Ferroptosis and sepsis-induced myocardial injury: a prospective clinical study

Abstract

Background

Sepsis-induced myocardial injury (SIMI) is a common and severe complication of sepsis associated with increased mortality, yet its underlying mechanisms are not fully elucidated. Ferroptosis, an iron-dependent form of regulated cell death driven by lipid peroxidation and oxidative stress, has been implicated in preclinical models of septic cardiomyopathy. However, clinical evidence in humans remains limited, and the specificity of related biomarkers requires careful interpretation.

Methods

This prospective observational study enrolled 180 ICU-admitted sepsis patients (91 with SIMI and 89 without) from January to June 2025. Serum ferroptosis-related biomarkers, including malondialdehyde (MDA), lipid peroxidation (LPO), glutathione (GSH), reactive oxygen species (ROS), and ferrous iron (Fe²⁺), were measured at 24 h post-admission. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curves, and independent predictors were identified via multivariable logistic regression adjusting for clinical confounders.

Results

Compared with the non-SIMI group, SIMI patients showed significantly altered ferroptosis-related markers (all P < 0.001): higher MDA [6.08 (5.38–6.67) vs. 4.47 (3.92–5.30) nmol/mL], LPO, ROS, and Fe²⁺, and lower GSH. ROC analysis demonstrated moderate-to-good diagnostic discrimination for SIMI (AUC 0.746–0.825), with Fe²⁺ showing the highest AUC (0.825), followed by ROS (0.793) and MDA (0.787); these outperformed BNP but were inferior to cTnI. Multivariable logistic regression identified Fe²⁺ (OR = 11.883, P < 0.001) as the only independent ferroptosis-related predictor of SIMI, alongside cTnI and APACHE II score. The full model exhibited excellent discrimination (AUC = 0.989) and good calibration (Hosmer-Lemeshow P = 0.935).

Conclusions

Elevated ferroptosis-related biomarkers reflecting systemic oxidative stress and iron dysregulation are significantly associated with SIMI, with Fe²⁺ emerging as an independent predictor. These biomarkers provide additive diagnostic value beyond traditional cardiac markers but are not specific for ferroptosis as a regulated cell death mechanism and likely represent broader redox imbalance in severe sepsis. These findings support further investigation into iron-targeted and oxidative stress-modulating strategies in septic cardiac dysfunction.

Introduction

Sepsis is a life-threatening condition characterized by a dysregulated systemic inflammatory response to infection, frequently leading to multi-organ dysfunction and high mortality rates [1]. Recent studies have highlighted the critical role of gut microbiota and host metabolic reprogramming in sepsis tolerance and organ injury [2]. Among the complications of sepsis, sepsis-induced myocardial injury (SIMI) is a critical contributor to adverse outcomes, with studies reporting an incidence of up to 50% in septic patients and a significantly elevated risk of mortality [3]. SIMI is marked by myocardial cell damage and cardiac dysfunction, driven by complex mechanisms including inflammation, oxidative stress, and various forms of programmed cell death, such as apoptosis, ferroptosis, necroptosis, autophagy, and pyroptosis [4]. Emerging evidence indicates that gut microbiota-derived metabolites play important roles in modulating distant organ injury through systemic metabolic and immune regulation [5]. Of these, ferroptosis, an iron-dependent form of regulated cell death, has emerged as a pivotal mechanism in the pathogenesis of myocardial injury [6, 7].

Ferroptosis is characterized by the accumulation of lipid peroxides and iron-dependent oxidative damage, which disrupts cellular redox homeostasis and leads to cell death [8]. This process has been implicated in various cardiovascular diseases, including hypertrophic cardiomyopathy, myocardial infarction, ischemia/reperfusion injury, and heart failure, where hallmark molecular features of ferroptosis—such as reduced glutathione (GSH) levels and increased lipid peroxidation—are consistently observed [9–11]. In the context of sepsis, preclinical studies have demonstrated that ferroptosis contributes significantly to myocardial damage. For instance, Li et al. found that in lipopolysaccharide (LPS)-induced septic cardiomyopathy in mice, ferroptosis was a key driver of cardiomyocyte injury, and its inhibition markedly ameliorated cardiac dysfunction [12]. Similarly, Wang et al. reported that targeting ferroptosis through pharmacological interventions reduced myocardial injury in septic animal models, highlighting its therapeutic potential [13].

The molecular underpinnings of ferroptosis involve the interplay of iron metabolism, lipid peroxidation, and oxidative stress. Notably, gut microbiota-derived metabolites have been shown to regulate ferroptosis through modulation of key pathways such as glutathione peroxidase 4 (GPX4) expression [14]. Ferroptosis is initiated by increased transport of polyunsaturated fatty acids, leading to lipid overload, followed by iron-mediated catalysis of lipid peroxidation, which generates excessive reactive oxygen species (ROS) and disrupts cellular integrity [15]. Key ferroptosis-related biomarkers, such as acyl-CoA synthetase long-chain family member 4 (ACSL4), glutathione peroxidase 4 (GPX4), and prostaglandin-endoperoxide synthase 2 (PTGS2), have been identified as critical regulators in sepsis and SIMI [16]. Recent studies have shown that dysregulated expression of these biomarkers in septic myocardial tissue correlates with disease severity and poor prognosis, suggesting their utility as diagnostic and prognostic indicators [17].

Despite these advances, the role of ferroptosis in human SIMI remains underexplored, with most evidence derived from animal models or bioinformatics analyses. Clinical studies investigating the dynamic changes of ferroptosis-related biomarkers in septic patients are scarce, and their integration with traditional myocardial injury markers (e.g., troponin I, B-type natriuretic peptide) and inflammatory markers (e.g., procalcitonin, C-reactive protein) has not been systematically evaluated. This prospective clinical study aims to address these gaps by examining the role of ferroptosis in SIMI through the dynamic assessment of ferroptosis-related biomarkers in a cohort of septic patients. By constructing a prognostic prediction model, we seek to elucidate the clinical significance of ferroptosis in SIMI and provide novel insights into its diagnostic and therapeutic implications.

