Association of phthalate exposure and endogenous hormones with self-reported sleep disruptions: results from the Midlife Women’s Health Study

Abstract

Objective:

Follicle-stimulating hormone and estradiol (E₂) have been associated with sleep in midlife women, however, few studies have examined the association of other hormones or environmental chemical exposure such as phthalates, with self-reported sleep quality. We assessed the relationship of self-reported sleep with hormones and phthalates.

Methods:

In total, 762 women (aged 45–54 y, 459 premenopausal, and 303 perimenopausal) from the Midlife Women’s Health Study answered self-reported questions regarding the frequency of sleep disturbances, insomnia, and restless sleep. Serum E₂, progesterone, testosterone, serum hormone binding globulin, free E₂ index, free testosterone index, E₂:progesterone, and E₂:testosterone were measured. Summary measures of phthalate mixtures, including the phthalates from plastic sources (sumPLASTIC), personal care products (sumPCP), di-(2-ethyhexyl) phthalate (sumDEHP), anti-androgenic phthalates (sumAA), and all phthalate metabolites measured (sumALL), were calculated from urinary phthalate metabolites. Ordinal logistic regression was used to fit each outcome sleep measure with all hormones and summary phthalates.

Results:

Progesterone and testosterone were significantly negatively associated with the frequency of sleep disturbances and insomnia. Free testosterone index was also negatively associated with insomnia frequency. E₂:progesterone was positively associated with frequency of sleep disturbances and restless sleep in self-reported nonsmokers. SumPCP and sumALL were significantly negatively associated with frequency of sleep disturbances, insomnia, and restless sleep. SumDEHP and sumPLASTIC were negatively associated with insomnia frequency. Further, the direction of association between phthalates and sleep appears to be dependent on the quartile of phthalate exposure. All significant associations between phthalates and sleep were in self-reported nonsmokers or former smokers.

Conclusions:

Our study supports previous literature that hormones beyond follicle-stimulating hormone and E₂ are associated with sleep disruptions in menopause. Further, we are among the first to show that phthalate exposure is associated with sleep disruptions in midlife women.

Up to 60% of women in menopause transition experience sleep difficulties.¹ Further, 60%−80% of midlife women experience hot flashes and the risk of depression doubles during the transition.¹ These symptoms substantially impair a woman’s quality of life (QOL).² Furthermore, women with trouble falling asleep are at greater risk of developing persistent depression.³ When compared to midlife women without depression, midlife women with depression have worse health outcomes, require more medical care, and have increased absenteeism.⁴ Moreover, midlife women with more depressive or insomnia symptoms report reduced productivity at work and an increase in indirect and direct healthcare costs⁴ compared to women without depression or insomnia.²

The changing hormonal milieu is likely one mechanism underlying the adverse menopausal symptoms.⁵ As the ovary ages, the ovarian follicular pool decreases. This results in altered secretion of gonadal steroids, including estradiol (E₂) and progesterone.^5,6 Further, reduced inhibitory feedback of E₂ on the pituitary leads to rising concentrations of follicle stimulating hormone (FSH).^5,7 Collectively, these age-related changes result in reduced circulating ovarian hormones and higher FSH in the late menopause transition.

In midlife women, reproductive hormones associated with self-reported sleep quality. It is generally observed that as FSH increases and E₂ decreases, self-reported sleep quality gets worse.^8–11 Importantly, these findings follow the pattern that as women progress through menopause, FSH increases and E₂ decreases. An imbalance of hormones may also contribute to the increased prevalence of poor sleep quality. In support of this, one study found that there was a decrease in the time it takes to wake up after sleep onset as the amount of testosterone relative to E₂ increases.¹⁰ Another study found that changes in the E₂:progesterone and E₂ + estrone:progesterone ratios were associated with increased insomnia frequency among midlife women.¹²

Many studies have shown a significant association between estrogens and poor sleep;^8,9,13 however, this is not a consistent finding.^14,15 Additionally, several studies found that rising FSH concentrations are associated with poorer sleep quality in midlife women.^8,10,11 Alternatively, other studies have not found any associations between FSH and sleep in populations of menopausal women.^15–17 Notable differences in study design likely explain some of the discrepancies observed. For example, some studies sample across all menopause stages, whereas others are only able to work with women from one stage of menopause. Further, some use objective measures of sleep quality (ie polysomnography or actigraphy),^10,11,18,19 whereas others select to use self-reported measures.^8,9,13,16,17

Exposure to endocrine disrupting chemicals (EDCs) is one largely unexplored variable that may help explain the increased prevalence of sleep difficulties in midlife women. Phthalates are a major class of environmentally pervasive endocrine disruptors to which humans and animals are exposed daily. Phthalates are industrial plasticizers and chemical stabilizers commonly found in polyvinyl chloride plastic products and consumer products, including food packaging, clothes, cosmetics, medical bags and tubing, and children’s toys.²⁰ Phthalates such as dibutyl phthalate or diethyl phthalate are used in personal care products as chemical stabilizers for fragrances.^21,22 Phthalate plasticizers are noncovalently bound to plastic, and therefore can leach into the environment.²⁰ Normal human exposure can range from 3 to 30 μg/kg/day, with occupational exposure occurring around 0.21 to 2.1 mg/kg/d, and medical exposure estimated to be around 8.5 mg/kg/d.²³ Phthalate metabolites are detectable in women’s serum and ovarian follicular fluid.^24–26 Also, women have higher serum phthalate metabolite concentrations compared to men. This indicates there may be a potential sex difference in exposure and impact of phthalates on human health.²⁷

One study has investigated the association between phthalate exposure and hot flashes. This study found that increased exposure to phthalates from personal care products significantly increased the odds of ever having hot flashes, having a hot flash within the past 30 days, and having daily hot flashes.²⁸ Importantly, phthalates have been associated with sleep quality in other adult populations. Specifically, research from the National Health and Nutrition Examination Survey (NHANES) shows that increased urinary concentrations of the metabolite monocyclohexyl phthalate is associated with increased odds of waking up at night.²⁹ Furthermore, phthalates exposure is associated with depression in older adult populations of men and women.^30,31 Therefore, it is possible for phthalates to either be directly or indirectly associated with sleep in midlife women.

Phthalates are known to modulate the hormones associated with sleep and depression. In animal studies, increased prenatal exposure to di-(2-ethylhexyl) phthalate (DEHP) increased serum E₂ concentrations and decreased serum testosterone, FSH, and inhibin B concentrations in aged female mice.³² Further, acute adult exposure to DEHP reduced E₂, as well as increased progesterone and FSH in female mice.³³ Furthermore, acute adult exposure to diisononyl phthalate reduced E₂ and testosterone in female mice. Multiple phthalate metabolites have been associated with serum hormone concentrations, including estriol (a metabolite of E₂), progesterone:estriol ratio, testosterone, E₂, progesterone, and anti-Müllerian hormone in adult women.^25,26 Although the associations between phthalate exposure and hormones in midlife women have not been studied, it is possible that phthalate exposure contributes to altered hormone profiles in this population.

