Cholesterol esterification and atherogenic index of plasma correlate with lipoprotein size and findings on coronary angiography.

We examined the association between rate of cholesterol esterification in plasma depleted of apolipoprotein B-containing lipoproteins (FERHDL), atherogenic index of plasma (AIP) [(log (TG/HDL-C)], concentrations, and size of lipoproteins and changes in coronary artery stenosis in participants in the HDL-Atherosclerosis Treatment Study. A total of 160 patients was treated with simvastatin (S), niacin (N), antioxidants (A) and placebo (P) in four regimens. FERHDL was measured using a radioassay; the size and concentration of lipoprotein subclasses were determined by nuclear magnetic resonance spectroscopy. The S+N and S+N+A therapy decreased AIP and FERHDL, reduced total VLDL (mostly the large and medium size particles), decreased total LDL particles (mostly the small size), and increased total HDL particles (mostly the large size). FERHDL and AIP correlated negatively with particle sizes of HDL and LDL, positively with VLDL particle size, and closely with each other (r = 0.729). Changes in the proportions of small and large lipoprotein particles, which were reflected by FERHDL and AIP, corresponded with findings on coronary angiography. Logistic regression analysis of the changes in the coronary stenosis showed that probability of progression was best explained by FERHDL (P = 0.005). FERHDL and AIP reflect the actual composition of the lipoprotein spectrum and thus predict both the cardiovascular risk and effectiveness of therapy. AIP is already available for use in clinical practice as it can be readily calculated from the routine lipid profile.

knowledge about the lipoprotein structure and the link between particle diameter and total core lipid content. Lipoprotein size subpopulations were defi ned as follows: large VLDL/chylomicrons (>60 nm), medium VLDL (35-60 nm), small VLDL (27-35 nm), large LDL (21.2-23 nm), small LDL (18-21.2 nm), large HDL (8.8-13 nm), medium HDL (8.2-8.8 nm), and small HDL (7.3-8.2 nm). Measurement of FER HDL was described in detail previously ( 4,11,19 ). Briefl y, apoB-containing lipoproteins are precipitated from EDTA plasma (that can be stored at 2 20°C up to 4 months or at 2 70°C for up to 6 years without changes in absolute values of FER HDL ) by phosphotungstic acid and MgCl 2. To the supernatant, which contains plasma with HDL only, is added a fi lter paper disk containing a trace of 3-H cholesterol. After an overnight incubation at 4°C, the disk is removed and the plasma with labeled HDL is heated to 37°C and incubated for 30 min (the esterifi cation reaction is always linear over this time period). After the incubation, lipids are extracted by ethanol, ethanol evaporated, and with added internal standards of cholesterol and cholesteryl oleate, separated by TLC. Spots of cholesterol and cholesteryl oleate are visualized by iodine, spots cut from TLC plates, and transferred to scintillation vials. The radioactivity is estimated by liquid scintillation counting. The fractional esterifi cation rate is calculated from radioactivity in spots of free and esterifi ed cholesterol as percentages of HDL-C esterifi ed per h. AIP ( 12 ) was calculated as logarithmically transformed ratio of molar concentrations of TG and HDL-C [log (TG/HDL-C)] in plasma ( 20 ).

Statistical analysis
Statistical analysis was performed using SPSS. 15 .0 and R ( 21 ) software. The data are presented as means ± SD both before and during treatment for the four treatment groups. For descriptive purposes, the differences between measurements taken before and after treatment were tested by paired t -test within the four groups. The effect of treatment on FER HDL and AIP was analyzed by one-way ANOVA. We tested the hypotheses that the mean values after 1 year of treatment are equal against the alternative that they differ at least for one treatment.
To investigate the correlations between FER HDL and AIP on one hand and particle sizes and concentrations on the other hand, we calculated bivariate correlation coeffi cients for basal values of all subjects in the study and partial correlation coefficients for values obtained after treatment to eliminate the infl uence of the various treatments. To determine the after-treatment relationships between these measurements, we fi tted two linear regression models with FER HDL and AIP as the response variables and particle sizes and concentrations as explanatory variables.
We assessed association of changes in the coronary artery stenosis with FER-HDL, AIP, and other variables by logistic regression model. The progression of the coronary artery stenosis, defi ned as positive change versus no change or regression (i.e., dichotomous outcome) was considered as a response variable and the fi nal model was found by the forward selection procedure ( 21 ). The initial set of explanatory variables was as follows: AIP, FER HDL , total LDL and HDL cholesterol, triglycerides, apoAI, apoB, HDL, LDL, VLDL particle sizes, and HDL, LDL, and VLDL subpopulations' concentration. All models were adjusted for treatment.

