Analytical approach:
The analysis was based on a modified version of the validated Center for Outcomes Research (CORE) Diabetes Model. A lifetime horizon (up to 35 years) was considered. The authors stated that the perspective of the health care payer was adopted.
Effectiveness data:
Data on treatment efficacy, which was the key clinical input, and the baseline characteristics of patients came from the Prospective Pioglitazone Clinical Trial in Macrovascular Events (PROactive) trial, which was a multinational, prospective, double-blind, placebo, randomised controlled trial (RCT), with 5,238 patients in 321 centres. Other clinical inputs were already incorporated in the simulation model and were based on published sources (for example the UK Prospective Diabetes Study). Assumptions were required to extrapolate the clinical trial data to the patient's lifetime. The key model input was the change in glycated haemoglobin (HbA
1c) with pioglitazone compared with placebo at one, two, and three years and its impact on diabetes complications.
Monetary benefit and utility valuations:
The utility values associated with specific health states were already in the simulation model and were from the Cost of Diabetes in Europe – Type 2 (CODE-2) study.
Measure of benefit:
Quality-adjusted life-years (QALYs) and life-years (LYs) were the summary benefit measures and were discounted at an annual rate of 2.5%.
Cost data:
The economic analysis included the costs of drugs, patient management, and diabetes-related complications, including death. All the costs associated with diabetes complications were reported. Most of the data on resource use were derived from the PROactive trial. The costs were from published Swiss sources. They were in Swiss francs (CHF) and the price year was 2005. Future costs were discounted at an annual rate of 2.5%.
Analysis of uncertainty:
One-way sensitivity analyses were carried out on the key model inputs, such as the discount rate, time horizon, efficacy of pioglitazone, and utility values. Both published and assumed values were used in the alternative scenarios. A Monte Carlo simulation based on nonparametric bootstrapping was also carried out and cost-effectiveness acceptability curves were generated.