Associations of Menopausal Vasomotor Symptoms with Fracture Incidence

Abstract

Context:

Vasomotor symptoms (VMS) are common. Whether VMS are associated with fracture incidence or bone mineral density (BMD) levels is unknown.

Objective:

This study aimed to examine associations of baseline VMS with fracture incidence and BMD.

Design:

This was a prospective observational study with mean (SD) followup of 8.2 (1.7) years (1993–2005).

Setting:

Forty United States clinical centers.

Participants:

We examined data from Women's Health Initiative Clinical Trial participants (n = 23 573) age 50–79 years not using menopausal hormone therapy, and 4,867 participants of the BMD sub-study.

Interventions:

None.

Main Outcome Measures:

We measured baseline VMS, incident adjudicated fractures, and BMD (baseline, annual visits 1, 3, 6, and 9).

Results:

After adjustment for baseline age, body mass index, race/ethnicity, smoking, and education, the hazard ratio for hip fracture among women with baseline moderate/severe VMS (vs no VMS) was 1.78 (95% confidence interval [CI], 1.20–2.64; P = .01). There was no association between VMS and vertebral fracture. VMS severity was inversely associated with BMD during followup (P = .004 for femoral neck, P = .045 for lumbar spine). In repeated measures models, compared with women who reported no VMS, women with moderate/severe VMS had 0.015 g/cm² lower femoral neck BMD (95% CI, −0.025–−0.005) and 0.016 g/cm² lower lumbar spine BMD (95% CI, −0.032–−0.004).

Conclusions:

Women with moderate/severe VMS have lower BMD and increased hip fracture rates. Elucidation of the biological mechanisms underlying these associations may inform the design of preventive strategies for at-risk women prior to occurrence of fracture.

Vasomotor symptoms (VMS), ie, hot flashes and/or night sweats, are common in women during the menopause transition. VMS peak in late perimenopause, when women have had no menses for at least 3 months but not yet experienced 12 months of amenorrhea (1). During the late perimenopause, approximately 60% of women experience VMS (2). VMS may persist well past the final menstrual period. Four years after the final menstrual period, nearly 50% of women are still symptomatic, and 10% of women continue having VMS 12 years or more after the final menstrual period (2). The peak prevalence of VMS coincides with the time period of accelerated bone loss at the hip and spine (3). Given the high prevalence and potentially long exposure to VMS, it is clinically important to examine whether these symptoms are indicative of adverse bone health.

A few studies examined associations between VMS and bone mineral density (BMD); they were cross sectional (4–8), brief (9, 10), or focused on premenopausal infertile women (11). Cross-sectional studies (4, 5, 7, 12), and a 30-month study (10) of peri- and postmenopausal women suggested that spine BMD is lower among women with VMS than among women without VMS. A longitudinal study of younger women progressing through the menopausal transition using repeated measures analysis found that the presence of VMS was inversely associated with spine and femoral neck BMD (13). Likewise, little is known about associations between VMS and fracture risk. A longitudinal study of women who had osteoporosis at baseline found no association of “bothersome” VMS with vertebral or nonvertebral fracture risk (14). To our knowledge, no prospective studies have examined these associations in a large cohort of postmenopausal women not selected for having osteoporosis at baseline.

Our goal was to examine associations of baseline VMS with subsequent BMD and fracture incidence using data from the Women's Health Initiative, a large cohort of United States postmenopausal women (15). We hypothesized that VMS would be associated with lower BMD and higher fracture rates. We also hypothesized that selective estrogen receptor modulator use, physical functioning, incident cancer, and sleep disruption (potentially leading to falls) may be potential mechanisms involved in associations between VMS and fracture risk.

Materials and Methods

Participants

The Women's Health Initiative (WHI) was conducted at 40 Unites States clinical centers (16, 17). The clinical trials (WHI-CT) and the observational study (WHI-OS) enrolled postmenopausal women age 50–79 years at baseline who were free from serious medical conditions (15, 18). The three WHI-CTs evaluated menopausal hormone therapy (HT), calcium and vitamin D supplementation, and a low-fat eating pattern (15).

