AIMS: Glycaemic variability (GV) is associated with mortality in acutely ill patients, but the mechanism is unknown. The objective of this study is to determine whether common approaches to insulin therapy have distinct effects on GV and autonomic tone. METHODS: Hospitalized patients with diabetes were randomized to short-term intravenous (IV) or physiologic subcutaneous (SQ) insulin. Heart rate variability (HRV) and cardiac impedance (pre-ejection period, PEP) were used to estimate parasympathetic and sympathetic tone, respectively. GV was measured using a continuous glucose monitor. RESULTS: Mean glucose tended to be lower initially in the SQ group (N = 16) compared with the IV group (N = 17) on day 1 (10.5 vs. 8.6 mmol/l, p = 0.05), but became non-significant during the transition off of the infusion. There was no difference in glycaemic lability index (GLI), continuous overlapping net glycaemic action (CONGA) or coefficient of variation (CV) on day 1, but by day 2, these measures were higher in the IV group (p < 0.05 for all). PEP was higher in the SQ group during (110 vs. 123 ms, p = 0.02) and after the intervention (104 vs. 126 ms, p = 0.004). Hypoglycaemia was similar in both groups. There were only small differences in HRV. Post-treatment PEP was inversely correlated with log GLI (r = -0.41, p = 0.03) but not other measures. CONCLUSIONS: Short-term IV insulin is associated with an increase in multiple GV measures compared with optimal SQ insulin. However, GLI was the only predictor of PEP. Further research is needed to determine if interventions that minimize GV improve outcomes in the hospital.