Usefulness of Novel Atherogenic Lipid Indices for the Evaluation of Metabolic Status Leading to Coronary Heart Disease in a Real-World Survey of the Japanese Population

We evaluated the usefulness of two novel cholesterol-triglyceride subgroup (CTS) indices, CTSqlt and CTSqnt, that potentially reflect the metabolic status regarding risk of coronary heart disease (CHD) using a retrospective longitudinal study of the Japanese general population. We recruited 12,373 individuals from the annual users of our healthcare center. Among them, the first onset of CHD was recorded in 131 individuals between April 2014 and March 2020. The multivariate Cox proportional hazards regression analyses for all normalized lipid indices revealed that the CTSqnt index showed a comparable hazard ratio for the CHD outcome to non-high-density lipoprotein cholesterol (nonHDL-c) and triglycerides. The HR of the CTSqlt index was significantly lower than for CTSqnt, but still comparable to that for low-density lipoprotein cholesterol (LDL-c). In comparison with the other indices, CTSqlt is more sensitive to risk increment while the index value increases. Linear regression analyses for the CTS indices and previously known lipid indices suggest that the CTSqnt and CTSqlt indices reflect the quantity of atherogenic lipoproteins and particle size (quality) of smaller and denser LDLs, respectively. Furthermore, the CTSqnt/HDL-c index can be used as a comprehensive risk indicator that may represent the status of lipid metabolism determined by the CTSqlt and CTSqnt indices and thus may be useful for screening. The CTS indices can be used to evaluate the metabolic status of individuals, which may increase the risk of future CHD.


Introduction
Low-density lipoprotein (LDL) delivers cholesterol and other lipids from the liver, the primary production site of a precursor of LDL, very low-density lipoprotein (VLDL), to the periphery. In contrast, high-density lipoprotein (HDL) transports cholesterol from the periphery to the liver. This transport of lipoproteins between the liver and periphery maintains lipid homeostasis [1,2]. Lipoproteins with other densities, apoprotein types, and lipid composition do not disturb the lipid homeostasis in the physiological condition. However, excess amounts of normal or aberrant lipoproteins disturb the homeostasis, leading to abnormal lipid deposition in the periphery and liver. Cholesterol deposition in the arterial wall is one of the typical features of atherosclerosis. Although atherosclerosis generates through complex mechanisms [3,4], it is obvious that excess amount of LDL-cholesterol (LDL-c) is one of the causes of atherosclerosis and subsequent coronary heart disease (CHD). Statins improves the CHD outcome by decreasing the LDL-c levels. However, the LDL-c levels are only partially reduced and up to 40% of statin-treated patients still develop CHD [5,6]. Furthermore, many CHD patients may not have significantly increased LDL-c levels. Therefore, researchers have investigated other factors that may explain the residual risk of CHD caused by atherosclerosis. Recent studies have focused on blood triglyceride levels, which are derived from TG-rich lipoproteins (TRLs), including VLDL and chylomicrons [6][7][8]. VLDL and chylomicron macromolecules, however, are too big to pass through the arterial endothelium from the blood stream to enter into the arterial wall. Thus, they are not cholesterol sources in the physiological condition, but metabolic disturbances produce smaller TRL molecules, called remnant lipoproteins, that can cross the arterial wall. Moreover, persistence of high blood TG level makes the LDL molecules smaller and denser by the activity of lipases to catabolize TGs. The small dense LDL (sdLDL) thereby produced is easily oxidized to generate the highly harmful aberrant lipoprotein, oxidized LDL [9]. Because both remnant lipoproteins and sdLDL are highly atherogenic, their levels are clinically important to prevent or manage atherosclerotic diseases. However, the methods of laboratory measurement of remnant-cholesterol (remnant-c) and sdLDL-cholesterol (sdLDL-c) levels are not standardized as the routine laboratory tests. Therefore, alternative markers are required to estimate cholesterol levels in the atherogenic lipoproteins.