For the current study, analyses of the associations between VMS and fracture were performed using data from participants of the WHI-CT (Figure 1). Of the 68 132 women enrolled onto the WHI-CT, we excluded data from 588 participants for whom we lacked data regarding VMS severity. Because HT use can influence fracture risk, data from 35 217 participants who reported prior or current HT use, and 13 816 who were randomly assigned to receive HT (conjugated equine estrogens alone or conjugated equine estrogens + medroxyprogesterone acetate), were also excluded. Thus, our analytic sample size for incident fractures was 23 573 participants.

Figure 1.

Flow diagram of analytic cohorts.

Using data from the WHI BMD cohort, we analyzed associations between VMS and BMD. At the time of enrollment, participants at three of the 40 clinical centers (Tucson/Phoenix, Arizona; Pittsburgh, Pennsylvania; and Birmingham, Alabama; n = 10 833) underwent hip and anteroposterior lumbar dual-energy x-ray absorptiometry on a Hologic QDR2000 or 4500W machine (Hologic) using standard protocols for positioning and analysis (19–21). Quality assurance methods included cross-clinic calibration phantoms, further evaluation of scans with specific problems, and review of a random sample of scans (22).

Of 10 833 CT and OS participants from the WHI BMD cohort, 10 358 had at least one BMD measurement at baseline and/or followup. Among these, 5419 participants were excluded because of current or prior use of HT. (Women who reported never having used HT at baseline tended not to initiate HT during followup). A further 72 were missing baseline VMS severity data. Thus, the sample size for our BMD analyses was 4867 participants (Figure 1). Postrandomization BMD measures were censored for participants who were assigned to either HT preparation (n = 626).

Each institution obtained human subjects committee approval. All participants provided written informed consent.

Outcomes: Fracture incidence and BMD

Information regarding incident fractures was self reported semiannually. Self-reported fractures were then confirmed based on medical record adjudication. The questionnaire item asked participants whether they had experienced fracture events since the previous visit: “Has a doctor told you for the first time that you have a new broken, crushed, or fractured bone?” Which bone did you break?” Response choices included: hip, upper leg (not hip), pelvis, knee (patella), lower leg or ankle, foot (not toe), spine or back (vertebra), lower arm or wrist, hand (not finger), elbow, and upper arm or shoulder. For this analysis, we classified fractures into the following categories: hip, spine, and nonvertebral (including hip).

The BMD outcome was absolute BMD (g/cm²), measured at baseline, and annual followup visits 1, 3, 6, and 9 for the WHI-CT, and at baseline and annual followup visits 3, 6, and 9 for the WHI-OS. If information was missing regarding one or more BMD measurements for a participant, the participant still contributed to the model the BMD measurements that were available.

Predictor: Vasomotor symptom assessment

The primary exposure variable was VMS severity. Information regarding VMS at baseline was obtained on the baseline questionnaire. The questionnaire item stated, “Below is a list of symptoms people sometimes have. For each item, mark the one oval that best describes how bothersome the symptom was during the past 4 weeks for you.” Response choices were: symptom did not occur, mild (symptom did not interfere with usual activities), moderate (symptom interfered somewhat with usual activities), and severe (symptom was so bothersome that usual activities could not be performed). The symptom list included separate items for hot flashes and night sweats. As expected, the overlap between hot flashes and night sweats was high. Therefore, in this analysis, we made the a priori decision to define VMS as hot flashes and/or night sweats. Because of the low prevalence of severe VMS, we combined the moderate and severe categories of VMS into a single category for purposes of statistical analysis.

Other questionnaire-based information

We obtained information regarding age, race/ethnicity, education, family income, age at menopause, hysterectomy, oophorectomy, smoking, alcohol consumption, physical activity, previous fracture, sleep disturbance, incident falls, and medication use from baseline self-assessment questionnaires. Age at menopause was defined as previously described (23). Sleep disturbance was characterized using the validated WHI sleep disturbance construct (24). Participants were classified as never, past, or current users of HT based on questionnaire responses. Past HT use was defined as use of an estrogen or progestogen-containing pill or transdermal patch for 3 months or longer following menopause.

