Treatment recommendations based on network meta-analysis: Rules for risk-averse decision-makers

Area of Science:

• Health Economics
• Decision Analysis
• Biostatistics

Background:

• Network meta-analysis (NMA) typically recommends a single treatment based on highest expected value (EV).
• Existing methods may not adequately address decision-maker risk aversion or uncertainty.
• Need for ranking systems that penalize uncertainty and identify clinically relevant treatment options.

Purpose of the Study:

• To develop and evaluate novel approaches for multiple treatment recommendations in NMA.
• To introduce and assess loss-adjusted expected value (LaEV) as a risk-averse decision criterion.
• To compare LaEV with existing methods like GRADE and probability-based rankings.

Main Methods:

• Introduction of loss-adjusted expected value (LaEV) and definition of desirable ranking properties.
• Development of a two-stage recommendation process incorporating minimal clinically important difference (MCID).
• Comparative analysis using stylized examples and 10 National Institute of Health and Care Excellence (NICE) NMAs.

Main Results:

• LaEV reliably produces valid rankings under uncertainty and possesses desirable properties.
• LaEV-based recommendations were more conservative than standard EV, recommending fewer treatments (median 2 fewer).
• GRADE rules showed anomalies, sometimes failing to recommend the highest EV/LaEV treatment and privileging uncertain options.

Conclusions:

• LaEV provides a theoretically sound, conservative, and implementable method for risk-averse decision-making in NMA.
• A two-stage approach using MCID ensures clinically appropriate numbers of recommended treatments.
• LaEV offers a superior alternative to standard EV, GRADE, and probability-based rankings for complex treatment comparisons.