Original Article

Cancer Treatments and Their Side Effects Are Associated With

Aggravation of Insomnia: Results of a Longitudinal Study
e Savard, PhD1,2,3; Hans Ivers, PhD1,2,3; Marie-He
Jose lène Savard, PhD2,3; and Charles M. Morin, PhD1

BACKGROUND: Insomnia affects between 30% to 60% of patients with cancer but to the authors’ knowledge little is known regard-
ing factors associated with its development. It has been postulated that adjuvant cancer treatments and their side effects could trig-
ger sleep disturbances in this population but empirical evidence is lacking. The goal of the current study was to assess, separately in
patients with breast and prostate cancer, the effect of adjuvant treatments on the evolution of insomnia symptoms and the mediating
role of somatic symptoms. METHODS: As part of a population-based epidemiological study, patients with breast cancer (465
patients) and prostate cancer (263 patients) completed at baseline (perioperative period) and 2 months, 6 months, 10 months, 14
months, and 18 months later the Insomnia Severity Index (ISI) and a questionnaire assessing various somatic symptoms. RESULTS: In
patients with breast cancer, radiotherapy (overall effect) and chemotherapy (at 2 months), but not hormone therapy, were associated
with increased insomnia severity, whereas androgen deprivation therapy was related to increased insomnia in patients with prostate
cancer. In patients with breast cancer, the effect of chemotherapy and radiotherapy on insomnia was found to be significantly medi-
ated by a variety of somatic symptoms, whereas night sweats had a particularly marked mediating role for hormone therapy, both
in patients with breast and prostate cancer. CONCLUSIONS: The findings of the current study indicate that cancer treatments
and their side effects contribute to the aggravation of insomnia symptoms. Side effects of cancer treatments should be monitored
more closely and managed as effectively as possible to prevent the occurrence or aggravation of insomnia. Cancer 2015;000:000-
000. VC 2015 American Cancer Society.

KEYWORDS: cancer, insomnia, sleep, cancer treatment, longitudinal, side effects.

INTRODUCTION
Insomnia is a common problem in patients with cancer. Although insomnia is highly prevalent before any treatment is
administered,1 longitudinal data suggest that there are factors occurring during the cancer care trajectory that trigger its de-
velopment. A recent population-based epidemiological study revealed that insomnia affected up to 59% of patients during
the perioperative period.2 However, 14.4% of patients had a first incidence and 19.5% had a recurrence of insomnia
symptoms during the course of the study, for a total incidence rate of 31.8%.3
Although causes of insomnia comorbid with cancer are most likely multifaceted,4,5 it has been postulated that ad-
juvant treatments play a significant role.6,7 To the best of our knowledge, only a few longitudinal studies to date have
focused on this issue. Although lack of an association has also been noted,8 receiving chemotherapy has been found to
predict an increase in sleep disturbance during the cancer care trajectory.1,9 Two studies of patients with breast cancer
found that radiotherapy was associated with a temporary augmentation of sleep difficulties10,11 and one study revealed
that intensity-modulated radiotherapy for the treatment of nasopharyngeal cancer was associated with an increased
rate of clinically significant sleep impairments.12 In patients with prostate cancer, the introduction of androgen depri-
vation therapy was found to be related to increased insomnia symptoms.13 Hence, it would appear that all 3 of these
adjuvant treatments can increase the risk of insomnia, although studies disentangling their respective effects are
lacking.

Corresponding author: Jos ee Savard, PhD, Centre de recherche du CHU de Qu

ebec, 11 C^
ote du Palais, Qu
ebec, Qu
ebec, Canada, G1R 2J6; Fax: (418) 691-2971;
josee.savard@psy.ulaval.ca
1
School of Psychology, Laval University, Quebec City, Quebec, Canada; 2Centre de recherche du CHU de Qu
ebec, Quebec City, Quebec, Canada; 3Cancer Research
Center, Laval University, Quebec City, Quebec, Canada

We wish to acknowledge the important contribution of Julie Villa, Aude Caplette-Gingras, Marie-Solange Bernatchez, Val erie Tremblay, Lucie Casault, Caroline
Desautels, Genevi eve Dumont, Dave Flanagan, Nathalie Gagnon, Catherine Gonthier, Geneviève Laurent, Marie-Eve Le May, Julie Maheux, Marie-Esther Paradis,
Sylvie Perron, Julie Roy, Sophie Ruel, Elaine Th
eriault, Claudia Trudel-Fitzgerald, and Maude Villeneuve who were involved in the recruitment and assessment of
the participants or the data entry and the study coordination, as well as the participants who volunteered their time for this study.