Weight and height were measured using standardized protocols. Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared.

Statistical analysis

We examined baseline characteristics across categories of baseline VMS severity after adjustment for age, BMI, and smoking.

We analyzed associations between baseline VMS and incident fractures in the WHI-CT cohort. We analyzed associations between baseline VMS severity (none, mild, moderate/severe) and time to fracture using Cox proportional hazards regression, adjusting for covariates chosen a priori: baseline age, BMI, race/ethnicity, smoking, and education, and stratified by age and prior history of fracture. Hip, spine, and nonvertebral fractures each served as outcomes of separate models, where statistical significance was based on a 2-degree-of-freedom test for association.

In subgroup analyses, we tested for interactions between VMS and baseline characteristics chosen a priori: age, BMI, years since menopause, race/ethnicity, and physical activity. Statistical significance was based on a test of interaction. Twenty comparisons were made (5 outcomes × 4 subgroups); thus, one interaction was expected to be significant by chance alone. Additional post-hoc analyses investigated effect modification by baseline diabetes and history of cardiovascular disease.

Additional analyses were carried out to elucidate potential mechanisms of associations between VMS and fracture risk. Time-dependent covariates (ie, tamoxifen/selective estrogen receptor modulator use, sleep disturbance, physical functioning, incident cancer, number of incident falls) were added one at a time to the Cox regression model in which hip fracture was the outcome.

We analyzed associations between baseline VMS and BMD in the WHI BMD cohort. To do this, we examined associations between baseline VMS severity and longitudinal measures of BMD using repeated measures regression with an unstructured variance-covariance matrix, adjusting for baseline age, BMI, race/ethnicity, BMD scanner, smoking, and education. We compared mean BMD at baseline and during followup (averaged over followup visit) by baseline VMS severity categories using two 1-degree-of-freedom tests for trend. Femoral neck BMD and spine BMD each served as outcomes of separate regression models. Based on a priori hypotheses, we examined the associations of baseline VMS with BMD by subgroups of baseline age (50–59, 60–69, 70–79 y) and BMI (< 25, 25–<30, and ≥ 30 kg/m²) (2, 25). Statistical significance was based on a test for trend of the interaction of VMS severity with BMD at baseline and followup; one interaction was expected to be statistically significant by chance alone.

Because of the known influence of breast cancer therapy on both VMS and BMD, we performed a sensitivity analysis in which we repeated the regression models after excluding women with a history of breast cancer (n = 133).

All analyses were conducted by using SAS software version 9.3 (SAS Institute) and R software version 2.15 (R Foundation for Statistical Computing). All P values were two-sided and P ≤ .05 were regarded as statistically significant.

Results

Characteristics of the participants

Compared with excluded participants, participants in the analytic sample tended to be younger (mean age, 62.3 vs 63.5 y), thinner (mean BMI, 28.6 vs 29.6 kg/m²), and white (82 vs 78%) (data not shown).

At baseline, 15 418 participants (65%) reported having no VMS, 5766 (24%) reported mild VMS, and 2378 (10%) reported moderate/severe VMS (Table 1). Mean (SD) age of the participants of the fracture sample at baseline was 63.5 (7.0) years. Women in the fracture cohort who had moderate/severe VMS at baseline tended to be younger, were more likely to be non-White, and were more likely to have undergone bilateral oophorectomy compared with women with no VMS or mild VMS (all P < 0.05).

Table 1.

Baseline Characteristics of the Fracture Cohort (n = 23 573) by Severity of Baseline Vasomotor Symptoms