Identifying risk factors underlying disrupted sleep in midlife women is essential to identify potential targets for behavioral or therapeutic intervention. Therefore, the objectives of this study were to 1) describe the associations between serum hormone concentrations or summary measures of phthalate exposure and frequency of sleep disruptions and 2) identify whether hormones mediate the association between phthalate exposure and frequency of sleep disruptions among participants of the Midlife Women’s Health Study. Further, we aimed to understand the influence of self-reported smoking history on these associations. We hypothesized that increased phthalate exposure would be associated with increased frequency of sleep disruptions. We further hypothesized that serum reproductive hormones, particularly E₂ or testosterone, would mediate these associations.

METHODS

Study design and participants

All participants gave written informed consent according to the procedures provided by the University of Illinois and Johns Hopkins University Institution Review Boards, both of which approved this research. The Midlife Women’s Health Study is a longitudinal cohort study of the risk factors for hot flashes among midlife women (45–54 y) that began in 2006. This study enrolled residents of the Baltimore metropolitan region, including Baltimore and its surrounding counties. The parent study design is described in detail elsewhere.³⁴ Briefly, women within the selected age range (45–54 y) were recruited through mass mail. Women who were interested in participating were screened via telephone and an initial appointment was scheduled for a clinic visit if eligibility criteria were met. Women were eligible if they were aged between 45 and 54 years, had an intact uterus and ovaries, had at least three menstrual periods in the past 12 months, were not on hormone therapy, were not pregnant, and did not have a history of ovarian, endometrial, or breast cancer.

Women who were eligible and interested in participating in the study were asked to make an initial clinic visit to a Johns Hopkins clinical site. During this first clinic visit, the participant completed a 26-page study survey that obtained details on demographics, reproductive history, hormonal contraceptive use, physical symptoms, hormone therapy use, medical history, sleep characteristics, and health behaviors (smoking, alcohol use, vitamin use, eating habits). During the visit, each woman was given the survey in a quiet, private area and was instructed to complete the entire survey. Clinic staff were available if participants had questions regarding its completion. Women also submitted blood and urine samples, and had height, weight, and blood pressure measured at this clinic visit. All clinic visits were scheduled in the morning to account for diurnal fluctuations in hormones.^35–38 After the first visit, participants visited the clinic weekly for the following 3 weeks. Each of these weeks, the women provided an additional serum and urine samples. Blood and serum samples were stored at −80°C for 1–5 years until measurement of hormone concentrations and phthalate metabolites as described below.

Measures of self-reported sleep and depressive symptoms

In the study survey, women were asked to self-report their subjective frequency of different types of sleep disruptions via three questions in the questionnaire as previously described.¹³ Importantly, the present study is a secondary analysis of an existing dataset; the parent study aimed to identify risk factors for hot flashes among midlife women. Therefore, we were limited to these three variables for quantifying self-reported frequency of disrupted sleep.

To quantify the frequency of sleep disturbances, we asked women, “Please indicate how frequently you experienced sleep disturbances during the past year.” To quantify the frequency of insomnia, we asked women, “Please indicate how frequently you experienced insomnia during the past year.” Women answered both of these questions 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). Importantly, these questions were similar to those from the Midlife in the United States study and have been validated.^13,39 Additionally, we quantified the frequency of restless sleep by asking women to answer, “During the past week my sleep was restless,” on a four-point Likert scale (rarely, some of the time, moderately, or most of the time). Importantly, self-reported sleep is not equivalent to objective measures of (ie polysomnography or actigraphy).^40–42 However, self-reported sleep measures reflect how women feel about their sleep quality. Therefore, they are still an accurate measure for QOL.^8,40

Depressive symptoms were assessed using the Center for Epidemiologic Studies-Depression Scale (CES-D). The CES-D is a 20-item Likert scale questionnaire that asks participants to rate how often they have specific feelings within the last week.^43–45 The summary score of these questions indicates potential presence of depression. A score at or above 16 indicates clinically relevant depressive symptoms.

Covariate data collection

Data on smoking, hot flashes, QOL, and menopause status were collected using the study survey from the first visit. Similar to previous studies in midlife women,^15,46,47 self-reported smoking history (never smoked, former smoker, current smoker) was used to determine smoking status.

Women were asked a detailed history of their hot flashes, including whether women had weekly hot flashes, severity of hot flashes, and hot flashes at night. Hot flashes were defined in the survey 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.” For our analysis, we determined whether women experienced hot flashes at night using the following question: “On average, how many hot flashes do you experience every night (between 9:00 PM and 6:00 AM)?” with “none,” “1,” “2,” “3,” “4 or more,” or “I don’t know.” Women who answered “none” were categorized as “no”, whereas women who answered “1,” “2,” “3,” or “4 or more” were categorized as “yes.”

QOL was assessed with Cantril’s Self-Anchoring Ladder of Life.^48,49 Specifically, they were given the description, “Here is a ladder representing the ‘Ladder of Life.’ The top of the ladder represents the best possible life for you. The bottom of the ladder represents the worst possible life for you. Answer questions a through c.” After this description were three questions, reading: “a. On which step of the ladder do you feel you personally stand at the present time.”; “b. On which step would you have stood five years ago?”; and “c. Thinking about your future, on which step do you think you will stand about five years from now?”. Participants indicate a number from 0 (worst possible life) to 10 (best possible life) for each of the three questions. Because the results of QOL were skewed, responses of 1–6 were categorized as ‘low’ QOL, 7–8 categorized as ‘moderate’ QOL, and 9–10 were categorized as ‘high’ QOL.⁴⁸

Menopause status was defined as follows: premenopausal women were those who had their last menstrual period within the past 3 months and experienced 11 or more periods within the past year; perimenopausal women were those who reported (1) their last menstrual period within the past year, but not within the past 3 months or (2) their last menstrual period within the past 3 months and experienced 10 or fewer periods within the past year; and postmenopausal women were those who reported not having a menstrual period within the past year. Women who were postmenopausal at the first clinic visit were removed from the cohort and were not included in the study. Only data from premenopausal and perimenopausal women were used in the present study.

Measurement of serum hormones

Participant sera samples were collected at each of four consecutive weekly visits during the first year of the study. For each weekly sample, E₂, progesterone, testosterone, and sex hormone binding globulin (SHBG) were measured with enzyme-linked immunosorbent assays (DRG International, Springfield, NJ). The values of SHBG were used to calculate the free versus bound serum hormone concentrations (E₂ and testosterone), as the nonbound hormones are bioactive. The minimum detection limits and intra-assay variation (CV) were as follows: E₂ 7 pg/mL, 3.3 ± 0.17% CV; progesterone 0.1 ng/mL, 2.1 ± 0.65% CV; testosterone 0.04 ng/mL, 2.2 ± 0.56% CV; and SHBG 0.1 nmol/L, 2.4 ± 0.67% CV. For values lower than the limit of detection for the assay, we used the limit of detection as the hormone value. Each sample was run in duplicate within the assay.

The free-E₂ index was calculated with the following formula: 100*([E₂*0.003671]/SHBG). The free-testosterone index was calculated using the following formula: 100*([Testosterone*3.467]/SHBG). These indices indicate how much free E₂ and free testosterone is available in the sample, as previously described.¹² Additionally, the ratio of E₂ to progesterone (E₂:progesterone) and E₂ to testosterone (E₂:testosterone) were determined by dividing E₂ by progesterone or testosterone, respectively. For all women’s hormone measures, we calculated the geometric mean of the four samples to account for individual variability (thus establishing normality).

