1 INTRODUCTION

Common infections can develop into severe bloodstream infections (BSIs) and further into sepsis, a major cause of death worldwide. Approximately 50 million individuals develop severe sepsis and 11 million die every year from this.1 Sepsis is characterized by life-threatening organ failure due to a dysregulated host immune response to infection. Recent studies have shown that the risk of BSI is increased in relation to adverse levels of modifiable risk factors such as high body mass index (BMI),2, 3 smoking,2 severe depression,4 low physical activity, 2 and low iron levels.5

Previous studies on the association between thyroid function and infectious disease have primarily involved patients admitted to the intensive care unit (ICU),6, 7 and low levels of thyroid hormones have been associated with higher mortality in the ICU.6 Changes in thyroid levels during critical illness, known as nonthyroidal illness syndrome (NTIS), is thought to be a result of the body adapting to an increased use of energy and activation of the innate immune system.8 Studies on ICU patients therefore do not inform about possible association between thyroid function levels and the risk of developing BSI, and population-based studies are needed to examine this association. Accumulating evidence suggests that both hyperthyroidism and hypothyroidism may influence several immune functions like the antibody production, lymphocyte proliferation, phagocytosis, cytokine secretion, chemotaxis, and generating reactive oxygen species.9, 10 Hyperthyroidism and hypothyroidism may affect the immune system in opposite directions. Lymphocyte proliferation is one example, where the proliferation may increase during hyperthyroidism, but decrease during hypothyroidism. None of these studies have adjusted for smoking as an important confounder, known to influence the risk of both BSI and thyroid disease.11

Present evidence is also conflicting on whether hypo- or hyperthyroidism increases or reduces the risk of infection. Hyperthyroidism has been associated with both increased and reduced risk of infection. One small study of 254 elderly patients found that hyperthyroidism increased the risk of pulmonary and urinary tract infections after hip fracture surgery in comparison with euthyroid patients.12 While another small study of 93 healthy participants found that higher serum concentrations, within normal physiologic ranges of the thyroid hormones T3 and T4, were associated with enhanced innate and adaptive immunity and greater responsiveness to immune stimuli.13 In an animal study, hyperthyroid rats had a milder course of sepsis compared with hypothyroid or normothyroid rats.14

Also, for hypothyroidism there are conflicting results regarding infection risk. In a small cohort of 235 peritoneal dialysis patients, the authors found that lower fT4 level was a predictor of mortality attributable to infection.15 In 32,289 patients with periprostetic joint infections hypothyroidism was a risk factor for disease development,16 and in another study hypothyroidism was associated with increased risk of bacteriuria after urodynamic examination.17 On the other hand, an animal study showed that gestational hypothyroidism altered the offspring's immune response and lung-physiology, resulting in improved resistance to respiratory bacterial infections.18

Increased knowledge about modifiable causes of BSI is important for this potentially deadly disease, and where survivors often experience sequalae.19 The aim of this study was to prospectively examine the association between thyroid function and the risk of BSI and BSI-related mortality in a large population-based cohort. We emphasized associations related to differences within the physiological range of thyroid function, indicated by the reference range of thyroid-stimulating hormone (TSH), and additionally examined associations related to abnormal TSH levels or thyroid disease.

2 MATERIALS AND METHOD 2.1 Study population

This study is based on data from the second survey of the Norwegian the second trøndelag health study (HUNT) Study (HUNT2, 1995–1997). All inhabitants 20 years or older (n = 94,194) receding in the Nord-Trøndelag region were invited, and 65,237 (69.3%) participated. They answered comprehensive questionnaires on lifestyle and health-related factors and took part in a clinical examination that included blood sampling.20 For the purpose of this study, we included all participants who were selected for TSH measurements; all women aged ≥40 years, a 50% random sample of men aged ≥40 years, and a 5% random sample of women and men aged 20–40 years.21 In total, 34,169 participants had TSH and BSI information (see Figure 1).

