Most immunoassays applied to drugs in human plasma do not use an extraction of analyte. To compensate for interferences due to plasma proteins or salts, standards are prepared in drug-free plasma. Because the concentration of plasma components varies from one subject to another, it is likely that the drug-free plasma is not representative of the potential interference in each plasma. Using two immunoassays, for a steroid (nomegestrol acetate) and a heptapeptide (BN 52080), the authors have shown that tracer binding to the antibody may vary significantly between plasma from different subjects. Intersubject variability of tracer-antibody binding was 21.6% (coefficient of variation for 25 subjects) for nomegestrol acetate. When the same plasma were spiked with the steroid at a concentration corresponding to the central part of the standard curve, the recovery was between 39 and 215%. Intersubject variability in tracer binding was lower (7.7%) for the peptide immunoassay, but still affected accuracy. The authors show that this problem is common to direct immunoassays for other drugs and must be solved in assay development.