Abstract
Sepsis accounts for one in three hospital deaths. Higher concentrations of high‐density lipoprotein cholesterol (HDL‐C) are associated with apparent protection from sepsis, suggesting a potential therapeutic role for HDL‐C or drugs, such as cholesteryl ester transport protein (CETP) inhibitors that increase HDL‐C. However, these beneficial clinical associations might be due to confounding; genetic approaches can address this possibility. We identified 73,406 White adults admitted to Vanderbilt University Medical Center with infection; 11,612 had HDL‐C levels, and 12,377 had genotype information from which we constructed polygenic risk scores (PRS) for HDL‐C and the effect of CETP on HDL‐C. We tested the associations between predictors (measured HDL‐C, HDL‐C PRS, CETP PRS, and rs1800777) and outcomes: sepsis, septic shock, respiratory failure, and in‐hospital death. In unadjusted analyses, lower measured HDL‐C concentrations were significantly associated with increased risk of sepsis (p = 2.4 × 10−23), septic shock (p = 4.1 × 10−12), respiratory failure (p = 2.8 × 10−8), and in‐hospital death (p = 1.0 × 10−8). After adjustment (age, sex, electronic health record length, comorbidity score, LDL‐C, triglycerides, and body mass index), these associations were markedly attenuated: sepsis (p = 2.6 × 10−3), septic shock (p = 8.1 × 10−3), respiratory failure (p = 0.11), and in‐hospital death (p = 4.5 × 10−3). HDL‐C PRS, CETP PRS, and rs1800777 significantly predicted HDL‐C (p < 2 × 10−16), but none were associated with sepsis outcomes. Concordant findings were observed in 13,254 Black patients hospitalized with infections. Lower measured HDL‐C levels were significantly associated with increased risk of sepsis and related outcomes in patients with infection, but a causal relationship is unlikely because no association was found between the HDL‐C PRS or the CETP PRS and the risk of adverse sepsis outcomes.
Study Highlights.
WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?
Higher concentrations of circulating high‐density lipoprotein cholesterol (HDL‐C) are associated with apparent protection from sepsis, suggesting a potential therapeutic role for HDL‐C. However, these beneficial clinical associations might be due to confounding.
WHAT QUESTION DID THIS STUDY ADDRESS?
Is there a causal relationship between low HDL‐C and increased risk of sepsis and sepsis‐related adverse outcomes?
WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?
This study confirmed the significant association between low HDL‐C and increased risk of sepsis, but a causal relationship is unlikely because there was no association between genetically predicted HDL and the risk of sepsis or sepsis‐related adverse outcomes in patients admitted to the hospital with infection.
HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?
Low HDL‐C is associated with an increased risk of sepsis, but genetic analyses do not support a causal relationship.
INTRODUCTION
Sepsis, a complication of serious infection, is associated with organ failure and high mortality 1 and is one of the leading causes of death, contributing to ~250,000 deaths annually in the United States. 2 Hospital costs for sepsis exceed $24 billion annually, accounting for 13% of US hospital costs; sepsis is the single most expensive diagnosis for Medicare, accounting for ~8% of all expenditures. 3 , 4 , 5 Moreover, sepsis has additional long‐term societal costs because patients who survive remain at increased risk of death and have significantly impaired quality of life. 6 , 7 Yet, there are no specific effective treatments for sepsis; thus, there is great enthusiasm for new approaches to predict risk and treat patients.
The effect of high‐density lipoprotein (HDL) on sepsis and related outcomes is an area of therapeutic interest. In animal models, ApoA1‐null mice with HDL deficiency had decreased clearance of lipopolysaccharide (LPS) and increased risk of experimental septic death; in contrast, transgenic mice with high HDL levels were protected from LPS and had improved survival. 8 , 9 , 10 HDL cholesterol (HDL‐C) levels—the measure of HDL obtained clinically—drop rapidly in patients with sepsis 11 ; further, lower levels of HDL‐C are associated with early onset of sepsis, increased risk of organ failure, and higher sepsis mortality. 12 , 13 However, epidemiologic studies of lipid levels that have reported these clinical associations with HDL suffer from potential unaccounted comorbidities, residual confounding, and reverse causality. A previous study we performed illustrated this problem in relation to levels of low‐density lipoprotein cholesterol (LDL‐C) and sepsis. 14 As reported by others, we found that lower measured LDL‐C levels were indeed associated with increased risk of sepsis; however, the association was due to comorbidities, because a Mendelian randomization (MR) approach and appropriately adjusted clinical approaches indicated that LDL‐C did not influence sepsis or its outcomes risk directly.
Genetic approaches akin to MR performed in large biobanks that have genotypes and phenotypes can overcome the inherent limitations of epidemiologic studies. These approaches use genetic variants as proxies of exposure (e.g., HDL‐C) and test their effects on an outcome. Because genetic variants are randomly allocated during conception, genetic approaches are less vulnerable to confounding and reverse causation. A study using this approach suggested that elevated genetically determined HDL‐C, and, particularly, levels determined by the CETP gene, conferred a protective effect on the risk of both infection and sepsis mortality. 15 , 16 However, a potential HDL‐based therapy is most likely to be used in patients hospitalized with infection to prevent progression to sepsis. Thus, the question whether HDL plays a role in the transition from serious infection to sepsis remains important.
Leveraging a de‐identified electronic health record (EHR) repository linked to a DNA biobank at Vanderbilt (BioVU), we set out to test and contrast the associations between HDL‐C (measured and genetically predicted) and the risk of progression from infection to sepsis as well as adverse sepsis‐related outcomes in patients admitted to hospital with infection. We specifically also tested CETP genetic variants because medications targeting this gene have been developed and effectively increase HDL‐C levels, thus offering potential for drugs’ repurposing to treat sepsis.
METHODS
Data sources
Data were obtained from the Synthetic Derivative that contains a de‐identified copy of the EHR for all patients at Vanderbilt University Medical Center (VUMC), a large teaching hospital (~3.4 million individual records as of December 2021), and from genomewide genotyping that is available for ~90,000 BioVU patients on the Infinium Multi‐Ethnic Genotyping Array (MEGA). Both International Classification of Disease, ninth revision, Clinical Modification (ICD‐9‐CM) and tenth revision (ICD‐10‐CM) and Current Procedural Terminology (CPT) codes were used for outcomes and cohort construction and covariates. The study was approved by the VUMC Institutional Review Board.
