Trajectory Analysis of Sleep Maintenance Problems in Midlife Women Before and After Surgical Menopause: The Study of Women’s Health Across the Nation (SWAN)

Abstract

Objective:

Investigate temporal patterns of sleep maintenance problems in women who became surgically menopausal (hysterectomy with bilateral oophorectomy) before their final menstrual period and examine whether pre-surgery trajectories of sleep maintenance problems are related to problems staying asleep post-surgery.

Methods:

Longitudinal analysis of sleep self-reports collected every 1–2 years from 1996–2013 from 176 surgically menopausal women in the Study of Women’s Health Across the Nation (SWAN), a 7-site community-based, multi-ethnic/multi-racial, cohort study. Median follow-up was 15.3 years (4.2 years pre-surgery, 10.2 years post-surgery). Group-based trajectory modeling was used to identify patterns of problems staying asleep, and the pre-surgery trajectories were used to predict similar post-surgery sleep problems.

Results:

4 trajectory patterns of sleep maintenance problems were identified: low (33.5% of women), moderate (33.0%), increasing during pre-surgery (19.9%), and high (13.6%). One-fifth of women reported a pre-surgery increase in these problems. Post-surgically, problems staying asleep remained associated with similar levels of pre-surgical problems, even after adjusting for post-surgical early morning awakening, frequent vasomotor symptoms, and bodily pain score (β_low = −1.716, β_moderate = −1.144, β_increasing = −0.957, β_high = −1.021; all Ps < 0.01).

Conclusions:

Sleep maintenance problems were relatively stable across time post-surgery. These data are remarkably consistent with our trajectory results across the natural menopause, suggesting that pre-surgical assessment of sleep concerns could help guide women’s expectations post-surgically. While reassuring that sleep complaints do not worsen post-surgically for most surgically menopausal women, referral to a sleep specialist should be considered if sleep symptoms persist or worsen after surgery.

Introduction

Hysterectomy is common, with approximately 37% of American women having had a hysterectomy by age 60 years, at least half of whom undergo concurrent bilateral oophorectomy,^1,2 i.e., are “surgically menopausal”.³ Hysterectomy with or without bilateral oophorectomy has been linked with poor sleep.^4–6 Little is known about the longitudinal patterns and progression of self-reported sleep problems in surgically menopausal women, both prior to and following surgical menopause. Given the associations of sleep health with psychological and physical health and functioning in aging persons,^7–8 particularly women, this is a major gap in the literature.

Difficulty staying asleep due to repeated awakenings during the night is a common sleep problem related to the menopause transition (MT).^4,9 However, we found that not all naturally postmenopausal women experience a rise in sleep problems during the MT. Although most women’s sleep maintenance problems were stable or increased slowly from pre-final menstrual period (FMP) to post-FMP, only one trajectory group (15% of the sample) demonstrated an early and rapidly increasing pre-FMP rise in problems staying asleep.¹⁰ In a recent cross-sectional analysis, surgically menopausal women (oophorectomy status not specified) had more than double the risk of clinical insomnia symptoms relative to naturally menopausal women.¹¹

Group based trajectory analysis is a statistical procedure for recognizing and visualizing different patterns of temporal change in a characteristic or outcome.¹² This modeling procedure uncovers unobserved heterogeneity in the data, and distinct longitudinal patterns (termed “latent trajectories”) can be identified, to reveal (in this article) subgroups of women who share a similar course in their sleep maintenance problem.¹²

Our objectives were to investigate in surgically menopausal women the heterogeneity of temporal patterns of sleep problems across the MT and to examine whether the pattern (trajectories) of self-reported problems staying asleep pre-surgery are related to sleep maintenance problems post-surgery after adjusting for covariates known to be associated with sleep problems.

Methods

Study Design and Participants

SWAN is a multi-ethnic/multi-racial, community-based, cohort study of the menopausal transition. Initiated in 1996, 3,302 women were enrolled at seven sites: Boston, MA, Chicago, IL, Detroit area, MI, Los Angeles and Oakland, CA, Newark, NJ, and Pittsburgh, PA. Study design and recruitment of the SWAN cohort have been described in detail.¹³ Briefly, each site recruited white women and a racial/ethnic minority group. Eligible women were 42–52 years, premenopausal or early perimenopausal, had an intact uterus and at least one ovary, had at least one menstrual period in the previous three months, and were not pregnant/lactating or using any sex steroid hormones in the three months preceding the baseline interview. Extensive data on psychological, social and health parameters were collected at baseline and at follow-up visits scheduled at approximately 12- to 18-month intervals. Institutional review board approval was obtained at each SWAN study site and written informed consent was obtained from participants at each visit.

Procedure and Measures

Participant selection for inclusion in the analytic sample is shown in Figure 1 (STROBE diagram). Women who reported a natural menopause were excluded from this analysis. Between 1996–2013, 283 of 3302 SWAN women who had not yet reached a natural FMP self-reported a surgically induced menopause (Figure 1). Self-report regarding surgery was supplemented by medical records if available, and if the data conflicted, medical record data were used to determine surgical categorization. Of the 283 women, 184 (65.0%) reported having a hysterectomy with bilateral oophorectomy (“surgical menopause”) and 99 (35.0%) women were excluded due to not meeting this criterion. No woman reported having a bilateral oophorectomy without a concomitant hysterectomy. Eight surgically menopausal women were excluded because of insufficient sleep assessments (i.e., did not have at least 1 observation pre-surgery and at least 1 post-surgery) for the trajectory analyses, leaving 176 women (78.4% with medical record confirmation) in the analytic sample. Compared with these 176 women, the 107 women excluded from the analysis were similar on most baseline characteristics examined, with a few exceptions. Excluded women were 0.6 years younger, a slightly higher percentage completed at least some college and a slightly lower percentage completed high school or less, a smaller percentage were married and a larger percentage were unpartnered, a smaller percentage were premenopausal and a larger percentage were early perimenopausal at baseline, and a smaller percentage drank alcohol moderately or less). There was no significant difference in self-reported trouble falling asleep, sleep maintenance problems, early morning awakening, or any sleep problem.

