Trajectories of Blood Pressure in Midlife Women: Does Menopause Matter?

Abstract

Background:

Whether changes in blood pressure (BP) over women’s midlife are more driven by chronological aging or the menopause transition (MT) has been debated. We sought to determine whether women can be classified into distinct trajectory groups based on pattern and level of systolic (SBP), diastolic blood pressure, pulse pressure (PP), and mean arterial pressure (MAP) over the MT, and to assess whether menopause-related factors predict the group and/or level of BP measures.

Methods:

Participants were from the Study of Women’s Health Across the Nation (SWAN). Group-based trajectory modeling was used to identify women who shared distinct BP trajectories over time relative to menopause onset and to assess associations of menopause-related factors with trajectory group and/or level of BP measures. An accelerated rise relative to menopause onset suggests a menopause contribution.

Results:

The study included 3,302 multi-racial/ethnic women with BP measures over 17 follow-up visits (baseline age[SD]: 46.3[2.7]). Women were classified into either low, medium, or high trajectory group in each BP measure. The low SBP, PP, and MAP trajectories (in 35%, 53%, and 28% of the cohort, respectively) were rising slowly before menopause but showed a significant accelerated rise 1 year after menopause, indicating a menopause contribution. The remaining BP trajectories were rising up until menopause and either continued with the same rise or declined after menopause. A younger menopause age predicted the low SBP, PP, and MAP trajectories. A greater follicle-stimulating hormone level predicted lower SBP and PP levels, while vasomotor symptoms occurrence predicted higher SBP, PP, and MAP levels over time. Estradiol did not predict trajectory or level of any BP measure.

Conclusions:

Distinct BP trajectories over the MT exist that revealed a group of women whose SBP, PP, and MAP trajectories are consistent with a menopause contribution. Our findings support frequent monitoring of BP during the MT.

Graphical Abstract

INTRODUCTION

Hypertension is an important modifiable risk factor for premature cardiovascular disease (CVD), yet 43% of adult women have hypertension.^1,2 Blood pressure (BP) increases with age in both sexes, but women’s BP surpasses that of men after midlife.³ This phenomenon has generated interest in investigating the role the menopause transition (MT) may play in BP over midlife.⁴

After controlling for chronological aging, previous cross-sectional studies on the association between the MT and BP were inconclusive and findings were inconsistent.⁴ Conflicting findings may have resulted from heterogeneity among the studied populations.⁴ In fact, in White women of normal weight, postmenopausal women had a higher BP compared with their premenopausal counterparts.^5,6 However, a similar association was not observed in overweight White women.^7,8 In contrast, studies on overweight Asian populations tend to show an association between the MT and BP,^9,10 while BP was similar across MT stages in studies of normal weight Asian women.^11,12 Longitudinal studies of multi-ethnic populations mainly found a linear trajectory of BP over the MT with no obvious inflection at menopause, consistent with aging effect.^13–15 However, the literature thus far has described changes over the MT assuming that women share a common BP trajectory. With existing racial/ethnic differences in hypertension and obesity in midlife,¹ BP in women may have diverse and distinct patterns over the MT. The previous conflicting findings suggest that BP is possibly vulnerable to menopause or menopause-related factors only in women with certain characteristics.

Declines in estradiol, increases in follicle-stimulating hormone (FSH), and the occurrence of vasomotor symptoms are the hallmarks of the MT.^16,17 It has been hypothesized that these menopause-related factors and age at menopause may influence BP.^10,18–20 However, associations between these menopause-related factors and BP trajectories over the MT have not been evaluated thoroughly. Additionally, studies of the associations between the MT and BP measures mainly focused on systolic (SBP) but not diastolic blood pressure (DBP), pulse pressure (PP), or mean arterial pressure (MAP).⁴ The Study of Women’s Health across the Nation (SWAN) is a cohort study with prospective and comprehensive assessment of the MT and concurrent measurements of the associated physical and hormonal changes. Using data from SWAN, we sought to determine whether women can be classified into distinct trajectory groups based on pattern and level of SBP, DBP, PP, and MAP over time relative to the final menstrual period (FMP). A trajectory with an accelerated rise within one year of the FMP suggests a menopause contribution. We additionally sought to determine whether age at menopause, and time-varying estradiol, FSH, and vasomotor symptoms predicted the trajectory pattern and/or level of BP.

METHODS

Data Availability.

All data and materials have been made publicly available at the National Institute on Aging and can be accessed at https://agingresearchbiobank.nia.nih.gov.

Study population

SWAN is an ongoing multi-ethnic, multi-site longitudinal study designed to examine physical, biological, and psychosocial changes in women during midlife. Detailed design and methods were presented elsewhere.²¹ In brief, 3,302 women were recruited between 1996 and 1997 from seven sites (Detroit, MI; Boston, MA; Chicago, IL; Oakland, CA; Los Angeles, CA; Newark, NJ; and Pittsburgh, PA). To enroll in SWAN, women had to be aged 42–52 years, have an intact uterus and at least one ovary with menstrual bleeding within the past three months, not be pregnant or breast-feeding, and not have used hormone therapy in the past three months. Women self-identified as a member of one of five racial/ethnic groups: White (all sites), Black (Boston, Chicago, Detroit, Pittsburgh), Chinese/Chinese American (Oakland), Hispanic (Newark) or Japanese/Japanese American (Los Angeles). The institutional review board at each site approved the study protocol, and all participants signed informed consent before participation. Women in the current study were followed for up to 17 visits. Data from all SWAN women (n=3,302) were included in this analysis.