Flowchart

2.2 Thyroid function measures

Blood samples were sent to the Hormone Laboratory, Aker University Hospital in Oslo for analysis of serum TSH. If TSH was >4.0 mU/L, thyroid peroxidase antibodies (TPOAb) were measured. Details of the laboratory methods and quality control are previously published.21 We categorized participants into six groups defined by their measured serum levels of TSH: four groups of TSH within the reference range (0.5–1.4, 1.5–2.4, 2.5–3.4, 3.5–4.5 mU/L), as well as TSH below (<0.5 mU/L) or above (>4.5 mU/L) the reference range.22, 23 The upper and lower TSH groups are defined by the limits commonly used to categorize (subclinical) thyroid dysfunction. As for differences within the reference range, we chose to classify into groups of equal width (~1 mU/L) as has been done in previous HUNT studies on thyroid function24, 25 and in the international Thyroid Studies Collaboration.26 Persons self-reported being diagnosed with hyperthyroidism, hypothyroidism, goitre, or use of thyroid-related medication (levothyroxine and thionamides), thyroid surgery or radioactive iodine11 at HUNT2 baseline. We categorized self-reported thyroid illness in two ways; (i) thyroid illness (yes/no) and (ii) thyroid illness-subcategories of hyperthyroidism, hypothyroidism, or other thyroid disease.

2.3 Outcome ascertainment

Participants were followed-up for future BSI defined by the growth of one or more microbes from blood culture combined with clinical evidence of systemic infection in need of hospitalization.27 BSI was identified at the two community hospitals in Nord-Trøndelag Hospital Trust (HNT HF); Levanger Hospital and Namsos Hospital or at the tertiary referral hospital St. Olavs hospital, Trondheim. All HUNT participants are referred to these hospitals in event of systemic infections like BSI, and all BSIs are prospectively included in HNT HF's-sepsis registry.28 In this registry only positive blood cultures from analysis using BACTEC 9240 Vacutainer Culture Bottles (Becton Dickinson Diagnostic Instrument Systems), which in 2010 was replaced by BACTEC FX 98% were included.29, 30 Furthermore, blood cultures solely with microorganisms commonly associated with skin contamination such as coagulase negative Staphylococcus species, Corynebacterium species, and Cutibacterium species were not considered as BSI27 unless that bacterium grew in two sets of blood culture or in one blood culture combined with additional growth from intravenous catheter or prosthesis. In addition, all patients underwent a clinical examination when hospitalized which confirms that 98% of the included BSI cases are validated to have sepsis according to the sepsis 2 criteria.27, 29, 30 Time at the risk of BSI started at the date of participation in HUNT2 (1995–1997) for participants referred to Levanger or St. Olavs hospital. As BSI follow-up in Namsos Hospital started in September 1999, HUNT2 participants belonging to Namsos Hospital started their time at risk at that time. A total of 1188 participants were not included in the study as follow-up time in the HNT registry ended (due to death or migration) before the start of follow-up time of BSI. We performed analysis in 31,413 participants without known thyroid disease (see Figure 1) and in a separate analysis participants we included the persons with known thyroid disease (n = 3476).

Our main outcome was a first-time event after the start follow-up of BSI. Secondary outcomes included BSI-related mortality and bacteria-specific BSI (i.e., Escherichia coli, Staphylococcus aureus, and Streptococcus pneumoniae). BSI-related mortality was defined as death within 30 days of a diagnosed BSI. The participants were followed until each outcome and censored at death, migration from the region or at end of follow-up (31 December 2011). Dates of migration and death were obtained from the Norwegian population registry.