We engaged in medical check-up programs, including occupational healthcare programs for workers and personal healthcare programs for the general population in Shiga prefecture, Japan, and provided useful information regarding healthcare management and disease prevention. Since cardiovascular diseases provoked by atherosclerosis are leading causes of death and disability, and lead to personal suffering and socioeconomic loss, we focused on encouraging individuals without CHD but with risk factors for CHD to modify their lifestyle to prevent CHD and stroke through our check-up service. Lipid biomarkers are useful, but the ideal biomarker should be easily measured by routine laboratory tests at a low cost. For this purpose, we measured LDL-c, HDL-c, TG, and nonHDL-c levels, in addition to encouragement of smoking cessation, control of blood pressure and blood glucose levels with suitable exercise and diet, and advice to eliminate other risk factors. However, the prediction of future risk using the traditional lipid biomarkers is not always accurate because the biomarkers are components of biomolecules that are in a dynamic balance and vary between individuals. Additionally, these biomarkers are essential for the homeostasis in the physiological condition and are not only biomarkers for atherosclerosis. We aimed to identify biomarkers that reflect lipid metabolism in the pathological condition and can be easily measured. While providing healthcare information to individuals using our service, we identified some patterns on the scatter plot of TG against LDL-c for more than 10,000 individuals in a year. The distribution appeared to be separated into subgroups characterized by two indices calculated with simple formulas and designated as the cholesterol-triglyceride-subgroup (CTS) indices. Herein, we investigated the usefulness of CTS indices for predicting CHD outcome by comparing them with previously known lipid indices in a retrospective longitudinal study, and explored the basis of these indices in lipid metabolism. The aim of this study was to evaluate the usefulness of novel atherogenic indices that potentially reflect the metabolic status of apparently healthy people that increase the risk of atherosclerotic disorders, including CHD. For a more precise comparison among the lipid indices, including traditional lipid parameters (such as TC, LDL-c, HDL-c, and TG), we performed the Box-Cox transformation for each lipid index, to convert skewed distribution into a normal distribution. Then, we compared the HRs per one standard deviation (1 SD) of the indices.

Study Design
This observational study included participants who underwent an annual health checkup at our facility between April 2013 and March 2020. Most participants were residents of the Shiga prefecture or neighboring cities, Japan. The participants in this study were recruited from annual users of our service from April 2019 to March 2020. Then, we retrospectively excluded the participants who already had CHD history in 2013 or who did not have complete baseline data in 2013 as described below. We identified individuals with a history of CHD (disease group; 406 individuals) and without a history of CHD (control group; 15,546 individuals) who used our service between April 2019 and March 2020. The CHD history was checked annually during an interview or through medical records using a questionnaire that collected information on age at onset, current medication use for CHD, and general items (e.g., age, sex, and current or past smoking). We excluded individuals with CHD history or without complete baseline data (including age, sex, smoking status, systolic and diastolic blood pressures, blood glucose levels (fasting blood glucose level and/or HbA1c), and TG, total cholesterol (TC), LDL-c, and HDL-c levels) collected in 2013. In total, 131 individuals were included in the disease group, with CHD onset reported in 2014-2020, and 12,242 individuals in the control group. In this study, we estimated the ability of lipid indices to predict the first onset of CHD that did not prevent patients continuing to use our services.

Measurements
Lipid, blood glucose, and HbA1c levels were measured by routine laboratory tests. LDL-c was measured by the selective solubilization method using MetaboLead LDL-c (Kyowa Medix, Tokyo, Japan). Non-HDL-c was calculated as TC minus HDL-c. Atherogenic lipoprotein cholesterols were calculated as calculated sdLDL-c plus nonLDL-nonHDL cholesterol. SdLDL-c was calculated as previously reported [10]. NonLDL-nonHDL cholesterol was calculated as TC minus LDL-c minus HDL-c.
Two CTS indices, designated CTS qlt and CTS qnt , were calculated as follows: Of the 12,373 study participants, blood specimens were collected after more than 10 h of fasting from 10,297 participants (109 from the disease group) and residual 2076 specimens were collected within 10 h after the last meal. Blood glucose and HbA1c levels were categorized into lower (L), middle (M), and higher (H) categories (Table S1); the higher category was used for the regression analyses. Systolic and diastolic blood pressures were categorized into lower (L), middle (M), and higher (H) categories (Table S1); the higher category was used for regression analyses.

