Sleep Disturbances among Cancer Survivors

Abstract

Purpose:

We investigated sleep disturbances among cancer survivors compared to similarly aged women without cancer history.

Methods:

We identified 2,067 women with a history of cancer other than breast or non-melanoma skin cancer at enrollment in the Sister Study, a US-wide cohort of women with a family history of breast cancer. Cancer survivors were matched with up to 5 cancer-free women (N=9,717) on age at enrollment. An index age (for covariate classification) was defined as the age at cancer diagnosis for survivors and the same age for their matched comparators. Sleep disturbances included duration, sleep medication usage, insomnia symptoms, long sleep-latency onset (≥30 minutes to fall asleep), frequent night awakenings (waking ≥3/night, ≥ 3 times/week), frequent napping (≥ 3 times/week), and a composite outcome of ≥1sleep disturbance. Multivariable linear regression (effect estimate, 95% confidence interval (CI)) and logistic regression (odds ratio, OR, 95% CI) were used for continuous and dichotomous outcomes, respectively.

Results:

At enrollment, cancer survivors were on average 13.8 years (range=0, 62) from diagnosis. After adjustment for age at enrollment and depression, diabetes, hypertension, and menopausal status prior to the index age, sleep disturbances were generally not more common among cancer survivors compared to those without cancer. However, among cancer survivors, those >2 years from diagnosis were more likely to report ≥1 sleep disturbance (OR=1.44; 1.07, 1.93) compared to survivors 0-2 years from diagnosis.

Conclusion:

Addressing sleep disturbances may improve well-being for cancer survivors.

Introduction:

The National Cancer Institute (NCI) defines a cancer survivor as one who remains alive and continues to function from diagnosis until the end of life.¹ Approximately 75% of cancer survivors experience sleep disturbances.^2-5 Common sleep disturbances in survivors include short sleep duration, long sleep-onset latency (the amount of time it takes to fall asleep once in bed), and frequent night awakenings.^5,6 Consequences of disturbed sleep include daytime fatigue and depletion of energy and optimal cognition.⁷

The high prevalence of sleep disturbances in survivors may be attributed to the cancer itself, the emotional impact of the diagnosis, treatments, or side effects of treatments.^{5 8} Sleep disturbances may also be under-reported among survivors based on the prioritization of other, more pressing health concerns.^5,9-11 The prevalence of difficulty sleeping among newly diagnosed survivors ranges from 30-50%.⁸ Sleep disturbances typically decline following active treatment; however, some reports indicate disturbances are still 11-32% more prevalent in survivors one year post treatment than in the non-cancer population.^12,13 The prevalence of sleep disturbances in survivors 2-5 years post treatment is as high as 23-44%.^8,14-16

Understanding the long-term impact of cancer on sleep quality may lead to more timely interventions to improve sleep, reduce the likelihood of adverse conditions associated with sleep disturbances, and improve quality of life. The majority of existing literature on this topic focuses on survivors within 5 years of diagnosis; yet, sleep disturbances have the potential to persist decades after diagnosis.^13,17,18 Therefore, we evaluated multiple measures of disturbed sleep among survivors (0-61 years post-diagnosis) and women without cancer in the Sister Study cohort. The objective of this study is to estimate the burden of sleep disturbances among cancer survivors compared to their non-cancer counterparts. Additionally, among survivors we evaluated if burden of sleep disturbances differs based on time since cancer diagnosis.

Materials and Methods:

Study Population

The Sister Study is a nationwide cohort of 50,884 women in the U.S. including Puerto Rico (enrolled 2003-2009).¹⁹ Eligible women were aged 35-74 years, had a sister with breast cancer, but did not have a previous breast cancer diagnosis themselves. Women with a history of other types of cancer remained eligible for enrollment. The cohort was established to learn more about genetic and environmental risk factors for breast cancer.¹⁹ At enrollment, study participants completed computer-assisted telephone interviews (CATI) that collected information on medical history, lifestyle factors, and sleep disturbances. This study is a cross-sectional analysis of information reported on the enrollment survey for the Sister Study cohort.

