Component network meta-analysis compared to a matching method in a disconnected network: A case study.

Area of Science:

• Biostatistics
• Clinical Epidemiology
• Health Services Research

Background:

• Network meta-analysis (NMA) typically requires connected evidence networks from randomized controlled trials (RCTs).
• Disconnected networks pose challenges, often addressed by incorporating non-randomized comparative evidence, which introduces potential bias.
• Previous methods, like those by Schmitz et al., used observational studies to bridge network gaps.

Purpose of the Study:

• To reanalyze multiple myeloma treatment data using component network meta-analysis (CNMA) models based solely on RCTs.
• To evaluate CNMA models as an alternative to propensity score or matching-adjusted indirect comparisons that use observational data.
• To compare CNMA model results with those from a matching method using observational studies.

Main Methods:

• Component network meta-analysis (CNMA) models were applied to RCT data with common treatment components across disconnected networks.
• Forward and backward strategies were employed for selecting appropriate CNMA models.
• Results from CNMA models were compared to those from a matching-adjusted indirect comparison method.

Main Results:

• CNMA models demonstrated a good fit to the analyzed data.
• Treatment rankings derived from CNMA models were similar, though not identical, to those obtained using the matching method.
• The CNMA approach successfully linked disconnected networks using only RCT evidence.

Conclusions:

• CNMA models are a valuable tool for researchers facing disconnected treatment networks with shared components.
• These models provide a robust alternative to methods requiring supplementary observational data, mitigating bias concerns.
• CNMA models, implemented in the R package netmeta, offer a reliable RCT-only approach for evidence synthesis.