Presence of vasomotor symptoms is associated with lower bone mineral density

Study Sample

We used data from the Study of Women’s Health Across the Nation (SWAN) study, a prospective study characterizing health status across the menopause transition in a multi-racial/ethnic community-based cohort of 3302 women. Recruitment strategy and eligibility criteria have been described in detail elsewhere ¹⁵. Participants were aged 42-52 years at baseline, had an intact uterus, had at least one intact ovary, were not currently using exogenous reproductive hormones, were not pregnant or lactating, and had experienced at least one menstrual period in the 3 months preceding the baseline visit. In addition to recruiting Caucasian women, each of the 7 study sites recruited women of another ethnic group: African American women (Boston, Massachusetts; Chicago, Illinois; Pittsburgh, Pennsylvania; Detroit, Michigan), Hispanic women (Hudson County, New Jersey), Chinese women (Oakland, California), and Japanese women (Los Angeles, California). The SWAN bone substudy, which measured BMD annually, consisted of Caucasian, African American, Chinese, and Japanese women recruited at 4 of the SWAN sites: Boston, Pittsburgh, Detroit, Oakland, and Los Angeles Study. The study was approved by each study site’s institutional review board, and written informed consent was obtained from each participant.

For this study, we analyzed data spanning baseline through annual visit 5 from participants of the SWAN bone substudy who underwent at least one BMD measurement and at least one contemporaneous assessment of hot flashes and night sweat symptoms during the observation period. To clearly characterize the VMS-BMD association prospectively across the stages of menopause, during follow-up we censored women at the time they reported having hysterectomy, bilateral oophorectomy, or pregnancy/breastfeeding. We also excluded women taking medications that might influence bone density (corticosteroids, alendronate, etidronate, raloxifene, calcitonin, or calcitriol) and women taking exogenous hormones by censoring them starting at the time they first reported taking the exclusionary medications, resulting in an analytic sample of 2213 participants.

Bone Mineral Density Measurements

Lumbar spine, total hip, and femoral neck BMD was measured at baseline and annually using Hologic (Hologic, Inc., Bedford, MA) 2000 (Pittsburgh and Oakland sites) or 4500A (Boston, Detroit, and Los Angeles sites) densitometers. Reproducibility of hip measurements was optimized using Osteodyne (Research Triangle Park, NC) positioning devices ¹⁶. At all study sites, quality control measures consisted of daily anthropomorphic spine phantoms, cross-site and cross-time calibration with a Hologic spine phantom, and an on-site review of all scans according to pre-designated criteria. Five percent of scans, as well as scans with potential problems, were reviewed by Synarc, Inc. (Waltham, MA), resulting in a designation of each scan as acceptable, requiring reanalysis, or rejected ³.

Serum Estradiol and Follicle-Stimulating Hormone (FSH) Measurements

In women who were not postmenopausal, a fasting blood samples was obtained annually in the 2- to 5-day window of the early follicular phase of the menstrual cycle ³. If irregular menstrual cycles precluded an early follicular phase blood draw, blood sampling was performed without respect to the timing of menstrual bleeding. Estradiol was assayed using a semi-automated, competitive ACS:180 (e2-6) Immunoassay ¹⁷. At an estradiol level of 50 pg/mL, the inter-assay coefficient of variation was 10.6% and the intra-assay coefficient of variation was 6.4% ³. Serum FSH concentrations were measured with a 2-site chemiluminometric immunoassay ³. The assay used two monoclonal antibodies, each directed to different regions on the FSH beta-subunit. For the FSH assay, the inter-assay coefficient of variation was 6.0%, and the intra-assay coefficient of variation was 12.0% at an FSH level of 15 IU/L ³. Blood samples were taking in the early follicular phase in 84% of premenopausal, 75% of early perimenopausal, and 6% of late perimenopausal participants.

Questionnaire-based and Anthropometric Measurements

At baseline and at each annual follow-up visit, women completed a questionnaire that included a symptom checklist. The symptom checklist was worded as follows: “Below is a list of common problems which affect us from time to time in our daily lives. Thinking back over the past two weeks, please circle the number corresponding to how often you experienced any of the following”. Response choices were not at all, 1-5 days, 6-8 days, 9-13 days, or every day. For each visit, we classified women who had hot flashes or night sweats as having vasomotor symptoms (VMS). We classified women with VMS for 6 or more days in the past 2 weeks as having frequent VMS.