Characteristic	Severity of Baseline Vasomotor Symptoms						P Valuea
	None (n = 15 418)		Mild (n = 5766)		Moderate/Severe (n = 2389)
	N	%	N	%	N	%
Age at screening, y							<.001
50–54	1093	7.1	1036	18.0	612	25.6
55–59	2341	15.2	1497	26.0	662	27.7
60–69	7911	51.3	2481	43.0	844	35.3
70–79	4073	26.4	752	13.0	271	11.3
Race/ethnicity							<.001
White	12 746	82.7	4202	72.9	1363	57.1
Black	1490	9.7	1102	19.1	735	30.8
Hispanic	570	3.7	247	4.3	208	8.7
American Indian	63	0.4	32	0.6	13	0.5
Asian/Pacific Islander	355	2.3	106	1.8	28	1.2
Unknown	194	1.3	77	1.3	42	1.8
Education							<.001
≤High school/GED	3716	24.2	1503	26.3	815	34.4
School after high school	5805	37.9	2219	38.8	946	40.0
≥College degree	5806	37.9	1998	34.9	606	25.6
Family income							<.001
<20K	2905	20.2	1025	19.0	654	29.5
20–<35K	4016	27.9	1396	25.9	513	23.1
35–<50K	3113	21.6	1147	21.3	404	18.2
50–<75K	2528	17.5	1016	18.9	356	16.1
≥75K	1850	12.8	798	14.8	289	13.0
US region							.008
Northeast	4825	31.3	1728	30.0	676	28.3
South	3415	22.1	1381	24.0	720	30.1
Midwest	3560	23.1	1423	24.7	562	23.5
West	3618	23.5	1234	21.4	431	18.0
Time since menopause, y							<.001
<5	1102	8.7	1254	25.2	622	29.8
5–<10	1988	15.7	1030	20.7	414	19.9
10–<15	2750	21.7	963	19.4	338	16.2
≥15	6830	53.9	1728	34.7	711	34.1
Hysterectomy	4129	26.8	1518	26.3	824	34.5	.53
Bilateral oophorectomy	1362	9.1	474	8.4	262	11.4	.01
Smoking status							<.001
Never	8097	53.2	2922	51.2	1118	47.5
Past	5970	39.2	2246	39.4	942	40.0
Current	1164	7.6	535	9.4	294	12.5
Treated diabetes (pills or shots)	851	5.5	326	5.7	191	8.0	.11
Alcohol consumption							.002
Non/past drinker	4687	30.6	1755	30.7	898	38.0
<1 drink/wk	5275	34.5	2037	35.7	829	35.1
1–14 drinks/wk	4869	31.8	1750	30.6	581	24.6
>14 drinks/wk	470	3.1	171	3.0	56	2.4
Moderate-strenuous physical activity ≥20 min, episodes/wk							.05
No activity	2665	18.9	1099	21.1	517	23.4
Some activity	6419	45.5	2361	45.4	1000	45.3
2–<4 episodes/wk	2229	15.8	797	15.3	297	13.5
≥4 episodes/wk	2802	19.9	941	18.1	393	17.8
Osteoporosis prescription medication	302	2.0	75	1.3	26	1.1	.43

	Severity of Baseline Vasomotor Symptoms						P Value
	None (n = 15 418)		Mild (n = 5766)		Moderate/Severe (n = 2389)
	Mean	sd	Mean	sd	Mean	sd
Age at screening, y	64.9	6.7	61.2	6.8	59.9	6.9	<.001
BMI, kg/m2	29.5	6.2	29.6	6.1	30.7	6.3	<.001
Total calcium intake, mg	1104.1	674.4	1059.6	663.7	1029.2	745.0	.37
Total vitamin d intake, IU	339.4	260.7	326.4	266.4	305.4	251.5	.14
Physical functioning scale (RAND36)	80.9	19.5	80.9	20.2	75.2	23.6	<.001
Total energy expenditure/wk from phys act, MET-h	10.4	12.5	9.8	12.4	9.4	12.8	.38

Abbreviations: GED, General Educational Diploma; MET-hrs, metabolic equivalent of task (hours/week); RAND36, 36-item Short-Form Health Survey.

^aP values adjusted for age, BMI, race/ethnicity, and smoking.

Likewise, women in the BMD cohort with moderate/severe VMS tended to be younger and non-White (all P < 0.05; Table 2). Average (SD) followup was 8.2 (1.7) years.

Table 2.

