Low-density lipoprotein subclasses over the menopause transition and risk of coronary calcification and carotid atherosclerosis: The SWAN Heart & HDL Ancillary studies

Abstract

Objective:

Perimenopausal women experience a steep increase in low-density lipoprotein cholesterol (LDL-C) that is related to a higher risk of carotid plaque later in life. LDL subclasses have been linked to cardiovascular diseases (CVD) beyond LDL-C, promising a better risk stratification. We aim to characterize changes in LDL subclasses and assess their associations with presence of coronary artery calcification (CAC score ≥10) and carotid intima-media thickness (cIMT) over the menopause transition (MT), and by menopause stage.

Methods:

Nuclear magnetic resonance spectroscopy LDL subclasses were measured for a maximum of 5 time-points. CAC and cIMT were measured for a maximum of 2 time-points. LOESS plots, linear mixed effect models and generalized estimating equations were used for analyses.

Results:

The study included 471 women (baseline-age 50.2±2.7 years; 79.0% Pre-/Early perimenopausal), of whom 221 had data on CAC or cIMT. LDL subclasses increased over the MT while intermediate density-lipoprotein particles declined. In adjusted models, higher total LDL particles (LDL-P) and ApoB were associated with greater CAC prevalence and greater cIMT. Although none of the associations were modified by menopause stage, higher LDL-C, ApoB and total LDL-P were associated with greater cIMT during peri- or postmenopause stages, while higher LDL-C, and small LDL-P were associated with greater CAC prevalence, mainly during perimenopause.

Conclusions:

During the MT, women experience significant increases in LDL subclasses found to be related to greater cIMT levels and CAC prevalence. Whether these changes could better predict future risk of hard CVD events beyond LDL-C remains a research question to address.

Introduction

The menopause transition (MT) is a critical stage for midlife women when they experience alterations in hormone levels, metabolism, body composition, and risk for cardiovascular diseases (CVD)¹. Lipids adversely change during the MT. In the Study of Women’s Health Across the Nation (SWAN), women showed a significant accelerations in low-density lipoprotein cholesterol (LDL-C) and apolipoprotein B (ApoB) within 1 year of the final menstrual period (FMP)². Other studies have observed similarly higher levels of LDL-C in postmenopausal compared with premenopausal women^3,4. High LDL-C is a significant risk factor for CVD⁵. The reported increases in LDL-C during midlife² could contribute to the increased risk of CVD reported after menopause¹.

Despite the effort made to lower LDL-C either through lifestyle modifications and/or the use of lipid-lowering therapy, CVD prevalence remains high among those with normal or low levels of LDL-C⁶. This suggests that high LDL-C alone may not account for all atherosclerotic CVD risk, and that residual lipids and lipoprotein pathways may serve as important markers of CVD⁷. Recently, several studies have focused on characterizing LDL subclasses and their associations with CVD risks^8–10. However, none of these studies characterized the changes in LDL subclasses and their related CVD risk in women over the MT.

Measures of subclinical atherosclerosis such as coronary artery calcium score (CAC) and carotid intimal-medial wall thickness (cIMT) are significant predictors of future CVD risk^11,12. Multiple studies showed associations between LDL subclasses and subclinical atherosclerosis measures. In the Multi-Ethnic Study of Atherosclerosis (MESA), very small and small LDL particles (LDL-P) were positively associated with CAC progression beyond the traditional lipids¹³. Positive associations of higher total LDL-P and small LDL-P with CAC progression were reported in individuals with moderate or high risk of CVD^14,15. Additionally, in participants with moderate risk of CVD, higher LDL peak time (an indicator of greater average LDL density) was associated with greater cIMT progression¹⁶.

Very limited literature exists on the relationship between LDL subclasses and subclinical atherosclerosis measures in women during midlife. A cross-sectional study found a positive association of smaller LDL-P size and higher LDL-P concentration with higher CAC score in postmenopausal women from the Healthy Women Study¹⁷.

The Study of Women’s Health Across the Nation (SWAN) Heart and HDL ancillary studies quantified subclinical measures of atherosclerosis and LDL subclasses, respectively, over the MT and thus, provide an excellent opportunity to fill current gaps in our understanding of how LDL subclasses might contribute to women’s CVD risk at midlife. Using data from SWAN Heart and HDL ancillary studies, we characterized changes in LDL subclasses over the MT. We also examined whether changes in LDL subclasses were associated with a higher risk of CAC prevalence (CAC Agatston score ≥ 10) and a greater cIMT during the MT, and by menopause stage.

