HDL-C Levels Are Associated with Clinical Outcomes and Hospital Mortality in Critically Ill Patients with Sepsis: A Prospective Cohort Study

1. Introduction

Critical illness encompasses life-threatening conditions that require prompt intensive monitoring, support, and treatment of vital organ functions. Sepsis and acute respiratory distress syndrome (ARDS) represent two of the most common causes of critical illness among patients admitted to the intensive care unit (ICU) and are major contributors to multiple organ failure, with persistently high morbidity and mortality.

ARDS is a life-threatening form of acute hypoxemic respiratory failure characterized by oxidative stress, dysregulated airway and systemic inflammatory cascades and immune activation [1,2,3,4]. To date, no effective pharmacotherapies have been established, and lung-protective mechanical ventilation remains the cornerstone of evidence-based management for ARDS, with demonstrated benefits in improving clinical outcomes [5]. Despite significant advances in ARDS management, morbidity and mortality remain unacceptably high. Early identification of potential risk factors and prognostic determinants during the course of critical illness is therefore essential and may offer opportunities to reduce the burden of ARDS and improve patient outcomes.

Lipid synthesis, transport, and metabolism are frequently disrupted during the acute phase of critical illness, thereby affecting immune function, hormone and vitamin production, as well as cell membrane integrity and signaling pathway. High-density lipoprotein (HDL) is a complex lipoprotein particle that exerts antioxidant, anti-inflammatory, anti-apoptotic, antithrombotic, and immunomodulatory properties [6,7]. Hypocholesterolemia commonly occurs during critical illness and has been associated with disease severity and adverse clinical outcomes. However, the underlying pathophysiological mechanisms are multifactorial, complex, and not yet fully elucidated.

Systemic inflammation and oxidative stress, which are hallmarks of critical illness with sepsis and key contributors to ARDS pathogenesis, may induce structural and compositional modifications of HDL particles, resulting in dysfunctional HDL with impaired antioxidant and anti-inflammatory properties. Such alterations may reduce the protective functions of HDL and contribute to endothelial dysfunction, dysregulated immune responses, and organ injury. Although high-density lipoprotein cholesterol (HDL-C) is commonly used as a quantitative measure of HDL, it may also reflect broader alterations in HDL metabolism and functionality during critical illness.

The relationship between HDL-C levels and clinical outcomes in critically ill sepsis patients, particularly among those with and without ARDS, remains incompletely understood. Therefore, this study aimed to investigate the association between conventional lipid parameters, especially HDL-C, and clinical outcomes in critically ill patients with sepsis.

2. Materials and Methods

2.1. Study Design and Patients.

This prospective observational study was conducted between October 2020 and July 2025 in the ICU of a tertiary referral center, Chang Gung Memorial Hospital (CGMH), Linkou branch, Taiwan. Critically ill patients admitted to the ICU with the infection source of pneumonia or extra-pulmonary sepsis underwent screening. The inclusion criteria of ARDS was defined according to the Berlin criteria [8]. Patients who met the Berlin criteria at ICU admission were classified into the ARDS group, whereas those who did not meet the criteria were classified into the non-ARDS group. Participants were further categorized according to survival status (survivors vs. non-survivors) and HDL-C levels (high vs. low HDL-C groups based on the optimal ROC-derived cutoff value).

Eligible patients were prospectively enrolled after informed consent had been obtained from the patients or their legally authorized representative. The exclusion criteria were: (1) age < 20 years; (2) a moribund condition with refractory multiple organ failure and an expected survival of less than 24 hours; and (3) inability to obtain informed consent. Due to operational and logistical limitations, not all subjects at ICU admissions were screened for eligibility. Enrollment was dependent on research staffing and operational feasibility, resulting in only a subset of potentially eligible ICU patients underwent active screening and enrollment assessment during the study period. This study was conducted in accordance with the Declaration of Helsinki, and ethical approval was obtained from the Institutional Review Board of CGMH for Human Research (CGMH IRB No. 202000760A3, 202100595A3, 202201833A3, and 202300897A3).

