The Association of the Levels of High-Density Lipoprotein and Apolipoprotein A1 with SARS-CoV-2 Infection and COVID-19 Severity: An Analysis of the N3C Database

Simple Summary This study analyzed data from the National COVID Cohort Collaborative, the largest US COVID-19 database, to determine if an association exists between COVID-19 and blood levels of high-density lipoprotein, and high-density lipoprotein’s main protein component, apolipoprotein A1. Our findings suggest that individuals with higher levels of high-density lipoprotein and apolipoprotein A1 were less likely to be infected with SARS-CoV-2 and were less likely to develop severe COVID-19 requiring hospitalization or invasive medical interventions. They were also less likely to develop acute kidney injury. On the other hand, the presence of underlying health issues (comorbidities) increased the risk of developing severe COVID-19 and acute kidney injury. African American and Hispanic populations were more likely to experience severe disease, while smoking and being male appeared to reduce the incidence of infection but increased the risk of developing severe disease and acute kidney injury. This study is the first to explore the association of high-density lipoprotein and apolipoprotein A1 with COVID-19 outcomes using US population data. Abstract This study analyzed data from the National COVID Cohort Collaborative (N3C) database to investigate whether high-density lipoprotein (HDL) and its major protein component, apolipoprotein A1 (apoA1), are associated with severe COVID-19 sequelae, specifically acute kidney injury (AKI) and severe COVID-19 disease as defined by the infection resulting in hospitalization, extracorporeal membrane oxygenation (ECMO), invasive ventilation, or death. Our study included a total of 1,415,302 subjects with HDL values and 3589 subjects with apoA1 values. Higher levels of both HDL and apoA1 were associated with a lower incidence of infection as well as a lower incidence of severe disease. Higher HDL levels were also associated with a lower incidence of developing AKI. Most comorbidities were negatively correlated with SARS-CoV-2 infection, presumably due to the behavioral changes that occurred as a result of the precautions taken by individuals with underlying comorbidities. The presence of comorbidities, however, was associated with developing severe COVID-19 disease and AKI. African American and Hispanic populations experienced worse outcomes, including a higher incidence of infection and the development of severe disease, as well as AKI. Smoking and being male were associated with a lower incidence of infection, while they were risk factors for the development of severe disease and AKI. The results on cholesterol and diabetes drugs warrant further research, given that the database included multiple drugs in each category impeding for analysis of specific medications. Despite the current limitations in the N3C data, this study is the first to investigate the roles of HDL and apoA1 on the outcomes of COVID-19 using the US population data.


Data and Data Sources
The access to the N3C database was approved by the investigators' IRB Review Board on 6 September 2022 under #1936197-2: The study of lipoproteins and their components in SARS-CoV-2 (N3C DUR ID: DUR-44A1395). For the current study, we extracted the data for the period between January 2018 and November 2020. The data after November 2020 were not used to avoid confounding with the COVID-19 vaccination effects.
As of 21 March 2023, the N3C database encompassed 18,416,222 subjects. Of those, 5,017,750 received COVID-19 tests before 30 November 2020. Of those, we focused on the subjects whose medical records prior to their COVID-19 tests included either HDL or apoA1 values. There were 1,415,302 subjects with an HDL value and 3589 subjects with an apoA1 value. The subject data were then merged with data on comorbidities and drug use. Of the subjects who had an HDL value recorded, 18.36% (n = 259,905) were diagnosed with COVID-19. Of the subjects with an apoA1 value recorded, 17.86% (n = 641) were diagnosed with COVID-19. The PRISMA flow diagram summarizing the data extraction process is shown in Figure 1.

Population and Variables
The three outcome variables were considered, including (i) SARS-CoV-2 infection; (ii) severe COVID-19 outcome, i.e., the infection involving hospitalization, and extracorporeal membrane oxygenation (ECMO), invasive ventilation, or death; and (iii) the infection resulting in an acute kidney injury (AKI) requiring dialysis. HDL and apoA1 values recorded prior to the (positive or negative) COVID-19 tests were the two key variables tested for associations with the three outcome measures. The duration between HDL/apoA1 and COVID-19 tests varied by subject. The average duration between HDL and COVID-19 tests was 546 days (SD = 309), while the average duration between apoA1 and COVID-19 tests was 402 days (SD = 318).