Methods

Study participants

SWAN is an ongoing, multi-ethnic, multi-site, community-based, longitudinal study of women transitioning through menopause. The complete design of the SWAN study has been published previously¹⁸. In brief, 3302 women (42 to 52 years old) were recruited between 1996 and 1997 at 7 clinical sites across the United States (Boston, Chicago, the Detroit area, Los Angeles, Newark, Pittsburgh, and Oakland California) who met the following criteria: had an intact uterus and at least one ovary, not be pregnant or lactating at recruitment time, had at least one menstrual period and not on hormone therapy at least 3 months before recruitment.

The SWAN HDL study is an ancillary study to SWAN focusing on characterizing changes in HDL subclasses, lipid content and function during the MT. A total of 558 women from the SWAN cohort were selected to participate in SWAN HDL if they had no history of CVD, at least one visit before the FMP and two visits after the FMP with available blood samples at the selected visits (1461 observations). Lipoprotein subfractions were measured among all SWAN HDL participants using stored blood samples collected over the MT, corresponding to SWAN parent study visits 1, 3–9, and 12. Women without available FMP dates (n=87) were excluded from analysis characterizing LDL subclasses over time relative to FMP; leaving a total of 471 women (1288 observations) for this analysis. For the analysis linking LDL subclasses with subclinical measures of atherosclerosis over the MT, SWAN HDL participants who were also part of the SWAN Heart ancillary study, another ancillary to SWAN that aimed to characterize subclinical atherosclerosis among midlife women at the Pittsburgh and Chicago sites, were included; a total of 221 participants (417 observations) with a maximum of 2 time points of data.

The study protocol was approved by the institutional review board at each site and all participants signed the informed consent forms at each study visit.

Subclinical Atherosclerosis Measures

CAC measurements and quantification.

Calcification of the coronary arteries was measured at two visits using an Imatron C-150 Ultrafast scanner (GE-Imatron, South San Francisco, CA). The first pass aimed to identify the landmarks for the coronary scans. The second pass was performed to provide 30 to 40 contiguous 3-mm-thick transverse images of the coronary arteries from the level of the aortic root to the apex of the heart. All scans were saved and then scored using the Agatston method¹⁹ implemented by a DICOM workstation and software by AcuImage, Inc (South San Francisco, California). The presences of calcification were identified if there were at least 3 contiguous pixels with a radiodensity >130 Hounsfield units (HU). CAC was quantified for the 4 major arteries and then summed to get a total Agatston score. The scans were scored by a technologist under the supervision of a cardiologist. This method had high reproducibility when measured in 40 consecutive participants with wide range of calcium. Intraclass correlation for CAC scores was 0.99²⁰.

cIMT.

cIMT was assessed using a Toshiba SSA-270A scanner (Toshiba American Medical Systems, Tustin, CA) and a Hewlett-Packard 5500 scanner (Hewlett-Packard, Andover, MA). B-mode images were obtained from the near and far walls of the distal common carotid artery (1 cm proximal to the carotid bulb), the far wall of the carotid bulb (from the point where the near and far walls of the common carotid artery are no longer parallel and extending to the flow divider), and the far wall of the internal carotid artery (distal 1 cm from the flow divider). For each of these locations, a 1 cm segment of the lumen-intima interface and the media-adventitia interface were electronically traced using computer-assisted measurements. A total of 140 data points were generated for each location. Average of the readings in each location was calculated, and then readings were averaged across all locations to obtain the average cIMT value. The intraclass correlation coefficient was 0.98 and determined from repeated reads on 20 cIMT scans²¹.

Blood measurements

Blood sample collections.

Participants were fasting overnight for at least 10 hours before blood draws. Blood samples were collected within 2–5 days of their menstrual cycle. A random sample was obtained within 90 days of the scheduled follow-up visits when a timed sample was not collected due to less regular menstrual cycles. The blood samples were stored at −4°C before being separated and were then frozen at −80°C and shipped on dry ice to Medical Research Laboratories (MRL, Lexington, KY).

Lipoprotein subclasses.

