Changes in Regional Fat Distribution and Anthropometric Measures Across the Menopause Transition

Abstract

Context

The relation between the menopause transition (MT) and changes in regional fat distribution is uncertain.

Objective

To determine whether the MT is associated with the development of central adiposity.

Design

Longitudinal analysis from the Study of Women’s Health Across the Nation, spanning 1996-2013 (median follow-up 11.8 years).

Setting

Community-based.

Participants

380 women with regional body composition measures by dual energy X-ray absorptiometry. Mean baseline age was 45.7 years; racial/ethnic composition was 16% Black, 41% Japanese and 43% White.

Outcomes

Changes in android, gynoid and visceral fat and waist and hip circumferences.

Results

Android fat increased by 1.21% per year (py) and 5.54% py during premenopause and the MT, respectively (each P < 0.05). Visceral and gynoid fat began increasing at the MT, annualized changes were 6.24% and 2.03%, respectively (each P < 0.05). Postmenopausal annual trajectories decelerated to 1.47% (visceral), 0.90% (android), and -0.87% (gynoid), (all non-zero, P < 0.05). Waist girth grew during premenopause (0.55% py), the MT (0.96% py), and postmenopause (0.55% py) (all non-zero, P < 0.05; not statistically different from each other). Hip girth grew during premenopause (0.20% py) and the MT (0.35% py) (each non-zero, P < 0.05; not statistically different from each other) and decelerated to zero slope in postmenopause. Results are for the White referent; there were statistically significant differences in some trajectories in Black and Japanese women.

Conclusions

The MT is associated with the development of central adiposity. Waist or hip circumferences are less sensitive to changes in fat distribution.

menopause, body composition, anthropometrics, race, ethnicity

That obesity increases the risk of chronic conditions such as cardiovascular disease, metabolic disorders, and cancer is well-established (1). Recent investigations have identified the role of regional fat distribution, apart from general obesity, as a risk factor for negative health outcomes (2-5). For example, independent of total body fat or body mass index, greater visceral adipose tissue increases the risks of cardiovascular disease, diabetes and cancer (2, 4, 5).

The Study of Women’s Health Across the Nation (SWAN) found that the menopause transition (MT) led to accelerated gains in total body fat mass and simultaneous losses in total body lean mass, thereby contributing to women’s obesity risk (6). However, the question remains: “Does the MT also influence regional fat distribution?” It is plausible that the MT, and its corresponding alterations in sex steroid and gonadotropin levels, promotes a shift towards male-typical central obesity (7-9). While most human investigations of fat distribution and menopause support the hypothesis that the postmenopausal state is associated with greater amounts of central (android or visceral) fat (10-15), data emanate from cross-sectional reports (10, 11, 14, 15) or small longitudinal studies with short follow-up periods and few, observed transitions from premenopause to postmenopause (12, 13, 16). Larger, longer term, longitudinal studies such as SWAN can address this knowledge gap.

Therefore, the overarching goal of this analysis is to determine whether the MT influences regional (android, gynoid, and visceral) fat distribution and anthropometric measures (waist and hip circumferences). Using longitudinal data from the Los Angeles and Boston Sites of SWAN, we examine these outcomes in relation to the number of years before or after the final menstrual period (FMP). The premise underlying this analytic approach is that trajectories of MT-related characteristics will change abruptly before the date of the FMP and/or will slow down after it, thereby supporting a relation between the MT and that phenotype (6, 17). We aim to describe (if present): 1) the timing of onset and offset of accelerated increases or decreases in regional fat distribution (visceral, android, and gynoid) and anthropometric measures (waist and hip circumferences) in relation to the date of the FMP; 2) if accelerated gains or losses are found, to quantify the rate and amount of each during the interval spanning 5 years before to 7 years after the FMP; and 3) whether the starting levels of each outcome, race/ethnicity, or age at FMP influence rates of change in each of the outcomes.

Material and Methods

Study sample

SWAN is a multi-site, community-based, longitudinal cohort study, which had the following baseline eligibility criteria: aged between 42 and 52 years, intact uterus and ≥1 intact ovary, not using HT, ≥1 menses in the 3 months prior to screening, and self-identification with one of 5 eligible ethnic/racial groups (18). Seven SWAN clinical sites enrolled 3302 participants, 5 sites measured whole body composition (N = 2349). Initial visits occurred in 1996-1997 and follow-up visits began in 1998. The current analysis extends through the 13^th follow-up visit, which finished in 2013.

To provide regional body composition data for this project, 2 of the 5 sites, Boston and Los Angeles, retrieved and further analyzed archived whole body composition scans. We attempted to re-analyze body composition scans for regional analysis only from women who had a known FMP date. Thus, the maximum number of potential participants was 518 (Fig. 1). Among these women, 45 had body composition scans that were irretrievable, principally due to corrupt media, and regional readings on 5 of their scans were not usable (due to technical issues, such as boundary line placement). Finally, participants had to have at least 2 usable regional body composition readings; 460 women had ≥2 retrievable and analyzable body composition scans. In addition, because we normalized outcomes to their values closest to 3 years prior to the FMP (rationale detailed in Data Analysis), participants had to have their first regional body composition reading prior to their FMP date. In 69 women, the first regional measurement occurred after the FMP date, yielding 391 participants. Finally, we excluded 11 women whose regional body composition values at the normalization visit were outliers (greater than 3 SD above or below the mean), resulting in a final analysis sample of 380 women. In this analysis sample, we included observations that occurred starting at 5 years before and ending at 7 years after the FMP (eliminating the lower and upper 5% bounds of the distribution of time prior to and after the FMP in the sample). The total number of observations in the analysis sample was 3341. Sites obtained IRB approval and participants provided written informed consent.

Figure 1.

Derivation of the analysis sample: Regional Body Composition in Relation to the Final Menstrual Period (FMP) in the Study of Women’s Health Across the Nation (SWAN).