DOI: 10.1002/cncr.29244, Received: July 3, 2014; Revised: October 16, 2014; Accepted: December 17, 2014, Published online Month 00, 2015 in Wiley Online
Library (wileyonlinelibrary.com)

Cancer Month 00, 2015 1

Original Article
There are several plausible mechanisms through
which cancer treatments could lead to sleep disturbances.
Although other mechanisms are possible (eg, anticipatory
anxiety, behavioral changes such as day napping that
impair circadian rhythms, or chemotherapy-induced
inflammation),4,14-16 adjuvant treatments can also induce
sleep disturbances through some of their negative side
effects. Nocturnal hot flashes (due to chemotherapy and
hormone therapy17), urinary incontinence (eg, due to
radiotherapy to the urogenital area), and gastrointestinal
symptoms (eg, chemotherapy-induced nausea) as well as
pain (eg, associated with the use of aromatase inhibitors in
both men and women16) are all very likely to negatively
affect sleep quality.6,7,18 However, to our knowledge, very
little research data are available to help delineate their role
in precipitating insomnia comorbid with cancer.
The goals of this secondary analysis of a previous ep-
idemiological study by Savard et al2,3 were to assess, sepa-
rately in individuals with breast and prostate cancer, 1) the
effect of adjuvant treatments on the evolution of insomnia
symptoms and 2) to what extent somatic symptoms medi-
ated the relationship between adjuvant treatments and
insomnia symptoms. It was hypothesized that all adjuvant
treatments would be associated with increased insomnia
through the mediating effect of somatic symptoms.
MATERIALS AND METHODS
Participants
Recruitment
Inclusion criteria for the main study were: 1) a first diagnosis
of nonmetastatic cancer (a few patients with American Joint
Committee on Cancer (AJCC) TNM staging system stage
IV prostate cancer were included but they had no distant me-
tastases); 2) patients were scheduled to undergo curative sur-
gery; 3) patients were aged 18 to 80 years; and 4) patients
were able to read and understand French. Exclusion criteria
were: 1) the administration of neoadjuvant cancer treatment;
2) upcoming surgery was part of brachytherapy for prostate
cancer; 3) presence of severe cognitive impairments (eg,
Alzheimer disease) or severe psychiatric disorder (eg, psycho-
sis); 4) presence of a sleep disorder other than insomnia (eg,
sleep apnea); and 5) presence of severe visual, hearing, or lan-
guage defects.
Participants were recruited at the CHU de Quebec
from January 2005 to May 2007. Patients meeting the ini-
tial inclusion criteria were approached by a research assist-
ant on the day of their preoperative visit and eligible
patients were invited to provide their written consent. The
study was approved by the ethics review board of the study
institution. Of the 3196 patients approached at the clinics,
2 Cancer
1519 were excluded and 715 refused to participate in the
study, thus giving a participation rate for the larger study
of 57.4% (962 patients; for a detailed flowchart see the
study by Savard et al3). For the purpose of the current
analysis, only the 2 larger subgroups of patients could be
included (ie, patients with breast cancer [465 patients]
and prostate cancer [263 patients]) (Table 1).
Study Design
This study used a prospective longitudinal design com-
prising 6 time points: baseline (perioperative phase; T1)
and 2 months (T2), 6 months (T3), 10 months (T4), 14
months (T5), and 18 months (T6). Overall, 72.7% of the
patients with breast cancer (338 patients) and 69.6% of
patients with prostate cancer (183 patients) completed all
6 assessments (Mean (M) 5 5.2).
Measures
Demographics, health behaviors, and cancer
characteristics
Demographics, medical comorbidity, medication use,
and health behaviors (smoking status, alcohol and caffeine
use, and physical activity) were collected using a question-
naire. Cancer-related data (eg, cancer site and stage and
adjuvant treatments received) were obtained from the
patient’s medical record.
Insomnia Severity Index
The Insomnia Severity Index (ISI)19 is a 7-item question-
naire evaluating insomnia severity (eg, difficulties falling
asleep). Each item is rated using a 5-point Likert scale
ranging from 0 (not at all) to 4 (very much), for a total
score ranging from 0 to 28. The French-Canadian version
of the ISI was validated in the context of cancer.20 A cutoff
score of 8 was found to indicate the presence of clinical
levels of insomnia.20
Physical Symptoms Questionnaire
The Physical Symptoms Questionnaire is adapted from
the Memorial Symptom Assessment Scale21 and assesses
18 somatic symptoms commonly associated with cancer
and its treatment. Only the frequency scores (from 0 indi-
cating never to 4 indicating almost constantly) were ana-
lyzed. Symptoms had to meet the following criteria to be
investigated as possible mediators: 1) be reported by
20% of the sample; 2) have a mean frequency >2 on the
scale of 0 to 4; 3) 1 cancer treatment was found to be sig-
nificantly associated with the symptom; and 4) the symp-
tom exhibited at least a small association (beta >.10)22
with ISI scores. Seven symptoms were retained as possible
mediators: headache (breast cancer), dyspnea (both),
Month 00, 2015
 Cancer Treatments and Insomnia/Savard et al

TABLE 1. Participant Characteristics by Cancer Site

Breast Cancer n 5 465 Prostate Cancer n 5 263

Variable Mean (SD) No. (%) Mean (SD) No. (%)

Age, y 54.9 (9.6) 61.7 (6.4)