Methods

Study population

This prospective study enrolled 180 patients with sepsis, comprising 91 patients with sepsis-induced myocardial injury (SIMI) and 89 patients without SIMI (No-SIMI), admitted to the Intensive Care Unit (ICU) of Shanghai General Hospital between January 2025 and June 2025. The inclusion criteria for enrolled patients were: (1) Diagnosis of sepsis meeting the Sepsis-3.0 criteria [1]; (2) Confirmed infection focus; (3) Age ≥ 18 years; (4) ICU stay ≥ 24 h. Exclusion criteria included: (1) History of acute myocardial infarction, coronary artery disease, cardiomyopathy, valvular heart disease, or congenital heart disease; (2) History of cardiopulmonary resuscitation, defibrillation, or cardioversion; (3) Pulmonary embolism, chronic anemia, trauma, or burns; (4) Concomitant autoimmune diseases or malignancies; (5) Pregnancy or lactation.Acute coronary syndrome was excluded via clinical history, ECG, and coronary angiography where indicated.

Sepsis-3.0 was defined as life-threatening organ dysfunction due to infection with a Sequential Organ Failure Assessment (SOFA) score ≥ 2 compared to baseline [1]. SIMI was defined as the presence of myocardial injury with cardiac dysfunction, meeting either of the following criteria:

(1) LVEF < 50% on echocardiography at ICU admission; OR (2) LVEF decrease ≥ 10% from baseline (within 3 months) [18].

Patients were classified as the SIMI group if they met either criterion, while the non-SIMI group included patients with sepsis but without evidence of myocardial injury (normal LVEF and cTnI). Patients with pre-existing chronic heart failure (LVEF < 50% within 3 months) were excluded [4].

Primary outcome

Sepsis-induced myocardial injury (SIMI).

Secondary outcomes

1. All-cause mortality within 28 days;

2. 28-day survival time;

3. Need for renal-replacement therapy during ICU stay;

4. Prognostic accuracy (AUC) of ferroptosis markers for 28-day mortality.

Reagents

Ferroptosis-related biomarkers were measured to assess lipid peroxidation and oxidative stress. Reactive oxygen species (ROS) react with polyunsaturated fatty acids in biomembranes, enzymes, and nucleic acids, generating lipid peroxidation products such as malondialdehyde (MDA) and 4-hydroxynonenal (HNE), which alter membrane fluidity and permeability, leading to cellular dysfunction [7]. The following assay kits were used: Lipid Peroxide (LPO) (E-BC-K176-M, China), Fe²⁺ (E-BC-F101, China), Glutathione (GSH) (E-BC-K030-M, China), MDA (E-EL-0060, China), and ROS (Shanghai Shuangying Biotechnology Co., Ltd., China).

Data collection

Upon ICU admission, baseline data were collected, including demographic information (sex, age, comorbidities), cardiac function parameters (left ventricular ejection fraction [LVEF] cardiac output [CO], cardiac index [CI]), cardiac biomarkers (troponin I [cTnI], B-type natriuretic peptide [BNP]), inflammatory markers (C-reactive protein [CRP], interleukin-6 [IL-6], procalcitonin [PCT]), and ferroptosis markers (malondialdehyde [MDA], lipid peroxidation [LPO], glutathione [GSH], reactive oxygen species [ROS], ferrous iron [Fe²⁺]). Blood samples for ferroptosis markers were collected at 24 h post-ICU admission in sterile heparin tubes, centrifuged at 3200 rpm for 15 min at 4 °C, and stored at -80 °C. Echocardiography was performed within 6 h of ICU admission.

Ethical standards

The study was approved by the Medical Ethics Committee of Shanghai General Hospital (Approval No. [2025]KY012) and adhered to the Declaration of Helsinki. Patient confidentiality was maintained by anonymizing data, using only hospitalization numbers for validation. Informed consent was obtained from all participants and their legal guardians.

Clinical follow-up

Follow-up was conducted via telephone at 28 days post-enrollment to assess patient survival status.

Statistical analysis

Data were analyzed using SPSS 26.0 (IBM, Armonk, NY). Normality was assessed using the Shapiro-Wilk test. Variables with skewed distribution (MDA, LPO, ROS, Fe²⁺, GSH, cTnI, BNP, PCT, IL-6, etc.) were analyzed using non-parametric tests and presented as median (IQR). Normally distributed variables were presented as mean ± SD. Categorical variables were reported as frequencies (percentages) and compared using chi-square tests. Differences between SIMI and No-SIMI groups were assessed using the Mann-Whitney U test for skewed variables and independent t-tests for normally distributed variables. Multivariable logistic regression was performed to identify independent predictors of SIMI, adjusting for ferroptosis markers, cardiac biomarkers, and disease severity scores (full model variables listed in Table 3). Model performance was evaluated by area under the receiver operating characteristic curve (AUC), Hosmer-Lemeshow test for calibration, and 5-fold cross-validation for internal validation. A p-value ≤ 0.05 was considered statistically significant.

Table 3.

