The plasma protein binding of three model compounds was investigated using a variant of equilibrium dialysis, denoted comparative equilibrium dialysis (CED), and the results were compared with those obtained with ultrafiltration (UF). In CED, the buffer that the plasma is dialysed against in traditional equilibrium dialysis is replaced by, for example, plasma from other species. The CED method has the advantage that the unbound concentration (C(u)) does not need to be measured, which can be difficult for drugs with extremely small unbound fractions. Instead, the ratio of the total drug concentration (C(tot)) on either side of the dialysis membrane at equilibrium is a direct measure of the relative binding properties of the two plasma types. For the first model compound, having an unbound fraction (f(u)) of about 0.05% in human plasma, the time to reach equilibrium was too long (> or =40 h) to make the CED technique feasible in practice. For the second model compound, the more weakly bound drug NAD-299 (with an unbound fraction of about 2% in human plasma), the CED equilibration times were considerably shortened (< or =16 h), and the technique was applied to plasma from three different species. Large discrepancies between the CED and UF results were seen, CED always giving rise to much lower C(tot) differences than expected from the UF results. It is suspected that this discrepancy was due to equilibration between the dialysis chambers of all plasma components with a molecular weight less than the cut-off of the membrane. This equilibration causes altered binding properties compared to the initial plasma. When performing ultrafiltration on plasma where drug was added to untreated plasma or added to blank plasma that was equilibrated against plasma from the same or from another species, the change of binding properties was confirmed. To ensure that the results were not specific for NAD-299, a third model compound, tolterodine, was also included. The same trends as for NAD-299 were seen. Because of the long equilibration times for compounds with high protein binding and, in particular, the suspected partial mixture of low molecular weight compounds from the two plasma types and the subsequent change of binding properties, we cannot recommend the CED method as a tool for studying relative protein binding.