation, we will discuss many of the challenges that NMA scientists are facing including methods to evaluate its assumptions, synthesize randomized and non-randomized evidence, use of individual participant data, population adjustment methods, missing data, ranking interventions etc.