Statistical Analyses
We intended to compare hazard ratios (HRs) per 1 SD of the lipid indices for the CHD outcome, because one unit in the whole range was significantly different between the lipid indices. However, the distribution of some lipid indices (e.g., CTS qlt ) was extremely skewed and was therefore not suitable for use to calculate the SD to evaluate the HRs. Therefore, we used the Box-Cox transformation method to transform the distribution of the lipid index close to normal distribution [11]. The Box-Cox transformation was as follows: where x is the original value of the lipid index before the transformation. We determined the appropriate λ value for each index using p values from the Kolmogorov-Smirnov test and a Q-Q plot analysis ( Figure S1 for CTS qlt index). Using the transformed lipid indices, we performed Cox proportional hazards regression analyses to obtain the HRs per 1 SD. In the multivariate regression analyses, age, sex, smoking history, and categories of blood pressure and blood glucose levels were adjusted. Difference in the HR values for the two normalized indices was evaluated using the Welch's test.
For the latter analyses, we divided the population in this study into three group (groups 1, 2, and 3 from lower to higher) divided by 33.3 percentile and 66.6 percentile for each lipid index that is not transformed. Then, we estimated the risk increment compared to group 1. For other analyses, Chi-square test and Mann-Whitney U test were used for the categorical variables and continuous variables, respectively. All analyses were performed using the EZR software (Saitama Medical Centre, Jichi Medical University, Saitama, Japan), which is a modified version of R commander, designed to add statistical functions for biostatistics [12]. Table 1 shows the summary of the study population. Age, sex, smoking history, and categories of blood pressure and blood glucose levels significantly differed between the disease group and control group. To precisely estimate the HRs of the lipid indices, we adjusted them in the multivariate regression analyses. The HR for CTS qnt index was not significantly different from those of the TG and nonHDL-c in the multivariate Cox proportional hazards regression analyses ( Table 2). The HR for the CTS qlt index, which was lower than that for CTS qnt (p < 0.05), was comparable to that of the traditional atherogenic index, LDL-c. The HR for TC was the lowest among the indices evaluated, but the 95% confidence interval (CI) was greater than 1, indicating that the increase in TC levels increased the CHD risk. In contrast, the 95% CI for HDL-c was lower than 1, so the increase in HDL-c levels decreased the CHD risk, similar to previous studies [2]. We observed little effects of medications for HRs of the lipid indices in our model.  The combination of a lipid parameter with HDL-c, in which a parameter is divided by HDL-c, enhances the predictive ability of the index by increasing the HR for CHD. Both the LDL-c/HDL-c and TG/HDL-c indices have been proposed as good indicators of atherogenicity [13][14][15][16]. Therefore, the LDL-c/HDL-c and TG/HDL-c indices are likely to be better predictors of CHD, as evidenced by their HRs (Table 3). In this study, we found that the combination of CTS qnt and HDL-c showed comparable ability with LDL-c/HDL-c and TG/HDL-c to predict the CHD outcome. There was no statistically significant difference among HRs of these three indices. To estimate relative risk in a lipid parameter with skewed distribution, we divided the target population into quartiles (i.e., G-1, G-2, and G-3; divided by 33.3 percentile and 66.6 percentile for CTS qlt , TG/HDL-c, CTS qnt , and nonHDL-c) ( Table 4). The HRs in the higher group (G-3) and lower group (G-1) were similar in all indices, but the HR in the middle group (G-2) of the CTS qlt index was considerably higher. Notably, the magnitude of change in the HR between G-1 and G-2 and between G-1 and G-3 is more important in each index rather than their absolute values. This result was confirmed by graphical analyses using Kaplan-Meier curves ( Figure S2). In addition, we confirmed the proportionality of hazards using a log-log plot. These results suggested that the CTS qlt index was more sensitive for predicting the CHD risk, especially for the middle group.