Baseline Characteristics of the BMD Cohort (n = 4867) by Severity of Baseline Vasomotor Symptoms

Characterisic	Severity of Baseline Vasomotor Symptoms												P Valuea
	None (n = 3161)				Mild (n = 1157)				Moderate/Severe (n = 549)
	N	%	Mean	sd	N	%	Mean	sd	N	%	Mean	sd
Age at screening, y													<.001
50–54	251	7.9			238	20.6			142	25.9
55–59	437	13.8			264	22.8			127	23.1
60–69	1507	47.7			451	39			199	36.2
70–79	966	30.6			204	17.6			81	14.8
Race/ethnicity													<.001
White	2517	79.6			759	65.6			252	45.9
Black	381	12.1			276	23.9			213	38.8
Hispanic	180	5.7			91	7.9			66	12
American Indian	54	1.7			17	1.5			10	1.8
Asian/Pacific Islander	7	0.2			3	0.3			3	0.5
Unknown	22	0.7			11	1			5	0.9
Education													<.001
≤High school/GED	1083	34.4			395	34.4			251	46.2
School after high school	1134	36			414	36.1			181	33.3
≥College Degree	929	29.5			338	29.5			111	20.4
Family income													.002
<20K	946	32.3			343	32.7			234	46.7
20–<35K	880	30			305	29			125	25
35–<50K	503	17.2			177	16.9			65	13
50–<75K	360	12.3			138	13.1			50	10
≥75K	243	8.3			87	8.3			27	5.4
US region													.08b
Northeast	1277	40.4			406	35.1			143	26
South	801	25.3			403	34.8			246	44.8
West	1083	34.3			348	30.1			160	29.1
Time since menopause, y													<.001
<5	183	7.6			215	22.7			115	25.2
5–<10	308	12.8			183	19.3			71	15.6
10–<15	487	20.2			167	17.6			75	16.4
≥15	1427	59.3			383	40.4			195	42.8
Hysterectomy	1029	32.6			405	35			245	44.7			.18
Bilateral oophorectomy	315	10.3			117	10.6			67	12.8			.07
Smoking status													.01
Never	1797	57.6			638	55.8			272	50.5
Past	1072	34.4			408	35.7			201	37.3
Current	251	8			98	8.6			66	12.2
Treated diabetes (pills or shots)	215	6.8			75	6.5			75	13.7			.22
Alcohol consumption													.13
Non/past drinker	1276	40.7			495	43.2			290	53.8
<1 drink/wk	1027	32.7			372	32.5			146	27.1
1–14 drinks/wk	749	23.9			249	21.7			84	15.6
>14 drinks/wk	85	2.7			30	2.6			19	3.5
Moderate-strenuous physical activity ≥20 min, episodes/wk													.26
No activity	523	18.5			226	22.3			125	25.7
Some activity	1234	43.8			446	43.9			219	45
2–<4 episodes/wk	420	14.9			128	12.6			65	13.3
≥4 episodes/wk	643	22.8			215	21.2			78	16
Osteoporosis prescription medication	55	1.7			10	0.9			5	0.9			.51
Age at screening, y			65.3	7			61.8	7.5			60.6	7.6	<.001
BMI, kg/m2			28.5	6.1			29.1	6			30.7	6.7	.02
Total calcium intake, mg			1085.5	672.4			1055.6	691.4			942.2	592.8	.49
Total vitamin d intake, IU			313.4	252.9			298.2	288.5			272.4	259	.96
Physical functioning scale (RAND36)			80	20			78	20.9			69.9	25.7	<.001
Total energy expenditure/wk from phys act, MET-h			11.5	14.5			10.4	13.2			9.1	13	.34

Abbreviations: GED, General Educational Diploma; MET-h, metabolic equivalent of task (hours/week); RAND36, 36-item Short-Form Health Survey.

^aP adjusted for age, BMI, race/ethnicity, BMD scanner, and smoking.

^bRegion and BMD scanner are confounded so P was not adjusted for scanner.

