Diabetes, a condition where your body has trouble managing the sugar (glucose) in your blood) is a long-term disease affecting millions of people globally. In the UK, about 90% of diabetics have type 2 diabetes (T2DM). This type of diabetes worsens over time, leading to poor blood sugar control, and can cause severe issues like heart attacks and kidney failure if not managed well.
Many T2DM patients need two or more medications to keep their blood sugar in check. Doctors choose these medications based on the patient's other health problems, personal traits, and preferences, using their knowledge and existing research. However, there's little research on the best timing for adding extra medications if blood sugar remains uncontrolled. Understanding if starting a drug earlier affects the risk of future complications is challenging, as such clinical trials would need many participants and often exclude patients with other conditions like depression or Alzheimer's disease.
To address these gaps, we can analyze past data to make informed decisions. This study aims to help doctors choose the best treatment for each patient based on their unique characteristics like age, Body Mass Index, other health conditions, and lab results.
Recent guidelines for type 2 diabetes (T2D) treatment emphasize the need to tailor pharmacotherapy based on patient comorbidities, rather than just HbA1c levels. Still, many physicians prefer metformin as the first-line treatment. Clinical decisions about intensifying diabetes therapy rely on research evidence, clinical expertise, and patient preferences, yet guidelines often exclude patients with severe or multiple comorbidities. We aim to develop a precision management approach by assessing individual risks associated with specific antihyperglycemic medications, potentially leading to a clinical decision support tool that improves personalized treatment and health outcomes. In this retrospective cohort study between 2012-2021, new users of metformin who either receive a first intensification step or stay on metformin until the occurrence of the outcomes of interest will constitute our study population. Since cardiovascular outcomes are one of the most important treatment outcomes to consider when prescribing antihyperglycemic therapies, we will focus on this outcome to initially test these hypotheses in the current study. We will obtain estimates for the probability of myocardial infarcton/stroke/all-cause mortality under each treatment strategy for each individual conditional on their characteristics using g-formula's. First, we will model the distribution of an outcome and covariates over time using maximum likelihood methods. Second, Monte Carlo simulation is used to approximate the weighted average under each treatment strategy. By generating indicators for a given outcome for individuals, a dataset is created which includes individuals trajectories had everyone followed treatment strategy. An interventional prediction model will be developed using the counterfactual trajectories calculated in the previous step. We will fit a model to the training sets (95% random sample of the counterfactual trajectories) simulated under each treatment strategy to predict the risk of outcome events had everyone received treatment under that strategy. We will use a nonparametric bootstrap to calculate prediction intervals for each individual risk estimate.
Myocardial infarction (MI), stroke, all-cause mortality
Daniala Weir - Chief Investigator - Utrecht University
Patrick Souverein - Corresponding Applicant - Utrecht University
David Liang - Collaborator - Utrecht University
Vasilis Antoniadis - Collaborator - Utrecht University
HES Admitted Patient Care;Practice Level Index of Multiple Deprivation