We have compared insulin concentrations measured by radioimmunoassay (RIA) in plasma from 50 fasting non-insulin-dependent diabetics (NIDDM) with those measured by a new monoclonal antibody-based two-site immunoradiometric assay (IRMA) of insulin (which has no significant cross-reaction with proinsulin-like molecules). We find that the RIA measures the sum of the insulin and proinsulin like molecules and that the IRMA insulin concentrations are 38% of those measured by the RIA in those diabetic subjects. We conclude that the importance of insulin deficiency in NIDDM may have been obscured by this error.