Nocturnal Hot Flashes,but Not Serum Hormone Concentrations,as a Predictor of Insomnia in Menopausal Women: Results from the Midlife Women's Health Study

Abstract

Background:

Sleep disruptions are among the most common symptoms experienced during menopause and can be associated with depression, hot flashes, and fluctuating hormones. However, few studies have examined how such risk factors influence sleep in midlife women in a network-based approach that will establish the complex relationship between variables.

Materials and Methods:

We used a Bayesian network (BN) to examine the relationship between multiple factors known to influence sleep and depression in midlife women, including hormone concentrations, hot flashes, and menopause status among participants of the longitudinal Midlife Women's Health Study. In year 1, 762 women (45–54 years of age) answered questions regarding the frequency of insomnia, hot flashes, and depression; 389 of the same women answered similar questions at year 4. We measured serum hormones and calculated free estradiol index, free testosterone index, and ratios of estradiol:progesterone, and estradiol:testosterone. For our model, we calculated the change in frequency of insomnia, depression, and covariates (body mass index, menopause status, hot flashes at night, and present quality of life) from year 1 to 4.

Results:

Using a BN, we found that self-reported hot flashes at night, and no other factors, were direct predictors of self-reported insomnia in year 1. Surprisingly, we did not identify an association between hormone concentrations and self-reported insomnia. Frequency of insomnia in year 4 was only predicted by frequency of insomnia in year 1, whereas frequency of depression in year 4 was predicted by year 4 insomnia and frequency of depression in year 1. No other factors were direct predictors of insomnia or depression in our model.

Conclusions:

Therefore, hot flashes at night, previous insomnia, and depression are stronger predictors of how women will self-report frequency of sleep disruptions and treatment may reduce menopausal sleep complaints.

Introduction

Several studies, including a recent meta-analysis,¹ show that women's sleep quality declines as they progress through premenopausal, perimenopausal, and postmenopausal stages.^2

–5 Additionally, women are two to five times more likely to experience depression during their perimenopausal transition compared with when they are premenopausal.⁶ Furthermore, risk factors known to influence sleep quality also change across the menopausal transition.^5,7,8 How these changes in risk factors contribute to the decline in sleep quality remains poorly understood. Importantly, women who report sleep difficulties during midlife often also report reduced quality of life and overall health.⁹ Therefore, it is critical to understand whether factors such as endogenous hormone concentrations and hot flashes are influencing depression and sleep quality through the menopause transition.

The change in reproductive hormone patterns during midlife is hypothesized to be responsible for the prevalence of poor sleep quality and depression in midlife women. During reproductive aging, a reduction in ovarian follicles results in a significant drop in estradiol and progesterone.^10,11 Consequently, a decline in the negative feedback at the level of the pituitary leads to the increase in follicle-stimulating hormone (FSH) observed during the menopausal transition and postmenopause period.^11,12 As FSH concentrations increase and estradiol concentrations decrease, women have increased odds of self-reporting impaired sleep quality.^5,13

–16 Furthermore, the ratio of estradiol to progesterone is positively associated with women self-reporting insomnia, indicating that women with relatively higher levels of estradiol to progesterone experience more frequent insomnia.¹⁶ However, these findings are not always consistent. Some studies find no association between FSH or estradiol and self-reported sleep quality,^17
–19 even when controlling for covariates such as race, psychological factors, perceived stress, and menopause status.

Midlife women with subjective or objective measures of hot flashes experience poorer subjective sleep quality,^{2,5,20

–23} including trouble falling asleep, staying asleep, and general sleep disturbances.²² Additionally, several studies have found associations between depressive symptoms and hot flashes in midlife women.^24
–26 Importantly, there is likely a strong bidirectional relationship between sleep and depression in midlife women. Indeed, studies using self-reported depressive symptoms measures such as the Center for Epidemiologic Studies Depression Scale find that increased frequency and severity of depression are associated with poorer subjective sleep quality.^{2,8,18,27
–29}

Many factors associated with sleep quality or depression, particularly hot flashes, are often associated with each another. A meta-analysis found that perimenopausal women with vasomotor symptoms were more likely to report depression than pre- or postmenopausal women.²⁶ The Penn Ovarian Aging Study found that 41% of women who did not have depression or hot flashes upon enrollment developed both hot flashes and depression during the menopausal transition.²⁴ Importantly, this study also found that depression may precede the onset of hot flashes, indicating that depression may be a critical risk factor for the development of vasomotor symptoms and sleep disturbances.