The Sister Study is overseen by the Institutional Review Board (IRB) of the National Institutes of Health. All participants provided written informed consent. This analysis was approved by the IRB at the University of North Carolina at Chapel Hill.

Sample selection

From the cohort of 50,884 participants, exclusions were made for nightshift work (N=2,469); missing data on sleep disturbances (sleep duration [N=74], night awakenings [N=1,182], sleep-onset latency [N=60]), and missing menopausal status (N=18); and unknown timing of cancer at enrollment (N=25). Although personal breast cancer diagnosis was an exclusion criterion for cohort enrollment, 89 women were diagnosed with breast cancer prior to completion of enrollment activities or had a prevalent diagnosis of lobular carcinoma in situ and were subsequently excluded; leaving 46,967 records for analyses. Of these, 2,452 reported a cancer history (non-breast) at enrollment.

For each cancer survivor, we selected up to 5 comparison participants from the cohort with no history of cancer matched on age at enrollment into the Sister Study. An index age was defined to correspond to the matched case’s age at first cancer diagnosis, and the corresponding age in the matched non-cancer sample. The 2,452 prevalent cancer cases were matched with to up to five cancer-free women on age at enrollment (N=10,273). Among the 2,452 prevalent cancer cases, the majority were able to be matched on age at enrollment with five comparators (98.4%); 0.7% were matched with 4 comparators, 0.5% were matched with 3, 0.1% were matched with 2, and 0.2% were matched with 1.

Depression, diabetes, hypertension, and menopausal status were categorized at the index age for each participant to separate conditions that preceded or were concurrent with the index age versus those that emerged subsequently. From both the cancer and non-cancer samples, we excluded participants who reported the use of sleep medications (including an antihistamine or melatonin to aid with sleep) prior to the index age (N=385 cases, 556 non-cases), leaving a final analytic sample of 2,067 participants with a history of cancer and 9,717 without.

Exposure Assessment

The exposure in this study is prevalent cancer history self-reported at enrollment into the Sister Study cohort. To be in the exposed group, participants self-reported at least one (non-breast) cancer diagnosis prior to completion of cohort enrollment activities. Participants self-reported cancer type and age at diagnosis. Time since cancer diagnosis was calculated as the difference between age at study enrollment and age at diagnosis and categorized as 0 to <2 years; 2-4.9 years; 5-9.9 years; 10-19.9 years; and >20 years to 62 years. Participants with more than one previous cancer (N=165), were categorized according to their earliest cancer.

Outcome Assessment

The outcome of interest was sleep quality as reported on the Sister Study enrollment questionnaires. Participants reported their normal sleep patterns and their average number of hours and/or minutes of sleep. We categorized sleep duration as short sleep (<7 hours/night), recommended sleep (7-9 hours/night) and long sleep (>9 hours/night) based on recommendations from the National Sleep Foundation (NSF).²⁰ Participants were asked how long it takes them to fall asleep on average, with responses including <15 minutes, 15 minutes to half an hour, more than half an hour but less than one hour, and one hour or more. Sleep-onset latency is defined as the amount of time it takes to fall asleep from the time the lights go out.²¹ Most healthy adults have a sleep latency of 10-20 minutes.²² Sleep-onset latency was dichotomized as <30 minutes and ≥30 minutes to fall asleep. Participants were asked, “When asleep, how often do you wake for any reason?” and provided categorical responses including: every or most days/nights; three or four days/nights per week; one or two days/nights per week; one to three days/nights per month; less than once a month; and never. Additionally, they were asked, “On the days/nights when you wake, how many times do you usually wake?” which was answered numerically. This was used to construct dichotomous frequent night awakenings, defined as normally waking ≥3 times per night, ≥3 nights per week according to the NSF’s sleep quality recommendations and previous literature.^23,24 Insomnia symptoms included frequent night awakenings or long sleep-onset latency.^25,26 Participants were also asked to report “have you taken prescription or over the counter medication in the past six weeks to help you fall asleep or stay asleep?” and if they answered yes, they were asked to report the name of the medication and the age of initiation. Frequent napping was defined as napping ≥3 times per week.^25,26

Covariates

Using information from the enrollment survey, we defined the following covariates at the index age to assess self-reported medical history prior to cancer diagnosis or the corresponding age in the comparison group: depression, diabetes, hypertension, and menopausal status. Sleep medication use was defined as use between the index age and study enrollment.

