Risk analysis of 2-year mortality in elderly male military veterans with non-thyroidal illness syndrome

Background: Elderly patients with non-thyroidal illness syndrome (NTIS) have relatively severe symptoms and a poor prognosis. However, there are few studies on the correlation between NTIS and mortality among hospitalised elderly patients. This study aimed to investigate the characterization of NTIS inpatients with biochemical indicators and mortality prediction. Methods: In the present study, 931 male veteran inpatients ≥ 60 years of age who visited our hospital from January 2012 to December 2013 were selected and divided into the NTIS group (n = 193) and normal thyroid function (non-NTIS) group (cid:0) n = 738). Following propensity score matching to match the two groups according to age and body mass index, the NTIS group and non-NTIS group consisted of 192 and 660 patients, respectively. Data on biochemical indicators and mortality were collected. Results: Patients had more primary care and more respiratory disease and chronic kidney disease in the NTIS than in the non-NTIS group. Serum total protein (TP), albumin (Alb), prealbumin (PA), haemoglobin (Hb), uric acid (UA), triglyceride (TG), and high-density lipoprotein cholesterol (HDLC) levels were signicantly lower, and fasting blood glucose (FBG) and urea nitrogen (UN) levels were higher, in the NTIS than in the non-NTIS group. Triiodothyronine, TP, Alb, and PA levels correlated positively with the Hb level and negatively with FBG, UN, and creatinine (Cr) levels. The free T 3 level correlated positively with TP, Alb, PA, Hb, and UA levels and negatively with FBG, UN, and Cr levels. A lower free T 3 level was associated with increased all-cause mortality after adjusting for covariates. Patients in the NTIS group had a lower survival rate at 6-month (78.65% vs. 97.73%), 1-year (68.23% vs. 96.97%) and 2-year (64.58% vs. 86.52%). Receiver operating characteristic curve (ROC) analysis showed that a cut-off


Background
Many changes in thyroid hormones occur as a result of illness or nutritional deprivation. These changes consist of decreased levels of serum triiodothyronine (T 3 ) and/or thyroxine (T 4 ), without an increase in the thyroid-stimulating hormone (TSH) level, and usually accompanied by elevated reverse T 3 (rT 3 ) [1,2]. The combination of these ndings is termed non-thyroidal illness syndrome (NTIS), euthyroid sick syndrome, or low T 3 syndrome, indicating a systemic disease outside the thyroid that causes abnormal levels of thyroid hormones and is often considered a compensatory mechanism for the body. NTIS has been reported in patients with acute and chronic illnesses, including infectious diseases, cardiovascular and gastrointestinal diseases, cancer, and trauma [3], which are quite common in patients in intensive care units (ICUs) [4,5]. In previous studies, NTIS showed a high sensitivity and speci city for predicting patient mortality in ICU patients [6]. Serum T 3 levels further decrease as the severity of disease progresses. However, among hospitalised elderly patients in the general ward, data on the correlation between NTIS and patient mortality are lacking.
Elderly patients often have multiple chronic diseases and a poor nutritional status; thus, NTIS is quite common in these patients. Tognini [7] reported that among elderly patients (≥ 65 years of age) hospitalised for acute illness, the prevalence of NTIS was 31.9% and signi cantly associated with acute renal failure, New York Heart Association classi cation IV heart failure, and metastatic cancer. We identi ed elderly patients who were hospitalised in our hospital. Triiodothyronine, total T 4 , free T 3 , free T 4 , thyrotropin and serum levels of total protein (TP), albumin (Alb), prealbumin (PA), urea nitrogen (UN), creatinine (Cr), uric acid (UA), fasting blood glucose (FBG), blood lipids, alanine transaminase (ALT), aspartate transaminase (AST), and haemoglobin (Hb) were analysed, and 6-month, 1-year and 2-year all-cause mortality rates based on free T 3 levels was assessed. This study investigated the characterization of NTIS patients with biochemical indicators and mortality prediction that may provide a basis for predicting disease outcome and mortality in hospitalised elderly male patients with NTIS.

Patients
The inclusion criteria for the study were as follows: 1) ≥ 60 years of age, 2) male, 3) hospitalised between January 2012 and December 2013, and 4) complete data on thyroid function obtained within 1 day of admission. The exclusion criteria were as follows: 1) diagnosed with thyroid disease such as hyperthyroidism, hypothyroidism, subclinical hyperthyroidism, subclinical hypothyroidism, or Hashimoto's thyroiditis, 2) taking amiodarone, euthyroid, glucocorticoid, dopamine, or interferon, and 3) history of thyroid surgery. Because the patients in our hospital are Chinese military veterans, there were few female patients. As such, only male patients were selected as research subjects to reduce the in uence of sex bias on the results. Within 2 years, a total of 9,832 patients were admitted to the hospital, 931 of whom were enrolled in the present study. All patients were followed up for 2 years, until December 2015. The subjects were divided based on the presence of NTIS into the NTIS group and normal thyroid function (non-NTIS) group. Because the baseline age and body mass index (BMI) in the two groups differed signi cantly, the groups were matched by age and weight using the propensity score matching (PSM). After matching, 852 patients were included in the study: 192 in the NTIS group and 660 in the non-NTIS (normal thyroid function) group. At the end of the 2-year follow-up, 157 patients had died and 45 were lost to follow-up (5.28%). The time and cause of death were recorded for all deceased patients.

