Increased
glucose variability (GV) is an independent risk factor for mortality in the
critically ill; unfortunately, the optimal
insulin therapy that minimizes GV is not known. We simulate the
glucose-
insulin feedback system to study how stress
hyperglycemia (SH) states, taken to be a non-uniform group of physiologic disorders with varying
insulin resistance (IR) and similar levels of
hyperglycemia, respond to the type and dose of subcutaneous (SQ)
insulin. Two groups of 100 virtual patients are studied: those receiving and those not receiving continuous enteral feeds. Stress
hyperglycemia was facilitated by doubling the gluconeogenesis rate and IR was stepwise varied from a borderline to a high value.
Lispro and
regular insulin were simulated with dosages that ranged from 0 to 6 units; the resulting GV was analyzed after each
insulin injection. The numerical model used consists of a set of non-linear differential equations with two time delays and five adjustable parameters. The results show that
regular insulin decreased GV in both patient groups and rarely caused
hypoglycemia. With continuous enteral feeds and borderline to mild IR,
Lispro showed minimal effect on GV; however, rebound
hyperglycemia that increased GV occurred when the IR was moderate to high. Without a nutritional source,
Lispro worsened GV through frequent
hypoglycemia episodes as the injection dose increased. The inferior performance of
Lispro is a result of its rapid absorption profile; half of its duration of action is similar to the
glucose ultradian period. Clinical trials are needed to examine whether these numerical results represent the
glucose-
insulin dynamics that occur in intensive care units, and if such dynamics are present, their clinical effects should be evaluated.