Low T3 syndrome is a strong predictor of poor outcomes in patients with community-acquired pneumonia

Low T3 syndrome was previously reported to be linked to poor clinical outcomes in critically ill patients. The aim of this study was to evaluate the predictive power of low T3 syndrome for clinical outcomes in patients with community-acquired pneumonia (CAP). Data for 503 patients were analyzed retrospectively, and the primary end point was 30-day mortality. The intensive care unit (ICU) admission rate and 30-day mortality were 8.3% and 6.4% respectively. The prevalence of low T3 syndrome differed significantly between survivors and nonsurvivors (29.1% vs 71.9%, P < 0.001), and low T3 syndrome was associated with a remarkable increased risk of 30-day mortality and ICU admission in patients with severe CAP. Multivariate logistic regression analysis produced an odds ratio of 2.96 (95% CI 1.14–7.76, P = 0.025) for 30-day mortality in CAP patients with low T3 syndrome. Survival analysis revealed that the survival rate among CAP patients with low T3 syndrome was lower than that in the control group (P < 0.01). Adding low T3 syndrome to the PSI and CURB-65 significantly increased the areas under the ROC curves for predicting ICU admission and 30-day mortality. In conclusion, low T3 syndrome is an independent risk factor for 30-day mortality in CAP patients.


Results
Patient characteristics. A total of 1640 patients hospitalized with pneumonia derived from the community were evaluated in this retrospective study. Eighty-two patients who had a history of thyroid diseases and 114 patients receiving immunosuppressant therapy or presenting with human immunodeficiency virus (HIV) infection were excluded. Another 941 patients who did not complete a serum thyroid hormone test also were excluded. Finally, a total of 503 hospitalized CAP patients were eligible for analysis in this retrospective observational study. The demographic characteristics of these patients are presented in Table 1. The mean age was 63 ± 18 years, and 275 patients (54.7%) were male. Overall, 193 (38.4%) of 503 cases were accompanied by one or more coexisting illnesses, including congestive heart failure (CHF) (12.9%), chronic obstructive pulmonary disease (COPD) (8.5%), chronic renal disease (5.6%), chronic liver disease (2.4%), cerebrovascular disease (7.8%), malignant tumor (4.2%), diabetes mellitus (9.1%), and septic shock (6.4%). In addition, 160 (31.8%) patients had low T3 syndrome. The 30-day mortality and ICU admission rate in the included population were 6.4% and 8.3%, respectively. The median length of stay (LOS) was 11 days (interquartile range [IQR], 7-16 days; Table 1). To exclude potential selection bias, we analyzed the baseline characteristics of patients who did or did not participate in a thyroid hormone test. We found that neither the 30-day mortality nor ICU admission rate showed any significant difference between two groups (Supplemental Table S1).

Univariate and multivariate analysis of risk factors for ICU admission and 30-day mortality.
Compared to those who survived beyond 30 days, non-survivors were older; more likely to also have CHF, COPD, hypoalbuminemia, thrombocytopenia, hyponatremia, low T3 syndrome, and septic shock; and showed a higher C reactive protein (CRP) level, a higher blood urea nitrogen (BUN) level, lower arterial pH, and higher  Table 2). The results showed that age, CHF, low T3 syndrome, platelet count < 10 6 /μl and serum sodium < 130 mmol/l are independent risk factors for 30-day mortality. Also, CHF, COPD, pulse >125/min, albumin (ALB) < 30 g/dl, and platelet count < 10 6 /μl are independent risk factors for ICU admission.
Low T3 syndrome is associated with increased 30-day mortality. The characteristics of CAP patients with or without low T3 syndrome are listed in Table 3