Time anchored to FMP.

At each visit, women provided the date of their most recent menstrual period which enabled retrospectively assigning the final menstrual period (FMP) date as the date of the participant’s last menstrual period before 12 consecutive months of amenorrhea. However, we did not observe the FMP in 1346 (41%) women due to hormone therapy or hysterectomy. Because SWAN is a study of the menopause transition and since hormone therapy and hysterectomy are widely used treatments in the midlife women population, FMP dates for all participants is essential to characterizing menopause-related changes and to generalizing results to women represented by SWAN. Therefore, SWAN multiply imputed the unobserved FMP dates using a wealth of demographic and longitudinal characteristics known in the literature to influence age at the FMP. Our method of imputation and analysis accounted for the uncertainty in the imputation by incorporating within and between imputation variability in the estimates.²² In addition to increased sample size and precision, the advantages of reducing bias by including both women with observed and imputed FMP dates outweigh errors due to imputing the data.²² A fuller description of the imputation process is in the Supplemental Materials.

Blood pressure measurements.

BP measurements were obtained at all visits except the 14^th follow-up visit. Women had not smoked or consumed any caffeinated beverage within 30 minutes of obtaining BP measurements. BP measurements were taken after at least 5 minutes of rest using the appropriate cuff size based on arm circumference. The participants were seated with feet flat on the floor (legs uncrossed) and refraining from talking during the measurements. A standard mercury sphygmomanometer was used to record SBP and DBP at the first and fifth phase Korotkoff sounds, respectively. Two sequential values were recorded with a 2-minute period between measurements. The average of the two measures was used in this analysis. To account for the effect of BP medications on BP values in visits when women reported using anti-hypertensive medications, we added 10 and 5 mmHg to the observed SBP and DBP values, respectively.^3,23 PP, a BP measure reflecting arterial compliance, was calculated as: PP= SBP-DBP. MAP measures the average arterial pressure throughout one cardiac cycle and was calculated as: MAP= DBP+(PP/3). Measures of BP were treated as time-varying variables in analyses.

Blood assays.

To allow for a standardized hormonal milieu, a fasting blood sample was drawn during the early follicular phase of the menstrual cycle (day 2–5) if women were still menstruating. If a timed sample could not be obtained, because menstrual cycles became less regular over time or due to menopause, a random fasting sample was taken within 90-days of the corresponding visit. All samples were maintained at 4°C until separated and then were frozen at −80°C and shipped in dry ice to the central certified laboratory (Medical Research Laboratories, Highland Heights, Kentucky).

Endogenous sex hormones were measured at the University of Michigan Endocrine Laboratory using the Automated Chemilumisence System −180 automated analyzer (Bayer Diagnostics Corp., Norwood, MA). Estradiol was measured using a modified, off-line Automated Chemilumisence System: 180 (E2–6) at the baseline visit, the 1^st–10^th, 12^th, and 13^th follow-up visits. The lower limit of detection (LLD) was between 1 and 7 pg/mL. The inter- and intra-assay coefficients of variation were 10.6% and 6.4%, respectively.²⁴ FSH was measured by a modification of a manual assay kit (Bayer Diagnostics) utilizing two monoclonal antibodies directed to different regions on the beta subunit at the baseline visit, the 1^st–10^th, 12^th, 13^th, and 15^th follow-up visits. The LLD was between 0.4–1.0 mIU/mL. The inter-and intra-assay coefficients of variation were 11.4% and 3.8%, respectively.²⁴ Since estradiol and FSH show dynamic changes throughout the menstrual cycle, cycle-day of blood draw (within 2–5 days of the menstrual cycle or not) was considered when analyzing these hormones. Estradiol and FSH were treated as time varying variables in analyses.

Total cholesterol was analyzed using enzymatic methods on a Hitachi 747 analyzer (Boehringer Mannheim Diagnostics, Indianapolis, IN), and high-density lipoprotein cholesterol was isolated using heparin-manganese. Low-density lipoprotein cholesterol (LDL-C) was calculated using Friedewald equation.²⁵ LDL-C was obtained at baseline, 1^st, 3^rd–7^th, 9^th, 12^th,13^th, and 15^th follow-up visits and was treated as a time varying variable in analyses.

Study covariates.

