We measured the free thyroxin (FT4) concentration in serum from 79 patients with nonthyroidal illnesses (NTI), using two analog RIAs and two calculation methods, and compared the results with those by ultrafiltration (UF) as a reference. The percentage of NTI patients with subnormal concentrations of FT4 was 10.1% by UF; 20.3% by mathematical calculation (MC) according to mass-action theory; 19.0% by FT4 index (FT4I), calculated by the triiodothyronine-albumin uptake method expressed as the bound:free ratio; 37.2% by Amerlex M FT4 (P less than 0.001 vs UF, P less than 0.025 vs MC); and 33.3% by Coat-A-Count FT4 with albumin blocker (P less than 0.001 vs UF, P less than 0.025 vs MC). No significant correlations were found between FT4 (UF) and T4-binding proteins (TBP) in serum, but FT4 (MC) correlated with T4-binding globulin (TBG) (r = 0.383, P less than 0.001), FT4I with TBG (r = 0.345, P less than 0.001), FT4 (Amerlex M) with albumin (r = 0.491, P less than 0.001), and FT4 (Coat-A-Count) with TBG (r = 0.636, P less than 0.001). All FT4 measurements correlated well with each other, but correlations were stronger between both calculated FT4 values and FT4 (UF), r = 0.702 to 0.760, than between both values for RIA and FT4 (UF), r = 0.475 to 0.665. These results indicate that in comparison with the UF method, both calculated FT4 values are little affected by TBP concentrations, and FT4 (MC) or FT4I by T3-albumin uptake may be more appropriate than analog RIA for evaluation of circulating FT4 concentration in NTI.