Association Between the Non-HDL-Cholesterol-to-HDL-Cholesterol Ratio and 28-Day Mortality in Septic Patients: A Cohort Study Based on the eICU Collaborative Research Database

Background: Dyslipidemia contributes to the development and progression of cardiovascular disease. However, the potential association between non-high-density lipoprotein-cholesterol-to-high-density lipoprotein-cholesterol (nonHDLc/HDLc) ratio and mortality in septic patients is unclear. Methods: This was a retrospective cohort study of patients with sepsis in the eICU Collaborative Research Database (eICU-CRD) from 208 distinct ICUs across the United States between 2014 and 2015. All-cause mortality within 28-days after ICU admission. A multivariable logistic regression model was used to estimate the risk of death. Result: Of the 724 patients with a median age of 68 years, 43 (5.94%) died within 28 days after ICU admission. The association between the nonHDLc/HDLc ratio and the risk of all cause mortality was J shaped, and a high level was associated with increased risk of all cause mortality. The mortality rate increased when the nonHDLc/HDLc ratio higher than the turning point ( ≥ 3.41) with an adjusted odds ratio (OR) of 1.34 (95% CI: 1.07–1.67, P=0.010) for every 1 increment of nonHDLc/HDLc ratio. With the per-SD increase in the nonHDLc/HDLc ratio, the OR of mortality was 1.79 (95% CI: 1.15–2.80, P=0.010) when the nonHDLc/HDLc ratio was ≥ 3.41. The trend of sensitivity analysis was consistent with the main analysis. Conclusion: For patients with sepsis, the association between the nonHDLc/HDLc ratio and the 28-day mortality risk was J shaped. A higher level of nonHDLc/HDLc ratio was associated with an increased risk of 28-day mortality. These ndings need to be conrmed in other studies.

Conclusion: For patients with sepsis, the association between the nonHDLc/HDLc ratio and the 28-day mortality risk was J shaped. A higher level of nonHDLc/HDLc ratio was associated with an increased risk of 28-day mortality. These ndings need to be con rmed in other studies.

Background
Sepsis is a common and lethal syndrome. The systemic in ammatory response is biologically complex, redundant, and activated by infectious and non-infectious triggers. Its manipulation can bring bene ts and harm [1]. Therefore, a closer examination of the phenotypes and subphenotypes of patients who exhibit strong survival signals in sepsis may enable us to understand novel mechanisms for improving treatment.
High-density lipoprotein-cholesterol (HDLc) or low-density lipoprotein-cholesterol (LDLc) cholesterol is inversely associated with coronary heart disease (CHD) [2,3] and can respond differently to changes in diet and treatment. A poole of 458 population-based studies involving 82.1 million participants in 23 countries in Asian and Western found that HDL cholesterol increased in many Western countries, Japan and South Korea [4]. The non-HDLc-to-high-HDLc (nonHDLc/HDLc) ratio can be obtained from the standard lipid pro le without additional cost, and is highly correlated with levels of LDL particle number [4], which has been con rmed by multiple studies shows that it is a strong cardiovascular risk marker [5,6]. We previously shown that nonHDLc/HDLc ratio is an independent risk factor for the development of chronic kidney disease (CKD) [7].
A recent clinical study showed that during a median follow-up of 1.72 years, the mortality rate associated with the range of LDLc/HDLc ratios was U-shaped in hypertensive patients [8]. However, there is a lack of evidence to guide the emergency management of patients with sepsis.
We hypothesized that in patients with sepsis, even high and low nonHDLc/HDLc ratios are associated with a higher risk of all-cause mortality within 28-days after admission to the ICU. In this retrospective multicenter cohort study, we used the eICU Collaborative Research Database (eICU-CRD), which participated in the Philips Healthcare eICU program from 208 distinct ICUs in the United States between 2014 and 2015. We aimed to explore the threshold of the nonHDLc/HDLc ratio where risk of death signi cantly increases is a high priority in patients with sepsis.

Data Source
We extracted data from the eICU Collaborative Research Database (eICU-CRD) [9] in accordance with the data usage agreement of the PhysioNet Review Board (our record ID: 40859994). From 2014 to 2015, all data was automatically stored through the Philips Healthcare eICU program and retrieved electronically [9]. The eICU-CRD has been used for observational research [10][11][12]. Ethical approval from our local ethics committee is not required, as this is a retrospective analysis based on an anonymous database of researchers.

Study population
Brie y, all patients diagnosed with sepsis on admission to the ICU were included. Sepsis was de ned as suspected or documented infection plus an acute increase in SOFA score greater than 2 points [13]. Recorded on the Acute Physiology and Chronic Health Evaluation (APACHE) IV dataset [14]. The following exclusion criteria were used: (1) Not rst ICU admission; (2) ICU stay < 48 hours; (3) 18 years old; (4) Missing ICU outcome; (5) Missing total cholesterol after ICU asmission and system error. The study owchart was presented in Fig. 1.

