PEA: Review of Models Used in Economic Analyses of New Oral Treatments for Type 2 Diabetes Mellitus
Economic models are considered to be important, as they help evaluate the long-term impact of diabetes treatment. To date, it appears that no article has reviewed and critically appraised the cost-effectiveness models developed to evaluate new oral treatments [glucagon-like peptide-1 (GLP-1) receptor agonists and dipeptidyl peptidase-4 (DPP-4) inhibitors] for type 2 diabetes mellitus (T2DM).
This study aimed to provide insight into the utilization of cost-effectiveness modelling methods. The focus of our study was aimed at the applicability of these models, particularly around the major assumptions related to the clinical parameters (glycated haemoglobin [A1c], systolic blood pressure [SBP], lipids and weight) used in the models, and subsequent clinical outcomes.
MEDLINE and EMBASE were searched from 1 January 2004 to 14 February 2013 in order to identify published cost-effectiveness evaluations for the treatment of T2DM by new oral treatments (GLP-1 receptor agonists and DPP-4 inhibitors). Once identified, the articles were reviewed and grouped together according to the type of model. The following data were captured for each study: comparators; country; evaluation and key cost drivers; time horizon; perspective; discounting rates; currency/year; cost-effectiveness threshold, sensitivity analysis; and cost-effectiveness analysis curves.
A total of 15 studies were identified in our review. Nearly all of the models utilized a health care payer perspective and provided a lifetime horizon. The CORE Diabetes Model, UK Prospective Diabetes Study (UKPDS) Outcomes Model, Cardiff Diabetes Model, Centers for Disease Control and Prevention (CDC) Diabetes Cost-Effectiveness Group Model and Diabetes Mellitus Model were cited. With the exception of two studies, all of the studies made significant assumptions surrounding the impact of GLP-1 receptor agonists or DPP-4 inhibitors on clinical parameters and subsequent short- and long-term outcomes. Moreover, often the differences in the clinical parameters were relatively small (e.g. 1 or 2 mmHg in blood pressure) and would not be considered by many as clinically important. Yet, the impact of these small clinical changes often resulted in large lifetime changes in health outcomes in the models. In particular, many studies assumed that changes in weight associated with the therapies would equate to improved outcomes, despite limited evidence for this assumption. Although the new oral treatments were regarded as cost effective in most studies based upon the studies reviewed, the validity of these projections, particularly for the longer time frames, is questionable. Indeed, although most of these studies have been conducted in the last 5 years, recent trial evidence has already questioned the validity of most of these studies.
It is clear that a number of changes are required in the evaluation of diabetes therapies. First and foremost, the basic models need to be updated to include contemporary important clinical trial data assessing hard clinical outcomes in patients with diabetes. Second, there should be less emphasis on 40-year or lifetime costs and consequences of the therapies and a greater focus on short-term (5-year) and intermediate-term (10-year) outcomes. Practice is continually evolving, and the probability that these models would provide any valid predictions beyond 10 years is remote. Third, all modellers should immediately remove the basic assumption that small clinically inconsequential changes in A1c, SBP, lipids and weight result in major clinical improvements in patients. Future models should aim to include all relevant treatment outcomes, whether these relate to effects on underlying diabetes and its complications or to short- or long-term side effects of treatment. We need to explore why cost-saving interventions could benefit further from adding patient characteristics, which may be able to better predict the use of lower-cost alternatives. Moreover, the vast array of different clinical, cost and utility data used in the different models reviewed makes it apparent that a uniform methodology should be developed for diabetes economic models. In this manner, future models could be run using the same data, which would allow for more acceptable comparability between studies.