Early levothyroxine sodium administration and clinical outcomes in patients with sepsis: a MIMIC-IV database analysis

Abstract

Sepsis is frequently complicated by non-thyroidal illness syndrome (NTIS). Although thyroid hormones are essential for cardiovascular and respiratory stability, their supplementation in sepsis remains controversial due to conflicting evidence. This study aimed to evaluate the impact of thyroid hormone supplementation on 28 day mortality, mechanical ventilation, vasopressor use, ICU length of stay, and changes in SOFA scores in patients with sepsis. We analyzed 20,231 septic patients from the Medical Information Mart for Intensive Care Database IV (MIMIC-IV) database, comparing those treated with levothyroxine sodium within seven days to a control group using 1:4 propensity score matching. Early levothyroxine sodium administration was significantly associated with increased 28 day mortality in both the original (HR 2.48, 95% CI 1.96–3.15; P < 0.001) and propensity score matching (PSM) cohorts (HR 2.38, 95% CI 1.75–3.23; P < 0.001). The treatment group required higher cumulative norepinephrine equivalents (P = 0.006) and longer vasoactive support duration. Multistate modeling revealed that the treatment group had fewer days alive without mechanical ventilation (17.8 vs. 21.8 days; P < 0.001) and fewer days discharged alive (11.3 vs. 13.4 days; P = 0.006). In patients with less severe sepsis, levothyroxine sodium administration should be approached with particular caution, as it may be associated with adverse clinical outcomes.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-43822-z.

Introduction

Sepsis is characterized as life-threatening organ dysfunction resulting from a dysregulated host response to infection¹. Recent data indicate that sepsis causes 21.4 million annual deaths, accounting for 31.5% of total global mortality and remaining a leading cause of morbidity and mortality worldwide^2,3. Specific and effective therapies for sepsis and its associated organ dysfunction remain currently limited⁴.

Sepsis patients frequently manifest non-thyroidal illness syndrome (NTIS)^5,6, characterized by an early decline in triiodothyronine (T3) and elevated free triiodothyronine (rT3), alongside normal or decreased levels of thyroxine (T4) and thyroid-stimulating hormone (TSH)^7,8. Decreased thyroid hormone levels correlate with organ dysfunction severity and 28 day mortality, and T3 is an independent risk factor for sepsis and septic shock^9–13. Additionally, study show that NTIS is a risk factor for prolonged mechanical ventilation¹⁴. However, thyroid hormone supplementation in patients with sepsis remains controversial¹⁵. Thyroid hormone decline in sepsis was considered an adaptation¹⁶. Supplementation therapy may increase oxygen demand and arrhythmia risk^17,18. While studies in cardiovascular diseases provide a physiological basis^19,20., the clinical benefit of thyroid hormone supplementation for septic patients remains a subject of ongoing investigation.

Thyroid hormones are essential regulators of cardiovascular homeostasis and play a critical role in maintaining physiological cardiac function²¹. These maintain myocardial energy supply during stress by enhancing fatty acid oxidation (FAO), adenosine triphosphate (ATP) production, and mitochondrial biogenesis²². Clinical studies demonstrate that thyroid hormone supplementation significantly improves cardiac function and hemodynamic status in patients with acute myocardial infarction or chronic heart failure^23,24. In animal models of septic cardiomyopathy (SCM), T3 pretreatment alleviates myocardial inflammation, fibrosis, and calcium dysregulation²⁵. Evidence regarding thyroid hormone supplementation in sepsis remains limited. A small-scale study found that such supplementation improved survival in patients with decreased both T3 and T4, but increased mortality in those with isolated low T3 ²⁶. The finding underscores the complexity of thyroid hormone supplementation in sepsis, necessitating further research to define the optimal timing, target populations, and therapeutic regimens. Furthermore, the scarcity of registered randomized controlled trials in this field limits the availability of reliable evidence in the near term. Therefore, we utilized the MIMIC database to evaluate real-world evidence regarding thyroid hormone supplementation therapy in sepsis and its association with 28 day mortality, mechanical ventilation, and vasopressor requirements.

