Pro‐ and Anti‐Inflammatory Responses and Clinical Outcomes in Critically Ill Patients With Endotheliopathy: A Cohort Study

Caroline Humble; Martin Schønemann‐Lund; Peter Bruun‐Rasmussen; Morten H Bestle; Pär I Johansson; Lone M Poulsen; Ole Mathiesen

doi:10.1111/aas.70117

. 2025 Sep 8;69(9):e70117. doi: 10.1111/aas.70117

Pro‐ and Anti‐Inflammatory Responses and Clinical Outcomes in Critically Ill Patients With Endotheliopathy: A Cohort Study

Caroline Humble ^1,^2,^✉, Martin Schønemann‐Lund ³, Peter Bruun‐Rasmussen ², Morten H Bestle ^3,⁴, Pär I Johansson ^2,⁴, Lone M Poulsen ¹, Ole Mathiesen ^1,⁴

PMCID: PMC12417775 PMID: 40922409

ABSTRACT

Background

Multiple organ dysfunction syndrome (MODS) in critical illness involves dysregulated immune and inflammatory responses, endotheliopathy, and coagulation activation. We investigated how three types of endotheliopathy biomarkers relate to pro‐ and anti‐inflammatory responses and clinical outcomes in intensive care unit (ICU) patients.

Methods

In this secondary, explorative analysis of a prospective single‐centre cohort (n = 459), we assessed associations between endotheliopathy biomarkers (syndecan‐1, soluble thrombomodulin (sTM), platelet endothelial cell adhesion molecule‐1 (PECAM‐1)) and inflammatory biomarkers (pro‐inflammatory: IFN‐ϒ, IL‐1β, IL‐2, IL‐6, IL‐8, IL‐12p70, TNF‐α; anti‐inflammatory: IL‐4, IL‐10, IL‐13) at ICU admission using linear regression. Associations with 30‐day clinical outcomes were analysed using linear and Cox regression. All models were adjusted for age, sex, septic shock, pre‐ICU surgery and chronic disease.

Results

Higher levels of all three endotheliopathy biomarkers were associated with higher levels of inflammatory biomarkers. PECAM‐1, however, showed no significant association with IFN‐ϒ, IL‐1β and IL‐12p70. IL‐4 was excluded from linear regression due to > 50% imputed values. Higher levels of all three endotheliopathy biomarkers were significantly associated with increased mean and maximum modified Sequential Organ Failure Assessment (mSOFA) scores over 30 days, as well as with renal, hepatic, and coagulation failure, and 30‐day all‐cause mortality. Only sTM was significantly associated with cardiovascular failure; none were significantly associated with respiratory failure. Higher levels of sTM were associated with the highest levels of inflammatory biomarkers, the largest increases in mean and maximum mSOFA scores, and the highest hazard ratios for organ failure and 30‐day all‐cause mortality, compared with syndecan‐1 and PECAM‐1.

Conclusion

In this cohort of critically ill ICU patients, endotheliopathy was associated with (1) higher levels of pro‐ and anti‐inflammatory biomarkers at ICU admission and (2) MODS, single organ failure, and 30‐day all‐cause mortality. Among the three endotheliopathy biomarkers, sTM demonstrated the most consistent and strongest associations with both inflammatory biomarkers and clinical outcomes. These findings are exploratory and should be interpreted as hypothesis‐generating.

Editor's Comment

In this analysis of different biomarkers in a critically ill cohort, associations are demonstrated between markers related to endothelial stress, cytokines related to modulation of inflammation, and severity of illness scores.

1. Introduction

Critical illness, including sepsis and septic shock, is prevalent [1, 2] and burdens patients and society [2, 3]. While estimating the global incidence of critical illness is challenging [1], sepsis alone accounts for 48.9 million cases annually, contributing to about 20% of global deaths [4]. In intensive care units (ICUs), approximately 30% of patients have sepsis or septic shock, with a 30‐day mortality rate of around 35% [5]. For those with septic shock, in‐hospital mortality can exceed 50% [6].

Multiple organ dysfunction syndrome (MODS) significantly contributes to the morbidity and mortality of critically ill patients [7, 8]. The Sepsis‐3 definition highlights this by defining sepsis as life‐threatening organ dysfunction caused by a dysregulated host response to infection [9]. While our understanding of the complex mechanisms of MODS remains limited, it is acknowledged to involve dysregulated immune and inflammatory responses, along with coagulation activation [8, 10, 11]. However, the persistent lack of effective treatments targeting these dysregulated systems suggests that immune and inflammatory responses alone cannot fully explain the pathophysiology of MODS, further emphasising the complexity of this condition [12].

Shock‐induced endotheliopathy (SHINE) has been proposed as a unifying mechanism for MODS in critically ill patients with septic, traumatic, cardiogenic and neurogenic shock [13]. In SHINE, injurious insults, such as severe trauma or sepsis, cause increased sympathetic activation and catecholamine release (e.g., epinephrine, norepinephrine), which leads to dose‐dependent damage to the microvascular endothelial cell membranes (endotheliopathies) [14], ultimately contributing to MODS [13].

Key biomarkers in SHINE include syndecan‐1, soluble thrombomodulin (sTM), and platelet endothelial cell adhesion molecule‐1 (PECAM‐1). Syndecan‐1 indicates shedding of the glycocalyx [15], sTM reflects cleavage of thrombomodulin and perturbation of the anticoagulant protein C system [16, 17, 18, 19], and PECAM‐1 indicates disruption of endothelial intercellular junctions [20, 21]. Loss of glycocalyx and downregulation of the protein C system lead to a procoagulant microvasculature, promoting microvascular thrombosis that impairs oxygen delivery, resulting in hypoxia and MODS [10, 13]. Capillary leakage from glycocalyx loss and intercellular junctional disruption further impairs oxygen delivery and permits infiltration of inflammatory cells and pathogen‐ and damage‐associated molecular patterns (PAMPs, DAMPs), exacerbating MODS [13, 22]. PAMPs are microbial molecules, while DAMPs are released from damaged host cells. The innate immune system recognises PAMPs and DAMPs through pattern recognition receptors, initiating diverse immune and inflammatory responses [23]. Endotheliopathy is independently linked to MODS [24], delayed mechanical ventilation liberation (MV) [25], higher mortality rates during MV [25] and overall mortality [24, 25] in ICU patients.

Although endotheliopathy [24, 25, 26, 27, 28, 29, 30, 31, 32, 33] and systemic inflammation [34, 35, 36, 37, 38, 39, 40] have been extensively studied in critical illness, they have primarily been investigated in isolation. A limited number of studies [41, 42, 43, 44, 45] have examined their interplay, typically focusing on selected inflammatory biomarkers such as interleukin‐6 (IL‐6), IL‐8, tumour necrosis factor‐α (TNF‐α) and IL‐10. To our knowledge, no previous study has comprehensively assessed the association between distinct types of endotheliopathies, measured via syndecan‐1, sTM and PECAM‐1, and a broad panel of both pro‐ and anti‐inflammatory biomarkers in relation to clinical outcomes in a large ICU cohort. We believe that such an integrated approach may offer novel insights into the pathophysiology of MODS.

This study's primary objective was to investigate the association between three types of SHINE endotheliopathies, assessed by syndecan‐1, sTM, PECAM‐1 and 10 key pro‐ and anti‐inflammatory biomarkers at ICU admission. The secondary objective was to investigate the association between these three SHINE endotheliopathy biomarkers and the clinical outcomes of MODS, single organ failure and all‐cause mortality.

2. Methods

2.1. Study Design, Setting and Participants

This secondary, explorative analysis of the Metabolomics study [25], a prospective single‐centre cohort study, follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [46]. Other studies based on the Metabolomics study include [47, 48, 49]. The study included acutely admitted ICU patients (≥ 18 years) at the mixed ICU at Copenhagen University Hospital—North Zealand (November 2016–June 2019) with an expected stay of > 24 h. Exclusion criteria included inability to obtain informed consent or if the treating clinician deemed active treatment futile. A trial guardian approved participation before consent was obtained from relatives and/or patients. A total of 577 patients were enrolled. Blood samples were collected at baseline (time of inclusion) and daily up to day five. Baseline blood samples were analysed for syndecan‐1, sTM and PECAM‐1.

We included patients with (1) baseline Ethylene Diamine Tetra Acetic acid (EDTA) plasma measurements for syndecan‐1, sTM, and PECAM‐1; and (2) baseline EDTA plasma samples with one previous freeze–thaw cycle (FTC) for inflammatory biomarker analysis. Samples with fewer or more FTCs were excluded to enhance inter‐sample comparability of inflammatory biomarkers [50].

The Metabolomics study and the present study were approved by the Scientific Ethics Committee for the Capital Region of Denmark (H‐17027963 and H‐22041368). The Metabolomics study was approved by the Danish Data Protection Agency (I‐suite nos. 04,673 and 04,674), and the present study by the Centre for Data Registration for the Capital Region of Denmark (P‐2022‐483).

2.2. Measurements

EDTA blood was centrifuged at 3000 rounds per minute at 5°C for 10 min directly after collection. EDTA plasma was separated from the blood, aliquoted, and stored at −80°C [25, 47].

2.2.1. Endotheliopathy Biomarkers

Baseline syndecan‐1, sTM and PECAM‐1 were measured in uniplicate using enzyme‐linked immunosorbent assay (ELISA) (syndecan‐1 and sTM: Diaclone SAS, PECAM‐1: R&D System) undergoing one to three FTCs (syndecan‐1: one FTC, sTM: two FTCs, PECAM‐1: three FTCs) [25, 47].

