Insulin sensitivity tools for critical care.
Thesis DisciplineMechanical Engineering
Degree GrantorUniversity of Canterbury
Degree NameMaster of Engineering
Stress induced hyperglycaemia is prevalent in critical care. Since the landmark paper published by Van den Berghe et al. (2001) a great deal of attention has been paid to intensive insulin therapy in an ICU setting to combat the adverse effects of elevated glucose levels and poor glycaemic control. Glycaemic control protocols have been extensively developed, tested and validated within an ICU setting. However, little research has been conducted on the effects of a glycaemic control protocol in a less acute ward setting. There are many additional challenges presented in a ward setting, such as the variation in meals and levels of activity between patients, from day to day and throughout the day.
A simple compartment model is used to describe the nature of insulin and glucose metabolism in patients of the Cardiothoracic Ward (CTW). A stochastic model of the fitted insulin sensitivity parameter is generated for this cohort and validated against cohorts of similar characteristics. The stochastic model is then used to run simulations of predictive control on 7 CTW patients, which shows significantly tighter glucose control than what is obtained with regular clinical procedures. However, the rate of severe hypoglycaemia is an unacceptably high 4.2%. The greatest challenge in maintaining tight glycaemic control in such patients is the consumption of meals at irregular times and of inconsistent quantities.
Insulin sensitivity was compared to extensive hourly clinical data of 36 ICU patients. From this data a sepsis score of value 0-4 was generated as gold standard marker of sepsis. Comparing the sepsis score to insulin sensitivity found that insulin sensitivity provides a negative predictive diagnostic for sepsis. High insulin sensitivity of greater than Si = 8 x 10⁻⁵ L mU⁻¹ min⁻¹ rules out sepsis for the majority of patient hours and may be determined non-invasively in real-time from glycaemic control protocol data. Low insulin sensitivity is not an effective diagnostic, as it can equally mark the presence of sepsis or other conditions.