Sex (female) 465 (100) 0 (0.0)
Marital status
Married/cohabitating 286 (62.0) 205 (78.5)
Single 54 (11.7) 13 (5.0)
Separated/divorced/widowed 121 (26.3) 43 (16.5)
Education
Primary diploma or less 24 (5.2) 25 (9.7)
High school diploma 192 (41.7) 90 (35.0)
College degree 120 (26.1) 65 (25.3)
University degree 124 (27.0) 77 (30.0)
Annual family income (in Canadian dollars)
<$20,000 72 (18.7) 24 (10.4)
$20,001-$40,000 114 (29.6) 75 (32.5)
$40,001-$60,000 73 (19.0) 59 (25.5)
$60,001-$80,000 61 (15.8) 33 (14.3)
$80,001 65 (16.9) 40 (17.3)
Current occupation
Working (full/part time) 201 (43.6) 96 (36.9)
Family work 32 (6.9) 1 (0.4)
Sick leave 79 (17.1) 18 (6.9)
Retired 138 (29.9) 142 (54.6)
Unemployed 11 (2.4) 3 (1.2)
Time since initial diagnosis, mo 1.6 (0.9) 3.9 (2.7)
AJCC Cancer stage of disease
0 44 (9.5) 0 (0.0)
I 198 (42.6) 2 (0.8)
II 148 (31.8) 160 (60.8)
III 57 (12.3) 83 (31.6)
IVa 0 (0.0) 18 (6.8)
Unspecified 18 (3.9) 0 (0.0)
Cancer treatments received during the studyb
Radiotherapy 367 (85.6) 10 (4.2)
Chemotherapy 232 (54.1) 2 (0.8)
Hormone therapy 309 (72.0) 25 (10.4)
Other (eg, trastuzumab) 37 (8.6) 0 (0.0)
Type of chemotherapyc
AC 85 (36.6)
FEC plus docetaxel 36 (15.5)
FEC 35 (15.1)
AC plus paclitaxel plus gemcitabine 14 (6.0)
AC plus docetaxel 13 (5.6)
AC plus paclitaxel 9 (3.9)
Other/not specified 40 (17.2) 2 (100.0)
Type of hormone therapyc,d
Anastrozole 148 (47.9) 0 (0.0)
Tamoxifen 146 (47.2) 4 (16.0)
Letrozole 28 (9.1) 0 (0.0)
Exemestane 9 (2.9) 0 (0.0)
Bicalutamide 0 (0.0) 19 (76.0)
Goserelin 3 (1.0) 16 (64.0)
Leuprolide 0 (0.0) 3 (12.0)
Insomnia Severity Index
Baseline 10.3 (5.9) 305 (66.2)e 6.4 (5.6) 96 (36.9) e

2 mo 9.7 (5.6) 264 (62.6) 7.0 (5.2) 89 (38.7)

6 mo 8.5 (5.8) 203 (51.0) 5.9 (5.3) 66 (29.0)
10 mo 7.7 (5.4) 166 (43.1) 5.4 (4.9) 57 (26.4)
14 mo 7.3 (5.4) 149 (40.6) 5.8 (5.1) 64 (30.9)
18 mo 7.0 (5.1) 137 (38.8) 5.3 (4.5) 52 (25.2)