The interaction between multiple factors and their influence on sleep as well as how these factors change over the menopausal transition remains largely understudied. Some studies have examined how various symptoms, including sleep, cluster together in midlife women.^30

–35 However, there is a lack of analyses specifically focusing on what collective symptoms influence sleep in midlife women. In this study, utilizing a novel, network-based approach, we examined the relationship between risk factors known to influence sleep in participants of the Midlife Women's Health Study (MWHS). Specifically, we used a Bayesian network (BN) analysis that identifies the most likely relationship between variables while considering all factors included in the model. We tested the overall hypothesis that changes in variables, such as depression, hormone concentrations, and hot flashes over the course of the menopausal transition, are associated with the frequency of self-reported insomnia in midlife women.

Materials and Methods

Description of study and overall sampling methods

The MWHS is a longitudinal cohort study designed to identify risk factors for menopausal symptoms, among midlife women. The participants of the MWHS submitted written informed consent according to the procedures provided by the Institutional Review Boards of the University of Illinois and Johns Hopkins University, both of which approved this research. The parent study design is described elsewhere.³⁶ Briefly, women were eligible for the parent study if they were between 45 and 54 years of age and had intact ovaries and uterus and were enrolled if they were not postmenopausal (i.e., they reported having three menstrual periods within the last 12 months). Women were excluded if they were taking hormone replacement therapy or hormonal contraception, had a history of reproductive cancers, or if they were currently pregnant.

Women who were enrolled in the study made an initial clinic visit where they completed a questionnaire regarding demographics, reproductive history, menstrual cycles, hormone therapy and contraceptive use, medical therapy, physical symptoms, sleep characteristics, depressive symptoms, and health behaviors (including smoking habits, alcohol use, and amount of activity). Women also submitted a blood and urine sample and had height and weight measured. Following the first visit, women returned once a week for 3 more weeks to provide urine and blood samples. Furthermore, all clinic visits were scheduled in the morning to account for diurnal fluctuations in hormones.^37

–40

For the next 3 years, participants returned to the clinic once a year and completed follow-up questionnaires. These questionnaires repeated previous questions about factors such as menstrual cycles, health status, lifestyle, depressive symptoms, and sleep. Additionally, blood and urine samples were taken weekly for 4 weeks each year. We used the biographical, hormone concentrations, and survey responses from year 1 and 4 for this study.

Description of sleep and depression measures

At each clinic visit, women self-reported the frequency of their subjective insomnia and depressive symptoms through three separate questions as previously described.⁸ Women were asked, “Please indicate how frequently you experienced insomnia during the past year,” and “Please indicate how frequently you experienced depression during the past year.” The response to these questions were on a five-point Likert scale (never, less than once per month, one to four times per month, two to four times per week, or more than five times per week). Insomnia and depression were assessed at each annual clinic visit. For the present study, responses to the insomnia and depression questions at years 1 and 4 were used for analysis. The questions were based on questions from the Midlife in the United States study and have been previously validated.^8,9

Frequency of insomnia and depression were dichotomized as less than monthly (<Monthly) or ≥Monthly in year 1 and 4. Specifically, women were categorized as <Monthly if they responded, “Never,” or “Less than once per month.” Women were categorized as ≥Monthly if they responded, “One to four times per month,” “Two to four times per week,” or “More than five times per week.”

Measures of serum hormones

Estradiol, progesterone, and testosterone levels were measured in duplicate using enzyme-linked immunosorbent assays (DRG International, Springfield, NJ, USA). The minimum detection limits and intraassay coefficients of variation were as follows: estradiol 7 pg/mL, 3.3 ± 0.17%CV; progesterone 0.1 ng/mL, 2.1 ± 0.65%CV; and testosterone 0.04 ng/mL, 2.2 ± 0.56%CV. Limit of detection was used for the hormone value for values lower than the limit of detection for the assay. All hormone values were log-transformed to meet normality assumptions.

Covariate data collection

Covariates included race, menopause status, body mass index (BMI), having hot flashes at night, smoking status, and present quality of life (QOL). Covariates were assessed at each annual clinic visit. For the present study, covariate data from years 1 and 4 were used for analysis.

In each annual survey, hot flashes were defined as, “a sudden feeling of heat in the face, neck, or upper part of the chest. Hot flashes are often accompanied by reddening or flushing of the skin followed by sweating and chills.” The self-reported number of hot flashes at night was determined by the question, “On average, how many hot flashes do you experience every night (between 9:00 p.m. and 6:00 a.m.)?”

