Association of HbA1c with Time in Range, Time in Tight Range, Empirically Derived Continuous Glucose Monitoring Glucose Ranges, and Mean Glucose.

OBJECTIVE

To evaluate how continuous glucose monitoring (CGM) glucose thresholds, sampling duration, and statistical modeling influence associations between CGM-derived range metrics and HbA1c, and to compare empirically derived glucose ranges with time in range (TIR) and time in tight range (TITR).

RESEARCH DESIGN AND

METHODS

Out of 8160 HbA1c-matched CGM datasets from 1584 adults, CGM data from the 14, 30, and 90 days preceding each HbA1c measurement were evaluated. Associations between HbA1c and 7790 CGM glucose ranges were assessed across varying lower and upper glucose thresholds. Linear and log-linear mixed-effects models adjusted for age and race with participant-level random intercepts were compared using root mean square error and marginal R2. Participant-level derivation (80%) and validation (20%) cohorts were used for cross-validated threshold selection and discrimination analyses for HbA1c ≤7.0%.

RESULTS

Ninety-day CGM sampling showed the strongest associations with HbA1c. Log-linear models frequently provided comparable or improved fit, particularly for glucose ranges with upper thresholds below 180 mg/dL. Stronger HbA1c associations clustered around upper glucose thresholds of 155-175 mg/dL. In validation analyses, empirically derived ranges demonstrated strong discrimination of HbA1c ≤7.0% (area under the receiver operating characteristic curve [AUC] 0.909-0.915), compared with 0.903 for TIR and 0.905 for TITR, although absolute differences were small. Mean glucose demonstrated the strongest explanatory performance for continuous HbA1c (marginal R2 = 0.74) and similarly strong discrimination (AUC 0.913).

CONCLUSIONS

CGM-HbA1c relationships are influenced by glucose range, sampling duration, and statistical modeling. Our findings reinforce TIR and TITR as robust CGM metrics with strong associations with HbA1c, while also showing that selected empirically derived glucose ranges can provide modestly stronger statistical alignment. These results support the continued use of TIR and TITR while highlighting how glucose range definition may refine interpretation of CGM-based endpoints in future research.