Response to Comment on Garvey et al. Association of Baseline Factors With Glycemic Outcomes in GRADE: A Comparative Effectiveness Randomized Clinical Trial. Diabetes Care 2024;47:562–570

We thank Cardoso et al. (1) for responding to our paper, which analyzed baseline factors associated with the glycemic outcome in the Glycemia Reduction Approaches in Diabetes: A Comparative Effectiveness Study (GRADE) (2). Cardoso et al. note several patient attributes that predict differential glycemic responses in type 2 diabetes (T2D) based on excellent studies using the observational U.K. primary care data set and the TriMaster (Study of a DPP4 Inhibitor, SGLT2 Inhibitor, and Thiazolidinedione as Third-Line Therapy in Patients With Type 2 Diabetes) trial.

Our GRADE analyses explored a wide array of baseline characteristics associated with our primary outcome, defined as confirmed HbA_1c ≥7.0% over an average of 5 years in patients who were treated with metformin at baseline and were randomized to sitagliptin, glimepiride, liraglutide, or glargine insulin as add-on medications. In univariate analyses, multiple baseline factors were associated with the primary glycemic outcome at years 1 and 4. Many of these positive associations agree with observations in the publications referenced in the comment letter. However, in multivariate and classification and regression tree (CART) analyses at year 4, only treatment group, younger age, higher HbA_1c, and fasting glucose were statistically significant factors in these models. The original observation in GRADE indicating modest clinical superiority of liraglutide and glargine over glimepiride and sitagliptin at year 4 remained after controlling for baseline factors, and no factors were associated with the differential effectiveness of the medications.

We share enthusiasm with Cardoso et al. (1) regarding individualized selection of diabetes medications based on clinical characteristics in T2D. Many of the studies referenced by Cardoso et al. were of ≤1-year duration. One exception is a study using the U.K. primary care data set (3), which demonstrated that the superior effectiveness of sulfonylureas in males with BMI <30 kg/m² over females had dissipated by year 3, while the superiority of thiazolidinediones (not studied in GRADE) in women with BMI >30 kg/m² over males persisted for 5 years. More studies are required that assess factors that predict differential responsiveness over a longer term (i.e., >1 year). T2D is a chronic progressive disease, and the predictive value of baseline characteristics on the more short-term differential effects of specific medicalizations is likely to change over time. We take the point of Cardoso et al. that the primary glycemic outcome in GRADE of HbA_1c ≥7.0% is somewhat rigorous and that findings assessing a higher HbA_1c binary outcome or glycemia as a continuum may provide different results. However, HbA_1c <7.0% is the target recommended in most evidence-based guidelines for long-term glycemic control to prevent diabetes complications.

We have one point of clarification. Cardoso et al. (1) state that we found a lack of association between insulin secretion and response to the glucagon-like peptide 1 receptor agonist liraglutide, in disagreement with previous reports. However, we did find that reduced insulin secretion was highly associated with reduced effectiveness of liraglutide at both years 1 and 4and that this was true for the other medications as well. However, this effect did not persist in multivariate analyses and could not be used to predict differential effectiveness among the four add-on medications.