Characteristics of CTS Indices
To understand the possible biological basis of the CTS qlt and CTS qnt indices, we investigated their relationships with previously known lipid indices. The CTS qlt index showed the best correlation with TG/HDL-c in the logarithmically transformed forms, in which the coefficient of determination was 0.8593 for log(TG/HDL-c) as a response variable ( Figure 1A). The log(TG/HDL-c) has been designated as an "atherogenic index in plasma (AIP)" and a good predictor for CHD [17]. More importantly, AIP showed very good correlation with the LDL particle size [18], with larger AIP values suggesting the production of potentially atherogenic smaller and denser LDLs. For the CTS qnt index, we considered a combination of calculated sdLDL-c and nonLDL-nonHDL cholesterol as a response variable, which is designated as an "atherogenic lipoprotein cholesterol" in this study. Surprisingly, the CTS qnt index and atherogenic lipoprotein cholesterol showed an extremely high correlation coefficient ( Figure 1B). Thus, the CTS qnt index appears to be another marker of the atherogenic lipoprotein cholesterol, but it should be noted that this index does not indicate the quantity of the lipoprotein cholesterol. In addition, we found that sdLDL-c can be estimated using the following formula: 0.2 LDL-c + 0.1 TG (Pearson's correlation coefficient = 0.9863; 95% CI = 0.9961-0.9964; p < 0.001). Therefore, nonLDL-nonHDL-c is expressed by the residual 0.05 TG (Pearson's correlation coefficient = 0.876; 95% CI = 0.872-0.880; p < 0.001). To understand the correlations between CTS qlt , CTS qnt , and CTS qnt /HDL-c, we observed the relationship among these indices using a three-dimensional graphical analysis (Figure 2), in which an axis of the CTS qlt index is logarithmically expressed. The Pearson's correlation coefficients for the relationships are as follows: 0.780 (95% CI = 0.773-0.787; p < 0.001) for CTS qnt and log(CTS qlt ); 0.904 (95% CI = 0.901-0.907; p < 0.001) for CTS qnt and CTS qnt /HDL-c; and 0.763 (95% CI = 0.756-0.770; p < 0.001) for log(CTS qlt ) and CTS qnt /HDLc. As shown in Figure 2, the distribution was linear, not distorted. Within the limited range of the CTS qnt /HDL-c (e.g., 0.6-0.7 in Figure 2), the shape of the distribution appeared like a vertical column on the CTS qlt -CTS qnt plane. This result suggests that the CTS qnt /HDL-c index represents the metabolic status of individuals restricted within the range of CTS qnt and CTS qlt . In other words, the CTS qnt /HDL-c index provides some information regarding the metabolic status of an individual, which can be further explored by the use of CTS qnt and CTS qlt indices.

Discussion
The CTS index is comparable to the previously known lipid indices in terms of ability to predict the CHD outcome for the Japanese general population. In particular, the LDL-c/HDL-c and TG/HDL-c indices have been reported as good lipid predictors of atherogenicity. This study found that the CTS qnt /HDL-c index was comparable to the aforementioned indices. Interestingly, the formula of the CTS qnt /HDL-c index can be deformed as 0.2 × (LDL-c/HDL-c) + 0.15 × (TG/HDL-c), which is understandable as a consolidated index of LDL-c/HDL-c and TG/HDL-c with the addition of weights. These results indicate the CTS qnt /HDL-c index is as useful as the previously known atherogenic indices.
Both CTS qlt and CTS qnt indices have characteristic patterns of distribution and biochemical natures. The CTS qlt index may reflect lipoprotein particle size, as presumed from its close correlation with the TG/HDL-c index in their logarithmically transformed forms. The HR/1 SD for the CTS qlt index was less than TG/HDL-c (Tables 1 and 2). However, focusing on the change in the relative risk along with the increment (or decrement) of those indices, the CTS qlt index appears to be more sensitive to the changes over the middle group than the TG/HDL-c index, suggesting that the CTS qlt index is a better indicator (Table 4 and Figure S2). The CTS qnt index is a marker of atherogenic lipoprotein cholesterol and consists of calculated sdLDL-c and nonLDL-nonHDL-c. The nonLDL-nonHDL-c is recognized as the calculated remnant-c [19,20]; therefore, the atherogenic lipoprotein cholesterol in this study is a consolidated expression of sdLDL-c and remnant-c. The most important issue is that the formula of CTS qnt index includes TG. Hence, it is reasonable that both sdLDL-c and remnant-c are affected by the increased TG. Notably, both sdLDL-c and remnant-c are calculated only with cholesterol parameters. Srisawasdi et al. [10] proposed a complicated formula to estimate sdLDL-c using several cholesterol parameters by a regression analysis of the measured sdLDL-c. However, our results suggest that this formula includes a hidden TG parameter. For the calculated remnant-c, it may be assumed that this is an expression of cholesterol in VLDL and thus, it can be estimated by TG of VLDL, particularly in the fasting condition. The usefulness of both calculated sdLDL-c and remnant-c for predicting the risk of CHD has previously been reported [21,22]. It is suggested that the CTS qnt index proposed in this study is a consolidated and more useful indicator for atherogenicity than the isolated use of calculated sdLDL-c or remnant-c.