2.2. Data Collection

Demographic characteristics and underlying comorbidities were recorded for all participants. Blood samples were collected within 12 hours of ICU admission in 6 mL plastic tubes containing K₂EDTA as an anticoagulant (BD, Franklin Lakes, NJ, USA), and analyses were performed in the institutional biochemistry laboratory. Clinical and laboratory variables were obtained, including interleukin-6 (IL-6); conventional plasma lipid parameters, including HDL-C, low-density lipoprotein cholesterol (LDL-C), total cholesterol, and triglycerides; recent statin use within the preceding six months; Acute Physiology and Chronic Health Evaluation II (APACHE II) score; and Sequential Organ Failure Assessment (SOFA) score. The dates of hospital and ICU admission, mechanical ventilator initiation and liberation, ICU and hospital discharge, and time of death were recorded. In addition, the presence of shock, use of vasopressor, and requirement for renal replacement therapy during the ICU stay were documented.

2.3. Outcome Measurements

Hospital mortality was defined as death from any cause during the index hospitalization. Survival status at 90 days after hospital discharge was determined through review of electronic medical records and hospital follow-up data. Complete follow-up information was available for all included patients. Mortality outcomes included 28-day, 60-day, and 90-day mortality, as well as all-cause hospital mortality. Additional clinical outcomes included the incidence of shock, use of vasopressor, development of acute kidney injury, requirement for renal replacement therapy, duration of mechanical ventilation, length of stay in the ICU, and total hospital length of stay.

2.4. Statistical Analysis

Comparisons between groups were performed using analysis of variance or the Student’s t test for normally distributed variables, and the Kruskal–Wallis test or Mann–Whitney U test for non-normally distributed variables, as appropriate. All continuous variables were reported as mean (standard deviation) for normally distributed variables or median (interquartile range) for non-normally distributed variables. Categorical variables were reported as frequency (proportion) and were compared using the chi-square test for equal proportions or Fisher’s exact test. Receiver operating characteristic (ROC) curve analysis was performed to evaluate the ability of HDL-C measured at ICU admission to discriminate between critically ill patients with and without ARDS. The optimal HDL-C cutoff value was determined using the Youden index (sensitivity + specificity − 1). The ROC-derived cutoff was subsequently used to dichotomize HDL-C levels in outcome analyses. Univariate analysis was examined to identify the risk factors associated with hospital mortality in the first step, followed by the construction of multivariate logistic regression models. To control for potential covariances, all candidate variables were selected based on clinically driven selection and included in the multivariable model using the Enter method. The results were expressed using odds ratio (OR) and 95% confidence interval (CI). Kaplan–Meier method was employed to estimate the cumulative incidence of hospital survival for each group and the difference in survival curve was assessed by log-rank test. All statistical analyses were performed using SPSS version 29.0 (IBM Inc., Armonk, NY), and a two-sided p value less than 0.05 was considered statistically significant.

3. Results

A total of 6,238 critically ill patients with sepsis admitted to the ICU were screened during the study period. Of these, 62 patients without ARDS and 223 patients with ARDS at ICU admission were included in the final analysis (Figure 1). The overall all-cause in-hospital mortality among the study population was 37.9%.

3.1. Comparisons Between Critically Ill Sepsis Patients with and Without ARDS

As shown in Table 1, there were no significant differences in age or gender distribution between patients with and without ARDS. Body mass index (BMI) was significantly higher in the ARDS group than in the non-ARDS group. The proportion of pneumonia was significantly higher in the ARDS group. Underlying comorbidities were generally comparable between the two groups.

Both APACHE II and SOFA scores were significantly higher in the ARDS group than in the non-ARDS group (both p < 0.05). Inflammatory markers including C-reactive protein (CRP) and IL-6 were also significantly elevated in patients with ARDS (all p < 0.05).

Among conventional lipid parameters, levels of HDL-C, LDL-C, and total cholesterol were significantly lower in the ARDS group compared with the non-ARDS group (all p < 0.05), whereas triglyceride levels were significantly higher in the ARDS group (p = 0.008). Statin use was observed in 7.6% of patients with ARDS, whereas no patients without ARDS had received statin therapy. Patients in the ARDS group had significantly higher rates of shock and use of vasopressor. All-cause in-hospital mortality was significantly higher in the ARDS group than in the non-ARDS group (43.0% vs. 19.4%, p = 0.001).

3.2. Comparisons Between Critically Ill Survivors and Non-Survivors

As shown in Table 2, no significant differences in age or gender distribution were observed between survivors and non-survivors. BMI was significantly higher among survivors. The proportion of immunocompromised status was more prevalent among non-survivors (p < 0.05).