Hospitalized infection cohort
We studied individuals older than 18 years with race assigned as White or Black in the de‐identified EHRs admitted to the hospital with an infection between January 2000 and August 2020 (Figure 1). The day of hospital admission was designated day zero. Infection was defined as having a billing code indicating infection (Table S1) and receiving an antibiotic (Table S2) within 1 day of hospital admission (i.e., on days −1, 0, or + 1), as described previously, 14 , 17 using ICD‐9‐CM and ICD‐10‐CM codes for infection based on the criteria of Angus et al. 18 , 19 excluding viral, mycobacterial, fungal, and spirochetal infections (Table S1). If a patient had more than one qualifying episode of infection, only the first episode was included. We excluded individuals undergoing cardiac surgery, those with cardiogenic shock, those admitted for organ transplantation (Table S3), and those who had no relevant laboratory values (creatinine, bilirubin, and platelets) on days −1, 0, or +1. We also excluded patients who had a positive test or ICD code (U07.1) for coronavirus disease 2019 (COVID‐19) on days −1, 0, or +1. Within the cohort of patients who had been hospitalized with infection, we further identified two nonexclusive subcohorts (Figure 1): (1) those who had at least one HDL‐C measurement before admission (the median of measured HDL‐C values before admission was calculated for each patient); and (2) those who had genomewide genotype available (the cohort with genetically predicted HDL‐C).
FIGURE 1.
Algorithm to identify patients hospitalized with infections. COVID‐19, coronavirus disease 2019; CPT, Clinical Procedure Code; ICD, International Classification of Disease; VUMC, Vanderbilt University Medical Center.
Outcomes
For both the genetically predicted HDL‐C and the measured HDL‐C analyses, the primary outcome was the development of sepsis in patients with infection and secondary outcomes were severe sepsis or septic shock, respiratory failure, and in‐hospital death.
Sepsis was defined by the Sepsis‐3 criteria of concurrent infection and organ dysfunction (Figure S1). 3 , 19 The validity of the Sepsis‐3 definition has been confirmed in real‐world hospital settings, 3 , 20 and Sepsis‐3 criteria were optimized and validated for application across EHR systems from 409 hospitals. 3 We have adapted, validated, and applied this EHR‐based Sepsis‐3 algorithm to our de‐identified EHRs (Table S4). 14 The algorithm uses billing codes and clinical criteria and had a sensitivity of 69.7% and specificity of 98.1% in the validation study, 3 and performed well in the Vanderbilt de‐identified EHRs (Table S4). 14 , 17 The vast majority of cases of sepsis (87%) are present on admission to hospital 3 ; therefore, to minimize the confounding effects of sepsis occurring secondary to procedures or events in the hospital, we studied sepsis occurring within 1 day of hospital admission (days −1, 0, and +1).
In brief, individuals with an infection code who received an antibiotic met the definition of sepsis if they had either ICD codes for septic shock or severe sepsis (ICD‐9‐CM, 995.92 and 785.52; ICD‐10‐CM, R65.20 and R65.21) because these are highly specific (99.3%), 3 or met any Sepsis‐3 organ dysfunction criterion (Figure S1). 3 Criteria for organ dysfunction included: (1) cardiovascular failure: defined as the use of the vasopressor, which we extracted as use of norepinephrine, or use of the vasopressors dobutamine or dopamine not related due to stress echocardiography (CPT codes 78452, 93015, 93018, 93016, 93017, and 93351) and with at least two mentions of any of the keywords (infection, sepsis, and septic); (2) respiratory failure: defined by ICD and CPT codes for ventilation and admission to an ICU (Table S5); (3) renal failure: defined by a doubling or greater increase of baseline creatinine (baseline creatinine was defined as the lowest creatinine between 1 year before admission and hospital discharge); (4) hepatic failure: defined as a total bilirubin greater than or equal to 34.2 μmol/L (2 mg/dl) that had doubled from baseline (baseline bilirubin was defined as the lowest total bilirubin occurring between 1 year before admission and hospital discharge); and (5) hematologic failure: defined as a platelet count <100,000 /microl and greater than or equal to 50% decline from a baseline that must have been greater than or equal to 100,000 (the baseline value was the highest platelet count occurring between 1 year before admission and hospital discharge). 14 In‐hospital deaths were defined as patients who had their death recorded in the EHR within the index hospital stay.
Covariates
Covariates extracted from the EHR for each patient included median body mass index (BMI) and median HDL‐C, LDL‐C, and triglycerides (TGs); values were calculated using measurements before the index hospital admission; age was obtained at the time of the index hospital admission. A weighted Charlson/Deyo comorbidity score was calculated 21 , 22 using relevant diagnostic codes in the year before the index hospital admission (Table S6) grouped into the 17 Charlson/Deyo comorbidity categories. 23 , 24 , 25 Principal components (PCs) for ancestry were calculated using common variants (minor allele frequency [MAF] > 1%) with high variant call rate (>98%), excluding variants in linkage and regions known to affect PCs (HLA region on chromosome 6, inversion on chromosome 8 [8135000–12,000,000], and inversion on chromosome 17 [40,900,000–45,000,000], GRCh37 build). It was important to consider variables particularly associated with both exposure (HDL) and outcome (sepsis), because these are most likely to confound the association. To this end, we included broadly relevant covariates that potentially influence the risk of sepsis and alter HDL levels, such as age, sex, and comorbidity scores. We included LDL‐C levels as a covariate because previous studies showed measured LDL levels were associated with sepsis outcomes (but not casually related) 14 , 26 and LDL and HDL levels may be inversely correlated. Length of EHR reflects the length of time each patient was in the VUMC system and reflects the duration of follow‐up. An individual with a long follow‐up has more time in which to develop sepsis and will have more information about comorbidities.
Genotyping and single‐nucleotide polymorphism imputation
Genotyping was performed on the Infinium MEGAchip. We excluded DNA samples: (1) with call rate <95%; (2) with wrongly assigned sex; or (3) unexpected duplication. We performed whole genome imputation (Appendix S1) and filtered variants with (1) low imputation quality (r 2 < 0.3), (2) MAF <0.5%, and (3) variants with MAF different from the Haplotype Reference Consortium reference panel (MAF differences >0.3).
HDL‐C polygenic risk score
A previously validated polygenic risk score (PRS) for HDL‐C from the polygenic score catalog was applied to the cohort with genotypes available. 27 This PRS for HDL‐C uses 120 single‐nucleotide polymorphisms (SNPs) independently associated (p < 5 × 10−8) with HDL‐C in the meta‐analysis of genomewide studies performed by the Global Lipids Genetics Consortium (GLGC; Table S7) 27 , 28 , 29 ; all 120 SNPs were available in our genotyped cohort. The HDL‐C PRS was calculated for each individual by adding the number of minor alleles (0, 1, or 2) weighted for the effect size (β) of the SNP‐HDL association. Cholesteryl ester transfer protein (CETP) genetic variants and HDL‐C levels. To isolate the effects of CETP we also constructed a CETP PRS (Table S8). 16 The associations between PRSs and measured HDL‐C levels were validated in an independent cohort of 25,009 White patients who had both measured HDL‐C levels in their EHRs and genomewide genotype available (HDL PRS, r = 0.24, p < 2.2 × 10−16; CETP PRS, r = 0.15, p < 2.2 × 10−16; rs1800777, r = 0.07, p < 2.2 × 10−16); none of these patients were included in the HDL PRS cohort.