Figure 1:

STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) Diagram: Participant selection for data analysis, included and excluded women.

Dependent variables:

At each assessment, women self-reported frequency of sleep problems in the past 2 weeks, which was dichotomized as “no” (<3 times a week) or “yes” (≥3 times a week) for each of three sleep problems (trouble falling asleep, waking several times, early morning awakening). This dichotomization is consistent with the criterion for insomnia in both the Diagnostic and Statistical Manual of Mental Disorders, 5th edition.¹⁴ and the International Classification of Sleep Disorders, 3rd edition.¹⁵ “Waking several times” (problems staying asleep/sleep maintenance problems) is the focus of the analysis because it was the most frequent sleep problem, and almost as prevalent as the composite variable ‘any sleep problem’, and so represented the most uniform measure for examining sleep changes before and after surgical menopause.

Independent variables:

The primary independent variables were the pre-surgery sleep trajectory groups for sleep maintenance problems.

Covariates:

The primary covariates were selected based on previously identified associations with sleep maintenance problems, and all were time-varying, including (a) two other individual sleep problems (trouble falling asleep, early morning awakening), (b) VMS (hot flashes/flushes and/or night sweats) frequency in the past 2 weeks (0, no symptoms; 1, fewer than 6 days; 2, 6–14 days), and (c) serum reproductive hormone concentrations (follicle stimulating hormone, FSH; estradiol, E2). FSH and E2 were measured on days 2–5 of a spontaneous menstrual cycle occurring within 60 days of recruitment at the baseline visit, and annually thereafter. If a day 2–5 specimen could not be obtained, or once the participant was surgically menopausal, a random fasting specimen was taken within 90 days of the annual assessment.¹⁶ Because not all pre-surgical reproductive hormone samples were collected within the intended window, cycle day of blood draw also was included as a covariate (dichotomized as day 2–5 versus not day 2–5 or unknown).

Other time-varying covariates included marital status (married/living with a partner or unpartnered), body mass index (BMI, kg/m²), depressive symptoms (Center for Epidemiologic Studies-Depression scale (CES-D)¹⁷ score (sum of all 20 items minus the sleep item), and anxiety symptoms score (sum of 4 items derived from a menopausal symptom questionnaire, each item scored 0 (not at all) to 4 (every day over 14 days)),^18,19 psychosocial variables (social support,²⁰ financial strain (very or somewhat hard to pay for basics versus not hard at all),²¹ self-assessed health (very good/excellent versus poor/fair/good),²² number of medical conditions (0, 1, or 2 or more conditions), the Short Form-36 bodily pain score (SF-36 pain score; sum of two items assessing the level of pain severity and extent that pain interferes with function; higher scores indicate less pain),²² health-related behaviors (current smoking, alcohol and caffeine use, physical activity), and medication use (hormone therapy [HT], medications for nervous conditions [eg, antidepressants, tranquilizers], sedatives [including medication taken to sleep], and medications taken for pain).

Sociodemographic covariates were time-invariant: self-identified race/ethnicity, study site, education (completed high school or less versus more than a high school diploma), and age at surgery.

Data Analysis

Analyses were centered on surgery date (time 0) with two-tailed P values < 0.05 considered statistically significant (SAS 9.3; SAS, Cary, NC).

To describe trajectory patterns of problems with sleep maintenance problems across the MT, from pre-surgery through post-surgery, group-based trajectory modeling was used (PROC TRAJ).^23–25 We examined whether 3 or 4 groups fit the data best, allowing linear, quadratic, and cubic terms for time in each group. Higher order terms were eliminated one by one, starting with the least significant term in any group. Bayes Information Criterion (BIC) was used to select the number of distinct trajectory groups.²³ Analyses were adjusted for age at surgery and site/race indicators.

To answer the question of whether pre-surgery sleep predicts sleep problems post-surgery, each woman’s sleep reports were grouped into pre- and post-surgery time periods. Describing temporal patterns across the entire observation period required at least 1 observation pre-surgery and 1 observation post-surgery. Predicting post-surgery patterns required at least 2 observations pre-surgery and 2 post-surgery; this was a within group analysis and used models with an intercept for each group separately instead of choosing one group as a reference. Trajectory analyses were repeated for the pre-surgery period, adjusted for age at surgery and site/race indicators. Four sleep maintenance problem groups were identified based on only pre-surgery observations: (low prevalence, moderate prevalence, increasing prevalence, and high prevalence). In the best-fitting trajectory model for this restricted time pre-surgery, only one non-constant time term was retained, and it was a linear term for the increasing prevalence group. Prevalences of sleep problems post-surgery were analyzed using repeated-measures log-binomial regression analysis with the generalized estimating equations (GEE) method.²⁶ The log-binomial model with robust standard errors was used, yielding relative risk estimates.²⁶ The main effects of group, time, and the interaction of group and time were included in Model 1. The model then was adjusted for time-varying early morning awakening and trouble falling asleep, VMS, and hormones (E2 and FSH). Models without the group by time interaction fit the data as well as those with the interaction. Therefore, we present only main effects models. For ease of interpretation, and as the question of interest is whether pre-surgery sleep problems predict those in post-surgery, we chose to use models with an intercept for each group separately instead of choosing one group as a reference. Single covariates were added to the model, and those which were significant at P ≤ 0.10 were included in the final model. Financial strain rather than income was added to the model due to collinearity and more missing data for income. The New Jersey (NJ) site did not complete in-person follow-up visits 7 and 8, but in-person assessments resumed at visit 9. Because sensitivity analyses excluding NJ women (N = 10) showed almost identical results, they were included in the primary analysis.