2.4 Possible confounding factors

Information on possible confounders were retrieved from the HUNT2 data collection at baseline. Marital status was self-reported and categorized as (1) married/partner, (2) never married, and (3) separated/widower/divorced. Level of education was categorized as (1) <10 years, (2) 10–12 years, and (3) >12 years of schooling based on the HUNT question “What is your highest level of fulfilled education?”. Information on smoking status was obtained from several HUNT questions on past and current smoking, and participants were categorized as (1) current, (2) former, and (3) never smoking. The following chronic diseases were categorized as no/yes: (1) Cardiovascular disease (CVD) as a self-reported history of acute myocardial infarction, angina pectoris, stroke, and/or intermittent claudication surgery. (2) Lung disease as participants reporting having productive cough continuously for more than 3 months each year the last 2 years. (3) Chronic kidney disease (CKD) as an estimated glomerular filtration rate (eGFR) <60 ml/min per 1.73 m2. eGFR was estimated from recalibrated creatinine values using the Modification of Diet in Renal Disease (MDRD)-formula. (4) Cancer as saying yes to ever having been diagnosed with cancer. Weight and height were measured by trained nurses with the participants wearing light clothes and no shoes and rounded to the nearest half kilogram and centimetre, respectively. BMI was calculated as weight (kg) divided by the squared value of height (m2) 20 and categorized into six groups according to the World Health Organization (WHO) recommendations <18.5, 18.5–24.9, 25.0–29.9, 30.0–34.9, 35.0–39.9, and ≥40.0 kg/m2.

2.5 Statistical analysis

Descriptive statistics were calculated according to TSH categories. We imputed missing data using multiple imputation by fully conditional specification (FCS), generating a total of 10 complete datasets in the “mi chained”-routine. We then compared these results, with results from complete case analysis. We used Cox regression to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) for each BSI outcome (i.e., first BSI, bacteria-specific BSI, and BSI-related death) between categories of TSH using TSH of 0.5–1.4 mU/L as the reference category.25 Similarly, we estimated the risk of BSI associated with TSH as a continuous variable within the reference range (i.e., 0.5–4.5 mU/L). All analyses were adjusted for potential confounding by age (as the timescale), sex, marital status, education, smoking, and BMI. The proportional hazards assumption was evaluated by testing Schoenfeld residuals and by graphical inspection of log–log plots. There was no clear violation of the assumptions other than for sex, which was adjusted for using a stratified procedure and for education where we used the time varying option in all adjustments. We also performed an analysis investigating whether self-report of thyroid illness (yes) or subdivided into hyperthyroid, hypothyroid, or other thyroid disease was associated with increased risk of BSI compared with participants with no known thyroid disease. We performed several sensitivity analyses to assess the robustness of the results: (1) All analysis were repeated to assess possible competing risk by death from all causes using the Fine and Gray method.31 (2) Since comorbidity could mediate, rather than confound the association between TSH and BSI risk, this was adjusted for in separate analyses. (3) Log–log plots indicated a somewhat stronger association in younger than in older age groups, therefore we conducted analyses stratified by baseline age < versus ≥ 65 years and examined possible statistical interaction in a likelihood ratio test of a product term of the TSH and age categories. (4) We repeated all analysis in a sample restricted to never-smokers to reduce possible residual confounding by smoking. (5) The main analysis was repeated after excluding the first 3 and 5 years of follow-up to avoid possible confounding due to pre-existing diseases and conditions. (6) We tested linear as well as quadratic trends for the associations of TSH levels and the risk of BSI outcomes. (7) Finally, evolving hypothyroidism is more likely if TPOAb are present, and levels of TPOAb were available for those with TSH > 4.0 mU/L. Thus, we estimated the risk of each BSI outcome in people with TSH > 4.0 mU/L with and without high TPOAb levels (> 200 U/ml). All analyses were performed using STATA version 16 (StataCorp).