Phenome‐wide association studies
The phenome‐wide association studies (PheWAS) were conducted to identify clinical phenotypes that associate with measured HDL‐C, the HDL‐C PRS, the CETP PRS, and rs1800777. Specifically, we grouped each individual's ICD codes into phecodes following an established protocol (Appendix S1). 30 , 31 Median measured HDL‐C levels were extracted and analyzed for 313,478 White individuals and we also constructed HDL‐C PRS for 65,592 BioVU White individuals. PheWAS of measured HDL‐C were adjusted for sex, year of birth, length of EHRs; PheWAS of HDL‐C PRS, CETP PRS, and rs1800777 were additionally adjusted for sex, year of birth, length of EHRs, and five PCs. We analyzed a total of 1782 (measured HDL‐C) and 1739 (HDL‐C PRS, CETP PRS, and rs1800777) phecodes with more than 20 cases. The p values <2.8 × 10−5 (0.05/1782) and 2.9 × 10−5 (0.05/1739) were considered significant for measured HDL‐C, and the genetic predictors (HDL‐C PRS, CETP PRS, and rs1800777), respectively.
Statistical analyses
Values are shown as mean ± SD.
Measured HDL‐C
Logistic regression was used to test the associations between measured HDL‐C levels and outcomes, with adjustment for age (at hospital admission), sex, length of EHR, weighted Charlson/Deyo comorbidity score, LDL‐C level, TG levels, and BMI.
HDL‐C PRS and CETP PRS
Logistic regression was used to test the associations between the HDL‐C PRS (and CETP PRS) and outcomes with adjustment for age (at hospital admission), sex, length of EHR, and five PCs for ancestry. For all analyses, odds ratios (ORs) were estimated per increase of one SD and 95% confidence intervals (CIs) for the association between measured HDL‐C levels (before hospital admission) or PRS and the outcomes. The p values <0.05 were considered as statistically significant. In a sensitivity analysis, we tested the measured HDL‐C value closest to hospital admission. Similar to the analyses performed in the entire cohort, we performed post hoc sensitivity analyses in the 3007 individuals who had both genotype and measured HDL‐C levels. For illustrative purpose, we estimated ORs and 95% CIs for quartiles of HDL‐C levels and HDL‐C PRS (Table S9) using the highest quartile as reference and with adjustment as in the primary analyses. Analyses were also performed separately in the cohort of 13,254 Black patients hospitalized with infection of whom 2566 had HDL‐C measurements before admission and 2342 had genotype to test if the observations in White patients were generally concordant in Black patients.
RESULTS
Study cohort
There were 73,406 White patients hospitalized with infection (37,076 women and 36,330 men). They had an average EHR length of 8.2 ± 7.8 years (mean ± SD) and an average age at hospital admission of 55.3 ± 18.1 years (mean ± SD). During the hospital stay, 21,955 developed sepsis. Of the 73,406 White patients with infection, 11,612 had HDL measurements before hospital admission and 12,377 had genomewide genotypes available (3007 patients were in both cohorts; Table 1). There were 13,254 Black patients hospitalized with infection; 2566 had HDL‐C measurements before admission, and 2342 had genotypes available (Table S10).
TABLE 1.
Demographic characteristics in White patients
| Demographic variables | Measured HDL‐C (N = 11,612) | Genetically predicted HDL‐C (N = 12,377) |
|---|---|---|
| Sex | ||
| Female, N | 5686 (48.97%) | 6350 (51.30%) |
| Male, N | 5926 (51.01%) | 6027 (48.70%) |
| Age, years, mean ± SD | 62.03 ± 15.39 | 57. 10 ± 16.87 |
| EHR, years, mean ± SD | 11.48 ± 7.76 | 11.12 ± 7.72 |
| BMI, kg/m2, mean ± SD | 29.52 ± 10.19 | 29.50 ± 26.96 |
| HDL, mmol/L, mean ± SD | 1.17 ± 0.46 | − |
| LDL, mmol/L, mean ± SD | 2.43 ± 0.97 | − |
| TGs, mmol/L, mean ± SD | 1.91 ± 2.08 | − |
| Sepsis and sepsis‐related outcomes | ||
| Sepsis, N | 3439 (29.62%) | 3790 (30.62%) |
| Septic shock, N | 1084 (9.34%) | 1442 (11.65%) |
| Respiratory failure, N | 707 (6.09%) | 999 (8.07%) |
| In‐hospital death, N | 444 (3.82%) | 465 (3.76%) |
| Comorbidities | ||
| Myocardial infarction, N | 2309 (19.88%) | 1472 (11.89%) |
| Congestive heart failure, N | 2428 (20.91%) | 1594 (12.88%) |
| Peripheral vascular disease, N | 1383 (11.91%) | 987 (7.97%) |
| Cerebrovascular disease, N | 2229 (19.20%) | 1161 (9.38%) |
| Dementia, N | 355 (3.06%) | 196 (1.58%) |
| Chronic pulmonary disease, N | 2905 (25.02%) | 2546 (20.57%) |
| Rheumatic disease, N | 296 (2.55%) | 334 (2.70%) |
| Peptic ulcer disease, N | 180 (1.55%) | 143 (1.16%) |
| Mild liver disease, N | 213 (1.83%) | 260 (2.10%) |
| Diabetes with complication, N | 1883 (16.22%) | 887 (7.17%) |
| Diabetes without complication, N | 1418 (12.21%) | 915 (7.39%) |
| Hemiplegia or paraplegia, N | 90 (0.78%) | 210 (1.70%) |
| Renal disease, N | 1832 (15.78%) | 1718 (13.88%) |
| Any malignancy, including lymphoma and leukemia, except malignant, N neoplasm of skin | 2303 (19.83%) | 3459 (27.95%) |
| Moderate or severe liver disease, N | 413 (3.56%) | 593 (4.79%) |
| Metastatic solid tumor, N | 945 (8.14%) | 1975 (15.96%) |
| AIDS/HIV, N | 78 (0.67%) | 52 (0.42%) |
Note: For LDL‐C and HDL‐C, 1 mmol/L = 38.67 mg/dl; for TGs, 1 mmol/L = 88.57. mg/dl.