Results

Baseline characteristics

Table 1 displays the baseline characteristics for the 176 surgically menopausal women. Their mean age was 46.2 years and surgery occurred, on average, 5 years later at a mean age of 51.2 years.

Table 1:

Baseline characteristics of participants included in analysis of prevalence of sleep problems and trajectory patterns pre- and post-surgery

Characteristic a	Baseline Sample (n=176)
Age at surgery, years	51.2 ± 4.0
Age at baseline, years	46.2 ± 2.6
Body Mass Index (kg/m2), N=173	29.7 ± 8.1
Social Support, range 0–16	12.2 ± 3.5
Caffeine per day, mg, N=174	243.4 ± 262.0
Short Form-36 bodily pain score, range 0–100	65.6 ± 22.6
Center for Epidemiologic Studies-Depression scale score, excluding Sleep item, range 0–57	10.8 ± 8.9
Anxiety symptoms score, range 0–16	2.8 ± 2.3
Total physical activity, range 3–15, N=172	7.5 ± 1.8
Estradiol (pg/mL)	51.8 (33.1 – 82.8)
Follicle stimulating hormone (mIU/mL)	13.8 (10.2 – 22.1)
Site
Michigan	38 (21.6)
Boston	18 (10.2)
Chicago	32 (18.2)
Oakland	13 (7.4)
Los Angeles	25 (14.2)
New Jersey	14 (8.0)
Pittsburgh	36 (20.5)
Self-perceived health
excellent/very good	95 (54.0)
good/fair/poor	76 (43.2)
missing	5 (2.8)
Financial strain
very hard/somewhat hard to pay for basics	64 (36.4)
not hard to pay for basics	111 (63.1)
missing	1 (0.6)
Income
less than $20,000	18 (10.2)
$20,000 − < $35,000	25 (14.2)
$35,000 − < $50,000	37 (21.0)
$50,000 − < $75,000	44 (25.0)
$75,000 − < $100,000	16 (9.1)
$100,000 or more	30 (17.0)
missing	6 (3.4)
Race/Ethnicity
Black	60 (34.1)
Chinese	10 (5.7)
Hispanic	8 (4.5)
Japanese	9 (5.1)
White	89 (50.6)
Education
completed high school or less	37 (21.0)
completed at least some college	138 (78.4)
missing	1 (0.6)
Marital status
married or living as married	119 (67.6)
separated/divorced or widowed	51 (29.0)
missing	6 (3.4)
Menopausal status
premenopausal	86 (48.9)
early perimenopausal	87 (49.4)
indeterminant b	3 (1.7)
Medications c
nervous conditions	19 (10.8)
pain	83 (47.2)
sedatives/hypnotics	14 (8.0)
Smoking
current smoker	27 (15.3)
not current smoker	146 (83.0)
missing	3 (1.7)
Vasomotor symptoms
6–14 days/2 weeks	23 (13.1)
1–5 days/2 weeks	65 (36.9)
none	88 (50.0)
Missing	0 (0)
Alcohol use
14 or more drinks/week	23 (13.1)
less than 14 drinks/week	143 (81.3)
Missing	10 (5.7)
Medical conditions d
2 or more	95 (54.0)
1	51 (29.0)
0	30 (17.0)
Sleep Problems e
any	59 (33.5)
trouble falling asleep	22 (12.5)
sleep maintenance problems	50 (28.4)
early morning awakening	28 (15.9)

Data are mean ± SD, median (interquartile range), or n (%).

^aN=176 includes women who had at least 1 observation pre-surgery and at least 1 observation post-surgery. For continuous variables, number of participants with data shown in table if less than 176. For categorical variables, number (%) missing is included in table. Column percentages may not add up to 100% due to rounding.

^bStudy participants who were eligible for enrollment but ‘menopausal status’ category (premenopausal or early perimenopausal) not specified in database.

^cN = number of women who reported taking the medication. Medications were classified from product brand or generic names using a computerized medication dictionary (Iowa Drug Information Service (IDIS) Drug Vocabulary, College of Pharmacy, University of Iowa, Iowa City, IA).

Medication for nervous conditions includes the following IDIS categories: antidepressants, antipsychotics, other psychotherapeutic agents (e.g., lithium).

Pain medication includes the following IDIS categories: non-steroidal anti-inflammatory drugs (includes Cox-2 inhibitors), analgesics and other antipyretics (includes acetaminophen and salicylates), opioids, and relaxants.

Sedative medication includes the following IDIS categories: benzodiazepine and non-benzodiazepine antianxiety medications and sleeping pills, and barbiturates.

^dSpecific medical conditions: diabetes, anemia, hypertension, osteoporosis, hyperlipidemia, migraines, thyroid problems, osteoarthritis, stroke, heart attack, angina, cancer, fibroids.

^eReported at least 3 nights weekly in the previous 2 weeks.

Total duration of study participation was 1.9–16.5 years, with a median (interquartile range, IQR) = 15.3 (14.8–15.6) years (pre-surgery = 4.2 (2.2–7.2) years, post-surgery = 10.2 (6.7–12.9) years). Median (IQR) number of assessments was 13 (9–14), 4 (2–6) pre-surgery and 6 (4–10) post-surgery, and the total number of sleep assessments was 1,975 (806 pre-surgery; 1,169 post-surgery).