Abbreviations: AC, cyclophosphamide and doxorubicin; AJCC, American Joint Committee on Cancer; FEC, 5-fluorouracil, epirubicin, and cyclophosphamide;
SD, standard deviation.
a
Patients with stage IV cancer did not have distant metastases.
b
Some patients did not receive any adjuvant treatment whereas others received >1 treatment during the study.
c
The percentages computed over the number of users.
d
Patients could have received >1 treatment.
e
The percentage of patients displaying a clinical level of insomnia (Insomnia Severity Index 8).
Original Article
nausea (breast cancer), digestive problems (breast cancer),
night sweats (both), pain (both), and 2 items related to
urination (frequent and involuntary; r 5 0.48), which
were combined together (both).
Procedure
At each time point, participants were given a battery of
self-report scales including the ISI and the Physical Symp-
toms Questionnaire that they had to complete within the
next week and mail back. More details on the procedure
are available elsewhere.2
Statistical Analyses
Descriptive and inferential statistics were conducted using
SAS statistical software (version 9.3; SAS Institute Inc,
Cary, NC).23 The alpha level was fixed at 5% (2-tailed) for
all inferential tests. All participants with at least 1 available
time point on the main outcome (ISI) were included in the
analyses (728 participants) and no data imputation was per-
formed because the chosen statistical models are robust to
missing data. Several variables were investigated for a possi-
ble inclusion in statistical models as covariates (eg, demo-
graphics, cancer site and stage, medications, comorbidity,
and health behaviors) but none of them was retained
because none had at least a minimal correlation with the de-
pendent variable (ISI) (correlation coefficient [r] 5 60.30)
(a medium effect size [r 0.30] rather than statistical signif-
icance was deemed to be a more appropriate criterion
because of the large sample in the current study, which was
likely to yield numerous significant but fortuitous associa-
tions).24 In both groups, patients did not receive any adju-
vant treatment at baseline, and therefore these data could
not be used as dependent variables. In patients with breast
cancer, T5 and T6 could not be used either because none
of the patients received radiotherapy at those times, thus
leaving T2 to T4 for these analyses. Because <10 of the
patients with prostate cancer received radiotherapy and
chemotherapy at any time point of the study, only the use
of hormone therapy was analyzed (from T3 to T6) (Fig. 1).
Three binary independent variables, specifying
whether each participant was exposed to chemotherapy,
radiotherapy, and hormone therapy at each time point,
were created. Linear mixed models using a factorial
design (group [chemotherapy, radiotherapy, and hor-
mone therapy] 3 time) were performed to test the signif-
icance of temporal changes, while adjusting means for
missing data and taking into account individual baseline
(random effect). Main treatment and time effects and
treatment 3 time interactions were systematically tested
and interactions were decomposed using simple effects.
4 Cancer
Figure 1. Percentage of patients with (a) breast cancer and
(b) prostate cancer exposed to each treatment since the last
assessment at each time point are shown.
The best-fitting covariance matrix for each model was
selected based on the Bayesian information criterion.
Empirical “sandwich” estimators of standard errors for
fixed effects were used.
Longitudinal mediation analyses were computed
using an autoregressive mediation model,25 which permit-
ted the examination of both concurrent (associations of
variables all assessed at the same time point) and longitu-
dinal (effect of each treatment at 1 time point on the me-
diator and insomnia severity assessed at the subsequent
time point) mediation. For longitudinal relations, only
lag-1 associations (between consecutive time points) were
investigated, mainly because we were interested in investi-
gating the proximal factors triggering an aggravation of
insomnia. Paths tested are illustrated in Figure 2.
Three linear mixed models were performed for each
mediator (symptom) to estimate: 1) the total relationship
(treatment!insomnia); 2) the alpha relationship (treat-
ment!mediator); and 3) the beta (mediator!insomnia)
and direct (treatment!insomnia, controlling for the me-
diator) relationship. The overall mediation effect was esti-
mated as the product of the alpha and beta regression
coefficients and its standard error was computed accord-
ing to the procedure of MacKinnon et al (Z-prime for-
mula).26 The ratio of the mediation effect (alpha-beta
product) on its standard error yielded a Z-prime statistic,
which was compared against the appendix in MacKinnon
et al26 (ie, null hypothesis indicates no mediating effect;
N 5 500) to determine the significance of the mediating
Month 00, 2015

effect. The percentage of the total effect explained by the
mediated (indirect) effect was computed as the ratio of the
mediation standard coefficient on the standard coefficient
for the total treatment effect. A mediation was considered
a total mediation when only the mediation (indirect)
effect was statistically significant, whereas a partial media-
tion was determined when both mediation and directs
effects were significant.
RESULTS
Objective 1: Effect of Adjuvant Treatments on
the Evolution of Insomnia Severity
Breast cancer
In participants with breast cancer, the results of a normal
linear mixed model conducted on ISI scores revealed a sig-
Figure 2. Summary of the relationships tested with the longi-
tudinal autoregressive mediational model of the lagged (P)
and current (C) contributions of chemotherapy (C), radio-
therapy (R), and hormone therapy (H) to the Insomnia Sever-
ity Index, mediated by the frequency of a physical symptom,
is shown. The first letters of the abbreviations in circles desig-
nate the time point (P indicates previous assessment and C
indicates current assessment), whereas the second letters
designate the variable (C indicates chemotherapy; R, radio-
therapy; H, hormone therapy; S, symptom; and I, insomnia).
For example, PC indicates “previous chemotherapy.” The
mediation was tested by estimating 2 paths: treatment to
mediator (alpha relation) and mediator to insomnia (beta
relation). Indirect alpha relations between treatments and the
mediator were tested for previous treatments (PC, PR, and
PH!CS) and current treatments (CC, CR, and CH!CS). Indi-
rect beta relations between the mediator and insomnia were
tested for the previous mediator (PS!CI) and the current
mediator (CS!CI). Direct relations between treatments and
insomnia, controlling for the mediator effect, were tested for
previous treatments (PC, PR, and PH!CI) and current treat-
ments (CC, CR, and CH!CI). Autoregressive relations to con-
trol for temporal dependency were tested for both the
mediator (PS!CS) and insomnia (PI!CI).
Cancer Month 00, 2015
Cancer Treatments and Insomnia/Savard et al
nificant time effect (F(2,760) 5 9.77; P<.001) and a sig-
nificant and unique radiotherapy effect (F(1,345) 5 3.84;
P 5 .05), but no significant main effect for chemotherapy
(F(1,190) 5 0.79; P 5 .38) or hormone therapy
(F(1,175) 5 0.06; P 5 .80) after controlling for the effect
of the other 2 treatments. However, simple effects demon-
strated that women receiving chemotherapy had signifi-
cantly greater ISI scores at T2 than those not receiving this
treatment (M of 10.6 vs 9.1; F(1,760) 5 5.29 [P 5 .02])
(Fig. 3a), which corresponds to their peak exposure to
chemotherapy (Fig. 1a). Overall, the exposure to radio-
therapy was associated with significantly higher ISI scores
throughout the study (M of 9.0 vs 8.4) (Fig. 3b). No treat-
ment 3 time interaction was found to be statistically sig-
nificant (P 5 .23 for chemotherapy, P 5 .87 for
radiotherapy, and P 5 .66 for hormone therapy).
Prostate cancer
The mixed model analysis conducted among participants
with prostate cancer revealed a significant main effect for
hormone therapy (F(1,9) 5 5.01; P 5 .05) but no signifi-
cant time effect (F(1,610) 5 0.62; P 5 .60) or interaction
(F(1,610) 5 0.83; P 5 .48). Overall, participants
reported significantly higher ISI scores throughout the
study when exposed to hormone therapy (M of 7.3 vs 5.4)
(Fig. 3d). In addition, simple effects revealed that patients
receiving hormone therapy had significantly greater ISI
scores at T4 (P 5 .03), T5 (P 5 .04), and T6 (P 5 .03),
which corresponds to the peak exposure to hormone ther-
apy in that group (Fig. 1b).
Objective 2: Mediating Effect of Somatic
Symptoms in the Relationship Between
Adjuvant Treatments and Insomnia
Table 2 presents the results of mediation effects calculated
by the product method for both subgroups of partici-
pants. All significant relationships indicated worse insom-
nia symptoms in association with cancer treatments and
somatic symptoms. With the exception of one effect for
radiotherapy, all effects indicated a total mediation.
Breast cancer
Concurrent mediation effects were more frequent (10 sig-
nificant effects from 21 tests) than lagged ones (6 signifi-
cant effects from 21 tests). More specifically, the
concurrent effect of chemotherapy on insomnia symp-
toms was significantly mediated by urinary symptoms
(39.1% of the total effect explained by the mediation
effect), nausea (34.7% of the total effect), night sweats
(30.0% of the total effect), digestive symptoms (14.4% of
the total effect), and dyspnea (13.5% of the total effect).
5
Original Article

