The challenges and pitfalls of incorporating evidence from cardiovascular outcomes trials in health economic modelling of type 2 diabetes.

The clinical evidence base for evaluating modern type 2 diabetes interventions has expanded greatly in recent years, with numerous efficacious treatment options available (including dipeptidyl peptidase-4 inhibitors, glucagon-like peptide-1 receptor agonists and sodium-glucose cotransporter-2 inhibitors). The cardiovascular safety of these interventions has been assessed individually versus placebo in numerous cardiovascular outcomes trials (CVOTs), statistically powered to detect differences in a composite endpoint of major adverse cardiovascular events. There have been growing calls to incorporate these data in the long-term modelling of type 2 diabetes interventions because current diabetes models were developed prior to the conduct of the CVOTs and therefore rely on risk equations developed in the absence of these data. However, there are numerous challenges and pitfalls to avoid when using data from CVOTs. The primary concerns are around the heterogeneity of the trials, which have different study durations, inclusion criteria, rescue medication protocols and endpoint definitions; this results in significant uncertainty when comparing two or more interventions evaluated in separate CVOTs, as robust adjustment for these differences is difficult. Analyses using CVOT data inappropriately can dilute clear evidence from head-to-head clinical trials, and blur healthcare decision making. Calibration of existing models may represent an approach to incorporating CVOT data into diabetes modelling, but this can only offer a valid comparison of one intervention versus placebo based on a single CVOT. Ideally, model development should utilize patient-level data from CVOTs to prepare novel risk equations that can better model modern therapies for type 2 diabetes.