Both APACHE II and SOFA scores were significantly higher in non-survivors than in survivors (both p < 0.05). Inflammatory markers including CRP and IL-6 were elevated in non-survivors. Non-survivors also exhibited more severe hypoxemia, as reflected by lower PaO₂/FiO₂ ratios (p = 0.010).

Regarding conventional lipid parameters, HDL-C and total cholesterol levels were significantly lower in non-survivors than in survivors (all p < 0.05). The values of lower LDL-C and higher triglyceride was detected in non-survivors; however, these differences did not reach statistical significance. Statin use did not differ significantly between the two groups. Non-survivors had significantly higher rates of shock and use of vasopressor. The proportions of mechanical ventilation use and ARDS occurrence were both significantly higher among non-survivors.

3.3. Comparisons Between Critically Ill Sepsis Patients with High and Low Plasma HDL-C Levels

Participants were stratified into a high HDL-C group (n = 123, 43.2%) and a low HDL-C group (n = 162, 56.8%) based on the optimal cutoff value of 22.5 mg/dL at ICU admission (Table 3).

There were no significant differences between the two groups in terms of age or BMI. APACHE II and SOFA scores were significantly higher in the low HDL-C group than in the high HDL-C group (both p < 0.05). The low HDL-C group also exhibited significantly higher levels of inflammatory markers, including CRP and IL-6, and more severe hypoxemia, as reflected by lower PaO₂/FiO₂ ratios (all p < 0.05).

Regarding conventional lipid parameters, lower levels of HDL-C, LDL-C, and total cholesterol, along with higher triglyceride levels, were observed in the low HDL-C group compared with the high HDL-C group, with all differences reaching statistical significance (all p < 0.05).

3.4. Clinical Outcomes of Critically Ill Sepsis Patients Stratified by Plasma HDL-C Level

The 28-, 60-, and 90-day mortality, as well as all-cause in-hospital mortality, were all significantly higher in the low HDL-C group (all p < 0.05). Patients in the low HDL-C group also had significantly higher rates of shock, use of vasopressor, and acute kidney injury (all p < 0.05), whereas the requirement for renal replacement therapy did not differ significantly between the two groups. No significant differences were observed between the groups in the duration of mechanical ventilation, ICU length of stay, or total hospital length of stay (Table 4).

3.5. Factors Associated with Hospital Mortality

After adjusting for potential confounders, multivariable logistic regression analyses demonstrated that immunocompromised status, lower BMI and lower plasma HDL-C levels were independently associated with in-hospital mortality (Table 5). A significant inverse association was observed between plasma HDL-C levels and in-hospital mortality (adjusted OR 0.971 [95% CI 0.946–0.996], p = 0.024).

Notably, a plasma HDL-C level ≤ 22.5 mg/dL demonstrated the highest predictive value among all studied variables and was independently associated with in-hospital mortality (adjusted OR 2.330 [95% CI 1.257–4.317], p = 0.007). Furthermore, the 90-day survival rate was significantly higher in the high HDL-C group (> 22.5 mg/dL) compared with the low HDL-C group (≤ 22.5 mg/dL) (72.4% vs. 54.9%, p = 0.003, log-rank test) (Figure 2).

4. Discussion

The key finding of this prospective study is that plasma HDL-C demonstrated the greatest predictive value among all conventional lipid parameters and was independently associated with in-hospital mortality in critically ill patients with sepsis. Our findings suggest that plasma HDL-C measured at ICU admission may serve as an easily accessible and useful for early risk stratification in this population.

Metabolic derangements in glucose, amino acid, and lipid metabolism occur during both acute and prolonged critical illness and are closely associated with disease severity and clinical outcomes [9,10,11,12]. Critical illnesses, particularly sepsis, are characterized by a hypermetabolic and hypercatabolic state that disrupts lipid homeostasis. This disruption is marked by an immediate and sustained decline in plasma HDL-C, LDL-C, and total cholesterol levels, the magnitude of which correlates with the intensity of inflammation, disease severity, and mortality, whereas hypertriglyceridemia appears to have limited prognostic value. Potential underlying mechanisms include reduced dietary intake and intestinal fat absorption, decreased lipid synthesis, impaired cholesterol transport, increased metabolic consumption, and enhanced toxin scavenging [13,14,15,16].