Figure 3. Adjusted Insomnia Severity Index (ISI) scores are shown for participants exposed or not to each cancer treatment since
the time of the last assessment. Only time points with at least 10 participants exposed are shown: (a) chemotherapy in breast
cancer, (b) radiotherapy in breast cancer, (c) hormone therapy in breast cancer, and (d) hormone therapy in prostate cancer.

TABLE 2. Concurrent and Longitudinal Mediational Effects of Symptoms on the Relationship Between
Treatment and Insomnia, According to Treatment and Cancer Type

Breast Cancer (n 5 465) Prostate Cancer (n 5 263)

Chemotherapy Radiotherapy Hormone Therapy Hormone Therapy

Concurrent Longitudinal Concurrent Longitudinal Concurrent Longitudinal Concurrent Longitudinal

Symptoms Z0 (%) Z0 (%) Z0 (%) Z0 (%) Z0 (%) Z0 (%) Z0 (%) Z0 (%)

Headache NS 1.39a (73.0) T NS 0.95b (35.9) T 0.90b (15.4) T NS - -

Dyspnea 0.96b (13.5) T NS 1.64a (19.8) P NS NS NS 1.35a (18.0) T NS
a
Nausea 2.45 (34.7) T 1.08b (14.3) T NS 0.92b (3.9) T NS NS - -
Digestive 1.72a (14.4) T 1.56a (20.2) T NS NS 1.44a (21.8) T 1.09a (36.7) T - -
Urination 3.09a (39.1) T NS NS NS NS NS 1.55a (16.6) T NS
Night sweats 2.05a (30.0) T NS 1.12a (14.9) T NS 3.56a (100) T NS 1.83a (48.6) T 0.85b (45.8) T
Pain NS NS NS NS NS NS 1.25a (13.8) T NS

Abbreviations: %, percentage of the total relation accounted for by the mediation effect; NS, not significant; P, partial mediation; T, total mediation; Z0 , Z-prime
statistic for the alpha 3 beta mediation test.
a
P <.01.
b
P <.05.

The lagged (longitudinal) effect of chemotherapy on
insomnia was significantly mediated by headache (73.0%
of the total effect), digestive symptoms (20.2% of the total
effect), and nausea (14.3% of the total effect). The con-
current effect of radiotherapy on insomnia symptoms was
significantly mediated by dyspnea (19.8% of the total
effect [partial mediation]) and night sweats (14.9% of the
total effect). The lagged radiotherapy effect was signifi-
cantly mediated by headache (35.9% of the total effect)
and nausea (3.9% of the total effect). Finally, the concur-
rent effect of hormone therapy on ISI scores was signifi-
cantly mediated by night sweats (100% of the total

