Diabetes Mellitus is a chronic disease that affects an increasing number of people worldwide. The condition is characterised by high blood glucose levels, which can lead to several complications at short and long term, when they are not controlled. Clinicians need to adjust each patient’s treatment to the evolving condition over time, considering the individual characteristics and the high number of treatment alternatives available. Lifestyle modification and moderate weight loss are recommended both to prevent and delay the appearance of type 2 diabetes and its progression, while the use of monitoring devices is recommended to follow progression of glucose levels in both types of diabetes. The right choice of combined treatment alternatives is essential to avoid worsening of health outcomes which might need costly care interventions such as hospitalization or emergency events. This study aims to study whether patients can be grouped intro subtypes of diabetes to characterise treatment response, disease progression and costs; and to extract patterns of data to predict the best sequence of treatment for each patient to optimize long-term outcomes. Outcomes include glycaemic control, mortality, risk of complications and cost-effectiveness of treatment interventions.
This study will provide important information to help clinical decision making in practice and health technology assessment organizations in policy making.
A. Technical Summary (Max. 300 words)
Diabetes is a highly heterogeneous disease that needs to be managed by clinicians in a long-term basis following different sequences of treatment, mainly pharmacological, but also considering lifestyle and educational factors. Even though clinical guidelines provide recommendations about the most appropriate drugs for therapy adjustments, patient stratification and individual characteristics can help to optimise treatment and prevent disease progression and associated complications. This study aims to use routinely collected data from the Clinical Practice Research Datalink to extract meaningful patterns about subgroups of patients with diabetes and to predict effectiveness and cost-effectiveness of treatment sequences based on deep learning algorithms. Patients will be followed from one year before the first antihyperglycemic drug is prescribed and the sequence of subsequent treatments will be considered to predict effectiveness in terms of improvement of glycaemic control (changes in HbA1c), mortality and the risk to develop adverse events at prediction horizons of 1, 3, 5 and 10 years.
Patients will be linked to Hospital Episode Statistics to determine resource utilization and to calculate incremental cost-effectiveness ratios. This research will create a prognostic score to provide personalized treatment recommendations at baseline and help to allocate patients to the most beneficial treatment sequence. This could potentially help to improve decision-making of clinicians, better health outcomes for individuals, and reduced health care costs.
Hospital Episode Statistics (HES) Admitted Patient Care (APC) data will be used to determine resource use (Hospital admission, length of hospital stay, emergency admission) from hospital episode information. Patient Level Index of Multiple Deprivation will be used to characterize patients from a socioeconomic point of view.
Primary outcomes: change in HbA1c and all cause mortality.
Secondary outcomes: Frequency and severity of hypoglycaemia; frequency of abnormal fasting blood glucose values, change in body weight; changes in cardiovascular risk factors ((LDL-C, HDL-C, TG and TC) and blood pressure); retinopathy (specific lesions or macular changes, referable retinopathy, blindness/loss of vision, visual acuity); chronic kidney disease (eGFR, serum creatinine, proteinuria, microalbuminuria, dialysis, transplant); cardiovascular disease (myocardial infarction, heart failure, stroke, acute coronary syndrome, transient ischemic attack, peripheral arterial disease, revascularisation, stenting); foot complications (amputations, diabetic foot ulcers, Charcot osteoarthropathy, diabetic foot infection); rheumatoid arthritis; depression; dementia, severe mental illness (SMI); cognitive impairment; diabetic autonomic neuropathy; chronic obstructive pulmonary disease (COPD) disease; progression to insulin treatment. Healthcare resource use and associated costs (Hospital admission, length of hospital stay, primary care visits, medication use, emergency room visits registered in GP visits), use of medical devices (glucose meter, continuous insulin infusion system, continuous glucose monitoring, intermittent glucose monitoring)
Wim Goettsch - Chief Investigator - Utrecht Institute for Pharmaceutical Sciences
Patrick Souverein - Corresponding Applicant - Utrecht University
Francisco Javier Somolinos Simón - Collaborator - Universidad Politécnica de Madrid (Polytechnic University of Madrid)
Gema Garcia Saez - Collaborator - Universidad Politécnica de Madrid (Polytechnic University of Madrid)
Jose Tapia Galisteo - Collaborator - Universidad Politécnica de Madrid (Polytechnic University of Madrid)
Junfeng Wang - Collaborator - Utrecht University
Li Jiu - Collaborator - Utrecht Institute for Pharmaceutical Sciences
Maria Elena HERNANDO - Collaborator - Universidad Politécnica de Madrid (Polytechnic University of Madrid)
HES Admitted Patient Care;Patient Level Index of Multiple Deprivation