IN PATIENTS with Type 2 diabetes (T2D), achieving optimal glucose control is imperative for managing the condition and reducing the risk of complications. However, selecting the optimal glucose-lowering therapy can be challenging due to the many available options. A new study has developed and validated a five-drug class model, designed to predict the optimal drug class for an individual, based on clinical data. The model aims to help clinicians make personalised treatment decisions, identifying which drug class would best lower glycated haemoglobin (HbA1c) over a 12 month period. The findings suggest that using this model can lead to significant improvements in glycaemic control and risks of long term complications, such as renal progression and microvascular diseases.
The study utilised observational data from the Clinical Practice Research Datalink (CPRD) in England, incorporating over 100,000 drug initiation events. The model was validated using data from three clinical trials and in two CPRD validation cohorts, one based on geography and the other on time period. The researchers matched patients receiving model-predicted optimal therapy with those on non-optimal therapy and observed a mean reduction in HbA1c of 5.3 mmol/mol (95% CI: 4.9–5.7) in the CRPD geographical validation cohort, and 5.0 mmol/mol (95% CI: 4.3–5.6) in the CRPD temporal validation group compared to the non-optimal group.
The results also indicated that patients on model-optimal optimal therapy had significantly improved outcomes, including a 38% lower 5 year risk of glycaemic failure (confirmed HbhA1c≥69 mmol/mol) (adjusted hazard ratio [aHR] 0.62; 95% CI: 0.59–0.64) and reductions in the 5-year risks of major adverse cardiovascular events or heart failure (MACE-HF) outcomes (aHR 0.85; 95% CI: 0.76–0.95), renal progression (aHR 0.7; 95% CI: 0.64–0.79), and macrovascular complications (aHR 0.86; 95% CI: 0.78–0.96). Despite these results, only 15.2% of drug initiations in the overall CPRD cohort were in line with the model’s predictions, highlighting the gap between model predictions and real-world practice.
In conclusion, the study demonstrates that a predictive model based on clinical data can significantly improve the management of T2D by helping clinicians choose the optimal treatment for individual patients. While the model shows promise in improving glycaemic control and reducing diabetes complications, its integration into clinical practice will need to address existing gaps, such as underuse of optimal therapies, and account for factors beyond glycaemia, including drug cost and availability.
Reference
Dennis JM et al. A five-drug class model using routinely available clinical features to optimise prescribing in type 2 diabetes: a prediction model development and validation study. Lancet. 2025;405(10480):701-14.
