Serum thyrotropin (TSH) concentrations were measured serially in 14 heart-transplant recipients (group 1) and 21 patients undergoing coronary artery bypass surgery (group 2), all without thyroid disease, and randomly in 158 patients hospitalized for various other nonthyroidal illnesses, including 144 judged euthyroid (group 3), six with increased FT4 and (or) T3 (group 4), and eight classified hypothyroid by conventional tests. The serial measurements indicated profound fluctuations. In group 1, TSH was subnormal in 21% of studies and increased in 10%. In group 2, corresponding abnormalities were found in 7% and 13%, respectively. Transiently low or high TSH tended to be associated with normal free thyroxin (FT4), prolonged subnormal TSH (greater than 1 week) with subnormal FT4. By contrast, subnormal TSH plus elevated FT4, or high TSH plus low FT4, were not encountered, making it unlikely that they occur by chance in severely ill patients who are not also hyper- or hypothyroid. In group 3, a suppressed TSH (plus borderline high FT4, T3/FT3) identified four cases of subclinical hyperthyroidism; however, another 11% of patients had subnormal and 10% had above-normal TSH, paired with normal FT4 and no evidence of thyroid disease. In group 4, suppressed TSH confirmed hyperthyroidism in five of six patients, and all in group 5 had increased TSH. We conclude that, in the hospital setting, sensitive TSH measurement can help to detect or confirm mild hyperthyroidism, but the positive predictive value of TSH alone may be as low as 35%.