HDL exerts multiple biological functions, including promoting macrophage cholesterol efflux and reverse cholesterol transport from peripheral tissues to the liver, maintaining endothelial function by stimulating nitric oxide production, and providing antioxidative effects through the inhibition of LDL oxidation. In addition, HDL has anti-inflammatory, anti-apoptotic, antithrombotic, and immunomodulatory properties. Notably, HDL particles can bind and neutralize lipopolysaccharide and circulating cytokines via scavenger receptor class B type 1, the principal hepatic HDL receptor [6,7,15,17,18,19,20,21,22,23,24].

Plasma HDL-C levels have been shown to correlate with the severity of organ failure and clinical outcomes in patients with critical illness, including those with sepsis and septic shock [25,26,27,28,29,30]. However, whether conventional lipid parameters differ between critically ill sepsis patients with and without ARDS remains unclear. In the present study, patients who had ARDS at ICU admission exhibited significantly higher levels of inflammatory markers, including CRP and IL-6, as well as greater disease severity, reflected by higher APACHE II and SOFA scores, compared with those without ARDS. Notably, patients with ARDS also demonstrated significantly lower levels of HDL-C, LDL-C, and total cholesterol (all p < 0.05).

In multivariable regression analyses, HDL-C remain independently associated with in-hospital mortality beyond inflammatory (CRP and IL-6) and severity markers (PaO₂/FiO₂). HDL could provide additional biological properties including regulation of lipid metabolism and transport, hormone synthesis, endotoxin neutralization, modulation of cytokine release, and maintenance of vascular homeostasis. This complementary role may explain that HDL reflect pathophysiological processes beyond systemic inflammation and why HDL-C remained independently associated with adverse outcomes even after adjustment for inflammatory and severity markers. Our findings suggest that critically ill sepsis patients may experience profound dysregulation of inflammatory and immune responses, accompanied by progressive organ dysfunction during the acute phase of critical illness. These processes could lead to alterations in lipid metabolism, including increased lipoprotein consumption for endotoxin neutralization and enhanced utilization of cholesterol for steroidogenesis and other metabolic demands associated with the stress response to critical illness, thereby contributing to hypocholesterolemia and are at an increased risk of mortality.

In addition, previous studies have reported that underweight status is associated with an increased risk of mortality, whereas overweight and obesity are associated with a lower risk of mortality among critically ill patients [31,32]. Consistent with these findings, our multivariable regression analyses demonstrated that a lower BMI was independently associated with increased hospital mortality. However, the mean BMI of our study population (23.6 kg/m²) was within the normal range, and patients were not further stratified according to BMI categories.

A previous study showed that an HDL-C level below the cutoff of 20 mg/dL was significantly associated with mortality in patients with severe sepsis. Similarly, another study revealed that an HDL-C level below a cutoff of 25.1 mg/dL was significantly associated with multiple organ dysfunction and death in patients with suspected sepsis [25,33]. In our study, critically ill sepsis patients with low plasma HDL-C levels (≤ 22.5 mg/dL) exhibited significantly higher inflammatory markers, including CRP and IL-6, more severe hypoxemia, greater organ failure severity, and lower levels of HDL-C, LDL-C, and total cholesterol compared with those with higher HDL-C levels. In addition, patients in the low HDL-C group experienced worse clinical outcomes, including a significantly higher risk of all-cause in-hospital mortality, shock, vasopressor use, and acute kidney injury (all p < 0.05). These findings suppose that critically ill sepsis patients with more exaggerated inflammatory responses and multiple organ dysfunction may consume greater amounts of HDL particles due to their endotoxin-scavenging and anti-inflammatory properties, leading to reduced circulating HDL-C levels. This process may, in turn, contribute to poorer clinical outcomes.

