Ij

ONLINE EXCLUSIVE
This material is protected by U.S. copyright law. Unauthorized reproduction is prohibited. To purchase quantity reprints,
please e-mail reprints@ons.org or to request permission to reproduce multiple copies, please e-mail pubpermissions@ons.org.
Pain, Sleep Disturbance, and Fatigue
in Patients With Cancer: Using a Mediation Model
to Test a Symptom Cluster
Susan L. Beck, PhD, APRN, FAAN, William N. Dudley, PhD,
and Andrea Barsevick, DNSc, RN
Key Points . . .
Purpose/Objectives: To test whether sleep disturbance mediates the
effect of pain on fatigue.
Design: Cross-sectional.
Pain is related significantly to fatigue in individuals experienc-
Setting: Radiation therapy clinic, oncology ambulatory clinic, and
ing cancer pain.
inpatient oncology unit in an urban teaching hospital.
Sample: 84 patients with cancer with multiple primary diagnoses who
Although some of the effect of pain on fatigue is mediated by
were experiencing pain. Fifty-three percent were female and 92% were
sleep disturbance, pain has a direct effect on fatigue as well.
Caucasian, with a mean age of 54 years.
Strategies to improve sleep by effectively managing pain may
Methods: All participants completed a symptom questionnaire that
decrease fatigue.
included the Brief Pain Inventory­Short Form, the Pittsburgh Sleep Quality
Index, and the fatigue subscale of the Profile of Mood States questionnaire.
Multistage linear regression was used to test a mediation model.
Main Research Variables: Fatigue, pain, and sleep disturbance.
sleeplessness and fatigue. This article examines the symptom
Findings: Mediation analyses indicate that pain influences fatigue
cluster of pain, sleep disturbance, and fatigue in a sample of
directly as well as indirectly by its effect on sleep. About 20% (adjusted
84 patients with cancer-related pain who participated in a
R2 = 0.20) of the variation in fatigue is explained by pain. Thirty-five
percent of the variance in fatigue explained by pain was accounted for
study to examine the symptom experience.
by the mediation pathway.
Conclusions: Some of the effect of pain on fatigue is mediated by
Background
sleep disturbance, but pain has a direct effect on fatigue as well.
Implications for Nursing: Although the relationship can be ex-
In oncology, the study of symptoms has focused primar-
plained only partially by the commonsense point of view that people
ily on single symptoms, overall symptom burden, and, more
who are in pain lose sleep and naturally report more fatigue, this finding
recently, symptom pairs. Yet, experienced oncology special-
is important and leads to a potential intervention opportunity. Strate-
ists know that individuals undergoing cancer therapy often
gies to improve sleep by better pain management may contribute to
experience multiple symptoms of different origin, pattern,
decreased fatigue.
and duration. Recent work (Dodd, Miaskowski, & Lee, 2004;
Dodd, Miaskowski, & Paul, 2001) has helped to refocus the
U
nrelieved symptoms significantly affect quality of
study of cancer symptoms on "symptom clusters." However,
life for individuals with cancer. Much of the research
although patients with cancer seem to experience clusters
to date has focused on single symptoms, yet most
of symptoms, the research into cancer symptoms has not
patients with cancer experience more than one symptom at a
time. Pain, sleep disturbance, and fatigue are three of the most
common symptoms; however, the relationship among these
Susan L. Beck, PhD, APRN, FAAN, is the associate dean for research
variables has not been studied fully. Two models regarding the
and William N. Dudley, PhD, is the director of applied statistics, both
association among these three constructs could be proposed.
in the College of Nursing at the University of Utah in Salt Lake City;
and Andrea Barsevick, DNSc, RN, is the director of nursing research
First, individuals who are in pain sleep less and show greater
at Fox Chase Cancer Center in Philadelphia, PA. (Submitted July
fatigue. Second, pain and fatigue are related but independent
2004. Accepted for publication November 3, 2004.)
of sleeplessness. In this second model, pain has a role in
fatigue that is above and beyond the relationship between
Digital Object Identifier: 10.1188/05.ONF.E48-E55
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E48
progressed to a point where healthcare providers know why
nocturnal sleep-wake disturbances range from 31%­75%
cancer symptoms form aggregates or clusters or what underly-
among people with cancer (Clark, Cunningham, McMillan,
ing processes cause symptoms to coexist.
Vena, & Parker, 2004). In a telephone survey of 150 outpa-
tients with cancer, 44% reported insomnia in the prior month
In other healthcare arenas, the examination of symptom
(Engstrom, Strohl, Rose, Lewandowski, & Stefanek, 1999).
clusters has been useful in the diagnosis and treatment of
The most common problem was awakening during the night
syndromes. For example, premenstrual symptoms have been
(90%) or sleep disturbance. Sleeping fewer hours (85%) and
shown to form clusters that vary across the menstrual cycle,
difficulty returning to sleep (75%) also were prevalent. Differ-
distinguishing among women with and without premenstrual
ences in prevalence may be related to type of cancer and, for a
problems (Woods, Mitchell, & Lentz, 1999). With regard to
significant proportion of patients, may exist prior to diagnosis
chronic pain, individuals with varying symptom patterns have
(Clark et al., 2004).
been shown to benefit from different approaches to therapy
(Geisser, Perna, Kirsch, & Bachman, 1998).
