New compounds showing hypoglycaemic activity have been designed through a computer aided method based on quantitative structure-activity relationship (QSAR) and molecular connectivity. After calculation of topological indices for a set of 89 compounds including active and inactive with regards to hypoglycaemic action, linear discriminant analysis was performed so that a useful model to predict such an activity was achieved. Later on, the discriminant model was applied on a huge database so that fourteen compounds were selected as potential new hypoglycaemics. From them, just five were finally selected for experimental test on expected hypoglycaemic activity. Among the selected compounds, l-arabitol, Acid blue 161, 1,4-butanediol diglycidil ether and Acid red 151 stand out, which are comparable in potency to standard drugs such as tolbutamide. Acid blue has a glycaemia profile similar to that of tolbutamide but does not lead to a severe hypoglycaemic condition, while the profile of the other agents is near normality.