Quantification of analogues of human insulin in biological matrices is complicated by differences in their immunoreactivity and the presence of both the analogue and endogenous concentrations of insulin in test samples. To facilitate pharmacokinetic comparisons of carboxyl-terminal B-chain analogues of human insulin, we undertook development of a sensitive ELISA. The ELISA detection method was optimized systematically to permit routine analysis of 10-microl serum samples. Accordingly, a noncompetitive 'sandwich' chemiluminescent ELISA was validated for the quantification of carboxyl-terminal B-chain insulin analogues in human serum over a concentration range from 5 to 3125 pM. The mean bias (RE%) within the validated range varied from -10.3 to 4.3%, with an intermediate precision (inter-assay CV%) from 4.2 to 11.5%. The two-sided 90% expectation tolerance interval for total measurement error was within +/-25% of the nominal concentration for all levels of validation samples. Insulin lispro, human insulin, proinsulin, despentapeptide insulin (DPI) and porcine insulin displayed comparable crossreactivity in the ELISA. Potential utility of the new assay for insulin bioanalysis in nonhuman species was investigated by assessing the pharmacokinetic profile of DPI in rats following administration of a single subcutaneous dose. The sensitive chemiluminescent detection method is simple to perform and should be readily adaptable for ELISAs of other therapeutic proteins.