Methods

Data source

The retrospective observational study used the Medical Information Mart for Intensive Care MIMIC-IV (version 3.1) database, which was developed and maintained by the Laboratory for Computational Physiology at the Massachusetts Institute of Technology (MIT). The MIMIC-IV database integrated comprehensive clinical information from patients in the intensive care unit (ICU) of the Beth Israel Deaconess Medical Center in Boston, Massachusetts, USA, from between 2008 and 2022 ²⁷. One author, Jing Chu, passed the Collaborative Institutional Training Initiative Examination and obtained permission to extract data (certification number: 69733112). The dataset used in this research was newly and independently extracted using Structured Query Language (SQL) from the raw database specifically for this study. MIMIC-IV is a de-identified, credentialed database hosted on PhysioNet. The Institutional Review Board at the Beth Israel Deaconess Medical Center and the Institutional Review Board of the Massachusetts Institute of Technology approved the creation of this research resource. According to the official documentation on the PhysioNet website, the Institutional Review Board at the Beth Israel Deaconess Medical Center also granted a waiver of informed consent. We confirm that all methods were performed in accordance with the relevant guidelines and regulations.

Participants

Our study included patients diagnosed with sepsis based on the MIMIC-IV database. According to the Sepsis-3 guidelines, sepsis was defined as a suspected or documented infection coupled with an acute increase of ≥ 2 points in the Sequential Organ Failure Assessment (SOFA) score¹. The inclusion criteria were: (1) age ≥ 18 years; (2) first time in ICU; (3) time in ICU ≥ 24 h; (4) diagnosis of Sepsis 3.0 within 72 h of ICU admission; and (5) administration of levothyroxine sodium within 7 days after the confirmation of sepsis or no use of any levothyroxine sodium medications. Any patient with a prior history of hypothyroidism or who received a diagnosis during their current ICU stay was excluded from the study.

Data extraction and outcomes

Clinical data were retrieved from the MIMIC-IV database using SQL. To ensure statistical robustness, only variables with sufficient completeness were included in the final analysis. These variables included age, gender, pulse oximetry (SpO2), and laboratory results for white blood cell count (WBC), and serum creatinine, as well as arterial lactate from blood gas analysis. Initial severity was assessed using the Acute Physiology Score III (APS III), the Sequential Organ Failure Assessment (SOFA) score, and the Charlson Comorbidity Index (CCI). Chronic comorbidities, such as hypertension, diabetes, cancer, and respiratory, cardiovascular, liver, renal, or neurological diseases, were identified using ICD-9 and ICD-10 codes. Baseline SOFA scores and laboratory parameters were obtained from the first recorded values within the initial twenty four hours of ICU admission. To assess clinical progression, the delta SOFA score was calculated as the difference between the SOFA score on the day of levothyroxine sodium administration and the following day. For the control group, a matching index date was assigned based on the medication start day of the treated patient. Daily clinical data were extracted from the database to ensure a standardized evaluation of organ dysfunction changes across both groups. For all study variables including clinical data, laboratory findings, and severity scores, we only included the first recorded value within the first 24 h of ICU admission.

The primary exposure was the administration of levothyroxine sodium within seven days of Sepsis-3 diagnosis compared with a control group that received no levothyroxine sodium therapy. Our primary outcome was 28 day mortality. Secondary outcomes primarily focused on cumulative norepinephrine equivalent (NEE) doses, vasoactive support duration^28,29, ventilator-free days, and discharge status. Additionally, ICU length of stay and the delta SOFA score were recorded. The delta SOFA score was calculated as the difference between the SOFA score on the day after and the day of medication to assess organ dysfunction progression.