Participants were also asked about their QOL using the Cantril's Self-Anchoring Ladder of Life.^41,42 Participants indicated a number ranging from 0 (indicating worst possible life) to 10 (indicating best possible life). Due to the results of QOL being skewed, women responding 1–6 were categorized as having low QOL, 7–8 was moderate QOL, and 9–10 was high QOL.⁴¹

Menopause status was determined using the following categorization: (1) women who had their most recent menstrual period in the last 3 months and experienced ≥11 periods within the past year were categorized as premenopausal; (2) women who had reported their last menstrual period (a) within the past year, but not within the past 3 months or (b) within the last 3 months, but experienced ≤10 periods within the past year were considered perimenopausal; and (3) women who reported not having a menstrual period within the last year were considered as postmenopausal. Women who were considered postmenopausal at the first clinic visit were not included in the present study, with only pre- and perimenopausal women beginning the study in the first year.

BN analysis

To identify what variables are the most likely predictors of self-reported sleep quality in our population, we used a BN model similar to others previously described for this population.⁴³ Briefly, a BN model indicates connections between observed values for variables entered into the model. Parent nodes are the source of an arrow, while the child node is the variable to which the arrow points. A child node can have multiple parent nodes. Importantly, the values of a child node depend on a combination of the parent nodes. The type of variable of the child node determines the statistical model used to assess the relationship between the parent and child nodes. Specifically, a continuous child node uses linear regression, a binary child uses logistic regression, and a count child uses Poisson regression.

Possible nodes for this model include race (white vs. nonwhite), and year 1 and 4 values for hormones, BMI, smoking status, menopause status, present QOL, and hot flashes at night. Additionally, depression and insomnia in years 1 and 4 were also included as possible parent nodes. All models were fit using the bnlearn package⁴⁴ in R version 3.6.2 (Dark and Stormy Night).⁴⁵

Next, we identified all possible configurations (i.e., all possible parent–child combinations) and computed the likelihood of each configuration based on the data observed in this population. As we were only interested in whether year 1 variables predicted responses in year 4, all variables from year 4 were banned from being parent nodes for year 1 variables. Additionally, race was banned from being a potential child node of any variable. Smoking status and menopause status from year 1 were the only potential parent node for smoking status and menopause status in year 4, respectively. Hot flashes at night, depression, insomnia, and present QOL were banned from being parent nodes for all hormone nodes. Finally, insomnia and depression were banned from being parent nodes for hot flashes at night. All other possible connections were left as potential pathways. The optimal network based on all possible associations between variables was plotted. Year 1 and 4 data were separated and color coded.

Vectors were plotted and indicate strong relationships between the parent and child nodes. To understand what variables influenced our nodes of interest (i.e., depression and insomnia in years 1 and 4), we determined conditional probabilities based on the final network configuration.

Results

Description of study sample

Of 775 women in the MWHS, 742 in year 1 provided samples for hormone measures and responded to the survey questions used in this study, whereas, in year 4, 389 women provided samples for hormone measures and completed the survey questions. In year 1, 305 women were premenopausal, and 463 women were perimenopausal. Postmenopausal women were excluded from enrolling in the study. By year 4, 144 women were still premenopausal, 124 women were perimenopausal, and 120 women were postmenopausal. The population characteristics for both years of this study sample are shown in Table 1A.

Table 1A.

Selected Characteristics of the Study Sample at Year 1 (n = 772) and Year 4 (n = 389)

	Year 1		Year 4
	N	%	n	%
Race
White	504	67.9	262	69.5
Black	238	32.1	115	30.5
Menopause status
Pre-	305	39.7	144	37.1
Peri-	463	60.3	124	32.0
Post-	0	0	120	30.9
Hot flashes at night
No	499	70.9	190	57.4
Yes	205	29.1	141	42.6
Smoking status
Nonsmoker	422	54.6	207	53.2
Former smoker	271	35.1	147	37.8
Current smoker	80	10.3	35	9.0
Insomnia
Never	228	29.7	109	28.1
Rarely (<1/month)	193	25.2	93	24.0
Sometimes (1–4/month)	163	21.3	98	25.3
Frequently (2–4/week)	113	14.7	64	16.5
Regularly (>5/week)	70	9.1	24	6.2
Depression
Never	273	35.6	165	42.4
Rarely (<1/month)	280	36.5	116	29.8
Sometimes (1–4/month)	142	18.5	79	20.3
Frequently (2–4/week)	45	5.9	15	3.9
Regularly (>5/week)	27	3.5	14	3.6
	Year 1		Year 4
	Median	IQR	Median	IQR
BMI	26.5	(23.0, 32.1)	27.5	(23.6, 32.7)
Present QOL	8	(7, 9)	8	(7, 8)
Hormones
Estradiol (pg/mL)	49.62	(31.83, 71.24)	35.56	(19.32, 60.55)
Progesterone (ng/mL)	0.63	(0.23, 1.19)	0.19	(0.10, 0.66)
Testosterone (ng/mL)	0.2	(0.18, 0.41)	0.36	(0.27 0.50)
Estradiol:progesterone	0.08	(0.4, 0.17)	0.14	(0.07, 0.26)
Estradiol:testosterone	0.16	(0.10, 0.28)	0.09	(0.06, 0.15)

