Adding insulin directly into infusion bags seems to be a useful method for controlling hyperglycemia in patients under ternary parenteral nutrition (TPN). Its efficacy is assessed by glycemic monitoring but few data are available on insulin stability in this situation. Among the various methods for quantifying insulin levels in human serum, the immunoassay ones seemed potentially appropriate for a TPN admixture containing high lipid concentrations. We sought to identify and validate which of two immunoassay methods was the better to quantify human insulin and consequently be adapted to studying its stability in a TPN admixture. Two immunoassay methods to quantify recombinant human insulin were assessed in industrial TPN: an immunoradiometric assay (IRMA) and an immunoelectrochemiluminometric assay (IECMA). Validation trials for both methods were based on the accuracy profile method. Interference with immunometric assays due to the high lipidic content of TPN was eliminated through an improved preparation protocol using a bovine serum albumin (BSA) diluted in phosphate buffer saline (PBS). The relative total error of IECMA varied from 1.74 to 4.52% while it varied from -0.32 to 8.37% with IRMA. Only IECMA provided an accuracy profile with a 95% confidence interval of calculated-tolerance limits falling between the chosen acceptance limits (i.e., total error ≤±10%). IECMA combined with a BSA dilution is a simple and semi-automatic method that provides an accurate quantification of human insulin in a TPN admixture without any interference from lipids.