Study design and analytical approach

To minimize selection bias and confounding, propensity score matching (PSM) was performed using the ‘MatchThem’ package with the ‘within’ approach. Propensity scores were estimated via a multivariable logistic regression model based on baseline covariates. Patients were then matched in a 1:4 ratio using a nearest-neighbor algorithm and a 0.01 caliper without replacement. Covariate balance was confirmed using the absolute standardized mean difference (SMD), with an SMD below 0.1 considered indicative of adequate balance. Regarding data integrity, variables with more than 25% missing data were excluded unless clinically essential. Remaining missing values were addressed through multivariate imputation by chained equations (MICE) using the ‘mice’ package with the predictive mean matching (PMM) method to generate five imputed datasets. Finally, estimates across the imputed datasets were pooled according to Rubin’s rules.

Statistical analysis

Statistical analyses were performed using R software (version 4.5.1). Normally distributed continuous variables were expressed as mean ± standard deviation (SD), and compared between groups using the Student’s t-test. Non-normally distributed continuous variables were presented as median [interquartile range (IQR)], and compared using the Wilcoxon rank-sum test. Categorical variables were expressed as frequencies and percentages [n (%)], and compared using the Chi-square test or Fisher’s exact test as appropriate. Survival outcomes were visualized with the Kaplan-Meier method and compared using the log-rank test. Multivariable Cox proportional hazards models were employed to identify independent risk factors for 28 day mortality and estimate hazard ratio (HR) with 95% confidence interval (CI). A multistate model was used to assess ventilator-free days and discharge status, with the results visualized through a stacked probability plot³⁰. A P value of less than 0.05 was considered statistically significant.

Results

Population and baseline characteristics

In the MIMIC-IV cohort, eligible ICU admissions with Sepsis-3 were identified according to the predefined criteria. The study included a total of 20,231 ICU patients with Sepsis-3, consisting of 237 received levothyroxine sodium and 19,994 did not (Fig. 1). Before Propensity Score Matching, significant differences were observed in baseline characteristics between the two groups. Compared with the control group, patients received levothyroxine sodium had a higher proportion of females (49.8% vs. 37.5%, P < 0.001) and exhibited higher illness severity, indicated by higher APSIII scores (median 53.00 vs. 46.00, P < 0.001), SOFA scores (P = 0.037), lactate levels (P = 0.014), and creatinine levels (P < 0.001). Additionally, the prevalence of nephropathy (P = 0.008) and neurological disease (P < 0.001) was significantly higher in the treatment group. After Propensity Score Matching, all covariates were well balanced, with absolute standardized mean differences < 0.05 (Table 1). A Love plot illustrated the distribution of absolute standardized mean differences before and after matching (Figure S2).

Fig. 1.

Flowchart of construction of the cohort. This flow diagram illustrates the systematic screening process of participants derived from the Medical Information Mart for Intensive Care IV (MIMIC-IV v3.1) database. Initially, 41,295 adult patients diagnosed with Sepsis-3 were identified. After restricting the cohort to the first ICU admission, 27,881 patients remained. A total of 7650 patients were subsequently excluded based on the following pre-specified criteria: (1) age under 18 years (n = 0); (2) ICU length of stay less than 24 h (n = 2775); (3) Sepsis-3 diagnosis occurring more than 72 h after ICU admission (n = 1301); and (4) presence of pre-existing hypothyroidism (n = 3341). The final analytical cohort consisted of 20,231 patients. Participants were then stratified into two groups: the Treatment Group (n = 237), defined as patients who received levothyroxine sodium within 7 days following their sepsis diagnosis, and the Control Group (n = 19994), consisting of patients with no record of levothyroxine sodium administration. Abbreviations: ICU, intensive care unit; Medical Information Mart for Intensive Care, MIMIC.

Table 1.

Baseline characteristics before and after Propensity Score Matching.

