We performed a methodological comparison of free triiodothyronine (FT3) estimates in patients with liver cirrhosis and renal failure. Patients were classified in terms of severity of illness on the basis of their total triiodothyronine, total thyroxine and reverse triiodothyronine profiles. FT3 levels, measured in direct dialysis, microchromatography, labelled analogue and two-step immunoextraction assays were significantly (P < 0.01) lower than the control group in all patient categories. However, FT3 measured by a labelled antibody radioimmunoassay was significantly reduced only in the most severely ill sub-group of patients. In a further group of patients on long-term amiodarone therapy for cardiac disease all FT3 methods, with the exception of the labelled antibody radioimmunoassay and an analogue method, yielded significantly (P < 0.01) reduced levels. A significant negative association between FT3 and subject age was demonstrated for all methods except the labelled antibody radioimmunoassay, and a weak but significant negative correlation between log thyrotropin and FT3 was only seen with this assay. Three methods demonstrated a correlation (P < 0.02) with albumin levels in patients with the 'low T3 syndrome'. In this group, albumin had a predictive value (P < or = 0.02) for four out of six assays as determined by stepwise variable selection. Our findings suggest that users of FT3 assays should exercise caution in interpreting results in non-thyroidal illness and amiodarone treated patients, as there are method-related differences in the profiles obtained.