6 Cancer Month 00, 2015


effect), digestive symptoms (21.8% of the total effect),
and headache (15.4% of the total effect), whereas the
lagged effect was significantly mediated by digestive
symptoms only (36.7% of the total effect).
Prostate cancer
Again, concurrent effects were more frequent (4 signifi-
cant effects from 4 tests) than lagged effects (only 1
significant effect). The concurrent effect of androgen de-
privation therapy on symptoms of insomnia was signifi-
cantly mediated by night sweats (48.6% of the total
effect), dyspnea (18.0% of the total effect), urinary
symptoms (16.6% of the total effect), and pain (13.8% of
the total effect), whereas the lagged effect was signifi-
cantly mediated by night sweats only (45.8% of the total
effect).
DISCUSSION
The current longitudinal study assessed, within the con-
text of breast and prostate cancer, the role of adjuvant
treatments in the evolution of insomnia symptoms over
an 18-month period and the mediating role of somatic
symptoms potentially caused by these treatments. In
patients with breast cancer, the findings indicated that
chemotherapy (at T2) and radiotherapy (overall effect),
but not hormone therapy, were associated with an aggra-
vation of insomnia symptoms after controlling for the
effect of the other 2 treatments. In this group, the effect of
chemotherapy and radiotherapy on insomnia levels was
significantly mediated by a variety of somatic symptoms
likely to be their side effects. Although hormone therapy
was not found to be associated with a significant exacerba-
tion of insomnia symptoms in that group, its effect was
significantly mediated by night sweats. In patients with
prostate cancer, androgen deprivation therapy was consis-
tently associated with increased insomnia symptoms, an
effect that was strongly mediated by night sweats. Given
that nearly all mediating effects in both groups were total
mediations, this suggests that somatic symptoms explain a
major percentage of the relationship between cancer treat-
ments and insomnia.
In patients with breast cancer, chemotherapy was
associated with increased insomnia symptoms, an effect
that was significant at T2 when 51.3% of the women were
receiving this treatment. These results are consistent with
some evidence pointing to a deleterious effect of chemo-
therapy on sleep.1,27-29 Although other mechanisms are
possible (eg, psychological reaction, disruption of circa-
dian rhythms, or immune response4,30), the current
results supporting the mediating role of a variety of so-
Cancer Month 00, 2015
Cancer Treatments and Insomnia/Savard et al
matic symptoms suggest that this negative impact is sig-
nificantly due to its side effects, in particular headache,
nausea and digestive symptoms, urination, and night
sweats.
Some chemotherapeutic agents can induce headache
(eg, 5-fluorouracil) and a bidirectional relationship
between headache and insomnia appears to exist.31
Indeed, the discomfort associated with headache may
interfere with falling and staying asleep throughout the
night, but disturbed sleep can also cause headache. A dele-
terious impact of chemotherapy-induced nausea and di-
gestive symptoms on sleep has already been shown.32
However, it should be noted that this effect could also be
explained by the use of antiemetic medications that can al-
ter sleep as well (eg, dexamethasone33). Increased urina-
tion, another possible side effect of chemotherapy, was
associated with greater severity of insomnia, which is con-
sistent with an increasing body of literature indicating a
link between nocturia and poor sleep in the general popu-
lation.34 Finally, findings regarding night sweats are con-
sistent with studies performed within the context of breast
cancer using objective measures (polysomnography and
sternal skin conductance) supporting a concurrent associ-
ation between nocturnal hot flashes and various sleep
impairments.35,36
In the subsample of patients with breast cancer,
radiotherapy was found to be associated with a signifi-
cantly increased severity of insomnia overall. The strong-
est mediating effect was found for headache. Because
irradiating the breast is not likely to induce headache, this
may indicate an inverse relationship in which headache is
induced by insomnia. Other significant mediating effects
were noted for dyspnea, night sweats, and nausea, but
they explained a smaller percentage of the total effect.
Overall, somatic symptoms, at least those assessed in the
current study, appear to explain to a lesser extent the rela-
tionship between radiotherapy and insomnia in patients
with breast cancer. Future studies should investigate the
possible mediating role of fatigue. Fatigue is significantly
correlated with insomnia within the context of radiother-
apy,37 and our previous findings using the same sample
revealed that fatigue was a significant predictor of
insomnia.38
Hormone therapy also was found to be associated
with significantly greater insomnia scores, but only in
men with prostate cancer. In that group, the effect was sig-
nificant overall and at each time point, except T3, when
only 6.9% of patients received this treatment. Although
analyses revealed a small mediating role of other symp-
toms (ie, dyspnea, urinary symptoms, and pain), the most
7
Original Article
influential mediator was night sweats. Findings of the cur-
rent study are consistent with those of another recent lon-
gitudinal study that demonstrated a deleterious effect of
androgen deprivation therapy for prostate cancer on
insomnia levels through the mediating role of night sweats
and hot flashes.13 It is interesting to note that even though
hormone therapy was not associated with significantly
increased insomnia in patients with breast cancer in the
current study, night sweats explained 100% of the rela-