At each visit, participants completed self and interviewer-administered questionnaires that included assessment of sociodemographic and life-style factors and medical history. Age, race/ethnicity, education, income, financial strain, smoking status, alcohol consumption, presence of hot flashes or night sweats, hormone therapy, use of anti-diabetic medications, and steroids were self-reported. BP medications were self-reported and interviewer-verified from medication container labels. Body mass index (BMI) was calculated as weight/height². Hot flashes and night sweats were measured at all visits except the 16^th follow-up visit. Women reporting either hot flashes or night sweats were considered as having vasomotor symptoms. A detailed description of diabetes definition is presented in the Supplemental Materials. Vasomotor symptoms and diabetes were treated as time-varying variables. Unless specified, other study covariates were obtained from the baseline visit. Please see the Major Resources Table in the Supplemental Materials.

Statistical analysis.

To disentangle the MT contribution to BP from chronological aging, time in years was anchored to the FMP.¹⁴ Data points 13 years before and 15 years after the FMP were dropped from the analysis due to a small number of observations at these time points. We used group-based trajectory modeling (GBTM) to detect distinct trajectories of SBP, DBP, PP, and MAP over the MT, separately. GBTM uses finite mixture modeling to identify distinct groups that share common developmental trajectories while accounting for classification uncertainty.²⁶ To allow for flexibility in estimating the shapes of BP trajectories, we tested linear, quadratic, cubic, and quartic terms of time and retained the highest term, when significant, with the corresponding lower terms.²⁶ We determined the number of groups based upon Bayesian Information Criterion, reasonable scientific plausibility, and whether model with an additional group identified further trajectory features. For simplicity in building this classification model, no covariate was included at this stage of the analysis. Model fit statistics are presented in the Supplemental Materials. For women with imputed FMP dates, we used the mean of the multiply imputed FMP dates for the GBTM.

Within each SBP, DBP, PP, and MAP trajectory group identified via GBTM, a piecewise-linear pattern with an inflection point and an accelerated rise within one year of the FMP was suggestive of a menopause contribution. Potential inflection points around the FMP were identified via plotting repeated measures of SBP, DBP, PP, and MAP in each trajectory group using locally weighted scatterplot smoothing (LOWESS). A piecewise-linear random coefficients model was then used to test the year of inflection initially identified via LOWESS. If no significant inflection point was identified, a linear model was utilized. The random coefficients model adjusted for the fixed effects of age at menopause, race/ethnicity, study site, and baseline BMI. Established techniques of analyzing multiply imputed data were used in the random coefficients model.²²

We explored predictive associations of menopause-related factors with BP trajectories that showed a significant accelerated rise within one year of the FMP. Using a single GBTM with the same specification (number of respective BP groups and terms of time), we used multinomial logistic and multiple linear regression to estimate the odds ratio of trajectory group membership for baseline variables and beta coefficient of BP level for time-varying covariates, respectively. This model adjusted for race/ethnicity, study site, and baseline BMI, and time-varying hormone therapy, LDL-C, and diabetes. Menopause-related factors hypothesized to predict membership in a BP trajectory group were age at menopause (categorized into: <51 years, ≥51 and ≤53 years, and >53 years) and baseline estradiol, FSH, and vasomotor symptoms occurrence. Menopause-related factors hypothesized to predict BP measurement levels within each trajectory group were time-varying estradiol, FSH, and vasomotor symptoms occurrence. When we ran the model that explored predictive associations of menopause-related factors with PP trajectories, estimates of race were unstable due to small cell count and, therefore, we combined Chinese, Hispanic, and Japanese into “Other”.

We compared baseline characteristics between women with observed and imputed FMP dates and conducted sensitivity analyses restricted to women with an observed FMP. We also ran models without adding 10 and 5 mmHg to the observed SBP and DBP values, respectively, when women reported using anti-hypertensive medications. Additionally, as sensitivity analyses, we adjusted the classification and the random coefficients model for baseline age and respective BP measure. All analyses were performed with SAS 9.4 (SAS Institute, Cary NC) with a significance level set at 0.05.

RESULTS

Distinct blood pressure trajectories.

Using 32,967 observations of data spanning 28 years across the MT with a median follow-up of 19.1 years, women were classified into three distinct trajectory groups in SBP, DBP, PP, and MAP (figure 1).

Figure 1. Trajectories of blood pressure over the menopause transition (n=3302).

Blood pressure represents the predicted mean at each time point within each trajectory group from group-based trajectory modelling. No factors were included in the model. Bands represent 95% CI. Panel A: systolic blood pressure; Panel B: diastolic blood pressure; Panel C: pulse pressure; Panel D: mean arterial pressure.

The low-accelerated rise SBP, PP, and MAP trajectories (in 35%, 53%, and 28% of the cohort, respectively) were rising slowly before the FMP but showed an accelerated rise 1 year after the FMP. The low-linear rise DBP trajectory (28%), all the medium-linear rise trajectories (SBP: 48%, DBP: 49%, PP: 38%, and MAP: 48%), and the high-continued rise PP trajectory (9%) showed a monotonic rise over the MT. The high-slow decline SBP (17%) and the high-accelerated decline DBP (23%) and MAP (24%) trajectories were rising before and showed a slow and an accelerated decline after the FMP, respectively.