Variable		Before Propensity Score Matching				After Propensity Score Matching
		Control group 19,994 case	Treatment group 237 case	P value	SMD	Control group 929 case	Treatment group 235 case	P value	SMD
Admission_age (years)		66.74[55.51, 77.69]	67.70 [54.67, 78.24]	0.797	0.016	66.68 [55.69, 77.53]	67.71 [55.09, 78.37]	0.755	0.038
Gender (%)	M	12,497 (62.5)	119 (50.2)	< 0.001	0.25	466 (50.2)	119 (50.6)	0.93	0.022
	F	7497 (37.5)	118 (49.8)			463 (49.8)	116 (49.4)
APS III score		46.00 [34.00, 62.00]	53.00 [38.00, 68.00]	< 0.001	0.256	50.00 [38.00, 68.00]	52.00 [38.00, 68.00]	0.452	0.031
SOFA score		2.00 [1.00, 4.00]	3.00 [1.00, 5.00]	0.037	0.093	2.00 [1.00, 4.00]	3.00 [1.00, 5.00]	0.516	0.013
CCI		5.00 [3.00, 7.00]	5.00 [2.00, 8.00]	0.969	0.002	5.00 [3.00, 7.00]	5.00 [2.00, 8.00]	0.664	0.027
SpO2(%)		98.00 [95.00, 100.00]	98.00 [95.00, 100.00]	0.393	0.047	98.00 [96.00, 100.00]	98.00 [95.00, 100.00]	0.592	0.037
Lactate (mmol/L)		1.90[1.30, 2.90]	2.30 [1.40, 3.30]	0.014	0.146	1.90 [1.30, 3.00]	2.10 [1.30, 3.20]	0.318	0.023
WBC(×109/L)		11.80 [8.30, 16.40]	12.65 [8.62, 16.98]	0.097	0.12	12.70 [8.80, 17.90]	12.60 [8.70, 16.85]	0.724	0.028
Creatinine serum(mg/dL)		1.00 [0.70, 1.60]	1.20 [0.80, 1.90]	< 0.001	0.148	1.10 [0.80, 1.80]	1.20 [0.80, 1.90]	0.103	0.013
Hypertension (%)	No	7342 (36.7)	101 (42.6)	0.071	0.121	374 (40.3)	99 (42.1)	0.655	0.03
	Yes	12,652 (63.3)	136 (57.4)			555 (59.7)	136 (57.9)
Diabetes (%)	No	17,327 (86.7)	208 (87.8)	0.689	0.033	819 (88.2)	206 (87.7)	0.973	0.016
	Yes	2667 (13.3)	29 (12.2)			110 (11.8)	29 (12.3)
Cancer (%)	No	16,943 (84.7)	202 (85.2)	0.906	0.014	793 (85.4)	200 (85.1)	1	0.01
	Yes	3051 (15.3)	35 (14.8)			136 (14.6)	35 (14.9)
Respiratory disease (%)	No	6935 (34.7)	68 (28.7)	0.063	0.129	270 (29.1)	68 (28.9)	1	0.008
	Yes	13,059 (65.3)	169 (71.3)			659 (70.9)	167 (71.1)
Cardiovascular disease (%)	No	3154 (15.8)	40 (16.9)	0.709	0.03	155 (16.7)	40 (17.0)	0.947	0.018
	Yes	16,840 (84.2)	197 (83.1)			774 (83.3)	195 (83.0)
Liver disease (%)	No	16,343 (81.7)	204 (86.1)	0.102	0.118	801 (86.2)	202 (86.0)	0.852	0.035
	Yes	3651 (18.3)	33 (13.9)			128 (13.8)	33 (14.0)
Nephropathy disease (%)	No	10,294 (51.5)	101 (42.6)	0.008	0.178	395 (42.5)	99 (42.1)	0.791	0.021
	Yes	9700 (48.5)	136 (57.4)			534 (57.5)	136 (57.9)
Neurological disease (%)	No	10,428 (52.2)	96 (40.5)	< 0.001	0.235	396 (42.6)	96 (40.9)	0.795	0.032
	Yes	9566 (47.8)	141 (59.5)			533 (57.4)	139 (59.1)

Continuous variables are presented as median [interquartile range (IQR)], and categorical variables are presented as frequencies and percentages [n (%)]. Comparisons between groups were performed using the Wilcoxon rank-sum test for continuous variables and the Chi-square test or Fisher’s exact test for categorical variables. The Standardized Mean Difference (SMD) was used to evaluate covariate balance, with an SMD below 0.1 considered indicative of adequate balance. Statistical significance was defined as P < 0.05.