Baseline and annual follow-up questionnaires served as the source of information regarding age, ethnicity, current medication use, and current menopause status. We defined exogenous hormone therapy as estrogen pills, progestin pills, progesterone pills, estrogen patches, birth control pills, estrogen injections, tamoxifen pills, diethyl-stilbesterol pills, progestin injection, fertility medications, hormone creams or hormone suppositories. Menopausal status was defined as follows based on self-reported menstrual cycle characteristics recalled over the past year: premenopausal (menstruation in the past 3 months with no change in menstrual regularity in past year), early perimenopausal (menstruation in past 3 months with decreased regularity in past year), late perimenopausal (no menses for 3-11 months), and postmenopausal (no menses for past 12 months.

At baseline and at each annual follow-up visit, height and weight were measured using calibrated electronic or balance beam scales and stadiometers.

Statistical Analysis

For baseline and each annual follow-up visit, we assessed the mean lumbar spine, total hip, and femoral neck BMD, prevalence of VMS, and menopausal stage. We estimated the duration of time each women spent in each menopause stage category using an algorithm that took into account self-reported menopause stage at each annual interview (Finkelstein et al, submitted for publication).

We performed repeated measures mixed models with BMD of each anatomical site (lumbar spine, total hip, and femoral neck) as the outcome. For the main exposure, for each visit, we created an index that accounted for whether the participant reported having VMS (any vs. none in the past 2 weeks) as well as her menopausal status (premenopausal, early peri-, late peri-, or post-menopausal). The resulting 8 categories for the main exposure variable were: premenopausal without VMS, premenopausal with VMS, early perimenopausal without VMS, early perimenopausal with VMS, late perimenopausal without VMS, late perimenopausal with VMS, postmenopausal without VMS, and duration of being postmenopausal with VMS. Using the 8-category main exposure variable, we performed two sets of repeated measures mixed models. In the first set of models, we controlled for baseline age, cumulative time spent in each menopausal stage, baseline VMS, baseline menopausal status, race/ethnicity, and study site. In the second set of models, we additionally controlled for baseline weight, baseline height, and change in weight since baseline. Finally, we repeated the models to account for frequency of VMS as the main exposure by using a similar 8-category indicator for frequent VMS (≥6 days vs. ≤5 days in past 2 weeks) within each menopausal stage. While we controlled for the main effect of race/ethnicity, we did not explore effect modification by race/ethnicity, as we had no reason to believe that the association between VMS and BMD would vary by race/ethnicity, and adding the effect modification term to the models would reduce our power to detect the association of interest.

In secondary analyses, we explored the role of circulating estradiol and FSH in the association between VMS and BMD, we sequentially added log-transformed baseline and follow-up serum estradiol and FSH levels and an indicator of whether each serum sample was taken outside of the 2-5-day follicular phase window to the repeated measures mixed models described above. All analyses were performed using SAS version 9.1.3 (SAS Institute, Cary, NC).

Characteristics of the Participants

After censoring participants at the time they reported hysterectomy, bilateral oophorectomy, pregnancy, breastfeeding, or use of medications influencing BMD, the analytic sample consisted of 2213 women who had complete information regarding at least 1 BMD measurement at any anatomical site of interest and at least 1 concurrent VMS assessment (Table 1). Compared to the participants of the SWAN Bone Substudy sites who were excluded from our analytic sample, the participants comprising the analytic sample were significantly less likely to have prior fracture since age 20 (18% vs. 27%, p=0.0007) or to have anorexia or bulimia (1% vs. 3%, p=0.02); they also had lower BMI (28 vs. 32 kg/m², p<0.0001)(data not shown). However, the prevalence of smoking and alcohol consumption, ethnicity, menopause status, and family history of osteoporosis was not different in included vs. excluded participants. Because SWAN is a prospective study of women undergoing the menopause transition, no women were late perimenopausal or postmenopausal at baseline (Table 2). Participants were on average 46 years-old at study entry. The ethnic composition of the analytic sample was as follows: 28% African American, 49% Caucasian, 11% Chinese, 12% Japanese (Table 2). The proportions of women with frequent VMS over the 5 years of follow-up ranged 4.6%-9.3% among premenopausal women, 13%-15% among early perimenopausal women, 30%-39% among late perimenopausal women, and 31%-36% among postmenopausal women (Table 3).