The lipoprotein subclasses were analyzed by Nuclear Magnetic Resonance (NMR) spectroscopy at LabCorp (Morrisville, NC) by the NMR LipoProfile-3 algorithm²², using the 400 MHz proton NMR spectrum platform. Lipoprotein subclasses of different diameters generate unique NMR signals at specific shapes and frequencies. The signal amplitude of each subclass is proportional to the number of particles emitting the particular signal, which enables the calculation of the subclass particle concentrations from the signal amplitude. To obtain the amplitude contributed by each lipoprotein subclass, the line shape of the signal envelope was modeled by summing the lipoprotein subclasses signals. The subclass particle concentrations were calculated by multiplying the derived subclass signal areas by their specific signals to particle conversion factors, and further grouped into intermediate-density lipoprotein particles (IDL-P), large LDL-P, and small LDL-P. The total LDL-P was calculated by summing the concentrations of all subclasses. The average LDL-P size was calculated by adding the multiplicative products of the diameter of each subclass and its relative mass percentage from NMR signal amplitude. The three identified LDL subclasses were IDL-P (23–29 nm), large LDL-P (20.5–23 nm), and small LDL-P (18–20.5 nm). To ensure valid results, stored samples that had been frozen at −80°C and never been thawed were used. The intra-assay and inter-assay coefficients of the variation for LDL-P concentrations and size ranged from 0.5% to 11.9% and 0.6% to 13.2%, respectively²².

Triglycerides, HDL-C and LDL-C.

Fasting triglycerides were measured using the Hitachi 747–200 clinical analyzer at Medical Research Lab (MRL, Lexington, KY) at SWAN visit 1 and follow-up visits 3–7 or by using the ADVIA assay at the University of Michigan (UM, Ann Arbor, MI) at SWAN visits 9 and 12. The HDL cholesterol (HDL-C) was quantitated using EDTA treated plasma at MRL at SWAN visit 1 and follow-up visits 3–7 or analyzed using the ADVIA Direct-HDL Cholesterol method at UM at SWAN visits 9 and 12. The results from UM were calibrated to be comparable to values obtained from MRL. The estimated LDL-C was calculated at all visits using the Friedewald equation when the corresponding triglycerides were below 400 mg/dL²³.

Apolipoprotein A1 and B.

ApoA1 and ApoB were measured in plasma using immunonephelometry using Behring reagents at MRL (Michigan, MI) for SWAN visit 1 and follow-up visits 3–7 or analyzed using reagents from Beckman-Coulter (Brea, Ca) at University of Pittsburgh Heinz Lab (UP, PA) for SWAN follow-up visits 9 and 12. Results from UP were calibrated to be comparable to values obtained from MRL.

Study covariates

Information on age, menopause status, body mass index (BMI), systolic blood pressure (SBP), current smoking status, and time since FMP was collected at each SWAN visit. Race/ethnicity was self-reported at the SWAN baseline visit. The reported race/ethnicity categories were White, Black, Hispanic, Chinese, and Japanese. BMI (Kg/m²) was calculated as the ratio of the measured weight (Kg) relative to measured height² (m²). Smoking status was based on participant self-report and categorized as current smokers vs. past/never smokers.

Age was calculated as the differences between visit date and birth date. Menopause status was determined based on the self-reported menstrual bleeding patterns and medical history. Participants were categorized into the following categories: 1) pre- (no change in menstrual bleeding and cycles) or early perimenopausal (changes in menstrual bleeding and had menstrual cycle within the past 3 months), 2) Late perimenopausal (one menstrual cycle within the past 12 months but no menstrual in the past 3 months, 3) postmenopausal (no menstrual cycle within the past 12 months due to natural menopause), or 4) unknown (menopausal status unknown due to hormone therapy use). FMP was identified as the date of the last reported menstrual period at the visit before the visit that women were first classified as postmenopausal.

Using the National Cholesterol Education Program Adult Treatment Panel III (NCEP-ATP III) diagnostic criteria, women were diagnosed with MetS if they presented with at least three of the five criteria: abdominal obesity (defined as waist circumference ≥ 80 cm for Japanese/Chinese women, and ≥ 88 cm for others), hypertriglyceridemia (fasting triglycerides ≥ 150 mg/dL), low HDL-C (< 50 mg/dL), impaired fasting glucose (fasting glucose ≥ 100 mg/dL or antidiabetic medication use), and hypertension (defined as systolic blood pressure ≥130 mmHg or diastolic blood pressure ≥85 mmHg or taking any blood pressure medication).

