Background: In severe illness, plasma selenium levels are decreased; a decreased activity of the selenoenzyme 5 0 -deiodinase has been hypothesized to contribute to low tri-iodothyronine (T3) levels in non-thyroidal illness (NTI) syndrome in these patients. Objective: To analyse the influence of selenium substitution on thyroid hormone metabolism in patients with severe sepsis. Design: A prospective, randomized, controlled study at the medical internal intensive care unit of the University of Munich. Results are for 41 consecutive patients with severe sepsis with an APACHE II score . 15. Patients received either sodium selenite (500 mg/day for the first 3 days, reducing to 250 and then 125 mg/day every 3 days) or a placebo. Results: At study entry, APACHE II score and demographics were identical in both groups. The mean levels of TSH, free tri-iodithyronine and total T3, as well as plasma selenium and selenium-dependent peroxidase (GSH-Px) activity, were decreased. Plasma selenium and GSH-Px activity were normalized on days 3, 7 and 14 in patients receiving selenium (n ¼ 21), but remained below normal in the control patients. Patients receiving selenium had a better clinical outcome and thyroid hormone levels normalized earlier. Thyroid hormone levels increased in patients who showed clinical improvement, independent of selenium levels or selenium substitution. Conclusions: Selenium substitution in patients with NTI improves morbidity, but has no direct effect on the free and total thyroid hormones. In severely ill patients, decreased deiodinase activity due to low plasma selenium levels seems unlikely. After clinical revival, TSH and then the thyroidal hormones normalize independently of selenium substitution.