Prevalence of sleep problems

Figure 2 graphically displays prevalence by years centered at the time of surgery, and Table 2 shows these data grouped separately for baseline, pre-surgery, and post-surgery. At baseline, 33.5% reported one or more of the 3 sleep problems at least three times a week in the previous 2 weeks. Sleep maintenance, the most prevalent problem, was reported by 28.4% of the sample at baseline, 33.1% pre-surgery, and 43.3% post-surgery, with prevalence increasing across time pre-surgery and stable post-surgery.

Figure 2:

Prevalence of sleep problems at least 3 nights per week reported before and after surgery (N = 176). ^a

^a X-axis: vertical dashed line centered at the time of surgery (0). Negative numbers indicate years before surgery and positive numbers indicate years after surgery. Legend embedded within figure indicates the line pattern associated with each of the 3 individual types of self-reported sleep problems and the composite group for any sleep problem.

Table 2:

Percent of women with sleep problems at baseline and pre- and post-surgery

	Baseline (n=176)	Pre- Surgery (n=176)	Post- Surgery (n=176)
Sleep maintenance problems	28.4	33.1	43.3
Trouble falling asleep (i.e., sleep onset)	12.5	14.1	18.3
Early morning awakening	15.9	18.4	21.4
Any sleep problem	33.5	39.6	50.5
No sleep problems	66.5	60.4	49.5
Sleep maintenance problems only	13.6	14.2	22.1
Trouble falling asleep only	2.8	3.1	3.5
Early morning awakening only	1.7	3.2	2.9
Only SM & TFA	1.1	3.8	3.5
Only SM & EMA	5.7	8.0	7.2
Only TFA & EMA	0.6	0.1	0.8
All 3 sleep problems	8.0	7.1	10.5

SM, Sleep Maintenance problems; TFA, Trouble Falling Asleep; EMA, Early Morning Awakening.

Column numbers represent the percentage (%) for each problem reported at least 3 nights weekly in the previous 2 weeks. The percentages for “Pre‐Surgery” and “Post-Surgery” present the average percent of women reporting the sleep problem(s) per visit during the “Pre‐Surgery” and “Post-Surgery” period, respectively.

First 5 rows display percentages for each specific sleep problem regardless of co-occurring additional sleep problem(s). Single indicators of sleep problems are shown in the next 3 rows, labeled “Sleep Maintenance problems only” etc. Combined sleep problems (“SM & TFA,” “SM & EMA,” “TFA & EMA,” and “All”) are shown in the last 4 rows.

For example, averaged across all pre-surgery visits, 33.1% of women reported Sleep Maintenance problems when they were pre-surgery (row 1, column 2). This percentage is based on 14.2% reporting SM only, 3.8% reporting SM and TFA, 8.0% reporting SM and EMA, and 7.1% reporting all 3 sleep problems, for a total prevalence of = 33.1% with Sleep Maintenance problems.

Trajectories for sleep maintenance problems

Trajectory groups: Figure 3 displays four distinct trajectory groups for sleep maintenance problems across the entire observation period for the 176 women. All but two women contributed more than two time points to these trajectories, and only four women contributed just three time points. The four groups identified include a group with a low prevalence (33.5%), a group with a moderate prevalence (33.0%), a group with an increasing prevalence (19.9%), and a group with a high prevalence (13.6%). Thus, only 20% of women demonstrated a pattern of increasing sleep maintenance problems across surgical menopause, while the others demonstrated stable levels of low, medium, or high prevalence relative to their pre-surgery sleep maintenance problems.

Figure 3:

Trajectories for sleep maintenance problems at least 3 nights per week reported before and after surgery (N = 176).^a,b

^a Groups were determined by trajectory analysis (see text for details) using the entire range of observed data.

Legend embedded within figure indicates the line pattern for each of the 4 trajectory groups: low prevalence of sleep maintenance problems (N=59; 33.5%); moderate prevalence of sleep maintenance problems (N=58; 33.0%); increasing prevalence of sleep maintenance problems (N=35; 19.9%); high prevalence of sleep maintenance problems (N=24; 13.6%).

^b X-axis: vertical dashed line centered at the time of surgery (0). Negative numbers indicate years before surgery and positive numbers indicate years after surgery.

Table 3 displays the baseline characteristics for the four pre-surgery trajectory groups (N=136). The 136 women included in this subsample completed a median (IQR) 13 (10–14) sleep assessments (pre-surgery = 5 (3–7), post-surgery = 7 (4–9) years), and the total number of sleep assessments was 1,623 (694 pre-surgery; 929 post-surgery). Overall, these four groups differed significantly on age at surgery (p < 0.01), anxiety symptoms (p = 0.04), physical activity (p = 0.03), medications for nervous symptoms (p = 0.03) and pain (p < 0.05), VMS (p = 0.02), and self-reported sleep problems (any sleep problem and sleep maintenance problems (both p < 0.01), and early morning awakening (p = 0.01)). They did not differ on age at baseline, BMI, social support, SF 36 pain score, E2, FSH, medical health conditions, self-reported health, depressive symptoms, lifestyle behaviors (caffeine, alcohol, smoking), race/ethnicity, or other sociodemographic or socioeconomic characteristics (all p values > 0.05).

