We studied various tests of thyroid function in sick patients with nonthyroidal illness (NTI) in order to determine the utility of each test for differentiating these patience from a group with hypothyroidism. We evaluated each test in 22 healthy volunteers who served as controls, 20 patients with hypothyroidism, 14 patients admitted to medical intensive care unit whose serum T4 was less than 5 micrograms/dl, 13 patients with chronic liver disease, 32 patients on chronic hemodialysis for renal failure, 13 ambulatory oncology patients receiving chemotherapy, 16 pregnant women, 7 women on estrogens, and 20 hyperthyroid patients. On all samples, we measured serum T4, the free T4 index by several methods, free T4 by equilibrium dialysis, free T4 calculated from thyronine-binding globulin (TBG) RIA, free T4 by three commercial kits (Gammacoat, Immophase, and Liquisol), T3, rT3, and TSH (by 3 different RIAs). Although all of the methods used for measuring free T4 (including free T4 index, free T4 by dialysis, free T4 assessed by TBG, and free T4 assessed by the 3 commercial kits) were excellent for the diagnosis of hypothyroidism, hyperthyroidism, and euthyroidism in the presence of high TBG, none of these methods showed that free T4 was consistently normal in patients with NTI; with each method, a number of NTI patients had subnormal values. In the NTI groups, free T4 measured by dialysis and the free T4 index generally correlated significantly with the commercial free T4 methods. Serum rT3 was elevated or normal in NTI patients and low in hypothyroid subjects. Serum TSH provided the most reliable differentiation between patients with primary hypothyroidism and those with NTI and low serum T4 levels.