Low HDL-C levels are a common biomarker across a range of disease states; however, the precise mechanisms underlying the decline in HDL-C during critical illness remain incompletely understood. In our multivariable regression analyses, lower plasma HDL-C was the only lipid parameter independently associated with an increased in-hospital mortality. These findings underscore the potentially central role of HDL-C in the pathophysiology of critical illness. Taken together, systemic inflammation and oxidative stress are known to induce compositional and functional changes in HDL particles, resulting in impaired antioxidant and anti-inflammatory properties. In the present study, patients with ARDS, non-survivors, and those with lower HDL-C levels exhibited more severe hypoxemia, as reflected by lower PaO₂/FiO₂ ratios, along with greater systemic inflammation, as evidenced by higher CRP and IL-6 concentrations, compared with patients without ARDS, survivors, and those with higher HDL-C levels. These findings suggest that increased oxidative stress and inflammatory burden during sepsis may contribute to HDL dysfunction, potentially impairing its protective biological functions. Although HDL functionality was not directly assessed, our findings support the potential role of HDL dysfunction (e.g., impaired antioxidant/anti-inflammatory properties) as a mechanism that may explain the observed associations that increased oxidative stress and inflammatory burden may cause HDL dysfunction, which in turn may promote pro-inflammatory activities and contribute to endothelial injury, dysregulated immune responses, and worse clinical outcomes. Our results also suggest that plasma HDL-C measured at ICU admission may serve as a simple, early, and clinically useful for early risk stratification. Furthermore, these findings raise the possibility that early therapeutic strategies targeting HDL metabolism or function may represent a novel approach to improve outcomes in critically ill patients.

Although high HDL-C concentrations have traditionally been considered protective against cardiovascular and inflammatory diseases, their causal role remains uncertain, and extremely elevated levels may be detrimental to human health. Emerging evidence suggests a U-shaped relationship between HDL-C levels and clinical outcomes, with both low and excessively high HDL-C concentrations (>80 mg/dL in men and >100 mg/dL in women) associated with increased risk of adverse outcomes and all-cause mortality—a phenomenon referred to as the “HDL cholesterol paradox” [34,35,36]. The structure and function of HDL particles are highly heterogeneous and complex. Accordingly, the biological effects of HDL-C may depend not only on its circulating quantitative concentration but also on qualitative characteristics, including particle size and number, density, charge and shape, functional assay, apolipoprotein composition, and cholesterol efflux capacity [17,20,23,24,29,37,38,39]. These considerations highlight that HDL-C concentration alone may not fully capture its functional capacity, and further investigation into HDL functionality may provide deeper insights into its role in critical illness.

This study has several limitations. First, this was a single-center observational cohort study conducted at a tertiary medical center in Taiwan, which may limit the generalizability of our findings. As a tertiary referral medical center, our institution may care for patients with greater illness severity and a different case-mix than other populations and institutions. Moreover, the lack of internal and external validation may have reduced the robustness and external validity of the study findings. Second, as patients were enrolled during the acute phase of critical illness characterized by a hypercatabolic state, conventional plasma lipid parameters were measured only at ICU admission; thus, dynamic changes over time were not assessed. Third, this study cohort represents only a very small proportion of the overall ICU cohort with a highly selected subpopulation and does not adequately reflect the general ICU population. Enrollment depended on study screening procedures, informed consent, and operational feasibility, which may have further contributed to selection bias. The proportion of ARDS cases was relatively high, which warrants cautious interpretation of the results. Accordingly, our findings may be more appropriately interpreted as an association analysis within a selected cohort. Finally, given the observational design, our study was intended to evaluate associations between conventional plasma lipid parameters at ICU admission and clinical outcomes, and causality cannot be inferred. Although low HDL-C levels were independently associated with in-hospital mortality, HDL-C may primarily function as a surrogate marker of systemic inflammation and critical illness severity rather than a direct causal factor. The biological function of HDL cannot be fully elucidated by measuring absolute plasma HDL-C concentrations alone. Both acute and chronic systemic inflammation, as well as oxidative stress, may induce structural and conformational modifications of HDL particles, leading to potentially impaired antioxidant and anti-inflammatory functions—commonly referred to as dysfunctional HDL—which may further exacerbate inflammation, oxidative stress, and endothelial injury [23,24]. We did not assess genetic factors or the complex heterogeneity of HDL particle composition and function, nor did we investigate the precise cellular and molecular mechanisms linking lipid metabolism to the pathophysiology of sepsis or ARDS. Future studies incorporating functional assays and mechanistic approaches are warranted to better clarify these relationships. Specifically, investigating whether the functional heterogeneity of HDL particles or specific HDL subclasses is causally linked to the pathophysiology of sepsis or ARDS will help determine their potential as early prognostic indicators or therapeutic targets.

5. Conclusions

Our findings demonstrate that plasma HDL-C levels at ICU admission are independently associated with in-hospital mortality in critically ill patients with sepsis, suggesting that HDL-C may serve as a simple, inexpensive, and useful for early risk stratification.