Multivariable logistic regression model

Variables	B	SE	Wald	pvalue	Exp(B)	95%CIofExp(B)
const	-12.8238	4.8536	6.9808	0.008	0	(-22.3366, -3.3109)
SOFA	0.2068	0.1643	1.585	0.208	1.230	(-0.1152, 0.5288)
APACHEII	-0.1796	0.081	4.92	0.027	0.836	(-0.3383, -0.0209)
cTnI(µg/l)	62.6764	13.3064	22.1863	<0.001	1.660E + 27	(36.5963, 88.7566)
BNP(pg/ml)	0.0001	0.0007	0.0092	0.924	1.000	(-0.0013, 0.0014)
MDA(nmol/ml)	0.5998	0.4087	2.1535	0.142	1.822	(-0.2013, 1.4009)
LPO(ng/ml)	-0.0362	0.089	0.1653	0.684	0.965	(-0.2105, 0.1382)
GSH(ng/ml)	0.0037	0.3418	0.0001	0.991	1.004	(-0.6661, 0.6736)
ROS(IU/ml)	0.0033	0.0039	0.6965	0.404	1.003	(-0.0044, 0.011)
Fe2+(µmol/l)	2.4751	0.7436	11.0784	0.001	11.883	(1.0176, 3.9326)

Note: MDA = Malondialdehyde; ROS = Reactive Oxygen Species; APACHE II = Acute Physiology and Chronic Health Evaluation II; SOFA = Sequential Organ Failure Assessment; PCT = Procalcitonin; OR = Odds Ratio; CI = Confidence Interval

Results

Patient demographics and baseline characteristics

A total of 180 sepsis patients were enrolled (SIMI, n = 91; No-SIMI, n = 89) (Fig. 1). Baseline characteristics are summarized in Table 1. Age [65 (54, 72) vs. 62 (48.5, 69.5) years, P = 0.154] and sex distribution (63.7% male vs. 65.2%, P = 0.841) were comparable between groups. Comorbidity rates were similar except for a higher prevalence of chronic digestive diseases in the No-SIMI group (20.2% vs. 8.8%, P = 0.029). Noradrenaline use was significantly higher in the SIMI group (60.4% vs. 35.9%, P = 0.001). Disease severity scores were higher in SIMI patients: SOFA 6 (4, 10) vs. 5 (3, 6.5), P < 0.0001; APACHE II 16 (11, 23) vs. 14 (9.5, 19), P = 0.026. The SIMI group also showed higher 28-day mortality (39.6% vs. 15.7%, P = 0.001), shorter 28-day survival time [28 (12, 28) vs. 28 (28, 28) days, P = 0.001], and greater need for red blood cell transfusion (33.0% vs. 14.6%, P = 0.004). Need for dialysis was numerically higher in SIMI (20.9% vs. 10.1%, P = 0.063).

Fig. 1.

Study flow diagram showing patient enrollment and stratification

Table 1.

Baseline demographic and clinical characteristics of patients with and without sepsis-induced myocardial injury (n = 180)

Characteristics	SIMI(N = 91)	No-SIMI(N = 89)	p	Total(N = 180)
Sex, % male	58(63.7)	58(65.2)	0.841	116(64.4)
Age, years	65 (54,72)	62(48.5,69.5)	0.154	64 (51.25, 70.75)
Comorbidities, n (%)
Hypertension	42(46.1)	36(40.4)	0.456	78(43.3)
Diabetes	16(17.6)	18(20.2)	0.651	34(18.9)
Immune diseases	8(8.8)	5(5.6)	0.411	13(7.2)
CKD	14(15.4)	6(6.7)	0.065	20(11.1)
Chronic digestive diseases	8(8.8)	18(20.2)	0.029*	26(14.4)
COPD	15(16.5)	11(12.3)	0.431	26(14.4)
Infection Site
Intra-Abdominal, n (%)	28 (30.8)	36 (40.4)	0.230	64 (35.6)
Pulmonary, n (%)	59 (64.8)	52 (58.4)	0.465	111 (61.7)
Urinary tract, n (%)	4 (4.4)	5 (5.6)	0.973	9 (5.0)
Skin/soft tissue/joint, n (%)	5 (5.5)	6 (6.7)	0.970	11 (6.1)
Bloodstream, n (%)	6 (6.6)	9 (10.1)	0.559	15 (8.3)
Central nervous system, n (%)	4.0 (4.4)	7.0 (7.9)	0.509	11.0 (6.1)
Pathogen of infection
Negative bacteria, n (%)	26 (28.6)	32 (36.0)	0.368	58 (32.2)
Positive bacteria, n (%)	20 (22.0)	14 (15.7)	0.379	34 (18.9)
Virus, n (%)	16 (17.6)	9 (10.1)	0.217	25 (13.9)
Fungus, n (%)	7 (7.7)	16 (18.0)	0.065	23 (12.8)
Noradrenaline use, (%)	55(60.4)	32(35.9)	0.001**	87(48.3)
Mechanical ventilation, (%)	60(65.9)	47(52.8)	0.095	107(59.4)
SOFA	6(4,10)	5(3, 6.5)	<0.001***	5.5(4,8)
APACHE II	16(11,23)	14(9.5, 19)	0.026*	15(10.5,20.75)
Mortality (%)	36(39.6)	14(15.7)	0.001**	50(27.8)
28 days survival time (days)	28(12, 28)	28(28,28)	0.001**	28(17.25, 28)
Need for dialysis, (%)	19(20.9)	9(10.1)	0.063	28(15.5)
Red blood cells transfused, (%)	30(33.0)	13(14.6)	0.004**	43(23.9)
Iron supplementation, (%)	16(17.6)	23(25.8)	0.179	39(21.7)

CKD, chronic kidney disease; COPD, chronic obstructive pulmonary disease; APACHE, Acute Physiology And Chronic Health Evaluation; SOFA, sequential organ failure assessment. Values are expressed as the mean ± standard deviation or the median and interquartile range. *P < 0.05,**p<0.01,***p<0.001.For normally distributed data, the mean ± standard deviation is used, and for non-normal distribution, the median, lower quartile and upper quartile are used

