OBJECTIVES: To validate outcomes from a diabetes model through comparison to results reported in trials published in the peer-reviewed literature. METHODS: Diabetes is a debilitating, costly disease that continues to increase in prevalence. Computer simulation models that estimate the impact of interventions on behalf of patients with diabetes are pivotal tools in improving care and evaluating place in therapy and cost-effectiveness of new treatments. The model was developed using the latest, best available evidence from the literature. It includes a set of complication submodels, a continuous-time HbA1c model, and a treatment model that can replicate recently published consensus algorithms. Additionally, the model incorporates treatment specific adverse events, patient adherence to therapy, and estimates of the patient population with a durable response. Random sampling from distributions from trial cohort characteristics is performed to build a patient profile. Each patient is simulated over the trial timeframe. Complications included: macrovascular-heart (coronary heart disease+congestive heart failure), microvascular- stroke, microvascular (renal+neuropathy+retinopathy), mortality, and overall complications rates. Scatter plots of the model predicted results versus the results reported for numerous trial populations in the literature (including 7 studies from ACCORD, ASPEN, and ADVANCE) were constructed. Linear regression estimates were calculated with adjusted correlation coefficients as an estimate of model validity. RESULTS: The predicted model outcomes were generally acceptably accurate as judged by adjusted correlation coefficients (macrovascular-heart, 0.9118; macrovascular-stroke, 0.5388; microvascular, 0.9508; mortality, 0.9808; overall complications, 0.9334). CONCLUSIONS: The diabetes modeling framework possesses the necessary flexibility to perform broad population analysis and important subgroup analyses. The validation exercise, in which the model simulates published cohorts, adequately predicts observed rates of complications.
Furiak, N., Bansal, M., Gahn, J. C., Klein, R. W., & Smolen, H. J. (2012). PDB18 Validation of a Diabetes Modeling Framework. Value in Health, 15(4), A173. https://doi.org/10.1016/j.jval.2012.03.939