Systolic blood pressure (SBP) was measured on the right arm using a standardized protocol. Participants were required to be seated and feet flat on the floor for at least 5 minutes before measurements. Women were instructed not to smoke or consume any caffeinated beverage within 30 minutes prior to their measurements. A standard sphygmomanometer was used to record SBP at the first Korotkoff sound. Two sequential measures were taken with a minimum 2-minute rest between measures. The average values of the last two measures were used.

Statistical analysis

Non-parametric locally weighted regression with scatter smoothing (LOESS) plots were used to examine the trajectories of LDL subclasses, LDL-C, and ApoB over the MT and to identify inflection points. For simple presentation, the identified segments (i.e., between successive inflection points) were labeled as premenopause, perimenopause or postmenopause stage. The annual changes of each analyte during the different time segments (e.g., menopause stage) were estimated using piecewise linear mixed effect models. The association between each LDL metric and cIMT over time was assessed using a linear mixed effect model. Coronary artery calcium (CAC) was categorized and defined as present when the CAC Agatston score was at least 10. The association between each LDL metric and prevalence of CAC were evaluated using a generalized estimating equation model for binary outcomes. To assess whether the association of each analyte with cIMT levels and CAC prevalence varied by menopause stage, an indicator variable for stage (premenopause, perimenopause or postmenopause) was created according to the identified inflection points. An interaction term for this indicator with each analyte was included in the linear mixed effect or generalized estimating equation models to test effect modification by time relative to the FMP for the association between the LDL metrics and subclinical CVD outcomes. Effect modification by race/ethnicity was also tested for the association between the LDL metrics and subclinical CVD outcomes. Analyses were performed using SAS version 9.4 (SAS Institute, Cary, CA).

Results

The study included 471 midlife women of whom more than 50% were White. Women were mainly pre- or early perimenopausal (79.19%) and aged 50.20±2.67 years old at first available visit. The mean±SD age for the subset of women with available cIMT or CAC measures (N=221) was 51.13±2.80 years, and 60.18% of them were pre- or early perimenopausal at first available visit with a CAC or cIMT measure (Table 1). The prevalence of CAC≥10 was 23.70%. The median (Q1, Q3) of the continuous CAC score was 0 (0, 7.21).The mean±SD of cIMT was 0.68±0.10mm. The summary statistics of other demographics, CVD risk factors and LDL metrics for both groups are listed in Table 1.

Table 1.

Characteristics of study participants at first available visit with LDL metric data

Variable	Sample of all women with data on LDL subclasses N=471	Subsample of women with data on LDL subclasses and subclinical measures of atherosclerosis N=221a
Age, years, mean (SD)	50.20(2.67)	51.13(2.80)
Race, n (%)
White	244(51.80%)	139(62.90%)
Black	130(27.60%)	82(37.10%)
Hispanic	3(0.60%)	0(0%)
Chinese	52(11.04%)	0(0%)
Japanese	42(8.92%)	0(0%)
Education, n(%)
High School or less	83(17.70%)	34(15.38%)
Some college	143(30.49%)	66(29.86%)
College/Postgraduate	243(51.81%)	121(54.75%)
Menopausal Status, n(%)
Pre-/ Early perimenopausal	373(79.19%)	133(60.18%)
Late Peri-menopausal	37(7.86%)	24(10.86%)
Postmenopausal (Natural)	52(11.04%)	57(25.79%)
Unknown due to HT	9(1.91%)	7(3.17%)
BMI, kg/m2, mean (SD)	28.10(6.64)	29.54(6.47)
Systolic blood pressure, mmHg, median (Q1, Q3)	112(104, 125)	117(107, 128)
Current smoking, n (%)	49(10.47%)	29(13.18%)
Anti-lipid medication use, n (%)	13(2.76%)	13(5.88%)
Time since FMP, years, median (Q1, Q3) b	−2.08(−3.54, −0.77)	−1.22(−3.30, 0,81)
Triglycerides, mg/dL, median (Q1, Q3)	94(70, 130)	104(76, 141)
LDL-C, mg/dL, mean (SD)	112.48(31.03)	120.66(33.98)
ApoB, mg/dL, mean (SD)	105.46(26.34)	110.00(28.30)
HDL-C, mg/dL, mean (SD)	59.37(14.16)	55.68(12.86)
apoA1, mg/dL, mean (SD)	163.62(26.18)	160.58(25.26)
Total LDL-P, nmol/L, mean (SD)	1148.31(375.16)	1251.06(406.31)
Small LDL-P, nmol/L, median (Q1, Q3)	505.50(323.29)	567.25(330.65)
Large LDL-P, nmol/L, median (Q1, Q3)	371.14(211.11)	404.09(223.62)
IDL-P, nmol/L, mean (SD)	271.65(140.82)	279.72(149.37)
LDL Size, nm, mean (SD)	20.96(0.62)	21.00(0.64)
CAC score ≥10, n (%) a	-	50(23.70%)
CAC score, median (Q1, Q3) a	-	0.00(0.00, 7.21)
cIMT, mm, mean (SD) a	-	0.68(0.098)