Body composition and anthropometric outcomes

This analysis addresses 5 outcomes: 3 measures of DXA-acquired regional fat distribution (android, gynoid, and visceral fat) and 2 anthropometric measures (waist and hip girths). Android and gynoid fat estimates are derived from DXA whole body composition scans (Hologic, Inc., Waltham, Massachusetts). We used DXA-measured total body fat (not a primary outcome) to compute the percentages of total fat represented by the regional fat estimates and to test whether regional fat mass changes were distinct from each other and from overall fat gain. A detailed description SWAN’s whole body composition protocols is published (6). Germane to this study, SWAN conducted longitudinal, within-site calibration of DXA body composition by scanning at least 30 volunteers on both original and replacement DXA machines when clinical centers upgraded their hardware; values at all sites were standardized to NHANES Apex (6). SWAN did not have access to a body composition phantom, thus did not conduct cross-site body composition calibration. Therefore, unadjusted, absolute values (in grams) of body fat may vary between the UCLA and MGH sites; adjustment for site and race/ethnicity addresses this issue. Trajectories of regional fat are computed as within-woman percent change over time and are, therefore, comparable across sites. Precision estimates for total fat mass values acquired with a 4500A or a Discovery model range between 1.5% and 2.1%. For percent fat, precision is between 1.0% and 1.3% (6). To compute regional fat, the software automatically outlines the android and gynoid regions, after the technologist places the whole body boundary lines. The android region of interest (ROI) is a quadrilateral. Arm cut lines define the android lateral margins; the horizontal pelvic cut line demarcates the inferior margin and the superior boundary is a second horizontal line placed at 20% of the distance from the pelvic line to the neck cut line. The gynoid ROI is a quadrilateral with lateral borders delineated by leg cut lines. The gynoid upper boundary is a horizontal line located below the horizontal pelvic line, by a distance of 1.5 times the height of the android ROI. The gynoid lower border is below the upper gynoid bound, at a distance of 2.0 times the height of the android ROI. Android and gynoid percent fat are computed automatically. To compute visceral fat, software identifies the subcutaneous fat ring, inner abdominal muscle wall, and visceral cavity at the level of the 4^th lumbar vertebral body. Abdominal subcutaneous fat is estimated from the subcutaneous fat on each side of the abdominal cavity and subtracted from the total fat in the abdominal region to yield visceral fat. (See the DAPA measurement tool kit for an illustration of regional body composition boundary lines (19). DXA visceral fat area measurements obtained with this method were highly correlated (r = 0.93) and linearly related to visceral fat area measured by computed tomography (20). For waist and hip measures, participants wore hospital gowns and removed shoes; if preferred, measures were taken over light clothing (attire was consistent over time). Waist and hip circumferences were measured to the nearest 0.1 cm using inelastic tape over non-restrictive undergarments. Waist was the narrowest part of the torso. Hip circumference was taken at the maximum excursion of the buttocks.

Primary predictor

The primary exposure was the number of months before or after the FMP at the time of the DXA or anthropometric measurement (FMP-time); we calculated FMP time using month and year of the FMP and month and year of each DXA or anthropometry. FMP date is the last menstrual bleeding date immediately prior to the first visit when the participant was postmenopausal. The date of a non-surgical FMP is specified in retrospect, after the occurrence of 12 months of amenorrhea.

Other predictors

Age at FMP (years), self-defined race/ethnicity (Black, Japanese, White), self-reported menstrual bleeding patterns, and systemic HT use (yes/no, time varying) were obtained from standardized questionnaires and interviews.

Data analysis

We computed the mean values of descriptive characteristics of the study sample and tested for differences using unpaired t-tests. To analyze change in the 3 measures of regional fat distribution and the 2 anthropometric outcomes relative to FMP-time, we used a previously described method (6, 17). Briefly, using LOESS plots, we identified the functional form of each outcome’s trajectory in relation to time from FMP. Next, we created piece-wise-linear regressions to determine the best knot placement for the parametric outcome trajectories. Finally, we fit piece-wise-linear regression with fixed knots to estimate each outcome’s rate of decline or increase during each phase of the trajectory and the influence of covariates on that trajectory.

We fit mixed effects linear regression to fit piece-wise-linear growth curves to repeated measurements of normalized values of each of the 5 outcomes (in separate models) as functions of time before or after FMP, using linear splines with 2 fixed knots, including random effects for the intercept and 3 slopes. We normalized the outcomes to their value at the visit closest to 3 years prior to the FMP, because values of some outcomes were already on an upward trajectory prior to the MT and because, after normalization, the units of slope are annual percent change. Throughout the manuscript, the terms “starting level” or “starting value” refer to the measurement closest to 3 years prior to FMP.

In an analysis that uses FMP-time, rather than menstrually based clinical stages, to examine the relation of a given outcome to the MT, the MT is operationalized as the time during which an accelerated change in the outcome occurs. This accelerated change interval is defined by the times at which the knots occur in the piecewise regression (6, 17). Our prior work in other biological systems shows that optimal knot locations vary by outcome: The rise in bone turnover markers begins 2 years prior to the FMP, while bone mineral density (BMD) decline starts later, 1 year prior to FMP, consistent with remodeling imbalance preceding BMD loss (17, 21). In the current analysis, optimum knots (and therefore operationally defined MT intervals) also varied by outcome. They were FMP minus 1 year and FMP plus 1.5 years (android and gynoid); FMP minus 3 years and FMP plus 1.5 years (visceral fat); FMP minus 1 year and FMP plus 5 years (hip circumference). For waist circumference, we could not identify knot locations that minimized residual variance; we therefore modeled it using the same as those for android fat (its DXA counterpart). If waist circumference knots were not needed, this would manifest in the final model as no difference between segment slopes.

We added starting level of each outcome, age at FMP and race/ethnicity to the models, as fixed effects on the intercept and 3 slopes, to assess how each influenced the rate of change in the outcome during each segment. The model also adjusted for SWAN site (a fixed effect on intercept and slopes) and time-varying HT use (a fixed effect on intercept). We were missing information about HT use at 21 observations (follow-up visits), which represents 0.6% of observations. These 21 observations were not included in the final multivariable models.