Measurement of urinary phthalate metabolites and summary phthalate measures

Women also provided urine samples once per week for four consecutive weeks. These four urine samples were pooled for analysis. Urine samples were analyzed with high-performance liquid chromatography negative-ion electrospray ionization-tandem mass spectrometry (HPLC-MS/MS) to analyze target phthalate metabolites. Specifically, the following individual phthalate metabolites were measured: monomethyl phthalate, mono-(3-carboxypropyl) phthalate (MCPP), monoethyl phthalate (MEP), monobutyl phthalate (MBP), monobenzyl phthalate (MBzP), monoisobutyl phthalate (MiBP), mono-(5-carboxy-2-ethylpentyl) phthalate (MCEPP), mono-(2-ethyl-5-oxohexyl) phthalate (MEOHP), mono-(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), and monoethylhexyl phthalate (MEHP). To account for any potential hydration differences between participants and the volume of the donated urine sample, all values (ng/mL) were normalized to the specific gravity value of the sample as described in previous studies.²⁸ The metabolites were selected as they are the major urinary metabolites of common phthalate parent compounds, including diethyl phthalate, DEHP, dibutyl phthalate, diisobutyl phthalate, dioctyl phthalate , and benzyl butyl phthalate.^28,31,50,51 Additionally, these metabolites or their parent compounds have been associated with altered reproductive outcomes or behaviors in animal models and epidemiological studies.^{28,32,33,52–56} For values lower than the limit of detection, we used a value half of the limit of detection.⁵⁷

Humans are exposed to mixtures of phthalates that contain different metabolites and parent compounds with varying toxicities. Therefore, we quantified five summary measures that estimate exposure to relevant phthalate mixtures, as previously described.²⁸ First, we calculated the estimated exposure to urinary phthalate metabolites from parent compounds present in personal care products (PCP) by summing MBP (MW 222 μg/mmol) and MEP (MW 194 μg/mmol). The sumPCP estimates exposure to phthalates from PCPs, including shampoos, conditioners, perfumes, and nail polishes.^21,58 We calculated the estimated exposure to DEHP by summing the most common urinary metabolites, including mono-(5-carboxy-2-ethylpentyl) phthalate (MW 308 μg/mmol), MEOHP (MW 292 μg/mmol), MEHHP (MW 294 μg/mmol), and MEHP (MW 278 μg/mmol). DEHP is one of the most commonly used phthalate plasticizers in polyvinyl chloride plastics that are found in consumer products, building materials, and medical equipment and tubing.^58,59 We calculated the sum of urinary phthalate metabolites from parent compounds with known anti-androgenic (sumAA) biological activity in the body,^60,61 including MBP, MBzP (MW 256 μg/mmol), and MiBP (MW 222 μg/mmol). Then, we calculated the estimated exposure to phthalates from plastic sources (sumPLASTIC) by summing sumDEHP with MCPP (MW 252 μg/mmol) and MBzP (MW 256 μg/mmol).^23,58,62,63 Finally, we created a summary measure of all phthalate metabolites (sumALL) by summing all metabolites measured in this study (MEHHP, MEHP, MEOHP, MECPP, MCPP, MBzP, MEP, MBP, and MiBP).

Statistical analysis

The relationship between the geometric mean of each hormone measure (E₂, progesterone, testosterone, SHBG, free E₂ index, free testosterone index, E₂:progesterone, and E₂:testosterone) and each self-reported sleep measure was assessed with ordinal logistic regression (frequency, described above). The number of women missing hormone samples were as follows: E₂ (n = 1), testosterone (n = 1), progesterone (n = 1), and SHBG (n = 6). Geometric mean was used as opposed to arithmetic mean as it controls for individual variability across multiple samples.

Similarly, the relationship between the mean of each summary phthalate measure (sumPCP, sumAA, sumDEHP, sumPLASTIC, and sumALL) and each self-reported sleep measure was assessed with ordinal logistic regression. The number of women missing summary phthalate measures were as follows: sumPCP (n = 33), sumDEHP (n = 48), sumAA (n = 75), sumPLASTIC (n = 75), and sumALL (n = 90).

For the ordinal logistic regression, we fit unadjusted models and adjusted models. We selected covariates based on previous work that estimated the risk factors for impaired sleep in midlife women. Specifically, we adjusted for menopause status (pre- and perimenopausal), Body Mass Index (BMI), self-reported hot flashes at night, CES-D score, self-reported smoking status, and present QOL. Women who were identified as postmenopausal at the first clinic visit were excluded from the analysis. Importantly, other studies have also found an association between sleep quality and depression,^9,15,16 hot flashes at night,^64,65 BMI,¹⁵ menopause status,^8,9,17,66 QOL or perceived health,^9,15 and smoking.⁹ Hence the selection of these variables as covariates. Age was not entered into the logistic regression models, as it was colinear with menopause status in our population.

Self-reported smoking status is significantly associated with sleep,^13,67 phthalates,⁶⁸ and hormones.^69–71 Therefore, we also planned to stratify our adjusted models by self-reported smoking status a priori. In these analyses, we adjusted for all of the same covariates described above, excluding smoking status.

A planned mediation analysis was performed to determine if hormones mediated the association between summary phthalate exposure and frequency of sleep disruptions. We a priori elected to focus on hormones and summary phthalate measures that were significantly associated with our sleep measures. In the analyses described above, we observed significant associations between progesterone and sleep disruptions, as well as sumPCP and sleep disruptions. Therefore, we investigated whether progesterone mediated the association between sumPCP and self-reported frequency of sleep disturbances, insomnia, and restless sleep. This was done by fitting two separate models. The first model was an ordinal logistic regression to estimate the association between sumPCP concentrations and the three sleep measures. This model was adjusted by menopause status (pre- and perimenopausal), BMI, hot flashes at night, CES-D score, and self-reported smoking status. The second model was identical to the first in all aspects except that it also included progesterone score as an additional covariate. We then calculated the difference in the estimated effect of sumPCP between the two models. If progesterone did mediate the association between sumPCP and sleep, we would anticipate a substantial decrease in the second model relative to the first model.⁷²

Many EDCs, including phthalates,^27,32,56,73 have a nonmonotonic dose response as well as low dose effects.⁷⁴ Therefore, we were also interested in identifying any potential impact of lower exposure of phthalates in our population. We calculated the quartiles of each summary phthalate and grouped women into either Quartile 1 (Q1) or Quartiles 2–4 (Q2–4). Q1 represents women with relatively lower exposure, whereas Q2–4 represents women with relatively high exposure. We then performed similar adjusted ordinal logistic regression as described above and stratified by self-reported smoking status (nonsmokers, former smokers, and current smokers).

For all statistical analyses, ordinal logistic regression models were fit with the MASS package⁷⁵ in R version 3.6.0 (Planting of a Tree).⁷⁶ All results are presented as β coefficients and 95% confidence intervals. Importantly, β coefficients represent the predicted likelihood of a change of the outcome (ie frequency of sleep disruptions) when the predictor variable (ie hormones or summary phthalates) changes by 1 unit; CI of β coefficients that do not cross 0 are considered statistically significant.

RESULTS

Study participants

Data from the first year of enrollment were included in this study. Of 775 women, 762 provided samples for both hormone and phthalate measures. Of the 762 women, 459 were premenopausal and 303 were perimenopausal. The population characteristics are shown in Table 1, serum hormone concentrations are shown in Table 2, and urinary phthalate metabolite concentrations and summary phthalate measures are shown in Table 3. Importantly, we compared exposure in our women to women from a national sample (from the NHANES). We found exposure to individual phthalate metabolites was similar between NHANES and our population. This implies that the associations we observed are based on environmentally relevant exposure profiles.