tionship between this treatment and insomnia, thereby
emphasizing the importance of this symptom as a risk fac-
tor for insomnia in this group as well. The finding that
hormone therapy was not associated with a significant
exacerbation of insomnia in patients with breast cancer is
surprising but may be due to some ceiling effect. Indeed,
women with breast cancer had already, on average, clinical
levels of insomnia before receiving hormone therapy (at
T2), thus leaving less room for further aggravation of their
symptoms because of its introduction.
The current study is characterized by several
strengths, including the use of a large population-based
sample and of a longitudinal design with 6 time points
over the entire cancer care trajectory. Given that treatment
regimens and side effect profiles may vary across cancers,
the separate investigation of the most common cancer
types is also a strength. One limitation is the relatively
large time frame between the study measurements, thus
making causal inferences difficult. Indeed, participants
were considered to be exposed to a particular treatment if
they had received it since the time of the last assessment.
Given the large intervals between time points (2-4
months), this means that treatments may have ended days
and even weeks before insomnia (and somatic symptoms)
was measured, thus potentially weakening the strength of
the association between treatments and the severity of
insomnia. This may also explain why concurrent associa-
tions were more frequent than longitudinal ones. Finally,
given the large variety of treatments received in the cur-
rent study sample, the analyses could not assess the differ-
ential impact of each type of chemotherapy regimen and
hormone therapy.
Although the etiology of insomnia comorbid with
cancer is most likely multifactorial, the results of the
current study suggest that adjuvant cancer treatments
are associated with increased severity of insomnia through
the mediating effect of their side effects. The results of
the current study highlight the importance of the appro-
priate management of the side effects of cancer treatments
to prevent the occurrence or aggravation of sleep
difficulties.
8 Cancer
FUNDING SUPPORT
Supported by a grant from the Canadian Institutes of Health
Research (MOP-69073).
CONFLICT OF INTEREST DISCLOSURES
Dr. Morin has served as consultant for Merck, Novartis, and
Valeant and received research support from Novartis for work per-
formed outside of the current study.
REFERENCES
1. Van Onselen C, Paul SM, Lee K, et al. Trajectories of sleep disturb-
ance and daytime sleepiness in women before and after surgery for
breast cancer. J Pain Symptom Manage. 2013;45:244-260.
2. Savard J, Villa J, Ivers H, Simard S, Morin CM. Prevalence, natural
course, and risk factors of insomnia comorbid with cancer over a 2-
month period. J Clin Oncol. 2009;27:5233-5239.
3. Savard J, Ivers H, Villa J, Caplette-Gingras A, Morin CM. Natural
course of insomnia comorbid with cancer: an 18-month longitudinal
study. J Clin Oncol. 2011;29:3580-3586.
4. Palesh O, Peppone L, Innominato PF, et al. Prevalence, putative
mechanisms, and current management of sleep problems during
chemotherapy for cancer. Nat Sci Sleep. 2012;4:151-162.
5. Ancoli-Israel S, Savard J. Sleep and fatigue in cancer patients. In:
Kryger MH, Roth T, Dement WC, eds. Principles and Practice of
Sleep Medicine. St Louis, MO: Elsevier Saunders; 2011:1416-1421.
6. Savard J, Savard MH. Insomnia and cancer: prevalence, nature, and
nonpharmacologic treatment. Sleep Med Clin. 2013;8:373-387.
7. Yue HJ, Dimsdale JE. Management of specific symptoms: sleep and
cancer. In: Holland JC, Breitbart W, Jacobsen PB, et al, eds. Psy-
chooncology. 2nd ed. New York: Oxford University Press; 2010:
258-269.
8. Sanford SD, Wagner LI, Beaumont JL, Butt Z, Sweet JJ, Cella D.
Longitudinal prospective assessment of sleep quality: before, during,
and after adjuvant chemotherapy for breast cancer. Support Care
Cancer. 2013;21:959-967.
9. Liu L, Fiorentino L, Natarajan L, et al. Pre-treatment symptom clus-
ter in breast cancer patients is associated with worse sleep, fatigue
and depression during chemotherapy. Psychooncology. 2009;18:187-
194.
10. Omne-Ponten M, Holmberg L, Burns T, Adami HO, Bergstrom R.
Determinants of the psycho-social outcome after operation for breast
cancer: results of a prospective comparative interview study following
mastectomy and breast conservation. Eur J Cancer. 1992;28A:1062-
1067.
11. Wengstrom Y, Haggmark C, Strander H, Forsberg C. Perceived
symptoms and quality of life in women with breast cancer receiving
radiation therapy. Eur J Oncol Nurs. 2000;4:78-88.
12. Mo YL, Li L, Qin L, et al. Cognitive function, mood, and sleep
quality in patients treated with intensity-modulated radiation therapy
for nasopharyngeal cancer: a prospective study. Psychooncology. 2014;
23:1185-1191.
13. Savard J, Hervouet S, Ivers H. Prostate cancer treatments and their
side effects are associated with increased insomnia. Psychooncology.
2013;22:1381-1388.
14. Levin TT, Alici Y. Psychiatric disorders: anxiety disorders. In: Hol-
land JC, Breitbart W, Jacobsen PB, et al, eds. Psychooncology. 2nd
ed. New York: Oxford University Press; 2010:324-331.
15. Liu L, Marler MR, Parker BA, et al. The relationship between fa-
tigue and light exposure during chemotherapy. Support Care Cancer.
2005;13:1010-1017.
16. Park JY, Lee SK, Bae SY, et al. Aromatase inhibitor-associated mus-
culoskeletal symptoms: incidence and associated factors. J Korean
Surg Soc. 2013;85:205-211.
17. Buijs C, de Vries EG, Mourits MJ, Willemse PH. The influence of
endocrine treatments for breast cancer on health-related quality of
life. Cancer Treat Rev. 2008;34:640-655.
Month 00, 2015