BMI, body mass index; IQR, interquartile range; Peri-, perimenopausal; Pre-, premenopausal; Post-, postmenopausal; QOL, quality of life.

Table 1B.

Categorization of Insomnia and Depression Outcomes in Year 1 and 4

	Year 1		Year 4
	n	%	n	%
Insomnia
<Monthly	421	54.9	202	52.1
≥Monthly	346	45.1	186	47.9
Depression
<Monthly	553	72.1	281	72.2
≥Monthly	214	27.9	108	27.8

<Monthly includes women who self-reported insomnia or depression either “never” or “rarely.”

≥Monthly includes women who self-reported insomnia or depression either “sometimes,” “frequently,” and “regularly.”

Categorization of sleep and depression measures

In year 1, 45.1% of women reported regular or frequent (i.e., ≥Monthly) insomnia and 27.9% of women reported regular or frequent depression (Table 1B). This distribution remained similar in year 4, with 47.9% of women reporting ≥Monthly insomnia and 27.8% of women reporting ≥Monthly depression.

Overview of BN

We were interested in identifying how variables may be interacting to influence the frequency of sleep disruptions in our study population. Therefore, we performed a BN, with the final BN model presented as the best-fit model given the sampled data (Fig. 1). Light gray nodes and arrows indicate variables from year 1, whereas dark gray nodes and arrows indicate variables from year 4. Nodes are directly impacted by the arrows pointing to it. We found race and both year 1 smoking and menopause status influenced hormone concentrations in year 1. Hormones in year 4 were influenced by more variables, including race, year 1 and year 4 smoking status, year 4 menopause status and BMI, and year 1 testosterone and estradiol:testosterone ratio. Importantly, year 4 BMI, menopausal status, and smoking status were dependent upon their respective year 1 values; therefore, year 4 hormones were indirectly dependent upon those variables, as well.

FIG. 1.

Bayesian network of selected characteristics across years 1 and 4. Network includes all selected characteristics from years 1 (light gray) and 4 (dark gray). Arrow indicates a strong association between the two variables. Dashed arrows specifically indicate association between race and other variables (for ease of visualization). No arrow indicates either weak or lacking relationship.

Unexpectedly, we found that hot flashes at night was the only factor directly associated with insomnia in year 1 (Fig. 1). This was despite including other factors in the BN known previously to be associated with insomnia, including endogenous hormone levels, menopause status, and smoking status. Hot flashes at night in year 1 were also dependent upon menopause status, making insomnia indirectly dependent on menopause status. Depression in year 1 and insomnia in year 4 were dependent upon year 1 insomnia, and thus all the parent nodes leading to year 1 insomnia (i.e., hot flashes at night and menopause status). Depression in year 4 was influenced by insomnia in year 4 and depression in year 1, and thus all parent nodes leading to year 1 insomnia and depression.

Relationship among hot flashes at night, depression, and insomnia

We were particularly interested in what factors influenced depression and insomnia. Thus, we determined the direction and size of impact of parent nodes on select outcomes using conditional probabilities. These probabilities were based upon what parent nodes directly influenced each symptom from the BN. We determined the probabilities of women reporting insomnia in year 1 or 4 and depression in year 4 (Table 2). In each table, the number of women reporting at least monthly symptoms in each category are presented (n) with the probability (prob) of reporting monthly as opposed to less than monthly. Specifically, we observed that women who self-reported hot flashes at night in year 1 were more likely to report more frequent insomnia in year 1. Furthermore, we observed that women who report more frequent insomnia in year 1 were more likely to report frequent insomnia in year 4.

Table 2.