Associations between VMS and fracture incidence

After adjusting for baseline age, BMI, race/ethnicity, smoking, and education, baseline VMS severity was associated with increased hazard ratio (HR) for fracture (Table 3). Among women reporting baseline moderate/severe VMS (compared with no VMS), the HR for hip fracture was 1.78 (95% confidence interval [CI], 1.20–2.64, P = .01). Higher VMS severity was associated with increased HR for nonvertebral fractures (P = .04). However, VMS severity was not associated with the incidence of clinical spine fractures. VMS were not associated with lower arm/wrist fractures (data not shown).

Table 3.

Multivariable^a-Adjusted Risk of Fracture Associated with Severity of Baseline Vasomotor Symptoms in the Fracture Study Population (n = 23 573), Stratified by Age and BMI

Fracture Site	Subgroup	Severity of Baseline Vasomotor Symptoms
		None		Mild				Moderate/Severe				P Valueb
		N	%	N	%	HR	(95% CI)	N	%	HR	(95% CI)
Hip	Overall	265	0.21	53	0.11	0.93	(0.69–1.27)	29	0.15	1.78	(1.20–2.64)	.01
	Age, y											.90
	50–59	13	0.04	8	0.04	1.08	(0.43–2.70)	5	0.05	1.76	(0.61–5.11)
	60–69	85	0.13	21	0.11	0.92	(0.57–1.49)	10	0.15	1.63	(0.84–3.16)
	70–79	167	0.53	24	0.41	0.90	(0.57–1.41)	14	0.69	1.93	(1.11–3.34)
	BMI											.33
	Normal (<25)	88	0.28	21	0.19	1.26	(0.76–2.09)	7	0.19	1.53	(0.70–3.35)
	Overweight (25–30)	120	0.28	21	0.13	0.79	(0.50–1.27)	12	0.19	1.90	(1.03–3.50)
	Obese (≥30)	57	0.11	11	0.06	0.75	(0.38–1.48)	10	0.11	1.95	(0.98–3.87)
Spine	Overall	226	0.18	66	0.14	1.16	(0.87–1.54)	21	0.11	1.09	(0.68–1.75)	.59
	Age, y											.07
	50–59	29	0.10	13	0.06	0.71	(0.37–1.37)	3	0.03	0.39	(0.12–1.28)
	60–69	84	0.13	31	0.16	1.34	(0.88–2.02)	7	0.11	1.12	(0.52–2.42)
	70–79	113	0.36	22	0.38	1.20	(0.75–1.91)	11	0.54	1.74	(0.91–3.34)
	BMI											.17
	Normal (<25)	76	0.24	15	0.13	0.82	(0.46–1.48)	7	0.19	1.40	(0.63–3.10)
	Overweight (25–30)	75	0.17	24	0.15	1.27	(0.79–2.02)	7	0.11	1.45	(0.65–3.19)
	Obese (≥30)	74	0.15	27	0.14	1.38	(0.88–2.18)	7	0.07	0.75	(0.32–1.74)
Otherc	Overall	2240	1.94	673	1.51	0.93	(0.85–1.02)	289	1.57	1.12	(0.98–1.27)	.04
	Age, y											.77
	50–59	387	1.41	253	1.23	0.91	(0.78–1.07)	133	1.30	1.07	(0.87–1.31)
	60–69	1115	1.88	295	1.57	0.92	(0.80–1.04)	109	1.75	1.15	(0.95–1.41)
	70–79	738	2.55	125	2.32	1.00	(0.82–1.22)	47	2.52	1.14	(0.84–1.55)
	BMI											.49
	Normal (<25)	647	2.26	169	1.61	0.89	(0.74–1.06)	54	1.58	0.92	(0.69–1.23)
	Overweight (25–30)	798	2.00	266	1.73	1.03	(0.89–1.19)	101	1.72	1.21	(0.97–1.50)
	Obese (≥30)	785	1.68	236	1.27	0.88	(0.76–1.02)	130	1.46	1.15	(0.95–1.40)

^aAdjusted for baseline age, BMI, race/ethnicity, smoking, education, and stratified by age group and prior history of fracture.

^bP corresponds to a 2-degree-of-freedom test of association for the 3-level nominal categorical predictor.

^cIncludes hip fracture.