Diagnostic criteria for NTIS
Decreased serum triiodothyronine and/or free T 3 , normal or mildly reduced total T 4 or free T 4 , and normal TSH levels were the diagnostic criteria for NTIS [8].
Statistical analysis PSM (1:4) was used to account for demographic differences in the selected participants. SPSS 22.0 was used for the statistical analysis. Normally distributed data are expressed as means ± standard deviation (`x ± SD) and non-normally distributed data as quartiles. Normally distributed data were compared between the two groups using t-test or t'-test. The rank sum test was used to compare the data distribution between the two groups. All data were evaluated using two-sided tests. Correlations were determined using Pearson correlation analysis for normally distributed data and Spearman's test for non-normally distributed data. Overall survival at 6-month, 1-year and 2-years was estimated using the Kaplan-Meier method. Bivariate and multivariate Cox proportional hazards models were used to evaluate the risk factors associated with patient mortality.
Statistical signi cance was set at P < 0.05.

Comparison of baseline data between the NTIS and non-NTIS groups
To minimize research bias, PSM was used to match the two groups according to two major factors that in uence metabolic indicators and mortality: age and BMI. After matching, statistically signi cant differences in age and BMI were not observed (see Additional le 1). And after matching by age and BMI, the triiodothyronine, free T 3 , total T 4 (all P < 0.001), and TSH levels (P = 0.001) were signi cantly lower in the NTIS group than in the non-NTIS group, moreover, the percentage of patients with primary care was much more in the NTIS group than in the non-NTIS group (P < 0.001), and that for patients admitted with respiratory disease and accompanied by respiratory disease and chronic kidney disease were more in the NTIS group than in the non-NTIS group (P < 0.001, Table 1). After matching by age and BMI, TP, Alb, PA, Hb, UA, TG, and HDLC levels were signi cantly lower, while UN and FBG levels were higher, in the NTIS group than in the non-NTIS group ( Table 2).   Table 3). Association of the free T 3 level with mortality according to Cox proportional hazards models Univariate Cox proportional hazards models showed that the free T 3 level was associated with reduced patient mortality. Among the other biochemical indicators, Alb and Hb levels were negatively, whereas FBG, total cholesterol (TC), and ALT levels were positively, correlated with patient mortality. Multivariate models adjusted for other confounding factors showed that a lower free T 3 level in elderly male patients is associated with all-cause mortality. Other indicators including Alb and Hb levels were negatively correlated and the ALT level positively correlated with patient mortality. Patients with respiratory diseases (RD), chronic kidney disease (CKD), and tumours had a higher mortality rate ( Table 4). The free T 3 level and 2-year mortality rate of all the patients were also determined. According to a receiver operating characteristic (ROC) analysis of the free T 3 level and 2-year mortality rate, a cut-off free T 3 level of 3.45 pmol/L yielded the highest Youden index (0.327), with a sensitivity of 0.675 and speci city of 0.642 (Figure 1).