Variables
The eICU database includes demographic records, physiological indicators of bedside monitors, diagnosis via International Classi cation of Diseases, 9th Edition, Clinical Modi cation (ICD-9-CM) codes, and other laboratory data obtained during routine medical care.
All subject data within the rst 24 hours after hospitalization admission were collected from eICU-CRD. The physiological variables, including temperature (°C), respiratory rate, heart rate (HR) and mean arterial pressure (MAP), were obtained from the table apacheApsVar. Baseline characteristics such as age, gender, ethnicity, and weight were collected from the tables of patient and apachePatientResult. The laboratory indices of total cholesterol, triglycerides, HDLc, and LDLc from laboratory tables. Non-HDL-c refers to total cholesterol (TC) minus HDLc, including cholesterol in atherogenic lipoproteins.

Outcomes
The outcome of the study was all-cause ICU mortality within 30 days after admission to the ICU. In the supplemental analysis, we also analyzed 14-days mortality rate after admission to the ICU.

Statistical analysis
Continuous variables are described as means ± SD or median and interquartile ranges (IQR). Categorical data is presented as number and percentage. The difference according to the tertiles of the nonHDLc/HDLc ratio was compared using one-way analysis of variance (ANOVA) for continuous data and Chi-squared tests for categorical variables.
We applied a Generalized Additive Model (GAM) to investigate dose-response relationship between the nonHDLc/HDLc ratio and mortality (Fig. 1). We applied logistic regression model to estimate the association between the nonHDLc/HDLc ratio and 28-day mortality. The results were presented as odds ratios (ORs) with its 95% con dence intervals (95% CIs). Crude regression estimates are presented, as well as estimates adjusted for covariates. We selected these confounders on the basis of their association with the outcomes of interest or changes in effect estimates of more than 10% [15]. After considering the clinical signi cance, we selected to adjust the following covariates: age (years), sex, weight, heart rate, SOFA score, and site of infection.
We further applied two-piece-wise linear regression model to examine the threshold effect of the nonHDLc/HDLc ratio on mortality ( Table 3). The turning point for the nonHDLc/HDLc ratio was determined using "exploratory" analyses, which is to move the trial turning point along the pre-de ned interval and pick up the one which gave maximum model likelihood. We also performed a log-likelihood ratio test and compared the one-line linear regression model with two-piece-wise linear model. As described in previous analysis [16,17]. Adjusted for age (years), sex, weight, heart rate, SOFA score, and site of infection. CI, con dence interval; OR, odds ratio.
To examine the robustness of results, we conducted sensitivity analyses. Dummy variables were used to indicate missing covariate values, which was performed when continuous variables are missing more than 1%. The two-sided alpha level was set at 0.05. All the statistical analysis was performed using the EmpowerStats (www.empowerstats.com, X&Y solutions, Inc. Boston MA) and R software version 3.6.1 (http://www.r-project.org).

Baseline characteristics
Data from 724 patients were analyzed. The median age was 68 years (IQR 58-78 years). 334 patients (46.1%) were women. Table 1 compares the patient's demographic, vital signs, laboratory, site of infection, and severity of illness through the tertiles of the nonHDLc/HDLc ratio. Compared with subjects in the lowest tertile of the nonHDLc/HDLc ratio, subjects in the highest tertile were younger and had higher admission weight. Unadjusted association between baseline variables and 28day mortality
Using the generalized additive model, the J shaped association between the nonHDLc/HDLc ratio and 30day mortality was detected ( Table 3). The linear regression model and a two-piece-wise linear regression model were compared, and the P value of the log-likelihood ratio test is < 0.037. This result indicates that the two-piece-wise linear regression model should be used to t the model.
The trend of sensitivity analysis was consistent with the main analysis. In the supplementary analysis, we also analyzed 14-day mortality rate, and the result were generally similar to our main results ( Figure  S1 and Table S1). Dummy variables were used to indicate missing covariate values. Similar results were obtained after considering the impact of missing data (Table S2).

