Cross-Validation of the Glycated Hemoglobin HbA1c Translator: A Predictive Tool to Evaluate the Relationship Between Diabetes Complications and Associated Healthcare Costs
Author(s)
Osvaldo Ulises Garay, MSc1, Anamaria V. Olivieri, MSc2, Anuj Narang, MPharm3, Ahmad Samir Teima, BSc4, Paco Cerletti, PhD5.
1Roche Diagnostics International, Rotkreuz, Switzerland, 2IQVIA, Basel, Switzerland, 3IQVIA, Guragaon, India, 4Roche Diagnostics Middle East, Dubai, United Arab Emirates, 5Roche Diagnostics International, Basel, Switzerland.
1Roche Diagnostics International, Rotkreuz, Switzerland, 2IQVIA, Basel, Switzerland, 3IQVIA, Guragaon, India, 4Roche Diagnostics Middle East, Dubai, United Arab Emirates, 5Roche Diagnostics International, Basel, Switzerland.
OBJECTIVES: Among insulin-treated people with type 1 (T1D) and type 2 diabetes (T2D), improvements in HbA1c are associated with a delayed incidence of diabetes complications (e.g. cardiovascular events, retinopathy, renal disease). The HbA1c Translator is a predictive tool developed by Fortwaengler et al. (2017) to estimate how changes in HbA1c levels relate to the occurrence of diabetes complications and their associated costs. This study aimed to cross-validate predictions from an updated version of the HbA1c Translator with the IQVIA Core Diabetes Model (CDM).
METHODS: Baseline characteristics (e.g. age, HbA1c level, diabetes duration) were aligned between the HbA1c Translator and the CDM. A hypothetical −1% point reduction in HbA1c was simulated vs no reduction. Cumulative incidences and costs of complications were predicted over a 5-year period. Concordance between predictions was assessed by plotting predicted incidences of complications between the HbA1c Translator and the CDM. Predicted outcomes in people with T1D and T2D pre- and post-intervention were compared using ordinary least squares linear regression lines (OLS-LRL) with zero intercepts; values above or below 1.0 indicated over- or underestimation, respectively.
RESULTS: A small underestimation of complications was observed vs the CDM in pre-intervention results for T1D and a small overestimation was observed for T2D (OLS-LRL slopes: T1D = 0.9, T2D = 1.1). A moderate overestimation was observed in the post-intervention predictions (OLS-LRL slopes: T1D = 1.3, T2D = 1.2). The total projected per-person cost savings over 5 years predicted by the HbA1c Translator vs the CDM were: 1089 US dollars (USD) vs 872 USD (24.9% overestimation) for T1D, and 2246 USD vs 1929 USD (16.4% overestimation) for T2D.
CONCLUSIONS: Cross-validation of the modified HbA1c Translator vs the CDM demonstrated high reliability of results, with a moderate overestimation of predicted incidences and corresponding per-person cost savings in T1D and T2D over a 5-year period.
METHODS: Baseline characteristics (e.g. age, HbA1c level, diabetes duration) were aligned between the HbA1c Translator and the CDM. A hypothetical −1% point reduction in HbA1c was simulated vs no reduction. Cumulative incidences and costs of complications were predicted over a 5-year period. Concordance between predictions was assessed by plotting predicted incidences of complications between the HbA1c Translator and the CDM. Predicted outcomes in people with T1D and T2D pre- and post-intervention were compared using ordinary least squares linear regression lines (OLS-LRL) with zero intercepts; values above or below 1.0 indicated over- or underestimation, respectively.
RESULTS: A small underestimation of complications was observed vs the CDM in pre-intervention results for T1D and a small overestimation was observed for T2D (OLS-LRL slopes: T1D = 0.9, T2D = 1.1). A moderate overestimation was observed in the post-intervention predictions (OLS-LRL slopes: T1D = 1.3, T2D = 1.2). The total projected per-person cost savings over 5 years predicted by the HbA1c Translator vs the CDM were: 1089 US dollars (USD) vs 872 USD (24.9% overestimation) for T1D, and 2246 USD vs 1929 USD (16.4% overestimation) for T2D.
CONCLUSIONS: Cross-validation of the modified HbA1c Translator vs the CDM demonstrated high reliability of results, with a moderate overestimation of predicted incidences and corresponding per-person cost savings in T1D and T2D over a 5-year period.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
PT28
Topic
Methodological & Statistical Research
Disease
Diabetes/Endocrine/Metabolic Disorders (including obesity)