Abbreviations: BMI, body mass index; EHR, electronic health record; TGs, triglycerides.
Measured HDL‐C levels and sepsis or sepsis‐related adverse outcomes
Among 11,612 White patients with HDL measurements, the median HDL‐C level was 1.17 ± 0.46 mmol/L (mean ± SD). In unadjusted analyses, higher measured HDL‐C levels were significantly associated with reduced risk of sepsis (OR 0.62, 95% CI 0.56–0.68; p = 2.36 × 10−23), septic shock (OR 0.58, 95% CI 0.50–0.68; p = 4.12 × 10−12), respiratory failure (OR 0.59, 95% CI 0.49–0.71; p = 2.82 × 10−08), and in‐hospital death (OR 0.49, 95% CI 0.39–0.63; p = 1.02 × 10−08). After adjusting for age, sex, EHR length, comorbidity score, LDL‐C levels, TG levels and BMI, the associations with sepsis (OR 0.83, 95% CI 0.74–0.94, p = 2.6 × 10−03), septic shock (OR 0.78, 95% CI 0.64–0.93, p = 8.14 × 10−03), in‐hospital death (OR 0.58, 95% CI 0.42–0.78, p = 4.49 × 10−04), and respiratory failure (OR 0.82, 95% CI 0.65–1.04; p = 0.11) were all markedly attenuated (Table 2). A sensitivity analysis using the HDL‐C level closest to the index hospital admission showed comparable results: in unadjusted analysis, the HDL‐C level measured closest to index hospital admission was significantly associated with all outcomes and after adjustment the strength of the associations was attenuated but remained significant for sepsis, septic shock, and in‐hospital mortality but not respiratory failure (Table S11).
TABLE 2.
Associations between HDL‐C (measured or PRS) and sepsis and sepsis‐related outcomes in White patients
| Predictors | Phenotypes | Unadjusted | Adjusted 1 a | ||
|---|---|---|---|---|---|
| Odds ratio | p value | Odds ratio | p value | ||
| (A) Measured baseline HDL‐C, N = 11,612 | Sepsis | 0.62 (0.56–0.68) | 2.36 × 10−23 | 0.83 (0.74–0.94) | 2.6 × 10−3 |
| Septic shock | 0.58 (0.50–0.68) | 4.12 × 10−12 | 0.78 (0.64–0.93) | 8.14 × 10−3 | |
| Respiratory failure | 0.59 (0.49–0.71) | 2.82 × 10−8 | 0.82 (0.65–1.04) | 0.11 | |
| in‐hospital death | 0.49 (0.39–0.63) | 1.02 × 10−8 | 0.58 (0.42–0.78) | 4.49 × 10−4 | |
| (B) GLGC‐based HDL‐C PRS, N = 12,377 | Sepsis | 0.99 (0.96–1.03) | 0.74 | 0.99 (0.95–1.03) | 0.70 |
| Septic shock | 1.00 (0.95–1.06) | 0.92 | 1.00 (0.95–1.06) | 0.97 | |
| Respiratory failure | 1.04 (0.98–1.12) | 0.21 | 1.04 (0.97–1.11) | 0.25 | |
| in‐hospital death | 1.00 (0.91–1.10) | 0.97 | 0.99 (0.90–1.09) | 0.86 | |
| (C) CETP PRS, N = 12,377 | Sepsis | 1.01 (0.98–1.05) | 0.52 | 1.01 (0.97–1.05) | 0.53 |
| Septic shock | 1.02 (0.96–1.07) | 0.58 | 1.02 (0.96–1.07) | 0.58 | |
| Respiratory failure | 1.05 (0.98–1.12) | 0.16 | 1.04 (0.98–1.11) | 0.18 | |
| in‐hospital death | 1.01 (0.92–1.10) | 0.86 | 1.01 (0.92–1.11) | 0.85 | |
| (D) rs1800777, N = 12,377 | Sepsis | 1.00 (0.87–1.15) | 0.98 | 1.01 (0.87–1.16) | 0.94 |
| Septic shock | 0.97 (0.80–1.19) | 0.77 | 0.97 (0.79–1.19) | 0.74 | |
| Respiratory failure | 1.19 (0.93–1.56) | 0.17 | 1.21 (0.94–1.57) | 0.15 | |
| in‐hospital death | 0.93 (0.68–1.31) | 0.66 | 0.94 (0.68–1.34) | 0.73 | |
Abbreviations: BMI, body mass index; EHR, electronic health records; GLGC, Global Lipid Genetics Consortium; HDL‐C, high‐density lipoprotein cholesterol; PCs, principal components for ancestry; PRS, polygenic risk score; TGs, triglycerides.
(A) Measured HDL‐C was adjusted for age, sex, EHR length, comorbidity score, LDL‐C, TGs, and BMI; (B) HDL‐C PRS, (C) CETP PRS, and (D) rs1800777 were adjusted for age, sex, EHR length, and five PCs.
Analyses in Black patients or sensitivity analyses in White patients with both genotype and measured HDL‐C were generally consistent with those in White patients: the unadjusted associations between measured HDL‐C and sepsis, septic shock and in‐hospital death were significant, but these associations were attenuated after adjusting for confounders (Tables S12 and S13).
HDL‐C PRS and sepsis or sepsis‐related outcomes
In White patients, the HDL‐C PRS was significantly associated with measured HDL‐C levels (N = 25,009, r = 0.24, p < 2.2 × 10−16) but not with the risk of sepsis (OR 0.99, 95% CI 0.96–1.03; p = 0.74), septic shock (OR 1.00, 95% CI 0.95–1.06; p = 0.92), respiratory failure (OR 1.04, 95% CI 0.98–1.12; p = 0.21), or in‐hospital mortality (OR 1.00, 95% CI 0.91–1.10; p = 0.97). Results were similar after adjusting for age, sex, EHR length, and five PCs (sepsis: OR 0.99, 95% CI 0.95–1.03, p = 0.70; septic shock: OR 1.00, 95% CI 0.95–1.06, p = 0.97; respiratory failure: OR 1.04, 95% CI 0.97–1.11, p = 0.25; and in‐hospital mortality: OR 0.99, 95% CI 0.90–1.09; p = 0.86; Table 2). Findings from the post hoc sensitivity analyses in 3007 White patients with HDL‐C PRS were consistent with the main analyses (Table S13). Analyses in Black patients were consistent with those in White patients: there was no association between genetically predicted HDL‐C levels and sepsis or sepsis‐related outcomes (Table S12).