2.2.2. Inflammatory Biomarkers

Baseline pro‐ (IFN‐ϒ, IL‐1β, IL‐2, IL‐6, IL‐8, IL‐12p70, TNF‐α) and anti‐inflammatory (IL‐4, IL‐10, IL‐13) biomarkers were measured in uniplicate using electrochemiluminescent sandwich immunoassays (V‐PLEX Proinflammatory Panel 1 Human Kit, Mesoscale Discovery, Rockville, MD, USA). EDTA plasma was twofold diluted and underwent one additional FTC, resulting in two FTCs in total (IL‐4: three FTCs due to kit error). Inter‐assay coefficient of variation (CVs) was < 15% for all inflammatory biomarkers; intra‐assay CVs were not calculated due to lack of internal controls [51].

The same inflammatory biomarkers were measured in 21 healthy controls as part of an internal validation using the same protocol (i.e., in uniplicate, twofold diluted, and two FTCs in total). Only sample storage duration, defined as time in days from blood sampling and freezing to analysis of the inflammatory biomarkers, was recorded for healthy controls (87.0 days), compared to a median of 1853 days (IQR 1560–2097, range 1367–2318) for ICU patients. All samples used the same kit LOT number (Additional File S1: Certificate of Analysis) and followed the manufacturer's instructions (Additional File S1: Supporting Information).

2.3. Statistical Analysis

A pre‐registered statistical analysis plan (SAP) [52] defined variables, outcomes, and statistical methods (Additional File S2). For all models, the explanatory variables of interest were the three endotheliopathy biomarkers, syndecan‐1, sTM and PECAM‐1, treated as three separate continuous explanatory variables. We present the results associated with an elevation in syndecan‐1, sTM and PECAM‐1 from the 25th to the 75th percentile (syndecan‐1: 21.0–114.6 ng/mL, sTM: 4.8–17.9 ng/mL, PECAM‐1: 11.2–14.5 ng/mL). All models were adjusted for the confounders identified using a directed acyclic graph (DAG) [53], unless otherwise stated. The adjustment set contained: sex, age, acute infection at ICU admission, septic shock at ICU admission, surgery before ICU admission and chronic disease (Additional File S2). Sepsis and septic shock were defined according to the third international consensus definitions for sepsis and septic shock [9], and the remaining variables were adjudicated by one author (MSL) based on information from the electronic health records (EHR) [25]. Chronic disease was defined as having ≥ 1 of the following: heart failure, diabetes, obesity, cirrhosis, chronic kidney disease or metastatic cancer.

2.3.1. Endotheliopathy and Inflammatory Responses

For this study's primary objective, the 10 inflammatory biomarkers served as the explorative outcomes. Linear regression was used for each inflammatory biomarker (log‐transformed). Interactions with sex, age and septic shock were tested. A post hoc sensitivity analysis included adjustments for systemic glucocorticoid at ICU admission and sample storage duration in addition to the confounders identified using a DAG [53].

2.3.2. Inflammatory Biomarkers

Post hoc comparisons were made between healthy controls and ICU patients (Wilcoxon rank‐sum test, linear regression adjusted for sample storage duration), and between ICU patient subgroups (without acute infection versus those with sepsis; sepsis versus septic shock). Linear regression, adjusted for age and sex, was used for the latter two subgroup comparisons, with sample storage age added in a sensitivity analysis. All models used log‐transformed inflammatory biomarkers.

2.3.3. Endotheliopathy and Clinical Outcomes

For this study's secondary objective, the explorative outcomes over 30 days included mean and maximum modified Sequential Organ Failure Assessment (mSOFA, excluding the neurological component [54, 55] with a possible range of 0–20) scores as a measure of MODS, time to a SOFA sub‐score ≥ 3 [56] or death (composite outcome), time to a SOFA sub‐score ≥ 3 [56] (single organ failure), and time to all‐cause mortality within 30 days. We added the maximum mSOFA score post hoc, defined as the maximum total mSOFA score experienced over 30 days from study inclusion. Linear regression was used for mean and maximum mSOFA scores; Cox regression for the time‐to‐event outcomes. Cause‐specific hazard ratios (HRs) were reported for single organ failure and the competing risk of death. Thirty‐day absolute risks were calculated from cause‐specific HRs [57]. Interactions with sex, age and septic shock were tested in the linear regression models for mean mSOFA and in the Cox regression models for all‐cause mortality.

Continuous variables are presented as medians with interquartile ranges (IQR), and categorical variables as counts and percentages (%). p‐values and 95% confidence intervals (CIs) were adjusted for multiple testing using the false discovery rate (FDR) [58, 59] with FDR‐adjusted two‐sided p‐values < 0.05 considered statistically significant. Missing data were handled using multivariate imputation by chained equations (MICE) [60], except for missing SOFA sub‐scores that were imputed using the last observation carried forward (LOCF) approach [61, 62]. For inflammatory biomarkers, we used fixed imputation corresponding to the lower limit of detection (LLOD) for values < LLOD, and to the upper limit of detection (ULOD) for values > ULOD, to preserve the relative ranking [63]. Inflammatory biomarkers with > 50% imputed values were excluded post hoc from statistical models to preserve analytical validity. All exclusions are detailed in the Results section. Analyses were performed in R version 4.3.0 [64].

3. Results

We included 459 of the 577 patients from the Metabolomics cohort (Figure 1).

Flowchart of the inclusion‐ and exclusion process. §Missing measurements of the inflammatory biomarkers, resulting from an analytical error in the laboratory. The eight patients are considered missing completely at random (MCAR) due to the nature of the error. §§The study cohort presented here has 80% overlap with the cohort (n = 459) presented in the papers by Schønemann‐Lund et al. [25, 47], but it is not identical.

Baseline characteristics are presented in Table 1. The median age was 71.0 years (IQR 62.0–79.0 years), 42.7% of the ICU patients were female, 48.1% had a history of chronic disease, 23.3% had undergone surgery, 75.4% had an acute infection, 55.3% had sepsis and 19.0% had septic shock. The median simplified acute physiology score (SAPS) 3 score was 63.0 (IQR 55.0–73.0) and the median SOFA score was 8.0 (IQR 5.0–10.0). The median time from hospital admission to ICU admission was 0.0 days (IQR 0.0–3.0), and the median time from ICU admission to study inclusion (i.e., baseline blood sampling) was 4.2 h (IQR 1.5–9.9). The median concentrations were 45.3 ng/mL (IQR 21.0–114.6 ng/mL) for syndecan‐1, 9.7 ng/mL (IQR 4.8–17.9 ng/mL) for sTM, and 12.9 ng/mL (IQR 11.2–14.5 ng/mL) for PECAM‐1.

TABLE 1.

ICU patient characteristics at baseline (n = 459).

Variables	Levels		Missing values (%)
Age, years	Median (IQR)	71.0 (62.0–79.0)	0 (0.0)
Sex, female	n (%)	196 (42.7)	0 (0.0)
History of stroke	n (%)	43 (9.4)	0 (0.0)
History of COPD	n (%)	144 (31.4)	0 (0.0)
History of hypertension	n (%)	235 (51.2)	0 (0.0)
History of myocardial infarction	n (%)	50 (10.9)	0 (0.0)
History of heart failure	n (%)	41 (8.9)	0 (0.0)
History of chronic kidney disease	n (%)	63 (13.7)	0 (0.0)
History of cirrhosis	n (%)	26 (5.7)	0 (0.0)
History of diabetes	n (%)	109 (23.7)	0 (0.0)
History of obesity	n (%)	56 (12.2)	269 (58.6)
History of metastatic cancer	n (%)	17 (3.7)	0 (0.0)
History of chronic disease ^a	n (%)	221 (48.1)	151 (32.9)
Use of systemic glucocorticoids	n (%)	70 (15.2)	126 (27.5)
Type of admission
Medical	n (%)	352 (76.7)	0 (0.0)
Surgical	n (%)	107 (23.3)	0 (0.0)
Acute infection	n (%)	346 (75.4)	0 (0.0)
Site or source of infection
Abdominal	n (%)	67 (14.6)	0 (0.0)
Neurological	n (%)	8 (1.7)	0 (0.0)
Other/unknown	n (%)	39 (8.5)	0 (0.0)
Pulmonary	n (%)	199 (43.4)	0 (0.0)
Skin/soft tissue	n (%)	16 (3.5)	0 (0.0)
Urinary	n (%)	17 (3.7)	0 (0.0)
Sepsis	n (%)	254 (55.3)	0 (0.0)
Septic shock	n (%)	87 (19.0)	0 (0.0)
Mechanical ventilation
NIV	n (%)	146 (31.8)	0 (0.0)
Invasive	n (%)	206 (44.9)	0 (0.0)
Vasopressors or inotropes ^b	n (%)	146 (31.8)	0 (0.0)
Renal replacement therapy	n (%)	14 (3.1)	83 (18.1)
KDIGO score ≥ 2	n (%)	115 (25.1)	0 (0.0)
SAPS 3	Median (IQR)	63.0 (55.0–73.0)	11 (2.4) ^c
SOFA score	Median (IQR)	8.0 (5.0–10.0)	78 (17.0) ^c
Syndecan‐1, ng/mL	Median (IQR)	45.3 (21.0–114.6)	0 (0.0)
sTM, ng/mL	Median (IQR)	9.7 (4.8–17.9)	0 (0.0)
PECAM‐1, ng/mL	Median (IQR)	12.9 (11.2–14.5)	0 (0.0)
C‐reactive protein, mg/L	Median (IQR)	63.8 (15.2–148.4)	29 (6.3)
White blood cells, ×10⁹/L	Median (IQR)	12.8 (9.2–17.7)	28 (6.1)
Time from hospital admission to ICU admission, days	Median (IQR)	0.0 (0.0–3.0)	0 (0.0)
Time from ICU admission to study inclusion, hours	Median (IQR)	4.2 (1.5–9.9)	0 (0.0)

Open in a new tab

Abbreviations: COPD, chronic obstructive pulmonary disease; ICU, intensive care unit; KDIGO, kidney disease improving global outcomes; NIV, non‐invasive ventilation; PECAM‐1, platelet endothelial cell adhesion molecule 1; SAPS, simplified acute physiology score; SOFA, sequential organ failure assessment; sTM: soluble thrombomodulin.