White, Chinese, and Japanese women were more likely to follow the low-accelerated rise SBP, PP and MAP trajectories. Hispanic women were more likely to follow the medium-linear rise SBP, PP, and MAP trajectories. Black women were more likely to follow the high-slow decline SBP, the high-continued rise PP, and the high-accelerated decline MAP trajectories. Women in the low-accelerated rise SBP, PP, and MAP trajectories had a borderline normal weight with a BMI of 24.6, 25.9, and 24.6 kg/m², respectively. Table 1 and tables S1–S3 show baseline characteristics by SBP, DBP, PP, and MAP trajectory groups, respectively.

Table 1.

Baseline characteristics of study participants by SBP trajectory group

Variable*	Total, n=3302	Low-accelerated rise 1172 (35%)	Medium-linear rise 1574 (48%)	High-slow decline 556 (17%)	P-value†
Age (years)	46.34 ± 2.69	46.04 ± 2.60	46.39 ± 2.67	46.82 ± 2.84	6.67×10−8
Race/ethnicity, N (%)					1.00×10−30
White	1551 (47.0)	709 (60.5)	710 (45.1)	132 (23.7)
Black	934 (28.3)	150 (12.8)	450 (28.6)	334 (60.1)
Chinese	250 (7.6)	134 (11.4)	96 (6.1)	20 (3.6)
Hispanic	286 (8.7)	25 (2.1)	209 (13.3)	52 (9.4)
Japanese	281 (8.5)	154 (13.1)	109 (6.9)	18 (3.2)
Education, N (%)					2.60×10−29
Less than high school	1551 (47.0)	709 (60.5)	710 (45.1)	132 (23.7)
Some College	934 (28.3)	150 (12.8)	450 (28.6)	334 (60.1)
College Degree or higher	250 (7.6)	134 (11.4)	96 (6.1)	20 (3.6)
Income, N (%)					6.34×10−27
<$35,000	1004 (31.3)	239 (20.9)	522 (34.1)	243 (45.1)
$35,000-$75,000	1308 (40.7)	491 (43.0)	615 (40.1)	202 (37.5)
>$75,000	900 (28.0)	411 (36.0)	395 (25.8)	94 (17.4)
Financial Strain, N (%)	1312 (40.0)	356 (30.5)	669 (42.9)	287 (51.8)	2.00×10−18
Smoking Status, N (%)					2.82×10−8
Never	1890 (57.3)	702 (60.0)	907 (57.6)	281 (50.5)
Past	830 (25.2)	309 (26.4)	391 (24.8)	130 (23.4)
Current	580 (17.6)	159 (13.6)	276 (17.5)	145 (26.1)
Alcohol consumption (drinks per week), N (%)					0.00049
None	1553 (49.6)	524 (47.1)	727 (48.4)	302 (58.1)
<1 drinks	326 (10.4)	106 (9.5)	172 (11.5)	48 (9.2)
1–7 drinks	813 (25.9)	304 (27.3)	401 (26.7)	108 (20.8)
>7 drinks	441 (14.1)	178 (16.0)	201 (13.4)	62 (11.9)
Systolic blood pressure (mmHg)	117.85 ± 17.03	104.72 ± 8.49	119.37 ± 10.96	141.22 ± 17.73	1.00×10−30
Diastolic blood pressure (mmHg)	75.48 ± 10.52	68.89 ± 7.63	76.87 ± 8.77	85.38 ± 11.07	1.00×10−30
Pulse pressure (mmHg)	42.38 ± 12.07	35.83 ± 7.23	42.50 ± 9.68	55.82 ± 14.79	1.00×10−30
Mean arterial pressure (mmHg)	89.60 ± 11.75	80.84 ± 7.15	91.03 ± 8.40	103.99 ± 11.75	1.00×10−30
Blood pressure medication, N (%)	471 (14.3)	29 (2.5)	184 (11.7)	258 (46.4)	1.00×10−30
Body mass index (kg/m2)	28.26 ± 7.21	24.63 ± 4.92	29.24 ± 6.94	33.24 ± 8.15	1.00×10−30
Diabetes, N (%)	152 (4.6)	10 (0.9)	74 (4.7)	68 (12.2)	7.07×10−25
Low-density lipoprotein cholesterol (mg/dl)	116.11 ± 31.01	111.47 ± 29.10	118.35 ± 30.76	119.81 ± 34.41	2.10×10−9
Age at FMP (years)	52.15 ± 2.89	51.88 ± 2.71	52.34 ± 2.95	52.18 ± 3.05	0.00024
Vasomotor symptoms, N (%)	1287 (39.2)	377 (32.4)	631 (40.3)	279 (50.5)	4.18×10−12
Estradiol (pg/mL), Median (Q1, Q3)	55.2 (33.0, 88.7)	57.6 (34.2, 89.6)	55.2 (32.8, 88.4)	50.3 (31.8, 85.2)	0.15
Follicle-stimulating hormone (mIU/mL), Median (Q1, Q3)	15.9 (10.8, 26.4)	16.2 (10.9, 27.2)	15.7 (10.7, 26.1)	16.2 (10.9, 25.5)	0.23
Hormone therapy, N (%)	6 (0.2)	1 (0.1)	4 (0.3)	1 (0.2)	0.59

^*Mean ± SD is presented unless otherwise specified

^†Analysis of variance (ANOVA) was used to compare means for all continuous normally distributed variables. Kruskal-Wallis test was used to compare medians for estradiol and follicle-stimulating hormone. Chi-square test was used to compare frequency distributions for categorical variables

BP group-specific pattern: linear vs piecewise-linear change.

