Insulin agents available for the treatment of diabetes mellitus include conventional insulins and insulin analogues. Insulin analogues were developed to mimic more closely the separate bolus and basal components of insulin secretion.1 Rapid-acting (bolus or mealtime) and longacting (basal or background) analogue formulations are available. This new class of drugs has been promoted as providing more flexible treatment schedules and a reduced risk of hypoglycemia relative to conventional insulins.
1The cost of insulin analogues exceeds that of conventional insulins.2,3 More than US$7.3 billion was spent globally on the purchase of insulin products in 2005 -an increase of 19% over the previous year.4 It has been suggested that the increased expenditure was due to both the increasing prevalence of diabetes and the increased use of insulin analogues. 5 We performed an analysis of the cost-effectiveness of insulin analogues compared with conventional insulins in the management of type 1 or type 2 diabetes in adults.
MethodsThe economic model We used the Center for Outcomes Research Diabetes Model 6 to calculate the cost-effectiveness estimates. This model, described in detail by Palmer and colleagues, 6 has been validated against published clinical and epidemiologic studies (Figure 1). 7 Using data derived from the published literature, the model uses mathematical equations to determine the diabetes-related complications that would occur throughout a patient's life span. 6 The equations take into consideration risk factors such as age and hemoglobin A 1c levels, as well as patient characteristics, type of diabetes and history of diabetesrelated complications. 6 For type 1 and type 2 diabetes, correlation analyses produced R 2 estimates of 0.9778 and 0.8861, which demonstrate that simulations in the Center for Outcomes Research Diabetes Model 6 provide a reasonably accurate representation of patient outcomes in real-life settings. 7 We derived the clinical effects of therapy (hemoglobin A 1c , mild to moderate hypoglycemia and severe hypoglycemia), required as inputs for the model, from meta-analyses of randomized controlled trials (Table 1). [8][9][10] We compared rapidacting insulin analogues (insulin aspart and insulin lispro) with regular human insulin. We compared long-acting analogues (insulin glargine and insulin detemir) with neutral protamine Hagedorn insulin. For treatment comparisons, we