Serum total triiodothyronine (T3) as a predictor of mortality and morbidity in critically ill patients and its correlation of predictability with acute physiology and chronic health evaluation II score: A prospective observational study.

BACKGROUND: The aim is to assess the prognostic value of total T3, total T4, and thyroid-stimulating hormone among critically ill patients admitted to the medical intensive care unit (ICU) in association with mortality and its correlation with the acute physiology and chronic health evaluation II (APACHE II) score.
METHODS: : Our prospective observational study consists of 257 patients without known thyroid diseases admitted to the medical ICU. The baseline characteristics of the patients were recorded, including the APACHE II score and thyroid hormone levels at ICU admission. Based on the primary outcome of mortality, we analyzed the data by appropriate statistical methods. A P < 0.05 was considered significant.
RESULTS: Of the 257 patients included in the study, 47 (18.28%) succumbed to their illnesses. A significant difference in T3 levels (P < 0.001), T4 levels (P < 0.001), and APACHE II score (P < 0.001) was found between the survivors and the nonsurvivors. There was negative correlation observed between T3 and APACHE II score (r = -0.448, P < 0.001) and T4 and APACHE II score (r = -0.221, P ≤ 0.001). Multivariate logistic regression analysis determined T3 to be the only independent predictor of ICU mortality among thyroid hormones. The area under the curve (AUC) for T3 (0.811 ± 0.04) was almost equal to that of the APACHE II score (0.858 ± 0.029). The duration of ICU stay and hospital stay in patients with low T3 was significantly higher compared to patients with normal T3.
CONCLUSION: Serum T3 is a good indicator for predicting mortality and morbidity among critically ill patients. Copyright:

INTRODUCTION

Critical illness is any disease process causing physiological instability that may lead to disability or death within minutes or hours.[1] Such illness is known to cause various hormonal disturbances, including abnormality in thyroid function. This alteration in thyroid function in critically ill patients with no previous thyroid disorder is called nonthyroidal illness syndrome (NTIS) (aka euthyroid sick syndrome or low T3 syndrome). NTIS has been shown to correlate with the severity of disease and increased mortality.

Predicting mortality and morbidity among critically ill patients would enable risk stratification and more intensive care of these patients. While one such method of risk stratification is by using validated scoring systems such as the acute physiology and chronic health evaluation II (APACHE II) score that makes use of various clinical and biochemical parameters, another method is by studying the levels of certain serum markers as predictors. The role of thyroid hormones in predicting mortality among such patients is yet to be ascertained. Thus, we carried out a prospective observational study among critically ill patients admitted in the medical intensive care unit (ICU) to assess the prognostic value of total T3, total T4, and thyroid-stimulating hormone (TSH) in association with mortality and its correlation with the APACHE II score.

METHODS

Study setting and participants

We conducted our study in a secondary care hospital located in the Udupi district of Karnataka in South India. This study was a noninterventional, single-center, prospective observational study in critically ill patients admitted to the medical ICU between August 2016 to August 2017. Our study received ethical approval from the Manipal University Ethics Committee (IEC No. MUEC/008/2016-17). We obtained written informed consent from all the patients or their family members/surrogate. Our manuscript adheres to the STROBE guidelines. All patients >18 years and without prior thyroid disorder were included. Pregnant women and patients on amiodarone or dopamine that could potentially affect thyroid function were excluded.

Data collection

On admission to ICU, all the enrolled patients were subjected to baseline blood investigations including complete blood count, hemoglobin A1C, blood sugars, renal function tests, liver function tests, blood gas analysis, and other relevant tests. Thyroid function tests, i.e., total triiodothyronine (T3), total thyroxine (T4), TSH, were measured within 2 h of ICU admission. T3, T4, and TSH were measured using Elecsys electrochemiluminescence assay (Cobas e 411 analyzer; Roche Diagnostics; Mannheim, Germany). Testing was performed according to the manufacturer's instructions. The normal reference ranges of serum hormone concentrations in our laboratory are as follows: T3 (0.80–2.0 ng/ml); T4 (5.50–12.2 μg/dl); TSH, (0.3 μIU/ml to 4.5 μIU/ml). Based on history, clinical examination, and the worst parameters within the first 24 h of admission, acute physiology score, chronic health score, Glasgow Coma Scale score, and Age score were calculated to get the APACHE II score. Charlson Comorbidity Index (CCI) was calculated for each patient. The patients were followed up till discharge/death and duration of ICU stay, and the total duration of hospital stay was noted. The primary outcome was death in the ICU due to any cause. The association of thyroid hormones with mortality and morbidity and their correlation of predictability with the APACHE II score was assessed.