^{^a}

Defined as having ≥ 1 of the following: heart failure, diabetes, obesity, cirrhosis, chronic kidney disease or metastatic cancer.

^{^b}

Including the use of norepinephrine, epinephrine, phenylephrine, vasopressin analogues, dopamine, dobutamine, milrinone, levosimendan and ephedrine (n = 1 got ephedrine exclusively).

^{^c}

I.e. incomplete score.

3.1. Endotheliopathy and Inflammatory Responses

We found a significant association between higher levels of syndecan‐1, sTM and PECAM‐1 and higher levels of all inflammatory biomarkers, except for the associations between PECAM‐1 and IFN‐ϒ, IL‐1β and IL‐12p70 (Table 2, Figure S2). We excluded IL‐4 from the linear regression analysis due to a high percentage of imputed values (77.8%) (Table 3). Sensitivity analysis adjusting for systemic glucocorticoids and sample storage duration did not alter results (Table S1). Higher levels of sTM were associated with the highest levels of the nine inflammatory biomarkers, compared with syndecan‐1 and PECAM‐1. Age showed a weak but significant negative interaction with the associations of syndecan‐1 with IL‐6, IL‐8, and IL‐10 and of sTM with IL‐6 and IL‐8 (Table S2–S4).

TABLE 2.

Adjusted ^a estimates ^b for the associations between syndecan‐1, sTM, PECAM‐1 and 10 inflammatory biomarkers (n = 459).

	Syndecan‐1, 25th versus 75th percentile		sTM, 25th versus 75th percentile		PECAM‐1, 25th versus 75th percentile
	Estimate (95% CI)	p	Estimate (95% CI)	p	Estimate (95% CI)	p
IFN‐ϒ	1.30 (1.05–1.62)	0.02	1.79 (1.31–2.43)	< 0.001	1.10 (0.81–1.48)	0.55
IL‐1β	1.30 (1.09–1.54)	< 0.01	2.16 (1.71–2.73)	< 0.001	1.18 (0.97–1.44)	0.11
IL‐2	1.30 (1.06–1.58)	0.01	2.90 (2.23–3.75)	< 0.001	1.41 (1.15–1.74)	< 0.001
IL‐4 ^c	NA	NA	NA	NA	NA	NA
IL‐6	1.74 (1.40–2.15)	< 0.001	4.17 (3.09–5.61)	< 0.001	1.53 (1.20–1.95)	< 0.001
IL‐8	2.09 (1.77–2.48)	< 0.001	3.45 (2.69–4.43)	< 0.001	1.71 (1.41–2.07)	< 0.001
IL‐10	1.64 (1.33–2.03)	< 0.001	2.78 (2.05–3.77)	< 0.001	1.62 (1.29–2.05)	< 0.001
IL‐12p70	1.38 (1.15–1.66)	< 0.001	2.31 (1.79–2.98)	< 0.001	1.18 (0.94–1.47)	0.16
IL‐13	1.63 (1.38–1.93)	< 0.001	2.50 (1.95–3.21)	< 0.001	1.24 (1.01–1.53)	0.04
TNF‐α	1.56 (1.38–1.77)	< 0.001	2.74 (2.31–3.27)	< 0.001	1.37 (1.19–1.59)	< 0.001

Open in a new tab

Note: Syndecan‐1, sTM and PECAM‐1 were analysed as continuous variables, and the inflammatory biomarkers were log‐transformed in the linear regression models. p‐values and 95% CIs were adjusted for multiple testing according to the FDR approach.

Abbreviations: CI, confidence interval; FDR, false discovery rate; ICU, intensive care unit; IFN, interferon; IL, interleukin; PECAM‐1, platelet endothelial cell adhesion molecule 1; sTM, soluble thrombomodulin; TNF, tumour necrosis factor.

^{^a}

All models were adjusted for age, sex, acute infection, septic shock, surgery before ICU admission and chronic disease.

^{^b}

The estimates represent the fold change in the inflammatory biomarkers when syndecan‐1, sTM, and PECAM‐1 change from the 25th to the 75th percentile (syndecan‐1: 21.0–114.6 ng/mL, sTM: 4.8–17.9 ng/mL, PECAM‐1: 11.2–14.5 ng/mL).

^{^c}

Linear regression modelling was not performed for IL‐4 due to the high percentage of imputed values (77.8%).

TABLE 3.

Differences in inflammatory biomarkers ^a between healthy controls and ICU patients.

Healthy controls (n = 21)			ICU patients (n = 459)
	Median (IQR) ^b	Imputed values (%) ^c	Median (IQR) ^b	Imputed values (%) ^c	p‐value (healthy controls versus ICU patients) ^d
IFN‐ϒ	5.34 (3.58–6.85)	0.0	7.67 (2.75–23.36)	12.2	0.43
IL‐1β ^e	0.01 (0.01–0.01)	90.5	0.47 (0.25–0.97)	26.1	NA
IL‐2	0.19 (0.14–0.36)	9.5	0.90 (0.37–2.07)	8.7	< 0.001
IL‐4 ^e	0.01 (0.01–0.02)	33.3	0.01 (0.01–0.02)	77.8	NA
IL‐6	0.61 (0.48–1.03)	0.0	32.91 (10.46–201.91)	14.8	< 0.001
IL‐8	2.91 (2.17–3.41)	0.0	23.75 (9.91–81.89)	8.7	< 0.001
IL‐10	0.22 (0.17–0.31)	0.0	5.21 (1.55–20.29)	2.6	< 0.001
IL‐12p70	0.14 (0.09–0.19)	28.6	0.45 (0.30–1.05)	22.7	< 0.001
IL‐13	0.59 (0.38–0.97)	28.6	7.36 (3.49–19.64)	6.7	< 0.001
TNF‐α	1.41 (1.29–1.45)	0.0	3.62 (1.96–7.61)	0.2	< 0.001

Open in a new tab

Abbreviations: CI, confidence interval; FDR, false discovery rate; ICU, intensive care unit; IFN, interferon; IL, interleukin; IQR, interquartile range; LLOD, lower limit of detection; NA, not applicable; TNF, tumour necrosis factor; ULOD, upper limit of detection.

^{^a}

The unit is pg/mL for all inflammatory biomarkers.

^{^b}

Medians (IQR) were calculated based on the complete data sets (healthy controls, n = 21) and (ICU patients, n = 459), i.e., including imputed values.

^{^c}

Values below the plate‐specific LLOD and above the plate‐specific ULOD were imputed with a fixed value corresponding to the plate‐specific LLOD and ULOD, respectively.

^{^d}

Wilcoxon rank‐sum test. p‐values were adjusted for multiple testing according to the FDR approach.

^{^e}

Wilcoxon rank‐sum tests were not performed due to the high percentage of imputed values for IL‐1β (90.5% in healthy controls) and IL‐4 (77.8% in ICU patients).

3.2. Inflammatory Biomarkers

3.2.1. Healthy Controls Versus ICU Patients

Table 3 lists median levels and percentages of imputed values for the inflammatory biomarkers in healthy controls (n = 21) and ICU patients (n = 459). Details of ranges, LLODs and ULODs are provided in Table S5. ICU patients had significantly higher levels of IL‐2, IL‐6, IL‐8, IL‐10, IL‐12p70, IL‐13 and TNF‐α compared to healthy controls. Sensitivity analysis adjusting for sample storage duration confirmed these findings, with IFN‐γ also being significantly higher in ICU patients (Table S6). Due to high percentages (> 50%) of imputed values, Wilcoxon rank‐sum tests and linear regression modeling were not conducted for IL‐1β and IL‐4.

3.2.2. ICU Patient Subgroups: Without Acute Infection Versus Sepsis, and Sepsis Versus Septic Shock

Table 4 lists median levels of inflammatory biomarkers for patients without acute infection (n = 113), sepsis (n = 254) and septic shock (n = 87). Septic ICU patients had significantly higher age‐ and sex‐adjusted levels of IFN‐ϒ, IL‐1β, IL‐2, IL‐8 and IL‐13 than those without acute infection. After adjusting for sample storage duration in the sensitivity analysis, IL‐6, IL‐10 and TNF‐α were also significantly higher (Table S7). ICU patients with septic shock had significantly higher levels of all inflammatory biomarkers, except for IFN‐ϒ, compared with septic ICU patients. The sensitivity analysis did not alter this result (Table S7). A linear regression model for IL‐4 was not fitted due to a high percentage of imputed values (77.8%).

TABLE 4.

Age‐ and sex‐adjusted estimates ^a for the difference in inflammatory biomarkers ^b between ICU patients without acute infection and sepsis and between ICU patients with sepsis and septic shock ^c .

