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

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

This single-center prospective observational study included 285 critically ill patients with sepsis in the ICU, of whom 223 had ARDS, and 62 did not. The researchers observed lower HDL-C levels in patients with ARDS and in non-survivors, and linked these levels to hospital mortality. Key strengths include prospective data collection and a sizable ARDS subgroup. However, I have fundamental questions about the novelty and clinical usefulness of these findings.

Many manuscripts already examine the link between HDL-C and mortality. Notably, prior studies have indicated that lower HDL-C levels are associated with poorer outcomes in sepsis (Cirstea et al., 2017; Taylor et al., 2024). Researchers have also examined the relationship between lipid profiles and clinical outcomes in septic patients with and without ARDS. Although this study includes a larger ARDS subgroup and uses early-admission measurements, it does not clearly explain how its findings advance existing knowledge or improve clinical risk assessment beyond established practices.

My major Methodological concerns are,

Table 5 does not clearly indicate whether the analysis adjusted for illness-severity scores such as SOFA or APACHE II, even though these scores differ between survivors and non-survivors. As a result, it is unclear whether HDL-C provides prognostic value beyond disease severity.
The methods section states that the 22.5 mg/dL cutoff was chosen to identify ARDS but was then used to predict mortality; this cutoff needs to be separately evaluated and validated for mortality prediction. The claim that HDL-C has better predictive value is not supported by formal comparisons with established clinical models.
The choice is also concerning. Of 6,238 ICU sepsis admissions screened, only 285 were included, primarily because of operational issues or lack of consent. This could introduce selection bias and limit the results' generalizability.
Further, the study lacks both internal and external validation, making it difficult to apply these findings broadly or in actual clinical practice.

Due to its limited novelty and ongoing analytical issues, I cannot recommend publishing this manuscript. The findings seem mostly confirmatory; however, the manuscript does not show that HDL-C offers clinically meaningful prognostic information beyond current severity measures. I encourage the authors to clarify the specific research gap and demonstrate prognostic value beyond current clinical measures before considering publication elsewhere.

Comments on the Quality of English Language

Need to be improved.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The article is a prospective cohort study involving 285 critically ill patients, the desing of the stady is correct, tables and results are clearly displayed.
The patient population in the Section 2.1. is not clealy defined. I presume these are COVID-19 patients with pneumonia. Futhermore, the majory of patients in the ARDS group appaer to be immunocompromised and are reciiving corticosteroid therapy.
What is your definition of immunocompromised in this cohort? Do you know the type, dose and duration of steroid therapy? How mignt immunocompromise and baseline steroid use confound the interpretation of HDL-C and ARDS outome? Are there any similar studies in COVID patients?
The Discussion section should be also include studies that perforemed serial blood sampling to monitor plasma HDL-C levels and kinetics, wtha their results showed, and wheter serial measurements offer an adventage over a single determination.
In the literature, I also found some meta-analyses of plasma HDL-C level monitoring in sepsis, and those results also be briefly predented in the Discusion Section
In connection with above, the Conclusion Section should be adapted accordingly.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Table 5 shows the adjusted OR only for BMI, immunocompromised status, and HDL-C, while SOFA and APACHE II are even absent from the univariate analysis. However, patients with low HDL-C were much more severe (SOFA 9.3 vs 6.1; APACHE II 17.8 vs 13.9), so the association may only reflect the severity of the disease. The authors should include SOFA or APACHE II in the model, albumin, sex, and shock, justify the choice of variables, and report the complete model.
The threshold of 22.5 mg/dL was established to distinguish ARDS from non-ARDS, then it was applied to mortality, without validation (acknowledged in line 368) and without reported AUC. The claim about the "highest predictive value" (lines 240–241) is not supported. The authors should report the AUC for mortality, internally validate the threshold, and compare HDL-C with SOFA and APACHE II.
The cohort is predominantly composed of patients with pneumonia (91%) and ARDS (78%). The authors acknowledge the selection in the Limitations section (lines 372–378), but the title, abstract, and conclusions refer to sepsis in general. These should be adjusted.
The Methods (lines 95–98) state that not all patients were evaluated, while the Results (line 148) state that all 6,238 were evaluated. The authors need to clarify the figure and correct Figure 1.
The definition of sepsis and the timing of sample collection relative to its onset are not specified (line 104).
The study covers the pandemic period, and 77% of patients received corticosteroids, but COVID-19 is not mentioned. The authors should report the proportion of patients with COVID-19 and include it in the analysis.
The factors that modify lipids in the ICU are not discussed: corticosteroids, propofol, parenteral nutrition, and hemodilution. It is not specified whether LDL-C was measured or calculated.
References 9 and 10 appear to come from the same cohort. The authors should declare the overlap and specify whether the lipid analysis was planned from the beginning.
Lines 185–186 state that CRP and IL-6 are elevated in the deceased, but Table 2 shows no difference (p = 0.617 and p = 0.878). The text and argumentation in the discussion need to be corrected.
The point from which mortality at 28, 60, and 90 days is calculated is unclear: Table 4, section 2.3 (line 117), and Figure 2 describe it differently. The authors need to clarify it.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Dear Authors,

Thank you for revising the manuscript and providing detailed responses to the previous comments. I appreciate that you included SOFA in the multivariable analysis, conducted a separate ROC analysis for hospital mortality, and clarified the enrollment process—these enhancements improve the manuscript. Nevertheless, it still needs further refinement before acceptance.

Major comments

Adding SOFA improves the analysis. However, an adjusted association alone does not confirm that HDL-C improves mortality prediction. I recommend comparing models that include and exclude HDL-C, either by internally validating their performance or by removing claims of proven clinical utility.
The 17.5 mg/dL cutoff was selected and tested within the same cohort and remains exploratory. Please emphasize that HDL-C is continuous and confirm the cutoff's predictive value.
Provide a list of all variables in the final regression model, outline the criteria for their inclusion, and specify the sample size along with how missing data were managed. Verify that references to adjustments for CRP, IL-6, and PaO₂/FiO₂ align with the described analyses.
The clarification regarding enrollment does not address potential selection bias. Define the source population, clarify why many patients were not assessed, compare enrolled and non-enrolled patients where feasible, and align the exclusion criteria with the reported exclusions
Please clearly differentiate between 28-, 60-, and 90- day all-cause mortality and hospital mortality. Include details on post-discharge follow-up and censoring, ensure consistent definitions throughout the manuscript, and add numbers at risk beneath the survival curves.
The ARDS subgroup primarily provides descriptive data due to unadjusted comparisons. Please clarify its exact contribution.
Since HDL function was not assessed, the results cannot confirm HDL dysfunction or any therapeutic effects. Please treat these mechanisms as hypotheses and revise the statements to indicate significantly higher CRP or IL-6 levels in non-survivors.
IL-6 data were available for 205 of 285 patients. Please specify the subgroup denominators in Tables 2 and 3 and indicate whether missing data affected any regression or sensitivity analyses.
Define shock and acute kidney injury operationally and clarify whether these outcomes were present at admission or developed after HDL-C sampling.
Table 4 states that patients receiving maintenance hemodialysis were excluded from the renal replacement therapy outcome. Please clarify how you applied this for exclusion.
I suggest including detailed methods for the lipid assay, as well as the timing of the severity-score assessment in relation to blood sampling, if possible.