Laboratory findings, ferroptosis markers, and hemodynamic variables

Laboratory parameters are presented in Table 2. Ferroptosis-related markers were markedly altered in the SIMI group (all P < 0.001):MDA: 6.08 (5.38–6.67) nmol/mL vs. 4.47 (3.92–5.30) nmol/mL; LPO: 26.14 (23.18–28.62) ng/mL vs. 19.15 (16.14–23.01) ng/mL; GSH: 5.26 (4.75–5.75) ng/mL vs. 6.79 (6.04–7.82) ng/mL; ROS: 633.23 (553.86–708.18) IU/mL vs. 500.95 (450.24–542.84) IU/mL; Fe²⁺: 2.85 (2.67–3.35) µmol/L vs. 2.22 (1.92–2.48) µmol/L; Inflammatory markers PCT [6.00 (1.10, 39.11) vs. 2.02 (0.69, 7.43) ng/mL, P = 0.006] and IL-6 [220.80 (53.47, 1000.00) vs. 102.10 (29.09, 298.00) pg/mL, P = 0.010] were significantly higher in SIMI, while CRP and TNF-α showed no difference. Cardiac injury markers cTnI and BNP were markedly elevated in SIMI (both P < 0.001). Endotoxin levels were higher in SIMI [0.07 (0.03, 0.14) vs. 0.05 (0.03, 0.10) EU/mL, P = 0.029].

Table 2.

Laboratory parameters and ferroptosis-related markers in patients with and without myocardial injury (n = 180)

Laboratoryresults	SIMI(N = 91)	No-SIMI(N = 89)	p	Total(N = 180)
MDA (nmol/ml)	6.08(5.38–6.67)	4.47(3.92–5.30)	< 0.001***	5.38(4.29–6.37)
LPO (ng/ml)	26.14(23.18–28.62)	19.15(16.14–23.01)	< 0.001***	23.09(17.69–27.28)
GSH (ng/ml)	5.26(4.75–5.75)	6.79(6.04–7.82)	< 0.001***	5.89(4.97–7.16)
ROS (IU/ml)	633.23(553.86-708.18)	500.95(450.24-542.84)	< 0.001***	553.37(480.56-661.81)
Fe²⁺ (µmol/l)	2.85(2.67–3.35)	2.22(1.92–2.48)	< 0.001***	2.60(2.14–2.98)
PCT (ng/ml)	6.00 (1.10, 39.11)	2.02 (0.69, 7.43)	0.006**	3.12 (0.81, 19.18)
CRP (mg/L)	151.60 (100.00, 224.90)	154.40 (89.50, 227.00)	0.923	152.75 (91.75, 224.95)
Endotoxin (Eu/ml)	0.07 (0.03, 0.14)	0.05 (0.03, 0.10)	0.029*	0.05 (0.03, 0.12)
IL-6 (pg/ml)	220.80 (53.47, 1000.00)	102.10 (29.09, 298.00)	0.010*	121.50 (40.05, 379.27)
cTNI (µg/l)	0.14 (0.08, 0.68)	0.02 (0.01, 0.03)	<0.001***	0.05 (0.02, 0.14)
BNP (pg/ml)	295.00 (128.50, 644.00)	75.00 (37.00, 133.00)	<0.001***	132.00 (58.53, 389.75)
TNF-α (pg/ml)	19.70 (13.80, 33.10)	17.80 (12.40, 29.00)	0.235	19.45 (13.35, 30.80)
LVEF (%)	43.81 ± 3.27	57.76 ± 3.71	<0.001***	50.71 ± 7.81
LVEF ≥ 50 (%)	1.1	100.0	<0.001***	50.0
LVEF<50 (%)	98.9	0.0	<0.001***	50.0
SBP (mmHg)	116.20 ± 21.71	116.89 ± 20.60	0.827	116.54 ± 21.11
DBP (mmHg)	67.02 ± 12.92	66.73 ± 13.65	0.883	66.88 ± 13.25
MAP (mmHg)	84.30 (74.70, 92.00)	83.70 (78.30, 92.00)	0.736	84.30 (75.60, 92.08)
CO (L/min)	8.04 (7.05, 8.51)	8.42 (7.18, 9.49)	0.022*	8.33 (7.15, 9.10)
CI (L/min/m²)	4.28 (4.00, 5.16)	4.86 (4.26, 5.54)	0.005**	4.53 (4.03, 5.44)
P (bpm)	100.00 (86.00, 116.00)	100.00 (86.75, 112.50)	0.708	100.00 (86.00, 115.75)

Note: ROS, reactive oxygen species; MDA, malonaldehyde༛LPO, lipid peroxidation༛GSH, glutathione ༛PCT, Procalcitonin; CRP, C-reactive protein; IL-6, Interleukin-6;cTNI, Cardiac troponin I; BNP, B-type natriuretic peptide; TNF-α,Tumor necrosis factor-alpha; LVEF, Left ventricular ejection fraction, SBP, Systolic Blood Pressure; DBP, Diastolic Blood Pressure; MAP, Mean Arterial Pressure; CO, Cardiac Output; CI, Cardiac Index; *p<0.05, **p<0.01, ***p<0.001༛For normally distributed data, the mean ± standard deviation is used, and for non-normal distribution, the median, lower quartile and upper quartile are used

Left ventricular ejection fraction (LVEF) was significantly lower in the SIMI group (43.81 ± 3.27% vs. 57.76 ± 3.71%, P < 0.001), with 98.9% of SIMI patients having LVEF < 50% compared to 0% in the No-SIMI group. Hemodynamic parameters showed lower cardiac output [CO: 8.04 (7.05, 8.51) vs. 8.42 (7.18, 9.49) L/min, P = 0.022] and cardiac index [CI: 4.28 (4.00, 5.16) vs. 4.86 (4.26, 5.54) L/min/m², P = 0.005] in the SIMI group, while blood pressure and heart rate were comparable (all P > 0.05).