RESULTS
Changes in the concentration and particle size of lipoproteins on treatment with S+N and S+N+A Table 1 summarizes the data on lipoprotein subpopulations before and after 1 year of therapy with the four treatment regimens. The table also shows results of paired (apo)B-depleted plasma (FER HDL ) and atherogenic index of plasma [log (TG/HDL-C)] (AIP) refl ect the size of LDL and HDL subpopulations and closely correlate with each other over a wide range of plasma lipid values (10)(11)(12)(13). AIP is, of course, a transformation of triglyceride (TG)/ HDL-C that better meets the assumption of normality of the errors in the statistical model being used to describe the treatment effects than does the untransformed variable.
The value of both FER HDL and AIP can be seen in the context of intravascular cholesterol transport: FER HDL measures esterifi cation rate of cholesterol by lecithin: cholesterol acyltransferase within HDL differently sized subpopulations. In small HDLs the esterifi cation rate is high but large particles reduce it ( 10,14 ). The destination of newly produced cholesteryl esters (CEs) is also linked to subpopulations size and with added internal standards of unesterifi ed cholesterol and cholesteryl oleate. Large HDLs reduce esterifi cation rate and serve as the most effective vehicle for delivery of CE via scavenger receptor class B type 1 to catabolic sites in liver and steroidogenic tissues ( 15 ). The close association of FER HDL with AIP can be explained by TG participation in the production of large VLDL and small dense LDLs and have also been proposed to be the major determinants of cholesterol esterification/transfer and HDL remodeling in particles that regulate the esterifi cation rate.
The potential of FER HDL and AIP to predict CV risk was shown in the study of 1,108 patients who underwent coronary angiography ( 16 ). The relationships between FER HDL or AIP and CV risk have been well established ( 12,16,17 ). However, the changes of these risk biomarkers with different therapies and their relation to treatment outcomes have not been studied.
In this study, we related the changes on coronary angiography in HATS to the values of FER HDL and AIP and investigated their relation to lipoprotein subpopulations in patients on different therapeutic regimens.

Patients
The rationale, methods, and results of HATS have been described in detail ( 9 ). The study tested the hypothesis that a decrease in serum LDL-cholesterol (LDL-C) with a simultaneous increase in HDL-C induced by the statin-niacin combination therapy provides greater benefi ts than treatment with either placebo or antioxidants. One hundred and sixty patients were divided into four groups and each group was treated with one of four regimens: simvastatin plus niacin (S+N), antioxidants (A), simvastatin, niacin, and antioxidants (S+N+A ), or placebos (P). Patients underwent coronary angiography before and after 3 years of treatment. Plasma samples obtained at baseline and at 1 year on therapy were examined in the present analysis.

Laboratory assays
Analyses of plasma lipids and apolipoproteins were previously described ( 9 ). The average particle sizes of HDL, LDL, and VLDL subpopulations were determined by NMR spectroscopy ( 18 ). Particle concentrations (nmol/L for VLDL and LDL; µmol/L for HDL) were calculated for each subclass based on existing values after 1 year of treatment were compared by one-way ANOVA. In both cases, the hypothesis that mean values are the same in all groups was rejected ( P < 0.001). Compared with placebo, antioxidant therapy had no effect, whereas S+N and S+N+A treatment decreased AIP and FER HDL signifi cantly.