apoA1: apolipoprotein A1; ApoB: apolipoprotein B; BMI: body mass index; CAC: coronary artery calcium score; cIMT: carotid intimal-medial wall thickness; FMP: final menstrual period; HDL-C: high-density lipoprotein cholesterol; HT: hormone therapy; IDL-P: intermediate-density lipoprotein particles; LDL: low-density lipoprotein; LDL-C: low-density lipoprotein cholesterol; LDL-P: low-density lipoprotein particles.

^aCAC n=211, cIMT n=217

^bTime since FMP was calculated as the difference between the SWAN study visit date and the FMP date. Negative time since FMP indicated SWAN study visit date was prior to the FMP.

Correlations among LDL measures are presented in Supplemental Table 1. LDL-C was highly correlated with ApoB and total LDL-P, and moderately correlated with small LDL-P, Large LDL-P and IDL-P. ApoB and Total LDL-P were highly correlated with each other and were correlated more with small vs. large LDL-P.

Changes in LDL metrics relative to premenopause, perimenopause and postmenopause stages

Adjusting for study site, race/ethnicity, and age at FMP, LDL subclasses increased during the perimenopause stage while IDL-P declined, Table 2 and Figure 1. The reported annual changes significantly differ from those occurred during the premenopause stage, Table 2. Interestingly, LDL size significantly increased only during the premenopause stage with the reported change significantly differed from those in the peri/postmenopuase stage, Table2.

Table 2.

Annual changes in LDL metrics by menopause stage

LDL metrica	Premenopauseb	Perimenopauseb	Postmenopauseb
	β(SE)c	β(SE)c	β(SE)c
LDL-C, mg/L	−0.95(0.85)	4.71(0.45) d,e	−0.64(0.31)
ApoB, mg/L	−0.65(0.76)	4.92(0.42) d,e	1.49(0.30)
Total LDL-P, umol/L	−11.24(10.25)	38.11(5.48) d,e	−3.56(3.81)
Small LDL-P, umol/L	−8.44(6.17)	20.10(8.30) d,e	−3.13(4.10)
Large LDL-P, umol/L	3.55(5.01)	21.50(3.75) d,e	−2.84(4.30)
IDL-P, umol/L	3.73(2.16)	−3.47(1.43)d
LDL Size, nm	0.04(0.01)	0.002(0.005)d

ApoB: apolipoprotein B; IDL-P: intermediate-density lipoprotein particles; LDL: low-density lipoprotein; LDL-C: low-density lipoprotein cholesterol; LDL-P: low-density lipoprotein particles.

^aNumber of observations and Sample sizes for LDL-C: obs=1117, n=458; ApoB: obs=1172, n=459; Total LDL-P: obs=1213, n=460; Small LDL-P: obs=1213, n=460; Large LDL-P: obs=1213, n=460; IDL-P: obs=1213, n=460; LDL size: obs=1204, n=460.

^bTo simplify presenting the results, menopause stages were used to label the segments based on the inflection cut points identified by the LOESS plots, and each cut points reflected the years relative to the final menstrual period. For the metrics with three segments, the cut points were: LDL-C: −2, 2.5; ApoB: −2, 2.5; Total LDL-P: −2, 2.5; Small LDL-P: −0.5, 2.5; Large LDL-P: −1, 4; The cut points for LDL metrics with two segments were: IDL-P: 1; LDL size: 0. There were no significant cut points identified to separate peri- and postmenopause stages.

^cAdjusted for age at FMP, study site and race.

^dSignificantly differs from premenopause stage.

^eSignificantly differs from postmenopause stage.

Bold indicates significantly differs from zero at P ≤ 0.05.

