Fewer than 20% of type 2 diabetes patients on most effective medicine for them, model shows

A model designed to predict the best medicines to choose for patients with type 2 diabetes (T2DM) after metformin has found that only 18% of patients were taking the optimal medicine for them.

Results of a study, published in The Lancet on 25 February 2025, show that “in recent years since 2019, only 3,986 (17.8%) of 22,452 therapy initiations in England were of the model-predicted optimal drug class in clinical practice”.

The study authors added that the model, which optimises therapy, was associated with a 38% reduction in the likelihood of glycaemic failure over five years, “which would substantially increase the time on stable glucose-lowering therapies before additional intensification is required”.

Metformin is the most common first treatment for T2DM, but there a five other glucose-lowering medicines available: dipeptidyl peptidase-4 inhibitors, glucagon-like peptide-1 receptor agonists, sodium–glucose co-transporter-2 inhibitors, sulfonylureas and thiazolidinediones.

The model, developed with funding from the Medical Research Council and National Institute for Health Research, predicts the relative glycaemic effectiveness — in terms of the glycated haemoglobin (HbA_1c) level after 12 months — of the five alternative medicines.

It was developed using data from the GP and hospital records of 1 million people in the UK with T2DM, along with 100,107 drug initiations from the Clinical Practice Research Datalink (CPRD).

The model used nine clinical characteristics, including the age, sex, BMI and total cholesterol of people with T2DM, to help predict response to the medicines. It was then validated against observational data of people with T2DM who were initiated on one of the medicines between 2004 and 2020 in England, as well as individual-level data from three randomised medicines trials in T2DM.

“In the CPRD cohorts from England, we found an approximate 5 mmol/mol benefit in 12-month HbA_1c in taking model-predicted optimal therapy,” the authors said.

They noted that the model “is aimed at directly informing clinical decisions on treatment, and unlike alternative approaches, is not aimed at advancing aetiological understanding of diabetes”.

The authors added that before the model is implemented in a specific population or setting, “we recommend that model performance is carefully assessed, with a particular focus on the average glycaemic effectiveness of each drug class, which might be population-specific”.

John Dennis, associate professor of health data science at the University of Exeter, who led the study, said: “We have developed a completely new personalised approach for diabetes treatment that could benefit everyone with T2DM in the UK and worldwide.

“For the first time, our model allows people living with T2DM to quickly identify the best treatment to manage their blood sugar levels, helping reduce their risk of diabetes complications. This offers a major advance on the current approach to choosing diabetes medications.”

Andrew Hattersley, professor of molecular medicine at the University of Exeter and one of the study’s authors, said: “Critically, our model can be implemented in clinical care immediately and at no additional cost. This is because it uses simple measures such as sex, weight and standard blood tests that are performed routinely.

“We hope that we can roll out the model quickly to make it available to help people with T2DM in the UK and across the world.”

Elizabeth Robertson, director of research and clinical at Diabetes UK, said: “This innovation using routine clinical data could help countless people with type 2 diabetes to get their blood sugars levels into a safe range, significantly reducing their risk of devastating diabetes complications and easing the burden of living with this relentless condition.

“If shown to be effective in practice and widely adopted by health services in the UK and globally, this tool could mark the most significant advance in type 2 diabetes care in more than a decade, improving health outcomes for millions.”  

Data published in June 2024 by the National Diabetes Audit Programme showed that the number of patients at risk of T2DM in England had increased by more than half a million in 2023, compared with the year before.