2.6 Ethics statement

This study has been approved by the ethical committee for medical research (REK 2018/1819/REKmidt and REK 2012/153) and by the HUNT data access committee. From the data access committee in HNT HF we have authorization to use and merge the HUNT data file with the HNT BSI registry. Participation in HUNT2 was voluntary. Participants signed a written consent to data collection and to linking their data to other registers. HUNT-participants were recommended to consult their general practitioner within 2–3 weeks if their TSH levels indicated probable hypothyroidism or hyperthyroidism and within 2 years if their TSH levels indicated the risk of hypothyroidism or hyperthyroidism.32

3 RESULTS

Baseline characteristics of the study sample stratified by TSH levels are shown in Table 1. Of the 34,619 participants, a total of 1,179 persons had a first-time BSI during a median of 14.5 years of follow-up.

Table 1. Baseline characteristics (in %) of the 34,619 participants with information on thyroid levels at inclusion to HUNT2 Variable TSH levels <0.5 mU/L 0.5–1.4 mU/L 1.5–2.4 mU/L 2.5–3.4 mU/L 3.5–4.5 mU/L >4.5 mU/L n=1463 n=13,947 n=12,136 n=4001 n=1458 n=1612 Total 4.5 40.3 35.1 11.6 4.2 4.7 Sex Female 80.6 65.8 66.0 69.4 73.7 76.3 Male 19.4 34.3 34.1 30.6 26.3 23.7 Smoking Never 40.9 36.5 45.8 50.5 53.4 49.3 Former 27.3 28.0 29.2 30.9 31.2 32.9 Current 31.9 35.5 25.0 18.6 15.4 17.8 Marital status Never married 8.9 9.4 9.7 9.8 9.5 9.4 Married/partner 64.6 69.6 69.2 67.5 67.2 66.8 Separated/divorced/widower 26.5 21.1 21.2 22.7 23.3 23.8 Education <10 years 55.0 47.7 48.8 53.7 55.9 59.0 10–12 years 30.6 35.8 34.1 31.0 28.8 28.3 >12 years 14.4 16.6 17.1 15.3 15.3 12.7 Comorbidityb CVD 10.6 9.2 10.4 12.1 12.2 12.7 CKD 7.2 4.7 6.3 8.2 9.6 11.4 Lung disease 9.0 9.3 9.4 9.0 7.9 8.7 Cancer 9.9 5.0 5.1 6.2 7.1 5.1 Diabetes 5.9 36.8 35.0 13.3 4.5 4.4 Thyroid illness 50.4 7.2 6.0 9.9 14.8 24.6 Mean age 59.6 56.4 58.3 60.7 62.1 62.0 Mean BMI kg/m2 26.8 26.4 27.1 27.4 27.6 27.6 Abbreviations: BMI, body mass index; CI, confidence interval; CKD, chronic kidney disease; CVD, cardiovascular disease; HR, hazard ratio; TSH, thyroid-stimulating hormone. 3.1 TSH levels within the reference range and risk of BSI

Overall, there was little evidence that TSH levels were associated with the risk of first-time BSI in the population with no known thyroid disease (Table 2). The HR associated with one unit (mU/L) increase in TSH within the reference range was 0.97 (95% CI: 0.90–1.04), p from test of linear trend .93. The test for cubic trend did not show any signs of U-formed relationship between TSH levels and BSI. Compared with the reference category of TSH 0.5–1.4 mU/L, persons with a TSH levels 3.4–4.5 mU/L had a slightly lower risk of BSI (HR: 0.90, 95% CI: 0.67–1.21) but the CIs were broad and compatible with no difference in risk.

Table 2. Hazard ratios (HR) with 95% confidence intervals (CIs) for first time BSI according to TSH level at baseline TSH levels (mU/L) Age and sex adjusted Multivariably adjustedb Person-years at risk No. of events HR HR 95% CI <0.5 8497 29 1.09 1.07 0.74–1.56 0.5–1.4 164,165 458 1.00 1.00 Reference 1.5–2.4 141,904 440 1.00 1.01 0.89–1.15 2.5–3.4 43,870 146 0.92 0.94 0.78–1.14 3.5–4.5 14,640 50 0.89 0.90 0.67–1.21 >4.5 14,278 56 1.00 1.02 0.77–1.35 TSH per mU/L within reference range 387,353 1.179 0.97 0.97 0.90–1.04 p for linear trend .921 .929 p for quadratic trend .502 .516 Abbreviations: BMI, body mass index (kg/m2); BSI, blood stream infection; CI, confidence interval; HR, hazard ratio; TSH, thyroid-stimulating hormone. 3.2 Thyroid disease and risk of BSI

