Statistical considerations in the quantitation of serum immunoglobulin levels using the enzyme-linked immunosorbent assay (ELISA).

Several methods for analyzing ELISA data have been evaluated using optical density values derived by reacting serial two-fold dilutions of a rhesus monkey (Macaca mulatta) reference serum with dilutions of peroxidase-conjugated monospecific antisera to human IgG and IgM in the micro-ELISA 'sandwich' technique using microplates coated with appropriately diluted antisera to human IgG and IgM. Representation of optical density as a linear function of log serum dilution was shown to be inappropriate and potentially misleading. Weighted non-linear least squares analysis of a 4-parameter logit was demonstrated to be inappropriate because it required the specification of a somewhat arbitrary variance function and weight estimates varied substantially depending on the function used. Representation of optical density as a 4-parameter logistic function with estimation carried out on the log scale was shown to be the most appropriate procedure for determining the concentration of antigens or antibodies by the ELISA method.