To test whether regional fat mass changes during the MT were distinct from each other and from overall fat gain, we computed percent change per year in total fat and each of the regional fat estimates during the MT interval. For each, we identified a value (A) at the visit that occurred close to 1 year prior to the FMP and a subsequent value (B) that occurred closest to 1.5 year after the FMP (6-month latitude allowed for each). We computed percent change per year in each of the 4 fat variables as 100*(Value B-Value A) / [Value A*(time B-time A)] and computed mean slope (95% CI) for each. We compared mean regional fat mass slopes using a t-test.

A priori, we hypothesized that the MT would have unfavorable effects on body composition. In post-hoc SS calculations, with a sample size of 59 Black and 163 White women, we had 80% power to detect (with 95% confidence, or 2-sided alpha of 0.05) Black-White differences of 0.425 standard deviation (SD) in each outcome. The SD was 14.8% for visceral fat, 9.2% for android fat, and 5.4% for gynoid fat. Therefore, the analysis sample had power to find a statistically significant Black-White difference of 6.3% in visceral fat, 3.9% in android fat, and 2.3% in gynoid fat.

We ran analyses in SAS version 9.2 and used 2-sided alpha of 0.05 for statistical significance.

Results

Sample characteristics

The analysis sample consisted of 380 participants, 59 Black, 158 Japanese and 163 White women. Mean age at first on-study observation was 45.7 years (standard deviation [SD], 2.5 years) and mean age at which FMP occurred was 52.8 years (SD, 2.5 years). Table 1 summarizes premenopausal values (at the visit closest to 3 years before the FMP, the visit used for normalization) of regional fat and anthrompometric characteristics of the analysis sample, overall and by race/ethnicity. Mean total grams of body fat, visceral fat, and android fat regional fat were greatest in Black, intermediate in White, and least in Japanese women. However, the percentage of total fat contained in the visceral region was greatest in Japanese and similar in Black and White women. Median follow-up time was 11.8 years (inter-quartile range [IQR] 5.5; maximum 16.5) and median number of outcome measurements per woman was 10 (IQR 4.0, maximum 13).

Table 1.

Descriptive Characteristics of the Analytic Sample

Characteristic a,b,c	Overall (N = 380)	Black (N = 59, 16%)	Japanese (N = 158, 41%)	White (N = 163, 43%)
Age (y)
At study baseline d	45.68(2.54)	45.39(1.99)	46.21(2.48)	45.27(2.68)
At final menstrual period	52.83(2.53)	53.36(2.45)	52.53(2.29)	52.93(2.76)
Total Body Fat (kg)	24.82(9.83)	31.91(10.66)	19.19(6.02)	27.71(9.65)
Regional Body Composition
Visceral fat (kg)	0.43(0.25)	0.52(0.24)	0.36(0.21)	0.46(0.26)
Visceral fat (%)	1.67(0.61)	1.61(0.54)	1.78(0.61)	1.57(0.60)
Android fat (kg)	1.87(0.96)	2.43(1.12)	1.46(0.67)	2.07(0.97)
Android fat (%)	7.27(1.48)	7.34(1.66)	7.32(1.43)	7.20(1.47)
Gynoid Fat (kg)	4.62(1.51)	5.47(1.56)	3.68(0.)	5.23(1.48)
Gynoid Fat (%)	19.29(2.91)	17.64(2.75)	19.67(2.69)	19.50(2.99)
Anthropometric Measures
Weight (kg)	67.90(16.12)	82.26(17.56)	57.51(9.35)	72.69(14.47)
Height (cm)	160.43(13.65)	162.93(6.42)	154.79(18.39)	164.97(6.31)
Waist circumference (cm)	80.96(13.07)	91.98(14.21)	74.40(8.98)	83.21(12.53)
Hip circumference (cm)	101.14(11.21)	108.48(11.51)	94.11(6.82)	105.20(10.80)

^aValues in table are means (standard deviations)

^bData for anthropometric and body composition characteristics are from the SWAN visit that occurred closest to 3 years prior to the final menstrual period (see Methods for rationale). For regional fat, values represent percentage of total body fat.

^cBold typeface indicates that means or percent values in Black or Japanese participants are statistically significantly different from those values in White participants (P ≤ 0.05)

^dAge at the first body composition observation (analysis sample baseline)

Change in visceral, android, and gynoid fat mass in relation to FMP-time

Table 2 details changes in the 3 regional fat mass outcomes by race/ethnicity and according to the starting value of each outcome. Associations are shown for each of 3 time segments, ie, premenopause, MT, and postmenopause. Associations are also shown for the cumulative change in each regional fat mass outcome over the 12-year period. The referent woman is White, does not use HT, is aged 52.8 years at FMP (analysis sample average), and has a starting value of each outcome equal to that of the analysis sample mean. To compute the model-predicted fat mass slopes (and 12-year change) in non-White women and in women with starting values 1 SD greater than the sample average, we add the effect size estimates for race/ethnicity or starting value to the slopes in the referent woman. The race/ethnicity specific CI shown are for the effect size differences from the White referent. When statistically significant racial/ethnic differences are observed, we provide (below) the race/ethnicity specific rate of change (and 95% CI) for women in that racial/ethnic group who did not use HT, were 52.8 years old at the FMP (analysis sample average), and had starting value of each outcome equal to that of the analysis sample mean. These are Japanese referent and Black referent women, respectively. The last column of Table 2 lists the model-predicted 12-year cumulative change in each fat mass outcome for Japanese referent and Black referent women.

Table 2.