The low-accelerated rise SBP, PP, and MAP trajectory groups showed a piecewise-linear pattern with a significant inflection point and an accelerated rise 1 year after the FMP, consistent with an MT contribution (table 2). However, the other two SBP, PP, and MAP trajectories and the three DBP trajectories were not MT-related as they showed a linear pattern or a piecewise-linear pattern with a decline close to the FMP.

Table 2.

Annual Changes in Blood Pressure measures (mmHg) in Time Segments, by Trajectory Groups, (n=3302) ^*

Blood pressure measure	Trajectory Group	n	Inflection point	Change in blood pressure (95% CI)		P-Value for difference
				Before inflection	After inflection
Systolic	High-slow decline	556	FMP	0.50 (0.24, 0.76)	−0.12 (−0.32, 0.08)	0.0010
	Medium-linear rise	1574	N/A	0.67 (0.63, 0.72)		N/A
	Low-accelerated rise	1172	1 year after FMP	0.23 (0.16, 0.29)	0.77 (0.71, 0.84)	5.43×10−24
Diastolic	High-accelerated decline	770	FMP	0.10 (−0.02, 0.21)	−0.35 (−0.42, −0.27)	4.97×10−8
	Medium-linear rise	1613	N/A	0.13 (0.10, 0.16)		N/A
	Low-linear rise	919	N/A	0.13 (0.10, 0.16)		N/A
Pulse pressure	High-continued rise	303	N/A	0.34 (0.18, 0.50)		N/A
	Medium-linear rise	1250	N/A	0.59 (0.55, 0.63)		N/A
	Low-accelerated rise	1749	1 year after FMP	0.22 (0.17, 0.26)	0.54 (0.49, 0.58)	1.36×10−17
Mean arterial pressure	High-accelerated decline	781	FMP	0.24 (0.10, 0.38)	−0.25 (−0.34, −0.16)	5.35×10−7
	Medium-linear rise	1593	N/A	0.32 (0.29, 0.35)		N/A
	Low-accelerated rise	928	1 year after FMP	0.12 (0.06, 0.17)	0.39 (0.34, 0.43)	3.08×10−10

^*Models adjusted for age at the FMP, site, race/ethnicity, and SWAN baseline body-mass index

Menopause-related predictors of SBP, PP, and MAP trajectories.

Relative to the medium-linear rise group (aging-related trajectory) and women with age at menopause between 51 and 53 years (a range encompassing the mean of 52.15 years), women with a younger age at menopause and White women were more likely to follow the low-accelerated rise SBP, PP, and MAP trajectories (table 3, table S4, and table 4). Vasomotor symptoms occurrence predicted a higher BP level over time within the low-accelerated rise and medium-linear rise SBP and PP trajectories and in all MAP trajectories. Greater FSH predicted a lower BP level over time within the medium-linear rise SBP and in all PP trajectories. Baseline or time-varying estradiol did not predict BP trajectory groups or levels, respectively.

Table 3.

Predictors of SBP trajectory group membership and level across the menopause transition, (n=3302) ^*

Variables‡	SBP trajectory group, n (%)
	Low-accelerated rise 1172 (35%)	Medium-linear rise 1574 (48%)	High-slow decline 556 (17%)
Predictors of SBP trajectory group membership, OR (95% CI) §
Race/ethnicity
White	1 [ref]	[ref]	1 [ref]
Black	0.30 (0.22, 0.42)	[ref]	3.08 (2.28, 4.17)
Hispanic	0.29 (0.08, 1.13)	[ref]	3.33 (1.11, 10.01)
Japanese	0.44 (0.27, 0.70)	[ref]	1.63 (0.69, 3.86)
Chinese	0.37 (0.23, 0.61)	[ref]	2.84 (1.03, 7.84)
Baseline BMI (per 1-SD increase)	0.33 (0.28, 0.39)	[ref]	1.50 (1.33, 1.70)
Age at menopause
Before 51 years	1.73 (1.31, 2.28)	[ref]	1.06 (0.78, 1.45)
Between 51 and 53 years	1 [ref]	[ref]	1 [ref]
After 53 years	0.75 (0.58, 0.98)	[ref]	1.34 (1.00, 1.80)
Baseline estradiol† (per 1-SD increase)	1.01 (0.89, 1.14)	[ref]	1.11 (0.96, 1.28)
Baseline FSH† (per 1-SD increase)	1.00 (0.87, 1.14)	[ref]	1.10 (0.95, 1.26)
Baseline vasomotor symptoms	0.80 (0.64, 1.00)	[ref]	1.21 (0.95, 1.54)
Predictors of SBP level over time within each trajectory group, beta (SE) §
Estradiol† (per 1-SD increase)	0.11 (0.12)	−0.12 (0.13)	0.14 (0.23)
FSH† (per 1-SD increase)	−0.16 (0.10)	−0.39 (0.12)	−0.06 (0.22)
Vasomotor symptoms	0.92 (0.30)	1.17 (0.30)	0.96 (0.52)
Hormone therapy use	0.65 (0.43)	0.57 (0.42)	−1.57 (0.84)
LDL-C (per 1-SD increase)	0.39 (0.16)	0.30 (0.16)	0.42 (0.24)
Diabetes	4.93 (0.93)	5.36 (0.56)	4.92 (0.80)