We investigated the relationships between CTS qlt , CTS qnt , and CTS qnt /HDL-c ( Figure 2). CTS qlt and CTS qnt showed related metabolic conditions, but differed in their biochemical nature. The CTS qnt index reflects the quantity of the atherogenic lipoproteins, whereas the CTS qlt index probably reflects the size of LDL macromolecules. These two indices can be illustrated using a scattered plot with log(TG) as the X axis and log(LDL-c) as the Y axis (Figure 3), in which gray dots are the individual subjects. Considering an individual with high CTS qnt index (red-colored star symbol) who is administered a treatment that reduces LDL-c without decreasing TG (blue-colored star symbol), the CTS qnt index will definitely decrease, but the CTS qlt index will increase. This suggests that a treatment targeting only the LDL-c without affecting blood TG levels may be less beneficial and may even be harmful. In this point of view, reduction of TG appears to be very important together with decrement of LDL-c by statin treatment. Though several pharmacological interventions including use of fibrate, omega-3 fatty acid, or niacin have been attempted as the TG-lowering therapy [7], non-pharmacological approaches including adjustment of quantity and quality of daily diets, controlled exercise to consume inner fat, and maybe nutraceuticals, such as fish oils should be encouraged for apparently healthy individuals in every health check-up service. The CTS qnt /HDL-c index may represent the status of lipid metabolism determined by the CTS qlt and CTS qnt indices in an individual. Therefore, we may use the CTS qnt /HDL-c index as a comprehensive indicator of the metabolic status. Herein, we described novel atherogenic indices that are potentially useful to predict the risk of CHD. Our study revealed that three CTS indices are comparable to previously reported atherogenic indices in terms of their prediction ability for CHD. More importantly, it is suggested that these CTS indices are better indicators of the metabolic status that predisposes to atherosclerosis than the previously known indices and lipid parameters. Thus, the CTS indices may be superior to the previously known lipid indices for the evaluation of the metabolic status of individuals that may lead to CHD in the future. This information may be used to prevent CHD by advising the appropriate lifestyle changes.
The CTS indices proposed in this study are comparable indicators of the CHD risk to previously reported atherogenic indices, but the CTS qlt and CTS qnt indices more directly reflect the metabolic status that predisposes to CHD than the previously known indices. The CTS qnt /HDL-c index can be recommended as a screening indicator. However, we need more comprehensive and controlled prospective studies involving precisely diagnosed clinical entities, including severe or fatal CHD, to confirm our results. Additionally, we need to evaluate how well the CTS indices reflect the properties of the atherogenic lipoproteins as compared to direct biochemical measurements.
Supplementary Materials: The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/healthcare10040747/s1, Figure S1: The Q-Q plots for the CTS qlt index before (A) and after (B) the Box-Cox transformation. The λ value is -0.2; Figure S2: Kaplan-Meier curves of cumulative CHD incidence in the groups of CTS qlt (A), TG/HDL-c (B), CTS qnt (C), and nonHDL-c (D). The study population is divided by 33.3 and 66.6 percentiles of each index. Groups 1, 2, and 3 are shown as blue, green, and red lines, respectively; Table S1: Categories of blood pressure and blood glucose levels.
Author Contributions: I.M. contributed to the study design, data analyses, and manuscript preparation. M.A., Y.T. and M.N. collected the data via the health check-up service. Y.K. contributed to the study management. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no grant from any funding agency in the public, commercial, or not-for-profit sectors.
Institutional Review Board Statement: The present study was conducted according to the principles of the Declaration of Helsinki, and approval from the Institutional Review Board of JCHO Shiga Hospital (No. 2020-04).
Informed Consent Statement: Patient consent was waived due to the anonymization of the personal data.

Data Availability Statement:
The data presented in this study are available from the corresponding author on reasonable request. The data are not publicly available due to the restriction from the Institutional Review Board of JCHO Shiga Hospital.