18. Savard J. Insomnia. In: Davis MP, Feyer P, Ortner P, Zimmermann
C, eds. Supportive Oncology. 1st ed. Philadelphia: Elsevier Saunders;
2011:187-199.
19. Morin CM. Insomnia: Psychological Assessment and Management.
New York: The Guilford Press; 1993.
20. Savard MH, Savard J, Simard S, Ivers H. Empirical validation of the
Insomnia Severity Index in cancer patients. Psychooncology. 2005;14:
429-441.
21. Portenoy RK, Thaler HT, Kornblith AB, et al. The Memorial
Symptom Assessment Scale: an instrument for the evaluation of
symptom prevalence, characteristics and distress. Eur J Cancer. 1994;
30A:1326-1336.
22. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nd
ed. Hillsdale, NJ: Lawrence Erlbaum Associates; 1988.
23. SAS Institute Inc. SAS/STAT 9.3 User’s Guide. Cary, NC: SAS
Institute Inc; 2011.
24. Frigon JY, Laurencelle L. Analysis of covariance: a proposed algo-
rithm. Educ Psychol Meas. 1993;53:1-18.
25. MacKinnon DP. Longitudinal mediation models. In: MacKinnon
DP, ed. Introduction to Statistical Mediation Analysis. New York:
Lawrence Erlbaum Associates, Taylor & Francis Group; 2008:193-
236.
26. MacKinnon DP, Lockwood CM, Hoffman JM, West SG, Sheets V.
A comparison of methods to test mediation and other intervening
variable effects. Psychol Methods. 2002;7:83-104.
27. Savard J, Liu L, Natarajan L, et al. Breast cancer patients have pro-
gressively impaired sleep-wake activity rhythms during chemother-
apy. Sleep. 2009;32:1155-1160.
28. Chen ML, Yu CT, Yang CH. Sleep disturbances and quality of life
in lung cancer patients undergoing chemotherapy. Lung Cancer.
2008;62:391-400.
29. Jim HS, Small B, Faul LA, Franzen J, Apte S, Jacobsen PB. Fatigue,
depression, sleep, and activity during chemotherapy: daily and intra-
Cancer Month 00, 2015
Cancer Treatments and Insomnia/Savard et al
day variation and relationships among symptom changes. Ann Behav
Med. 2011;42:321-333.
30. Wood LJ, Weymann K. Inflammation and neural signaling: etiologic
mechanisms of the cancer treatment-related symptom cluster. Curr
Opin Support Palliat Care. 2013;7:54-59.
31. Tran DP, Spierings EL. Headache and insomnia: their relation
reviewed. Cranio. 2013;31:165-170.
32. Osoba D, Zee B, Warr D, Latreille J, Kaizer L, Pater J. Effect of
postchemotherapy nausea and vomiting on health-related quality of
life. The Quality of Life and Symptom Control Committees of the
National Cancer Institute of Canada Clinical Trials Group. Support
Care Cancer. 1997;5:307-313.
33. Ling MH, Perry PJ, Tsuang MT. Side effects of corticosteroid ther-
apy: psychiatric aspects. Arch Gen Psychiatry. 1981;38:471-477.
34. Bliwise DL, Friedman L, Hernandez B, Zeitzer JM, Kushida CA,
Yesavage JA. Nocturia reported in nightly sleep diaries: common
occurrence with significant implications? Health Psychol. 2014;33:
1362-1365.
35. Savard J, Davidson JR, Ivers H, et al. The association between noc-
turnal hot flashes and sleep in breast cancer survivors. J Pain Symp-
tom Manage. 2004;27:513-522.
36. Savard MH, Savard J, Caplette-Gingras A, Ivers H, Bastien C. Rela-
tionship between objectively recorded hot flashes and sleep distur-
bances among breast cancer patients: investigating hot flash
characteristics other than frequency. Menopause. 2013;20:997-1005.
37. Donovan KA, Jacobsen PB. Fatigue, depression, and insomnia: evi-
dence for a symptom cluster in cancer. Semin Oncol Nurs. 2007;23:
127-135.
38. Trudel-Fitzgerald C, Savard J, Ivers H. Which symptoms come first?
Exploration of temporal relationships between cancer-related symp-
toms over an 18-month period. Ann Behav Med. 2013;45:329-337.
9