Statistical analyses

Characteristics of study participants were described with mean values and standard deviations (SD) for continuous variables, and counts and percentages for categorical variables. Multivariable linear regression was used to estimate differences in mean sleep duration (beta coefficients) and 95% confidence intervals (CI) according to cancer history. Multivariable logistic regression was used to estimate odds ratios (OR) and 95% CI for dichotomous sleep disturbances including long sleep-onset latency and frequent night awakenings. As matching was not performed based on the outcome (sleep quality), unconditional logistic regression was used with multivariable adjustment for age at enrollment, depression, diabetes, hypertension, menopausal status, and sleep medication usage.²⁷ Polytomous logistic regression was used to calculate ORs and 95% CIs for three-category sleep duration: short sleep duration, long sleep duration, with recommended sleep duration as the reference. Model confounders were identified a priori using a directed acyclic graph²⁸ and included age at enrollment; depression, diabetes, hypertension, menopausal status at the index age; and sleep medication use after index age. We examined potential variation according to cancer type and survivorship intervals. We conducted two sensitivity analyses that alternately excluded: 1) participants who reported sleep medication use after the index age (as adjustment for sleep medication usage may not fully address potential confounding) and 2) women with a history of more than one cancer. Statistical analyses were completed using SAS 9.4 (Cary, NC) using Study data release 9.4.

Results

The analytic sample included 2,067 participants with a cancer history and 9,717 without (Table 1). The mean age at enrollment was 58.9 years (SD=8.7) among cancer survivors. The mean age at first cancer diagnosis was 45.0 years (SD=13.3), and the mean time from diagnosis to enrollment was 13.8 years (SD=11.6; range: <1 to 62 years). Survivors were more commonly non-Hispanic white (89.1% vs. 85.0%), were more often postmenopausal at the index age (45.3% vs. 37.6%), and more likely to have had diabetes (3.6% vs 2.2%) and hypertension (15.7% vs. 13.9%) prior to the index age (Table 1).

Table 1.

Characteristics of Sister Study Participants by Prevalent Cancer History

	History of Cancer (N=2,067)		No Cancer History (N=9,717)
	N	%	N	%
Age at enrollment (Mean, SD)	58.9	8.7	58.8	8.7
35 to <45	131	6.3	638	6.6
45 to <55	567	27.4	2682	27.6
55 to <65	764	37.0	3594	37.0
65-74	605	29.3	2803	28.9
Index age (Mean, SD)	45.0	13.3	44.6	13.3
Race/Ethnicity*
Non-Hispanic White	1841	89.1	8261	85.0
Non-Hispanic Black	99	4.8	772	7.9
Hispanic	58	2.8	432	4.5
Other	69	3.3	250	2.6
Depression prior to index age
Yes	222	10.7	972	10.0
No	1845	89.3	8745	90.0
Menopausal status at index age*
Premenopausal	1130	54.7	6064	62.4
Postmenopausal	937	45.3	3653	37.6
Diabetes prior to index age*
Yes	74	3.6	209	2.2
No	1993	96.4	9508	97.9
Hypertension prior to index age*
Yes	324	15.7	1347	13.9
No	1743	84.3	8370	86.1

^*There was a significant difference between the History of Cancer and No Cancer History using a chi-square test at an alpha level of 0.05

Among survivors, the most commonly reported cancers were melanoma (27.9%), cervical (19.4%), uterine (13.0%), thyroid (9.3%), ovarian (6.3%), colorectal (7.3%) and lymphoma (5.2%). Cancers with <5% representation were combined as “other cancers” (11.6%). Most survivors had one previous cancer (92.5%), 6.8% had two, and <1% had >2 previous cancers.