	Without acute infection (n = 113)	Sepsis (n = 254)	Septic shock (n = 87)	Without acute infection versus sepsis		Sepsis versus septic shock
	Median (IQR) ^d			Estimate (95% CI)	p	Estimate (95% CI)	p
IFN‐ϒ	5.44 (2.52–12.63)	8.10 (2.67–29.87)	8.41 (3.66–38.66)	1.91 (1.34–2.71)	< 0.001	1.28 (0.77–2.12)	0.35
IL‐1β	0.29 (0.25–0.55)	0.47 (0.25–0.85)	1.20 (0.52–3.83)	1.42 (1.01–2.00)	0.04	3.38 (2.42–4.72)	< 0.001
IL‐2	0.63 (0.24–1.19)	0.93 (0.39–1.86)	1.94 (0.76–4.86)	1.62 (1.11–2.36)	0.01	2.12 (1.43–3.14)	< 0.001
IL‐4 ^e	0.01 (0.007–0.01)	0.01 (0.007–0.02)	0.01 (0.007–0.02)	NA	NA	NA	NA
IL‐6	18.97 (8.26–73.19)	24.67 (9.32–142.24)	348.90 (63.83–729.00)	1.52 (0.94–2.43)	0.09	5.11 (3.25–8.03)	< 0.001
IL‐8	12.78 (6.95–35.22)	22.47 (9.81–57.42)	114.55 (25.33–683.00)	1.66 (1.14–2.41)	< 0.01	3.90 (2.69–5.67)	< 0.001
IL‐10	2.98 (1.01–11.41)	4.36 (1.55–15.13)	20.94 (6.51–93.40)	1.50 (0.95–2.35)	0.08	3.79 (2.47–5.83)	< 0.001
IL‐12p70	0.45 (0.30–0.64)	0.45 (0.30–0.87)	0.99 (0.44–8.00)	1.33 (0.90–1.97)	0.15	3.03 (2.11–4.36)	< 0.001
IL‐13	5.82 (2.66–9.29)	6.77 (3.51–17.67)	17.39 (6.25–102.97)	1.61 (1.13–2.29)	< 0.01	2.64 (1.85–3.77)	< 0.001
TNF‐α	2.80 (1.66–4.79)	3.29 (1.89–6.69)	10.37 (3.74–19.95)	1.29 (0.97–1.71)	0.08	2.67 (2.04–3.49)	< 0.001

Open in a new tab

Note: The inflammatory biomarkers were log‐transformed in the linear regression models. p‐values and 95% CIs were adjusted for multiple testing according to the FDR approach.

Abbreviations: CI, confidence interval; FDR, false discovery rate; ICU, intensive care unit; IFN, interferon; IL, interleukin; IQR, interquartile range; NA, not applicable; TNF, Tumour necrosis factor.

^{^a}

The estimates represent the fold change in median levels of inflammatory biomarkers between ICU patients without acute infection and those with sepsis, as well as between ICU patients with sepsis and those with septic shock.

^{^b}

The unit is pg/mL for all inflammatory biomarkers.

^{^c}

n = 5 patients with acute infection, but without sepsis or septic shock, were not included in the analysis.

^{^d}

Medians (IQR) are calculated based on the complete data set (ICU patients, n = 459), i.e., including imputed values.

^{^e}

Linear regression modeling was not performed for IL‐4 due to the high percentage of imputed values (77.8%).

3.3. Endotheliopathy and Clinical Outcomes

3.3.1. Mean Modified SOFA Score

The median mean mSOFA score over 30 days was 5.1 (IQR 3.8–7.3, range 0.0–17.1).

Higher levels of syndecan‐1 (1.46 units increase in mean mSOFA score from the 25th to 75th percentile [95% CI 1.34–1.58, p < 0.001]), sTM (3.06 units increase in mean mSOFA score from the 25th to 75th percentile [95% CI 2.60–3.60, p < 0.001]) and PECAM‐1 (1.14 units increase in mean mSOFA score from the 25th to 75th percentile [95% CI 1.10–1.18, p < 0.001]) were associated with an increase in mean mSOFA score. Age showed a weak but significant negative interaction on the association between syndecan‐1, sTM and mean mSOFA score. In contrast, septic shock significantly strengthened the associations of both syndecan‐1 and PECAM‐1 with mean mSOFA score (Table S8).

3.3.2. Maximum Modified SOFA Score

The median maximum mSOFA score over 30 days was 7.0 (IQR 5.0–9.0, range 0.0–18.0).

Higher levels of syndecan‐1 (1.60 units increase in maximum mSOFA score from the 25th to 75th percentile [95% CI 1.43–1.78, p < 0.001]), sTM (3.61 units increase in maximum mSOFA score from the 25th to 75th percentile [95% CI 3.02–4.32, p < 0.001]) and PECAM‐1 (1.32 units increase in maximum mSOFA score from the 25th to 75th percentile [95% CI 1.21–1.44, p < 0.001]) were associated with an increase in maximum mSOFA score.

3.3.3. Single Organ Failure or Death

Table 5 shows that higher levels of syndecan‐1, sTM, and PECAM‐1 were associated with a significantly increased hazard of renal failure or death, hepatic failure or death and coagulation failure or death. None of syndecan‐1, sTM, or PECAM‐1 was associated with a significantly increased hazard of respiratory failure or death, and only higher levels of sTM were associated with a significantly increased hazard of cardiovascular failure or death. Soluble thrombomodulin exhibited the highest HRs for the composite outcomes, compared with syndecan‐1 and PECAM‐1.

TABLE 5.

Adjusted ^a associations ^b of syndecan‐1, sTM and PECAM‐1 with the composite outcome of single organ failure or death, whichever came first (n = 459).

Outcome	Syndecan‐1, 25th versus 75th percentile		sTM, 25th versus 75th percentile		PECAM‐1, 25th versus 75th percentile
Outcome	HR (95% CI)	p	HR (95% CI)	p	HR (95% CI)	p
Respiratory failure or death, n_{total events} = 327, n_{respiratory failure} = 309, n_death = 18	1.11 (0.93–1.32)	0.26	1.15 (0.87–1.53)	0.33	1.21 (1.00–1.47)	0.05
Cardiovascular failure or death, n_{total events} = 320, n_{cardiovascular failure} = 307, n_death = 13	1.15 (0.98–1.36)	0.09	1.77 (1.40–2.25)	< 0.001	1.10 (0.88–1.36)	0.42
Renal failure or death, n_{total events} = 174, n_{renal failure} = 129, n_death = 45	1.58 (1.30–1.91)	< 0.001	3.40 (2.40–4.84)	< 0.001	1.62 (1.27–2.06)	< 0.001
Hepatic failure or death, n_{total events} = 103, n_{hepatic failure} = 16, n_death = 87	1.86 (1.43–2.43)	< 0.001	2.44 (1.52–3.90)	< 0.001	2.08 (1.45–2.98)	< 0.001
Coagulation failure or death, n_{total events} = 112, n_{coagulation failure} = 29, n_death = 83	1.62 (1.27–2.09)	< 0.001	2.22 (1.43–3.46)	< 0.001	1.65 (1.18–2.30)	< 0.01

Open in a new tab

Note: Syndecan‐1, sTM and PECAM‐1 were analysed as continuous variables in the Cox regression models. p‐values and 95% CIs were adjusted for multiple testing using the FDR approach.

Abbreviations: CI, confidence interval; FDR, false discovery rate; HR, hazard ratio; ICU, intensive care unit; PECAM‐1, platelet endothelial cell adhesion molecule 1; sTM, soluble thrombomodulin.

^{^a}

All models were adjusted for age, sex, acute infection, septic shock, surgery before ICU admission and chronic disease.

^{^b}

The HRs represent the fold change in the hazard of experiencing the composite outcome of single organ failure or death when syndecan‐1, sTM and PECAM‐1 change from the 25th to the 75th percentile (syndecan‐1: 21.0–114.6 ng/mL, sTM: 4.8–17.9 ng/mL, PECAM‐1: 11.2–14.5 ng/mL).

3.3.4. Single Organ Failure

Table 6 presents the cause‐specific HRs for single organ failures and the competing risk of death. Higher levels of syndecan‐1, sTM and PECAM‐1 were associated with a significantly increased hazard of renal failure, hepatic failure and coagulation failure. None of syndecan‐1, sTM, or PECAM‐1 was associated with a significantly increased hazard of respiratory failure, and only higher levels of sTM were associated with a significantly increased hazard of cardiovascular failure. Soluble thrombomodulin exhibited the highest cause‐specific HRs for the single organ failures, compared with syndecan‐1 and PECAM‐1. Thirty‐day absolute risks calculated from the cause‐specific HRs are presented in Table S9.

TABLE 6.

Adjusted ^a associations ^b of syndecan‐1, sTM and PECAM‐1 with single organ failure and the competing event of death before single organ failure (n = 459).

Outcome	Syndecan‐1, 25th versus 75th percentile		sTM, 25th versus 75th percentile		PECAM‐1, 25th versus 75th percentile
Outcome	HR (95% CI)	p	HR (95% CI)	p	HR (95% CI)	p
Respiratory failure, n_events = 309	1.07 (0.87–1.32)	0.53	1.10 (0.80–1.50)	0.58	1.14 (0.92–1.40)	0.23
Death before respiratory failure, n_events = 18	2.04 (0.91–4.59)	0.09	4.03 (0.84–19.36)	0.08	9.76 (1.77–53.90)	< 0.01
Cardiovascular failure, n_events = 307	1.16 (0.98–1.36)	0.09	1.78 (1.40–2.27)	< 0.001	1.10 (0.88–1.37)	0.41
Death before cardiovascular failure ^c , n_events = 13	NA	NA	NA	NA	NA	NA
Renal failure, n_events = 129	1.61 (1.29–2.02)	< 0.001	4.30 (2.81–6.58)	< 0.001	1.62 (1.22–2.16)	< 0.01
Death before renal failure, n_events = 45	1.49 (0.94–2.35)	0.09	1.67 (0.73–3.83)	0.23	1.61 (0.96–2.72)	0.07
Hepatic failure, n_events = 16	7.67 (2.72–21.59)	< 0.01	44.19 (4.07–479.33)	0.02	8.33 (2.40–28.94)	0.01
Death before hepatic failure, n_events = 87	1.51 (1.15–1.97)	< 0.01	1.73 (1.10–2.72)	0.04	1.74 (1.24–2.44)	< 0.01
Coagulation failure, n_events = 29	2.03 (1.19–3.48)	< 0.01	5.45 (1.71–17.34)	< 0.01	1.65 (0.80–3.42)	0.18
Death before coagulation failure, n_events = 83	1.51 (1.10–2.06)	< 0.01	1.63 (0.94–2.83)	0.08	1.64 (1.11–2.44)	0.01

Open in a new tab

Note: Syndecan‐1, sTM and PECAM‐1 were analysed as continuous variables in the competing risks regression models. p‐values and 95% CIs were adjusted for multiple testing using the FDR approach.