Conditional Probabilities of Selected Nodes from the Bayesian Network (Fig. 1)

(A) Probability of reporting ≥Monthly insomnia in year 1
Hot flashes at night Y1	n	(prob)
No	73	(0.37)
Yes	50	(0.60)
(B) Probability of reporting ≥Monthly insomnia in year 4
Insomnia year 1	n	(prob)
<Monthly	41	(0.26)
≥Monthly	92	(0.75)
(C) Probability of reporting ≥Monthly depression in year 4
	Depression in year 1
	<Monthly		≥Monthly
Insomnia year 4	n	(prob)	N	(prob)
<Monthly	18	(0.62)	10	(0.91)
≥Monthly	110	(0.22)	71	(0.81)

(A) Conditional probabilities of women reporting ≥Monthly insomnia in year 1 based on whether women reported hot flashes at night in year 1. (B) Conditional probabilities of women reporting ≥Monthly insomnia in year 4 based on whether women reported ≥Monthly insomnia in year 1. (C) Conditional probabilities of women reporting ≥Monthly depression in year 4 based on whether women reported <Monthly or ≥Monthly depression in year 1 or <Monthly or ≥Monthly insomnia in year 4.

n, number; prob, probability.

We also observed that as women reported more frequent insomnia in year 4 and depression in year 1, they were more likely to also report frequent depression in year 4. Interestingly, women who reported neither frequent insomnia in year 4 nor frequent depression in year 1 had a higher probability of reporting frequent depression in year 4 compared with women who reported frequent insomnia in year 4 and no frequent depression in year 1.

Discussion

By applying a BN, a method that evaluates the relationship between all potential factors included in the model, the present study identified self-reported hot flashes at night, and no other variables were directly associated with self-reported insomnia in participants of the MWHS. Furthermore, we identified that self-reported insomnia, and no other variables, were directly associated with depression in participants of the MWHS. This was despite accounting for additional variables known to be associated with sleep and depression in midlife women, including endogenous hormone concentrations, smoking status, and menopausal status. Specifically, we found that women who report frequent hot flashes at night in year 1 of the study were more likely to report frequent insomnia in year 1 compared with women who did not report frequent hot flashes. Additionally, we found that women who report frequent insomnia in year 1 or 4 are more likely to self-report frequent depression in year 1 or 4, respectively.

These findings suggest that, when taking into consideration the complex relationship between multiple potential factors in a network-based statistical model, hot flashes are the strongest predictors of insomnia, and frequency of insomnia is the strongest predictor of depression in participants of the MWHS.

One critical finding from our model is that endogenous hormone concentrations, including estradiol and progesterone, were not associated with sleep disruptions or depression. This is contrary to previous work indicating hormones are associated with self-reported sleep disruptions and depression among midlife women. However, not all studies show relationships between hormones and self-reported sleep.^{8,13,14,18,46} Furthermore, only a few studies have used a network-based approach to examine the relationship between multiple symptoms in midlife women while also considering hormone concentrations.^43,47,48 Therefore, it remains unclear if and how hormones are associated with self-reported sleep symptoms when analyzed in an approach that accounts for potential relationships between multiple variables.

We also did not identify menopause status as a direct predictor of insomnia or depression in our model. However, menopause status was a predictor of frequency of hot flashes in year 1. Due to the nature of the results of our BN, it is likely that menopause status is indirectly related to the frequency of insomnia in our population. This is consistent with previous studies that have identified associations between menopause status and menopausal symptoms, including poor sleep and hot flashes.^{1,2,5,19,49,50} Overall, the findings from our network analysis are similar to those findings from previous studies that have identified clusters of symptoms among menopausal women,^30

–35 despite having been examined using different study populations and statistical approaches.

The current findings expand previous reports from the MWHS and provide additional understanding of what covariates, including hot flashes at night, are associated with self-reported insomnia and depression in midlife women. Findings from our analysis suggest that women who self-reported hot flashes at night in year 1 were more likely to self-report more frequent insomnia in year 1. This is consistent with a previous MWHS analysis that showed increased menopausal symptoms, such as hot flashes and night sweats, predicted more frequent insomnia among peri- and postmenopausal women.⁸ Interestingly, in the present analysis, we did not observe that depression was a predictor of insomnia in year 1 or 4. Alternatively, we observed that women who report more frequent depression in year 4 reported more frequent insomnia in year 4 and more frequent depression in year 1. Additionally, we found that women who report frequent insomnia in year 1 were also more likely to report frequent insomnia in year 4.