For illustrative purposes, we compared the risk of sepsis, severe sepsis or septic shock, respiratory failure, and in‐hospital mortality within different quartiles for measured HDL‐C levels and the HDL‐C PRS in White patients (Figure 2). Patients in the lowest quartile of measured HDL‐C levels had a significantly higher risk of sepsis, septic shock, respiratory failure, and in‐hospital mortality compared to those in the highest quartile. After adjustment, the differences between the lowest and the highest HDL‐C quartiles were partially attenuated (Figure 2a,b). No significant difference in risk of sepsis or other outcomes was observed among HDL‐C PRS quartiles, with or without adjustment (Figure 2c,d).
FIGURE 2.
Association between HDL‐C quartiles and sepsis, septic shock, respiratory failure, and in‐hospital death in White patients. The associations between sepsis and its adverse outcomes and measured HDL‐C quartiles (a, b) and polygenic risk score quartiles (c, d). Analyses a and c are unadjusted, b was adjusted for age, sex, EHR length, comorbidity score, LDL‐C, TGs, and BMI, and d was adjusted for age, sex, EHR length, and PCs. BMI, body mass index; EHR, electronic health record; LCI, lower confidence interval; OR, odds ratio; PCs, principal components for ancestry; PRS, polygenic risk scores; TGs, triglycerides; UCI, upper confidence interval.
CETP genetic variants and sepsis or sepsis‐related outcomes
We further tested the associations between CETP variants that alter HDL‐C levels and sepsis. Both the CETP PRS and a functional variant (rs1800777) were significantly associated with measured HDL‐C (CETP PRS, p < 2.2 × 10−16, r = 0.15, N = 25,009; rs1800777, p < 2.2 × 10−16, r = 0.07, N = 25,009) but there was no significant association between the CETP PRS or rs1800777 and the risk of sepsis or related phenotypes (Table 2). Analyses in Black patients were generally consistent with those in White patients (Table S12). Post hoc analyses in 3007 White patients with both measured HDL‐C levels and HDL‐C PRS found no association between rs1800777 and sepsis. There was an association between the CETP PRS and sepsis in the direction opposite to the hypothesis (OR 1.10, 95% CI 1.02–1.19; p = 0.01; Table S13).
PheWAS of measured HDL‐C, HDL‐C PRS, and CETP PRS
Using data from the de‐identified EHRs (N = 313,478) White patients, median measured HDL‐C levels, were significantly associated with 1032 different clinical phenotypes (p < 2.8 × 10−5; Figure 3 ) For example, lower measured HDL‐C levels were associated with increased risk of septicemia (p < 1 × 10−308), type 2 diabetes (p < 1 × 10−308), obesity(p < 1 × 10−308), sepsis (p < 1 × 10−308), myocardial infarction (p < 1 × 10−308), coronary atherosclerosis (p < 1 × 10−308), chronic kidney disease (p < 1 × 10−308), and increased risk of hyperlipidemia (p < 1 × 10−308). As mentioned above, the HDL‐C PRS was significantly associated with measured HDL‐C levels (N = 25,009, r = 0.24, p < 2.2 × 10−16); it was also weakly associated with BMI (N = 51,000, r = −0.009, p = 0.04). However, the HDL‐C PRS was significantly associated with only seven phecodes (N = 65,592) in PheWAS. Lower predicted HDL‐C by PRS was associated with increased risk of hyperlipidemia (p = 4.7 × 10−20), mixed hyperlipidemia (p = 7.8 × 10−20), hyperglyceridemia (p = 5.2 × 10−15), hypercholesterolemia (p = 6.6 × 10−12), coronary atherosclerosis (p = 2.9 × 10−9), other chronic ischemic heart disease (p = 7.3 × 10−7), and angina pectoris (p = 4.6 × 10−6). All significant associations for the HDL‐C PRS and the top 100 associations for measured HDL‐C levels are shown in Table S14. We also conducted PheWAS analyses for the CETP PRS and rs1800777. The CETP PRS was significantly associated with mixed hyperlipidemia (p = 3.0 × 10−6) and hyperglyceridemia (p = 4.1 × 10−6; Table S14). None of the rs1800777 associations passed the PheWAS multiple correction adjustment.
FIGURE 3.
Phenome‐wide association analyses (PheWAS) of (a) measured HDL‐C and (b) HDL‐C PRS in White patients. Manhattan plots of PheWAS. The PheWAS of measured HDL‐C were adjusted for sex, year of birth, length of EHRs; PheWAS of HDL‐C PRS were adjusted for sex, year of birth, length of EHRs, and five PCs. Red horizontal line designates minimum p value for statistical significance after correction for multiple testing. The blue line represents a suggestive p value of 0.05. Annotation for measured HDL‐C PheWAS a is impossible because there were 1032 significant associated phenotypes. The top 100 hits are shown in Table S10. EHR, electronic health record; PCs, principal components for ancestry; PRS, polygenic risk scores.
DISCUSSION
The current study confirmed that HDL‐C levels measured before hospital admission were strongly associated with the risk of sepsis and sepsis‐related outcomes. However, genetic analyses using HDL‐C PRS, CETP PRS, and rs1800777 were not associated with the risk of sepsis or sepsis‐related outcomes, although they were associated with HDL‐C levels.
HDL represents an attractive therapeutic target for sepsis, and several lines of evidence support this idea. HDL binds and thus neutralizes lipopolysaccharide (also known as endotoxin), a mediator of sepsis, and increases its clearance. 32 , 33 In animal models of sepsis, treatment with reconstituted HDL protected from organ injury and increased survival, 8 , 34 , 35 and in clinical and epidemiologic studies lower HDL‐C levels were a strong predictor for increased risk of infection, worse sepsis outcomes, and higher mortality. 36 However, observational clinical and epidemiologic studies have potential problems, such as confounding, reverse causation, and unaccounted comorbidities.
Concordant with previous clinical reports, we found that higher HDL‐C levels measured before infection were significantly associated with reduced risk of sepsis and other sepsis‐related outcomes. 36 , 37 However, even rudimentary statistical adjustment that included a comorbidity score markedly attenuated the associations. Illustrating the potential for confounding to contribute to the clinical associations with measured HDL‐C levels, the PheWAS for measured HDL‐C was significant for 1032 phenotypes, including an apparent protective effect for many infection‐related diagnoses. Although the sample sizes differed, the PheWAS for both measured HDL (n = 313,478) and the HDL PRS (n = 65,592) were large and had good power to study common clinical phenotypes, as illustrated by the CIs around point estimates. For example, the phenotype septic shock (PheCode 994.21) had 4198 cases and 298,224 controls in the measured HDL‐C group, and there were 2104 cases and 56,334 controls in the HDL PRS group. In a post hoc power calculation, 38 , 39 for septic shock, we had greater than 80% power to detect an HDL difference between cases and controls of 1.4 mg/dl in the measured HDL‐C group and a difference of 2.0 mg/dl in the HDL PRS group. However, although the association between low HDL‐C levels and increased risk of sepsis is likely due to confounding, measured HDL‐C levels could represent a convenient clinically accessible composite measure of underlying comorbidities and state of health and may thus still be useful for prediction of risk in patients.
