ALTERNATIVE APPROACHES TO MODELLING HBA1C PROGRESSION IN TYPE 2 DIABETES AND THEIR IMPACT ON HEALTH ECONOMIC OUTCOMES

The trajectory of HbA1c over time is an influential component within type 2 diabetes mellitus (T2DM) models. This study investigated the impact of alternative approaches to modelling HbA1c progression on health economic outcomes. A systematic literature search was conducted to identify T2DM studies (duration ³ 2 years and published between January 2005 and December 2015) that reported change in HbA1c. A total of 68 articles were reviewed and 2 alternative approaches to modelling HbA1c change were derived: i) a mathematical model [MM] related to baseline, minimum/maximum HbA1c levels; ii) a multiple linear regression model [LM]. These were incorporated into the Cardiff Diabetes Model with 40-year time-horizon and compared to the UKPDS 68 HbA1c progression approach in predicting event rates, costs (UK£), quality-adjusted life years (QALYs) and time to therapy escalation. UKPDS baseline characteristics, published utility values and 2015 costs were used (both discounted at 3.5%). Using UKPDS 68 HbA1c progression the model predicted 231.8 cardiovascular and 6.3 microvascular events, and per-patient costs and QALYs of £8,843 and 12.65, respectively. The MM approach resulted in 243.1 cardiovascular and 7.1 microvascular events, and costs and QALYs of £9,067 and 12.55, respectively. The LM approach resulted in 197.7 cardiovascular and 4.3 microvascular events, and costs and QALYS of £8,351 and 12.87, respectively. Commencing from a baseline HbA1c of 7.0%, the time taken to reach 7.5% was 4, 4 and 15 years for UKPDS, MM and LM approaches, respectively and 12, 7 and 28 years to reach 8.5% for the UKPDS, MM and LM approaches, respectively. The choice of approach to modelling HbA1c trajectories in T2DM can significantly impact predicted health economic outcomes, particularly in relation to the timing of therapy escalation and should therefore be justified and evaluated by relevant sensitivity analysis to ensure results are robust.