Multivariable analysis of ferroptosis biomarkers in SIMI

Multivariable logistic regression (full model, N = 180) identified several independent predictors of SIMI (Table 3). Among ferroptosis markers, Fe²⁺ was significantly associated with SIMI (OR = 11.883, P < 0.001), while ROS (OR = 1.003, P = 0.404), MDA (OR = 1.822, P = 0.142), LPO (OR = 0.965, P = 0.684) and GSH (OR = 1.004, P = 0.991) showed no significant independent correlation. For disease severity scores, the APACHE II score was an independent protective factor (OR = 0.836, P = 0.027), whereas the SOFA score had no significant association with SIMI (OR = 1.230, P = 0.208). Moreover, cTnI was strongly correlated with SIMI (OR = 1.660 × 10²⁷, P < 0.001), and BNP presented no statistical significance (OR = 1.000, P = 0.924).The model demonstrated excellent discrimination (AUC = 0.989) and good calibration (Hosmer-Lemeshow test: χ²=2.99, P = 0.935) (Figs. 2 and 3). Internal validation using 5-fold cross-validation yielded a mean AUC of 0.892, confirming robust generalizability (Fig. 4).

Fig. 2.

ROC curve of the multivariable logistic regression model for predicting SIMI (AUC = 0.9891)

Fig. 3.

Calibration plot of the multivariable logistic regression model for sepsis‑induced myocardial injury (SIMI)

Fig. 4.

Internal validation ROC curves of the stepwise multivariable logistic regression model using 5-fold cross-validation for sepsis-induced myocardial injury (SIMI)

Diagnostic discrimination value of ferroptosis markers for SIMI

Receiver operating characteristic (ROC) analysis (Fig. 5) demonstrated that ferroptosis markers had varying diagnostic performance for SIMI (Table 4). Fe²⁺ showed the highest AUC [0.825 (0.714–0.936), cutoff 2.543 µmol/L, sensitivity 83.5%, specificity 82.0%], followed by ROS [0.793 (0.674–0.911)], MDA [0.787 (0.667–0.906)], LPO [0.746 (0.619–0.873)], and GSH [0.757 (0.632–0.882)]. cTnI exhibited the highest overall AUC [0.992 (0.966–1.000)], while BNP had an AUC of 0.774 (0.651–0.896). Fe²⁺, ROS, and MDA were superior to BNP but inferior to cTnI.

Fig. 5.

Receiver operating characteristic (ROC) curves evaluating the SIMI diagnostic discrimination ability of ferroptosis markers (MDA, LPO, ROS, GSH, Fe²⁺,BNP, cTNI) in 180 sepsis patients, with area under the curve (AUC) and P-values.MDA: Malondialdehyde; LPO: Lipid Peroxidation༛ROS: Reactive Oxygen Species༛GSH: Glutathione༛BNP: B-type Natriuretic Peptide༛cTnI: Cardiac Troponin I༛SIMI: Sepsis-Induced Myocardial Injury

Table 4.

Diagnostic performance of biomarkers for SIMI

Marker	AUC (95% CI)	Cutoff	Sensitivity	Specificity	PPV	NPV
MDA(nmol/ml)	0.787 (0.667–0.906)	5.364	0.780	0.753	0.763	0.770
LPO(ng/ml)	0.746 (0.619–0.873)	23.041	0.769	0.764	0.769	0.764
GSH(ng/ml)	0.757(0.632–0.882)	5.9668	0.802	0.775	0.785	0.793
ROS(IU/ml)	0.793 (0.674–0.911)	551.656	0.780	0.764	0.772	0.773
Fe2+(µmol/l)	0.825 (0.714–0.936)	2.543	0.835	0.820	0.826	0.830
cTnI(µg/l)	0.992 (0.966-1.000)	0.050	1.000	0.989	0.989	1.000
BNP(pg/ml)	0.774 (0.651–0.896)	126.000	0.769	0.730	0.745	0.756
APACHEII	0.591 (0.448–0.735)	23.000	0.286	0.910	0.765	0.555
SOFA	0.649 (0.510–0.789)	8.000	0.440	0.809	0.702	0.585

ROS, reactive oxygen species; DA, malonaldehyde; LPO, lipid peroxidation; GSH, glutathione; cTNI, Cardiac troponin I; BNP, B-type natriuretic peptide; SIMI: Sepsis-Induced Myocardial Injury.ApacheII: Acute Physiology and Chronic Health Evaluation II; SOFA: Sequential Organ Failure Assessment

Serum levels of ferroptosis markers stratified by survival status

Violin plots (Fig. 6) illustrated the distribution of ferroptosis markers in SIMI and No-SIMI groups, further stratified by 28-day survival/death. In both groups, MDA, LPO, and ROS levels were higher and GSH levels were lower in the death subgroup compared with the survival subgroup, with more pronounced shifts observed in the SIMI group. These patterns support the association of ferroptosis severity with both myocardial injury and poor outcome.

Fig. 6.

Violin plots comparing serum levels of ferroptosis markers (MDA, LPO, ROS, GSH) in SIMI (n = 91) and non-SIMI (n = 89) groups, stratified by survival and death subgroups, highlighting differences in marker distribution.MDA: Malondialdehyde; LPO: Lipid Peroxidation༛ROS: Reactive Oxygen Species༛GSH: Glutathione༛SIMI: Sepsis-Induced Myocardial Injury. *p<0.05, **p<0.01, statistically significant

Prognostic performance for 28-day mortality

ROC analysis for 28-day mortality (Fig. 7) showed moderate performance of ferroptosis markers (Table 5). APACHE II (AUC = 0.870) and SOFA (AUC = 0.876) had the highest predictive accuracy, followed by BNP (0.753) and cTnI (0.732). Ferroptosis markers showed lower AUCs: ROS 0.618, MDA 0.594, Fe²⁺ 0.594, LPO 0.552, and GSH 0.560.

Fig. 7.

