Association of thyroid function with risk of bloodstream infections: results from the prospective population-based HUNT Study in Norway

Introduction: Studies of patients admitted to intensive care units (ICU) with sepsis have shown that thyroid function abnormalities are associated with adverse outcomes. It has also been suggested that patients’ thyroid levels prior to critical illness could affect sepsis mor- tality, even at subclinical levels(1). As thyroid dysfunction is common in the general population, this may represent a modifiable risk factor for sepsis(2,3). To our knowledge, the association between thyroid sta- tus and risk of bloodstream infections (BSI) has not previously been investigated in a large population-based study. Objectives: To determine if thyroid hormone levels in the general population is associated with future risk of BSI and mortality. Methods: Thyroid function was measured in 35,263 Norwegian men and women participating in the second HUNT Study(HUNT2, 1995- 97). Thyroid function was defined as Thyroid Stimulating Hormone (TSH) levels combined with free thyroxine (FT4) and total triiodthyro- nine (T3). HUNT2 data was linked to prospective information from the Mid-Norway Sepsis Register (MNSR), on clinically relevant BSI from local and regional hospitals until 2011. TSH levels where classified into six categories. Most of the BSI patients had sepsis. We evaluated the risk of first-time BSI, of all causative agents and by the most common infecting bacteria E. coli, S. aureus and S. pneumoniae. Risk of mortal- ity was assessed within 30 days of BSI and and all cause-mortality dur- ing the whole follow-up period. We used Cox regression to estimate hazard ratios with 95% confidence interval, using TSH 0,5-1,4 mU/L as reference. Results: After excluding participants with known thyroid disease, 31,707 HUNT2 participants where eligible for follow-up of BSI. During a median of 14,8 follow-up years, 1,138 experienced at least one epi- sode of BSI and 198 died within 30 days. Table 1 show the hazard ratios adjusted for age, sex, socioeconomical status, BMI and smoking. The higher TSH levels showed some association with development of BSI, but the results have a low precision and the confidence interval is also in agreement with no association. The results where robust for further adjustments of comorbidities (data not shown). The observed pattern displayed in Table 1 did not change across follow-up of sepsis due to E. coli, S. aureus and S. pneumoniae or for 30-day mortality after BSI. Conclusion: We found an association between higher TSH levels and a lower risk of hospitalizations and mortality due to BSI. However, the width of the confidence intervals also coincides with no risk differ- ence. All participants with abnormal levels of TSH where advised for further investigations, leaving a possibility that our associations could have been stronger in an untreated population.