Table 3:

Baseline (pre-surgery) characteristics overall and by group membership for the 136 participants with at least 2 observations pre-surgery and at least 2 observations post-surgery

		Pre-Surgery Sleep Maintenance Problem Group
Characteristic a	Overall	Low	Moderate	Increasing	High	P-valueb
N (%)	136	53 (39.0)	41 (30.1)	17 (12.5)	25 (18.4)
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
Age at surgery, years	51.4 (3.8)	50.1 (3.0)	52.7 (3.8)	52.5 (4.5)	51.3 (3.8)	0.01
Age at baseline, years	46.1 (2.6)	45.8 (2.5)	46.4 (2.7)	46.3 (2.9)	46.3 (2.6)	0.73
Body Mass Index (kg/m2), N=133	28.9 (7.3)	29.3 (7.9)	28.1 (6.6)	29.5 (8.1)	29.1 (7.2)	0.87
Social support	12.3 (3.3)	12.3 (3.4)	11.8 (3.8)	12.5 (2.6)	13.1 (2.9)	0.50
Caffeine per day, mg	247.5 (274.0)	245.7 (312.6)	233.2 (198.8)	273.4 (277.7)	257.9 (305.2)	0.96
Short form-36 bodily pain score	67.7 (20.9)	74.3 (18)	62.5 (25)	65.9 (20.7)	63.2 (16.3)	0.24
Center for Epidemiologic Studies-Depression scale score, excluding sleep item	10.1 (8.3)	9.2 (7.3)	10.5 (8.8)	9.4 (7.7)	11.8 (9.5)	0.58
Anxiety symptoms score	2.7 (2.1)	2 (1.8)	3.1 (2.5)	3.2 (2.1)	2.9 (1.8)	0.04
Total physical activity, N=133	7.5 (1.7)	7.3 (1.6)	7.5 (1.8)	7.2 (1.6)	8.1 (1.9)	0.03
	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)	Median (IQR)
Estradiol (pg/mL)	51.3 (31.9, 81.7)	50.8 (39.6, 79.7)	47.9 (29.5, 69.1)	52.4 (27.1, 83.0)	51.4 (34.7, 88.2)	0.67
Follicle stimulating hormone (mIU/mL)	13.5 (10.4, 21.7)	12.8 (9.4, 18.4)	15.1 (10.9, 26.1)	14.7 (11.5, 18.6)	13.8 (10.8, 18.7)	0.54
	N (%)	N (%)	N (%)	N (%)	N (%)
Site
Michigan	27 (19.9)	12 (22.6)	4 (9.8)	4 (23.5)	7 (28.0)	0.27
Boston	13 (9.6)	3 (5.7)	4 (9.8)	3 (17.6)	3 (12.0)
Chicago	25 (18.4)	10 (18.9)	10 (24.4)	1 (5.9)	4 (16.0)
Oakland	11 (8.1)	6 (11.3)	1 (2.4)	1 (5.9)	3 (12.0)
Los Angeles	20 (14.7)	9 (17.0)	5 (12.2)	5 (29.4)	1 (4.0)
New Jersey	10 (7.4)	2 (3.8)	5 (12.2)	0 (0)	3 (12.0)
Pittsburgh	30 (22.1)	11 (20.8)	12 (29.3)	3 (17.6)	4 (16.0)
Overall self-rated health
excellent/very good	75 (55.1)	30 (56.6)	22 (53.7)	8 (47.1)	15 (60.0)	0.12
good/fair/poor	58 (42.6)	23 (43.4)	19 (46.3)	7 (41.2)	9 (36.0)
Missing	3 (2.2)	0 (0)	0 (0)	2 (11.8)	1 (4.0)
Financial strain
hard/somewhat hard to pay for basics	44 (32.4)	20 (37.7)	13 (31.7)	8 (47.1)	3 (12.0)	0.07
not hard to pay for basics	92 (67.6)	33 (62.3)	28 (68.3)	9 (52.9)	22 (88.0)
Income
<$20,000	9 (6.6)	4 (7.5)	4 (9.8)	1 (5.9)	0 (0)	0.69
$20,000 − <$35,000	21 (15.4)	10 (18.9)	4 (9.8)	4 (23.5)	3 (12.0)
$35,000 − <$50,000	31 (22.8)	14 (26.4)	10 (24.4)	3 (17.6)	4 (16.0)
$50,000 − <$75,000	35 (25.7)	15 (28.3)	8 (19.5)	3 (17.6)	9 (36.0)
$75,000 − < $100,000	14 (10.3)	3 (5.7)	7 (17.1)	1 (5.9)	3 (12.0)
$100,000 and up	23 (16.9)	6 (11.3)	7 (17.1)	5 (29.4)	5 (20.0)
Missing	3 (2.2)	1 (1.9)	1 (2.4)	0 (0)	1 (4.0)
Race/Ethnicity
Black	45 (33.1)	20 (37.7)	11 (26.8)	4 (23.5)	10 (40.0)	0.19
Chinese	9 (6.6)	6 (11.3)	1 (2.4)	1 (5.9)	1 (4.0)
Hispanic	4 (2.9)	1 (1.9)	3 (7.3)	0 (0)	0 (0)
Japanese	7 (5.1)	5 (9.4)	2 (4.9)	0 (0)	0 (0)
White	71 (52.2)	21 (39.6)	24 (58.5)	12 (70.6)	14 (56.0)
Education
completed high school or less	25 (18.4)	11 (20.8)	6 (14.6)	4 (23.5)	4 (16.0)	0.81
completed at least some college	111 (81.6)	42 (79.2)	35 (85.4)	13 (76.5)	21 (84.0)
Marital status
married or living as married	93 (68.4)	35 (66.0)	30 (73.2)	10 (58.8)	18 (72.0)	0.33
separated/widowed or divorced	39 (28.7)	17 (32.1)	11 (26.8)	5 (29.4)	6 (24.0)
Missing	4 (2.9)	1 (1.9)	0 (0)	2 (11.8)	1 (4.0)
Menopausal status
premenopausal	67 (49.3)	24 (45.3)	20 (48.8)	9 (52.9)	14 (56.0)	0.83
early perimenopausal	67 (49.3)	29 (54.7)	21 (51.2)	6 (35.3)	11 (44.0)
Indeterminant c	2 (1.5)	0 (0)	0 (0)	2 (11.8)	0 (0)
Medications (IDIS categories) d
nervous conditions	14 (10.3)	2 (3.8)	9 (22.0)	1 (5.9)	2 (8.0)	0.03
Pain	64 (47.1)	17 (32.1)	23 (56.1)	9 (52.9)	15 (60.0)	<0.05
sedatives	9 (6.6)	4 (7.5)	3 (7.3)	0 (0)	2 (8.0)	0.71
Smoking
current smoker	18 (13.2)	9 (17.0)	2 (4.9)	3 (17.6)	4 (16.0)	0.47
not currently smoking	115 (84.6)	43 (81.1)	37 (90.2)	14 (82.4)	21 (84.0)
Missing	3 (2.2)	1 (1.9)	2 (4.9)	0 (0)	0 (0)
Vasomotor symptoms
6–14 days/2 weeks	16 (11.8)	1 (1.9)	6 (14.6)	2 (11.8)	7 (28.0)	0.02
1–5 days/2 weeks	52 (38.2)	19 (35.8)	17 (41.5)	9 (52.9)	7 (28.0)
None	68 (50.0)	33 (62.3)	18 (43.9)	6 (35.3)	11 (44.0)
Alcohol use
14 or more drinks/week	19 (14.0)	6 (11.3)	6 (14.6)	2 (11.8)	5 (20.0)	0.49
less than 14 drinks/week	112 (82.4)	44 (83.0)	35 (85.4)	15 (88.2)	18 (72.0)
Missing	5 (3.7)	3 (5.7)	0 (0)	0 (0)	2 (8.0)
Medical conditions e|
2 or more	74 (54.4)	27 (50.9)	24 (58.5)	6 (35.3)	17 (68.0)	0.30
1	37 (27.2)	14 (26.4)	9 (22.0)	8 (47.1)	6 (24.0)
0	25 (18.4)	12 (22.6)	8 (19.5)	3 (17.6)	2 (8.0)
Sleep problems at least 3 nights weekly in the previous 2 weeks
Any	42 (30.9)	4 (7.5)	17 (41.5)	3 (17.6)	18 (72.0)	<0.01
trouble falling asleep	11 (8.1)	3 (5.7)	3 (7.3)	0 (0)	5 (20.0)	0.08
sleep maintenance problems	35 (25.7)	1 (1.9)	15 (36.6)	1 (5.9)	18 (72.0)	<0.01
early morning awakening	19 (14)	1 (1.9)	7 (17.1)	2 (11.8)	9 (36.0)	0.01