TABLE 1.

Selected characteristics of the study sample (n = 762)

	n	%
Menopause status
Pre-	459	60.2
Peri-	303	39.8
Hot flashes at night
Yes	203	29.1
No	495	70.9
Smoking status
Nonsmokers	419	54.7
Former smokers	269	35.1
Current smoker	78	10.2
	Median (IQR)
BMI	26.55 (23.08, 32.00)
Present QOL	8.00 (7.00, 9.00)
CES-D	7.00 (3.00, 14.00)
	n	%
Sleep disturbances
Never	111	14.6
Rarely (<1/mo)	176	23.2
Sometimes (1–4/mo)	214	28.2
Frequently (2–4/wk)	159	20.9
Regularly (>5/wk)	99	13.0
Insomnia
Never	227	29.9
Rarely (<1/mo)	191	25.1
Sometimes (1–4/mo)	163	21.4
Frequently (2–4/wk)	112	14.7
Regularly (>5/wk)	67	8.8
Restless sleep
Rarely	306	40.2
Some of the time	253	33.2
Moderately	114	15.0
Most of the time	89	11.7

BMI, body mass index; CES-D, Center for Epidemiologic Studies – Depression; IQR, interquartile range; Pre-, premenopausal; Peri-, perimenopausal; QOL, quality of life.

TABLE 2.

Endogenous hormone concentrations in the study sample (n = 762)

Hormone	Mean (IQR)
Estradiol (pg/mL)	56.79 (35.35, 81.79)
Progesterone (ng/mL)	1.59 (0.28, 3.09)
Testosterone (ng/mL)	0.29 (0.19, 0.41)
SHBG (pg/mL)	64.93 (44.64, 93.82)
Free Estradiol Index	0.32 (0.2, 0.52)
Free Testosterone Index	1.53 (0.91, 2.88)
Estradiol:Progesterone	0.08 (0.04, 0.17)
Estradiol:Testosterone	0.16 (0.10, 0.28)

IQR, interquartile range; SHBG, serum hormone binding globulin.

TABLE 3.

Range of phthalate metabolites and summary phthalate measures for study participants in the Midlife Women’s Health Study

Phthalates		MWHS (n = 762)		NHANES 2005-2006 (all females; n = 1,278)	NHANES 2007-2008 (all females; n = 1,310)
Parent	Metabolite	% ≥ LOD	Mediana (95% CI)	Median (95% CI)
DEHP	MEHHP	100	33.3 (30.9-36.2)	21.4 (19.5-23.4)	19.9 (16.8-23.3)
	MEHP	100	4.5 (4.2-5.0)	2.10 (1.90-2.40)	2.00 (1.70-2.40)
	MEOHP	100	11.95 (11.2-12.5)	13.8 (12.5-15.7)	11.7 (9.80-13.5)
	MECPP	100	25.6 (23.9-28.0)	31.7 (28.0-36.6)	31.0 (26.4-38.3)
DnOP	MCPP	100	2.53 (2.3-2.77)	1.80 (1.60-2.10)	2.60 (2.30-3.00)
BBzP	MBzP	100	9.25 (8.5-9.9)	7.70 (6.55-9.36)	7.99 (6.41-9.36)
DEP	MEP	99	96.95 (87.6-107)	109 (86.8-130)	82.4 (67.6-101)
DBP	MBP	100	19.8 (18.5-21.3)	20.4 (17.1-24.7)	20.8 (18.1-23.2)
DiBP	MiBP	100	16.25 (15.3-17.3)	5.50 (4.70-6.60)	7.40 (6.50-8.30)
Summary phthalates b		Median (IQR)
sumPCP		2.05 (1.00, 3.10)
sumDEHP		0.34 (0.24, 0.54)
sumAA		0.69 (0.52, 0.94)
sumPLASTIC		0.43 (0.31, 0.63)
sumALL		2.54 (1.54, 3.73)

AA, antiandrogenic; BBzP, benzyl butyl phthalate; CI, confidence interval; DBP, dibutyl phthalate; DEHP, di-(2-ethylhexyl) phthalate; DEP, diethyl phthalate; DiBP, diisobutyl phthalate; DnOP, dioctyl phthalate; IQR, interquartile range; LOD, level of detection; MBP, monobutyl phthalate; MBzP, monobenzyl phthalate; MCEPP, mono-(5-carboxy-2-ethylpentyl) phthalate; MCPP, mono-(3-carboxypropyl) phthalate; MEHHP, mono-(2-ethyl-5-hydroxyhexyl) phthalate; MEHP, monoethylhexyl phthalate; MEOHP, mono-(2-ethyl-5-oxohexyl) phthalate; MEP, monoethyl phthalate; MiBP, monoisobutyl phthalate; MWHS, Midlife Women’s Health Study; NHANES, National Health and Nutrition Examination Survey; PCP, personal care products.

^ang/mL adjusted for specific gravity.

^bSummary phthalate measures were calculated as follows:.

sumPCP is the sum of phthalate metabolites present in personal care products: MBP + MEP.

sumDEHP is the sum of DEHP metabolites: MECPP + MEHHP + MEHP + MEOHP.

sumAA is the sum of phthalate metabolites with known androgenic activity: sumDEHP + MBP + MBzP + MiBP.

sumPLASTIC is the sum of phthalate metabolites from plastic sources: sumDEHP + MCPP + MBzP.

sumALL is the sum of all calculated phthalate metabolites: sumDEHP + MCPP + MBzP + MEP + MBP + MiBP.

Association of hormone concentrations with self-reported frequency of sleep disturbances, insomnia, and restless sleep

In the unadjusted analysis progesterone and E₂:progesterone were significantly associated with all three self-reported sleep measures (sleep disturbances, insomnia, and restless sleep) (see Supplemental Table 1, http://links.lww.com/MENO/A628, demonstrating the estimated associations between serum hormone concentrations and subjective sleep measures). Lower concentrations of free E₂ (indicated by lower free E₂ index) were associated with increased frequency of sleep disturbances and restless sleep. Lower concentrations of testosterone were associated with increased frequency of sleep disturbances. Finally, lower serum concentrations of progesterone and E₂:progesterone were associated with increased frequency of all three measures of sleep disruptions. E₂, SHBG, free testosterone index, and E₂:testosterone were not associated with any sleep measure.

The adjusted analysis (Table 4) included BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL. There was a similar trend observed in the adjusted model as the unadjusted model. Specifically, lower concentrations of progesterone and testosterone were associated with increased frequency of sleep disturbances and insomnia. The frequency of insomnia increased when free testosterone index decreased. E₂, SHBG, free E₂ index, E₂:progesterone, and E₂:testosterone were not associated with any sleep measure after we adjusted for covariates. Further, no significant associations between any hormone and restless sleep was observed in the adjusted model. Marginal effect sizes for the significant associations from the adjusted model are presented in supplemental content (see Supplemental Table 2, http://links.lww.com/MENO/A628, demonstrating the marginal effect sizes for the significant estimated associations between serum hormone concentrations and self-reported frequency of sleep disruptions).

TABLE 4.

