Development and Validation of a Novel Patient Microsimulation Cost-Effectiveness Model to Evaluate Automated Insulin Delivery Systems in Type 2 Diabetes
Author(s)
Joshua Weinstein, PhD1, Colin Hopley, MSc1, Anthony Zara, MSc2, Kirk Szafranski, MSc2, Joseph Biskupiak, MBA, PhD3, Dan Carlow, MPH1.
1Insulet, Acton, MA, USA, 2EVERSANA, Burlington, ON, Canada, 3Department of Pharmacotherapy, University of Utah, Salt Lake City, UT, USA.
1Insulet, Acton, MA, USA, 2EVERSANA, Burlington, ON, Canada, 3Department of Pharmacotherapy, University of Utah, Salt Lake City, UT, USA.
OBJECTIVES: A novel patient-level microsimulation was developed, validated and used to assess the cost-effectiveness of a tubeless automated insulin delivery (AID) system + standard of care (SoC) compared to multiple daily injections (MDI) of insulin + SoC for adults with type 2 diabetes mellitus (T2DM) in the United States (US).
METHODS: The simulation evaluates the impact of improved T2DM outcomes over a specified time horizon using a cost utility approach. It was developed using risk equations and risk factor progression equations derived from the United Kingdom Prospective Diabetes Study (UKPDS) 82 and 90, respectively. Internal calibration and validation were performed by comparison to outcomes from UKPDS 82. External and cross validation involved comparison to studies and models included in the Mount Hood 9 (MH9) Challenge. The reference case considered a private payer perspective over 30 years, informed by a subgroup analysis from the SECURE-T2D study. Recent sources for costs and utilities with precedence in T2DM health technology assessment were used.
RESULTS: After calibration, internal validation demonstrated high precision (R2 = 0.99) and reasonable accuracy (mean absolute percentage error [MAPE] = 21%) for all-cause mortality, macrovascular events (R2 = 0.98; MAPE range, 9.5-37.4%), and microvascular events (R2 = 0.97; MAPE range = 8.4-54.6%). External validation indicated improved accuracy (MAPE = 28% versus 37%) compared to the models included in MH9. AID+SoC was associated with an increased cost and increased quality-adjusted life years (QALYs), resulting in $73,107 per QALY compared to MDI+SoC. This was consistent in probabilistic analysis, with a 100% probability of cost-effectiveness at a willingness-to-pay threshold of $100,000 per QALY.
CONCLUSIONS: The microsimulation sufficiently predicted outcomes in internal and external validation, and performed comparably with T2DM models that participated in the MH9 challenge. AID+SoC demonstrated as a cost-effective option compared with MDI+SoC for people with T2DM.
METHODS: The simulation evaluates the impact of improved T2DM outcomes over a specified time horizon using a cost utility approach. It was developed using risk equations and risk factor progression equations derived from the United Kingdom Prospective Diabetes Study (UKPDS) 82 and 90, respectively. Internal calibration and validation were performed by comparison to outcomes from UKPDS 82. External and cross validation involved comparison to studies and models included in the Mount Hood 9 (MH9) Challenge. The reference case considered a private payer perspective over 30 years, informed by a subgroup analysis from the SECURE-T2D study. Recent sources for costs and utilities with precedence in T2DM health technology assessment were used.
RESULTS: After calibration, internal validation demonstrated high precision (R2 = 0.99) and reasonable accuracy (mean absolute percentage error [MAPE] = 21%) for all-cause mortality, macrovascular events (R2 = 0.98; MAPE range, 9.5-37.4%), and microvascular events (R2 = 0.97; MAPE range = 8.4-54.6%). External validation indicated improved accuracy (MAPE = 28% versus 37%) compared to the models included in MH9. AID+SoC was associated with an increased cost and increased quality-adjusted life years (QALYs), resulting in $73,107 per QALY compared to MDI+SoC. This was consistent in probabilistic analysis, with a 100% probability of cost-effectiveness at a willingness-to-pay threshold of $100,000 per QALY.
CONCLUSIONS: The microsimulation sufficiently predicted outcomes in internal and external validation, and performed comparably with T2DM models that participated in the MH9 challenge. AID+SoC demonstrated as a cost-effective option compared with MDI+SoC for people with T2DM.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
EE326
Topic
Economic Evaluation
Disease
Diabetes/Endocrine/Metabolic Disorders (including obesity)