Receiver operating characteristic (ROC) curves evaluating the 28-day Death discrimination ability of ferroptosis markers (MDA, LPO, ROS, GSH, Fe²⁺,BNP, cTNI, ApacheII, SOFA) in 180 sepsis patients, with area under the curve (AUC) and P-values.MDA: Malondialdehyde; LPO: Lipid Peroxidation༛ROS: Reactive Oxygen Species༛GSH: Glutathione༛BNP: B-type Natriuretic Peptide༛cTnI: Cardiac Troponin I༛SIMI: Sepsis-Induced Myocardial Injury.ApacheII: Acute Physiology and Chronic Health Evaluation II; SOFA: Sequential Organ Failure Assessment

Table 5.

Diagnostic performance of biomarkers for 28-day death

Marker	AUC (95% CI)	Cutoff	Sensitivity	Specificity	PPV	NPV
MDA(nmol/ml)	0.594 (0.433–0.754)	5.787	0.580	0.669	0.403	0.806
LPO(ng/ml)	0.552 (0.389–0.714)	23.041	0.620	0.538	0.341	0.787
GSH(ng/ml)	0.560 (0.399–0.722)	5.734	0.640	0.592	0.376	0.810
ROS(IU/ml)	0.618 (0.459–0.776)	597.344	0.560	0.654	0.384	0.794
Fe2+(µmol/l)	0.594 (0.434–0.755)	2.467	0.720	0.492	0.353	0.821
cTnI(µg/l)	0.732 (0.587–0.876)	0.070	0.680	0.685	0.453	0.848
BNP(pg/ml)	0.753 (0.613–0.894)	200.000	0.760	0.708	0.500	0.885
APACHEII	0.870 (0.760–0.980)	17.000	0.880	0.738	0.564	0.941
SOFA	0.876 (0.769–0.984)	7.000	0.800	0.800	0.606	0.912

ROS, reactive oxygen species; MDA, malonaldehyde༛LPO, lipid peroxidation༛GSH, glutathione ༛cTNI, Cardiac troponin I; BNP, B-type natriuretic peptide; SIMI: Sepsis-Induced Myocardial Injury.ApacheII: Acute Physiology and Chronic Health Evaluation II; SOFA: Sequential Organ Failure Assessment

Comparison of predictive performance: model 1: baseline clinical scores (SOFA/APACHE II) + cTnI/BNP vs. model 2: model 1 + ferroptosis markers for SIMI

To enhance the clinical utility of ferroptosis-related biomarkers for predicting sepsis-induced myocardial injury (SIMI), we developed an optimized predictive model by integrating ferroptosis biomarkers(MDA/LPO/GSH/ROS/Fe²⁺)with established clinical severity scores(SOFA score, APACHE II score and the IL-6).

As shown in Fig. 8, although the AUC of Model 2 was slightly lower than that of Model 1, both the net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were positive, indicating that the ferroptosis markers indeed provided additional diagnostic value: NRI = 0.3763 (reclassification improvement): Model 2 can more accurately classify patients into the correct risk categories. IDI = 0.0518 (discrimination improvement): Model 2 performs better in differentiating cases from controls.

Fig. 8.

Comparison of predictive performance: Model 1: Baseline clinical scores (SOFA/APACHE II) + cTnI/BNP vs. Model 2: Model 1 + Ferroptosis markers for Sepsis-Induced Myocardial Injury (SIMI).ΔAUC = -0.0004, NRI = 0.3763,IDI = 0.0518

Sensitivity analysis

Among critically ill patients with SOFA score > 5

Multivariable logistic regression (sensitivity analysis restricted to patients with SOFA>5, N = 90) identified Fe²⁺ as an independent predictor of SIMI among ferroptosis markers (Table 6). Fe²⁺ was significantly associated with SIMI after adjustment for all covariates (OR = 62.983, P = 0.018). Other ferroptosis markers showed no significant independent correlation: MDA (OR = 2.034, P = 0.383), LPO (OR = 1.165, P = 0.241), GSH (OR = 1.626, P = 0.430), and ROS (OR = 1.003, P = 0.697). For disease severity scores, the SOFA score had no statistical significance (OR = 1.568, P = 0.290), while the APACHE II score was significantly correlated with SIMI (OR = 0.755, P = 0.029). Additionally, cTnI was strongly associated with SIMI (OR = 9.379 × 10¹⁶, P = 0.002), and BNP showed no significant association (OR = 1.001, P = 0.483).The model demonstrated excellent discrimination (AUC = 0.986) and good calibration (Hosmer-Lemeshow test: χ²=15.547, P = 0.030) (Figs. 9 and 10). Internal validation using 5-fold cross-validation yielded a mean AUC of 0.970, confirming robust generalizability (Fig. 11).

Table 6.

Multivariable logistic regression was performed for sensitivity analysis among critically ill patients with SOFA score > 5

Variables	B	SE	Wald	pvalue	Exp(B)	95%CIofExp(B)
const	-24.0881	9.7484	6.1057	0.014	0	(-43.1947, -4.9815)
SOFA	0.4495	0.4244	1.1217	0.290	1.568	(-0.3823, 1.2813)
APACHEII	-0.2804	0.1283	4.7765	0.029	0.756	(-0.5319, -0.0289)
cTnI(µg/l)	39.0798	12.7654	9.3721	0.002	9.379E + 16	(14.0601, 64.0996)
BNP(pg/ml)	0.0011	0.0015	0.4924	0.483	1.001	(-0.0019, 0.004)
MDA(nmol/ml)	0.7102	0.8139	0.7613	0.383	2.034	(-0.8851, 2.3054)
LPO(ng/ml)	0.1523	0.1299	1.3754	0.241	1.165	(-0.1022, 0.4069)
GSH(ng/ml)	0.4861	0.6155	0.6238	0.430	1.626	(-0.7202, 1.6925)
ROS(IU/ml)	0.0032	0.0081	0.1512	0.697	1.003	(-0.0128, 0.0191)
Fe2+(µmol/l)	4.1429	1.7575	5.5564	0.018	62.983	(0.6982, 7.5876)

Note: MDA = Malondialdehyde; ROS = Reactive Oxygen Species; APACHE II = Acute Physiology and Chronic Health Evaluation II; SOFA = Sequential Organ Failure Assessment; PCT = Procalcitonin; OR = Odds Ratio; CI = Confidence Interval

Fig. 9.