Adjusted model^a of ordinal logistic regression examining associations between serum hormone concentrations and self-reported frequency of sleep disturbances, insomnia, and restless sleep in midlife women

	Sleep disturbances β (95% CI)	Insomnia β (95% CI)	Restless sleep β (95% CI)
Hormones
Estradiol	−0.0022 (−0.0059 to 0.0014)	−0.0023 (−0.0060 to 0.0013)	−0.0029 (−0.0069 to 0.001)
Progesterone	−0.15 (−0.27 to −0.043)	−0.12 (−0.26 to −0.014)	−0.083 (−0.22 to 0.025)
Testosterone	−0.31 (−0.63 to −0.0072)	−0.34 (−0.69 to −0.023)	−0.20 (−0.57 to 0.14)
SHBG	0.00083 (−0.0031 to 0.0048)	−0.00055 (−0.0045 to 0.0034)	−0.00072 (−0.0048 to 0.0034)
Free Estradiol Index	−0.50 (−1.0 to 0.0046)	−0.50 (−1.0 to 0.029)	−0.46 (−1.0 to 0.095)
Free Testosterone Index	−0.031 (−0.073 to 0.011)	−0.050 (−0.11 to −0.0014)	−0.012 (−0.064 to 0.035)
Estradiol:Progesterone	0.55 (−0.13 to 1.20)	0.52 (−0.14 to 1.20)	0.36 (−0.34 to 1.0)
Estradiol:Testosterone	0.15 (−0.65 to 0.97)	0.30 (−0.53 to 1.10)	−0.52 (−1.40 to 0.34)

Values in bold indicate significant associations.

BMI, body mass index; CES-D, Center for Epidemiologic Studies – Depression; CI, Confidence interval; QOL, quality of life; SHBG, serum hormone binding globulin.

^aAdjusted for BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL.

Smoking is associated with both sleep^13,67 and hormones,^69–71 Therefore, we stratified our adjusted analysis by self-reported smoking status (Nonsmoker, n = 419; Former Smoker, n = 269; and Current Smoker, n = 78). We adjusted for BMI, menopause status, hot flashes at night, CES-D score, and present QOL. These results are shown in Table 5.

TABLE 5.

Association of serum reproductive hormone self-reported frequency of sleep disturbances, insomnia, and restless sleep in midlife women, stratified by self-reported smoking status

Hormones	Sleep disturbances b (95% CI)a	Insomnia b (95% CI)a	Restless sleep b (95% CI)a
Estradiol
Nonsmokers	−0.0036 (−0.0085 to 0.0012)	−0.0031 (−0.008 to 0.0017)	−0.0022 (−0.0076 to 0.0030)
Former smokers	−0.0020 (−0.0084 to 0.0039)	−0.0011 (−0.0076 to 0.0048)	−0.0040 (−0.011 to 0.0024)
Current smokers	0.000027 (−0.017 to 0.016)	−0.0082 (−0.025 to 0.0082)	−0.011 (−0.029 to 0.0064)
Progesterone
Nonsmokers	−0.12 (−0.26 to −0.014)	−0.085 (−0.24 to 0.022)	−0.097 (−0.27 to 0.031)
Former smokers	−0.31 (−0.57 to −0.052)	−0.25 (−0.54 to 0.019)	−0.12 (−0.43 to 0.18)
Current smokers	0.045 (−0.54 to 0.57)	−0.19 (−0.93 to 0.40)	0.20 (−0.41 to 0.75)
Testosterone
Nonsmokers	−0.33 (−0.69 to 0.03)	−0.40 (−0.81 to −0.015)	−0.22 (−0.67 to 0.20)
Former smokers	−0.42 (−1.1 to 0.22)	−0.20 (−0.91 to 0.44)	−0.25 (−1.0 to 0.43)
Current smokers	0.16 (−2.1 to 2.3)	−0.71 (−3.4 to 1.6)	−0.47 (−2.9 to 1.8)
SHBG
Nonsmokers	0.0030 (−0.0025 to 0.0086)	0.0017 (−0.0039 to 0.0073)	0.0018 (−0.0039 to 0.0075)
Former smokers	−0.0017 (−0.0078 to 0.0046)	−0.0028 (−0.0089 to 0.0033)	−0.0040 (−0.011 to 0.0025)
Current smokers	−0.0053 (−0.023 to 0.012)	−0.010 (−0.027 to 0.0064)	−0.0092 (−0.028 to 0.0094)
Free Estradiol Index
Nonsmokers	−0.68 (−1.3 to −0.061)	−0.77 (−1.5 to −0.10)	−0.59 (−1.3 to 0.12)
Former smokers	−0.20 (−1.2 to 0.80)	0.14 (−0.88 to 1.1)	0.051 (−1.0 to 1.1)
Current smokers	−0.53 (−2.8 to 1.6)	−0.84 (−3.2 to 1.3)	−1.4 (−4.3 to 0.95)
Free Testosterone Index
Nonsmokers	−0.035 (−0.080 to 0.0099)	−0.064 (−0.13 to −0.0094)	−0.026 (−0.087 to 0.027)
Former smokers	−0.027 (−0.16 to 0.10)	0.041 (−0.091 to 0.17)	0.099 (−0.044 to 0.24)
Current smokers	−0.098 (−0.55 to 0.35)	−0.26 (−0.74 to 0.20)	−0.28 (−0.78 to 0.20)
Estradiol:Progesterone
Nonsmokers	1.2 (0.26 to 2.2)	1.0 (0.061 to 2.0)	0.90 (−0.072 to 1.9)
Former smokers	0.027 (−1.2 to 1.2)	0.40 (−0.77 to 1.6)	−0.013 (−1.4 to 1.3)
Current Smokers	−0.18 (−1.7 to 1.2)	−0.21 (−1.7 to 1.2)	−0.26 (−2.0 to 1.2)
Estradiol:Testosterone
Nonsmokers	−0.34 (−1.6 to 0.92)	0.48 (−0.80 to 1.7)	−0.67 (−2.2 to 0.74)
Former smokers	0.42 (−0.73 to 1.6)	−0.0049 (−1.2 to 1.2)	−0.48 (−1.7 to 1.6)
Current smokers	0.26 (−3.7 to 4.2)	0.28 (−3.5 to 4.1)	−0.57 (−4.7 to 3.5)

Values in bold indicate significant associations.

BMI, body mass index; CES-D, Center for Epidemiologic Studies – Depression; CI, Confidence interval; QOL, quality of life; SHBG, serum hormone binding globulin.

^aAdjusted for BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL.

Lower concentrations of progesterone were associated with increased frequency of sleep disturbances in both nonsmokers and smokers. Lower testosterone concentrations were also associated with increased frequency of insomnia in nonsmokers, but not former or current smokers. Lower concentrations of free E₂ index were associated with increased frequency of sleep disturbances and insomnia in nonsmokers, but not former or current smokers. Higher E₂:progesterone ratio was associated with increased frequency of sleep disturbances, insomnia, and restless sleep in nonsmokers, but not former or current smokers. No significant associations between any hormone measures and self-reported sleep measures were identified in current smokers. Marginal effect sizes for the significant associations from the stratified model are presented in supplemental content (see Supplemental Table 3, http://links.lww.com/MENO/A628, demonstrating the marginal effect sizes for the significant estimated associations between serum hormone concentrations and self-reported frequency of sleep disruptions, stratified by smoking status).