Interestingly, we found that women who did not report frequent insomnia in year 4 and frequent depression in year 1 were more likely to report frequent depression in year 4 compared with women who only reported frequent insomnia in year 4 (and not frequent depression in year 1). It is possible that this difference in probability is due to sample size, or it could be due to additional factors that could impact depression that were not included in the model. Regardless, our overall findings suggest that how midlife women report insomnia and depression are predictors for how they will report the same symptoms in subsequent years, allowing for earlier intervention with symptoms that may negatively impact quality of life as women progress through menopause. These findings add to our understanding of what variables are associated with sleep and depression and highlight the importance of using network-based approaches to examining such complex relationships between variables.

One limitation of our study is that this was a secondary analysis and the only data available for insomnia and depression for both years 1 and 4 were based on limited questions given to participants. The questions used for insomnia and depression were not operationalized for study participants, potentially biasing sampling. Additionally, there are multiple possible reasons why some women were not retained across the 4 years of the study, including moving, losing interest, or increased burden of participating. However, another limitation of our study is that we are unable to completely examine the differences between women who did have year 4 data to those who did not have year 4 data, thus preventing us from gaining a full picture of the relationship between year 1 and 4. Furthermore, we averaged hormone samples from our participants over 4 consecutive weeks each year to control for stages of menstrual cycles. However, we did not obtain specific information regarding the stage of the menstrual cycle from women who were still cycling.

Although this is a caveat to the interpretation of our hormone analysis, the method of hormone sampling is consistent with many studies that evaluate the relationship between hormones and menopausal symptoms. Nevertheless, the geometric mean of hormones does not take into account the nuanced relationship between menstrual cycle stage and subjective sleep (Reviewed in Ref.⁵¹).

Conclusions

We used a BN approach to analyze the relationship between factors associated with menopause. We found that hot flashes at night are the strongest predictors of self-reported frequency of insomnia in participants of the MWHS. Furthermore, when multiple covariates are taken into consideration within the same model, hormone concentrations are not directly related to insomnia in our population. Future studies should use similar network-based approaches to better understand the associations between multiple menopausal symptoms. Based on the results of this study, women experiencing poor sleep during menopausal transition could address this in part by seeking evaluation of and treatment for hot flashes, particularly hot flashes at night.

Footnotes

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This research was supported by the National Institutes of Health (Grant No. R01 ES026956-03). C.C. was supported by the National Institutes of Health (Grant No. T32 ES007326). K.M.H. was supported by the Interdisciplinary Environmental Toxicology Program and the Carle Foundation Hospital Seed Grant.

References

, Lang

. Examining the relationship between subjective sleep disturbance and menopause: A systematic review and meta-analysis. Menopause, 2014; 21(12):1301–1318; doi:10.1097/GME.0000000000000240

Woods

, Mitchell

. Sleep symptoms during the menopausal transition and early postmenopause: Observations from the Seattle Midlife Women's Health Study. Sleep, 2010; 33(4):539–549.

Tom

, Kuh

, Guralnik

, et al. Self-reported sleep difficulty during the menopausal transition: Results from a prospective cohort study. Menopause, 2010; 17(6):1128–1135; doi:10.1097/gme.0b013e3181dd55b0

Shin

, Lee

, et al. Prevalence of insomnia and its relationship to menopausal status in middle-aged Korean women. Psychiatry Clin Neurosci, 2005; 59(4):395–402; doi:10.1111/j.1440-1819.2005.01391.x

Kravitz

, Zhao

, Bromberger

, et al. Sleep disturbance during the menopausal transition in a multi-ethnic community sample of women. Sleep, 2008; 31(7):979–990.

Bromberger

, Epperson

. Depression during and after the perimenopause: Impact of hormones, genetics, and environmental determinants of disease. Obstet Gynecol Clin North Am, 2018; 45(4):663–678; doi:10.1016/j.ogc.2018.07.007

Lampio

, Polo-Kantola

, Himanen

, et al. Sleep during menopausal transition: A 6-year follow-up. Sleep, 2017; 40(7); doi: 10.1093/sleep/zsx090

Smith

, Flaws

, Mahoney

. Factors associated with poor sleep during menopause: Results from the Midlife Women's Health Study. Sleep Med, 2018; 45:98–105; doi: 10.1016/j.sleep.2018.01.012

Karlson

, Gallagher

, Olson

, et al. Insomnia symptoms and well-being: Longitudinal follow-up. Health Psychol, 2013; 32(3):311–319; doi:10.1037/a0028186

10.

Hale

, Robertson

, Burger

. The perimenopausal woman: Endocrinology and management. J Steroid Biochem Mol Biol, 2014; 142:121–131; doi: 10.1016/j.jsbmb.2013.08.015

11.