A host of demographic, clinical, and other subject-level characteristics were the covariates for the infection as well as the outcomes of COVID-19. Demographic variables included age, sex, and race, while clinical variables were comorbidities known prior to the COVID-19 test, including obesity, diabetes complicated, diabetes not complicated, hypertension, a history of cerebrovascular, heart or lung disease, as well as end-stage kidney disease (ESRD). Other known or suspected risk factors for SARS-CoV-2 infection and

Population and Variables
The three outcome variables were considered, including (i) SARS-CoV-2 infection; (ii) severe COVID-19 outcome, i.e., the infection involving hospitalization, and extracorporeal membrane oxygenation (ECMO), invasive ventilation, or death; and (iii) the infection resulting in an acute kidney injury (AKI) requiring dialysis. HDL and apoA1 values recorded prior to the (positive or negative) COVID-19 tests were the two key variables tested for associations with the three outcome measures. The duration between HDL/apoA1 and COVID-19 tests varied by subject. The average duration between HDL and COVID-19 tests was 546 days (SD = 309), while the average duration between apoA1 and COVID-19 tests was 402 days (SD = 318).
A host of demographic, clinical, and other subject-level characteristics were the covariates for the infection as well as the outcomes of COVID-19. Demographic variables included age, sex, and race, while clinical variables were comorbidities known prior to the COVID-19 test, including obesity, diabetes complicated, diabetes not complicated, hypertension, a history of cerebrovascular, heart or lung disease, as well as end-stage kidney disease (ESRD). Other known or suspected risk factors for SARS-CoV-2 infection and outcomes of COVID-19 included smoking, pregnancy, dementia, depression, hemiplegia, HIV, and a history of a transplant. All variables used in this study, along with their corresponding measures, are listed in Appendix A.

Statistical Analysis
Demographic, clinical, and other relevant characteristics were statistically compared between those individuals who: (i) were infected or not infected with SARS-CoV-2; (ii) experienced severe or non-severe COVID-19 as defined above; and (iii) experienced AKI or no AKI as a result of COVID-19. These characteristics were compared in a bivariate fashion using either a t-test or Wilcoxon rank-sum test for continuous variables and Chisq. or Fisher's exact test for categorical variables, depending on the sample size and the distribution of the variable tested.
Three multivariable logistic regressions were performed: first, to investigate the risk factors for the SARS-CoV-2 infection, second to investigate the risk factors for the COVID-19 disease severity, and third to investigate the risk factors for developing AKI as a result of COVID-19. The key independent variables in these regressions were HDL and apoA1 values. Since the values of HDL and apoA1 were highly correlated (ρ = 0.70, p < 0.001), two separate regressions were run, one with HDL and another with apoA1 values, to address the multicollinearity between the two variables. Running two separate regressions also allowed us to retain more observations in the HDL regression since a large proportion of the subjects only had their HDL values recorded. A total of six logistic regressions were run, two regressions per dependent variable. All statistical analyses were performed using R. Statistical significance was defined by p ≤ 0.05 in the analysis unless it is specified otherwise. Following the N3C user agreement, all statistics involving less than 20 observations were presented as "<20" and the associated statistics, such as percentages, were also evaluated at n = 20 and were presented accordingly.