Abbreviations: CI, confidence interval; FDR, false discovery rate; HR, hazard ratio; ICU, intensive care unit; NA, not applicable; PECAM‐1, platelet endothelial cell adhesion molecule 1; sTM, soluble thrombomodulin.

^{^a}

All models were adjusted for age, sex, acute infection, septic shock, surgery before ICU admission, and chronic disease.

^{^b}

The HRs represent the fold change in the hazard of experiencing a single organ failure or the competing event of death before single organ failure, when syndecan‐1, sTM, and PECAM‐1 change from the 25th to the 75th percentile (syndecan‐1: 21.0–114.6 ng/mL, sTM: 4.8–17.9 ng/mL, PECAM‐1: 11.2–14.5 ng/mL).

^{^c}

There were too few events (n = 13) of death before cardiovascular failure to be able to calculate HRs.

3.3.5. All‐Cause Mortality

At 30 days, follow‐up for all‐cause mortality was complete, and a total of 150 patients had died. Higher levels of syndecan‐1 (HR 1.60 from the 25th to 75th percentile [95% CI 1.29–1.99, p < 0.001]), sTM (HR 2.10 from the 25th to 75th percentile [95% CI 1.47–3.00, p < 0.001]) and PECAM‐1 (HR 1.69 from the 25th to 75th percentile [95% CI 1.28–2.24, p < 0.001]) were associated with higher 30‐day all‐cause mortality. Septic shock significantly strengthened the association between syndecan‐1 and 30‐day all‐cause mortality (Table S10).

4. Discussion

In this cohort of critically ill ICU patients, higher levels of the endotheliopathy biomarkers syndecan‐1, sTM and PECAM‐1 were significantly associated with both pro‐ and anti‐inflammatory biomarkers at ICU admission. These findings highlight a strong link between endothelial dysfunction and systemic inflammation during the acute phase of critical illness. Higher biomarker levels were also associated with MODS, single organ failure, and 30‐day all‐cause mortality. Among the three, sTM showed the most consistent and strongest associations with inflammatory biomarkers and clinical outcomes, suggesting its potential as a robust indicator of disease severity and a key mediator in the pathophysiology of critical illness.

Our results align with prior studies linking endotheliopathy to systemic inflammation and adverse outcomes in both septic and non‐infectious critical illnesses. All three biomarkers were associated with MODS and mortality, consistent with findings in septic [24] and acute respiratory distress syndrome (ARDS) cohorts [33]. We also observed higher levels of endotheliopathy and pro‐inflammatory biomarkers, including IL‐6, IL‐8 and TNF‐α, in ICU patients with adverse outcomes, consistent with previous studies [41, 42, 43, 44, 45]. These observations further add to the growing body of evidence linking systemic inflammation to poor clinical outcomes [35, 36, 39, 40, 65, 66].

We observed elevated IL‐10 and IL‐13 levels in patients with higher endotheliopathy biomarker levels, contributing to the limited literature on anti‐inflammatory responses in critical illness [40]. This finding aligns with a trauma cohort study reporting an association between elevated syndecan‐1 and IL‐10 [41]. Post hoc analyses showed a graded increase in inflammatory biomarkers with disease severity, consistent with previous studies [34, 35, 36, 39, 40]. In contrast, for IL‐1β and TNF‐α specifically, a meta‐analysis found no significant differences across sepsis severities [40]. For IFN‐γ, we observed higher but statistically non‐significant levels in healthy controls compared with ICU patients and between the septic and septic shock subgroups. This contrasts with four single‐plex ELISA‐based studies [38, 39] and two multiplex‐based studies [37, 67], all of which reported significantly higher levels of IFN‐γ in septic patients compared with healthy controls. One multiplex‐based study reported a significant difference between severe sepsis and septic shock (n = 14 vs. n = 46) [36]; however, the authors did not adjust for multiple testing or specify how values outside the detection range were handled, which may limit the interpretability of their findings. These discrepancies may stem from methodological differences, particularly in assay platform and data handling. The MSD V‐plex multiplex platform, while advantageous for high‐throughput cytokine profiling, may differ in sensitivity, calibration and dynamic range compared to single‐plex ELISA and other multiplex platforms [51, 68]. Biological variability, including timing of sample collection, patient heterogeneity, and treatment effects, may also influence IFN‐γ levels. Our findings underscore the importance of assay‐specific interpretation and transparent reporting.

None of the endotheliopathy biomarkers showed statistically significant associations with respiratory failure, though trends suggested increased risk. This aligns with mixed findings from prior studies. Johansen et al. found no significant associations between syndecan‐1 or sTM and respiratory failure [24], while Sapru et al. [33] reported associations between sTM and the number of ventilator‐free days in ARDS patients. Similarly, Schønemann‐Lund et al. [25] showed significant associations between sTM, PECAM‐1 and delayed liberation from mechanical ventilation. Additionally, a small, underpowered RCT in ventilated COVID‐19 patients suggested a potential benefit from prostacyclin infusion, an endothelium‐protective therapy [69]. Our study's lack of strong associations may reflect limited power or the broad definition of respiratory failure, which may include extrapulmonary causes. An ongoing RCT on prostacyclin infusion in ventilated ICU patients with suspected pulmonary infection and severe endotheliopathy may provide further insights [70].

We found a significant association between endotheliopathy, particularly sTM, and renal failure, consistent with studies on refractory [31] and sepsis‐associated acute kidney injury (AKI) [32]. Although different diagnostic criteria [71] limit direct comparisons, the link between sTM and renal dysfunction aligns with the multifactorial pathophysiology of sepsis‐associated AKI, inclusive of microcirculatory dysfunction and inflammation [72]. Soluble thrombomodulin also showed the strongest associations with cardiovascular and hepatic failure, consistent with previous findings [24]. Our findings align with the established pathophysiology of cardiovascular failure in critical illness, involving both macro‐ and microvascular dysfunction. Hepatic failure incidence was 3.8%, consistent with previously reported rates in ICU populations [73, 74]. However, the small number of events contributes to the wide confidence intervals, necessitating cautious interpretation. Similarly, while sTM was associated with coagulation failure, the small number of events again limits interpretability. This association is biologically plausible, as loss of thrombomodulin promotes a procoagulant microvascular environment with increased platelet consumption.

Although syndecan‐1 and PECAM‐1 exhibited weaker associations with inflammation, MODS, and single organ failure, all three biomarkers were similarly associated with 30‐day all‐cause mortality.

This pattern may be attributable to their distinct roles in endothelial pathophysiology. Soluble thrombomodulin, released during severe endothelial injury, has been associated with coagulopathy and inflammation; syndecan‐1 indicates early glycocalyx degradation; and PECAM‐1 plays a role in leukocyte transmigration and endothelial integrity. Syndecan‐1 and PECAM‐1 may contribute to mortality through mechanisms not fully captured by conventional organ dysfunction scores. Additionally, baseline levels may underestimate endotheliopathy, as dynamic PECAM‐1 changes in respiratory failure have shown added prognostic value [47].

The cleavage of thrombomodulin into its soluble form impairs the protein C anticoagulant pathway, contributing to a procoagulant endothelial phenotype. Inflammatory responses further promote coagulation activation in the microvasculature [8]. These mechanisms may explain the stronger associations observed between sTM and both inflammatory biomarkers and clinical outcomes. Thrombomodulin also has anti‐inflammatory effects, as shown in both in vivo and in vitro studies [16, 17]. Our findings suggest these properties may be relevant in critically ill patients. The association between sTM and anti‐inflammatory cytokines may reflect a compensatory host response to severe inflammation or an unrecognised pro‐inflammatory role of thrombomodulin. Future mechanistic studies are needed to explore this further.

Strengths of our study include a pre‐registered statistical analysis plan, transparent inflammatory biomarker reporting, and standardised sample handling. Our study has several limitations. First, the observational design precludes causal inference, and findings should be interpreted as hypothesis‐generating. Second, inflammatory biomarkers were measured in uniplicate without internal controls, and prolonged storage [75] of many of the ICU patient samples may have further impacted the internal validity. Imputation was required for values outside detection limits, and some biomarkers (IL‐4, IL‐1β) were excluded from quantitative analysis due to high imputation rates. Collectively, these limitations underscore that the inflammatory biomarker levels should be interpreted as exploratory rather than confirmatory. To strengthen the generalisability of our findings, external validation in independent cohorts using the same multiplex assay and comparable patient populations is warranted. Furthermore, the observed discrepancies in IFN‐γ levels compared to ELISA‐based studies highlight the need for cross‐platform validation. Third, comparisons between healthy controls and ICU patients should be interpreted with caution due to differences in analysis timing, storage duration, markedly imbalanced sample sizes, and lack of demographic matching. Fourth, sampling within 24 h of ICU admission may not consistently capture patients' baseline endotheliopathy and inflammatory status, given variability in illness progression and treatment at ICU entry. Although the short median interval from hospital admission to ICU transfer (0.0 days) and from ICU admission to sampling (4.2 h) suggests inclusion at initial stages of their critical illness, generalisability remains limited. While we adjusted for septic shock (i.e., vasopressor use) and systemic glucocorticoid use at ICU admission, lack of pre‐ICU treatment data introduces potential residual confounding. Fifth, only baseline measurements were available, preventing assessment of biomarker trajectories. Lastly, while syndecan‐1, sTM and PECAM‐1 are widely used as markers of endotheliopathy, they may also be expressed by other cell types, potentially limiting specificity [21, 76, 77].