Sleep disturbances among cancer survivors compared to cancer-free women

Overall, 54.7% of cancer survivors and 56.8% of women without a cancer history reported experiencing ≥1 sleep disturbance (Table 2). Survivors were not more likely than their non-cancer counterparts to report ≥1sleep disturbance, overall (OR=0.91, 95% CI 0.83, 1.00) or by cancer type. Survivors were less likely to use sleep medications after the index age (9.6%) compared to their non-cancer counterparts (14.2%) (OR=0.66, 95% CI 0.56, 0.77). The prevalence of sleep medication usage varied from 6.1-13.5% across cancer types. Insomnia symptoms (difficulty falling or staying asleep) were similar between survivors compared to non-cancer counterparts overall (OR=1.05, 95% CI 0.94, 1.16) and by cancer type, with the exception of ovarian cancer. Among 131 women with a history of ovarian cancer, 35.9% reported insomnia symptoms. The OR for insomnia symptoms was 1.45 (95% CI 1.01, 2.09) for women with a history of ovarian cancer compared to the non-cancer sample (Table 2). We observed no meaningful variation in the individual components of insomnia symptoms (difficulty falling or staying asleep) according to cancer status or type (Supplemental Table 1). ORs for long sleep-onset latency and frequent night awakenings were elevated for participants with an ovarian cancer history (OR=1.34 and 1.17, respectively), but not statistically significant. Frequent napping did not vary between survivors and those without a cancer history (OR=0.95, 95% CI 0.82, 1.10)(Table 2). When we alternately compared ovarian cancer survivors to survivors of melanoma; cervical; uterine/endometrial; thyroid; colon/rectum; or lymphoma cancers, the OR for insomnia symptoms among ovarian cancer survivors ranged from 1.26-1.62 but was not statistically significant (Supplemental Table 3).

Table 2.

The association between cancer history and sleep disturbances overall and by cancer type (N=11,784)

	Total	At least 1 Sleep Disturbancea			Sleep Medication Use			Insomnia Symptomsd			Frequent Nappinge
	N	N (%)	OR b	95% CIb	N(%)	OR b	95% CIb	N(%)	OR c	95% CIc	N (%)	ORc	95% CIc
No History of Cancer	9,717	5520 (56.8)	Ref		1378 (14.2)	Ref		2746 (28.3)	Ref		1260 (13.0)	Ref
History of Cancer	2,067	1130 (54.7)	0.91	0.83, 1.00	198 (9.6)	0.66	0.56, 0.77	598 (28.9)	1.05	0.94, 1.16	259 (12.5)	0.95	0.82, 1.10
Cancer Type
Melanoma	577	305 (52.9)	0.85	0.72, 1.01	48 (8.3)	0.56	0.41, 0.75	166 (28.8)	1.05	0.87, 1.27	65 (11.3)	0.87	0.66, 1.13
Cervical	400	221 (55.3)	0.98	0.80, 1.20	54 (13.5)	0.89	0.66, 1.20	115 (28.8)	1.09	0.87, 1.36	43 (10.8)	0.96	0.69, 1.33
Uterine/endometrial	268	155 (57.8)	0.99	0.78, 1.28	20 (7.5)	0.52	0.33, 0.83	72 (26.9)	0.92	0.70, 1.21	45 (16.8)	1.20	0.86, 1.67
Thyroid	193	94 (48.7)	0.73	0.54, 0.97	15 (7.8)	0.50	0.29, 0.85	49 (25.4)	0.90	0.65, 1.25	22 (11.4)	0.965	0.60, 1.49
Colon/rectum	150	78 (52.0)	0.78	0.56, 1.08	14 (9.3)	0.67	0.39, 1.17	40 (26.7)	0.89	0.62, 1.29	19 (12.7)	0.82	0.50, 1.35
Ovarian Cancers	131	74 (56.5)	1.01	0.71, 1.44	8 (6.1)	0.47	0.23, 0.96	47 (35.9)	1.45	1.01, 2.09	16 (12.2)	0.94	0.55, 1.62
Lymphoma	108	64 (59.3)	1.12	0.76, 1.65	11 (10.2)	0.73	0.39, 1.37	33 (30.6)	1.14	0.76, 1.73	19 (17.6)	1.40	0.84, 2.32
Other Cancers	240	139 (57.9)	0.97	0.74, 1.26	28 (11.7)	0.84	0.57, 1.26	76 (31.7)	1.10	0.83, 1.45	30 (12.5)	0.78	0.52, 1.15