Associations between VMS severity and fracture incidence did not vary across subgroups of baseline age or BMI (Table 3). In addition, there was no interaction of VMS with years since menopause, race/ethnicity, baseline cardiovascular disease, or physical activity (data not shown).

The HR for hip fracture associated with moderate/severe VMS was attenuated by the inclusion of physical functioning in the Cox regression models, although VMS remained statistically significantly associated with hip fracture incidence; the HR (95% CI) changed from 1.78 (1.20–2.64) to 1.64 (1.09–2.46). None of the other potential confounders (tamoxifen/selective estrogen receptor modulator use, sleep disturbance, incident cancer, and incident number of falls) appreciably altered the HR for VMS (attenuation ranging between 0 and 6%), and VMS remained significantly associated with hip fracture incidence.

There were insufficient hip fracture events to allow stratification of results according to presence of diabetes mellitus or prevalent cardiovascular disease. Of the 29 hip fracture cases with moderate/severe symptoms at baseline, only one occurred in a participant who was treated for diabetes at baseline. In the subset of women without diabetes, the HR (95% CI) for hip fracture among women with moderate/severe VMS, compared with women without VMS, was 1.88 (1.25–2.81). VMS were not associated with diabetes, and adding diabetes as a covariate did not appreciably change the results: hip fracture HR (95% CI) was 0.93 (0.69–1.26) for mild VMS (vs none); and 1.77 (1.19–2.63) for moderate/severe VMS (vs none; P = .01).

In a sensitivity analysis, we excluded data from participants who reported use of selective estrogen receptor modulators (n = 1038) or prescription osteoporosis medications (n = 3621) at baseline or any time during the followup period. Results were similar to the results of the primary analyses (data not shown).

Associations between VMS and BMD

The baseline severity of VMS was not associated with baseline femoral neck or spine BMD. Associations between baseline VMS severity and baseline BMD did not significantly differ across categories of BMI, age, or years since menopause (interaction P > 0.05), except that associations of VMS severity with baseline lumbar spine BMD differed according to BMI category (interaction P = 0.04). In the subgroup of women with BMI ≥ 30 kg/m², compared with women who reported no VMS, women with moderate/severe VMS had lower baseline femoral neck BMD (−0.015 g/cm²; 95% CI, −0.029–−0.001 g/cm²) and baseline lumbar spine BMD (−0.024 g/cm²; 95% CI, −0.045–−0.002 g/cm²).

After adjusting for baseline age, BMI, race/ethnicity, smoking, and education, VMS severity was inversely associated with mean followup BMD both at the femoral neck and at the spine BMD (P for test of trend = .004 for femoral neck and .045 for spine). Compared with women who reported no VMS, women with moderate/severe VMS had 0.015 g/cm² lower femoral neck BMD (95% CI, −0.025–−0.005, g/cm²) and 0.016 g/cm² lower lumbar spine BMD (95% CI, −0.032–−0.004, g/cm²). Associations between VMS severity and BMD during followup did not differ by baseline age, years since menopause, or baseline BMI category (P > .05 for trend test). We show the trajectories of femoral neck BMD (Figure 2) and lumbar spine BMD (Figure 3) over time for each category of VMS severity (no VMS, mild VMS, moderate/severe VMS). Compared with participants without VMS, participants with moderate/severe VMS, had a steeper decline in femoral neck BMD and blunted increase in spine BMD.

Figure 2.

Femoral neck BMD by baseline vasomotor symptoms.

Figure 3.

Lumbar spine (L2–L4) BMD by baseline vasomotor symptoms.

In a sensitivity analysis that excluded women with a history of breast cancer at baseline (n = 133), associations between VMS and BMD were similar. Compared with women who reported no VMS, women with moderate/severe VMS had 0.013 g/cm² lower femoral neck BMD (95% CI, −0.024–−0.003, g/cm², P = .01) and 0.017 g/cm² lower lumbar spine BMD (95% CI, −0.033–−0.007, g/cm², P = .04).