Annualized Percentage Rates of Change in Visceral Fat, Android Fat, and Gynoid Fat in Relation to the Final Menstrual Period (FMP) and the Associations of Change Rates With Race/Ethnicity and Starting Fat Level, the Study of Women’s Health Across the Nation^a,b

	Annualized rates of change during each interval prior to and after FMP in referent, and associations with race/ethnicity and starting levelsc,d			12-year change in referent and association with race/ethnicity and starting levelf	Race/ethnicity-specific cumulative 12-year change
	Premenopause	Menopause Transition e	Postmenopause
Visceral Fat
Referentg	-1.85% (-3.97%, 0.28%)	6.24% (4.31%, 8.17%)	1.47% (0.30%, 2.65%)	32.5% (23.5%, 41.5%)	32.5% (23.5%, 41.5%)
Japaneseh	2.36% (-0.78%, 5.49%)	-7.04% (-10.22%, -3.86%)	1.08% (-0.69%, 2.86%)	-21.0% (-35.3%, -6.7%)	11.5% (1.8%, 21.2%)
Blackh	2.25% (-2.69%, 7.19%)	-1.62% (-5.37%, 2.14%)	-3.62% (-6.22%, -1.02%)	-22.7% (-41.8%, -3.6%)	9.8% (-9.0%, 28.6%)
Starting leveli	3.11% (1.79%, 4.43%)	-2.42% (-3.71%, -1.12%)	-0.95% (-1.79%, -0.10%)	-9.9% (-16.0%, -3.7%)	N/A
Android Fat
Referentg	1.21% (0.21%, 2.21%)	5.54% (3.65%, 7.44%)	0.90% (0.04%, 1.77%)	23.7% (17.5%, 29.9%)	23.7% (17.5%, 29.9%)
Japaneseh	-0.60% (-2.24%, 1.03%)	-6.34% (-9.46%, -3.22%)	0.32% (-1.00%, 1.65%)	-16.5% (-26.5%, -6.5%)	7.2% (0.5%, 14.0%)
Blackh	1.17% (-0.94%, 3.28%)	-2.43% (-6.18%, 1.33%)	-1.04% (-2.89%, 0.81%)	-7.1% (-19.9%, 5.7%)	16.6% (3.9%, 29.3%)
Starting level i	0.41% (-0.29%, 1.12%)	-2.90% (-4.27%, -1.53%)	-0.86% (-1.52%, -0.21%)	-10.3% (-14.8%, -5.9%)	N/A
Gynoid Fat
Referent	0.09% (-0.47%, 0.64%)	2.03% (0.96%, 3.09%)	-0.87% (-1.29%, -0.45%)	0.7% (-2.3%, 3.7%)	0.7% (-2.3%, 3.6%)
Japanese h	-0.13% (-1.05%, 0.79%)	-4.38% (-6.18%, -2.58%)	0.59% (-0.06%, 1.24%)	-8.2% (-13.1%, -3.3%)	-7.6% (-10.9%, -4.3%)
Black h	0.50% (-0.64%, 1.65%)	-0.56% (-2.64%, 1.52%)	-0.52% (-1.44%, 0.39%)	-2.3% (-8.4%, 3.9%)	-1.6% (-7.7%, 4.5%)
Starting level i	0.49% (0.11%, 0.87%)	-1.73% (-2.50%, -0.95%)	0.09% (-0.23%, 0.41%)	-1.9% (-4.0%, 0.3%)	N/A

^aSample size for each of the visceral fat, android fat, and gynoid fat models was 380 and the total number of observations (scans over time) for each outcome as 3441. There were 21 observations missing hormone therapy information, resulting in 3220 usable observations in each multivariable model.

^bSlopes (annualized change rates) and cumulative change estimates in the referent and in each racial/ethnic group that are shown in bold font are statistically significantly different from zero.

^cChange in fat values expressed as percentage of starting levels; 95% confidence intervals are shown in parentheses. Starting level obtained from a visit that took place ~3 years before the FMP (see Methods for rationale).

^dIn addition to race/ethnicity and starting level, the model also adjusts for age at FMP, hormone therapy use (time varying), and SWAN study site. Associations in bold font are statistically significantly different from zero. Neither hormone therapy nor age at FMP had significant associations with slopes. Unmodeled residual error standard deviation was 23.0% of starting level for visceral fat, 13.8% for android fat, and 8.4% for gynoid fat.

^eThe second segment of the piece-wise regresstion (the menopause transition [MT]) varied across the 3 fat measurements, The optimal knot locations for changes in slopes for each of these outcomes differed (see Methods). The MT was defined as the period spanning 3 years before to 1.5 years after the FMP for visceral fat, 1 year before to 1.5 years after the FMP for android fat, and 1 year before to 5 years after the FMP for gyneoid fat.

^f12 year period spanning from 5 years before to 7 years after the FMP

^gModel-predicted slopes (percentage of starting level gained per year) for the referent woman. The referent woman was defined as White, not taking hormone therapy, having age at FMP at the sample mean of 52.8 years, and starting levels of fat at sample mean (see Table 1 for mean starting values).

^hTo compute the model-predicted fat mass slopes (and 12-year change) in non-White women and in women with starting values 1 SD greater than the sample average, add the effect size estimates for race/ethnicity or starting value to the slopes in the referent woman.

ⁱPer standard deviation increment

In the White referent woman, visceral fat mass did not change during premenopause, increased 6.24% annually in the MT, and increased more slowly, at 1.47% per year in postmenopause; the latter 2 rates were statistically significantly different from zero [Table 2], and from each other. In the referent, android fat mass began to climb during premenopause, at a rate of 1.21% annually, followed by a nearly 5-fold acceleration during the MT (to an annual gain of 5.54%) and a 6-fold deceleration (to an annual increase of 0.90% yearly) in postmenopause. Rates in each menopause stage were statistically significantly different from zero; both the acceleration and the deceleration were also statistically significant. At the onset of the MT in the referent woman, gynoid fat started increasing, with a slope of 2.03% per year, but in postmenopause gynoid fat began declining, at an annual change rate of -0.87% (each rate statistically significantly different from zero and from each other).

In each transition stage, all 3 regional fat mass trajectories of Black participants did not differ statistically from those of White participants with one exception: visceral fat increase in postmenopause was significantly smaller in Black women [Table 2]. In fact, unlike in White women, visceral fat in Black women did not increase significantly in postmenopause; estimated annualized change in the Black referent woman was -2.15% (95% CI for Black-specific rate, -4.68% to +0.38%). Thus, although Black women gained visceral fat mass at a rate similar to White women during the MT, the Black-specific mean 12-year fat mass change was 9.8% (-9.0% to 28.6%), significantly different from that of White women, and not statistically different from zero change.