Genetic approaches akin to MR are powerful tools to reduce confounding and assess the relationships between lipids and clinical phenotypes. This approach previously suggested that the association between HDL‐C and cardiovascular disease (CVD) was confound by TG levels, 40 , 41 and explained why CETP inhibitors successfully increased HDL‐C levels but did not prevent CVD. 42 , 43 Using an LDL‐C PRS and MR, our group and others have reported no causal relationship between LDL‐C and sepsis despite the clinically observed association between low LDL‐C and increased sepsis risk. 14 , 26
Similarly, we found no association between an HDL‐C PRS and the risk of progression from infection to sepsis or its complications. In contrast, using similar approaches in UK biobank (UKBB) data, Trinder et al. recently reported a significant association between an HDL‐C PRS and risk of infection, antibiotic usage, and sepsis mortality. 15 Although they did not specifically address the risk of progression from infection to sepsis, their findings suggest that the relationship between HDL‐C and infection and sepsis could be causal. A few things might account for the different observations. First, Trinder et al. tested the risk of infection within a general population, and we focused on the risk of progression from infection to sepsis in a cohort of patients with infection admitted to a tertiary hospital. Second, although we both used the HDL‐C PRS developed using the GLGC genomewide association study data, 27 , 29 , 44 the subsets of genetic variants differed. We adopted a validated PRS, 27 , 29 and its performance in BioVU (r = 0.24, p < 2.2 × 10−16) was comparable to previous reports (correlation ranged between 0.27 and 0.28). 27 , 29 However, we cannot rule out the possibility that different PRS capture different aspects of the underlying mechanism of HDL regulation and also capture varying degrees of effect on LDL‐C and TG levels. It is possible that HDL quality and function (e.g., anti‐inflammatory and anti‐oxidant effects) are more important than HDL quantity in regulating the progression from infection to sepsis and that this is not captured in the existing genetic associations which were derived from HDL‐C levels. 45 Third, for reasons that are unclear, in UKBB the HDL PRS used was significantly associated with diabetes. Although the apparent protective effects of HDL on infection were still present after statistical adjustment for diabetes, the potential for confounding remains. Other genetic studies have not supported a causal relationship between HDL and diabetes. 46 In addition to diabetes, the HDL PRS was associated with BMI in UKBB, 15 and, in previous MR studies, elevated BMI was causally associated with the increased risk of both infection admission and mortality. 47 , 48 In BioVU, the HDL PRS was not significantly associated with diabetes and only weakly associated with BMI (r = −0.009, p = 0.04); in the PheWAS, only cardiovascular and lipid phenotypes were significantly associated, supporting lack of confounding. Fourth, the definition of sepsis used in the UKBB study 15 differed from the one we used in that it was based entirely on ICD codes and included the code of “septicemia,” a condition not always accompanied by sepsis. Last, UKBB represents a healthier population compared to BioVU. Patients in BioVU had lower HDL‐C levels (1.17 ± 0.46 mmol/L) than those in UKBB, (1.45 ± 0.38 mmol/L) which may lead to greater difficulties in detecting a small effect, such as the one reported in UKBB for HDL and the risk of infection. 15
As mentioned, although CETP inhibitors failed to reduce CVD events in clinical trials, they demonstrated an ability to increase HDL‐C levels and therefore could be considered as potential drugs to treat sepsis. The role of CETP during sepsis remains unclear. 16 In animal models, studies have variously found low 16 and high CETP levels to protect against poor sepsis outcomes. 49 , 50 A recent report using genetic approaches showed that both a CETP functional variant and CETP PRS associated with sepsis mortality. In the current study, however, we found no association between the CETP PRS and sepsis risk or in‐hospital mortality in our primary analyses. The findings in the post hoc sensitivity analysis in the 3007 participants who had both genotype and measured HDL levels were largely concordant with the main analysis. However, there was a nominally significant association between the CETP PRS and sepsis. This association was in the opposite direction (i.e., higher HDL was associated with higher risk) to the proposed hypothesis of higher HDL levels being beneficial. Therefore, we interpret the association as being an artifact of multiple statistical tests. Additional observational, clinical, and functional studies are needed to further elucidate the role of CETP during infection and sepsis. 50 , 51
The current study has several strengths. First, we applied a validated algorithm to identify patients with infection and sepsis from large de‐identified EHRs. Compared to only ICD codes, our approach has higher accuracy and less concern for misclassification. Second, we leveraged a large EHR repository with longitudinal data, which allowed us to extract comorbidity from patients’ charts before the infection admission—something that has seldom been done in other studies. Third, we analyzed both measured and genetically predicted HDL‐C levels to evaluate the causal relationships between HDL‐C and sepsis. Fourth, we also explored other clinical phenotypes associated with both measured and genetically predicted HDL‐C by conducting PheWAS.
There are also limitations. First, whereas both White and Black patients were included in the analysis, the statistical power was limited for Black patients. Second, our cohort was constructed before the COVID‐19 pandemic had a major effect and there were a limited number of patients with COVID‐19. Future analysis of the relationship between HDL‐C and COVID‐19 related illness will be interesting. Third, we cannot rule out the possibility that the HDL PRS did not capture some aspect of HDL function, such as its anti‐inflammatory effects.
In summary, we confirmed the strong association between clinically measured HDL‐C levels and sepsis risk; however, genetic analyses did not support an association between HDL and risk of progression from infection to sepsis or between HDL and sepsis outcomes in patients admitted to the hospital with infection.
AUTHOR CONTRIBUTIONS
G.L., L.J., V.E.K., A.O., A.I., A.L.D., L.L.D., C.S., M.F.L., C.P.C., N.C., W.Q.W., C.M.S., and Q.F. wrote the manuscript. G.L., C.M.S., and Q.F. designed the research. G.L., L.J., V.E.K., A.O., A.I., A.L.D., L.L.D., C.S., C.P.C., W.Q.W., C.M.S., and Q.F. performed the research. G.L., L.J., C.M.S., and Q.F. analyzed the data. C.S. and N.C. contributed to analytical tools.
