Type 2 Diabetes

Mats Karlsson, Maria Kjellsson

Diabetes is a chronic disease, affecting more than 220 million people worldwide and the “diabetic epidemic” is projected to affect 366 million people by 2030, of which more than 90% will suffer from type 2 diabetes (T2D). T2D occurs when the body does not effectively respond to insulin and is unable to produce enough insulin to account for the inefficient use of insulin. The result is elevated blood glucose which is toxic and eventually lead to complications; e.g. cardio-vascular diseases (CVD) and chronic kidney disease (CKD). The aim with most anti-diabetic treatment is symptomatic, bringing glucose down to healthy concentrations. Diagnosis of diabetes is mainly based on fasting plasma glucose (FPG) but also glycosylated haemoglobin (HbA1c). The success of treatments is assessed on both FPG and HbA1c but also on dynamic glucose after provocation studies. Our research stands on three legs: 1) models of dynamic glucose after glucose provocations, 2) models of dynamic HbA1c and 3) models of long-term complications.

Provocation studies are used to characterize the functionality of the glucose-insulin system and could vary greatly in design from clamping of glucose or insulin by variable rate infusions, graded glucose infusions, intravenous bolus or oral administration of glucose or meals challenges. We have developed several integrated models with simultaneous analysis of dynamic glucose and insulin after such provocations. These models, which include production, disposition and control (homeostatic) mechanisms of the system, have shown to realistically simulate outcomes of short dynamic glucose provocations at the raw data level. These models have been developed to describe both healthy and pre-diabetic subjects as well as patients with T2D and have been coupled with models of incretin hormone secretion, sub-cutaneous insulin absorption, glucose gut absorption, pre-hepatic insulin and hepatic extraction ratio of insulin. We are focusing our current research on including mechanism of glucagon release and glucagon effects on glucose and insulin as well as a fully mechanistic whole-body integrated glucose homeostasis model. This model should include elements of disease progression from healthy to overtly diabetic through the pre-diabetic state.

Long-term clinical trials in T2D patients mainly focus on HbA1c. HbA1c is the fraction of the haemoglobin, in red blood cells (RBC) that has been glycosylated. This is a naturally occurring reaction depending on the amount of glucose in plasma; the higher the glucose concentration, the higher the HbA1c. As the life-span of RBCs ranges from 2 to 4 months, the HbA1c supplies a measurement of the sustained glycaemic control. Several models have been developed in-house to describe the relationship between HbA1c and FPG or mean plasma glucose (MPG) either using empirical descriptions or mechanistic approaches using knowledge about RBC life-span. Also models acknowledging the mechanism of insulin sensitivity, glucose production and disposition and changes in beta-cell mass or function in relation to weight loss has been developed. Currently we are investigating how study design can be improved using model-based analysis.

The overall endpoint of most anti-diabetic treatments is to lower the risk of long-term complications, such as CVD, retinopathy and CKD. Long term studies commonly involve assessments of the risk of CVD or CKD progression in relation to elevated levels of HbA1c or FPG. We are developing parametric risk models, using data from the National Diabetes Registry (NDR, Sweden) to characterise the relationship between CVD/CKD and time-varying covariates such as HbA1c and other biomarkers, i.e. blood pressure and weight, with the aim to predict CVD/CKD after changes in key biomarkers.

All models have been developed for the purpose of being used to quantify changes in the system following interventions (drug administration, diet changes, etc.) and associate these changes with known or hypothesized mechanisms of impact of the system. Further the models are intended as tools for hypothesis generation regarding single or combined interventions as well as clinical trial design optimization.