Figure 1. Trajectories of LDL conventional and novel metrics over the menopause transition.

The vertical lines represent the identified inflection points which separate stages of MT. For metrics with 3 segments, the two inflection points separate the MT to pre-, peri-, and postmenopause. For metrics with 2 segments, the infection point separates pre- vs. peri-/postmenopause. The solid lines are locally weighted smoothing (LOESS), and the dashed lines are the corresponding lower and upper 95% confidence interval. The observations were restricted to 5 years before and 10 years after FMP due to limited sample sizes at the two ends.

ApoB: apolipoprotein B; IDL-P: intermediate-density lipoprotein particles; LDL: low-density lipoprotein; LDL-C: low-density lipoprotein cholesterol; LDL-P: low-density lipoprotein particles.

Longitudinal associations of LDL metrics with subclinical atherosclerotic measures over the MT

In adjusted models for race, study site and time varying age, BMI, SBP, anti-lipid medications and smoking, higher total LDL-P and ApoB were associated with greater cIMT and CAC prevalence over time. Higher levels of LDL-C were significantly associated with larger cIMT in the unadjusted model but not the adjusted model, Table 3. Additional adjustment of metabolic syndrome status did not change the main results, Supplemental Table 2.

Table 3.

Longitudinal associations of LDL metrics with cIMT and CAC prevalence over the MT

LDL metricsa	cIMT, mmb		CAC ≥ 10c
	β(SE)	P value	OR (95%CI)	P value
LDL-C, mg/L
Unadjusted	0.012(0.004)	0.006	1.203(0.981 – 1.476)	0.076
Adjustedd	0.008(0.004)	0.058	1.211(0.940 – 1.561)	0.139
ApoB, mg/L
Unadjusted	0.016(0.005)	0.001	1.378(1.109 – 1.712)	0.004
Adjustedd	0.013(0.005)	0.006	1.367(1.063 – 1.758)	0.015
Total LDL-P, umol/L
Unadjusted	0.016(0.005)	0.001	1.261(1.026 – 1.550)	0.028
Adjustedd	0.011(0.004)	0.012	1.329(1.031 – 1.714)	0.028
Small LDL-P, umol/L e
Unadjusted	0.007(0.006)	0.209	1.397(0.974 – 2.004)	0.069
Adjustedd	−0.001(0.006)	0.846	1.328(0.822 – 2.144)	0.247
Large LDL-P, umol/L e
Unadjusted	0.007(0.005)	0.175	1.072(0.826 – 1.390)	0.602
Adjustedd	0.007(0.005)	0.116	1.255(0.902 – 1.745)	0.178
IDL-P, umol/L
Unadjusted	0.003(0.004)	0.523	1.023(0.844 – 1.239)	0.819
Adjustedd	0.002(0.004)	0.689	1.018(0.792 – 1.308)	0.889
LDL Size, nm
Unadjusted	0.003(0.005)	0.514	0.847(0.686 – 1.045)	0.122
Adjustedd	0.006(0.004)	0.208	0.909(0.698 – 1.183)	0.477

ApoB: apolipoprotein B; CAC: coronary artery calcium score; cIMT: carotid intimal-medial wall thickness; IDL-P: intermediate-density lipoprotein particles; LDL: low-density lipoprotein; LDL-C: low-density lipoprotein cholesterol; LDL-P: low-density lipoprotein particles.

^aData presented as increase or decrease in cIMT per 1-SD increase in each LDL metric.

^bMean(SD) number of observations for the unadjusted models: 400(8); mean(SD) number of observations for the adjusted models: 385(7).

^cMean(SD) number of observations for the unadjusted models: 373(8); mean(SD) number of observations for the adjusted models: 360(8).

^dAdjusted for race, study site, and time varying age, BMI, systolic blood pressure, anti-lipid medications, smoking status.

^eLog transformed.

Bold indicates significantly differs from zero at P ≤ 0.05.

Longitudinal associations of LDL metrics with subclinical atherosclerotic measures by menopause stage or race

Menopause significantly modified the association of LDL-C on cIMT; none of the other associations between LDL metrics and subclinical CVD measures were modified by menopause stage. In adjusted models, higher LDL-C, ApoB and total LDL-P were associated with greater cIMT during peri- or postmenopause stages, while higher LDL-C, and small LDL-P were associated with greater CAC prevalence, mainly during perimenopause stage, Table 4. Additional adjustment of metabolic syndrome status did not change the main results, Supplemental Table 3. No significant interaction was found between LDL metrics and race in relation to subclinical CVD measures, data not shown.

