Abstract
BACKGROUND: The Early Treatment Diabetic Retinopathy Study Diabetic Retinopathy Severity Scale (DRSS) is a standard approach to measure diabetic retinopathy (DR) severity. Many clinical trials evaluating drug intervention for DR rely upon demonstration of a therapeutic effect through measurement of a 2- or 3-step improvement or progression on the DRSS; however, these binary endpoints require a relatively large sample size for a reliable estimate of therapeutic efficacy, especially when the SOC (eg, anti-VEGF) is used as a control. This study was designed to evaluate the sensitivity and statistical efficiency of detecting a drug effect in DR across different DRSS endpoints, and present alternative analytical approaches to enable smaller-size DR trials for detecting a reliable efficacy signal before moving into larger confirmatory DR trials. METHODS: Data from two randomized, double-blinded, controlled Phase III trials, that enrolled patients with decreased vision due to center-involved DME and the presence of macular edema documented on optical coherence tomography and simulated data, were used for this study. Changes in DRSS steps during a 3-month period from patients (n=205) with no active intervention were used to confirm the reliability of DRSS outcomes. A simulation study compared sensitivity and statistical efficiency across different DRSS endpoints. RESULTS: The standard deviation of step change between baseline and month 3 DRSS across different steps at baseline were all within 1 step, confirming the reliability of DRSS measure by each step. Efficiency of detecting reliable therapeutic efficacy was augmented when treatment effect in improvement and progression was evaluated together; highest sensitivity was observed when change in DRSS steps was used directly as an endpoint. CONCLUSION: DRSS step change may provide more robust sensitivity and statistical efficiency. It is therefore a more cost-effective endpoint for the detection of therapeutic efficacy signal in drug discoveries in DR.