Subcutaneous (s.c.) injections of identical insulin doses may lead to considerable intra- and inter-individual differences in the current metabolic control of patients with diabetes mellitus. This well-known variability of the metabolic effect of insulin hampers practical insulin therapy considerably. The aim of this review is to summarize the knowledge about this topic, with a special focus on the variability of insulin action after pulmonary administration of insulin. A number of studies have been published describing the variability of insulin absorption from the s.c. depot. Only in a few published studies has the variability of insulin action after s.c. administration been quantified. Under controlled experimental conditions s.c. injections of regular insulins result in an intra-individual coefficient of variation (CV) of 15-25% of certain pharmacodynamic summary measures--which characterize the metabolic effect of the applied insulin--in healthy subjects. The inter-individual variability was approximately 10% higher than the intra-individual variability. Subcutaneously injected intermediate- and long-acting insulin preparations were described to have an even greater variability (> 50%) than subcutaneously injected regular insulin. However, in a glucose clamp study s.c. application of NPH insulin led to an intra-individual CV in the range of 12-45% in healthy subjects. The reason for this discrepancy might be that the NPH insulin suspension was sufficiently shaken prior to drawing up the dose. Compared with conventional insulin formulations, rapid- and long-acting insulin analogues appear to have a similar variability, which means that, unfortunately, no considerable advantages in terms of variability were achieved by the invention of these novel insulin preparations. There are no appropriate studies available investigating the variability of the metabolic effect after s.c. insulin administration in patients with diabetes. The inhalation of insulin is a novel form of insulin administration that is currently under clinical development. The variability of the metabolic effect induced by the inhalation of insulin has up to now only been investigated in a small number of (published) studies. In a glucose-clamp study with healthy subjects the inhalation of an identical insulin dose on three study days led to an intra-individual variability that was comparable to that after s.c. injection of regular insulin. In a dose-response study with patients with type 1 diabetes the intra-individual CV was 34% for the area under the curve of the glucose infusion rate for 0-10 h. Studies with patients with type 2 diabetes have shown that the intra-individual CVs were within the range seen after s.c. insulin administration or even lower. In summary, the intra-individual variability of the metabolic effect observed after insulin application, be it subcutaneously injected or be it inhaled, is considerable and, therefore, hampers practical diabetes therapy. To date no means have been found that could lead to a clinically relevant reduction in the variable metabolic effect.