OBJECTIVES: Validation of a novel glucose-insulin-potassium (GIK) infusion algorithm to optimize metabolic control in hospitalized diabetic patients. SUBJECTS: We randomized 33 diabetic patients admitted to Sahlgrenska University Hospital with acute internal medicine diseases to either GIK infusion or multiple injection therapy (MIT). The GIK infusion rate and the MIT were controlled according to special algorithms. The treatment efficacy was evaluated through comparisons of capillary blood glucose eight times on day 2 of the study. RESULTS: The GIK infusion led to significantly lower mean blood glucose when compared with MIT [10.1 (9.0-11.2) vs. 12.3 (9.3-14.4) mmol L(-1), median and interquartile range, P < 0.01]. Four episodes of hypoglycaemia without loss of consciousness were recorded in the GIK group whereas no hypoglycaemic event occurred in the MIT group. A mean of 1 (1-3) episodes of blood glucose levels above 12.0 mmol L(-1) were recorded in the GIK group compared with 3.5 (1.5-5.0) in the MIT group, P < 0.01. CONCLUSIONS: The algorithm used for the GIK infusion gave an acceptable level of metabolic control and this insulin infusion protocol is safe enough to be used by the nursing staff on a general internal medicine ward.