^*Bold indicates P-value <0.05. Actual p-values are provided in table S12

^†Log-transformed

^‡Model further adjusted for site and cycle-day of blood draw. There were 107 (3%) women excluded from this model due to missing covariate values (BMI [42], baseline estradiol [8], baseline FSH [1], baseline vasomotor symptoms [18], LDL-C [38])

^§Estimates of SBP trajectory group membership and level are from the same model

LDL-C: low-density lipoprotein cholesterol, SBP: systolic blood pressure

Table 4.

Predictors of MAP trajectory group membership and level across the menopause transition, (n=3302) ^*

Variables‡	MAP trajectory group, n (%)
	Low-accelerated rise 928 (28%)	Medium-linear rise 1593 (48%)	High-accelerated decline 781 (24%)
Predictors of MAP trajectory group membership, OR (95% CI) §
Race/ethnicity
White	1 [ref]	[ref]	1 [ref]
Black	0.36 (0.25, 0.50)	[ref]	3.26 (2.44, 4.36)
Hispanic	0.28 (0.05, 1.66)	[ref]	2.99 (1.29, 6.91)
Japanese	0.52 (0.33, 0.83)	[ref]	2.06 (0.97, 4.38)
Chinese	0.50 (0.31, 0.81)	[ref]	5.23 (2.11, 12.95)
Baseline BMI (per 1-SD increase)	0.36 (0.31, 0.43)	[ref]	1.40 (1.25, 1.57)
Age at menopause
Before 51 years	1.37 (1.04, 1.80)	[ref]	0.91 (0.68, 1.21)
Between 51 and 53 years	1 [ref]	[ref]	1 [ref]
After 53 years	0.36 (0.25, 0.50)	[ref]	1.10 (0.84, 1.43)
Baseline estradiol† (per 1-SD increase)	0.28 (0.05, 1.66)	[ref]	1.07 (0.94, 1.21)
Baseline FSH† (per 1-SD increase)	0.52 (0.33, 0.83)	[ref]	0.97 (0.85, 1.11)
Baseline vasomotor symptoms	0.50 (0.31, 0.81)	[ref]	1.17 (0.94, 1.47)
Predictors of MAP level over time within each trajectory group, beta (SE) §
Estradiol† (per 1-SD increase)	0.03 (0.09)	−0.04 (0.09)	0.13 (0.14)
FSH† (per 1-SD increase)	−0.04 (0.07)	−0.09 (0.08)	0.08 (0.13)
Vasomotor symptoms	0.86 (0.22)	0.84 (0.20)	1.29 (0.31)
Hormone therapy use	0.63 (0.31)	0.13 (0.29)	−1.15 (0.48)
LDL-C (per 1-SD increase)	0.35 (0.11)	0.44 (0.10)	0.49 (0.15)
Diabetes	3.85 (0.57)	2.70 (0.39)	2.41 (0.49)

^*Bold indicates P-value <0.05. Actual p-values are provided in table S12

^†Log-transformed

^‡Model further adjusted for site and cycle-day of blood draw. There were 107 (3%) women excluded from this model due to missing covariate values (BMI [42], baseline estradiol [8], baseline FSH [1], baseline vasomotor symptoms [18], LDL-C [38])

^§Estimates of pulse pressure trajectory group membership and level are from the same model

LDL-C: low-density lipoprotein cholesterol, MAP: mean arterial pressure

Sensitivity analyses.

Women with imputed and observed FMP showed small differences in baseline demographic variables, smoking and BP variables (table S5). Analyses restricted to women with observed FMP resulted in similar conclusions (figure S1 and tables S6–S8). Not adding 10 and 5 mmHg to the observed SBP and DBP values, respectively, when women reported using anti-hypertensive medications resulted in similar overall conclusions (figure S2 and tables S9–S11). Our classification and random coefficients models showed comparable findings after adjusting for baseline age and respective BP measure (data not shown).