Correlation between FT3, FT4, and TSH levels and outcomes of CAP patients. Because low T3
syndrome was found to be an independent risk factor for 30-day mortality in the present study, we performed a receiver operating characteristic (ROC) curve analysis of the ability of FT3, FT4, and TSH levels to predict the 30-day mortality and ICU admission among CAP patients (Fig. 2). We found that the areas under the ROC curves (AUCs) for FT3 were 0.782 (95% CI 0.743-0.817) for predicting 30-day mortality and 0.778 (95% CI 0.739-0.814) for predicting ICU admission, and these values were superior to those for serum FT4 and TSH. The best FT3 cut-off points for predicting 30-day mortality and ICU admission were FT3 < 3.06 pmol/l and FT3 < 3.21 pmol/l, respectively. Next, we made low T3 syndrome as an additional 20 points to the PSI and 1 point to CURB-65 to create two new scores and compared their predictive abilities with those of the PSI and CURB-65 alone for 30-day mortality and ICU admission (Figs 3 and 4). Interestingly, both low FT3 + PSI and low FT3 + CURB-65 had significantly increased AUCs for predicting 30-day mortality, compared to the individual indices (0.780 vs 0.753, P = 0.035 and 0.758 vs 0.698, P = 0.003, respectively). In addition, low FT3 + CURB-65 had a significantly greater AUC for predicting ICU admission than CURB-65 alone (0.745 vs 0.703, P = 0.021). However, low FT3 + PSI did not show great accuracy than PSI alone for predicting ICU admission (0.757 vs 0.736, P = 0.084).

Discussion
CAP is one of the most common diseases requiring hospital admission and a major cause of death worldwide. It is important to identify patients with severe pneumonia and those who have the worst prognosis in order to provide the most appropriate treatments. Low T3 syndrome was reported to be highly prevalent in critically ill patients and has been proven to be a predictor of poor outcome in these patients [21][22][23] . Our study showed that low T3 syndrome is an independent risk factor for 30-day mortality among CAP patients, and the survival curves confirmed that CAP patients with low T3 syndrome had a lower survival rate than those without low T3 syndrome. We also found that the serum FT3 level offers better diagnostic accuracy than FT4 and TSH levels, and adding low FT3 criteria to the PSI or CURB-65 scores could significantly increased the AUCs for predicting 30-day mortality and ICU admission, compared to those for the individual scores. Low T3 syndrome has been identified as a useful adaptation in the negative feedback regulation to preserve energy when patients suffer from an acute critical illness 24 . However, it remains controversial whether low T3 syndrome represents a protective or a maladaptive response to prolonged critical illness 25 . During acute illness, the occurrence of low T3 syndrome can be explained by changes in thyroid hormone binding, peripheral thyroid hormone uptake and the expression and activity of the type-1 deiodinases (D1) and type-3 deiodinases (D3) 26 . As patients with acute critical illnesses usually experience concomitant fasting, the decreased thyroid hormone availability may reflect an adaptive attempt to reduce energy expenditure and, thus, appears to be beneficial 27 . In addition, the increased D3 activity could optimize the bacterial killing capacity of neutrophilic granulocytes 28 .
Unlike in the acute phase, the thyroid axis appears to be regulated by different mechanisms during prolonged critical illness. Low T3 syndrome in these patients is found to result from the low expression of the thyrotropin-releasing hormone (TRH) gene in the paraventricular nucleus (PVN), which impairs the hypothalamic stimulation of the thyrotropes 29 . Peripheral tissues adapt to the decreased thyroid hormones status by increasing levels of thyroid hormone transporters, local activation of thyroid hormone and gene expression of   the active receptor isoforms [30][31][32][33] . Different from acute critical illness, the low T3 level correlates with markers of muscle breakdown and of bone loss, which indicate a hypercatabolic maladaptive response in prolonged critical   illness 34 . In summary, low T3 syndrome during prolonged critical illness appears to differ from changes observed during acute critical illnesses, both in its mechanisms and in its influence on prognosis.
In the present study, low T3 syndrome was identified as an independent risk factor associated with poor 30-day mortality among CAP patients. We found that the magnitude of the decrease in the serum FT3 levels was correlated with the severity of CAP, and serum FT3 shows accurate prognostic value for predicting 30-day mortality and ICU admission among CAP patients. This may be attributed to large amounts of inflammatory cytokines, including tumor necrosis factor (TNF)-α , interleukin (IL)-1 and IL-6, which are present during critical pneumonia. These cytokines are considered to be putative mediators of low T3 syndrome 35,36 . Additionally, D3, which is viewed as the major thyroid hormone inactivating enzyme, was shown to be highly expressed in infiltrating neutrophilic granulocytes during bacterial infections 28 , which may lead to low T3 levels during an acute infection.
The PSI and CURB-65 scores have shown good diagnostic accuracy for predicting mortality among pneumonia patients in previous studies 8,9 , but limited accuracy for predicting ICU admission. Although the low serum FT3 was not included as a variable in any previous severity scores for CAP, low FT3 criteria indeed significantly improved the performance of PSI and CURB-65 for predicting 30-day mortality and ICU admission in CAP patients.
There are a few limitations in this study. First, this was a retrospective single-center study. Second, the result of endocrine-based testing may be related to the timing of sampling. It has been reported that serum thyroid hormone level may require 4 days to reach a nadir after the onset of critical illness 37 . In addition, dopamine, which can induce iatrogenic hypothyroidism in patients with critical illness 38,39 , was not excluded as a confounder in our study.
In conclusion, we found that low T3 syndrome is a reliable predictor of worse outcomes in hospitalized CAP patients. The serum FT3 level showed great performance for prediction of 30-day mortality and ICU admission in CAP patients, and was found to be even more diagnostically accurate than the PSI and CURB-65. The finding of the present study should be conformed in multicenter large prospective studies, and future studies are needed to elucidate the underlying mechanisms by which low T3 syndrome affects CAP patients.