In summary, our findings suggest that endotheliopathy biomarkers may have prognostic relevance in critical illness and could contribute to future efforts in risk stratification. Although our study does not assess prognostic performance or validate risk models, the observed associations highlight the need for further investigation. Future studies should explore the temporal dynamics of these biomarkers and inflammatory responses, their prognostic value for long‐term outcomes, and their responsiveness to interventions aimed at endothelial protection or repair. Mechanistic studies are also needed to clarify the causal pathways linking endotheliopathy with inflammatory responses, organ failure and mortality in critical illness.

5. Conclusion

Author Contributions

P.I.J. concepted the idea of the Metabolomics cohort, wrote the original protocol and oversaw the analyses of the endotheliopathy biomarkers. M.S.‐L. acquired clinical data. C.H., M.S.‐L., P.B.‐R., M.H.B., P.I.J., L.M.P. and O.M. concepted the idea of the present study. C.H. performed the statistical analyses and drafted the first version of the manuscript. All authors revised the manuscript for important intellectual content.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: Sensitivity analysis—Adjusted^a estimates^b for the associations between syndecan‐1, sTM, PECAM‐1 and 10 inflammatory biomarkers (n = 459)

Table S2: Interaction effect estimates^a of age, sex, and septic shock on the adjusted^b associations between syndecan‐1 and 10 inflammatory biomarkers (n = 459)

Table S3: Interaction effect estimates^a of age, sex, and septic shock on the adjusted^b associations between sTM and 10 inflammatory biomarkers (n = 459)

Table S4: Interaction effect estimates^a of age, sex, and septic shock on the adjusted^b associations between PECAM‐1 and 10 inflammatory biomarkers (n = 459)

Table S5: Details of inflammatory biomarker^a measurements for healthy controls and ICU patients

Table S6: Sensitivity analysis—Sample storage duration‐adjusted estimates^a for the difference in inflammatory biomarkers^b between healthy controls and ICU patients

Table S7: Sensitivity analysis—Age‐, sex‐ and sample storage duration‐adjusted estimates^a for the difference in inflammatory biomarkers^b between ICU patients without acute infection and sepsis and between ICU patients with sepsis and septic shock^c

Table S8: Interaction effect estimates^a for age, sex, and septic shock on the adjusted^b associations between syndecan‐1, sTM, PECAM‐1 and the mean modified SOFA score.

Table S9: Adjusted 30‐day absolute risk^a of single organ failure and the competing risk of death for syndecan‐1, sTM and PECAM‐1.

Table S10: Interaction effect estimates^a for age, sex, and septic shock on the adjusted^b association between syndecan‐1, sTM, PECAM‐1 and 30‐day all‐cause mortality.

AAS-69-0-s002.docx^{(265.8KB, docx)}

Data S1: aas70117‐sup‐0002‐Supinfo.pdf.

AAS-69-0-s003.pdf^{(1.3MB, pdf)}

Figure S1A: Scatterplots of the association between syndecan‐1, sTM, PECAM‐1 and log‐transformed IFN‐γ, IL‐1β and IL‐2 levels at ICU admission. Each point represents an individual observation. Vertical lines represent the 25th and 75th percentiles of syndecan‐1, sTM and PECAM‐1. A linear regression line is overlaid to illustrate the trend. ICU, intensive care unit; IFN, interferon; IL, interleukin; PECAM‐1, platelet endothelial cell adhesion molecule 1; sTM, soluble thrombomodulin.

Figure S1B: Scatterplots of the association between syndecan‐1, sTM, PECAM‐1 and log‐transformed IL‐6, IL‐8 and IL‐10 levels at ICU admission. Each point represents an individual observation. Vertical lines represent the 25th and 75th percentiles of syndecan‐1, sTM and PECAM‐1. A linear regression line is overlaid to illustrate the trend. ICU, intensive care unit; IL, interleukin; PECAM‐1, platelet endothelial cell adhesion molecule 1; sTM, soluble thrombomodulin.

Figure S1C: Scatterplots of the association between syndecan‐1, sTM, PECAM‐1 and log‐transformed IL‐12p70, IL‐13 and TNF‐α levels at ICU admission. Each point represents an individual observation. Vertical lines represent the 25th and 75th percentiles of syndecan‐1, sTM and PECAM‐1. A linear regression line is overlaid to illustrate the trend. ICU, intensive care unit; IL, interleukin; PECAM‐1, platelet endothelial cell adhesion molecule 1; sTM, soluble thrombomodulin; TNF, tumour necrosis factor.

AAS-69-0-s001.docx^{(2.2MB, docx)}

Acknowledgements

The authors wish to thank study nurses Lone Valbjørn and Sanne Lauritzen, research assistants Saif Al‐Haidar and Frederik H. Bestle, biomedical laboratory scientist Doris Schnülle Nelleman, as well as clinical trial coordinator Kristine H. Pedersen for their invaluable contribution to the study. The study received grants from Greater Copenhagen Health Science Partners, Region Zealand Health Science Research Fund, Clinical Academic Group (CAG) Center for Endotheliomics, The Danish Society of Anaesthesiology and Intensive Care Medicine (DASAIM), Intensive Care Symposium Hindsgavl, A. P. Moller Foundation and King Christian X Fund.

Humble C., Schønemann‐Lund M., Bruun‐Rasmussen P., et al., “Pro‐ and Anti‐Inflammatory Responses and Clinical Outcomes in Critically Ill Patients With Endotheliopathy: A Cohort Study,” Acta Anaesthesiologica Scandinavica 69, no. 9 (2025): e70117, 10.1111/aas.70117.

Funding: This work was supported by the Greater Copenhagen Health Science Partners, the Region Zealand Health Science Research Fund, Clinical Academic Group (CAG) Center for Endotheliomics, the Intensive Care Symposium Hindsgavl, the A. P. Moller Foundation, the King Christian X Fund and The Danish Society of Anaesthesiology and Intensive Care Medicine (DASAIM).

Data Availability Statement

The data that support the findings of this study are available from the authors upon reasonable request.