SARS-CoV-2 Infection
Tables 1 and 2 summarize the results of the bivariate statistical analyses for SARS-CoV-2 infection for samples in which HDL or apoA1 values were measured (HDL group and apoA1 group). Both HDL and apoA1 values were higher in the subjects who tested negative for COVID-19 (54 mg/dL vs. 51 mg/dL with p < 0.001 for the HDL group and 144 mg/dL vs. 139 mg/dL with p = 0.002 for the apoA1 group). Older age was negatively correlated with the infection (56 y/o vs. 54 y/o, p < 0.001) for the HDL group, while it was not significant for the apoA1 group (53.6 y/o vs. 53.0 y/o, p = 0.36), the latter result possibly obtained because of the smaller sample size and biased sampling of age groups in the apoA1 group. Sex did not correlate with the risk of infection (57% in both groups, p = 0.18) in the HDL group, while females had a higher risk of getting infected in the apoA1 group (46% vs. 52%, p of.01) possibly due to a selection bias that may be inherent in the apoaA1 group. Hispanics and Whites were more likely to get infected (4.8% vs. 6.6% for the HDL group, 5.2% vs. 5.5% for the apoA1 group in Hispanics; 69.5% 71.0% for the HDL group, and 66% vs. 71% for the apoA1 group in Whites; all p < 0.001 for the HDL group and p = 0.002 for the apoA1 group), while Asians were less likely to get infected (3.0% vs. 1.77% for the HDL group, and 4% vs. <3% for the apoA1 group, all p < 0.001 for the HDL group and p = 0.002 for the apoA1 group).  Three comorbidities were negatively correlated with the infection in both groups, including obesity (53% vs. 46% for the HDL group and 53% vs. 44% for the apoA1 group; both p < 0.001); chronic lung disease (29% vs. 20% for the HDL group, and 35% vs. 27% for the apoA1 group; both p < 0.001); HIV (1.2% vs. 0.7% for the HDL group, and 5.9% vs. <3.12% for the apoA1 group; both p < 0.001), and depression (29% vs. 21% for the HDL group, and 40% vs. 33% for the apoA1 group; both p < 0.001). In addition, complicated diabetes was negatively associated with infection for the HDL group (16% vs. 15%, p < 0.001), while uncomplicated diabetes was positively associated with infection in both groups (26.1% vs. 26.5% for the HDL group, and 28% vs. 36% for the apoA1 group; both p < 0.001). Most of the other comorbidities, including hypertension (57% vs. 50%), cerebrovascular disease (12% vs. 7%), congestive heart failure (12% vs. 8%), heart failure (15% vs. 11%), hemiplegia (3% vs. 2%), ESRD (2% vs. 1%) and transplant (1% vs. 0.6%), were negatively correlated with infection for the HDL group (p < 0.001), but not for the apoA1 group (p > 0.05). Smoking was consistently negatively associated with infection (10% vs. 4% for the HDL group and 22% vs. 8% for the apoA1 group; both p < 0.001). Pregnancy was negatively associated with infection for the HDL group (4.2% vs. 3.9%, p < 0.001), while it was not significant for the apoA1 group (p = 0.31). Finally, the use of cholesterol drugs was negatively associated with infection (10% vs. 5% for the HDL group and 11% vs. 4% for the apoA1 group; both p < 0.001), and the use of diabetes drugs was positively associated with infection (5% vs. 6% for the HDL group, and 5% vs. 9% for the apoA1 group; both p < 0.001).
Tables 3 and 4 present the results of the multivariable logistic regressions for the infection using HDL and apoA1 groups, respectively. The results were, overall, consistent with the findings of the bivariate analyses. After adjusting for covariates, high HDL reduced the risk of contracting the disease (OR = 0.98, p < 0.001). High values of apoA1 were also negatively associated with infection (OR = 0.99, p < 0.001). Chronic lung disease (OR = 0.78 with p < 0.001 for the HDL group, and OR = 0.76 with p = 0.011 for the apoA1 group), obesity (OR = 0.68 for the HDL group, and OR = 0.58 for the apoA1 group; both p < 0.001) and HIV (OR = 0.65 for the HDL group, and OR = 0.28 for the apoA1 group; both p < 0.001) significantly reduced the risk of infection, while uncomplicated diabetes was a risk factor for infection (OR = 1.24 with p < 0.001 for the HDL group, and OR = 1.57 with p = 0.001 for the apoA1 group) in both groups.   In the regression, hypertension was the only comorbidity with contradicting results between the two groups. For the HDL group, hypertension reduced the risk of the infection by 9% (OR = 0.91, p < 0.001), while it increased it in the apoA1 group (OR = 1.25, p = 0.04). Most of the other comorbidities (cardiovascular disease, congestive heart failure, dementia, depression, heart failure, and diabetes complicated) were not statistically significant for the apoA1 group (p > 0.05), but they were negatively correlated with the infection for the HDL group (all p < 0.001).