Association of summary phthalates with self-reported frequency of sleep disturbances, insomnia, and restless sleep

In the unadjusted analysis (see Supplemental Table 4, http://links.lww.com/MENO/A628, demonstrating the estimated associations between summary phthalates and subjective sleep measures), there were no significant associations between any of the summary phthalate measures (sumPCP, sumDEHP, sumAA, sumPLASTIC, and sumALL) and any self-reported sleep measures. The adjusted analysis (Table 6) included BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL as covariates. There were no significant associations between any summary phthalate measures and any self-reported sleep measures in the adjusted model.

TABLE 6.

Adjusted model of ordinal logistic regression analysis examining associations between sum phthalate measures and self-reported frequency of sleep disturbances, insomnia, and restless sleep in midlife women

	Sleep disturbances β (95% CI)a	Insomnia β (95% CI)a	Restless sleep β (95% CI)a
Phthalate measure
sumPCP	−0.058 (−0.15 to 0.036)	−0.040 (−0.14 to 0.055)	−0.051 (−0.15 to 0.047)
sumDEHP	0.13 (−0.35 to 0.60)	0.027 (−0.45 to 0.50)	0.11 (−0.38 to 0.59)
sumAA	0.068 (−0.32 to 0.45)	−0.097 (−0.49 to 0.29)	0.18 (−0.22 to 0.58)
sumPLASTIC	0.14 (−0.31 to 0.59)	0.014 (−0.44 to 0.46)	0.14 (−0.33 to 0.60)
sumALL	−0.050 (−0.15 to 0.046)	−0.027 (−0.12 to 0.069)	−0.025 (−0.13 to 0.074)

sumPCP is the sum of phthalate metabolites present in personal care products: MBP + MEP.

sumDEHP is the sum of DEHP metabolites: MECPP + MEHHP + MEHP + MEOHP.

sumAA is the sum of phthalate metabolites with known androgenic activity: sumDEHP + MBP + MBzP + MiBP.

sumPLASTIC is the sum of phthalate metabolites from plastic sources: sumDEHP + MCPP + MBzP.

sumALL is the sum of all calculated phthalate metabolites: sumDEHP + MCPP + MBzP + MEP + MBP + MiBP.

AA, antiandrogenic; BMI, body mass index; CES-D, Center for Epidemiologic Studies – Depression; CI, Confidence interval; DEHP, di-(2-ethylhexyl) phthalate; MBP, monobutyl phthalate; MBzP, monobenzyl phthalate; MCEPP, mono-(5-carboxy-2-ethylpentyl) phthalate; MCPP, mono-(3-carboxypropyl) phthalate; MEHHP, mono-(2-ethyl-5-hydroxyhexyl) phthalate; MEHP, monoethylhexyl phthalate; MEOHP, mono-(2-ethyl-5-oxohexyl) phthalate; MEP, monoethyl phthalate; MiBP, monoisobutyl phthalate; PCP, personal care products; QOL, quality of life.

^aAdjusted for BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL.

As smoking is also associated with phthalates,⁶⁸ we chose to stratify our adjusted model by self-reported smoking status. These results are shown in Table 7. Interestingly, we only observed significant associations between summary phthalate measures and self-reported frequency of sleep disturbances, insomnia, and restless sleep in former smokers. Specifically, we found that lower values of sumPCP and sumALL were associated with increased frequency of all three self-reported sleep measures in former smokers. Additionally, we found that lower values for sumDEHP and sumPLASTIC were associated with increased self-reported insomnia in former smokers, but not nonsmokers or current smokers. There was no significant association of sumAA measures with any sleep measure in this model. Marginal effect sizes for the significant associations from the adjusted model are presented in supplemental content (see Supplemental Table 5, http://links.lww.com/MENO/A628, demonstrating the marginal effect sizes for the significant estimated associations between summary phthalate measures and self-reported frequency of sleep disruptions, stratified by smoking status).

TABLE 7.

Association of sum phthalate measures and self-reported frequency of sleep disturbances, insomnia, and restless sleep in midlife women, stratified by self-reported smoking status

	Sleep disturbances β (95% CI)a	Insomnia β (95% CI)a	Restless sleep β (95% CI)a
sumPCP
Nonsmokers	0.021 (−0.10 to 0.15)	0.077 (−0.046 to 0.20)	0.050 (−0.080 to 0.18)
Former smokers	−0.20 (−0.37 to −0.034)	−0.23 (−0.41 to −0.063)	−0.26 (−0.45 to −0.082)
Current smokers	0.040 (−0.27 to 0.36)	−0.13 (−0.45 to 0.19)	0.011 (−0.32 to 0.33)
sumDEHP
Nonsmokers	0.51 (−0.093 to 1.1)	0.43 (−0.18 to 1.0)	0.48 (−0.15 to 1.1)
Former smokers	−0.73 (−1.5 to 0.086)	−0.88 (−1.7 to −0.049)	−0.77 (−1.7 to 0.094)
Current smokers	0.22 (−2.0 to 2.4)	0.76 (−1.5 to 3.0)	0.73 (−1.9 to 3.3)
sumAA
Nonsmokers	0.23 (−0.26 to 0.72)	0.14 (−0.37 to 0.64)	0.40 (−0.12 to 0.92)
Former smokers	−0.37 −1 to 0.31)	−0.68 (−1.4 to 0.0048)	−0.38 (−1.1 to 0.32)
Current smokers	0.55 (−1.0 to 2.1)	0.45 (−1.1 to 2.0)	1.0 (−0.80 to 2.8)
sumPLASTIC
Nonsmokers	0.47 (−0.099 to 1.0)	0.35 (−0.22 to 0.93)	0.47 (−0.13 to 1.1)
Former smokers	−0.62 (−1.4 to 0.17)	−0.83 (−1.7 to −0.025)	−0.69 (−1.6 to 0.14)
Current smokers	0.21 (−1.9 to 2.2)	0.66 (−1.4 to 2.7)	0.56 (−1.9 to 2.9)
sumALL
Nonsmokers	0.035 (−0.09 to 0.16)	0.096 (−0.026 to 0.22)	0.070 (−0.060 to 0.20)
Former smokers	−0.22 (−0.39 to −0.042)	−0.25 (−0.44 to −0.077)	−0.25 (−0.45 to −0.069)
Current smokers	0.0028 (−0.32 to 0.33)	−0.14 (−0.48 to 0.18)	0.041 (−0.31 to 0.38)

Values in bold indicate statistically significant associations.

sumPCP is the sum of phthalate metabolites present in personal care products: MBP + MEP.

sumDEHP is the sum of DEHP metabolites: MECPP + MEHHP + MEHP + MEOHP.

sumAA is the sum of phthalate metabolites with known androgenic activity: sumDEHP + MBP + MBzP + MiBP.

sumPLASTIC is the sum of phthalate metabolites from plastic sources: sumDEHP + MCPP + MBzP.

sumALL is the sum of all calculated phthalate metabolites: sumDEHP + MCPP + MBzP + MEP + MBP + MiBP.

