One approach of improving metabolic control in type I diabetic patients is the application of computer-aided procedures aimed at supporting the decision on optimal therapeutic regimens. To accomplish this, a complex strategy was developed which in an individual patient permits (1) the evaluation of metabolic data by means of statistical and graphical methods, and (2) the prediction of the outcome in feedback and in non-feedback-controlled insulin therapy. The latter is realized by means of simulation, employing a structured model of the glucose-insulin control system where the model parameters can either be identified individually or be taken at random. The practical applicability was validated in C-peptide-negative type I diabetic patients who were on intensified insulin injection therapy. The comparison between theoretical predictions and daily glycaemic profiles measured by the patients under ambulatory conditions showed close correspondence which justifies the application of this method as a clinical decision support.