Discussion
This retrospective cohort study found that higher nonHDLc/HDLc ratio was associated with a higher risk of 30-day mortality in patients with sepsis in the eICU-CRD database from 208 distinct ICUs across the United States between 2014 and 2015. The major nding was that the association between the nonHDLc/HDLc ratio and the risk of all cause mortality was J shaped, with high levels associated with an increased risk of all cause mortality. To our knowledge, this is the rst study to report the association between the nonHDLc/HDLc ratio and 28-day mortality in septic patients.
Most studies investigating the relationship between levels of LDLc and the risk of death found no association [18][19][20] or an inverse association [21][22][23]. A prospective population-based cohort study with 5 years of follow-up and a validation cohort of African Americans with 4.25-year follow-up show that neither HDLc nor LDLc is associated with mortality[18]. One data from 1948 to 1980 on 5209 men and women enrolled in the Framingham Heart Study found that the negative results in the oldest age group for all-cause morality appeared to be due to a negative relationship with LDL-C levels rather than the protective effect of high HDLc levels [19]. You et al. [24] included 356 patients with intracranial hemorrhage (mean follow-up = 0.22 years) found that the LDLc/HDLc ratio was negatively correlated with all-cause mortality. Liu et al. [25] recruited 3250 stroke patients and found a negative relationship between the LDLc/HDLc ratio and all-cause mortality. included 116,508 individuals from the general population, the association between HDL cholesterol and all-cause mortality was U shaped, with both extreme high and low HDL cholesterol levels associated with high mortality. No previous study has examined HDLc or LDLc levels associated with the lowest risk of mortality in patients with sepsis. To our knowledge, although the relationship between blood lipids levels and mortality has been concerned, the association between nonHDLc/HDLc ratio and 30-day mortality are still unclear, which prompted us to conduct the current study.
We used non-HDL cholesterol rather than LDL cholesterol because our database measures TC and HDL cholesterol and can be calculated by subtracting non-HDL cholesterol from it. Of the 724 patients included, 178 did not have LDL measurements. In addition, the most commonly used estimation method, the Friedewald equation, can be inaccurate [29]. That non-HDL and LDL cholesterol were correlated in studies with data on both variables (r = 0.93) [4]. In our study, the correlation coe cient between non-HDL . This is similar to our result that the mortality rate associated with non-HDLc / HDLc ratios is not liner.
A possible explanation for our ndings was that the association between low levels of nonHDLc and an increased risk of all-cause mortality could be explained by reverse causation. Debilitation and illness are hypothesized to cause a decrease in levels of cholesterol [31,32]. Nonetheless, cholesterol-related risks are more complex and involve the interplay of several factors such as cholesterol particle concentration, reverse cholesterol transport, and triglyceride-rich lipoproteins, to mention a few [33]. The J shaped association between nonHDLc/HDLc ratios and mortality may resemble the obesity paradox, which is largely explained by methodological issues, including reverse causation [34].
The current retrospective cohort study is based on the eICU-CRD database and may be important for understanding nonHDLc/HDLc ratios in septic patients (ie, when the focus is not limited to myocardial infarction or atherosclerotic cardiovascular disease). We investigated thresholds for nonHDLc/HDLc ratio levels that signi cantly increased the risk of death in patients with sepsis.

Study limitations
One limitation of this study is inherent to the observational nature of the study design, which lends itself subject to limitations that should be considered including confounding by indication. Our ndings are hypothesis-generating and do not imply causality. In our analysis, we adjusted for possible confounding factors, including age (years), gender, weight, heart rate, SOFA score, and site of infection. Despite this adjustment, some unmeasured confounding may remain. Additional limitations of our study include missing data for some variables. Nonetheless, we used contemporary methods to deal with missing data to minimize bias.
Another limitation is related to the fact that the diagnosis was based on the ICD-9 coding which the responsible physician found relevant, and we did not have information concerning causes of death. Since we are examining mortality over a short period after the date of visit to the ICU, we did not nd it bene cial to distinguish between cardiovascular and non-cardiovascular death.
Furthermore, lack of information on interventions during the initial stabilization may have in uenced nonHDLc/HDLc ratio levels and survival. It is noteworthy that the potential resulting from interventions would bias toward to the null, and thus result in an underestimation of the association between nonHDLc/HDLc ratio level and mortality.
Finally, we acknowledge that the participants were patients referred to the emergency department for some reason, limiting the generalization of ndings to other populations.

Conclusion
Data from the eICU-CRD database was used to identify 724 patients with sepsis. This study identi es a nonlinear dose-response relationship between nonHDLc/HDLc ratios and 28-day mortality. The probability of mortality rose rapidly when the nonHDLc/HDLc ratio higher than the turning point (may at 3.41). This retrospective cohort study showed that higher nonHDLc/HDLc ratios were associated with higher risk of 28-day mortality in patients with sepsis.  Figure 1 Study population Brie y, all patients diagnosed with sepsis on admission to the ICU were included. Sepsis was de ned as suspected or documented infection plus an acute increase in SOFA score greater than 2 points [13]. Recorded on the Acute Physiology and Chronic Health Evaluation (APACHE) IV dataset [14].
The study owchart was presented in Figure 1.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.