FUNDING INFORMATION
This study was supported by GM120523 (Q.F.), R01HL163854 (Q.F.), R35GM131770 (C.M.S.), HL133786 (W.Q.W.), P01HL116263 (M.F.L.), 1K01HL157755–01 (V.E.K.), and Vanderbilt Faculty Research Scholar Fund (Q.F.). The dataset(s) used for the analyses described were obtained from Vanderbilt University Medical Center's BioVU, which is supported by institutional funding, the 1S10RR025141–01 instrumentation award, and by the CTSA grant UL1TR0004 from NCATS/NIH. Additional funding provided by the NIH through grants P50GM115305 and U19HL065962. The authors wish to acknowledge the expert technical support of the VANTAGE and VANGARD core facilities, supported in part by the Vanderbilt‐Ingram Cancer Center (P30 CA068485) and Vanderbilt Vision Center (P30 EY08126). Role of the Funder/Sponsor: The funders had no role in design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
CONFLICT OF INTEREST
The authors declared no competing interests for this work.
Supporting information
Appendix S1
Table S1
ACKNOWLEDGMENTS
The first and corresponding authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Liu G, Jiang L, Kerchberger VE, et al. The relationship between high density lipoprotein cholesterol and sepsis: A clinical and genetic approach. Clin Transl Sci. 2023;16:489‐501. doi: 10.1111/cts.13462
REFERENCES
- 1. Cecconi M, Evans L, Levy M, Rhodes A. Sepsis and septic shock. Lancet Lond Engl. 2018;392(10141):75‐87. [DOI] [PubMed] [Google Scholar]
- 2. Hajj J, Blaine N, Salavaci J, Jacoby D. The “centrality of sepsis”: a review on incidence, mortality, and cost of care. Healthcare. 2018;6(3):90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Rhee C, Dantes R, Epstein L, et al. Incidence and trends of sepsis in US hospitals using clinical vs claims data, 2009‐2014. JAMA. 2017;318(13):1241‐1249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Liu V, Escobar GJ, Greene JD, et al. Hospital deaths in patients with sepsis from 2 independent cohorts. JAMA. 2014;312(1):90‐92. [DOI] [PubMed] [Google Scholar]
- 5. Paoli CJ, Reynolds MA, Sinha M, Gitlin M, Crouser E. Epidemiology and costs of sepsis in the United States‐an analysis based on timing of diagnosis and severity level. Crit Care Med. 2018;46(12):1889‐1897. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Quartin AA, Schein RMH, Kett DH, Peduzzi PN. Magnitude and duration of the effect of sepsis on survival. JAMA. 1997;277(13):1058‐1063. [PubMed] [Google Scholar]
- 7. Storgaard M, Hallas J, Gahrn‐Hansen B, Pedersen SS, Pedersen C, Lassen AT. Short‐ and long‐term mortality in patients with community‐acquired severe sepsis and septic shock. Scand J Infect Dis. 2013;45(8):577‐583. [DOI] [PubMed] [Google Scholar]
- 8. Levine DM, Parker TS, Donnelly TM, Walsh A, Rubin AL. In vivo protection against endotoxin by plasma high density lipoprotein. Proc Natl Acad Sci U S A. 1993;90(24):12040‐12044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Guo L, Ai J, Zheng Z, et al. High density lipoprotein protects against Polymicrobe‐induced sepsis in mice. J Biol Chem. 2013;288(25):17947‐17953. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Morin EE, Guo L, Schwendeman A, Li XA. HDL in sepsis ‐ risk factor and therapeutic approach. Front Pharmacol. 2015;6:244. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. van Leeuwen HJ, Heezius ECJM, Dallinga GM, van Strijp JAG, Verhoef J, van Kessel KPM. Lipoprotein metabolism in patients with severe sepsis. Crit Care Med. 2003;31(5):1359‐1366. [DOI] [PubMed] [Google Scholar]
- 12. Lekkou A, Mouzaki A, Siagris D, Ravani I, Gogos CA. Serum lipid profile, cytokine production, and clinical outcome in patients with severe sepsis. J Crit Care. 2014;29(5):723‐727. [DOI] [PubMed] [Google Scholar]
- 13. Cirstea M, Walley KR, Russell JA, Brunham LR, Genga KR, Boyd JH. Decreased high‐density lipoprotein cholesterol level is an early prognostic marker for organ dysfunction and death in patients with suspected sepsis. J Crit Care. 2017;1(38):289‐294. [DOI] [PubMed] [Google Scholar]
- 14. Feng Q, Wei WQ, Chaugai S, et al. Association between low‐density lipoprotein cholesterol levels and risk for sepsis among patients admitted to the hospital with infection. JAMA Netw Open. 2019;2(1):e187223. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Mark T, Walley KR, Boyd JH, Brunham LR. Causal inference for genetically determined levels of high‐density lipoprotein cholesterol and risk of infectious disease. Arterioscler Thromb Vasc Biol. 2020;40(1):267‐278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Trinder M, Wang Y, Madsen CM, et al. Inhibition of cholesteryl Ester transfer protein preserves high‐density lipoprotein cholesterol and improves survival in sepsis. Circulation. 2021;143:921‐934. [DOI] [PubMed] [Google Scholar]
- 17. Feng Q, Wei WQ, Chaugai S, et al. A genetic approach to the association between PCSK9 and sepsis. JAMA Netw Open. 2019;2(9):e1911130. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Angus DC, Linde‐Zwirble WT, Lidicker J, Clermont G, Carcillo J, Pinsky MR. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit Care Med. 2001;29(7):1303‐1310. [DOI] [PubMed] [Google Scholar]
- 19. Donnelly JP, Safford MM, Shapiro NI, Baddley JW, Wang HE. Application of the third international consensus definitions for sepsis (Sepsis‐3) classification: a retrospective population‐based cohort study. Lancet Infect Dis. 2017;17(6):661‐670. Available from: http://www.sciencedirect.com/science/article/pii/S1473309917301172. [Accessed April 26, 2017] [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Seymour CW, Liu VX, Iwashyna TJ, et al. Assessment of clinical criteria for sepsis: for the third international consensus definitions for sepsis and septic shock (Sepsis‐3). JAMA. 2016. Feb 23;315(8):762‐774. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. MDCalc . Charlson Comorbidity Index (CCI) [Internet]. Available from: https://www.mdcalc.com/charlson‐comorbidity‐index‐cci. [Accessed May 7, 2019]
- 22. Charlson/Deyo Score | National Cancer Data Base ‐ Data Dictionary PUF 2013. Available from: http://ncdbpuf2013.facs.org/content/charlsondeyo‐comorbidity‐index. [Accessed May 6, 2020].