In Japanese women, premenopausal and postmenopausal trajectories of visceral fat, android fat, and gynoid fat mass did not differ statistically from those of White women. However, during the MT, all three regional fat outcomes increased significantly less in Japanese than they did in White women [Table 2]. Unlike the gains observed in the White referent woman during the MT, neither visceral fat mass nor android fat mass increased in the Japanese referent women during the MT and gynoid mass decreased significantly. Specifically, visceral fat mass changed by -0.80% (95% CI -2.92% to 1.33%, Japanese-specific) per year, android fat annual change was -0.80% (95% CI -2.89% to 1.30%, Japanese-specific) annually, and gynoid fat yearly rate of change was -2.35% (95% CI -3.55% to -1.16%, Japanese-specific). The Japanese 12-year cumulative gains in visceral and android fat were approximately one-third those of White women (11.5% vs. 32.5% and 7.2% vs 23.7%, respectively) and were statistically significantly different from the cumulated gains in White participants. The average total 12-year change in gynoid fat was significantly negative in Japanese woman and no different from zero in the other two racial/ethnic groups.

In general, greater starting level of each of the regional body fat parameters was associated with faster premenopausal increases, but slower growth during and after the MT; however, this finding was not universal [Table 2]. The combined effect was of a smaller cumulative 12-year increase in visceral and android fat in women who started at greater levels, and no significant association of starting level with 12-year cumulative change in gynoid fat [Table 2]. During the menopause transition, per SD increment in starting level, annual visceral fat gain was 2.42% per year smaller, android fat gain averaged 2.90% per year smaller, and gynoid fat gain 1.73% smaller. In postmenopause, higher starting level also had a tempering effect on visceral and android fat trajectories, which were 0.95% and 0.86% lower per SD increment in starting values, respectively.

Changes in waist and hip circumferences in relation to the FMP

In the White referent woman, average waist circumference increased by 0.55% per year in premenopause, 0.96% per year in the MT, and 0.55% annually in postmenopause (each statistically significantly different from zero) [Table 3]. The MT slope was not statistically different from the premenopausal slope (P = 0.5) or the postmenopausal slope (P = 0.3); ie, the trajectory was linear across all three phases (data not tabulated). In the same referent, hip circumference grew linearly during the first two MT stages, at 0.20% and 0.35% per year during premenopause and the MT, respectively, but the rate of change was statistically no different from zero in postmenopause. The premenopausal and MT slopes did not differ statistically (P = 0.8), but the postmenopausal slope differed significantly from the slope during the MT (P = 0.01) (data not tabulated).

Table 3.

Annualized Percentage Rates of Change in Waist and Hip Circumference in Relation to the Final Menstrual Period (FMP) and the Associations of Change Rates With Race/Ethnicity and Starting Waist or Hip Level, the Study of Women’s Health Across the Nation^a,b

	Annualized rates of change during each interval prior to and after FMP in the referent and associations with race/ethnicity and starting levelsc,d			12-year change in referent and association with race/ethnicity and starting levelf	Race/ethnicity-specific cumulative 12 -year change
	Premenopause	Menopause Transition e	Postmenopause
Waist Circumference
Referent g	0.55% (0.33%, 0.77%)	0.96% (0.53%, 1.40%)	0.55% (0.35%, 0.75%)	7.6% (6.3%, 9.0%)	7.6% (6.3%, 9.0%)
Japanese h	-0.20% (-0.56%, 0.16%)	-0.90% (-1.62%, -0.18%)	-0.16% (-0.46%, 0.14%)	-3.9% (-6.1%, -1.8%)	3.7% (2.2%, 5.1%)
Black h	0.06% (-0.42%, 0.54%)	-0.24% (-1.12%, 0.63%)	-0.27% (-0.69%, 0.16%)	-1.8% (-4.6%, 1.0%)	5.8% (3.0%, 8.6%)
Starting level i	-0.04% (-0.20%, 0.12%)	-0.43% (-0.75%, -0.10%)	-0.07% (-0.22%, 0.08%)	-1.6% (-2.6%, -0.6%)	N/A
Hip Circumference
Referent	0.20% (0.02%, 0.38%)	0.35% (0.18%, 0.51%)	-0.02% (-0.25%, 0.20%)	2.8% (1.8%, 3.9%)	2.8% (1.8%, 3.9%)
Japanese h	-0.12% (-0.43%, 0.18%)	-0.35% (-0.63%, -0.08%)	0.29% (-0.03%, 0.62%)	-2.0% (-3.8%, -0.3%)	0.8% (-0.4%, 2.0%)
Black h	0.21% (-0.15%, 0.58%)	-0.08% (-0.41%, 0.24%)	-0.17% (-0.71%, 0.38%)	0.0% (-2.1%, 2.1%)	2.8% (0.7%, 5.0%)
Starting level i	0.06% (-0.07%, 0.20%)	-0.10% (-0.23%, 0.03%)	0.18% (-0. 02%, 0.37%)	-0.0% (-0.8%, 0.8%)	N/A

^aSample size for each of the waist circumference and hip circumference models was 376; 4 participants in the analysis sample were missing these outcome variables, thus the total number of waist and hip circumference observations over time was 3233. There were 21 observations missing hormone therapy information, resulting in 3212 usable observations in each multivariable model.

^bSlopes (annualized change rates) and cumulative change estimates in the referent and in each racial/ethnic group that are shown in bold font are statistically significantly different from zero.

^cChange in circumferences expressed as percentage of starting levels; 95% confidence intervals are shown in parentheses. Starting level obtained from a visit that took place ~3 years before the final menstrual period (FMP) (see Methods for rationale).

^dIn addition to race/ethnicity and starting level, the model also adjusts for age at FMP, hormone therapy use (time varying), and SWAN study site. Associations in bold font are statistically significantly different from zero. Neither hormone therapy nor age at FMP had significant associations with slopes. Unmodeled residual error standard deviation was 3.2% of starting level for waist circumference, and 2.3% for hip circumference.

^eThe second segment of the piece-wise regresstion (the menopause transition [MT]) varied between the 2 anthropometric measures. The optimal knot locations for changes in slopes were different (see Methods). The MT was defined the period spanning 1 years before to 1.5 years after the FMP for waist circumference, 1 year before to 5 years after the FMP for hip circumference.