(Abbreviations: SMD, Standardized Mean Difference; APS III, Acute Physiology Score III; SOFA, Sequential Organ Failure Assessment; CCI, Charlson Comorbidity Index; WBC, white blood cell)

Survival analysis and multivariable Cox proportional hazards model

The unadjusted Kaplan-Meier curves showed an adverse effect of levothyroxine sodium on 28 day survival, compared with patients in the control group (log-rank tests, P < 0.001) (Fig. 2). This survival disadvantage persisted in the propensity score-matched cohort, as visually confirmed by the survival curves (Figure S3). Consistent with these observations, in the multivariable Cox proportional hazards model of the original cohort, patients treated with levothyroxine sodium had significantly inferior survival compared with those without treatment (HR = 2.48, 95% CI = 1.96 to 3.15, P < 0.001) (Table S1). Analyses in the propensity score-matched cohort displayed almost the same tendency (HR = 2.38, 95% CI = 1.75 to 3.23, P < 0.001) (Table 2). Additionally, patients with higher illness severity (APSIII score and SOFA score) had adverse survival outcomes (P < 0.05).

Fig. 2.

The 28 day Kaplan-Meier survival curves before Propensity Score Matching. The plot compares the 28 day survival probability between the Treatment Group (cyan line) and the Control Group (red line) in the original cohort. The shaded areas surrounding the survival curves represent the 95% confidence intervals (CI). A log-rank test was used to compare the survival distributions between the two groups, showing a significant difference (p < 0.0001). The “Number at risk” table below the curves displays the number of patients remaining in each group at 7 day intervals. Statistical significance was defined as P < 0.05. Abbreviations: PSM, Propensity Score Matching.

Table 2.

Univariate and multivariate Cox proportional hazards regression analysis after Propensity Score Matching.

Variable	Univariate analysis		Multivariate analysis
	HR (95% CI)	P value	HR (95% CI)	P value
Medicine_status (Yes)	2.13 (1.60–2.82)	< 0.001	2.38 (1.75–3.23)	< 0.001
Admission_age	1.00 (0.99–1.01)	0.869	1.00 (0.98–1.03)	0.834
Gender (Male)	1.04 (0.78–1.40)	0.779	1.08 (0.73–1.62)	0.68
APS III score	1.03 (1.02–1.04)	< 0.001	1.03 (1.02–1.03)	< 0.001
SOFA score	1.14 (1.07–1.20)	< 0.001	1.00 (0.94–1.07)	0.917
CCI	1.05 (0.98–1.12)	0.177	1.01 (0.86–1.20)	0.852
SpO2	0.96 (0.92–1.01)	0.104	0.99 (0.93–1.05)	0.643
Lactate	1.17 (1.10–1.23)	< 0.001	1.09 (0.99–1.20)	0.067
WBC	1.03 (1.01–1.05)	0.013	1.00 (0.98–1.03)	0.774
Serum creatinine	1.11 (0.99–1.24)	0.081	0.95 (0.83–1.09)	0.49
Hypertension (Yes)	0.74 (0.55–0.99)	0.043	0.70 (0.51–0.97)	0.033
Diabetes (Yes)	1.14 (0.73–1.77)	0.556	0.93 (0.43–2.04)	0.846
Cancer (Yes)	1.33 (0.93–1.91)	0.121	1.24 (0.59–2.61)	0.525
Respiratory disease (Yes)	2.14 (1.19–3.85)	0.016	1.39 (0.82–2.36)	0.208
Cardiovascular disease (Yes)	0.90 (0.58–1.38)	0.611	1.06 (0.63–1.78)	0.831
Liver disease (Yes)	2.51 (1.61–3.92)	< 0.001	1.37 (0.61–3.08)	0.393
Nephropathy disease (Yes)	1.86 (1.28–2.70)	0.002	1.05 (0.70–1.57)	0.8
Neurological disease (Yes)	1.31 (0.69–2.51)	0.348	1.24 (0.55–2.76)	0.534

For categorical variables, the results for “Medicine_status (Yes)” are compared against the reference group “No”, and results for “Gender (Male)” are compared against the reference group “Female”. The multivariate model included all variables listed in the table. An HR > 1 indicates an increased risk of 28 day mortality. Statistical significance was defined as P < 0.05.