Data are n (%).

^aFor continuous variables, number of participants with data shown in table if less than 136. For categorical variables, number (%) missing is included in table. Column totals may not add up to 100% due to rounding.

^bP-value for continuous variables from Analysis of Variance except for estradiol and follicle stimulating hormone which used the Kruskal-Wallis test. P-value for categorical variables from Chi-square test. Significant associations are bolded.

^cStudy participants who were eligible for enrollment but ‘menopausal status’ category (premenopausal or early perimenopausal) not specified in database.

^dN = number of women who reported taking the medication. Medications were classified from product brand or generic names using a computerized medication dictionary (Iowa Drug Information Service (IDIS) Drug Vocabulary, College of Pharmacy, University of Iowa, Iowa City, IA) (see footnote to Table 1 for full description).

^eSpecific medical conditions: diabetes, anemia, hypertension, osteoporosis, hyperlipidemia, migraines, thyroid problems, osteoarthritis, stroke, heart attack, angina, cancer, fibroids..

Pre-surgery trajectories as predictors of post-surgery sleep (N = 136, Table 4): In the unadjusted within group analyses, the low group had the lowest (i.e., most negative) estimate post-surgery, indicating the lowest prevalence of sleep maintenance problems, relative to the other three groups (β_low = −1.424 95% CI = −1.749 - −1.099; β_moderate = −0.683, 95% CI = −0.905 - −0.460; β_increasing = −0.482, −0.762 - −0.203; β_high = −0.332, 95% CI = −0.518 - −0.146; all Ps ≤ 0.01). The intercepts for these four groups are ordered from low, moderate, increasing, to high. This ordering was preserved during the post-surgery follow-up period because time (unadjusted: β = −0.005, 95% CI = −0.026 – 0.017, p = 0.67) and the group by time interactions (data not shown in Table 4) were not statistically significant. These results indicate that the prevalence of sleep problems within each group did not change during the post-surgery follow-up, i.e., the trajectories of pre-surgery sleep problems continue similarly post-surgery.

Table 4:

Predicting post-surgical sleep maintenance problems from pre-surgical trajectories in the base (unadjusted) and the adjusted (partially and final) models (N=136) ^a