^aAny sleep disturbance includes at least one of the following outcomes: long sleep-onset latency, frequent night awakenings, long sleep duration, low sleep duration, frequent napping, or sleep medication use after index age

^bModels are adjusted for age at enrollment, depression at the index age, diabetes at the index age, hypertension at the index age, and menopausal status at the index age

^cModels are adjusted for age at enrollment, depression at the index age, diabetes at the index age, hypertension at the index age, menopausal status at the index age, and sleep medication at index age

^dDifficulty falling (≥30 minutes) or staying asleep (waking during the night ≥3 times a night, ≥3 times a week)

^eNapping ≥ 3 times/week

Average, short, and long sleep duration did not vary according to cancer status overall (Table 3). The mean sleep duration was 7.1 hours in both groups, and most women reported sleep duration in the recommended range (7-9 hours; 70.4% cancer, 70.7% non-cancer). Although the OR was elevated (2.47; 95% CI 1.06, 5.76) for long sleep duration among women with uterine/endometrial cancer, this estimate was based on only 6 survivors.

Table 3.

The association between cancer history and sleep duration overall and by cancer type (N=11,784)

	Continuous Sleep Duration			Categorical Sleep Duration
				Recommended 7-9 hours	Shorta < 7 hours			Longa > 9 hours
	N	Mean (SD)	Regression Parameter (95% Cl)b	N (%)	N (%)	ORb	95% CIb	N (%)	ORb	95% CIb
No History of Cancer	9,717	7.09 (1.11)	Ref	6867 (70.7)	2753 (28.3)	Ref		99 (1.0)	Ref
History of Cancer	2,067	7.11 (1.11)	0.02 (−0.03, 0.08)	1456 (70.4)	583 (28.2)	1.00	0.90, 1.11	28 (1.4)	1.42	0.93, 2.17
Cancer Type
No History of Cancer	9,717	7.09 (1.09)	Ref	7046 (70.2)	2892 (28.8)	Ref		98 (1.0)	Ref
Melanoma	577	7.14 (1.04)	0.06 (−.0.3, 0.15)	424 (73.5)	144 (25.0)	0.85	0.70, 1.03	9 (1.6)	1.60	0.80, 3.19
Cervical	400	7.09 (1.14)	0.00 (−0.11, 0.12)	272 (68.0)	125 (31.3)	1.15	0.92, 1.43	3 (0.8)	NA
Uterine/endometrial	268	7.03 (1.13)	−0.03 (−0.17, 0.10)	180 (67.2)	82 (30.6)	1.15	0.88, 1.50	6 (2.2)	2.47	1.06, 5.76
Thyroid	193	7.12 (0.99)	0.01 (−0.15, 0.17)	144 (74.6)	48 (24.9)	0.83	0.60, 1.16	1 (0.5)	NA
Colon/rectum	150	6.97 (1.33)	−0.12 (−0.30, 0.06)	96 (64.0)	50 (33.3)	1.31	0.93, 1.86	3 (2.8)	NA
Ovarian Cancers	131	7.15 (1.04)	0.05 (0.−14, 0.24)	91 (69.5)	39 (29.8)	1.08	0.74, 1.58	1 (0.8)	NA
Lymphoma	108	7.2 (1.15)	0.13 (−0.09, 0.34)	79 (73.2)	27 (25.0)	0.88	0.56, 1.36	2 (1.9)	NA
Other Cancers	240	7.13 (1.17)	0.04 (−0.10, 0.18)	170 (70.8)	68 (28.3)	0.99	0.74, 1.32	2 (0.8)	NA