Abbreviations: HR, hazard ratio; CI, confidence interval; APS III, Acute Physiology Score III; SOFA, Sequential Organ Failure Assessment; CCI, Charlson Comorbidity Index; SpO2, peripheral oxygen saturation

Vasopressor requirements and clinical course

Compared with the control group, patients treated with levothyroxine sodium required a significantly higher cumulative dose of norepinephrine equivalents (median, 266.05 vs. 143.89 mcg/kg, P = 0.006) and a longer duration of vasoactive support (median, 36.60 vs. 25.98 h, P = 0.017) (Fig. 3a and b; Table 3). In contrast, no significant difference was found between the two groups regarding ICU length of stay (P = 0.973) and the change in SOFA scores during the ICU stay (P = 0.223) (Fig. 3c and d; Table 3).

Fig. 3.

Comparison of secondary outcomes between the two groups after Propensity Score Matching. Cumulative dose of norepinephrine equivalents (mcg/kg). This panel illustrates the total dosage of vasopressors required by patients (measured in mcg/kg), indicating the intensity of hemodynamic support. Cumulative vasoactive support duration (hours). This boxplot displays the total time (in hours) patients remained on vasoactive medications, reflecting the duration of cardiovascular instability. Delta Sequential Organ Failure Assessment (SOFA) score. Represents the dynamic change in organ dysfunction from baseline to a post-treatment assessment; a higher delta value may indicate a greater degree of organ recovery or progression. Intensive Care Unit length of stay (ICU LOS) in days. Compares the total duration of ICU hospitalization (in days) between the two cohorts. In each boxplot, the thick horizontal line represents the median value, the box boundaries indicate the interquartile range (IQR, 25th–75th percentiles), and the whiskers extend to 1.5 times the IQR. Discrete points represent outliers. P-values were calculated using the Wilcoxon rank-sum test. Statistical significance was defined as P < 0.05. Abbreviations: NEE, norepinephrine equivalents; ICU, intensive care unit; los: length of stay; SOFA: Sequential Organ Failure Assessment; LOS, length of stay; IQR, interquartile range.

Table 3.

Comparison of hemodynamic support requirements between the two groups after Propensity Score Matching.

Outcome	Control Group 497 case	Treatment Group 146 case	P Value
Cumulative NEE dose (mcg/kg), median [IQR]	143.89 [25.21-597.43]	266.05 [61.93-679.59]	0.006
Duration of vasoactive support (hours), median [IQR]	25.98 [6.97–65.97]	36.60 [15.81–74.11]	0.017

Data are presented as median [interquartile range]. P-values were calculated using the Wilcoxon rank-sum test. Statistical significance was defined as P < 0.05.

Abbreviations: NEE, norepinephrine equivalents; IQR, interquartile range

Clinical trajectory and ventilatory outcomes

The clinical trajectory of patients was analyzed using a multistate model and visualized via a stacked probability plot (Fig. 4). Compared with the control group, patients treated with levothyroxine sodium spent significantly fewer days alive without mechanical ventilation (17.8 days vs. 21.8 days, P < 0.001) and fewer days discharged alive (11.3 days vs. 13.4 days, P = 0.006). The total duration of hospitalization was shorter in the treatment group (8.7 days vs. 10.8 days; P < 0.001). However, no significant difference was observed in the mean duration of invasive mechanical ventilation (2.2 days vs. 2.3 days, P = 0.62). The 28 day mortality rate was significantly higher in the treatment group (34.00% vs. 18.50%, P < 0.001), as illustrated by the larger red area representing death in the stacked probability plot (Table 4).