Sample size

Anticipating 15% of critically ill patients to succumb to their illness and at least 10% higher rate in low T3 group for the power of 80% and 95% confidence, a minimum of 250 critically ill patients were to be studied, considering the study by Harikumar et al.[2]

Data analysis

We used IBM SPSS, Version 20.0. (IBM Corp., Armonk, NY, USA) to perform the statistical data analysis. To assess the association of thyroid hormones with the outcome, T3, T4, and TSH were grouped as low, normal, and high according to our lab reference values (as described in the previous section). Mortality and morbidity between different groups were compared. Categorical data were expressed in absolute numbers and percentages. A Chi-square test was used to find the associations between categorical variables and outcomes. Mean and standard deviation was used to express normally distributed continuous data, and independent samples t-test was used to compare means between the two groups. For nonnormal distribution, the median and interquartile range was reported, and the Mann–Whitney U-test and Kruskal–Wallis analysis of variance were used to compare the median. Spearman's rho coefficient has been reported for the correlation between the APACHE II score and the levels of T3 and T4. Univariate and multivariate logistic regression analysis was carried out to identify independent variables predicting ICU mortality. Receiver operating curve (ROC) analysis was done to further assess the APACHE II score and T3 levels as predictors of mortality. A P < 0.05 was considered statistically significant.

RESULTS

Baseline characteristics

A total of 257 patients were included in the study, with female patients (52.1%) being marginally higher than males (47.9%). Th e age of patients ranged between 22 years to 93 years, with an average age of 66.99 ± 14.47 years. A large proportion of patients 152 (59.1%) were in the age group of 60–80 years. Among the 257 patients, 109 (42.41%) patients had low T3; 76 (29.57%) patients had low T4; 47 (18.28%) patients had high TSH. The median APACHE II score was significantly higher in males as compared to females [13 (9, 20) vs. 11(8, 17), P=0.017]. Infections (52.5%) followed by cardiovascular diseases (28.4%) were the most frequent cause for admission into the ICU [Supplementary Figure 1]. Among infections, respiratory tract infection and its complications (34.63%) were the most common. Some patients had multiple diagnoses at admission.

The baseline clinical and laboratory characteristics of the patients are listed in Table 1. Out of the 257 patients, 210 (81.71%) survived, and 47 (18.28%) expired. The mean age of the survivors was 66.39 ± 14.10 years, and nonsurvivors had a mean age of 69.38 ± 15.92 years. The difference between the mean age of the two groups was not statistically significant. The median CCI was 4 (2,5). A statistically significant difference in T3 levels (0.93 ± 0.30 vs. 0.59 ± 0.28, P < 0.001), T4 levels (6.89 ± 1.84 vs. 5.49 ± 2.16 < 0.001), median APACHE II score [11 (8, 15) vs. 21 (16, 29), P < 0.001], and median CCI [3 (2, 5) vs. 4 (3, 7) P = 0.006] was observed between the survivors and the nonsurvivors.

Table 1

Baseline characteristics and their association with outcome

Variable	Sample (n=257)	Survivors (n=210)	Nonsurvivors (n=47)	P
Gender
Female	134	119	15	0.002
Male	123	91	32
Age (mean±SD) years	66.99±14.47	66.39±4.10	69.68±15.92	0.159
Charlson comorbidity index (median, IQR)	4 (2, 5)	3 (2, 5)	4 (3, 7)	0.006
T3 ng/mL (mean±SD)	0.87±0.32	0.93±0.30	0.59±0.28	<0.001
T4 µg/dL (mean±SD)	6.64±1.98	6.89±1.84	5.49±2.16	<0.001
TSH µIU/mL (mean±SD)	2.94±2.55	2.95±2.55	2.87±2.6	0.848
APACHE II score (median, IQR)	12 (9, 18)	11 (8, 15)	21 (16, 29)	<0.001