Pain
In oncology, the study of multiple symptoms is complicated
Pain is experienced by patients undergoing cancer treat-
by differences in the origin of specific symptoms. Some symp-
toms are disease related, whereas others are related specifi-
ment as well as those with advanced disease. Approximately
cally to treatment. For example, nausea and vomiting might
33%­50% of patients with cancer experience pain at some
be caused by a bowel obstruction or chemotherapy. Other
time during the course of their illness; rates are higher in the
symptoms may be related to disease and treatment (e.g., fa-
palliative care setting (McGuire, 2004). A majority (62%) of
tigue), whereas still others result from noncancer comorbidity
people receiving treatment for bone metastases reported going
(e.g., pain caused by arthritis). Finally, the possibility exists
to sleep at night with moderate to severe pain (Miaskowski
that one or more symptoms could cause another symptom
& Lee, 1999). The prevalence of pain in specific populations
(e.g., sleep disturbance could cause fatigue).
includes 10%­12% of older adults with cancer during a one-
Given the complexity of cancer symptoms, bivariate cor-
year period (Given, Given, Azzouz, Kozachik, & Stommel,
relational models alone are unlikely to adequately describe
2001), 29% of women with lung cancer (Sarna, 1993), and
the nature of the relationships among them or lead to an
54% of a mixed cancer population (Glover, Dibble, Dodd, &
understanding of which processes lead to the clustering of
Miaskowski, 1995). Average ratings of pain were moderate
symptoms. Others have used multiple regression and factor
(Gaston-Johansson, Fall-Dickson, Bakos, & Kennedy, 1999;
analysis as approaches to exploring the complexity of these
Glover et al.; Miaskowski & Lee; Sarna).
relationships (Dodd et al., 2001; Gift, Jablonski, Stommel, &
Fatigue and Insomnia
Given, 2004). An alternate approach is to use more complex
Several studies have documented a correlation between
path analysis models such as "mediation" (Baron & Kenny,
fatigue and insomnia in patients with cancer (Broeckel, Ja-
1986) modeling, in which one symptom is proposed to in-
fluence another symptom through its relationship to a third
cobsen, Horton, Balducci, & Lyman, 1998; Carpenter et al.,
symptom or factor. A move from bivariate descriptive models
2004; Hann et al., 1997; Jacobsen et al., 1999; Longman,
to more complex explanatory models could yield important
Braden, & Mishel, 1999). In people treated with radiation
information about the nature of the relationships among
therapy for bone metastases, Miaskowski and Lee (1999)
symptoms and suggest new avenues for symptom manage-
found that improvement in fatigue from evening to morning
ment (Beck, 2004).
was associated with higher total sleep time, decreased night-
time awakening, and better sleep efficiency. Berger and Farr
Fatigue
(1999) found that less active women with breast cancer who
Fatigue has been described as the most frequent problem
took more naps and had more nighttime awakenings had
related to cancer and its treatment (Lawrence, Kupelnick,
higher levels of fatigue. These findings suggest that insomnia,
Miller, Devine, & Lau, 2004). Fatigue is distressing and inter-
and sleep disturbance in particular, could be a cause of fatigue
feres with quality of life regardless of diagnosis, treatment, or
in patients with cancer.
prognosis (Nail, 2002). Cancer-related fatigue has profound
Fatigue and Pain
effects on patients' ability to function in usual roles and activi-
Evidence also is accumulating of an association between
ties, causes delays in treatment, can linger for months or years,
fatigue and pain. A weak to moderate correlation between
and may be predictive of shorter survival in certain cancer
fatigue and pain has been demonstrated (Dodd et al., 2001;
populations (Patrick et al., 2003; Stasi, Abriani, Beccaglia,
Glover et al., 1995). Given et al. (2001) found that pain and
Terzoli, & Amadori, 2003).
fatigue were independent and additive predictors of symptom
Insomnia and Sleep Disturbance
burden. Patients who had fatigue and pain reported a greater
Insomnia is defined as difficulty initiating or maintaining
number of symptoms than those who had fatigue or pain alone
or those reporting neither symptom.
sleep or as nonrestorative sleep causing clinically significant
distress or impairment of social, occupational, or other areas
Insomnia and Pain
of functioning. Insomnia may include increased sleep la-
Only one study examined insomnia and pain in patients
tency (defined as time to fall asleep), reduced sleep efficiency
with cancer. Miaskowski and Lee (1999) reported that 62% of
(defined as time asleep or total time in bed), or increased
their sample indicated having moderate or severe pain; 75%
number and duration of awakenings (American Sleep Dis-
orders Association, 1990; Morin, 1993). Sleep disturbance
had decreased sleep efficiency (using a wrist actigraph mea-
sure). However, the relationship between these two symptoms
indicates the degree to which sleep may be disrupted by
was not reported.