Discussion
In the elderly population, thyroid hormone levels can help monitor health status, predict short-term and longterm clinical prognoses, predict disease severity, and assess quality of life and survival status. In previous studies, the frequency of thyroid dysfunction increased with advancing age in the hospitalised elderly patients. The prevalence of NTIS in hospitalised severely or debilitated elderly patients can be as high as 32%-62% [7,9,10]. In the present study, among the 931 elderly male patients hospitalised for various reasons, there were 193 NTIS patients (20.73% prevalence rate). After matching by age and BMI, the Alb levels were signi cantly lower, and the renal function indices and FBG levels higher, in the NTIS group than in the non-NTIS group. In addition, the 2-year survival rate was signi cantly lower in the NTIS group than in the non-NTIS group. A reduced free T 3 level was strongly associated with all-cause mortality in NTIS patients, a similar nding of previous studies [11,12].
NTIS is often associated with nutritional de ciencies or acute and chronic diseases. Protein and UA levels are indicators of nutritional status. Proteins also play an important role in the synthesis and transport of thyroid hormones. In several studies, the serum Alb level was reduced, and the free T 3 level was positively correlated with the Alb level, in patients with NITS [13,14]. In the present study, the Alb levels were also signi cantly reduced in the NTIS group compared with the non-NTIS group. Correlation analysis showed that TP, Alb, and PA levels decreased as the triiodothyronine and free T 3 levels decreased. Hypothetically, decreased Alb levels leads to a decrease in the conversion of T 4 to T 3 , resulting in a decrease in T 3 levels or a decrease in T 4 binding to the protein, which accelerates the removal of thyroid hormones [15,16]. In the present study, the free T 3 level was also positively correlated with the UA and Hb levels, further con rming that fasting and hunger can cause NTIS [5].
The T 4 level is strongly associated with CKD. NTIS is a common thyroid dysfunction in CKD patients [13,17], and its mechanism is associated with the kidney's involvement in the synthesis, secretion, and metabolism of thyroid hormones. In kidney disease, chronic metabolic acidosis and in ammatory factors lead to the inhibition of deiodinase activity, and the conversion of T 4 to T 3 in kidney tissue and other tissue types is reduced [18]. Hypothalamic-pituitary-thyroid axis dysfunction [19] combined with loss of T 4 in the urine causes triiodothyronine and total T 4 levels to decrease. A decrease in the glomerular ltration rate (GFR) reduces iodine excretion, resulting in an iodine-blocking effect (Wolff-Chaikoff effect) [19]. Song et al. [20] retrospectively analysed 2,284 subjects with normal TSH levels and found that as the estimated GFR (eGFR) decreased in CKD patients, the prevalence of low T 3 syndrome gradually increased; the eGFR was positively correlated with the serum T 3 level and was independent of age and serum protein levels. In another study, reduced free T 3 levels predicted an increased risk of cardiovascular events in CKD patients with proteinuria [21]. In patients with chronic haemodialysis, reduced free T 3 levels were a strong predictor of all-cause mortality [22]. In the present study, among hospitalised elderly male patients, the UN levels in the NTIS group were higher than in the non-NTIS group. Correlation analysis showed triiodothyronine and free T 3 levels were negatively correlated with UN and Cr levels. After adjusting for confounding factors, all-cause mortality was signi cantly increased in CKD patients.
Compared with young, short-term diabetic patients, the older patients are more likely to develop NTIS. Some studies have compared the thyroid function status of diabetic patients according to age, disease course, and blood glucose control status and found lower free T 3 levels in diabetic patients than in normal controls [23,24]; furthermore, the incidence of cardiovascular events in patients with type 2 diabetes and NTIS was signi cantly increased [25]. Aging, long diabetes duration, poor blood sugar control, and several complications can increase the prevalence of NTIS, especially in patients with diabetic nephropathy and ketoacidosis [26,27]. In the present study, triiodothyronine and free T 3 levels were also negatively correlated with the FBG level.
In this study, at the end of the 2-year follow-up, a total of 157 patients had died (18.43% mortality rate). Kaplan-Meier survival analyses showed that the survival rate was signi cantly lower in the NTIS group (64.58%) than in the non-NTIS group (86.52%). Cox proportional hazards models showed that after removing confounding factors, reduced free T 3 levels increased the risk of all-cause death. The ROC analysis showed that when using a free T 3 cut-off level of 3.45 pmol/L, the Youden index was highest, with a sensitivity of 0.675 and speci city of 0.642, indicating that when free T 3 levels are less than 3.45 pmol/L, the risk of death increases. The cut-off level in our study was 3.45 pmol/L which was near to the low limit of the normal range (2.76 pmol/L), indicated that for older men with a free T 3 level < 3.45 pmol/L in spite of higher than 2.76 pmol/L may also have a poor prognosis and merit close clinical attention, however, it is still need to be further demonstrated by larger studies. In a recent study of 1,190 patients with acute heart failure, the survival rates were signi cantly lower in patients with low free T 3 compared with normal levels, and a multivariate Cox proportional hazards model showed that a low free T 3 level was an independent predictor of mortality [28].
Studies on ICU patients and hospitalised chronic patients reported the free T 3 level to be an independent predictor of all-cause mortality [11,29]. Similar to previous studies, a lower free T 3 level in the present study was associated with a worse prognosis in elderly male patients with chronic diseases.
The present study had several limitations including failure to evaluate many factors that affect the patient prognosis. In this retrospective cohort study, thyroid hormone levels were measured at random in patients from different departments according to their doctor's discretion, not according to a standard indication. In addition, although age and BMI were matched between the two groups, the type of care received, treatment plan, treatment timing, severity of the patient's condition, and response to the treatment plan may have affected the patient's condition. Due to the small sample size, many in uencing factors were di cult to quantify. The patients were not strati ed according to the above-mentioned factors. In addition, only elderly male inpatients were analysed. Whether the above study results can be generalised to the general population requires further research.

Conclusion
In summary, among elderly male patients hospitalised for multiple causes, the 2-year survival rate was lower in the NTIS group than non-NTIS group. As the triiodothyronine and free T 3 levels decreased, the protein levels and renal function worsened, and the FBG level increased. A lower free T 3 level was associated with an increased risk of all-cause mortality.