Fig. 4.

Stacked probability plot for the multistate model after Propensity Score Matching. The plot illustrates the time-dynamic probabilities of patients being in four mutually exclusive states over 28 days for the Control (left) and Treatment (right) groups. The colored areas represent the following states: dead (red), hospitalized with mechanical ventilation (MV; light green), hospitalized without MV (dark green), and discharged alive (orange). At any specific time point on the x-axis, the vertical distance between the curves represents the state occupation probability (the proportion of patients in that state). The transition from the bottom to the top (orange to red) reflects the patients’ progression from recovery/discharge to mortality. Abbreviations: MV, Mechanical ventilation.

Table 4.

Estimates of mean time spent in clinical states and mortality rates derived from the multistate model.

Outcome	Treatment(N = 235)	Control(N = 929)	P_Value
Alive without MV	17.8 days	21.8 days	< 0.001
Alive with MV	2.2 days	2.3 days	0.62
Hospitalized Duration	8.7 days	10.8 days	< 0.001
Discharged Days	11.3 days	13.4 days	0.006
Mortality Rate	34.00%	18.50%	< 0.001

Hospitalized = Alive with MV+ Alive without MV.

This table provides quantitative estimates derived from the four-state multistate model visualized in Fig. 4. The values for “Alive without MV,” “Alive with MV,” “Hospitalized Duration,” and “Discharged Days” represent the expected mean number of days spent in each state during the 28 day follow-up period. The mortality rate represents the cumulative probability of death at day 28. P-values indicate the statistical significance of differences between the Treatment (n = 235) and Control (n = 929) groups. Statistical significance was defined as P < 0.05.

Abbreviations: MV, Mechanical ventilation

Discussion

In our retrospective study, propensity-matched analysis showed that levothyroxine sodium was associated with higher mortality and more ventilator and vasopressor use, suggesting that thyroid hormone supplementation in sepsis requires extreme caution.

A 2024 randomized trial showed that T3 supplementation reduced mortality in septic shock patients with combined low T3 and T4, but increased mortality in those with isolated low T3 ²⁶. In our study, the median SOFA score was 3 with an interquartile range of 1 to 5, reflecting a relatively mild population. While we considered distinguishing between sepsis and septic shock, such categorical labels may not accurately represent the specific severity of our cohort. According to the Sepsis 3.0 guidelines, septic shock is a more severe stage within the sepsis spectrum rather than a parallel category. Although shock is generally associated with higher severity, a patient in shock without multi-organ failure may still present with a SOFA score as low as 2 to 4. Therefore, clinical labels cannot simply substitute for the precise severity indicated by actual SOFA scores. These findings align with the 2024 RCT, suggesting that thyroid hormone supplementation may be harmful in patients. The administration of levothyroxine sodium in this study reflects real-world clinical practices at Beth Israel Deaconess Medical Center. Since NTIS is closely linked to organ failure severity and poor prognosis in sepsis, clinicians may attempt supplementation based on the hypothesis that thyroid hormones could help maintain cardiovascular stability and support myocardial function. However, the notably low administration rate of 1.2% in our cohort confirms that thyroid hormone therapy was not a routine institutional practice, but rather an individualized decision based on clinical judgment and empirical experience in the absence of international guidelines. This practice mirrors a broader clinical phenomenon; for instance, in regions such as China, some practitioners consider thyroid hormone supplementation a potentially supportive measure for sepsis patients despite the lack of consensus. Our study leverages these real-world data to objectively evaluate the clinical impact of such individualized treatment attempts. Notably, while the 2024 trial reported shortened mechanical ventilation duration in severely ill patients, our analysis of a generally milder population found no improvement in ventilator-free days. Administration of levothyroxine sodium also showed no benefits for secondary outcomes. It did not improve vasopressor requirements or the duration of mechanical ventilation. In theory, thyroid hormones enhance myocardial contractility and sensitivity to catecholamines³¹. However, these benefits are less evident in milder patients and may increase myocardial oxygen demand³². Thyroid hormones are essential for respiratory muscle function³³, but exogenous supplementation may increase metabolic rate and carbon dioxide production, leading to delayed ventilator weaning^34,35. These findings further suggest that thyroid hormone supplementation may be harmful in patients with less severe sepsis.