Burger

, Hale

, Dennerstein

, et al. Cycle and hormone changes during perimenopause: The key role of ovarian function. Menopause, 2008; 15(4 Pt 1):603–612; doi:10.1097/gme.0b013e318174ea4d

12.

Hall

. Endocrinology of the menopause. Endocrinol Metab Clin North Am, 2015; 44(3):485–496; doi:10.1016/j.ecl.2015.05.010

13.

Hollander

, Freeman

, Sammel

, et al. Sleep quality, estradiol levels, and behavioral factors in late reproductive age women. Obstet Gynecol, 2001; 98(3):391–397.

14.

Sowers

, Zheng

, Kravitz

, et al. Sex steroid hormone profiles are related to sleep measures from polysomnography and the Pittsburgh Sleep Quality Index. Sleep, 2008; 31(10):1339–1349.

15.

de Zambotti

, Colrain

, Baker

. Interaction between reproductive hormones and physiological sleep in women. J Clin Endocrinol Metab, 2015; 100(4):1426–1433; doi:10.1210/jc.2014-3892

16.

Gallicchio

, Schilling

, Romani

, et al. Endogenous hormones, participant characteristics, and symptoms among midlife women. Maturitas, 2008; 59(2):114–127; doi:10.1016/j.maturitas.2008.01.003

17.

Freeman

, Sammel

, Gross

, et al. Poor sleep in relation to natural menopause: A population-based 14-year follow-up of midlife women. Menopause, 2015; 22(7):719–726; doi:10.1097/GME.0000000000000392

18.

Pien

, Sammel

, Freeman

, et al. Predictors of sleep quality in women in the menopausal transition. Sleep, 2008; 31(7):991–999.

19.

Jones

, Zak

, Lee

. Sleep disturbances in midlife women at the cusp of the menopausal transition. J Clin Sleep Med, 2018; 14(7):1127–1133; doi:10.5664/jcsm.7208

20.

Ensrud

, Stone

, Blackwell

, et al. Frequency and severity of hot flashes and sleep disturbance in postmenopausal women with hot flashes. Menopause, 2009; 16(2):286–292; doi:10.1097/gme.0b013e31818c0485

21.

, Thurston

, Matthews

, et al. Are hot flashes associated with sleep disturbance during midlife? Results from the STRIDE cohort study. Maturitas, 2012; 71(1):34–38; doi:10.1016/j.maturitas.2011.10.003

22.

de Zambotti

, Colrain

, Javitz

, et al. Magnitude of the impact of hot flashes on sleep in perimenopausal women. Fertil Steril, 2014; 102(6):1708–1715 e1; doi: 10.1016/j.fertnstert.2014.08.016

23.

Hung

, Lu

, Ou

, et al. Menopause is associated with self-reported poor sleep quality in women without vasomotor symptoms. Menopause, 2014; 21(8):834–839; doi:10.1097/GME.0000000000000183

24.

Freeman

, Sammel

, Lin

. Temporal associations of hot flashes and depression in the transition to menopause. Menopause, 2009; 16(4):728–734; doi:10.1097/gme.0b013e3181967e16

25.

Freeman

. Associations of depression with the transition to menopause. Menopause, 2010; 17(4):823–827; doi:10.1097/gme.0b013e3181db9f8b

26.

de Kruif

, Spijker

, Molendijk

. Depression during the perimenopause: A meta-analysis. J Affect Disord, 2016; 206:174–180; doi: 10.1016/j.jad.2016.07.040

27.

Lampio

, Polo-Kantola

, Polo

, et al. Sleep in midlife women: Effects of menopause, vasomotor symptoms, and depressive symptoms. Menopause, 2014; 21(11):1217–1224; doi:10.1097/GME.0000000000000239

28.

Vousoura

, Spyropoulou

, Koundi

, et al. Vasomotor and depression symptoms may be associated with different sleep disturbance patterns in postmenopausal women. Menopause, 2015; 22(10):1053–1057; doi:10.1097/GME.0000000000000442

29.

Lampio

, Saaresranta

, Engblom

, et al. Predictors of sleep disturbance in menopausal transition. Maturitas, 2016; 94:137–142; doi: 10.1016/j.maturitas.2016.10.004

30.

Cray

, Woods

, Herting

, et al. Symptom clusters during the late reproductive stage through the early postmenopause: Observations from the Seattle Midlife Women's Health Study. Menopause, 2012; 19(8):864–869; doi:10.1097/gme.0b013e31824790a6

31.