AA, antiandrogenic; BMI, body mass index; CES-D, Center for Epidemiologic Studies – Depression; CI, Confidence interval; DEHP, di-(2-ethylhexyl) phthalate; MBP, monobutyl phthalate; MBzP, monobenzyl phthalate; MCEPP, mono-(5-carboxy-2-ethylpentyl) phthalate; MCPP, mono-(3-carboxypropyl) phthalate; MEHHP, mono-(2-ethyl-5-hydroxyhexyl) phthalate; MEHP, monoethylhexyl phthalate; MEOHP, mono-(2-ethyl-5-oxohexyl) phthalate; MEP, monoethyl phthalate; MiBP, monoisobutyl phthalate; PCP, personal care products; QOL, quality of life.

^aAdjusted for BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL.

Mediation analysis

We found significant associations between progesterone and all sleep measures, as well as sumPCP and all sleep measures in former smokers. Thus, we assessed whether progesterone mediated the association between sumPCP and self-reported frequency of sleep disturbances, insomnia, and restless sleep. We completed this analysis in our stratified adjusted models as described above. The results from this analysis are shown in Table 8. In this mediation analysis, progesterone was included in the covariates when we performed the adjusted analysis (“mediated” model). To show a significant mediation of progesterone, we looked for a substantial decrease in these associations. Specifically, we wanted to identify if the mediated β estimate approached 0. However, this was not observed in any level of our stratification. Therefore, we concluded that progesterone did not mediate the relationships between sumPCP and any sleep measure.

TABLE 8.

Association of sumPCP and self-reported frequency of sleep disturbances, insomnia, and restless sleep in midlife women and mediation by progesterone, stratified by smoking status

Nonsmokers	Sleep disturbances β (95% CI)a	Insomnia β (95% CI)a	Restless sleep β (95% CI)a
Unmediated	0.021 (−0.10 to 0.15)	0.077 (−0.046 to 0.20)	0.050 (−0.080 to 0.18)
Mediated	0.011 (−0.11 to 0.14)	0.065 (−0.059 to 0.19)	0.042 (−0.089 to 0.17)
Difference	−0.01	−0.012	−0.008
Former smokers
Unmediated	−0.20 (−0.37 to −0.034)	−0.23 (−0.041 to −0.26)	−0.26 (−0.045 to −0.082)
Mediated	−0.19 (−0.036 to −0.024)	−0.22 (−0.40 to −0.053)	−0.2 (−0.45 to −0.079)
Difference	0.010	0.010	0
Current smokers
Unmediated	0.040 (−0.27 to 0.36)	−0.13 (−0.45 to 0.19)	0.011 (−0.32 to 0.33)
Mediated	0.078 (−0.23 to 0.40)	−0.13 (−0.45 to 0.19)	0.027 (−0.31 to 0.35)
Difference	0.038	0	0.016

sumPCP is the sum of phthalate metabolites present in personal care products: MBP + MEP.

Difference is the difference between the unmediated and mediated β (95% CI) for each level of stratification. To identify a mediation by progesterone, specifically we are looking for a substantial decrease in these associations (as in the mediated β estimate would approach 0).

BMI, body mass index; CES-D, Center for Epidemiologic Studies – Depression; CI, Confidence interval; MBP, monobutyl phthalate; MEP, monoethyl phthalate; PCP, personal care products; QOL, quality of life.

^aAdjusted for BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL.

Analysis of phthalate quartiles

We wanted to identify whether there was a different direction of association among women exposed to lower exposure compared to higher exposures to summary phthalates because EDCs, including phthalates,^27,32,56,73 are known to act in a nonmonotonic (ie nonlinear) manner.⁷⁴ To address this, we completed ordinal logistic regression of the association between phthalates (measured in quartiles) and self-reported frequency of sleep disturbances, insomnia, and restless sleep. Similar to the analyses described above, we adjusted for BMI, menopause status, hot flashes at night, smoking status, CES-D score, and present QOL (see Supplemental Table 6, http://links.lww.com/MENO/A628, demonstrating the estimated associations between high and low quartiles of sum phthalate measures and self-reported sleep measures). We did not find any significant associations between any phthalate measure with any sleep measure in either Q1 or Q2–4.

We then selected to stratify our adjusted model by self-reported smoking status, as described above (see Supplemental Table 7, http://links.lww.com/MENO/A628, demonstrating the estimated associations between high and low quartiles of sum phthalate measures and self-reported sleep measures, stratified by smoking status). Significant relationships are plotted in Figure 1; however, for ease of visualization of the overall patterns, we did not plot the stratified analysis. These relationships are represented in boxplots with both the actual values and the values predicted from the regression.

FIG. 1.

Boxplots representing observed and predicted relationships between frequency of sleep disturbances or insomnia and select summary phthalates, separated based on quartiles of exposure. (A) Relationship between sumPLASTIC and frequency of self-reported sleep disturbances for women in quartile 1 (Q1, left) or quartiles 2–4 (Q2–4, right). (B) Relationship between Q1 of sumPCP (left) or sumPLASTIC (right) and frequency of self-reported insomnia. Observed values are based on n = 762 women; predicted values are from ordinal logistic regression assessing the association between summary phthalates (in quartiles) and the outcome of interest (frequency of sleep disturbances or insomnia). If no boxplot is represented at a specific level, it indicates that women are not predicted to self-report those values based on the current population. Boxplots represent the distribution of the data in several ways: the box is the interquartile range, with the lower portion of the box representing the 25th percentile and the highest portion of the box representing the 75th percentile. The line through each box is representative of the median phthalate value. The lines coming from each box show the minimum (bottom line) and maximum (top line) phthalate data within each level of the sleep disruption measure. Finally, individual points are representative of outliers in the dataset.

Overall, we found that Q2–4 of sumPLASTIC were significantly associated with sleep disturbances in nonsmokers, with frequency of sleep disturbances increased at higher exposure to sumPLASTIC. In former smokers, we found that Q1 of sumPCP was significantly associated with insomnia, and Q1 of sumPLASTIC was significantly associated with both sleep disturbances and insomnia. In former smokers, lower exposure to sumPCP and sumPLASTIC were associated with increased frequency of self-reported poor sleep measures. There was no significant association between any summary phthalate measures and restless sleep at any quartile and level of stratification. Marginal effect sizes for the significant associations from the adjusted and stratified model are presented in supplemental content (see Supplemental Table 8, http://links.lww.com/MENO/A628, demonstrating the marginal effect sizes for the significant estimated associations between summary phthalate measures based on quartile, stratified by smoking status).

DISCUSSION

Overall, our results support previously published literature that shows endogenous hormones are associated with self-reported sleep in midlife women. We found serum concentrations of progesterone and testosterone were negatively associated with self-reported frequency of sleep disturbances and insomnia. We also found the amount of free testosterone was negatively associated with self-reported frequency of insomnia. When we stratified by self-reported smoking status, we observed that progesterone was only associated with sleep disturbances in nonsmokers and former smokers. All other associations between hormones and frequency of sleep disturbances and insomnia were observed in nonsmokers. We also identified associations between summary measures of phthalate exposures and self-reported frequency of sleep disturbances, insomnia, and restless sleep. Specifically, we found that phthalates found in sumPCP and all phthalate metabolites measured (sumALL) were negatively associated with all three self-reported sleep measures. Further, phthalates from DEHP (sumDEHP) and plastic sources (sumPLASTIC) were also negatively associated with self-reported frequency of insomnia. All of these associations between summary phthalates and sleep were only observed in former smokers.

