The Virtual DCCT #3: Relationship of HbA1c and CGM Metrics with Cardiovascular Outcomes

Objective: Using a multistep machine-learning approach, the aim is to create virtual continuous glucose monitoring (CGM) traces from glycemic data collected in the Diabetes Control and Complications Trial (DCCT) to assess the relationship between CGM metrics and DCCT cardiovascular (CV) outcomes in people with type 1 diabetes. Research Design and Methods: Utilizing the virtual CGM traces created for each DCCT participant, as previously published, discrete Cox proportional hazard models were used to calculate hazard ratios (HRs) for the association between glycemic metrics (hemoglobin A1c [HbA1c] and virtual CGM) and 3 separate DCCT CV outcome definitions: (1) all DCCT-recorded events; (2) a restricted set of "hard" CV end points; and (3) a restricted set of major CV and major peripheral vascular events. Results: Mean HbA1c and CGM metrics reflective of hyperglycemia were consistently higher, and time-in-range (70-180 mg/dL) and time-in-tight-range (70-140 mg/dL) were consistently lower, in DCCT participants who experienced a CV outcome versus those who did not. For the outcome definition encompassing all CV events, specific adjusted HRs for a CV outcome per a 1 standard deviation (SD) change in glucose metrics were 1.29 for HbA1c with nearly identical values of 1.29-1.31 for relevant CGM metrics. A similar pattern was seen when assuming a 0.5 SD change in glucose metrics. Notably, there was no increased risk for experiencing a CV outcome as time-below-range increased, and in fact, there was a trend toward a slightly protective effect when assuming either a 1- or 0.5-SD change in virtual hypoglycemia metrics. Conclusions: Virtual CGM metrics are associated with CV outcomes in people with type 1 diabetes. These findings support the case for CGM metrics to be included as clinical trial primary endpoints for this population.