Females had a higher risk of infection (OR = 1.21 for the HDL group and OR = 1.39 for the apoA1 group; both p < 0.001). Asians had a lower risk of infection than Whites in both groups (OR = 0.45 in the HDL group, and OR = 0.24 for the apoA1 group; both p < 0.001) while African Americans (OR = 1.03, p < 0.001) and Hispanics (OR = 1.19, p < 0.001) had a greater risk of infection in the HDL group. Older age reduced the risk in both groups (OR = 0.997 with p < 0.001 for the HDL group, and OR = 0.99 with p = 0.04 for the apoA1 group). Pregnancy reduced the infection risk (OR = 0.82, p < 0.001) only in the HDL group, while smoking reduced the risk in both groups (OR = 0.39 for the HDL group, and OR = 0.32 for the apoA1 group; both p < 0.001). Finally, the use of statins decreased the odds by 51% for the HDL group (OR = 0.49, p < 0.001) and by 80% for the apoA1 group (OR = 0.20, p < 0.001), while diabetes drugs increased the odds by 35% for the HDL group (OR = 1.35, p < 0.001) and by more than 2.5 times for the apoA1 group (OR = 2.55, p = 0.001).

SAR-CoV-2 Severity
Tables 5 and 6 summarize the results of the bivariate statistical analyses for SARS-CoV-2 severity. Both HDL (52 mg/dL vs. 48 mg/dL, p < 0.001) and apoA1 (142 mg/dL vs. 135 mg/dL, p = 0.01) were negatively associated with the severity. Older age was positively associated with the severity in both groups (51 y/o vs. 61 y/o in the HDL group and 51 y/o vs. 55 y/o for the apoA1 group; both p < 0.001). Non-Whites, particularly African Americans, were more likely to have severe disease (11% vs. 19% in the HDL group and 11% vs. 20% for the apoA1 group; both p < 0.001). In contrast to the infection case, being female reduced the risk of having a severe disease in both groups (58% vs. 53% with p < 0.001 for the HDL group and 55% vs. 47% with p = 0.06 for the apoA1 group, which is significant only at the 10% level). In contrast to what was observed for infection, most comorbidities were risk factors for severity in both groups (p = 0.03 for obesity for the apoA1 group and p < 0.001 for all other statistically significant comorbidities). The only non-significant comorbidities were HIV (p = 0.52) and dementia (p = 0.14) for the apoA1 group, presumably due to the small sample sizes (both n < 20). Pregnancy was negatively correlated with severe disease (4.14% vs. 3.37%, p < 0.001) for the HDL group, but it was not statistically significant for the apoA1 group (p = 0.16). Conversely, smoking was a risk factor for severe disease for the HDL group (2.66% vs. 5.38%, p < 0.001), while it was not statistically significant for the apoA1 group (p = 0.210). Finally, the use of both cholesterol (3.27% vs. 8.82%, with p < 0.001 for HDL, <5% vs. <8% with p = 0.012 for the apoA1 group) and diabetes (4.23% vs. 8.82% with p < 0.001 for the HDL group, and 6.38% vs. 12.05% with p = 0.012 for the apoA1 group) drugs increased the risk of severe COVID-19.   Tables 7 and 8 present the results of the multivariable logistic regressions. Both high HDL and apoA1 values reduced the odds of severe COVID-19 (OR = 0.99 with p < 0.001 for HDL, and OR = 0.99 with p = 0.075 for apoA1 values), although apoA1 was statistically significant only at 10%. Non-Whites were more likely to experience severe disease in the HDL group (ORs ranging between 1.52, Asian, and 2.02, African American, all p < 0.001), although African Americans were the only group with a higher risk in the apoA1 group (OR = 1.74, p = 0.04). Older age and being male were risk factors (OR = 1.03 and OR = 0.88, respectively; both p < 0.001) for the HDL group. For the apoA1 group, both age and sex were statistically non-significant (OR = 1.00 with p = 0.59 and OR = 0.81 with p = 0.30, respectively). Similarly, both smoking and pregnancy (OR = 1.33 and OR = 1.76, respectively; both p < 0.001) were risk factors for the HDL group, while they were not statistically significant for the apoA1 group (OR = 1.00 with p = 0.43 and OR = 0.81 with p = 0. 80, respectively).