^aLong and short sleep duration are compared to recommended sleep duration in polytomous logistic regression

^bModels are adjusted for age at enrollment, depression at the index age, diabetes at the index age, hypertension at the index age, menopausal status at the index age, and sleep medication use at index age

Sleep disturbances among cancer survivors

There was a wide range of time since cancer diagnosis (<1 to 62 years). Among survivors, odds of ≥1 sleep disturbance were higher among those >2 years post-diagnosis (OR=1.44; CI 1.07-1.93) compared to those 0-2 from diagnosis. The odds of ≥1 sleep disturbance were elevated for all survivorship intervals (5-9.9, 10-19.9, and 20+ years) when compared to most recent survivors (0-2 years post-diagnosis) (Table 4). Odds ratios were elevated but not statistically significant for the use of sleep medication use (1.40), insomnia symptoms (1.28), long sleep-onset latency (1.16), and frequent night awakenings (1.28) among those >2 years post-diagnosis compared to those 0-2 years post-diagnosis (Table 4 and Supplemental Table 2). There was no difference in frequent napping between those >2 years post-diagnosis compared to those 0-2 years post-diagnosis(Table 4).

Table 4.

The association between time since diagnosis and sleep disturbances among cancer survivors (N=2,067)

	Total	At least 1 Sleep Disturbancea			Sleep Medication Use			Insomnia Symptomsd			Frequent Nappinge
	N	N(%)	OR b	95% CIb	N(%)	OR b	95% CIb	N(%)	OR c	95% CIc	N (%)	OR c	95% CIc
0-2 years	238	120 (50.4)	Ref		15 (6.3)	Ref		66 (27.7)	Ref		36 (15.1)	Ref
>2 years	1,829	1010 (55.2)	1.44	1.07, 1.93	183 (10.0)	1.40	0.79, 2.49	532 (29.1)	1.28	0.93, 1.77	223 (12.2)	0.99	0.65, 1.50
Time since Diagnosis
0 to <2 yrs	238	120 (50.4)	Ref		15 (6.3)	Ref		66 (27.7)	Ref		36 (15.1)	Ref
2-4.9 yrs	352	194 (55.1)	1.34	0.96, 1.88	21 (6.0)	0.93	0.47, 1.86	117 (33.2)	1.42	0.98, 2.05	42 (11.9)	0.89	0.54, 1.45
5-9.9 yrs	393	210 (53.4)	1.36	0.97, 1.91	35 (8.9)	1.45	0.76, 2.76	110 (28.0)	1.19	0.82, 1.73	45 (11.5)	0.90	0.55, 1.47
10-19.9 yrs	520	282 (54.2)	1.52	1.09, 2.13	59 (11.4)	1.1.82	0.97, 3.40	132 (22.1)	1.12	0.77, 1.62	66 (12.7)	1.18	0.73, 1.93
>20 years to 62 yrs	564	324 (57.5)	1.86	1.28, 2.69	68 (12.1)	1.89	0.97, 3.71	173 (30.7)	1.57	1.04, 2.35	70 (12.4)	1.20	0.69, 2.07

^aAny sleep disturbance includes at least one of the following outcomes: long sleep-onset latency, frequent night awakenings, long sleep duration, low sleep duration, or sleep medication use after index age

^bModels are adjusted for age at enrollment, depression at the index age, diabetes at the index age, hypertension at the index age, and menopausal status at the index age

^cModels are adjusted for age at enrollment, depression at the index age, diabetes at the index age, hypertension at the index age, menopausal status at the index age, and sleep medication at index age

^dDifficulty falling (≥30 minutes) or staying asleep (waking during the night ≥3 times a night, ≥3 times a week)

^eNapping ≥ 3 times/week

Among survivors, those furthest from cancer diagnosis (>20 years) appeared more likely to report <7 hours of sleep (30.7%) compared to those within 2 years of diagnosis (25.6%, OR=1.47 95% CI 0.97, 2.22), but this estimate was not statistically significant (Table 5).

