The Effect of Variable vs Fixed Feeding on Glycaemic Control in the Adult ICU: Virtual Trial Evaluation

Abstract

Elevated blood glucose (BG) concentrations (Hyperglycaemia) is a common complication in the adult intensive care unit (ICU), and is associated with increased morbidity and mortality. However, it has been shown that effective glycaemic control (GC) can reduce morbidity and mortality. STAR is a modelbased GC protocol that uniquely maintains normal BG by changing both insulin and nutrition interventions, and has been proven to be effective in controlling BG levels in the ICU. However, most GC protocols in the ICU only change insulin interventions, making the variable feed aspect of STAR less clinically desirable. This paper investigates two simpler feeding protocols as an alternative. Fixed feed (100% calorific goal) and stepped feed (60%, 80% and 100% calorific goal for the first 3 days of GC, and then 100% thereafter) protocols, in conjunction with the STAR protocol, are simulated with clinically validated virtual trials on a 221 virtual patient cohort. The GC safety, performance and workload is compared for each of the different feeding protocols. The variable and stepped feeding protocol achieved very similar GC performance and safety, with a per-patient median time in the targeted 4.4-8.0 mmol/L BG range of 89.0% vs. 88.3% respectively and the number of patients BG < 4.0 mmol/L being 77 vs. 78 respectively. In contrast, the fixed feeding protocol resulted in significantly poorer GC performance with 85.6% time in the targeted 4.4-8.0 mmol/L band (P<0.025). Both the fixed and stepped feeding protocols significantly reduced the number of feed changes required per day (6.4 variable vs. 0 fixed and 0.5 stepped, P<0.025). However, as a consequence a small increase in the number of BG measurements per day was seen (11.4 variable vs. 13.4 fixed vs. 12 stepped, P<0.025). Overall the stepped feeding protocol provides a simple alternative to the current variable feeding protocol, with similar GC safety and performance.