Variable	Estimate	SE	Lower 95% CI	Upper 95% CI	P-value
Unadjusted Model
Group
high	−0.332	0.095	−0.518	−0.146	<0.01
increasing	−0.482	0.142	−0.762	−0.203	0.01
moderate	−0.683	0.114	−0.905	−0.460	<0.01
low	−1.424	0.166	−1.749	−1.099	<0.01
Time (Years)	−0.005	0.011	−0.026	0.017	0.67
Partially Adjusted Model (all pre-specified covariates)
Group
high	−0.970	0.195	−1.352	−0.588	<0.01
increasing	−0.913	0.201	−1.308	−0.518	<0.01
moderate	−1.131	0.199	−1.521	−0.741	<0.01
low	−1.711	0.193	−2.090	−1.332	<0.01
Time (Years)	0.001	0.013	−0.023	0.026	0.92
Early morning awakening	0.691	0.114	0.468	0.913	<0.01
Trouble falling asleep	0.177	0.098	−0.014	0.369	0.07
Vasomotor symptoms
6–14 days/2 weeks	0.381	0.131	0.124	0.638	0.01
1–5 days/2 weeks	0.105	0.119	−0.129	0.338	0.38
none	Reference
Estradiol (pg/mL) b,c	0.019	0.067	−0.111	0.150	0.77
Follicle stimulating hormone (mIU/mL) b,c	−0.016	0.088	−0.189	0.157	0.86
Final Adjusted Model d
Group
high	−1.021	0.200	−1.413	−0.629	<0.01
increasing	−0.957	0.206	−1.360	−0.553	<0.01
moderate	−1.144	0.194	−1.524	−0.765	<0.01
low	−1.716	0.191	−2.089	−1.343	<0.01
Time (Years)	0.005	0.013	−0.020	0.030	0.69
Early morning awakening	0.668	0.112	0.448	0.887	<0.01
Trouble falling asleep	0.135	0.104	−0.068	0.338	0.19
Vasomotor symptoms
6–14 days/2 weeks	0.370	0.130	0.114	0.626	0.01
1–5 days/2 weeks	0.085	0.119	−0.148	0.317	0.48
none	Reference
Estradiol (pg/mL) b,c	0.041	0.068	−0.093	0.175	0.55
Follicle stimulating hormone (mIU/mL) b,c	0.014	0.087	−0.158	0.185	0.88
Short Form-36 bodily pain score c	−0.005	0.002	−0.010	−0.0001	0.04

SE, standard error; CI, confidence interval.

^aNumber of women in each group: high, n=25; increasing, n=17; moderate, n=41; low, n=53; significant associations are bolded.

^bLog transformed.

^cCentered.

^dIn the final adjusted model, the following covariates were not included because they were not significant at level 0.10: education, financial strain, marital status, body mass index, physical activity, self-rated health, number of medical conditions, depressive symptoms, anxiety symptoms, social support, alcohol use, caffeine use, smoking, menopausal hormone therapy use, medications for a nervous condition or pain, or sedative/hypnotic medications.

Adjusted model estimates for all four groups remained significant in the final model (Table 4; β_low = −1.716, 95% CI = −2.089 - −1.343; β_moderate = −1.144, 95% CI = −1.524 - −0.765; β_increasing = −0.957, 95%CI = −1.360 - −0.553; β_high = −1.021, 95% CI = −1.413 - −0.629; all Ps < 0.01). In the adjusted models, the ordering of the intercepts was not seen, likely because of covariate effects (partial: VMS, p = 0.01, early morning awakening, p < 0.01, trouble falling asleep, p = 0.07; final: VMS, p = 0.01, early morning awakening, p < 0.01, SF-36 bodily pain, p = 0.04). The estimate for time remained non-significant in the partial and final adjusted models (partial: β = 0.001, 95% CI = −0.023 – 0.026, p = 0.92; final: β = 0.005, 95% CI = −0.020 – 0.030, p = 0.69). Sleep maintenance problems were stable throughout the observation period across the menopausal transition for most women (87.5%). Women in the increasing prevalence group (12.5%) experienced a marked increase up to the time of surgery, rising from initial levels similar to the low prevalence group, and then stayed at the elevated level similar to the high group.

Estimates can be converted to relative risk (RR) for between groups comparisons. In the unadjusted model, the RR (95% CI) was 2.98 (2.13 – 4.17) for the high group. 2.56 (1.71 – 3.85) in the increasing group, and 2.10 (1.45 – 3.04) in the moderate group compared to the low prevalence group. In the adjusted model, these estimates were attenuated but remained statistically significant: 2.00 (1.42 – 2.84), 2.14 (1.45 – 3.16), and 1.77 (1.21 – 2.59).

In the final adjusted model (Table 4), early morning awakening (β = 0.668, 95% CI = 0.448 – 0.887, p < 0.01) and frequent VMS (β = 0.370, 95% CI = 0.114 – 0.626, p = 0.01) also were significantly associated with sleep maintenance problems post-surgery. The addition of other covariates (education, financial strain, marital status, body mass index, physical activity, self-rated health, number of medical conditions, depressive symptoms, anxiety symptoms, social support, alcohol use, caffeine use, smoking, menopausal hormone therapy use, medications for a nervous condition or pain, or sedative/hypnotic medications; all p values > 0.10; Table 4 footnote d) had little impact on the pre-surgery sleep trajectory parameter estimates for frequent nighttime awakenings and were not included in the final adjusted model. Only the bodily pain score also had a significant association (β = −0.005, 95% CI = −0.010 – −0.0001, p = 0.04); more pain was associated with sleep maintenance problems.

Sensitivity analysis:

We conducted this analysis on the trajectory creation that excluded 25 (18.4%) of the 136 women with only two time points pre-surgery. The analysis revealed four groups with the same patterns of low, moderate, increasing, and high sleep maintenance problems, and similar percentages in each group (data not shown).