Among the comorbidities, the following variables were risk factors in both groups: cerebrovascular disease (OR = 1.36 with p < 0.001 in the HDL group, and OR = 2.12 with p = 0.05 for the apoA1 group); diabetes complicated (OR = 1.25 with p < 0.001 in the HDL group, and OR = 2.40 with p = 0.01 for the apoA1 group); and hemiplegia (OR = 1.78 with p < 0.001 for the HDL group, and OR = 9.34 with p = 0.01 for the apoA1 group). All other comorbidities were risk factors only for the HDL group (ORs ranging between 1.03, being obese, and 3.13, ESRD, all p < 0.001). Finally, the use of cholesterol drugs increased the odds of contracting the severe disease by 31% (OR = 1.31, p < 0.001), while the use of diabetes drugs increased the odds of contracting the severe disease by 12% (OR = 1.12, p < 0.001) for the HDL group. For the apoA1 group, the use of a cholesterol drug was significant at the 10% level and was found to increase the odds of contracting the severe disease by 137% (OR = 2.37, p = 0.08).

Development of AKI
Tables 9 and 10 summarize the results of the bivariate statistical analysis for the development of AKI as a result of COVID-19. Overall, the findings were similar to those of the severity analysis. Higher HDL and apoA1 values reduced the likelihood of developing AKI (51 mg/dL vs. 46 mg/dL with p < 0.001 and 140 mg/dL vs. 135 mg/dL with p = 0.085, respectively), although apoA1 values were statistically significant only at the 10% level. Older age was positively associated with AKI in both groups (53 y/o vs. 66 y/o with p < 0.001 for the HDL group and 52 y/o vs. 58 y/o with p = 0.005 for the apoA1 group). African Americans were more likely to develop AKI (12% vs. 26%, p < 0.001) for the HDL group, while females were less likely to develop AKI in both groups (58% vs. 45% with p < 0.001 for the HDL group, and 53% vs. 41% with p = 0.06 for the apoA1 group) although it was statistically significant only at the 10% level for the apoA1 group. Almost all comorbidities were risk factors for AKI in both groups (p = 0.002 for obesity in apoA1 and p < 0.001 for all other comorbidities in both groups). The only insignificant comorbidities were HIV for the apoA1 group (p = 0.263), most likely due to the small number of HIV patients in the group (n < 20). Pregnancy was negatively associated with AKI (4.10% vs. 1.37%, p < 0.001) for the HDL group, but it was not statistically significant for the apoA1 group (p = 0.10). Smoking was a risk factor for AKI in both groups (3.14% vs. 8.11% for the HDL group and 6.59% vs. 31.25% for the apoA1 group; both p < 0.001). The use of cholesterol and diabetes drugs was a risk factor only for the HDL group (4% vs. 12% for cholesterol drugs and 5% vs. 12% for diabetes drugs; both p < 0.001).   Tables 11 and 12 present the results of the multivariable logistic regressions. Higher HDL values reduced the odds of AKI (OR = 0.99, p < 0.001), while apoA1 values were not statistically significant (OR = 1.00, p = 0.87). Non-whites, regardless of the racial group, were more likely to develop AKI in the HDL group (ORs ranging between 1.34, Hispanic, and 2.05, African American, all p < 0.001). However, for the apoA1 group, Hispanics were more likely to develop AKI (OR = 3.13, p = 0.05). Older age was a risk factor (OR = 1.03, p < 0.001) for the HDL group, while it was not statistically significant for the apoA1 group (OR = 1.00 with p = 0.99). Similarly, females were less likely to develop AKI (OR = 0.64, p < 0.001), while sex was not statistically significant for the apoA1 group (OR = 0.70, p = 0.31). Smoking was a risk factor in both groups (OR = 1.36 with p < 0.001 for the HDL group, and OR = 2.31 with p = 0.068 for the apoA1 group), although it was statistically significant only at the 10% level for the apoA1 group.  Among the comorbidities, the following variables were statistically significant in both groups: cerebrovascular disease (OR = 1.33 with p < 0.001 for the HDL group and OR = 2.35 with p = 0.05 for the apoA1 group); and hypertension (OR = 3.35 with p < 0.001 for the HDL group, and OR = 3.26 with p = 0.02 for the apoA1 group). All other comorbidities were statistically significant only for the HDL group (ORs ranging between 1.14, diabetes uncomplicated, and 2.13, diabetes complicated; both p < 0.001). The use of diabetes drugs reduced the risk of AKI by 8% (OR = 0.92, p = 0.01) in the HDL group, although it was not statistically significant for the apoA1 group (OR = 1.00, p = 0.33). The use of a cholesterol drug was not statistically significant in either group (OR = 1.02 with p = 0.46 for the HDL group, and OR = 0.70 with p = 0.87 for the apoA1 group).