References

1. Adhikari N. K., Fowler R. A., Bhagwanjee S., and Rubenfeld G. D., “Critical Care and the Global Burden of Critical Illness in Adults,” Lancet 376, no. 9749 (2010): 1339–1346. [DOI] [PMC free article] [PubMed] [Google Scholar]
2. Paoli C. J., Reynolds M. A., Sinha M., Gitlin M., and Crouser E., “Epidemiology and Costs of Sepsis in the United States—An Analysis Based on Timing of Diagnosis and Severity Level,” Critical Care Medicine 46, no. 12 (2018): 1889–1897. [DOI] [PMC free article] [PubMed] [Google Scholar]
3. Tiru B., DiNino E. K., Orenstein A., et al., “The Economic and Humanistic Burden of Severe Sepsis,” PharmacoEconomics 33, no. 9 (2015): 925–937. [DOI] [PubMed] [Google Scholar]
4. Rudd K. E., Johnson S. C., Agesa K. M., et al., “Global, Regional, and National Sepsis Incidence and Mortality, 1990–2017: Analysis for the Global Burden of Disease Study,” Lancet 395, no. 10219 (2020): 200–211. [DOI] [PMC free article] [PubMed] [Google Scholar]
5. Sakr Y., Jaschinski U., Wittebole X., et al., “Sepsis in Intensive Care Unit Patients: Worldwide Data From the Intensive Care Over Nations Audit,” Open Forum Infectious Diseases 5, no. 12 (2018): ofy313. [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Vincent J. L., Jones G., David S., Olariu E., and Cadwell K. K., “Frequency and Mortality of Septic Shock in Europe and North America: A Systematic Review and Meta‐Analysis,” Critical Care 23, no. 1 (2019): 196. [DOI] [PMC free article] [PubMed] [Google Scholar]
7. Pool R., Gomez H., and Kellum J. A., “Mechanisms of Organ Dysfunction in Sepsis,” Critical Care Clinics 34, no. 1 (2018): 63–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
8. Marshall J. C., “Inflammation, Coagulopathy, and the Pathogenesis of Multiple Organ Dysfunction Syndrome,” Critical Care Medicine 29 (2001): S99–S106. [DOI] [PubMed] [Google Scholar]
9. Singer M., Deutschman C. S., Seymour C. W., et al., “The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis‐3),” Journal of the American Medical Association 315, no. 8 (2016): 801–810. [DOI] [PMC free article] [PubMed] [Google Scholar]
10. Aird W. C., “The Role of the Endothelium in Severe Sepsis and Multiple Organ Dysfunction Syndrome,” Blood 101, no. 10 (2003): 3765–3777. [DOI] [PubMed] [Google Scholar]
11. Fajgenbaum D. C. and June C. H., “Cytokine Storm,” New England Journal of Medicine 383, no. 23 (2020): 2255–2273. [DOI] [PMC free article] [PubMed] [Google Scholar]
12. Opal S. M., Dellinger R. P., Vincent J. L., Masur H., and Angus D. C., “The Next Generation of Sepsis Clinical Trial Designs: What Is Next After the Demise of Recombinant Human Activated Protein C?*,” Critical Care Medicine 42, no. 7 (2014): 1714–1721. [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Johansson P. I., Stensballe J., and Ostrowski S. R., “Shock Induced Endotheliopathy (SHINE) in Acute Critical Illness ‐ a Unifying Pathophysiologic Mechanism,” Critical Care 21, no. 1 (2017): 25. [DOI] [PMC free article] [PubMed] [Google Scholar]
14. García de Lomana A. L., Vilhjálmsson A. I., McGarrity S., et al., “Metabolic Response in Endothelial Cells to Catecholamine Stimulation Associated With Increased Vascular Permeability,” International Journal of Molecular Sciences 23, no. 6 (2022): 3162. [DOI] [PMC free article] [PubMed] [Google Scholar]
15. Jedlicka J., Becker B. F., and Chappell D., “Endothelial Glycocalyx,” Critical Care Clinics 36, no. 2 (2020): 217–232. [DOI] [PubMed] [Google Scholar]
16. Okamoto T., Tanigami H., Suzuki K., and Shimaoka M., “Thrombomodulin: A Bifunctional Modulator of Inflammation and Coagulation in Sepsis,” Critical Care Research and Practice 2012 (2012): 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
17. Morser J., “Thrombomodulin Links Coagulation to Inflammation and Immunity,” Current Drug Targets 13, no. 3 (2012): 421–431. [DOI] [PubMed] [Google Scholar]
18. Boehme M. W. J., Galle P., and Stremmel W., “Kinetics of Thrombomodulin Release and Endothelial Cell Injury by Neutrophil‐Derived Proteases and Oxygen Radicals,” Immunology 107, no. 3 (2002): 340–349. [DOI] [PMC free article] [PubMed] [Google Scholar]
19. Ishii H. and Majerus P. W., “Thrombomodulin Is Present in Human Plasma and Urine,” Journal of Clinical Investigation 76, no. 6 (1985): 2178–2181. [DOI] [PMC free article] [PubMed] [Google Scholar]
20. Lertkiatmongkol P., Liao D., Mei H., Hu Y., and Newman P. J., “Endothelial Functions of Platelet/Endothelial Cell Adhesion Molecule‐1 (CD31),” Current Opinion in Hematology 23, no. 3 (2016): 253–259. [DOI] [PMC free article] [PubMed] [Google Scholar]
21. Privratsky J. R. and Newman P. J., “PECAM‐1: Regulator of Endothelial Junctional Integrity,” Cell and Tissue Research 355, no. 3 (2014): 607–619. [DOI] [PMC free article] [PubMed] [Google Scholar]
22. Vincent J. L., Ince C., and Pickkers P., “Endothelial Dysfunction: A Therapeutic Target in Bacterial Sepsis?,” Expert Opinion on Therapeutic Targets 25, no. 9 (2021): 733–748. [DOI] [PubMed] [Google Scholar]
23. Abbas A., Lichtman A., and Pillai S., “Innate Immunity: The Early Defense Against Infections,” in Basic Immunology: Functions and Disorders of the Immune System, 6th ed., ed. Abbas A., Lichtman A., and Pillai S. (Elsevier, 2020), 24–29. [Google Scholar]
24. Johansen M. E., Johansson P. I., Ostrowski S. R., et al., “Profound Endothelial Damage Predicts Impending Organ Failure and Death in Sepsis,” Seminars in Thrombosis and Hemostasis 41, no. 1 (2015): 16–25. [DOI] [PubMed] [Google Scholar]
25. Schønemann‐Lund M., Itenov T. S., Larsson J. E., Lindegaard B., Johansson P. I., and Bestle M. H., “Endotheliopathy Is Associated With Slower Liberation From Mechanical Ventilation: A Cohort Study,” Critical Care 26, no. 1 (2022): 33. [DOI] [PMC free article] [PubMed] [Google Scholar]
26. Johansson P. I., Henriksen H. H., Stensballe J., et al., “Traumatic Endotheliopathy: A Prospective Observational Study of 424 Severely Injured Patients,” Annals of Surgery 265, no. 3 (2017): 597–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
27. Johansson P. I., Haase N., Perner A., and Ostrowski S. R., “Association Between Sympathoadrenal Activation, Fibrinolysis, and Endothelial Damage in Septic Patients: A Prospective Study,” Journal of Critical Care 29, no. 3 (2014): 327–333. [DOI] [PubMed] [Google Scholar]
28. Ostrowski S. R., Haase N., Müller R. B., et al., “Association Between Biomarkers of Endothelial Injury and Hypocoagulability in Patients With Severe Sepsis: A Prospective Study,” Critical Care 19, no. 1 (2015): 191. [DOI] [PMC free article] [PubMed] [Google Scholar]
29. Ostrowski S. R., Pedersen S. H., Jensen J. S., Mogelvang R., and Johansson P. I., “Acute Myocardial Infarction Is Associated With Endothelial Glycocalyx and Cell Damage and a Parallel Increase in Circulating Catecholamines,” Critical Care 17, no. 1 (2013): R32. [DOI] [PMC free article] [PubMed] [Google Scholar]
30. Johansson P. I., Bro‐Jeppesen J., Kjaergaard J., Wanscher M., Hassager C., and Ostrowski S. R., “Sympathoadrenal Activation and Endothelial Damage Are Inter Correlated and Predict Increased Mortality in Patients Resuscitated After Out‐Of‐Hospital Cardiac Arrest. A Post Hoc Sub‐Study of Patients From the TTM‐Trial,” PLoS One 10, no. 3 (2015): e0120914. [DOI] [PMC free article] [PubMed] [Google Scholar]
31. Itenov T. S., Jensen J. U., Ostrowski S. R., et al., “Endothelial Damage Signals Refractory Acute Kidney Injury in Critically Ill Patients,” Shock 47, no. 6 (2017): 696–701. [DOI] [PubMed] [Google Scholar]
32. Katayama S., Nunomiya S., Koyama K., et al., “Markers of Acute Kidney Injury in Patients With Sepsis: The Role of Soluble Thrombomodulin,” Critical Care 21 (2017): 229. [DOI] [PMC free article] [PubMed] [Google Scholar]
33. Sapru A., Calfee C. S., Liu K. D., et al., “Plasma Soluble Thrombomodulin Levels Are Associated With Mortality in the Acute Respiratory Distress Syndrome,” Intensive Care Medicine 41, no. 3 (2015): 470–478. [DOI] [PMC free article] [PubMed] [Google Scholar]
34. Damas P., Canivet J. L., De Groote D., et al., “Sepsis and Serum Cytokine Concentrations,” Critical Care Medicine 25, no. 3 (1997): 405–412. [DOI] [PubMed] [Google Scholar]
35. Gogos C. A., Drosou E., Bassaris H. P., and Skoutelis A., “Pro‐ Versus Anti‐Inflammatory Cytokine Profile in Patients With Severe Sepsis: A Marker for Prognosis and Future Therapeutic Options,” Journal of Infectious Diseases 181, no. 1 (2000): 176–180. [DOI] [PubMed] [Google Scholar]
36. Bozza F. A., Salluh J. I., Japiassu A. M., et al., “Cytokine Profiles as Markers of Disease Severity in Sepsis: A Multiplex Analysis,” Critical Care 11, no. 2 (2007): R49. [DOI] [PMC free article] [PubMed] [Google Scholar]
37. Jekarl D. W., Kim J. Y., Ha J. H., et al., “Diagnosis and Prognosis of Sepsis Based on Use of Cytokines, Chemokines, and Growth Factors,” Disease Markers 2019 (2019): 1089107–1089111. [DOI] [PMC free article] [PubMed] [Google Scholar]
38. Kumar S., Gupta E., Kaushik S., Kumar Srivastava V., Mehta S., and Jyoti A., “Evaluation of Oxidative Stress and Antioxidant Status: Correlation With the Severity of Sepsis,” Scandinavian Journal of Immunology 87, no. 4 (2018): e12653. [DOI] [PubMed] [Google Scholar]
39. Kumar S., Gupta E., Srivastava V. K., et al., “Nitrosative Stress and Cytokines Are Linked With the Severity of Sepsis and Organ Dysfunction,” British Journal of Biomedical Science 76, no. 1 (2019): 29–34. [DOI] [PubMed] [Google Scholar]
40. Gharamti A. A., Samara O., Monzon A., et al., “Proinflammatory Cytokines Levels in Sepsis and Healthy Volunteers, and Tumor Necrosis Factor‐Alpha Associated Sepsis Mortality: A Systematic Review and Meta‐Analysis,” Cytokine 158 (2022): 156006. [DOI] [PubMed] [Google Scholar]
