Network meta-interpolation: Effect modification adjustment in network meta-analysis using subgroup analyses

Area of Science:

• Biostatistics
• Evidence Synthesis
• Clinical Epidemiology

Background:

• Effect modification (EM) can introduce bias in network meta-analysis (NMA).
• Current population adjustment methods for EM in NMA often require individual patient data and assume shared effect modification (SEM).
• These methods overlook valuable subgroup information available in aggregated network data.

Purpose of the Study:

• To introduce Network Meta-Interpolation (NMI), a novel method for adjusting NMA for EM.
• To develop an NMA adjustment method that utilizes subgroup analyses without assuming SEM.
• To evaluate NMI's performance against standard NMA, network meta-regression (NMR), and Multilevel NMR (ML-NMR).

Main Methods:

• NMI adjusts for EM by transforming subgroup- and study-level treatment effect (TE) estimates into TE and standard errors at common EM values.
• An extensive simulation study was conducted using two evidence networks with four treatments each.
• Simulations assessed performance under varying conditions, including departures from SEM, variable EM correlation, and different network/sample sizes.

Main Results:

• NMI demonstrated superior estimation accuracy across all simulated scenarios, outperforming standard NMA, ML-NMR, and NMR.
• In the base case (non-SEM), NMI achieved the lowest RMSE (0.228).
• NMI also showed consistent dominance in credible interval (CrI) coverage, indicating improved reliability.

Conclusions:

• Network Meta-Interpolation (NMI) is an effective method for addressing EM bias in NMA, particularly when subgroup data is available.
• NMI offers an advantage over existing methods by not requiring the SEM assumption.
• NMI provides a robust approach for handling study imbalance and maximizing the utility of available subgroup data in NMA.