Table 4.

Longitudinal associations of LDL metrics and subclinical atherosclerosis measures by menopause stage.

LDL metrics	cIMT, mma				CAC ≥ 10b
	Pre-c β(SE)d	Peri-c β(SE)d	Post-c β(SE)d	Interaction P-value	Pre-c OR(95%CI)d	Peri-c OR(95%CI)d	Post-c OR(95%CI)d	Interaction P-value
LDL-C, mg/dL	−0.01(0.01)	0.01(0.01)	0.03(0.01)	0.01	1.22(0.74 – 2.01)	1.63(1.34 – 2.34)	0.79(0.49 – 1.26)	0.09
ApoB, mg/dL	−0.002(0.01)	0.01(0.01)	0.02(0.01)	0.10	1.53(0.91 – 2.56)	1.39(0.92 – 2.12)	1.26(0.63 – 2.53)	0.88
Total LDL-P, μmol/L	0.004(0.01)	0.01(0.01)	0.01(0.01)	0.58	1.45(0.91 – 2.32)	1.39(0.88 – 2.20)	1.21(0.75 – 1.95)	0.85
Small LDL-P, μmol/L e	0.003(0.01)	−0.004(0.01)	−0.01(0.01)	0.66	1.04(0.71 – 1.52)	2.04(1.03 – 4.03)	1.23(0.36 – 4.20)	0.23
Large LDL-P, μmol/L e	−0.0002(0.01)	0.01(0.01)	0.01(0.02)	0.45	1.89(1.07 – 3.30)	1.29(0.83 – 1.99)	0.61(0.21 – 1.74)	0.31
IDL-P, μmol/L	0.01(0.01)	−0.002(0.01)		0.10	0.99(0.74 – 1.34)		1.03(0.71 – 1.48)	0.87
LDL size, nm	0.01 (0.01)	−0.0001(0.01)		0.16	1.05(0.80 – 1.37)		0.84(0.59 – 1.19)	0.22

ApoB: apolipoprotein B; CAC: coronary artery calcium score; cIMT: carotid intimal-medial wall thickness; IDL-P: intermediate-density lipoprotein particles; LDL: low-density lipoprotein; LDL-C: low-density lipoprotein cholesterol; LDL-P: low-density lipoprotein particles.

^aMean(SD) number of observations for the adjusted models: 385(7).

^bMean(SD) number of observations for the adjusted models: 360(8).

^cMenopause stages were determined based on the inflection cut points identified by the LOESS plots, and each cut points reflected the years relative to the FMP. For the metrics with three segments, the cut points were: Total LDL-P: −2, 2.5; LDL-C: −2, 2.5; ApoB: −2, 2.5: Large LDL-P: −1, 4; Small LDL-P: −0.5, 2.5. The cut points for LDL metrics with two segments were: LDL size: 0; IDL-P: 1. There were no significant cut points identified to separate peri- and postmenopause stages.

^dAdjusted for race, study site, and time varying age, BMI, systolic blood pressure, anti-lipid medications, smoking status.

^eLog transformed.

Bold indicates significantly differs from zero at P ≤ 0.05.

Discussion

The current study provides unique data characterizing changes in multiple measures of low-density lipoproteins and their associations with subclinical coronary and carotid atherosclerosis over the menopause transition.

During the perimenopause stage, women experienced significant rises in LDL-C, ApoB, total LDL-P, large LDL-P and small LDL-P that were significantly higher than those seen during the premenopause stage. Moreover, during perimenopause, IDL-P significantly declined while LDL size did not change. In adjusted models, higher total LDL-P and ApoB were significantly associated with greater CAC prevalence and cIMT. Except for LDL-C, none of the other associations were modified by menopause stage. However, higher LDL-C, ApoB and total LDL-P were significantly associated with greater cIMT during peri- or postmenopause stages, while higher LDL-C, and small LDL-P were significantly associated with greater CAC prevalence, mainly during perimenopause stage. The study suggests that the reported adverse changes in low-density lipoprotein subclasses found to be associated with higher atherosclerotic risks during perimenopause, could increase women’s risk to develop CVD later in life. Whether these changes could better predict future risk of hard CVD events beyond LDL-C remains a research question to address.