Discussion

In this prospective study, compared with women who did not have VMS, women with moderate/severe VMS had lower BMD over time at the femoral neck and lumbar spine, and increased rates of hip fractures during an average (SD) of 8.2 years of followup. These associations did not differ by baseline age or BMI category. Despite being younger and heavier than asymptomatic women, characteristics associated with higher BMD, women with moderate/severe VMS had a higher risk of hip fractures that was also independent of other established risk factors for fractures.

To our knowledge, only one study has examined associations of VMS with subsequent fracture rates. That study included postmenopausal women with osteoporosis by BMD or prevalent fractures and found no association between hot flashes “that bothered them or interfered with their lives” in the previous 6 months with vertebral or nonvertebral fractures (14). However, followup duration was only 3 years and information regarding fracture locations was not analyzed. Two thirds of participants were randomly assigned to receive raloxifene, which is known to cause hot flashes in some women. The etiology of these hot flashes may be different than menopause-related VMS and may explain the discrepancy between results of the prior study and ours. Other longitudinal studies have found that postmenopausal women with VMS had lower BMD than those without VMS (13, 14). An additional longitudinal study found an association between “sweating” frequency and more rapid rates of loss in forearm BMD (9).

Lower estradiol levels in women with hot flashes may partly explain the associations we found between VMS and decreased bone density. We and others have previously found that higher estradiol levels are associated with lower rates of bone loss over the menopausal transition (26–28), and some longitudinal studies have found lower estradiol levels among women with VMS (29–31). Associations of estrogen level with VMS are likely complex, and may also depend on whether women have residual intermittent ovulatory activity (32). Information on estrogen levels is not available for the present study.

A possible explanation for our findings may be related to balance. However, moderate/severe VMS were not associated with increased risk for lower arm/wrist fracture. Moreover, including incident falls as a covariate did not attenuate the VMS HR for hip fracture. Our analysis does suggest that impaired physical functioning may partially explain the association between moderate/severe VMS and hip fracture. Falls are multifactorial and it is likely that we saw no attenuation because of our inability to directly control for balance.

In this study, increases in spine BMD over time were more pronounced among women without VMS than among women with moderate/severe VMS, yet spine BMD increased over time in all groups. A previous cross-sectional study found that age and BMD were strongly inversely related at the hip, but not at the lumbar spine (33). This may be because typical anteroposterior projection used in dual-energy x-ray absorptiometry to measure spine BMD is confounded by degenerative changes due to aging. For example, spinal degeneration and hyperostosis (osteophytes), vertebral fracture, and aortic calcification can spuriously increase bone density (34). It is also possible that increases in spine BMD are driven by the subgroup of WHI participants who received calcium and vitamin D in the WHI Calcium + Vitamin D trial (35).

Our study has limitations. Participants of WHI were healthier than similarly aged women in the general population, and the incidence of fractures in this cohort was low. Also, VMS were self reported and assessed at a single baseline visit. However, we suspect that this means of assessing VMS reflects what occurs in routine clinical practice. We cannot rule out residual confounding. The associations among age, VMS, BMD, fracture, and HT use in an observational setting are complex. However, in the current study, it is unlikely that HT use confounded our results because we included only women who reported that they had never used HT at baseline, and the overwhelming majority of these women did not initiate HT during followup. Because BMD was available only in a subset of participants (n = 4867), we were unable to adjust for BMD in the fracture models. Insufficient numbers of hip fracture events precluded a more detailed investigation into the joint effects of hot flash severity and night sweat severity and their corresponding interactions. Finally, information regarding serum levels of estrogen, vitamin D, and bone turnover markers was not available.

Strengths of our study include the large number of participants, the long followup duration, the collection of information regarding both hot flashes and night sweats, the direct measurement of BMI (which is associated with both VMS and BMD), the adjudication of fracture endpoints, and the detailed information regarding fracture risk factors and anatomical locations of fractures.

In conclusion, we found that women with moderate/severe VMS have lower BMD and increased hip fracture rates over followup. VMS are common, experienced by at least 60% of women (2). Information regarding VMS is easily assessable by clinicians. Women with VMS may benefit from greater attention to healthy lifestyle habits to maintain bone health. Elucidation of the biological mechanisms underlying these associations may inform the design of preventive strategies for at-risk women prior to occurrence of fracture.