ROC curve of the multivariable logistic regression model for predicting SIMI for sensitivity analysis among critically ill patients with SOFA score > 5(AUC = 0.9857)

Fig. 10.

Calibration plot of the multivariable logistic regression model for sepsis‑induced myocardial injury (SIMI) for sensitivity analysis among critically ill patients with SOFA score > 5

Fig. 11.

Internal validation ROC curves of the stepwise multivariable logistic regression model using 5-fold cross-validation for sepsis-induced myocardial injury (SIMI) for sensitivity analysis among critically ill patients with SOFA score > 5

Discussion

The present prospective cohort study demonstrates a robust association between elevated ferroptosis-related biomarkers and sepsis-induced myocardial injury (SIMI) in critically ill patients. Patients with SIMI exhibited significantly higher serum levels of MDA, LPO, ROS, and Fe²⁺, together with lower GSH, independent of several clinical confounders in multivariable modeling. These findings extend preclinical evidence of ferroptosis involvement in septic cardiomyopathy to a human clinical setting and highlight the potential additive value of oxidative stress/iron dysregulation markers beyond conventional cardiac injury biomarkers.

Our results align with earlier experimental studies showing that lipid peroxidation and iron overload drive cardiomyocyte death in lipopolysaccharide-induced septic models [12, 13]. However, several critical caveats must be considered when interpreting these data as evidence of ferroptosis. Importantly, the biomarkers measured (MDA, LPO, ROS, GSH, Fe²⁺) are not specific to ferroptosis as a regulated cell death pathway; they primarily reflect broader systemic redox imbalance and oxidative stress, which are common in severe sepsis regardless of the presence of cardiomyocyte ferroptosis. These changes may arise from multi-organ dysfunction, hemolysis, inflammatory iron redistribution, or sepsis-associated mitochondrial dysfunction [19, 20]. Although Fe²⁺ remained an independent predictor of SIMI after extensive adjustment (including SOFA, APACHE II, cTnI and BNP), we cannot exclude residual confounding by overall disease severity, as SIMI patients had markedly higher baseline SOFA and APACHE II scores. The observed changes are systemic rather than cardiac-specific; without myocardial tissue biopsy or advanced imaging (e.g., cardiac magnetic resonance with T2* mapping for iron or lipid peroxidation imaging), we cannot confirm cardiomyocyte ferroptosis as the direct driver of injury [21]. This limitation is particularly relevant given that ferroptosis is increasingly recognized as a context-dependent process rather than a uniform cell-death pathway across organs [10].

Notably, the multivariable logistic regression model (which demonstrated excellent discrimination and good calibration upon internal validation) identified Fe²⁺, rather than MDA or ROS, as the only independent ferroptosis-related predictor among the five markers measured. This finding aligns with preclinical reports that emphasize ferrous iron as the central driver through Fenton reaction-mediated lipid peroxidation in controlled animal models of septic cardiomyopathy [6]. It suggests that, in the complex clinical setting of human sepsis, circulating ferrous iron may exert more direct pathogenic effects on myocardium than lipid peroxidation products such as MDA and ROS. Possible explanations include: (i) Fe²⁺ directly catalyzes Fenton chemistry, generating hydroxyl radicals that initiate lipid peroxidation locally within cardiomyocytes, whereas circulating MDA and ROS represent downstream systemic spillover that is more susceptible to dilution and metabolic clearance; (ii) after adjusting for disease severity scores (SOFA, APACHE II) which strongly correlate with systemic oxidative stress, the independent signal carried by Fe²⁺ may reflect iron-specific dysregulation rather than global redox imbalance; and (iii) the high collinearity among oxidative stress markers may attenuate the individual coefficients of MDA and ROS in the full model [22, 23]. The excellent calibration and internal validation strengthen confidence in the model, yet external validation in independent cohorts remains essential to rule out overfitting and confirm generalizability.

The association between Fe²⁺ and SIMI was further accentuated in the sensitivity analysis restricted to critically ill patients with SOFA score > 5. In this subgroup, Fe²⁺ remained an independent predictor of SIMI with a substantially higher odds ratio compared to the full cohort, while other ferroptosis markers again showed no significant independent correlation. The model continued to demonstrate excellent discrimination and acceptable calibration upon internal validation. This dose-response-like pattern—stronger association in more severely ill patients—supports the biological plausibility of iron-mediated myocardial injury in sepsis. Several mechanisms may explain this amplified effect in the higher-severity subgroup. First, patients with higher SOFA scores are more likely to have multiple organ dysfunction, hemolysis, and systemic inflammation, all of which can release iron from damaged tissues and red blood cells, leading to elevated circulating Fe²⁺ levels [19]. Second, sepsis-associated mitochondrial dysfunction is more pronounced in critically ill patients, and iron accumulation within mitochondria can directly trigger oxidative damage and ferroptotic cell death [20]. Third, the antioxidant defense capacity (including GSH and GPX4) may be more severely depleted in the sickest patients, rendering them more vulnerable to iron-mediated lipid peroxidation. These findings suggest that Fe²⁺ may be particularly useful as a risk stratification biomarker in the most severely ill septic patients, and that iron-targeted interventions (e.g., iron chelation) could be most beneficial in this high-risk subgroup.

