RaCE: A rank-clustering estimation method for network meta-analysis.

Area of Science:

• Biostatistics
• Health Services Research
• Evidence Synthesis

Background:

• Network meta-analysis (NMA) is essential for comparing multiple interventions and informing clinical decisions.
• Traditional NMA ranking methods can oversimplify treatment effects, leading to misleading conclusions due to uncertainty.

Purpose of the Study:

• To introduce a novel Bayesian rank-clustering estimation (RaCE) approach for NMA.
• To offer a more nuanced interpretation of intervention effectiveness by clustering treatments with similar outcomes, rather than solely identifying a single best intervention.

Main Methods:

• Developed a Bayesian rank-clustering estimation (RaCE) approach for NMA.
• Decoupled the clustering from the NMA modeling for flexibility across outcome types, modeling approaches, and estimation frameworks.
• Validated through simulation studies and an NMA of frontline immunochemotherapies for follicular lymphoma.

Main Results:

• RaCE effectively identifies rank-clusters even with significant uncertainty and overlapping intervention effects.
• The approach provides more reasonable interpretations compared to traditional single-ranking methods.
• Application to follicular lymphoma revealed clinically relevant clusters among treatments previously considered distinct.

Conclusions:

• RaCE enhances rank estimation and interpretability in NMA.
• This method facilitates evidence-based decision-making in complex intervention comparisons.
• RaCE offers a valuable tool for researchers synthesizing evidence on multiple interventions.