Discussion
The current study investigated the relationship of HDL and apoA1 with three different outcomes related to COVID-19, including SARS-CoV-2 infection, the severity of COVID-19, and the development of AKI as a result of COVID-19 in a single study. We investigated these associations using the N3C data, which constitutes the largest COVID-19 database containing data on individuals who were tested for COVID-19 in a clinical setting in the United States. Overall, our results were consistent with those of previous studies, suggesting that higher HDL and apoA1 levels protect against SARS-CoV-2 infection [5,6] as well as severe outcomes of COVID-19 [6,11,12]. More specifically, our results indicated that an increase in the values of HDL and apoA1 reduces the odds of the infection by 2% and 1%, respectively, after adjusting for demographic covariates and comorbidities. Higher values of HDL and apoA1 also reduced the odds of severe COVID-19 by 1%, although the impact of apoA1 was statistically significant only at 0.5 < p ≤ 0.1, likely due to the small sample size.
Our study also demonstrated that higher values of HDL were protective against the development of AKI. An increase in HDL levels reduced the odds of developing AKI by 1%, adjusting for covariates. Previous studies have shown that high total cholesterol and LDL levels are associated with accelerated progress toward ESRD, while the anti-thrombotic and anti-inflammatory properties of HDL may protect renal arteries and thus reduce the incidence of kidney damage [14,15]. Furthermore, higher HDL levels were shown to protect patients with sepsis or those following heart and vascular surgery from developing AKI [20]. Our results were consistent with these studies, as well as with in vivo studies of endotoxininduced renal injury in rats [21,22]. In our study, however, apoA1 was not found to be protective against AKI. While this result may be due to the sample size for the apoA1 group, it may also be attributable to inherent selection bias. Specifically, the patient group in which the apoA1 values were available may, on average, have had more severe cases, and if so, the results could be explained by the fact that patients with severe COVID-19, were treated with steroids, among other drugs. Steroids are known to have a profound effect on immune function as well as on normal physiological processes, including causing an increase in both cholesterol and triglyceride levels, in addition to resulting in sodium and water retention and increase blood pressure [23,24]. Steroid use has been shown to increase LDL and HDL, as well as apoA1 levels in patients with autoimmune disease [25], and short-term steroid use may increase the levels of lecithin cholesterol acetyl transferase (LCAT) and increase levels of HDL2 [26]. As such, the protective effect of apoA1 may simply be muted by the overwhelming stress of SARS-CoV-2 on the body and the potent physiological impact of treatment modalities such as corticosteroids on HDL subpopulations in this group.
It should be noted that HDL is a particle with a significant degree of heterogeneity in size, shape, and composition of associated apoproteins [27]. Moreover, a cholesterolbased test for HDL is unlikely to capture the complexity of the HDL particle and may miss the functional variability that is thought to arise from compositional differences of heterogeneous HDL populations [17,18]. Thus, although total plasma HDL levels are reported as a single number, this number may reflect the presence of varying percentages of the different subfractions in individuals. Additionally, these subfractions are likely to have different biological effects and may be impacted by external factors.
Moreover, although apoA1 and HDL cholesterol levels seem to have a similar association with the risk of cerebrovascular disease, when the levels of LDL cholesterol and triglycerides are added to the model, the associations can be impacted [27,28]. Such confounding effects have been seen in previous studies. In the Multi-Ethnic Study of Atherosclerosis (MESA) study, HDL particle number was associated with a lower risk of coronary heart disease [29], while the Dallas Heart Study and the Women's Health Study showed that HDL-cholesterol levels and HDL particle numbers were less correlated [30,31]. The observed differences in the protective effect of HDL in the N3C HDL population and the subpopulation with apoA1 measurements may thus reflect a more nuanced impact that the various subfractions of HDL are having on AKI and COVID-19.