However, as a retrospective analysis of the MIMIC-IV database, this study has limitations that require a cautious interpretation of the results. First, the treatment group included only about 200 patients because thyroid hormone supplementation is relatively rare in sepsis. While matching was used to balance the baseline data, this small sample size may still lead to bias. Furthermore, although we used strict protocols to adjust for initial differences, some factors from the original illness might still exist. The higher physiological stress in the treatment group could have influenced mortality in ways that statistical methods cannot fully capture. Consequently, these results should be interpreted as a preliminary clinical alert and a basis for new hypotheses rather than as final evidence of direct harm. Second, a theoretical limitation of this study is the focus on levothyroxine monotherapy. Sepsis is known to impair the peripheral conversion of T4 to T3 by altering deiodinase activity³⁶. Consequently, T4 supplementation might be considered conceptually inadequate for addressing the true hormonal deficit. However, the choice of T4 in our analysis was dictated by its predominant use within the MIMIC-IV database. This reflects the clinical reality that levothyroxine remains the most common and mature thyroid intervention in real world settings. While T3 is often preferred in controlled research environments^{25,26,37–39}, its routine use in general intensive care is limited. Evaluating T4 outcomes in this setting remains crucial because it represents the most frequent treatment strategy implemented by clinicians. Future research should investigate whether different formulations or combination therapies provide superior benefits for hormonal recovery and clinical outcomes in sepsis. Finally, T4 dosages and administration routes were not standardized. Also, data on thyroid hormone level changes before and after levothyroxine sodium treatment were not recorded. Therefore, our findings linking thyroid therapy to higher mortality and poor outcomes should be interpreted with caution. We also consider that thyroid hormone supplementation might delay the recovery of the hypothalamic pituitary thyroid axis. External hormones can suppress this natural feedback loop and potentially hinder clinical recovery. Due to the retrospective nature of our data, we could not monitor these dynamic endocrine changes after treatment. Our work remains a preliminary and qualitative exploration in a real world setting where treatment protocols were not uniform. This study is limited by its retrospective design. Although we used propensity score matching to balance the groups, this method shows associations instead of definite causes. These results serve to generate new hypotheses for future testing. A randomized controlled trial is the best way to evaluate how levothyroxine sodium affects patients with sepsis. Our findings suggest that early treatment may be harmful to patients with less severe illness. Future prospective studies are needed to identify the right patients and the correct timing for thyroid hormone therapy in the intensive care unit. Our findings primarily serve as a clinical alert. Thyroid hormone supplementation should be used with extreme caution in patients with sepsis. Future research must be conducted under strict monitoring. Continuous and dynamic tracking of thyroid hormone levels is essential.

Conclusion

In conclusion, for patients with sepsis, administration of levothyroxine sodium is associated with increased 28 day mortality. Our results show no clinical benefit for this population regarding survival or organ recovery. Consequently, thyroid hormone supplementation is not supported for patients with sepsis due to the observed risks of adverse outcomes.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

J.C and M.C contributed equally to this work and shared first authorship. J.Y.G was responsible for the study of concepts and design. J.C and M.C were responsible for data extraction and drafting the initial manuscript. Q.W.Y and Z.J.G were responsible for statistical analysis and data interpretation. K.X and S.L were responsible for literature retrieval and data validation. J.Y.G provided critical revision of the manuscript for important intellectual content. All authors had full access to all the data in the study and took responsibility for the integrity of the data and the accuracy of the data analysis. All authors read and approved of the final manuscript.

Funding

This study was funded by the Hebei Provincial Health Commission Key Scientific and Technological Research Program [20221205].

Data availability

The data that support the findings of this study are available from PhysioNet but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of PhysioNet.

Declarations

Competing interests

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.