- 23. Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD‐9‐CM administrative databases. J Clin Epidemiol. 1992;45(6):613‐619. [DOI] [PubMed] [Google Scholar]
- 24. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373‐383. [DOI] [PubMed] [Google Scholar]
- 25. Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD‐9‐CM and ICD‐10 administrative data. Med Care. 2005;43(11):1130‐1139. [DOI] [PubMed] [Google Scholar]
- 26. Walley KR, Boyd JH, Kong HJ, Russell JA. Low low‐density lipoprotein levels are associated with, but Do not causally contribute to, increased mortality in sepsis. Crit Care Med. 2019;47(3):463‐466. [DOI] [PubMed] [Google Scholar]
- 27. Lambert SA, Gil L, Jupp S, et al. The polygenic score catalog as an open database for reproducibility and systematic evaluation. Nat Genet. 2021;53(4):420‐425. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Do R, Willer CJ, Schmidt EM, et al. Common variants associated with plasma triglycerides and risk for coronary artery disease. Nat Genet. 2013;45(11):1345‐1352. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Kuchenbaecker K, Telkar N, Reiker T, et al. The transferability of lipid loci across African, Asian and European cohorts. Nat Commun. 2019;10(1):4330. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Wei WQ, Bastarache LA, Carroll RJ, et al. Evaluating phecodes, clinical classification software, and ICD‐9‐CM codes for phenome‐wide association studies in the electronic health record. PLoS One. 2017;12(7):e0175508. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Wu P, Gifford A, Meng X, et al. Mapping ICD‐10 and ICD‐10‐CM codes to Phecodes: workflow development and initial evaluation. JMIR Med Inform. 2019;7(4):e14325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Vesy CJ, Kitchens RL, Wolfbauer G, Albers JJ, Munford RS. Lipopolysaccharide‐binding protein and phospholipid transfer protein release lipopolysaccharides from gram‐negative bacterial membranes. Infect Immun. 2000;68:2410‐2417. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Wurfel MM, Kunitake ST, Lichenstein H, Kane JP, Wright SD. Lipopolysaccharide (LPS)‐binding protein is carried on lipoproteins and acts as a cofactor in the neutralization of LPS. J Exp Med. 1994;180(3):1025‐1035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Tanaka S, Genève C, Zappella N, et al. Reconstituted high‐density lipoprotein therapy improves survival in mouse models of sepsis. Anesthesiology. 2020;132(4):825‐838. [DOI] [PubMed] [Google Scholar]
- 35. McDonald MC, Dhadly P, Cockerill GW, et al. Reconstituted high‐density lipoprotein attenuates organ injury and adhesion molecule expression in a rodent model of Endotoxic shock. Shock. 2003;20(6):551‐557. [DOI] [PubMed] [Google Scholar]
- 36. Madsen CM, Varbo A, Tybjærg‐Hansen A, Frikke‐Schmidt R. Nordestgaard BG. U‐shaped relationship of HDL and risk of infectious disease: two prospective population‐based cohort studies. Eur Heart J. 2018;39(14):1181‐1190. Available from: https://academic.oup.com/eurheartj/advance‐article/doi/10.1093/eurheartj/ehx665/4710060. [Accessed December 19, 2017]. [DOI] [PubMed] [Google Scholar]
- 37. Scalsky RJ, Chen YJ, Desai K, O'Connell JR, Perry JA, Hong CC. Baseline cardiometabolic profiles and SARS‐CoV‐2 infection in the UK biobank. PLoS One. 2021;16(4):e0248602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Dupont WD, Plummer WD. Power and sample size calculations. A review and computer program. Control Clin Trials. 1990;11(2):116‐128. [DOI] [PubMed] [Google Scholar]
- 39. Dupont WD, Plummer WD. Power and sample size calculations for studies involving linear regression. Control Clin Trials. 1998;19(6):589‐601. [DOI] [PubMed] [Google Scholar]
- 40. Holmes MV, Asselbergs FW, Palmer TM, et al. Mendelian randomization of blood lipids for coronary heart disease. Eur Heart J. 2015;36(9):539‐550. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. White J, Swerdlow DI, Preiss D, et al. Association of Lipid Fractions with Risks for coronary artery disease and diabetes. JAMA Cardiol. 2016;1:692‐699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Barter PJ, Caulfield M, Eriksson M, et al. Effects of Torcetrapib in patients at high risk for coronary events. N Engl J Med. 2007;357(21):2109‐2122. [DOI] [PubMed] [Google Scholar]
- 43. Armitage J, Holmes MV, Preiss D. Cholesteryl Ester transfer protein inhibition for preventing cardiovascular events: JACC review topic of the week. J Am Coll Cardiol. 2019;73(4):477‐487. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Klarin D, Damrauer SM, Cho K, et al. Genetics of blood lipids among ~300,000 multi‐ethnic participants of the million veteran program. Nat Genet. 2018. Nov;50(11):1514‐1523. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Barker G, Leeuwenburgh C, Brusko T, Moldawer L, Reddy ST, Guirgis FW. Lipid and lipoprotein dysregulation in sepsis: clinical and mechanistic insights into chronic critical illness. J Clin Med. 2021;10(8):1693. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Haase CL, Tybjærg‐Hansen A, Nordestgaard BG, Frikke‐Schmidt R. HDL cholesterol and risk of type 2 diabetes: a Mendelian randomization study. Diabetes. 2015;64(9):3328‐3333. [DOI] [PubMed] [Google Scholar]
- 47. Butler‐Laporte G, Harroud A, Forgetta V, Richards JB. Elevated body mass index is associated with an increased risk of infectious disease admissions and mortality: a mendelian randomization study. Clin Microbiol Infect. 2021;27(5):710‐716. [DOI] [PubMed] [Google Scholar]
- 48. Rogne T, Solligård E, Burgess S, et al. Body mass index and risk of dying from a bloodstream infection: a Mendelian randomization study. PLoS Med. 2020;17(11):e1003413. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Venancio TM, Machado RM, Castoldi A, et al. CETP lowers TLR4 expression which attenuates the inflammatory response induced by LPS and Polymicrobial sepsis. Mediat Inflamm. 2016;2016:1784014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Quintão ECR, Cazita PM. Letter by Quintão and Cazita regarding article, “inhibition of cholesteryl Ester transfer protein preserves high‐density lipoprotein cholesterol and improves survival in sepsis”. Circulation. 2021;144(6):e120‐e121. [DOI] [PubMed] [Google Scholar]
- 51. Brunham LR, Trinder M, Rensen PCN, Boyd J. Response by Brunham et al to letter regarding article, "inhibition of cholesteryl Ester transfer protein preserves high‐density lipoprotein cholesterol and improves survival in sepsis". Circulation. 2021;144(6):e122. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Appendix S1
Table S1