Our study found that the presence of comorbidities, in general, reduced the odds of infection while it increased the odds of severe COVID-19 and AKI. Chronic lung disease and HIV, in particular, reduced the risk of infection by >20% and >30%, respectively. The protective role of these comorbidities most likely reflects the behavioral changes that occurred among the individuals with these conditions as a result of the lockdown. The priming of the immune system by these conditions may also contribute to lower infection rates. In addition, being infected by HIV may confer some protection against other viruses. Interestingly, antibodies that can cross-neutralize SARS-CoV-2 through binding to the heavily glycosylated spike protein have been found in HIV-infected individuals [10]. The only comorbidity found to be a risk factor for infection was uncomplicated diabetes. Complicated diabetes was, on the other hand, a risk factor for getting severe COVID-19 in both groups and for developing AKI in one group. This result may again reflect that the precautionary behavioral changes were more prevalent among those individuals with complicated diabetes. In prior studies, type 2 diabetes mellitus was reported to increase both susceptibility to infection and severity of COVID-19 [22,[32][33][34][35]. For COVID-19 severity, chronic lung disease and cerebrovascular disease were found to be risk factors. Both of these are well-established population risk factors for severe COVID-19, as reported in multiple large-scale studies [36][37][38]. For AKI, our risk factors were prior transplant and heart failure. An increased risk of AKI as a result of COVID-19 in transplant recipients is documented, including a meta-analysis that recommends accelerated vaccination programs for kidney transplant recipients [39]. Yet, the effectiveness of the COVID-19 vaccine in this immunocompromised cohort has been debated since the vaccination programs started in November 2020 [40][41][42]. Both cardiovascular disease and heart failure have been associated with AKI in previous studies [43]. As such, the physiological parameters that result in post-vaccination changes may impact some patients adversely, although the greater risks associated with getting COVID-19 are likely to outweigh the side effects of immune hyperactivity that are seen in a small number of vaccine recipients.
Previous studies have shown that males have a higher incidence of COVID-19 and develop more severe disease [44]. In our study as well, males were at a higher risk for severe COVID-19 and AKI. Yet, we saw a higher odds ratio for infections in women. This is inconsistent with the findings of other large studies focused on populations outside the US. Although recent studies have shown that both sexes are equally likely to get infected [44,45], the reason for the discrepancy remains unclear. Our results were consistent with the findings of the prior studies reporting that females are less likely to experience severe COVID-19 [44,46], as well as COVID-19-induced AKI [47]. Older age was weakly associated with a lower incidence of infection in the patient population with HDL values, but it was a risk factor for both severe COVID-19 and AKI. Given the collection dates of the data, it is likely that behavioral factors, such as wearing masks and staying indoors, reduced the incidence of SARS-CoV-2 in older subjects, but that once infected, consistent with other studies, the older patients did have more severe sequelae [19,48]. Similarly, smoking was associated with a lower incidence of infection in both HDL and apoA1 groups though it was a risk factor for both severe COVID-19 and AKI in one group. While this appears to be counterintuitive, nicotine may be playing a mechanistic role, and similar observations have been reported previously, warranting further research on the effects of nicotine [49].
It should be noted, however, that the deleterious effects of smoking outweigh any potential protection that it may confer. Pregnancy was found to be protective against infection in the HDL group while being a risk factor for severe COVID-19 in that population.
Prior studies were inconclusive about the effect of pregnancy on the incidence of infection and severity. Studies have suggested that pregnancy increases the risk of contracting COVID-19 due to the weakened immune system [50], while it may or may not increase the odds of developing severe COVID-19 disease. It has also been reported that pregnant women are more likely to be asymptomatic [51], while it has also been reported that pregnancy increases the risk of severe outcomes in COVID-19, especially when infected with the Delta (B.1.617.2) variant of the virus [52]. While it is likely that our null result for the incidence of infection reflects the effect of behavioral changes (greater caution and reduced public exposure) among pregnant women, further research is warranted on the role of pregnancy on the COVID-19 outcomes, especially as the virus continues to evolve. We found minority populations, especially African American and Hispanic populations, are at higher risk for infection, severe COVID-19 and AKI. While there is a general consensus that African American and Hispanic populations had a disproportionately high prevalence, hospitalization rate, and mortality from COVID-19 [29,53], other studies highlighted the differences in COVID-19 severity within ethnic groups. The reasons for these differences, which may be related to immune hyperactivity or other factors, remain to be explained [54].