Table 5.

The association between time since diagnosis and sleep duration among cancer survivors (N=2,067)

	Continuous Sleep Duration			Categorical Sleep Duration
	Total			Recommended 7-9 hours	Shorta < 7 hours			Longa > 9 hours
	N	Mean (SD)	Regression Parameter (95% Cl)b	N (%)	N (%)	ORb	95% CIb	N (%)	ORb	95% CIb
0-2 years	238	7.19 (1.18)	Ref	173 (72.7)	61 (25.6)	Ref		4 (1.7)	Ref
>2 years	1,829	7.10 (1.10)	−0.12 (−0.28, 0.04)	1283 (70.2)	522 (28.5)	1.20	0.86, 1.66	24 (1.3)	0.97	0.30, 3.14
Time since Diagnosis
0 to <2 yrs	238	7.19 (1.18)	Ref	173 (72.7)	61 (25.6)	Ref		4 (1.7)	Ref
2-4.9 yrs	352	7.08 (1.12)	−0.12 (−0.30, 0.06)	247 (70.2)	103 (29.3)	1.18	0.81, 1.72	2 (0.6)	NA
5-9.9 yrs	393	7.17 (1.01)	−0.06 (−0.24, 0.13)	288 (73.3)	99 (25.2)	1.00	0.68, 1.47	6 (1.5)	1.17	0.31, 4.48
10-19.9 yrs	520	7.07 (1.11)	−0.16 (−0.35, 0.03)	365 (70.2)	147 (28.3)	1.22	0.84, 1.78	8 (1.5)	1.38	0.36, 5.33
>20 years to 62 yrs	564	7.09 (1.14)	−0.19 (−0.39, 0.01)	382 (67.9)	173 (30.7)	1.47	0.97, 2.22	8 (1.4)	1.31	027, 6.24

^aLong and short sleep duration are compared to recommended sleep duration in polytomous logistic regression

^bModels are adjusted for age at enrollment, depression at the index age, diabetes at the index age, hypertension at the index age, menopausal status at the index age, and sleep medication use after index age

In sensitivity analyses excluding survivors who used sleep medication and those with more than one previous cancer, our interpretations remained unchanged from the from the main analysis (data not shown).

Discussion

Recognizing the importance of sleep for overall health, our analysis examined sleep disturbances among survivors compared to their non-cancer counterparts. We included common cancer types among women often less frequently represented in the survivorship and sleep literature^3,6,29,30. Few differences were observed in sleep disturbances in survivors compared to non-cancer counterparts. Among survivors, sleep disturbances were more common >2 years post-diagnosis. This pattern supports our hypothesis that sleep disturbances persist long after active treatment.

The prevalence of ≥1 sleep disturbance in this study (56.4%) is consistent with the 50.9% reported by Strollo et al.²⁹ among 1,903 9-year cancer survivors (diagnosed with breast, prostate, CRC, bladder, uterine, melanoma, non-Hodgkin lymphoma, kidney, lung, and ovarian cancers). A meta-analysis looking at sleep disturbance among survivors in active treatment and in the early survivorship interval (>3 months post treatment) reported a pooled prevalence of sleep disturbances of 60.7% with a range among individual studies of 15.3%³¹- 99.8%.^32,33 In this study we defined short sleep duration as<7 hours of sleep, as recommended by the NSF²³, however it should also be noted that in a recent meta-analysis on sleep duration as a risk factor for incident cancer, the definition of short sleep duration in the literature ranges from <5-<7 hours per night.³⁴