SD: Standard deviation; IQR: Interquartile range; APACHE II: Acute Physiology and Chronic Health Evaluation II; TSH: Thyroid-stimulating hormone; T3: Total triiodothyronine; T4: Total thyroxine

Correlation of thyroid profile with acute physiology and chronic health evaluation II score

To identify predictors of mortality, Spearman's rank-order correlation analysis was done between the associated variables (T3, T4, APACHE II score). There was moderate, negative correlation found between T3 and APACHE II score (r = −0.448, P < 0.001), while a mild negative correlation between T4 and APACHE II score (r = −0.221, P < 0.001) was present. This indicates the presence of low T3 and low T4 are associated with increased severity of illness.

A univariate logistic regression analysis [Table 2] was performed to determine the predictors of ICU mortality. The logistic regression model was statistically significant χ2(7) =102.616, P < 0.001. The model explained 53.6% (Nagelkerke R2) of the variance in ICU mortality and correctly classified 87.5% of cases. Among the thyroid hormones, only T3 was found to be a predictor for ICU mortality along with the APACHE II score. The logistic regression model, when adjusted for age, gender, and diabetes mellitus status of the patient, did not bring about significant changes in the results. A forward stepwise multivariate logistic regression analysis was performed with a criterion of 0.05 for entry and 0.10 for removal. It was observed that the probability of predicting ICU mortality by the APACHE II score increased with the addition of T3 (Cox and Snell R2= 0.304, Nagelkerke R2= 0.496) than by APACHE II Score alone (Cox and Snell R2= 0.255, Nagelkerke R2= 0.304).

Table 2

Univariate and multivariate logistic regression analysis for variables predicting intensive care unit mortality

Predictor	Crude			Adjusted*
	Odds ratio (95% CI)		P	Odds ratio (95% CI)	P
T3(ng/ml)	0.047 (0.007-0.321)		0.002	0.033 (0.004- 0.248)	0.001
T4 (μg/dl)	0.869 (0.661- 1.142)		0.314	0.862 (0.652- 1.140)	0.298
TSH (μIU/ml)	0.937 (0.794- 1.105)		0.437	0.924 (0.783- 1.090)	0.347
APACHE II Score	1.203 (1.122- 1.290)		<0.001	1.209 (1.122- 1.302)	<0.001
	Odds ratio	P	−2 log likelihood	Cox and Snell R2	Nagelkerke R2
Model 1
APACHE II score	1.253	<0.001	168.774	0.255	0.416
Model 2
T3	0.034	<0.001	151.278	0.304	0.496
APACHE II score	1.209	<0.001

*Adjusted for age, gender, diabetes mellitus status. Normal values: T3 (0.80- 2.0 ng/ml); T4 (5.50- 12.2 μg/dl); TSH, (0.3- 4.5 μIU/ml). APACHE II: Acute Physiology and Chronic Health Evaluation II; TSH: Thyroid-stimulating hormone; T3: Total triiodothyronine; T4: Total thyroxine; CI: Confidence interval

To further assess T3 and APACHE II as predictors of mortality, the ROC curve was used [Figure 1]. Computation of the area under the ROC [Table 3] revealed that T3 levels (AUC: 0.811) are equally good predictors as compared to the APACHE II score (AUC: 0.858) for predicting mortality among critically ill patients. In conclusion, the APACHE II score and T3 level are predictive of mortality in patients.

Figure 1

(a) Receiver operating curve for T3. (b) Receiver operating curve for acute physiology and chronic health evaluation II

Table 3

Receiver operating curve, area under the curve, sensitivity, and specificity for Acute Physiology and Chronic Health Evaluation II and total triiodothyronine

Test result variable(s)	Area	Asymptotic 95% confidence interval		Predictive factors cut-off point	Sensitivity (%)	Specificity (%)
		Lower bound	Upper bound
APACHE II	0.858	0.801	0.915	15.5*	80.90	77.10
T3	0.811	0.773	0.889	0.775**	76.60	75.20

*Positive if greater than or equal to; **Positive if less than or equal to. APACHE II: Acute Physiology and Chronic Health Evaluation II; T3: Total triiodothyronine

Morbidity assessment

The duration of the ICU stay and the total duration of hospital stay was considered as indicators of morbidity for patients who survived and got discharged.On comparing patients in T3, T4, and TSH groups [Supplementary Table 1], the median (inter quartile range) duration of both ICU stay and hospital stay in patients with low T3 was significantly higher compared to that of patients with normal T3. Using the Kruskal–Wallis test, the association was found to be statistically significant (P < 0.001). However, there was no significant difference in the duration of ICU and hospital stay within the T4 and TSH groups.