Correlations between AIP and FER HDL and concentrations and particle sizes of lipoproteins on treatment
We examined the relationship between FER HDL , AIP, and the lipoprotein particles in plasma baseline and on the various treatment regimens ( Table 2 ). At baseline, we used bivariate analysis, as the starting values of the patients were similar. The possible effects of the 1 year therapy were eliminated using appropriate adjustments. Table 2 shows that values of the correlation coeffi cients before and on treatment remained very close. FER HDL and AIP values correlated with each other at baseline ( r = 0.721); the partial correlation coeffi cient at 1 year of therapy (adjusted for treatment) was r = 0.729. The type of the treatment did not have a statistically signifi cant effect on the linear relation between other variables. There was a signifi cant correlation between FER HDL and AIP and the number of total and small LDL and total, large, medium, and size of VLDL. Highly signifi cant inverse correlations were observed in the atheroprotective variables such as large HDL. The inverse correlations were seen between LDL particle size and large LDL. Also signifi cant association was found of FER HDL and AIP with atherogenic apoB and atheroprotective apoAI.
t -tests between baseline and on treatment values performed for each treatment separately. Although the p -values are not adjusted for multiple comparisons, they indicate that there was a marked increase in the total HDL particles and decrease in the total LDL and VLDL particles induced by the S+N and S+N+A treatment. Total HDL increased mostly on account of large HDL (from 8.7% to 15.5% of total HDL in S+N, 6.% to 12.5% in S+N+A treatment). LDL decreased on account of small LDL particles whereas the number of large particles was not signifi cantly changed. On the contrary, the treatment by S+N and S+N+A reduced practically all large, medium, and small VLDL particles. The mean particle size of HDL and LDL signifi cantly increased by treatment with S+N and S+N+A. During these treatments, total cholesterol (TC), LDL-C, and TG markedly decreased while HDL-C increased. Placebo treatment had similar even lower signifi cant effects on routine lipid profi le with the exception of TG. Table 1 shows that after 1 year of treatment with S+N and S+N+A, FER HDL decreased from 30.73 ± 7.05 and 32.0 ± 7.53%/h at baseline, respectively and to 19.53 ± 6.76 ( 2 36%) and 21.96 ± 8.64%/h ( 2 31%,), respectively. AIP decreased from 0.43 ± 0.22 and 0.49 ± 0.24 at baseline to 0.13 ± 0.25 ( 2 71%) and 0.22 ± 0.31 ( 2 51%), respectively. The placebo group also showed a small decrease in FER-HDL ( 2 12.1%) whereas antioxidants had negligible effect. For the four treatment groups, the mean AIP and FER HDL

DISCUSSION
The objective of this study was to assess the relation between the novel biomarkers FER HDL and AIP and the distribution of lipoprotein subpopulations before and during lipid-lowering treatment in patients with coronary disease, low HDL-C, and normal LDL-C in HATS ( 9 ). We also studied the association between these markers and the changes in coronary artery stenosis.
The lipid-lowering treatment changed the sizes and concentrations of the lipoprotein subpopulations as well

