Loading ...

TOPLINE:

Researchers developed a model using routine clinical data to identify optimal glucose-lowering therapies for patients with type 2 diabetes (T2D). Individuals receiving model-predicted optimal therapy showed better glycaemic control, a lower need for additional glucose-lowering therapy, and a lower risk for diabetes complications than those on non-optimal therapy.

METHODOLOGY:

• Researchers developed and validated a five–drug class model to predict the relative glycaemic efficacy of five glucose-lowering drug classes: Dipeptidyl peptidase-4 inhibitors, glucagon-like peptide 1 receptor agonists (GLP-1 RAs), sodium-glucose co-transporter-2 inhibitors, sulfonylureas, and thiazolidinediones.
• They included data from the Clinical Practice Research Datalink (CPRD) database on patients with T2D aged 18-79 years who initiated one of five drug classes between 2004 and 2020; a total of 100,107 drug initiations were used to develop the model.
• The model utilised data on nine routinely available clinical features at the time of drug initiation, namely, age, duration of diabetes, sex, baseline A1c levels, body mass index, estimated glomerular filtration rate, high-density lipoprotein cholesterol levels, total cholesterol levels, and alanine aminotransferase levels.
• The validation assessed differences in observed glycaemic effectiveness between matched (1:1) groups receiving therapy that either aligned (concordant) or did not align (discordant) with the model-predicted optimal therapy.
• Optimal therapy was defined as the use of the drug class with the highest predicted glycaemic effectiveness.
• The primary outcome was the absolute 12-month A1c level after drug initiation; long-term outcomes included risks for glycaemic failure, all-cause mortality, major adverse cardiovascular events, renal progression, and microvascular complications.

TAKEAWAY:

• In the overall CPRD cohort (combined development and validation cohort), 32,305 (15.2%) of 212,166 drug initiations were identified as model-predicted optimal therapy.
• GLP-1 RAs were the most frequently predicted optimal drug class for achieving the best A1c levels at 12 months.
• In the development cohort, treatment initiations that aligned with the model's predictions resulted in a mean 12-month A1c level that was 5.3 mmol/mol lower than that with discordant treatment initiations. This difference closely matched the observed reduction of 5.1 mmol/mol (95% CI, 4.7-5.5) between the two groups.
• The 5-year risk for glycaemic failure was 38% lower in the model-concordant group than in the model-discordant group (adjusted hazard ratio, 0.62; 95% CI, 0.59-0.64). The model-concordant group also showed lower risks for major adverse cardiovascular events or heart failure, renal progression, and microvascular complications.

IN PRACTICE:

"The model is based solely on routinely collected clinical parameters, supporting low-cost application worldwide," the authors wrote.

"By addressing the aforementioned limitations, this model has the potential to transform personalised diabetes care globally, particularly in primary care, where it could alleviate workload pressures and enhance patient outcomes," experts wrote in an accompanying editorial.

SOURCE:

This study was led by John M. Dennis, PhD, Clinical and Biomedical Sciences, University of Exeter Medical School, Exeter, United Kingdom, and was published online on February 25, 2025, in The Lancet.

LIMITATIONS:

The study model was validated using clinical trial data for several drug classes; however, there was insufficient access to active comparator trial data for GLP-1 RAs, which may have affected the model's robustness for this drug class. The analysis relied on routine clinical data, which may have introduced bias or misclassification in defining predictors and study outcomes. Due to limited real-world data on treatment outcomes, the study model excluded paediatric patients and individuals aged 80 years or older.

DISCLOSURES:

This study was funded by the UK Medical Research Council. Some authors reported receiving research funding, support, grants, or personal fees from various pharmaceutical companies and institutions.

This article was created using several editorial tools, including AI, as part of the process. Human editors reviewed this content before publication.