DISCUSSION

The state of abnormal thyroid function seen in critical illness is known as NTIS. This alteration of the normal thyroid homeostasis in the critically ill occurs due to dysregulations in the hypothalamus-pituitary-thyroid axis and altered metabolism of thyroid hormones.[3] The changes seen in NTIS are dynamic. In the acute phase of critical illness, often, triiodothyronine (T3) levels are low with elevated rT3 and normal/low thyroxine (T4) and normal TSH. In prolonged critical illness; however, levels of TSH, T3, and T4 are all low.[456]

Considerable controversy exists on whether the thyroid hormone level alterations in NTIS are a physiological adaptation during illness to conserve energy, or whether they are maladaptive and require treatment.[67] Clinical studies show conflicting results with both advocates for and against the treatment of NTIS with thyroid supplementation.[89101112131415] At present, however, no evidence-based consensus or guideline advocates treatment of NTIS in the critically ill; treatment and management of underlying critical illness is the focus.

Some studies that were done among ICU patients, as well as subsets of critically ill patients with sepsis, acute coronary syndrome, neurological patients, multi-trauma patients, and surgical ICU patients, have shown that NTIS, particularly low levels of T3, correlate with disease severity and predict a poor prognosis.[2151617181920212223242526] Conversely, a few studies show no association between NTIS and prognosis.[272829]

In our study of 257 patients, most patients were geriatric, with ages between 60-80 years (59.1%) and a mean age of 66.99 ± 14.47 years. This is consistent with other similar studies among adult ICU patients.[216] The major cause of the critical illness was infectious diseases (52.52%), with respiratory tract infection and its complications being preponderant (34.63%). Cardiovascular diseases were found to be a major cause of critical illness leading to ICU admission in previous studies.[21617] The mortality in our study group was 18.28%, and it was found to be significantly higher among males as compared to females (26.01% vs. 11.1%, P = 0.002), which can be attributed to the greater severity of illness noted in males as described by median APACHE II score [13 (9, 20) vs. 11 (8, 17), P = 0.017]. A statistically significant difference in T3 levels was found between the survivors and the nonsurvivors (P < 0.001). This association of low thyroid hormones with increased severity of illness and mortality was confirmed by a moderate negative correlation of T3 with the APACHE II score (r = −0.448, P < 0.001). A comparison of our study with other similar studies among patients admitted in ICU is shown in Table 4.

Table 4

Comparison with similar studies

Author	Kumar et al.[2]	Chinga-Alayo et al.[17]	Wang et al.[16]	Gutch et al.[30]	Our study
Year	2013	2005	2012	2018	2016- 2017
Main outcome measures	Mortality	Mortality	Mortality	Mortality	Mortality and morbidity
Sample size	100	113	480	270	257
Age in years	58.7±16.9	-	71.71±15.52	38.99±18.32	66.99±14.47
Gender (male, female) %	52, 48	41.6, 58.4	59.79, 40.21	51.1, 48.9	47.9, 52.1
Thyroid parameters
Low T3 (%)	61 (61)	-	23 (4.79)	-	109 (42.41)
Low T4 (%)	14 (14)	-	53 (11.04)	-	76 (29.57)
Low TSH (%)	Low TSH7 (7)	-	High TSH17 (3.54)	-	Low TSH6 (2.3)
APACHE II score
Survivors	-	13±8	11.38±5.60	14.83±5.95	11 (8, 15)
Nonsurvivors	-	20±8	19.49±6.85	25.00±9.75	21 (16, 29)
Mortality (%)	36	22	19.16	30	18.28
Association of thyroid parameters with mortality
T3	Nonsurvivors had low meanT3 than the survivors (P=0.004)	Nonsurvivors had low mean T3 than the survivors (P<0.001)	Nonsurvivors had low mean T3 than survivors (P<0.0001)	Survivors had low mean T3 than nonsurvivors (P=0.007)	Nonsurvivors had low mean T3 than survivors (P<0.001)
T4	No significant difference in serum T4 levels between survivors and nonsurvivors (P=0.544)	No significant difference in serum T4 levels between survivors and nonsurvivors (P=0.23)	Nonsurvivors had low mean T4 than survivors (P<0.0001)	No significant difference in serum T4 levels between survivors and nonsurvivors (P=0.742)	Nonsurvivors had low mean T4 than survivors (P <0.001)
TSH	No significant difference in serum TSH levels between survivors and nonsurvivors (P=0.208)	Nonsurvivors had significantly low mean TSH than survivors (P<0.01)	Nonsurvivors had low mean TSH than survivors (P=0.0022)	No significant difference in serum TSH levels between survivors and nonsurvivors (P=0.640)	No significant difference in serum TSH levels between survivors and nonsurvivors (P=0.848)