Associations of FER HDL and AIP with changes in coronary artery stenosis
The HATS study participants were divided into two groups based on changes (negative vs. positive) in coronary artery stenosis after 3 years of treatment to show association between plasma lipoproteins and their subpopulations to the angiographic changes ( Table 3 ). Both FER HDL and AIP had higher values in the group with increased stenosis ( P < 0.001 and 0.008), together with increased total particles of LDL ( P < 0.007) and VLDL ( P < 0.033), small LDL ( P < 0.005), and large and medium VLDL ( P < 0.044 and 0.036). Although the total number of HDL particles was not signifi cant in relationship to the changes in stenosis, the decreased stenosis was characterized by an increase of large HDL particles ( P < 0.001) and reduction of large VLDL and small LDLs. From traditional lipids, namely TC, LDL-C, HDL-C, and TG, only HDL-C has shown signifi cant reduction in the progression group.
The forward stepwise logistic regression analysis (adjusted for treatment) of changes in the coronary artery stenosis showed that albeit the probability of progression was signifi cant in the fi rst step with FER, AIP, HDL-C, ApoA1, ApoB, large HDL, total and small LDL particles, total and medium VLDL particles, and sizes of LDL and HDL, this probability was best explained by FER HDL only ( Table 4 ) (odds ratio = 1.07, P = 0.005). No other variable was signifi cant in this model. When FER HDL was not included in the initial set of predictors in the model 2 selection procedure, the fi nal model adjusted for treatment contained again only one signifi cant predictor of probability of progression/regression, which was the concentration of the large HDL subpopulation (odds ratio = 0.80, P = 0.016). If AIP was tested in the model (adjusted for treatment), its p -value was borderline signifi cant ( P = 0.055).   vastatin in the HATS not only decreased the concentration of plasma LDL-C and increased HDL-C ( 9 ) but also changed favorably the distribution of HDL subpopulations by increasing the proportion of large HDL. These changes resulted in markedly decreased values of both AIP and FER HDL . Fibrates also decrease TGs and increase HDL but alter HDL distribution, in contrast to niacin, by increasing the proportion of small HDL and decreasing the large HDL ( 25 ).
In the logistic regression model adjusted (also nonadjusted) for treatment regimens, the probability of progression of the coronary artery stenosis ( Table 4 ) was best explained by changes in FER HDL with no other variable being signifi cant in this model. When FER HDL was not included in the set of initial predictors, the fi nal model adjusted for treatment again contained only one predictor, namely the ( 2 ) concentration of large HDL subpopulation. If FER HDL was replaced by AIP in the model, the p -value for AIP was borderline signifi cant ( P = 0.055). The effect of large HDL on the change of coronary stenosis, assessed by different methods, was previously reported ( 9,22 ).
We hypothesize that the increased number of the large HDL particles, while suppressing the esterifi cation rate of cholesterol, enhances the catabolism of the newly produced CE via scavenger receptor class B type 1. Our idea that differently sized HDL particles may affect the targeting of CE produced in plasma to either atherogenic or atheroprotective targets ( 26 ) is supported by the recent fi nding that rosuvastatin therapy may induce the regression of coronary atherosclerosis by raising plasma HDL-C, specifi cally by increasing HDL particle size ( 27 ). Thus, we believe that FER HDL is a good measure of the atherogenic (or atheroprotective) pathways. It is not surprising that AIP, which is also associated with the lipoprotein size and correlates highly with FER HDL , has a similar predicting potential as FER HDL .
Our results confi rm the importance of not only quantitative but also qualitative changes in HDL that occur with niacin treatment.
Although the concept of using either AIP or FER HDL in practice will have to be further confi rmed, a recent paper suggests that AIP may be of importance: a large study from Turkey found that AIP was the best predictor of hypertension, diabetes, and vascular events ( 28 ).
The authors thank Dr. J. Otvos for the NMR analyses.
as the values of FER HDL and AIP. Statin+niacin-containing regimens lowered the atherogenic VLDL and LDL and increased the protective HDL. There was a decrease in the proportion of large and medium sized VLDL and small dense LDL particles. On the other hand, there was an increase in the atheroprotective large HDL.
Treatment with antioxidants alone did not affect either FER HDL or AIP, or VLDL or LDL subpopulations. However, use of antioxidants either alone or in combination with S+N tended to increase small HDL particles ( Table  1 ). Placebo treatment had a minor positive effect on the distribution of the various subpopulations ( Table 1 ), signifi cant only for the increase of large HDL. More pronounced was the decrease in FER HDL and improvement of routine lipid profi le. This effect was probably related to the participants' compliance with lifestyle recommendations and to the protocol use of simvastatin among the roughly 8% of placebo patients with baseline LDL-C > 3.5mmol/L. The correlation between FER HDL and AIP was highly positive ( r = 0.717) both at baseline and at 1 year on treatment ( r = 0.729).
FER HDL and AIP strongly correlated with the size and concentration of individual lipoprotein subpopulations ( Table 2 ). Increased concentration of medium and large VLDL and small LDL particles resulted in higher FER HDL and AIP whereas the values of these parameters decreased with increasing concentration of large LDL and large HDL subpopulations. To further investigate the relation between FER HDL , AIP, and particle sizes, we used two linear regression models to assess the potential of the explanatory variables (not shown). The variability of AIP was best explained by all VLDL concentrations and VLDL size (positive effect) and concentrations of large HDL and large LDL (negative effect). The coeffi cient of determination was 0.75, which means that the model explained 75% variability of AIP. The variability of FER HDL was best explained by concentration of large HDL and HDL and VLDL particle sizes with coeffi cient of determination 0.62.
In the HATS, patients with normal LDL-C and low HDL-C level benefi ted signifi cantly from the combination treatment with simvastatin and niacin that resulted in regression of coronary atherosclerosis ( 9 ). As previously reported, niacin increases the large particle size of HDL (21)(22)(23), and decreases the small HDL subpopulations ( 22,23 ). Statins also increase the large a -1 HDL subpopulation ( 24 ). That was probably why the combination of niacin and sim- Model 1: Model found by stepwise selection procedure, initial set of predictor variables: FER HDL , AIP, TC, LDL-C, HDL-C, TG, apoAI, apoB, HDL (total, large, medium, small, size), LDL (total, large, small, size), VLDL (large, medium, small, size) and age. Model 2: Model found by stepwise selection procedure, the initial set of predictor variables without FER HDL , other variables the same as in Model 1. Models are adjusted for treatment.