In recent years, analogs of human insulin have been engineered with the aim of improving therapy for people with diabetes. To ensure that the safety profile of the human hormone is not compromised by the molecular modifications, the toxico-pharmacological properties of insulin analogs should be carefully monitored. In this study, we compared the insulin and IGF-I receptor binding properties and metabolic and mitogenic potencies of insulin aspart (B28Asp human insulin), insulin lispro (B28Lys,B29Pro human insulin), insulin glargine (A21Gly,B31Arg,B32Arg human insulin), insulin detemir (NN304) [B29Lys(epsilon-tetradecanoyl), desB30 human insulin], and reference insulin analogs. Receptor affinities were measured using purified human receptors, insulin receptor dissociation rates were determined using Chinese hamster ovary cells overexpressing the human insulin receptor, metabolic potencies were evaluated using primary mouse adipocytes, and mitogenic potencies were determined in human osteosarcoma cells. Metabolic potencies correlated well with insulin receptor affinities. Mitogenic potencies in general correlated better with IGF-I receptor affinities than with insulin receptor off-rates. The 2 rapid-acting insulin analogs aspart and lispro resembled human insulin on all parameters, except for a slightly elevated IGF-I receptor affinity of lispro. In contrast, the 2 long-acting insulin analogs, glargine and detemir, differed significantly from human insulin. The combination of the B31B32diArg and A21Gly substitutions provided insulin glargine with a 6- to 8-fold increased IGF-I receptor affinity and mitogenic potency compared with human insulin. The attachment of a fatty acid chain to LysB29 provided insulin detemir with reduced receptor affinities and metabolic and mitogenic potencies but did not change the balance between mitogenic and metabolic potencies. The safety implications of the increased growth-stimulating potential of insulin glargine are unclear. The reduced in vitro potency of insulin detemir might explain why this analog is not as effective on a molar basis as human insulin in humans.