Discussion

In this report, we observed four distinct trajectory patterns for sleep maintenance problems in midlife women who became surgically menopausal (Figure 3), with only 20% (n=17) of women experiencing a pattern of increasing sleep maintenance problems mainly during the pre-surgery period of observation, while all others experienced sleep maintenance problems at a stable level of low, moderate, and high prevalence, similar to that observed pre-operatively. Within each of these three trajectory groups, the pre-surgery trajectory was related to a similar pattern of sleep maintenance problems post-surgery. These data are remarkably consistent with our natural menopause trajectory analyses.¹⁰

Since surgery preceded their natural FMP, the relative stability across time post-surgery within each trajectory group suggests that the surgical intervention and its resultant hormone changes did not impact the pattern of sleep maintenance problems. Although one might anticipate that hysterectomy might improve sleep problems for some women whose hysterectomy indication is driven by poor quality-of-life, sleep disturbance per se is not an indication for hysterectomy, and resolution of the primary gynecologic concern warranting surgical intervention did not lead to better sleep maintenance in our sample. Fortunately, our data suggest that for most women it also did not worsen the problem.

In the increasing prevalence group, the trajectory for sleep maintenance problems rose steeply beginning about 3.5 years before surgery, almost reaching the level of the high trajectory group about 1.5 years after surgery, and followed a similar pattern across the postmenopause (Figure 3). It is plausible that the steep increase before surgery was driven by a gynecological problem that led these women to seek medical attention and surgery, but precise data are lacking to explore this relevant clinical issue. In contrast, continuing sleep maintenance problems post-surgery in the other three trajectory groups raises questions regarding factors perpetuating their sleep disturbance or contributing to its prior occurrence.

Disturbed sleep is a common postoperative symptom after hysterectomy,²⁷ but only two other studies report long-term follow-up data regarding persistent sleep symptoms. In one study of 962 midlife women in Great Britain,⁵ only those women who had a hysterectomy (with or without bilateral oophorectomy) remained at increased risk for moderate sleep difficulty (post-hysterectomy follow-up duration not specified). In the Melbourne Women’s Midlife Health Project,²⁸ the 39 women who had a hysterectomy were significantly more likely to complain of trouble sleeping 5–6 years after surgery than those who did not have a hysterectomy. Our data showing persistence of sleep complaints post-surgery (median follow-up = 10 years) demonstrate that this problem is not transient and that persistence of sleep problems warrants ongoing monitoring for women who are surgically menopausal.

In women who are premenopausal or perimenopausal, a bilateral oophorectomy induces an abrupt cessation of ovarian hormone secretion, leading to a decline in estradiol and progesterone, and increased likelihood of menopausal symptoms.¹ Although risk status for post-surgical VMS may differ between women who undergo hysterectomy with ovarian conservation and those who have a bilateral oophorectomy,²⁹ most studies have combined both into a “surgical menopause” group and compared them with women who experience natural menopause. Our analysis eliminates this potential confound because it was restricted to women who had both a hysterectomy and bilateral oophorectomy. Also, neither serum FSH nor E2 was associated with post-surgery sleep maintenance problems.

Strengths of SWAN and our longitudinal analysis include a long follow-up in a large community sample of women unselected for sleep problems, which enhances the generalizability of our sample and decreases the risk of selection bias. The large sample and number of observations provided sufficient statistical power to control for the effects of a number of covariates. Group-based trajectory modeling accounted for heterogeneity within the sample, allowing us to identify groupings that are both statistically sound and scientifically valid.³⁰

Our longitudinal data and analytic approach, using pre-surgery sleep trajectories to predict sleep maintenance problems post-surgery, allowed us to examine patterns of sleep changes associated with surgical menopausal by separating the effects of the surgical intervention from pre-existing sleep problems. Although this analysis does not provide individual level results, these data do identify subgroups that share similar courses of sleep maintenance problems and demonstrate patterns and persistence of sleep complaints over time among midlife women. These findings suggest that pre-surgical sleep disturbances and concerns should be assessed to help guide women post-surgically. These findings also are consistent with our natural menopause trajectory analyses, which showed four similar trajectory patterns, including one group (15%) that had a rapidly increasing rise in sleep complaints, and pre-FMP sleep maintenance patterns that predicted post-FMP sleep patterns for each group.¹⁰ It is conjecture but since mean ages of natural FMP (52.6 years) and surgical menopause (51.2 years) are similar, maybe sleep maintenance problems are more chronological age related or perhaps health/lifestyle/environmental factors play more of a role than type of menopause.

Use of self-report measures of sleep problems may be considered a limitation because they carry a risk of information bias, but they are the clinical gold standard for clinical management of sleep problems constituting insomnia that warrants treatment. Our three questionnaire items, which constitute a valid and reliable self-report measure of sleep disturbance, were developed in a multi-racial/ethnic sample of postmenopausal women aged 50–79 years, ^31,32 have been administered repeatedly since the SWAN baseline assessment in 1996–1997. Women were not screened for other sleep disorders as contributing to their sleep disturbance, but the purpose of this report was to examine trajectories of sleep maintenance problems regardless of the cause. Another limitation is that medical records to confirm the women’s reports of or the reasons for their hysterectomy and bilateral oophorectomy were not available for about one-fifth of our participants. However, the accuracy and reproducibility of self-reported surgical menopause have been validated in research³³ and clinical practice.³⁴ Finally, we note that our increasing sleep complaint trajectory group included only 17 women. Although this is not a large sample in absolute numbers, the grouping is based on trajectories that consist of many repeated measurements for each woman (median number per woman is 6 in the increasing group, for a total of 106 observations). Despite the long and frequent follow-up and the adjustment for a number of covariates which increases the variability, these findings should be replicated with a larger sample of surgically menopausal women to establish that the trajectories patterns we report represent meaningful and clinically relevant groups.

Conclusions

Our findings provide longitudinal data regarding patterns and persistence of sleep complaints over time among midlife women that may help guide women post-surgically. Clinicians and their patients might find it reassuring that for the majority of women who are surgically menopausal, sleep complaints do not worsen post-surgically, yet they should be referred to a sleep specialist if symptoms worsen or if they continue to have a moderate or high level of persistent problems staying asleep.