^f12 year period spanning from 5 years before to 7 years after the FMP

^gModel-predicted slopes (percentage of starting level gained per year) for the referent woman. The referent woman was defined as White, not taking hormone therapy, having age at FMP at the sample mean of 52.8 years, and starting levels of fat at sample mean (see Table 1 for mean starting values).

^hTo compute the model-predicted fat mass slopes (and 12-year change) in non-White women and in women with starting values 1 SD greater than the sample average, add the effect size estimates for race/ethnicity or starting value to the slopes in the referent woman.

ⁱPer standard deviation increment

Rates of change in waist and hip circumference among Black women were like those of the referent sample of White participants [Table 3]. Rates of change in Japanese women did differ significantly from those in White women, at both the waist and the hip, during the MT only. Unlike the patterns observed in the White referent, Japanese women’s anthropometric measures did not increase during the MT. Among Japanese women, mean annualized rates of change in the MT were -0.06% (Japanese –specific 95% CI, -0.43% to 0.55%) and -0.01% (Japanese-specific 95% CI, -0.19% to 0.18%) at the waist and hip, respectively. In the premenopausal and postmenopausal intervals, waist and hip circumference trajectories were similar in Japanese and White women. Thus, the cumulative increase in waist circumference was about 50% less in Japanese than in White women, 3.7% vs. 7.9%, respectively [Table 3]. In addition, unlike Black and White participants, Japanese women’s hip circumference did not increase significantly during the 12-years of observation.

Of 3341 observations, HT use was reported during only 35 (1.0%) and all use occurred after the FMP. (It is unfeasible to define the FMP while taking HT, thus premenopausal HT users were excluded). HT use did not predict change in any study outcome (data not shown).

The mean annualized percent change in total fat during the MT (defined as 1.5 prior to the FMP to 1 year after FMP) in the entire sample was 1.40% (95% CI 0.56, 5.25). During this same interval, yearly percent changes in regional fat values in the whole sample were: 4.15% (95% CI 2.07, 6.24) for visceral fat, 2.94% (95% CI 1.65, 4.23) for android fat, and 0.60% (95% CI -0.16, 1.35) for gynoid fat. Mean difference in percent change per year in android vs. visceral fat in the entire sample was -1.12 (P = 0.08), between android and gynoid fat was 2.34 (P < 0.0001), and between visceral and gynoid fat was 3.55 (P < 0.0001) (data not tabulated).

Fig. 2 recapitulates the model-predicted trajectories of visceral fat, android fat, and waist circumference during each phase (premenopause, MT, and postmenopause) in White (panel A), Black (panel B) and Japanese (panel C) women. The figure highlights that visceral fat begins to change earliest, ~3 years prior to the FMP, increasing similarly in White and Black women; in contrast, visceral fat does not increase during the MT in Japanese women. In the postmenopausal segment, when the White trajectory is still climbing (more slowly than during the MT) the Black trajectory is trending downward, and the Japanese trajectory turns upward (at similar to the White rate). MT-related changes in android fat are apparent later than those of visceral fat, beginning ~1 year before the FMP, but patterns are similar to those of visceral fat. During the MT, android fat trajectories in White and Black women are alike and increasing, whereas the android fat trajectory does not climb in Japanese women. In the postmenopause, when the White trajectory is upward, at a slower rate than during the MT, the Black trajectory is now flat, and the Japanese trajectory has turned upward (like that of White women). Fig. 2 highlights the contrast between the non-linearity in DXA regional fat estimates across the menopause and the linearity of waist circumference during that same period. There is a small, similar, linear rise in waist circumference across all stages in all racial/ethnic groups.

Figure 2.

Average trajectories of change in percent visceral fat, percent android fat and waist circumference in relation to the date of the final menstrual period, by race/ethnicity, Study of Women’s Health Across the Nation.

Fig. 3 depicts the model-predicted trajectories of gynoid fat and hip circumference during each reproductive phase in White (panel A), Black (panel B) and Japanese (panel C) women. Gynoid fat trajectories during the MT are similar to those of android fat: it begins to rise about a year prior to the FMP and is similar in Black and White women but declines (substantively and significantly) in Japanese women. Postmenopausally, gynoid fat declines slowly, significantly, and similarly in all racial/ethnic groups. Fig. 3 points out that gynoid fat change is non-linear across the 3 menopause segments, while hip circumference rises slightly and linearly, in all ethnic groups during premenopause and the MT, and flattens to a zero slope in postmenopause.

Figure 3.

Average trajectories of change in percent gynoid fat and hip circumference in relation to the date of the final menstrual period, by race/ethnicity, Study of Women’s Health Across the Nation.

Discussion

This study describes the longitudinal trajectories of regional fat distribution (visceral, android and gynoid) and anthropometrics (waist and hip circumference) during a 12-year observation period, as women progress from premenopause into postmenopause. Stages of the MT were operationalized using time relative to the FMP. In all 3 racial/ethnic groups represented, premenopause was not associated with change in visceral or gynoid fat, but there was a small (~1%) annual growth in android fat. Onset of the MT was accompanied by a striking change in the trajectories of regional fat measures: Black and White women manifested a nearly 5-fold increase in annual android fat gain and both visceral and gynoid fat started increasing. In contrast, Japanese women’s trajectories of visceral and android fat did not climb during the MT; moreover, Japanese women started losing gynoid fat in the MT. A substantial deceleration in the trajectories of each regional fat measure characterized the start of postmenopause: visceral fat gain slowed to ~1.5% yearly in White and Japanese women while Black women stopped gaining visceral fat. Unlike regional fat mass, waist girth increased slowly prior to, during, and following the MT; no MT-related rate acceleration or deceleration was evident. Hip girth also climbed slowly during premenopause and the MT, but it decelerated to zero rate in postmenopause. For the majority of regional fat estimates, a greater starting value lessened the rate of gain. Finally, the amounts by which visceral, android and gynoid fat mass changed during the MT were distinct from each other, supporting that the MT is a time of shifts in regional body fat deposition, rather than a homogenous, overall gain in fat mass.

