Network meta-analysis of health care interventions with different doses and combinations of components

• Applicant: Maria Petropoulou, Ph.D.
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 504730171

Systematic reviews and network meta-analyses (NMAs) are established tools for evidence synthesis in medical research. NMA combines direct and indirect evidence of randomized controlled trials for estimating the effects of several alternative healthcare interventions. Still, clinical interest in medical decision-making now also focuses on more complex settings, such as multicomponent interventions and dose-response relationships, which so far can hardly be handled in NMA. While interventions are often administrated in different doses, most published NMAs do not consider the exact doses in their statistical model, but instead, either lump or split all doses for defining intervention nodes in the network. These typical approaches do not provide the desired insights into potential dose-response effects, related to dose efficacy and safety. Bayesian methods for dose-response network meta-analysis have recently been proposed for considering the exact doses in the analysis. However, several questions are still open and a solution in the widely used frequentist NMA framework is missing. Furthermore, NMA methods for combining both the dose-response relationships and the different components of interventions (i.e., multicomponent interventions) are needed, e.g., the investigation of dose-response effects of combinations of drugs for preventing nausea and vomiting after general anaesthesia. However, such methods so far are available neither in a Bayesian nor a frequentist framework. To address this demand for NMA in increasingly complex settings, the project will develop a new frequentist framework that incorporates dose information into NMA, also extending component NMA techniques. This will be achieved by combining a graph-theoretical perspective on NMA with approaches for modelling dose-response relationships. To facilitate practical use, model diagnostics will be developed, such as for testing, quantifying, and visualizing the fit of dose-response functions. A categorization approach — distinguishing between low, recommended, and high doses — will be used for comparison. These results will provide the basis for enabling even more flexible models when incorporating both dose-response models and models for components of interventions into NMA. The methods will be implemented as add-ons to the widely used R package netmeta to facilitate broad uptake by the meta-analysis community. Methods development will be guided by two applications, specifically from nutrition research and general anaesthesia, in close collaboration with clinical experts.

• DFG Programme: Research Grants