Few studies have assessed the associations between LDL subclasses and menopause. These studies were limited by either the cross-sectional study design or the short follow-up time^17,24. As such, none could characterize the trajectories of LDL subclasses by anchoring the data to time elapsed since the FMP as done in this study. In an earlier and exploratory cross-sectional analysis among a small subset of SWAN women, late peri-/postmenopausal (n=34) women had more total LDL-P than pre-/early perimenopausal women (n=19)²⁵. In another study of 1,492 healthy UK midlife women assessed the associations of changes in multiple metabolic biomarkers including LDL subclasses with changes in menopause status over 2.5 years of follow up, greater increases in IDL-P, and LDL subclasses, and greater decreases in LDL size were consistently observed in all four groups with a change in menopause status compared to those who remained premenopausal over time²⁶. This study could determine to the exact stage of the MT when most changes in LDL subclasses take place since data on changes in bleeding pattern and timing of menopause were not collected. Our results align with the results from these two previous studies^25,26 supporting increases in LDL subclasses over the MT and additionally identifying perimenopause as a critical time of these changes.

We found increases in total as well as in large and small LDL-Ps during perimenopause. Interestingly, LDL size did not change during peri-/postmenopause stage. The increase in small LDL-P could confer greater CVD risk²⁷. Several large prospective studies showed that higher levels of small but not large LDL-P predicted incident CHD independent of LDL-C^28–30. Small LDL-P enriched in more atherogenic proteomics and lipidomic profiles, include an electronegative LDL species related to endothelial dysfunction, and are more susceptible to hydroperoxide formation²⁷. Cross-sectional analysis from the Healthy Women Study found a positive association of small LDL-P with CAC. Moreover, smaller LDL sizes and higher levels of total LDL-P were associated with higher CAC levels in postmenopausal women¹⁷. Consistent with that, we reported positive associations between total LDL-P and prevalence of CAC and cIMT level overtime. Moreover, we found that higher levels of small LDL-P were significantly associated with prevalence of CAC only during the perimenopause stage, however this result should be interpreted with caution given the large 95% CI. The consistency of results on LDL-P and ApoB was expected in the current study. It is well-known that each LDL-P has one ApoB molecule, and ApoB concentration directly reflects the plasma number of LDL-particles³¹.

The strengths of this study included the use of novel LDL metrics in addition to LDL-C to better understand the contribution of the MT to LDL and subclinical atherosclerosis risk, the well-characterized sample of midlife women with precise data on menopause stage, date of the final menstrual period, LDL measures, as well as other clinical characteristics repeatedly measured over the MT reducing chances for misclassification. Limitations for this study include, the relatively small sample sizes for the longitudinal analysis of each subclinical atherosclerosis measure and the lack of measures of LDL proteomics and lipidomics which could have provided more in-depth understanding of possible mechanistic pathways linking subclasses with CVD risk. Furthermore, the current analysis could not rule out residual confounders of variables that were not part of SWAN HDL ancillary study.

Although LDL-C has been a well-characterized risk factor for cardiovascular events³², the emerging concerns about the residual risks from LDL subclasses have gained more attention in the public³³. The complexity of LDL molecules and pathology, as reflected by our study and previous literature^34–36, highlights the need to better understand the composition, function and pathways that link to alleviated CVD risk. Our findings, with the aggregating body of knowledge about the novel LDL metrics, can aid identification of novel targets for LDL-based therapy. There is accumulating evidence relating lipid surrogate markers besides LDL-C, such as ApoB and total LDL-P, with management and prevention of atherosclerotic CVD. A recent meta-analysis showed significant lowered risk for cardiovascular mortality for every 10 mg/dL decrease in ApoB³⁷. Two randomized clinical trials of anti-ApoB oligonucleotides use showed significant decreases in all LDL-related parameters, such as total LDL-P, compared to the control group, among patients with higher CVD risk^38,39.

Conclusions

In conclusion, women experience significant atherosclerotic increases in LDL subclasses that increases their risk of having greater cIMT levels and higher CAC prevalence. The reported associations lend further support for the concept that the perimenopause stage is a critical stage for CVD risk acceleration and an important window of opportunity to implement early intervention strategies. The reported findings call for additional studies to investigate the prediction role, beyond LDL-C, of MT changes in LDL subclasses with future CVD events among women.