Finally, our results suggest that using cholesterol-lowering drugs, such as statins and gemfibrozil, reduces the incidence of infection while increasing the odds of developing severe COVID-19. Prior studies have suggested that statins may protect against severe COVID-19 because of their anti-inflammatory and immune-modulatory properties [55][56][57]. However, a study from an Italian hospital reported worse outcomes in COVID-19 patients who are taking statins, calling for caution in accrediting benefits to statin therapy for COVID-19 [58]. The beneficial impact of metformin has also been reported by multiple studies [59][60][61][62][63]. High blood sugar has a profound effect on the immune system, and, as such, it is not surprising that it impacts viral infections, including those with SARS-CoV-2 [64]. Surprisingly, our results suggest that taking diabetes drugs such as metformin increased the risk of both infection and severe COVID-19 while it was protective against AKI, but only in the HDL group. It is currently unclear why there is a discrepancy between our findings and those of others, but it may be that our results are confounded by the underlining comorbidities and thus may not show a statistical benefit from diabetes treatments without further adjustments, such as propensity score matching.
The current study has a few notable limitations. First, although the number of subjects in the N3C database was over 12 million, the sample size was significantly reduced when we restricted the sample to those subjects with complete information on comorbidities as well as data on drug use and lab test results. Moreover, the merging of the data may result in selection bias. HDL heterogeneity may also be a confounding factor. Some of the results, which were inconsistent with prior studies, may at least partially be attributable to the selection bias embedded in the final data or the inherent biological complexity of HDL and its subfractions. Further, N3C data include only those subjects who were tested for COVID-19 in a clinical setting. Thus, those individuals who contracted COVID-19 but did not utilize any healthcare services are automatically excluded. This means that most individuals with very minor and asymptomatic cases of COVID-19 are not captured in the analysis, even though this is true for most, if not all, studies done on this topic. We acknowledge that there is a time lag (up to 1 year) between the time when HDL and apoaA1 values were recorded and the time of COVID-19 testing. Thus, it is possible that HDL and apoaA1 levels changed before the COVID-19 testing. A shorter interval, such as 6 months, may have enhanced the validity of the results, although this would have reduced the sample size significantly, given that a majority of Americans take laboratory tests during their annual physical exams. Lastly, the current study excluded data after November 2020 to avoid the potential confounding effects of COVID-19 vaccinations and subsequent changes in the COVID-19 treatment regimen. As SARS-CoV-2 is constantly mutating to generate new variants, it is possible that the results presented here on the role of HDL and apoA1 may not hold for the newer variants of COVID-19. Continued investigation is needed to better understand the implications of these limitations on our findings.

Conclusions
The current study demonstrated that higher levels of HDL and apoA1 could reduce the risk of SARS-CoV-2 infection as well as the risks of developing severe cases of COVID-19 and AKI. These associations were established using N3C data, the most comprehensive US-based database that houses a broad range of individual-level data related to COVID-19. While the protective roles of HDL and apoA1 have been well established for the infection and the severity of COVID-19, our study is the first to show the role of HDL in preventing AKI.
Despite the aforementioned caveats, our study validated multiple prior findings in a single study, thereby showing the usefulness of a large population-level database. As SARS-CoV-2 is constantly mutating to generate new variants, it is possible that the results presented here on the role of HDL and apoA1 may not hold for the newer variants of COVID-19. Continued investigation is needed to better understand the implications of these limitations on our findings.

Informed Consent Statement: Not applicable.
Data Availability Statement: The National Institute of Health's National COVID Cohort Collaborative (N3C) data used in this study is available upon application at https://covid.cd2h.org/ (accessed on 11 November 2022). whose information is included within the data and the organizations (https://ncats.nih.gov/n3c/ resources/data-contribution/data-transfer-agreement-signatories (accessed on 11 November 2022)) and scientists who have contributed to the ongoing development of this community resource [https://doi.org/10.1093/jamia/ocaa196 (accessed on 11 November 2022)]. For more details on acknowledgments, please refer to Appendix B.

Conflicts of Interest:
The authors declare no conflict of interest.