Very little is known about the impact of phthalates on measures of sleep quality. Data from NHANES found one phthalate metabolite (monocyclohexyl phthalate) was associated with increased self-reported nighttime awakenings, whereas another (mono(3-carboxypropyl) phthalate) was associated with leg cramps during sleep in adult men and women (ages 18–85).²⁹ However, this study did not examine summary measures of phthalates. Importantly, this NHANES study also found associations between additional endocrine disruptors and sleep difficulties (ie waking up at night, leg jerks or cramps, daytime sleepiness),²⁹ including arsenic, polyfluoroalkyl compounds, pesticides, heavy metals, and polyaromatic hydrocarbons. Another NHANES study also found an association between urinary concentrations of bisphenol-A and decreased number of hours asleep for both adult men and women.⁷⁷ We found that summary measures of phthalates are associated with self-reported measures of sleep disruptions in midlife women. These observations, together with previous results showing endocrine disruptors are associated with sleep in adult men and women, reveal that environmental chemicals have the potential to influence sleep in adult populations.

Hormones, particularly E₂1 and FSH, are associated with self-reported sleep quality in menopausal women. Although we did not find a significant association between E₂ and any of our sleep measures, we are consistent with other studies that have not found a significant association between sleep and E₂.^{9,10,16,19,78} Importantly, our study did not control for the stage of menstrual cycle; this limitation is significant and must be taken into consideration when interpreting our results.

When one evaluates the relationship between hormones and sleep, it is important to consider differences in study design, such as longitudinal versus cross-sectional cohorts, methods of controlling for menstrual cycle, or assessed sleep measures. Additionally, it is possible that increased anovulatory cycles, particularly amongst perimenopausal women, are responsible for differences observed across studies. Women who have anovulatory cycles generally have a lack of hormone fluctuations and relatively blunted FSH, E₂, and progesterone concentrations across the month compared to women who experience ovulatory cycles.⁷⁹ Therefore, future research must address questions about hormone variability across ovulatory versus anovulatory cycles and how this variability contributes to the development of adverse symptoms.

In addition to the observations discussed above, we also found a significant negative association of free E₂ index and sleep disturbances and insomnia when we stratified by self-reported smoking status. We also identified associations between additional hormones and sleep disturbances and insomnia, including progesterone (negative association), testosterone (negative association), and E₂:progesterone (positive association). Interestingly, our progesterone results are what we predicted, as progesterone is considered a sleep-promoting hormone. However, work from de Zambotti et al⁸⁰ observed midlife women had increased nighttime awakenings and reduced slow wave sleep, indicating poorer sleep. These women were specifically recruited during their luteal phase, a period of relatively elevated progesterone. It is possible that aged brain areas involved in regulating sleep are altered in midlife women, contributing to differences in how hormones influence behavior within this population. Regardless, these results from our hormone analyses indicate that additional hormones, beyond E₂ and FSH, possibly modulate sleep quality and frequency of disrupted sleep among midlife women.

Importantly, our results were highly dependent upon self-reported smoking status. We did not find any significant associations between any hormone measures or summary phthalates and self-reported measures of sleep disruptions in women who were currently smoking at enrollment. We speculate that this could be due to the impact of smoking being more significant than the effect of hormones and phthalates on disrupted sleep in our population. The most significant associations between serum hormones and frequency of self-reported sleep disturbances and insomnia in our study were observed in women who identified as nonsmokers. We did not find a significant association between progesterone and sleep disturbances in both nonsmokers and current smokers. Importantly, smoking affects endogenous hormone concentrations in midlife women, particularly E₂, androgens, and SHBG.⁶⁹ Furthermore, smoking status also impacts sleep quality symptoms, including depression⁸¹ and hot flashes.^82,83 Therefore, a woman’s history with smoking could influence measures of sleep either directly by altering endogenous hormone concentrations, or indirectly by influencing additional symptoms. Furthermore, we did not find a significant association between phthalate exposure (in either Q1 or Q2–4) and frequency of any measures of disrupted sleep in current smokers. Smoking status may be an important confounding factor when one considers the impact of hormones and EDC exposure, particularly in the context of self-reported sleep disruptions and midlife women.

Many EDCs, including phthalates, have effects on both behavior and the brain at low doses or lower exposure.^{27,32,56,73,74,84} Therefore, we examined the differences between our lowest quartile of exposure and the higher quartiles in relation to their association with frequency of disrupted sleep. We did not find any significant associations in our unstratified analysis. However, when we stratified by self-reported smoking status, we found interesting patterns among nonsmokers and former smokers. In nonsmokers, phthalates from plastic sources were positively associated with sleep disturbances only in women with exposures in second-fourth quartiles (ie higher exposure). In former smokers, we found that lower concentrations (ie first quartile) of urinary phthalate metabolites from personal care products and plastics were negatively associated with sleep disturbances and insomnia. This suggests that smoking history influences the relationship between phthalates and self-reported frequency of disrupted sleep as discussed above. Additionally, this indicates that the direction of association between phthalates and frequency of disrupted sleep is dependent upon the amount of exposure. Specifically, we found that lower quartiles of exposure have a negative direction of association, whereas higher quartiles of exposure have a positive association.

It should be noted that our results are limited to pre- and perimenopausal women who are not taking any form of hormonal therapy, including botanical therapies, and who do not have a history of ovarian or breast cancer. Furthermore, our results only reflect women in natural menopause. This limits the generalizability for women who underwent surgical or medically induced menopause. Additionally, as this was a secondary analysis of an existing dataset, we were limited on available sleep measures, as well as measures of desired covariates. Specifically, we did not use an exhaustive index for assessing sleep quality, such as the Pittsburgh Quality Sleep Index or Epworth Sleepiness Scale, or sleep logs for overall sleep patterns. However, our questions about sleep disruptions were modeled after previously published work and are validated as accurate measures of self-reported sleep disruptions.^13,39 Also, we did not measure other important factors that influence sleep, such as caffeine, anxiety, and perceived stress. Therefore, we were unable to include them as covariates in our models. Therefore, to better understand the impacts of hormones and phthalate exposure on self-reported sleep quality, future studies must be designed specifically to control for limitations such as those addressed in this manuscript.

CONCLUSIONS

Overall, we found significant associations with frequency of sleep disturbances, insomnia, or restless sleep and specific endogenous hormones. Further, we identified that the frequency of these sleep disruptions is associated with summary measures of urinary phthalate concentrations. We speculate that exposures to phthalates, as well as other EDCs, may contribute to the development of adverse symptoms in midlife. There is growing evidence that EDCs have negative health implications; therefore, we urge clinicians and individuals to consider EDC exposure as a potential contributor to disease risk. However, it is important to note that self-reported smoking status is a critical factor that influences these significant associations, indicating a more complex relationship that needs to be further studied. We also found additional hormone measures associated with self-reported frequency of sleep disruptions beyond estrogens. Therefore, this warrants further study into additional hormones as well as ratios of hormone profiles and their associations with self-reported sleep measures in midlife women. Future work must examine the impact of phthalate exposure on menopause symptoms, including sleep. Further, additional studies should also explore the complex relationship between multiple factors, including sleep, depression, hormones, and EDCs, as well as the underlying mechanisms of how hormones and EDC exposure influence sleep, particularly in midlife women.

Supplementary Material

Supplemental digital content is available for this article. Direct URL citations are provided in the HTML and PDF versions of this article on the journal’s Website (www.menopause.org).