Table 3 shows an 30% (95% CI: 1.10–1.53) increased risk of first-time BSI in persons having a diagnosis of thyroid illness compared with not having such an illness. For hyperthyroidism, the risk was increased with an 57% (95% CI: 1.20–2.06) compared with not having any thyroid disease. However, also for hypothyroidism (HR: 1.20) and other thyroid diseases (HR: 1.19), the point estimates pointed towards the increased risk compared with not having thyroid disease, however, here the CIs were wide and inconclusive.

Table 3. Hazard ratios (HR) with 95% confidence intervals (CIs) for first time BSI in association with thyroid-related illness at baseline TSH levels (mU/L) Age and sex adjusted Multivariably adjustedb Person-Years at risk No. of events HR HR 95% CI No thyroidillness 387,353 1179 1.00 1.00 Reference Any thyroid illness 42,331 167 1.37 1.30 1.10–1.53 Hypothyroidism 19,726 73 1.25 1.20 0.95–1.53 Hyperthyroidism 11,541 55 1.64 1.57 1.20–2.06 Other thyroid conditions/treatment 9,896 36 1.19 1.19 0.85–1.66 Person-Years at risk No. of events Age and sex adjusted Multivariably adjustedb HR 95% CI HR 95% CI With no known thyroid disease 387,353 1179 1.00 Reference 1.00 Reference Diagnosed with thyroid disease 42,331 167 1.37 1.00–1.87 1.30 1.10–1.53 Abbreviations: BMI, body mass index (kg/m2); BSI, blood stream infection; CI, confidence interval; HR, hazard ratio. 3.3 TSH levels and risk of BSI in age-stratified analysis

There was some evidence that age could modify the association between TSH and the risk of BSI (pinteraction, .002). In stratified analyses, TSH within the reference range was associated with a decreased HR of 0.88 (95% CI: 0.78–1.00) per mU/L in people <65 years, whereas no association was observed in the oldest age group (HR: 1.02; 95% CI: 0.94–1.11) (Table 4). Participants <65 years with TSH 3.5–4.5 mU/L had a lower risk of BSI with HR of 0.58 (95% CI: 0.23–0.93), compared with those with TSH 0.5–1.4  mU/L, and participants <65 years with TSH > 4.5 mU/L had a lower risk of BSI compared with those with TSH 0.5–1.4 mU/L (HR: 0.49, 95% CI: 0.24–0.99). All cases of BSI aged <65 years and with TSH > 4.5 mU/L had evidence of autoimmune hypothyroidism with a TPOAb well over 200 IU/ml and the majority were smokers (data not shown).

Table 4. Hazard ratios (HR) with 95% confidence intervals (CIs) for first BSI associated with TSH (mU/L) levels stratified by age TSH levels (mU/L) Person-Years at risk No. of events Multivariably adjustedb HR 95% CI Participants < 65 years of age <0.5 5903 10 0.95 0.50–1.78 0.5–1.4 126,864 223 1.00 Reference 1.5–2.4 103,500 157 0.84 0.68–1.03 2.5–3.4 29,006 52 0.95 0.70–1.29 3.5–4.5 9226 8 0.58 0.23–0.93 >4.5 8680 8 0.49 0.24–0.99 TSH per mU/L within reference range 283,218 458 0.88 0.78–1.00 Participants≥65 years of age <0.5 2594 19 1.18 0.74–1.89 0.5–1.4 37,301 235 1.00 Reference 1.5–2.4 38,405 283 1.16 0.98–1.38 2.5–3.4