Association between Presence of any VMS and BMD

In models adjusted for baseline age, cumulative time in each menopausal stage, race/ethnicity, study site, and baseline menopause stage, BMD was consistently lower among women with any VMS than among women without VMS, regardless of menopausal stage. The pattern of lower BMD among women with any VMS compared to women without VMS persisted after further adjustment (for baseline height, baseline weight, and change in weight since baseline), was present at all menopausal stages, and was apparent at all three anatomical BMD sites. However, the anatomical BMD site at which the relation between BMD and VMS was statistically significant varied according to menopausal stage. Among postmenopausal women, BMD was statistically significantly lower among women with VMS than among women without any VMS at the lumbar spine (0.008 g/cm² lower, p=0.001) and at the total hip (0.005 g/cm² lower, p=0.008)(data not shown). In contrast, among early perimenopausal women and among premenopausal women, the difference in BMD between women with VMS and women without VMS was statistically significant at the femoral neck (premenopausal women 0.003 g/cm² lower with VMS vs. without VMS, p=0.03; early perimenopausal women 0.003 g/cm² lower with VMS vs. without VMS, p=0.0001)(data not shown).

The patterns of lower BMD among women with VMS compared to without VMS, and the patterns by anatomical site and menopause stage, persisted after additional adjustment for baseline height, baseline weight, and change in weight since baseline (the final model.)

We compared BMD among women with and without VMS after adjustment for baseline age, cumulative time in each menopausal stage, ethnicity, and study site. Differences in mean BMD among women with, compared to women without, VMS, were apparent at the lumbar spine, total hip, and femoral neck, and depended on menopausal stage (Figure 1). For example, compared to postmenopausal women without VMS, postmenopausal women with VMS had significantly lower BMD at the lumbar spine (1.043 g/cm² vs. 1.051 g/cm², p<0.05) and total hip (0.941 g/cm² vs. 0.946 g/cm², p<0.05)(Figures 1a, 1b). Compared to premenopausal women without VMS, premenopausal women with VMS had significantly lower femoral neck BMD (0.833 g/cm² vs. 0.836 g/cm², p <0.05) (Figure 1c). Compared to early perimenopausal women without VMS, early perimenopausal women with VMS had significantly lower femoral neck BMD (0.833 g/cm² vs. 0.837 cm², p<0.05)(Figure 1c).

Association between Frequent vs. Infrequent VMS and BMD

We also compared BMD among women with frequent VMS compared to women without frequent VMS. In models adjusted for baseline age, cumulative time in each menopausal stage, ethnicity, study site, baseline menopause status, baseline height, baseline weight, and weight change since baseline, BMD was consistently lower among women with frequent VMS than among women without frequent VMS. Lumbar spine BMD was statistically significantly lower among women with frequent VMS compared to women without frequent VMS among early perimenopausal women (0.004 g/cm² lower, p=0.003) and among postmenopausal women (0.006 g/cm² lower, p=0.01). Among early perimenopausal women, femoral neck BMD was significantly lower in women with frequent VMS compared to women without frequent VMS (0.003 g/cm² lower, p=0.02). Among early perimenopausal women, the difference in total hip BMD among women with frequent VMS compared to women without frequent VMS was not statistically significant.

Secondary analyses

To explore a possible role of circulating estradiol and/or FSH in the association between VMS and BMD, we added baseline and follow-up serum estradiol levels, as well as an indicator of whether each serum sample was taken outside of the 2-5-day follicular phase window, to our final repeated measures mixed models. The association between VMS and BMD was somewhat weakened by addition of the estradiol parameters, although the difference in mean femoral neck BMD between women with VMS compared to women without VMS remained the same before and after adjustment for current serum estradiol levels in premenopausal women and early perimenopausal women (data not shown). Further adjustment for FSH level did not appreciably influence the results.