Different Methods for Modelling Severe Hypoglycaemic Events: Implications for Effectiveness, Costs and Health...

Area of Science:

• Health Economics
• Biostatistics
• Pharmacoeconomics

Background:

• Clinical trials report severe hypoglycaemic events using various metrics, necessitating appropriate analytical models.
• Different network meta-analysis models exist for analyzing diverse data types from clinical trials.

Purpose of the Study:

• To evaluate the impact of different network meta-analysis models on effectiveness, cost, and health utility estimates for severe hypoglycaemic events.
• To inform economic modeling for basal insulin choice in type 1 diabetes mellitus.

Main Methods:

• Analysis of a dataset from a network meta-analysis on severe hypoglycaemic events in type 1 diabetes.
• Fitting models with binomial and Poisson likelihoods, and a shared-parameter model.
• Comparison of relative effects, costs, and disutility estimates across models.

Main Results:

• Relative treatment effects for severe hypoglycaemic events were consistent across different network meta-analysis models.
• Significant differences were observed in baseline event probabilities (logit: 0.07, complementary log-log: 0.17, Poisson: 0.29).
• These differences translated to variations in yearly costs (up to £110) and disutility (0.004).

Conclusions:

• The choice of network meta-analysis model has a limited impact on the relative treatment effects of severe hypoglycaemic events.
• Accurate baseline probabilities are essential for economic models to avoid misrepresenting costs and health outcomes.
• Careful selection of model parameters is vital for robust pharmacoeconomic evaluations in type 1 diabetes.