Apparent subadditivity of the efficacy of initial combination treatments for type 2 diabetes is largely explained by the impact of baseline HbA1c on efficacy

Aim To explain the subadditive efficacy typically observed with initial combination treatments for type 2 diabetes. Methods Individual subject data from 1186 patients with type 2 diabetes [mean glycated haemoglobin ( HbA1c ) = 8.8%] treated with metformin, canagliflozin or canagliflozin + metformin were used. The baseline HbA1c versus ΔHbA1c relationships for monotherapy arms were determined using analysis of covariance and then used to predict efficacy in the combination arms by modelling how applying one treatment lowers the ‘effective baseline HbA1c ’ for a second treatment. The model was further tested using data from several published combination studies. Results The mean ΔHbA1c levels were −1.25, −1.33, −1.37, −1.77 and −1.81% with metformin, canagliflozin 100 mg, canagliflozin 300 mg, canagliflozin 100 mg/metformin and canagliflozin 300 mg/metformin, respectively. Using the monotherapy results, the predicted efficacy for the canagliflozin/metformin arms was within 10% of the observed values using the new model, whereas assuming simple additivity overpredicted efficacy in the combination arms by nearly 50%. For 10 other published initial combination studies, predictions from the new model [mean (standard error) predicted ΔHbA1c = 1.67% (0.14)] were much more consistent with observed values [ ΔHbA1c = 1.72% (0.12)] than predictions based on assuming additivity [predicted ΔHbA1c = 2.19% (0.21)]. Conclusions The less‐than‐additive efficacy commonly seen with initial combination treatments for type 2 diabetes can be largely explained by the impact of baseline HbA1c on the efficacy of individual treatments. Novel formulas have been developed for predicting the efficacy of combination treatments based on the efficacy of individual treatments and the baseline HbA1c of the target patients.