In our analysis, the mean sleep duration in both survivors and their non-cancer counterparts was 7.1 hours of sleep/night. Our findings are similar to the 7-hour median sleep duration in >2 year survivors of breast, colorectal, prostate, and other cancers, and non-cancer spouses and friends, reported by Forsythe et al.¹³ in the National Health and Nutritional Examination Survey (NHANES), Slade et al. also reported a sleep duration of 7.1 hours/night for ≥6-year survivors.³⁵ When compared to individuals with no history of cancer, Slade et al. also did not identify meaningful differences in sleep outcomes by cancer type, except for survivors of gynecologic cancers.³⁵ In our analysis, women with ovarian cancer were more likely to report insomnia symptoms than women without cancer; however, we did not observe this pattern for other gynecologic cancers, such as endometrial cancer.

In our sample, survivors were unexpectedly less likely to use sleep medications compared to their non-cancer counterparts (9.6% vs. 14.2%). When we restricted to survivors, sleep medication use increased from 6.3% to 12.1% with time since diagnosis. The prevalence of sleep medication use among 10-19.9 year survivors was 11.4%, substantially lower than the 28% reported by Strollo et al. at 9 years after diagnosis.²⁹ However, the Strollo et al. study population included male participants (37.3%), and the mean age was slightly older (64.5 years).²⁹ Our finding that survivors did not report more sleep disturbances than their non-cancer peers contrasts with some previous studies. Much of the existing literature focuses on breast cancer survivors^12,31 or survivors within the first 5-years of diagnosis.^13,17,18 Our sample excluded breast cancer survivors and included survivors a mean of 13.8 years from diagnosis.

Our analysis relied on self-reported sleep data, which may underreport true sleep duration compared to measured sleep duration.³⁶ Polysomnography is the gold standard measurement for sleep duration but is infeasible to implement across a nationwide, community-dwelling sample.³⁷ We assessed the sleep disturbances available in our data; this study did not collect information on other sleep disturbances such as early morning awakenings. Results from this study may not be generalizable to male breast cancer survivors who were not included in the Sister Study cohort. This study included many common cancers, such as thyroid, colorectal, ovarian, and lymphoma. Our findings may not be applicable to survivors of other cancer types, such as lung, bladder, and kidney, which were not represented in our sample. To participate at enrollment, survivors needed to survive until Sister Study enrollment, potentially selecting for healthier cancer survivors, a phenomenon known as healthy survivor bias³⁸. Survivors of poorer health may have been less likely to enroll in the Sister Study and to have experienced more adverse sleep outcomes.

We lacked data regarding cancer staging, treatment, and recurrence. More advanced cancers may receive more aggressive treatment which could have a larger effect on sleep disturbances. Most commonly, but not exclusively, active treatment for cancer is administered in the first two years post-diagnosis.^{39 40} Our analyses did not account for cancer treatments received for primary disease or recurrence or progression, as this information was not collected. Despite these limitations, our analysis contributes valuable information on the prevalence of sleep disturbances among a wide range of cancer types and survivorship periods. This contribution expands the evidence base from previous research that has largely focused on breast cancer survivors.^{6,29,30,41-43}

Previous studies in cancer survivors have shown sleep disturbances to be positively associated with depression^4,44 and fatigue.^45,46 Randomized control trial-tested interventions, such as cognitive behavioral therapy⁴⁷, can effectively modify sleep duration in survivors to improve fatigue, depression, and anxiety ^47-49 and quality of life.^47,48 Sleep is a modifiable behavior and, therefore, a potential intervention point to improve the quality of life, physical andmental health of cancer survivors.

Conclusion

In this study, sleep disturbances were common in cancer survivors and women without cancer. Among cancer survivors, we saw an increase in sleep disturbances >2 years post-diagnosis. Cancer survivors may benefit from interventions to enhance or maintain sleep health to improve quality of life and overall health.

Highlights.

• Among cancer survivors, sleep disturbances were more common >2 years post diagnosis compared to 0-2 years post diagnosis

• Cancer survivors were no more likely to report sleep disturbances than non-cancer counterparts

• Cancer survivors may experience sleep disturbances long after active treatment