Predicting the HbA1c level following glucose-lowering interventions in individuals with HbA1c-defined prediabetes: a post-hoc analysis from the randomized controlled PRE-D trial

Purpose To investigate whether the prediction of post-treatment HbA 1c levels can be improved by adding an additional biomarker of the glucose metabolism in addition to baseline HbA 1c . Methods We performed an exploratory analysis based on data from 112 individuals with prediabetes (HbA 1c 39–47 mmol) and overweight/obesity (BMI ≥ 25 kg/m 2 ), who completed 13 weeks of glucose-lowering interventions (exercise, dapagliflozin, or metformin) or control (habitual living) in the PRE-D trial. Seven prediction models were tested; one basic model with baseline HbA 1c as the sole glucometabolic marker and six models each containing one additional glucometabolic biomarker in addition to baseline HbA 1c . The additional glucometabolic biomarkers included: 1) plasma fructosamine, 2) fasting plasma glucose, 3) fasting plasma glucose × fasting serum insulin, 4) mean glucose during a 6-day free-living period measured by a continuous glucose monitor 5) mean glucose during an oral glucose tolerance test, and 6) mean plasma glucose × mean serum insulin during the oral glucose tolerance test. The primary outcome was overall goodness of fit ( R 2 ) from the internal validation step in bootstrap-based analysis using general linear models. Results The prediction models explained 46–50% of the variation ( R 2 ) in post-treatment HbA1c with standard deviations of the estimates of ~2 mmol/mol. R 2 was not statistically significantly different in the models containing an additional glucometabolic biomarker when compared to the basic model. Conclusion Adding an additional biomarker of the glucose metabolism did not improve the prediction of post-treatment HbA 1c in individuals with HbA 1c -defined prediabetes.