Our study supports the thesis that the transition from pre- to postmenopause is associated with unfavorable changes in regional fat distribution, with gains in visceral and android fat and loss of gynoid fat (with racial/ethnic variations, discussed later). Earlier cross-sectional analyses reported an association between postmenopause (compared to premenopause) and greater central fat and one observed less leg fat in postmenopause; however, these studies rely on between-woman comparisons (10, 11, 14, 15). Two subsequent longitudinal cohorts observed an increase in visceral or intra-abdominal fat among women who transitioned from pre- to postmenopause (12, 13). Relatively small sample sizes (≤ 153), in conjunction with modest follow-up intervals (≤7 years), allowed the investigators to compare only dichotomized menopause classifications (pre- vs. postmenopause), rather than to estimate the influence of each transition stage (premenopause, MT, and postmenopause) (12, 13). A third cohort of 102 women followed for 5 years reported that increasing age, but not MT stage, predicted an increase in visceral fat; small numbers of transitions from pre- to perimenopause (N = 44) and peri- to postmenopause (N = 51) likely limited the capacity to find an age by stage interaction (16). The present study’s larger sample size, longer duration, and use of FMP-based exposure time allows it to expand upon prior work, showing that the critical period for gains in central adiposity is the MT, which begins a few years before women have become clinically postmenopausal. In the current study’s referent group, 3 years before the FMP, the slope of visceral fat accelerated greatly, to a rate of 6.24% per year and slowed to 1.47% yearly in postmenopause (starting 1.5 years after the FMP). A similar trajectory characterized android fat: slope accelerated to 5.54% annually beginning 1 year prior to the FMP and slowed to a rate of 0.90% per year 1.5 years after it. In addition, beginning 5 years after the FMP, gynoid fat began falling at a rate of 0.87% yearly.

Patterns of change over time in anthropometric measures were distinct from those observed for DXA-estimated regional fat. Unlike the abrupt, large, jumps in the growth rates of visceral and android fat at the start of the MT, the trajectory of waist circumference was gradual and without abrupt accelerations or decelerations. It rose by ~0.5% to ~1% annually in all 3 stages. Trajectories of hip circumference and gynoid fat were also discordant. A small, linear, 0.2 to 0.4% increase in hip circumference occurred during the premenopause and the MT, which diminished to a zero slope in postmenopause; this contrasts with the accelerated gain in gynoid fat at the onset of the MT and loss of gynoid fat after the MT.

A central question is whether there are changes in regional body composition related to the MT—that is, “Are there differential gains or losses in fat within body regions?” This study’s answer to that question is affirmative, based on the variation in regional fat trajectories. For example, during the MT, rate of change of visceral and android fat increased 6-fold, but the annual rate of gain in gynoid fat only doubled. However, the primary trajectory models did not directly compare region-specific rates of change. An additional analysis explicitly tested for differences between overall fat gain and fat gain within regions during the MT phase. It demonstrated that android and visceral fat gain rates were 5 to 6 times greater that of gynoid fat during the interval spanning 1.5 years before through 1 year after the FMP, supporting the hypothesis that the development of central adiposity, particularly, is associated with the MT.

This study found both similarities and differences in the racial/ethnic trajectories of regional fat change. In the MT phase, rates of gain in visceral and android fat were alike in Black and White women. This concurs with results from a 2-year, longitudinal report based on Black and White Chicago SWAN participants, which found that CT-assessed abdominal fat area increased at a similar rate in Black and White women, whose average starting age was 50 years (22). Although the Chicago-SWAN analysis used calendar time to compute rates of change, given the mean starting age, many women were experiencing the MT. Our analysis revealed a striking Black-White difference in postmenopause--White women continued to gain visceral fat but Black women did not. The mean, postmenopausal, Black-specific rate of change in visceral fat was -2.15%, which approached (but did not reach) statistical significance (95% CI: -4.68% to +0.38%), suggesting that their visceral fat may be actually declining. This postmenopausal drop in visceral fat among Black women may be one mechanism underlying cross-sectional reports that postmenopausal Black women have less visceral fat than do White women, despite the presence of greater total body fat in Black women (23, 24).

In Japanese women, patterns of change in regional fat depots during the MT were unlike those of Black and White women: Japanese participants’ visceral and android fat did not climb and gynoid fat declined. In postmenopause, Japanese fat mass trajectories became similar to those of the other groups: central fat components increased, while gynoid fat declined. Epidemiological studies report that the prevalence of central obesity (by body composition or waist-hip ratio [WHR]) is greater among women of Asian heritage than women of European origin, despite Japanese women being at low risk of overweight or obesity (25, 26). Notably, Japanese SWAN participants had higher premenopausal values of visceral fat (as a percentage of total fat) than did Black or White SWAN women, offsetting the Japanese advantage during the MT (ie, not gaining visceral fat). Additionally, the unique decline in gynoid fat during the MT among Japanese women may underlie the unfavorable WHR observed in epidemiological surveys.

Rate of gain in fat over time was computed as a percentage of starting fat value and greater starting value was associated with a lesser percent change. However, this does not mean that those who began at a higher fat level gained fewer absolute kilograms of fat. For example, if a woman began at the mean referent level of visceral fat, 0.43 kg, she gained 32.5% (0.14 kg) in the 12 years surrounding the FMP; a second woman, who started 1 SD higher (at 0.68 kg), gained 22.6% in the same period (0.15 kg). The second woman’s percentage change is less than that of the first woman’s, but both women had similar absolute gains, leaving their relative rankings unaltered.

The extreme decline in estradiol (E2) and rise in follicle stimulating hormone (FSH) that accompany the MT are plausible mechanisms of gains in central adiposity. The E2 trajectory is marked by a sharp decline starting ~2 years prior to the FMP and stabilizing ~2 years after it and the FSH trajectory is a mirror image of the E2 curve (8, 9). The trajectories of android and visceral fat mass (abrupt growth from a few years before through a few years after FMP) correspond to the FSH rise and are the reciprocals of the E2 curve. E2 may affect total and regional fat mass through many energy homeostasis pathways, including central nervous system control of food intake and energy expenditure, regulation of adipose tissue lipid storage and metabolism, and insulin sensitivity (27). Decline in E2 may preferentially lead to central fat accumulation and promote visceral adiposity (28, 29). Human ovarian suppression using a gonadotropin-releasing hormone agonists results in visceral, but not overall fat mass gain; when E2 is added back, the visceral fat gain is prevented (29). Murine studies with an FSH-blocking antibody demonstrate that, in ovarian-intact animals with unaltered serum E2 levels, FSH antibody reduces body fat, induces beiging of adipocytes (making them more metabolically active), and leads to greater thermogenesis (24).