DISCUSSION

Using data from SWAN, one of the largest and longest study-to-date of the menopause transition in the US, we identified three distinct groups of women based on their common trajectories of SBP, DBP, PP, and MAP. We identified a group of women, the low-accelerated rise SBP, PP, and MAP trajectory group, whose trajectory showed a piecewise-linear increase with an inflection point 1 year after the FMP consistent with an MT increase. These findings are in line with previous studies of the MT that showed progressive decrements in arterial compliance between the pre- and the peri-menopausal periods and accelerated increases in arterial stiffness 1 year before the FMP.^27,28 In women who followed a linear trajectory, however, BP was mainly consistent with a chronological aging effect. Menopause-related factors showed significant associations with BP trajectories, such that age at menopause before 51 years predicted the low-accelerated rise BP trajectories, greater FSH predicted a lower SBP and PP levels, and vasomotor symptoms occurrence predicted a higher SBP, PP, and MAP levels during the MT. Contrary to previous studies that found a linear BP trajectory over the MT consistent with aging effect, the current findings suggest that women cluster in different BP trajectories that differ in level, pattern, and relationship with menopause and its related factors. Our findings highlight the period of the MT when SBP, PP, and MAP are likely to show accelerated increases. Therefore, frequent and timely monitoring of CVD risk factors in women transitioning through menopause represents a window for counselling about lifestyle changes.^29,30

Although the low-accelerated rise SBP, PP, and MAP trajectories are consistent with a menopause contribution, that does not directly indicate that women following these trajectories are at high risk of elevated SBP or MAP or PP widening complications. However, previous studies have shown a graded log-linear association between SBP and CVD death down to a SBP of 115 mmHg without a clear threshold.³¹ Additionally, among healthy midlife men and women without previous histories of hypertension or other traditional CVD risk factors, the risk of incident CVD showed a stepwise increase with increasing SBP starting as low as 90 mmHg.³² Interestingly, sex-specific analysis of 27,542 community-based participants (54% women) showed that SBP of 100 to 109 mmHg relative to SBP <100 mmHg was associated with a greater CVD risk in women (hazard ratio 1.25) but not men, in whom the same risk was seen at SBP 130 to 139 mmHg.³³ The increased CVD risk associated with SBP increases also holds across a broad age spectrum, starting as young as 30 years old.³⁴ Whether the acceleration in SBP and MAP or widening of PP one year after the FMP in women following the low-accelerated rise trajectory predicts future clinical complications remains to be explored.

Women following the SBP high-slow decline and the DBP and MAP high-accelerated decline trajectories showed interesting phenomena of plateau and decrease after the FMP, respectively. We hypothesized that these women may have received more aggressive antihypertensive therapy after menopause, which may explain this plateau/decrease in BP measures. However, exploratory analyses including adjusting for aggressive antihypertensive therapy did not confirm our hypothesis. A previous analysis from the National Health and Nutrition Examination Survey noticed a similar SBP plateauing in Black and Mexican American women and a similar DBP decrease in White, Black, and Mexican American women around the age 60.³⁵ Our study may extends previous observations into a longitudinal context by showing that Black women with high SBP, who tended to follow the high-slow decline trajectory, may experience a plateau in SBP after menopause. Although requiring further investigation, this plateau in the high-slow decline SBP trajectory appears to be a different phenomenon from the late-life SBP decrease which greatly steeps around age 80.³⁶

In other work, greater serum levels of endogenous estradiol during the menstrual cycle or pregnancy are inversely related to SBP.³⁷ Additionally, surgically induced menopause increases SBP within a few weeks.³⁸ Thus, it is possible that estradiol depletion during the MT would be associated with a higher SBP. However, our results did not support this hypothesis. In fact, previous studies found no link between endogenous estradiol and SBP level or risk of incident hypertension.^39,40 It is hypothesized that the vasodilatory and cardio-protective actions of endogenous estradiol are mediated in part through its metabolites.⁴¹ 2-methoxyestradiol is an estradiol metabolite that has estrogen receptor-independent growth inhibitory effects on smooth muscle and endothelial cells.⁴¹ Interestingly, estradiol metabolites, but not estradiol, were associated with a lower SBP in postmenopausal women.⁴² These results highlight the complexity of the MT and future studies should compare estradiol metabolite levels and its potential effects on vascular function across the stages of the MT.

Age at menopause showed no,¹¹ or even inverse,¹⁰ associations with BP levels in previous cross-sectional studies. In one longitudinal study, the rate of change in SBP over the MT did not differ among women with different ages at menopause.⁴³ However, our results did not align with previous studies, as we showed that age at menopause has a direct association with SBP, PP, and MAP trajectory groups, such that a younger age at menopause predicted the low-accelerated rise SBP, PP, and MAP group. This disagreement across the studies may have resulted from the potential recall bias of age at menopause in cross-sectional studies or from our different analytic approach that assumed the presence of distinct BP trajectories over the MT.