Cray

, Woods

, Mitchell

. Identifying symptom clusters during the menopausal transition: Observations from the Seattle Midlife Women's Health Study. Climacteric, 2013; 16(5):539–549; doi:10.3109/13697137.2012.746657

32.

Woods

, Cray

, Mitchell

, et al. Endocrine biomarkers and symptom clusters during the menopausal transition and early postmenopause: Observations from the Seattle Midlife Women's Health Study. Menopause, 2014; 21(6):646–652; doi:10.1097/GME.0000000000000122

33.

Harlow

, Karvonen-Gutierrez

, Elliott

, et al. It is not just menopause: Symptom clustering in the Study of Women's Health Across the Nation. Womens Midlife Health, 2017; 3; doi: 10.1186/s40695-017-0021-y

34.

Woods

, Hohensee

, Carpenter

, et al. Symptom clusters among MsFLASH clinical trial participants. Menopause, 2016; 23(2):158–165; doi:10.1097/GME.0000000000000516

35.

Min

, Docherty

, Im

, et al. Identification of symptom clusters among midlife menopausal women with metabolic syndrome. West J Nurs Res, 2021;1939459211018824; doi: 10.1177/01939459211018824

36.

Schilling

, Gallicchio

, Miller

, et al. Current alcohol use is associated with a reduced risk of hot flashes in midlife women. Alcohol Alcohol, 2005; 40(6):563–568; doi:10.1093/alcalc/agh191

37.

Rossmanith

. Ultradian and circadian patterns in luteinizing hormone secretion during reproductive life in women. Hum Reprod, 1993; 8 Suppl 2(2):77–83; doi:10.1093/humrep/8.suppl_2.77

38.

Parikh

, Stolze

, Ozarda

, et al. Diurnal variation of steroid hormones and their reference intervals using mass spectrometric analysis. Endocr Connect, 2018; 7(12):1354–1361; doi:10.1530/EC-18-0417

39.

Bao

, Liu

, van Someren

, et al. Diurnal rhythm of free estradiol during the menstrual cycle. Eur J Endocrinol, 2003; 148(2):227–232.

40.

Vermeulen

. The hormonal activity of the postmenopausal ovary. J Clin Endocrinol Metab, 1976; 42(2):247–253; doi:10.1210/jcem-42-2-247

41.

Gallicchio

, Miller

, Zacur

, et al. Race and health-related quality of life in midlife women in Baltimore, Maryland. Maturitas, 2009; 63(1):67–72; doi:10.1016/j.maturitas.2009.02.001

42.

Cantril

The Pattern of Human Concerns. Rutgers University Press: New Brunswick, NJ, USA; 1965.

43.

Smith

, Gallicchio

, Flaws

. Understanding the complex relationships underlying hot flashes: A Bayesian network approach. Menopause, 2018; 25(2):182–190; doi:10.1097/GME.0000000000000959

44.

Scutari

. Bayesian network constraint-based structure learning algorithms: Parallel and optimized implementations in the bnlearn R package. J Stat Softw, 2017; 77(2):1–20; doi:10.18637/jss.v077.i02

45.

R Development Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna, Austria; 2013.

46.

Hatcher

, Smith

, Chiang

, et al. Association of phthalate exposure and endogenous hormones with self-reported sleep disruptions: Results from the Midlife Women's Health Study. Menopause, 2020; 27(11):1251–1264; doi:10.1097/GME.0000000000001614

47.

Avis

, Crawford

, Stellato

, et al. Longitudinal study of hormone levels and depression among women transitioning through menopause. Climacteric, 2009; 4(3):243–249; doi:10.1080/cmt.4.3.243.249

48.

Chiang

, Gallicchio

, Zacur

, et al. Hormone variability and hot flash experience: Results from the midlife women's health study. Maturitas, 2019; 119:1–7; doi: 10.1016/j.maturitas.2018.10.007

49.

Avis

, Stellato

, Crawford

, et al. Is there a menopausal syndrome? Menopausal status and symptoms across racial/ethnic groups. Soc Sci Med, 2001; 52:345–356.

50.

Berecki-Gisolf

, Begum

, Dobson

. Symptoms reported by women in midlife: Menopausal transition or aging?. Menopause, 2009; 16(5):1021–1029; doi:10.1097/gme.0b013e3181a8c49f

51.

Baker

, Lee

. Menstrual cycle effects on sleep. Sleep Med Clin, 2018; 13(3):283–294; doi:10.1016/j.jsmc.2018.04.002