APACHE II: Acute Physiology and Chronic Health Evaluation II; TSH: Thyroid-stimulating hormone; T3: Total triiodothyronine; T4: Total thyroxine

A significant association of low T3 with mortality was also seen in the other studies on general ICU patients.[21617] However, conflicting results were observed in the association of low T4 and low TSH with mortality compared to our study. Whereas in our study, low T4 showed a similar association with mortality in the study by Erick Chinga-Alayo et al. and Harikumar et al., there was no significant difference in serum T4 levels between survivors and nonsurvivors.[217] In the study by Feilong Wang et al., T3, T4, TSH, fT3, fT4, and APACHE II score were all found to be associated with mortality in ICU.[16] Of these, only fT3 and APACHE II scores were determined as independent predictors of mortality. In our study, we found the T3 and APACHE II score as independent predictors of mortality in ICU patients. Among the survived patients in this study, the patients with low T3 also had increased duration of ICU and hospital stay and higher APACHE II scores, indicating a possible association of NTIS with increased morbidity as well.

There is scarce published evidence available from the Indian subcontinent that assesses the prognostic value of total T3 among critically ill adult patients admitted to the medical ICU. A literature search among the studies in India showed only two such studies, both conducted in North India, with one large study of 270 patients and another of a smaller sample size of 100 patients.[230] In the study by Gutch et al., the mean total T3 levels were found to be lower among survivors as compared to nonsurvivors (P = 0.007).[30] However, the fT3 and fT4 levels were lower in nonsurvivors as compared to survivors.

Our study has certain limitations. (1) Although we have excluded patients of known thyroid illness prior to admission from history and clinical examination, the inclusion of some patients with intrinsic thyroid illness cannot be ruled out as we did not follow-up the patients after discharge. (2) Although we excluded patients who were taking amiodarone and those treated with dopamine, many other drugs such as corticosteroids, propranolol, barbiturates, benzodiazepines, phenytoin, lithium, octreotide, and furosemide may have interfered with thyroid function tests. However, it is difficult to adjust for all these potential confounders in clinical practice because so many drugs are involved, and many of them essential for management in critical patients. (3) Other thyroid parameters such as fT3 and fT4, rT3 were not analyzed due to financial constraints.

CONCLUSION

NTIS is common in patients with a critical illness. The presence of low T3 and low T4 are associated with increased severity of illness. Low T3 is a good independent predictor of mortality, almost comparable to APACHE II SCORE. The addition of T3 to the APACHE II score increased the predictability of the APACHE II score. In the developing countries, where financial constraints considerably influence management decisions, serum T3 can be utilized as a cheap and useful independent predictor of prognosis.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.

Research quality and ethics statement

The authors of this manuscript declare that this scientific work complies with reporting quality, formatting, and reproducibility guidelines set forth by the EQUATOR Network. The authors also attest that this clinical investigation was determined to require the Institutional Review Board/Ethics Committee review, and the corresponding protocol/approval number is 1212092-1. We also certify that we have not plagiarized the contents in this submission and have done a Plagiarism Check.