Limitations of this study include that garments worn and time of day may affect accuracy and precision of anthropometric measures. SWAN used uniform morning measurement times and asked women to wear similar garments each year, minimizing these potential influences. The sample size of Black women was a third of that of Japanese and White women, raising the question of whether we had a limited ability to detect Black-White differences. Detailed in the Methods, the study had power to find a statistically significant Black-White difference of 6.3% in visceral fat, 3.9% in android fat, and 2.3% in gynoid fat, each of which is smaller than the observed Japanese-White differences of 7.0%, 6.3%, and 4.4%, respectively. This study did not include Chinese and Hispanic SWAN participants, as the clinical sites that recruited these racial/ethnic groups did not measure regional body composition. We did not observe an effect of HT on study outcomes, but it is unfeasible to test for an HT effect during the MT, which requires a known FMP date. During premenopause and the MT, HT use obscures the date of natural menopause, thus women who use HT prior to the FMP are not included. After menopause, HT use was very rare; although postmenopausal HT users are in the analysis, only 1% of observations (35 of 3341) were in women using HT. To be detectable in this small subgroup, the HT effect size during postmenopause would have had to have been 0.47 SD (ie, a 27%, 24%, and 15% change in visceral, android, and gynoid fat, respectively). We did not conduct studies to compute precision estimates of total and regional fat mass measures by DXA; in adults, published values are in the range of 1.8% to 2.1% for total fat mass (30, 31). Finally, because the scope of the analysis required to describe and quantify the natural history of regional body composition change across the MT is large, we have not yet investigated the role of many potential risk or explanatory factors (eg, physical activity, diet, sex steroids, gonadotropins, metabolic factors, etc.). We will do so in future work. Enumerating the principal study strengths, we analyzed DXA-quantified regional body composition and measured anthropometrics contemporaneously, enabling us to see how their trajectories compared. Using an FMP-based time scale to describe the effect of the transition from pre- to postmenopause on study outcomes is a more discerning assessment of progress through the transition that is an analysis based on clinical MT stages (17). Inclusion of Black, Japanese and White women allowed us to observe how their trajectories of change in regional fat mass and anthropometrics differ, which will encourage studies of the mechanisms and meanings of this racial/ethnic variation.

In summary, in general, the transition from pre- to postmenopause is accompanied by an increase in central and a diminution in peripheral fat. Greater central fat storage is associated with a higher risk of cardiometabolic diseases (2, 4, 5). The magnitude of the observed gains in central fat depots are substantial and could have an important influence on cardiometabolic risk in women similar to those studied. The MT may be an opportune time to prevent, or lessen, these unfavorable shifts in fat mass. The unique trajectories of change in regional body fat among Black, White and Japanese women underscore the necessity of including race/ethnicity as a variable in mechanistic or interventional studies.

Acknowledgments

Financial Support: The Study of Women’s Health Across the Nation (SWAN) has grant support from the National Institutes of Health (NIH), DHHS, through the National Institute on Aging (NIA), the National Institute of Nursing Research (NINR) and the NIH Office of Research on Women’s Health (ORWH) (Grants U01NR004061, U01AG012505, U01AG012535, U01AG012531, U01AG012539, U01AG012546, U01AG012553, U01AG012554, and U01AG012495). This project was also supported by U19AG063720. The content of this article is solely the responsibility of the authors and does not necessarily represent the official views of the NIA, NINR, ORWH or the NIH.

Author contributions: Participant recruitment and enrollment (GAG; MHH; JSF); data management and cleaning (GAG, MHH, WH, ASK); analytic design (GAG, ASK); statistical analysis (GAG, MHH, WH, ASK); primary manuscript drafting (GAG, ASK); critical review and revision of manuscript (all).

Clinical Centers: University of Michigan, Ann Arbor—Siobán Harlow, PI 2011—present, MaryFran Sowers, PI 1994-2011; Massachusetts General Hospital, Boston, MA—Joel Finkelstein, PI 1999—present; Robert Neer, PI 1994 – 1999; Rush University, Rush University Medical Center, Chicago, IL—Howard Kravitz, PI 2009—present; Lynda Powell, PI 1994 – 2009; University of California, Davis/Kaiser—Ellen Gold, PI; University of California, Los Angeles—Gail Greendale, PI; Albert Einstein College of Medicine, Bronx, NY—Carol Derby, PI 2011—present, Rachel Wildman, PI 2010 – 2011; Nanette Santoro, PI 2004 – 2010; University of Medicine and Dentistry—New Jersey Medical School, Newark—Gerson Weiss, PI 1994 – 2004; and the University of Pittsburgh, Pittsburgh, PA—Karen Matthews, PI.

NIH Program Office: National Institute on Aging, Bethesda, MD—Chhanda Dutta 2016- present; Winifred Rossi 2012–2016; Sherry Sherman 1994 – 2012; Marcia Ory 1994 – 2001; National Institute of Nursing Research, Bethesda, MD—Program Officers.

Central Laboratory: University of Michigan, Ann Arbor—Daniel McConnell (Central Ligand Assay Satellite Services).

Coordinating Center: University of Pittsburgh, Pittsburgh, PA—Maria Mori Brooks, PI 2012—present; Kim Sutton-Tyrrell, PI 2001 – 2012; New England Research Institutes, Watertown, MA—Sonja McKinlay, PI 1995 – 2001.

Steering Committee: Susan Johnson, Current Chair, Chris Gallagher, Former Chair

We thank the study staff at each site and all the women who participated in SWAN.

The authors have declared that no conflict of interest exists.

Data Availability

Restrictions apply to the availability of data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.