Proposing patient-specific diabetes treatments using a mathematical theory of disease progression

Type 2 diabetes (T2D) is a disease where control of blood glucose fails, and the 8th leading cause of death in the world. A typical T2D trajectory is slow and irreversible. Such a trajectory can be understood using a mathematical model which combines a fast, negative feedback loop (where insulin controls glucose) with slow (months–years) pathogenesis processes. Traditionally, all T2D patients are given similar treatment, adjusted on the scale of months, consistent with this trajectory. We focus on a subtype of T2D, called Ketosis-prone diabetes (KPD) in which symptoms can develop rapidly and are partially reversible. We showed that these trajectories cannot be explained by existing models of T2D pathogenesis, and thus argue that this subtype must involve a new cellular mechanism, where insulin-secreting cells are temporarily deactivated by high glucose on an intermediate timescale (days–weeks). This intermediate timescale is neglected both in existing research datasets and current treatment strategies. We thus proposed and carried out a clinical study, using continuous glucose monitors (CGMs) to monitor glucose on the intermediate timescales necessary to test our model. Our model fits these data well, thus allowing us to extract patient-specific parameters describing heterogeneity in the mechanism of pathogenesis. Thus, our model proposes patient-specific treatment protocols, continuously adjusted in response to new CGM data.