Materials and Methods
Study population. We retrospectively evaluated the records of all hospitalized patients who were diagnosed with CAP between January 2010 and December 2013 at the Second Affiliated Hospital, Zhejiang University School of Medicine. According to the American Thoracic Society (ATS)/Infectious Disease of Society of America (IDSA) guidelines, the diagnosis of CAP is based on the presence of select clinical features (e.g., cough, fever, sputum production, and pleuritic chest pain) and is supported by imaging of the lung, usually by chest radiography, with or without supporting microbiological data 11,40 . Eligible patients were older than 14 years, and had available serum thyroid hormone levels. Patients were excluded if they had an overt history of hypothyroidism or hyperthyroidism diagnosed during admission, or also presented with HIV infection, had undergone organ transplantation, or were receiving immunosuppressant therapy (> 20 mg/day of prednisone or equivalent). The Ethics Committee at the Second Affiliated Hospital, Zhejiang University School of Medicine approved this study. All aspects of the study were performed in accordance with relevant guidelines and regulations 41 . Data collection and definition of severe CAP. Baseline clinical information was obtained for each patient, including age, gender, comorbid diseases such as CHF, COPD, chronic renal diseases, chronic liver diseases, cerebrovascular disease, malignancy, diabetes mellitus, antimicrobial pre-treatment, physical examination results, septic shock (severe sepsis plus hypotension that was not reversed with fluid resuscitation), laboratory data such as white blood cell (WBC) count and platelet count, and levels of CRP, ALB, BUN, glucose, serum sodium, FT3, FT4, and TSH, and radiologic findings (new or progressive pulmonary infiltrate) within 24 h after admission. To evaluate the severity of CAP, the PSI, CURB-65 and SCAP scores were independently calculated 8,9,42,43 . The thyroid hormone levels were measured on the next morning after admission using commercial radioimmunoassay kits (Abbott Architect i2000 analyzer). According to the manufacturer's instructions, the normal reference intervals used were: 3.1-7.5 pmol/l for FT3, 8.9-20.6 pmol/l for FT4, and 0.35-4.6 mIU/l for TSH. In accordance with the manufacturer's standard value, low T3 syndrome was defined by a low serum FT3 level (< 3.1 pmol/l) with low or normal serum FT4 and TSH levels (≤ 20.6 pmol/l and ≤ 4.6 mIU/l, respectively), and patients who had been diagnosed with hypothyroidism, hyperthyroidism, or thyroiditis were excluded 13,44 . The primary end point was 30-day mortality, which was defined as death due to any cause within 30 days from hospital admission. Secondary end points were LOS in the hospital and ICU admission. Statistical analysis. All continuous variables are presented as mean ± standard deviation or as median (IQR). Categorical variables are presented as a number (%). Comparisons between continuous variables were performed using the unpaired t-test. Qualitative/categorical variables were compared between groups using the chi-square test or Fisher's exact test. Univariate analysis and multiple logistic regression analysis were performed to determine independent risk factors for 30-day mortality and ICU admission. The ORs and corresponding 95% CI were calculated for all variables. The ROC curves for the ability of FT3, FT4 and TSH to predict the 30-day mortality and ICU admission were analyzed. The AUCs were compared using Delong-Delong nonparametric method. Kaplan-Meier analysis was used to estimate the probability of survival and comparison was made by Log-rank test. The best FT3 cut-off points for predicting 30-day mortality and ICU admission were determined by the Youden index. A P value of < 0.05 was considered statistically significant. The results were analyzed using SPSS version 18.0 for Windows (SPSS Inc., Chicago, IL, USA).