Women transitioning through menopause frequently report vasomotor symptoms.⁴⁴ A previous analysis in SWAN showed that premenopausal vasomotor symptoms predicted a higher risk of incident hypertension.¹⁸ Although the exact mechanism linking vasomotor symptoms with BP is not completely understood, evidence of greater sympathetic nervous system activation in women reporting vasomotor symptoms may be one plausible mechanism.⁴⁵ Our findings extend the previous SWAN analysis by showing vasomotor symptoms predicting a higher SBP, PP, and MAP over time regardless of the trajectory.¹⁸

In women, FSH is known for its gonadal functions of follicular growth initiation and maturation. It is hypothesized that FSH has extragonadal functions after finding FSH receptors being expressed on blood vessels,⁴⁶ adipose tissues,⁴⁷ and the liver.⁴⁸ In fact, greater FSH predicted a lower SBP in a cross-sectional analysis among Chinese women.¹⁹ We showed a similar inverse association between FSH and SBP and PP over time. It is hypothesized that FSH has an angiogenesis property that is mediated through increasing vascular endothelial growth factor (VEGF).⁴⁹ In turn, VEGF showed an inverted U-shaped association with CVD events with the lowest risk experienced at the lower and upper end of the distribution.⁵⁰ However, it is not understood how FSH may reduce SBP and PP; the underlying pathological mechanisms remain to be elucidated in future studies.

This work includes some limitations. We did not observe the FMP in 41% of women. Additionally, for women without an observed FMP, GBTM does not offer the option of including uncertainty of the imputation in determining trajectories of BP. However, a sensitivity analysis excluding women without observed FMP did not alter our conclusions. By follow-up visit 14, >99% of SWAN women had reached their FMP, and thus we defined attrition as women lost to follow-up before visit 14. We had a 28% attrition rate and BP trajectory groups were associated with attrition. However, relaxing the assumption that attrition is independent of the trajectories by building the missing data mechanism into the GBTM resulted in a similar group sizes and shapes. As suggested by Haviland et al.⁵¹, when trajectory groups are well separated, as in our case, attrition will effectively be at random, and building the missing data structure into the model is not needed. Conclusions related to the model that explored predictive associations of menopause-related factors with PP trajectories were limited after combining Chinese, Hispanic, and Japanese into “Others” due to small cell count in the high-continued rise PP group. Findings of temporal associations between the MT and BP measures do not imply causal relationships. Additionally, our findings are considered hypothesis generating and are susceptible to type-1 error. Future confirmatory studies should adjust for multiple testing.

Major strengths of our study include the careful prospective characterization of the timing of the FMP, concurrent measurements of endogenous sex hormones and menopause-related factors, the multiethnic composition of SWAN, and the analytic technique that allowed linking menopause related factors with distinct SBP, PP, and MAP trajectories.

Conclusions.

Distinct patterns of BP trajectories over the MT exist that revealed a group of women whose SBP, PP, and MAP trajectories are consistent with a menopause contribution to BP over midlife. Our findings support frequent monitoring of BP during the MT. Additionally, menopause-related factors including age at menopause, vasomotor symptoms, and FSH predicted SBP, PP, and MAP pattern and level over time. Future work should investigate potential mechanisms by which the MT may accelerate SBP, PP, and MAP increase.

NOVELTY AND SIGNIFICANCE.

What Is Known?

• Hypertension contributes to cardiovascular disease (CVD) and 43% of adult women are known to have hypertension.

• After midlife, blood pressure (BP) in women exceeds that of men, suggesting a contribution of the menopause transition.

• Declines in estradiol, increases in follicle-stimulating hormone (FSH), and the occurrence of vasomotor symptoms are the hallmarks of the menopause transition and are hypothesized to influence BP.

• However, the literature thus far has described global BP changes over the menopause transition assuming that women share a common BP trajectory.

What New Information Does This Article Contribute?

• Based on multiple years of BP measurements, women were classified based on the level and shape of their BP trajectory into three distinct groups: low, medium, and high.

• Only the low trajectories demonstrate a contribution of the menopause transition with a significant accelerated rise occuring 1 year after menopause.

• The other BP trajectories were rising before the menopause transition and either continued with same rise or declined after menopause.

• Age at menopause, vasomotor symptoms, and FSH, but not estradiol, predict SBP, PP, and MAP level and shape.

The literature on the contribution of the menopause transition to BP changes in midlife is not consistent. Some studies show that postmenopausal women had a higher BP compared with their premenopausal counterparts, while others showed comparable BP measures before and after menopause. Heterogeneity among the previously studied populations may have resulted in the conflicting findings. To study this question, women were clustered into different trajectories based on level and shape of their BP measures. Our study revealed a group of midlife women whose trajectory is consistent with a menopause contribution (the low trajectory), associated with an accelerated rise after menopause in SBP, PP, and MAP measurements. Women with earlier menopause are more likely to experience the low trajectory pattern. Both greater frequency of vasomotor symptoms and lower levels of follicle stimulating hormones are related to increases in BP levels over time. Our results support the timely monitoring of CVD risk factors prior to the menopause transition and the critical need to adopt heart healthy lifestyle as early as possible.

Nonstandard Abbreviations and Acronyms

BP

blood pressure

FMP

final menstrual period

FSH

follicle-stimulating hormone

GBTM

group-based trajectory modeling

LDL-C

low-density lipoprotein cholesterol

LLD

lower limit of detection

LOWESS

locally weighted scatterplot smoothing

MAP

mean arterial pressure

MT

menopause transition

PP

pulse pressure

SWAN

Study of Women’s Health across the Nation

VEGF

vascular endothelial growth factor