In the present study our interests focused on the evaluation of a high capacity assay (MEIA) which allows true insulin determinations in the absence of cross-reactivity with proinsulin-like molecules. This method was compared to a commercially available radioimmunoassay (RIA) for insulin determination. As the latter gives insulin levels which represent a mixture of insulin and proinsulin-like molecules, the proinsulin-like molecules were quantitated by subtracting the true insulin levels measured using MEIA from the total insulin levels obtained using RIA. These methods were applied for the analysis of blood samples drawn in 63 normal subjects, 16 obese subjects, 3 patients submitted to islet transplantation and 4 patients with insulinoma. The MEIA was precise, fully automated and time-saving, making its application on a routine basis particularly attractive. MEIA and RIA were equally able to correctly quantify human insulin molecules. On the contrary, the antibody present in the true insulin assay did not interact with proinsulin-like molecules, which were recognized even in the presence of increasing insulin levels. In normal subjects, the true and total insulin levels in the fasting state and at the time peak after glucose- or arginine-induced endogenous insulin release were well correlated at r = 0.88 and 0.89, respectively. Interestingly, total insulin values were overestimated by 10%-16% as compared with true insulin levels, which represent proinsulin values superimposable on previously reported data. Proinsulin-like molecules made up 50% of the total insulin in obese and transplanted patients, and about 70% in patients with insulinoma.(ABSTRACT TRUNCATED AT 250 WORDS)