41. Johansson P. I., Stensballe J., Rasmussen L. S., and Ostrowski S. R., “A High Admission Syndecan‐1 Level, a Marker of Endothelial Glycocalyx Degradation, Is Associated With Inflammation, Protein C Depletion, Fibrinolysis, and Increased Mortality in Trauma Patients,” Annals of Surgery 254, no. 2 (2011): 194–200. [DOI] [PubMed] [Google Scholar]
42. Calfee C. S., Janz D. R., Bernard G. R., et al., “Distinct Molecular Phenotypes of Direct vs Indirect ARDS in Single‐Center and Multicenter Studies,” Chest 147, no. 6 (2015): 1539–1548. [DOI] [PMC free article] [PubMed] [Google Scholar]
43. Mikacenic C., Hahn W. O., Price B. L., et al., “Biomarkers of Endothelial Activation Are Associated With Poor Outcome in Critical Illness,” PLoS One 10, no. 10 (2015): e0141251. [DOI] [PMC free article] [PubMed] [Google Scholar]
44. Robinson‐Cohen C., Katz R., Price B. L., et al., “Association of Markers of Endothelial Dysregulation Ang1 and Ang2 With Acute Kidney Injury in Critically Ill Patients,” Critical Care 20, no. 1 (2016): 207. [DOI] [PMC free article] [PubMed] [Google Scholar]
45. Zhang D., Li L., Chen Y., et al., “Syndecan‐1, an Indicator of Endothelial Glycocalyx Degradation, Predicts Outcome of Patients Admitted to an ICU With COVID‐19,” Molecular Medicine 27, no. 1 (2021): 151. [DOI] [PMC free article] [PubMed] [Google Scholar]
46. von Elm E., Altman D. G., Egger M., Pocock S. J., Gøtzsche P. C., and Vandenbroucke J. P., “The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement: Guidelines for Reporting Observational Studies,” Lancet 370, no. 9596 (2007): 1453–1457. [DOI] [PubMed] [Google Scholar]
47. Schønemann‐Lund M., Itenov T. S., Larsson J. E., Lindegaard B., Johansson P. I., and Bestle M. H., “Novel Subgroups in Acute Respiratory Failure Based on the Trajectories of Three Endotheliopathy Biomarkers: A Cohort Study,” Acta Anaesthesiologica Scandinavica 67, no. 7 (2023): 896–908. [DOI] [PubMed] [Google Scholar]
48. Johansson P. I., Henriksen H. H., Karvelsson S. T., et al., “LASSO Regression Shows Histidine and Sphingosine 1 Phosphate Are Linked to Both Sepsis Mortality and Endothelial Damage,” European Journal of Medical Research 29 (2024): 71. [DOI] [PMC free article] [PubMed] [Google Scholar]
49. Mortensen K. M., Itenov T. S., Stensballe J., et al., “Changes in Nitric Oxide Inhibitors and Mortality in Critically Ill Patients: A Cohort Study,” Annals of Intensive Care 14, no. 1 (2024): 133. [DOI] [PMC free article] [PubMed] [Google Scholar]
50. Huang W. Y., Kemp T. J., Pfeiffer R. M., Pinto L. A., Hildesheim A., and Purdue M. P., “Impact of Freeze‐Thaw Cycles on Circulating Inflammation Marker Measurements,” Cytokine 95 (2017): 113–117. [DOI] [PMC free article] [PubMed] [Google Scholar]
51. Leng S. X., McElhaney J. E., Walston J. D., Xie D., Fedarko N. S., and Kuchel G. A., “ELISA and Multiplex Technologies for Cytokine Measurement in Inflammation and Aging Research,” Journals of Gerontology Series A: Biological Sciences and Medical Sciences 63, no. 8 (2008): 879–884. [DOI] [PMC free article] [PubMed] [Google Scholar]
52. Humble C., Schønemann‐Lund M., Bruun‐Rasmussen P., et al., “Pro‐ and Anti‐Inflammatory Response in Critically Ill Patients With Endothelial Damage (ENDO‐FLAME),” OSF (2023), 10.17605/OSF.IO/CEWR9. [DOI] [Google Scholar]
53. Textor J., van der Zander B., Gilthorpe M. S., Liskiewicz M., and Ellison G. T., “Robust Causal Inference Using Directed Acyclic Graphs: The R Package “Dagitty”,” International Journal of Epidemiology 45, no. 6 (2016): 1887–1894. [DOI] [PubMed] [Google Scholar]
54. Gordon A. C., Santhakumaran S., Al‐Beidh F., et al., “Levosimendan for the Prevention of Acute Organ Dysfunction in Sepsis,” New England Journal of Medicine 375, no. 17 (2016): 1638–1648. [DOI] [PubMed] [Google Scholar]
55. Bestle M. H., Clausen N. E., Søe‐Jensen P., et al., “Efficacy and Safety of Iloprost in Patients With Septic Shock‐Induced Endotheliopathy‐Protocol for the Multicenter Randomized, Placebo‐Controlled, Blinded, Investigator‐Initiated Trial,” Acta Anaesthesiologica Scandinavica 64, no. 5 (2020): 705–711. [DOI] [PMC free article] [PubMed] [Google Scholar]
56. Vincent J. L., Moreno R., Takala J., et al., “The SOFA (Sepsis‐Related Organ Failure Assessment) Score to Describe Organ Dysfunction/Failure. On Behalf of the Working Group on Sepsis‐Related Problems of the European Society of Intensive Care Medicine,” Intensive Care Medicine 22, no. 7 (1996): 707–710. [DOI] [PubMed] [Google Scholar]
57. Ozenne B., Sørensen A., Scheike T., Torp‐Pedersen C., and Gerds T., “riskRegression: Predicting the Risk of an Event Using Cox Regression Models,” R Journal 9 (2017): 440–460. [Google Scholar]
58. Benjamini Y. and Hochberg Y., “Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing,” Journal of the Royal Statistical Society. Series B, Statistical Methodology 57 (1995): 289–300. [Google Scholar]
59. Altman D. G. and Bland J. M., “How to Obtain the P Value From a Confidence Interval,” BMJ 343 (2011): d2304. [DOI] [PubMed] [Google Scholar]
60. van Buuren S. and Groothuis‐Oudshoorn K., “Mice: Multivariate Imputation by Chained Equations in R,” Journal of Statistical Software 45 (2011): 1–67. [Google Scholar]
61. Lambden S., Laterre P. F., Levy M. M., and Francois B., “The SOFA Score‐Development, Utility and Challenges of Accurate Assessment in Clinical Trials,” Critical Care 23, no. 1 (2019): 374. [DOI] [PMC free article] [PubMed] [Google Scholar]
62. Vasilevskis E. E., Pandharipande P. P., Graves A. J., et al., “Validity of a Modified Sequential Organ Failure Assessment Score Using the Richmond Agitation‐Sedation Scale,” Critical Care Medicine 44, no. 1 (2016): 138–146. [DOI] [PMC free article] [PubMed] [Google Scholar]
63. Herbers J., Miller R., Walther A., et al., “How to Deal With Non‐Detectable and Outlying Values in Biomarker Research: Best Practices and Recommendations for Univariate Imputation Approaches,” Comprehensive Psychoneuroendocrinology 7 (2021): 100052. [DOI] [PMC free article] [PubMed] [Google Scholar]
64. R Core Team , R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2023), https://www.R‐project.org/. [Google Scholar]
65. Loisa P., Rinne T., Laine S., Hurme M., and Kaukinen S., “Anti‐Inflammatory Cytokine Response and the Development of Multiple Organ Failure in Severe Sepsis,” Acta Anaesthesiologica Scandinavica 47, no. 3 (2003): 319–325. [DOI] [PubMed] [Google Scholar]
66. Del Valle D. M., Kim‐Schulze S., Huang H. H., et al., “An Inflammatory Cytokine Signature Predicts COVID‐19 Severity and Survival,” Nature Medicine 26, no. 10 (2020): 1636–1643. [DOI] [PMC free article] [PubMed] [Google Scholar]
67. Khurana S., Kumari M., Bhardwaj N., Kumar S., Malhotra R., and Mathur P., “T‐Helper‐17, Regulatory T‐Helper Cells Related Serum Markers and IL‐13 in the Outcome of Polytraumatic Patients With Bacteremia,” Iranian Journal of Immunology 15, no. 4 (2018): 302–308. [DOI] [PubMed] [Google Scholar]
68. Platchek M., Lu Q., Tran H., and Xie W., “Comparative Analysis of Multiple Immunoassays for Cytokine Profiling in Drug Discovery,” SLAS Discovery 25, no. 10 (2020): 1197–1213. [DOI] [PubMed] [Google Scholar]
69. Johansson P. I., Søe‐Jensen P., Bestle M. H., et al., “Prostacyclin in Intubated Patients With COVID‐19 and Severe Endotheliopathy: A Multicenter, Randomized Clinical Trial,” American Journal of Respiratory and Critical Care Medicine 205, no. 3 (2022): 324–329. [DOI] [PMC free article] [PubMed] [Google Scholar]
70. Søe‐Jensen P., Clausen N. E., Bestle M. H., et al., “Efficacy and Safety of a 72‐h Infusion of Prostacyclin (1 Ng/Kg/Min) in Mechanically Ventilated Patients With Pulmonary Infection and Endotheliopathy—Protocol for the Multicenter Randomized, Placebo‐Controlled, Blinded, Investigator‐Initiated COMBAT‐ARF Trial,” Acta Anaesthesiologica Scandinavica 69, no. 1 (2025): e14565. [DOI] [PMC free article] [PubMed] [Google Scholar]
71. Kellum J., Lameire N., Aspelin P., et al., “KDIGO Clinical Practice Guideline for Acute Kidney Injury,” Kidney International. Supplement 2, no. 1 (2012): 1–138. [Google Scholar]
72. Zarbock A., Nadim M. K., Pickkers P., et al., “Sepsis‐Associated Acute Kidney Injury: Consensus Report of the 28th Acute Disease Quality Initiative Workgroup,” Nature Reviews. Nephrology 19, no. 6 (2023): 401–417. [DOI] [PubMed] [Google Scholar]
73. Kramer L., Jordan B., Druml W., Bauer P., and Metnitz P. G. H., “Incidence and Prognosis of Early Hepatic Dysfunction in Critically Ill Patients—A Prospective Multicenter Study,” Critical Care Medicine 35, no. 4 (2007): 1099–1104. [DOI] [PubMed] [Google Scholar]
74. Pierrakos C., Velissaris D., Felleiter P., et al., “Increased Mortality in Critically Ill Patients With Mild or Moderate Hyperbilirubinemia,” Journal of Critical Care 40 (2017): 31–35. [DOI] [PubMed] [Google Scholar]
75. de Jager W., Bourcier K., Rijkers G. T., Prakken B. J., and Seyfert‐Margolis V., “Prerequisites for Cytokine Measurements in Clinical Trials With Multiplex Immunoassays,” BMC Immunology 10 (2009): 52. [DOI] [PMC free article] [PubMed] [Google Scholar]
76. Hahn R. G., Patel V., and Dull R. O., “Human Glycocalyx Shedding: Systematic Review and Critical Appraisal,” Acta Anaesthesiologica Scandinavica 65, no. 5 (2021): 590–606. [DOI] [PubMed] [Google Scholar]
77. Li Y. H., Kuo C. H., Shi G. Y., and Wu H. L., “The Role of Thrombomodulin Lectin‐Like Domain in Inflammation,” Journal of Biomedical Science 19, no. 1 (2012): 34. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials