Pre-admission TSH levels predict long-term mortality in adults treated for hypothyroidism.

CONTEXT: Limited data is available regarding the association between pre-admission thyroid function and prognosis of hospitalized patients treated for hypothyroidism. OBJECTIVE: Evaluate an association between thyroid stimulating hormone (TSH) levels and mortality in hospitalized levothyroxine-treated patients. DESIGN AND SETTING: Observational data of patients admitted to medical wards between 2011 and 2013. TSH levels obtained up to 180 days prior to admission were stratified as follows: low (≤0.5 mIU/L), normal (0.5-5 mIU/L), high (>5 mIU/L). PATIENTS: Patients aged 60-80 years with available thyroid function tests were matched with controls without hypothyroidism. MAIN OUTCOME: All-cause mortality up to 66-months following discharge. RESULTS: One thousand and fifty seven patients (73% females, mean (SD) age 71 ± 6 years) were matched with controls without hypothyroidism. Mean hospital stay and in-hospital mortality were not different between groups. Mortality risk at the end-of-follow-up was 41% (438/1057) and 37% (392/1057) for patients with and without hypothyroidism (p < 0.05). TSH levels were classified as follows: low, 84 patients (8%); normal, 667 patients (63%); high, 306 patients (29%). Length of hospitalization and in-hospital mortality were not different between TSH categories. Mortality risk at the end-of-follow-up was 30, 39, and 50% with low, normal and elevated TSH, respectively. Adjusted hazard ratio (95% CI) of mortality at the end-of-follow-up was of 2.2 (1.2-3.8) for high vs. low TSH levels, and 1.4 (1.1-1.9) for high vs. normal TSH levels. CONCLUSION: In treated hypothyroid adult patients, increased TSH up to 6 months prior to admission is associated with increased mortality. Treatment should aim at achieving euthyroidism to improve survival.

PHI-DAC: protein homology database through dihedral angle conservation.

UNLABELLED: Finding related conformations in the Protein Data Bank is essential in many areas of bioscience. To assist this task, we designed a dihedral angle database for searching protein segment homologs. The search engine relies on encoding of the protein coordinates into text characters representing amino acid sequence, φ and ψ dihedral angles. The search engine is advantageous owing to its high speed and interactive nature and is expected to assist scientists in discovering conformation homologs and evolutionary kinship. The search engine is fast, with query times lasting a few seconds, and freely available at http://tarshish.md.biu.ac.il/∼samsona. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.