The diagnostic performance of ferroptosis-related markers for SIMI was moderate to good (Fe²⁺ showed the highest AUC, followed by ROS and MDA), outperforming BNP but remaining inferior to cTnI. For 28-day mortality, however, these markers showed only modest predictive value, substantially weaker than established severity scores (SOFA and APACHE II). These results underscore that while ferroptosis-associated pathways correlate with myocardial injury, their prognostic utility for survival is limited compared with composite clinical scores, possibly because mortality in sepsis is driven by multi-organ failure rather than isolated cardiac dysfunction [24].

Several limitations temper the interpretation of our findings. The single-center design and relatively modest sample size (n = 180) may reduce external validity, particularly across different ethnicities or sepsis phenotypes. A more critical limitation is the single time‑point measurement of ferroptosis-related biomarkers, which were assessed only once at 24 h after ICU admission. This single snapshot cannot capture the dynamic trajectories of these markers over the course of sepsis, nor can it establish temporal or causal relationships between changes in ferroptosis markers and the development or resolution of SIMI. Serial measurements would be required to determine whether rising Fe²⁺ levels precede myocardial injury (supporting causality) or merely reflect ongoing organ damage. Moreover, although we excluded patients with pre-existing cardiac disease, subclinical chronic conditions or unmeasured confounders (e.g., nutritional status affecting baseline GSH) cannot be entirely ruled out. Finally, the absence of more specific ferroptosis effectors (GPX4 activity, ACSL4 expression, or 4-HNE adducts) and lack of cardiac tissue validation preclude definitive mechanistic attribution.

Despite these limitations, our study provides the largest prospective clinical dataset to date linking oxidative stress and iron dysregulation markers with SIMI. The data support the hypothesis that ferroptosis-related pathways contribute to sepsis-associated cardiac dysfunction, with Fe²⁺ emerging as a key independent correlate—particularly among the most severely ill patients. Our findings emphasize the need for caution against over-interpretation of biomarker specificity but also highlight iron dysregulation as a potential therapeutic target. Future multicenter studies incorporating serial measurements, tissue-level validation, and targeted interventions (e.g., iron chelation or ferroptosis inhibitors such as ferrostatin-1 or liproxstatin-1) will be required to establish causality and therapeutic potential.

Clinical Implications: While these biomarkers do not serve as a specific diagnostic tool for ferroptosis itself, they help identify a high-risk phenotype of systemic oxidant stress in septic patients that may be responsive to broader antioxidant therapies, such as N-acetylcysteine or high-dose Vitamin C, which are not specific ferroptosis inhibitors. Future randomized trials should test these anti-oxidant strategies in patients with elevated ferroptosis-related markers and SIMI.

Conclusion

In conclusion, this prospective study demonstrates that ferroptosis-related biomarkers reflecting systemic oxidative stress and iron dysregulation are significantly associated with sepsis-induced myocardial injury and provide additive diagnostic information beyond traditional cardiac markers.However, these biomarkers should not be interpreted as direct evidence or specific diagnostic tools for ferroptosis as a regulated cell death mechanism, but rather as indicators of broader redox imbalance in severe sepsis.Our findings warrant further investigation into the precise role of ferroptosis pathways in human septic cardiomyopathy and support the rationale for future interventional trials targeting lipid peroxidation and iron homeostasis to improve cardiac outcomes in sepsis.

Abbreviations

SIMI

Sepsis-Induced Myocardial Injury

MI

Myocardial Injury

No-MI

No Myocardial Injury

cTnI

Cardiac Troponin I

LVEF

Left Ventricular Ejection Fraction

MDA

Malondialdehyde

LPO

Lipid Peroxidation

Fe²⁺

Ferrous Iron

GSH

Glutathione

ROS

Reactive Oxygen Species

PCT

Procalcitonin

IL-6

Interleukin-6

CRP

C-Reactive Protein

TNF-α

Tumor Necrosis Factor-Alpha

BNP

B-type Natriuretic Peptide

NT-proBNP

N-terminal pro-B-type Natriuretic Peptide

CK-MB

Creatine Kinase-MB

CVP

Central Venous Pressure

BP

Blood Pressure

MAP

Mean Arterial Pressure

PaO₂/FiO₂

Oxygenation Index

HR

Heart Rate

CBC

Complete Blood Count

SOFA

Sequential Organ Failure Assessment

APACHE II

Acute Physiology and Chronic Health Evaluation II

ICU

Intensive Care Unit

LPS

Lipopolysaccharide

ACSL4

Acyl-CoA Synthetase Long-Chain Family Member 4

GPX4

Glutathione Peroxidase 4

PTGS2

Prostaglandin-Endoperoxide Synthase 2

Author contributions

DC and YX contributed to the study conception and design. DC, HX, and PH contributed to data collection and analysis. DC drafted the manuscript. RW and YX critically revised the manuscript. All authors approved the final version.

Funding

This study was supported by the Clinical Research Innovation Plan of Shanghai General Hospital (CCTR-2025C09), Youth Program of the National Natural Science Foundation of China (82202423), the National Natural Science Foundation of China (Nos. 82402583), Noncommunicable Chronic Diseases-National Science and Technology Major Project (NO. 2023ZD0506502), the Key Supporting Discipline of Shanghai Healthcare System (NO. 2023ZDFC0102), Prevention and Control of Emerging and Major Infectious Diseases-National Science and Technology Major Project (2025ZD01904000), the Shanghai Jiao Tong University Cross-disciplinary Research Fund in Medicine and Engineering Key Project (2026), and the Fundamental Research Funds for the Central Universities (Grant No. YG2026ZD19).

Data availability

All data generated or analysed during this study are included in this published article.The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Ethical approval was obtained from the Shanghai General Hospital Institutional Review Board [Approval No. [2025]KY012]. All procedures performed in this study were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki Declaration and its later amendments. Written informed consent was obtained from all patients or their legal guardians prior to enrollment.

Consent for publication

The manuscript does not contain any individual person’s data in any form (including individual details, images, or videos). Therefore, consent for publication is not applicable.

Competing interests

The authors declare no competing interests.