environmental and personal factors. The prevalence rates for
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E49
Pain, Fatigue, and Sleep Insufficiency
Sample and Setting
All patients who were receiving care in three settings (out-
Dodd et al. (2001) studied the symptom cluster of pain, fa-
patient oncology, radiation therapy, and inpatient oncology) at
tigue, and sleep insufficiency in 93 patients with cancer. They
the University of Utah during a designated time period were
used a hierarchical multiple regression approach and found
screened for study eligibility and invited to complete a ques-
that age, pain, and fatigue explained 48% of the variance in
tionnaire designed to assess their symptom experience.
functional status. Sleep insufficiency was not a significant
factor. A limitation of the study was that symptom measures
Measures
were restricted to individual items on a quality-of-life instru-
A symptom questionnaire was comprised of several tools
ment. Gift et al. (2004) used a factor analytic approach in
with established psychometric properties. All tools were
patients with lung cancer; the symptom cluster of pain, sleep
framed within the context of "the past week."
disturbance, and fatigue did not emerge as a factor.
Pain: The symptom questionnaire included the Brief Pain
Summary
Inventory­Short Form (BPI), a self-report tool that is used
widely to measure pain (Daut, Cleeland, & Flanery, 1983). The
Together, these studies demonstrate the authors' initial under-
BPI is brief and easy to complete (Serlin, Mendoza, Nakamura,
standing of the problem of symptom clusters in a mixed group
Edwards, & Cleeland, 1995). The World Health Organization
of patients with varying diagnoses, treatment regimens, and
adopted the BPI to evaluate the effectiveness of national cancer
stages of disease. All provided some evidence to support the
pain relief programs. The intensity subscale of the BPI consists
association among two or more symptoms. However, previous
of four items that ask patients to rate their pain intensity as
research has not addressed a most crucial question: What is the
worst pain, least pain, and average pain in the past week, as well
nature of the relationship among a group of commonly experi-
as pain now. Each item is rated on a scale from 0 (no pain) to
enced symptoms? This limitation also is reflected in the current
10 (the worst pain imaginable).
stage of existing conceptual models. The Symptoms Experience
Evidence exists to support the reliability and validity of the
Model (Armstrong, 2003), the Theory of Unpleasant Symptoms
BPI. Test-retest reliability of the worst pain scale was 0.93
(Lenz & Pugh, 2003), and the Explanatory Model of Fatigue
over a two-day period in a sample of 20 hospitalized patients
(Berger & Walker, 2001) provide a framework for the study of
with cancer. Internal consistency for the intensity scale was
multiple symptoms. Within these models, an individual may
an alpha of 0.87 (Lin & Ward, 1995; Serlin et al., 1995). In
experience multiple symptoms at one time, each with varying
the current study's sample, the Cronbach's alpha for the BPI
levels of severity. Symptoms may be multiplicative in nature
was 0.875.
and may act as catalysts for the occurrence of other symptoms.
Sleep: The Pittsburgh Sleep Quality Index (PSQI), a self-
These mid-range theories propose that symptoms are related but
report questionnaire that assesses sleep quality and quantity,
generic; they do not specify any symptoms or any order to their
was used to measure sleep. The original version was designed
relationship. The current study thus builds on mid-level theory
to measure sleep reports during a one-month interval (Buysse,
but adds the specificity of systematically examining the pattern
Reynolds, Monk, Berman, & Kupfer, 1989). The 19-item
of relationships among three commonly reported symptoms
self-report questionnaire yields seven component scores:
in patients with cancer: pain, sleep disturbance, and fatigue.
subjective sleep quality, sleep latency, duration, habitual sleep
Fatigue is known to be the most prevalent symptom, yet pain,
efficiency, sleep disturbances, use of sleeping medication, and
sleep disturbance, or both could cause fatigue. This information
daytime dysfunction. A sleep disturbance score is computed
would be useful in targeting an intervention appropriately to the
by taking the sum of 10 items that assess the degree to which
root cause. An additional possibility is that one symptom could
a variety of factors have interfered with sleep.
affect another symptom indirectly through the mediating effect
The PSQI has reliability and validity when used in older
of a third symptom. In this article, the authors will examine the
adults (Buysse et al., 1991; Gentili, Weiner, Kuchibhatla, &
proposition that pain increases fatigue severity by causing sleep
Edinger, 1995), bereaved spouses (Reynolds et al., 1993),
disturbance. In this case, resolving or reducing fatigue and sleep
patients with panic disorder (Stein, Chartier, & Walker, 1993),
disturbance would require improved pain management.
and patients with phobias (Stein, Kroft, & Walker, 1993). A
Cronbach's alpha of 0.83 was reported for the Global Sleep
Purpose
Quality Scale. Psychometric evaluation of the PSQI in this
sample and another sample of patients with cancer undergoing
The aims of this study were to (a) examine the relationships
treatment supported its internal consistency reliability and con-
among three symptoms--pain, sleep disturbance, and fatigue,
struct validity (Beck, Schwartz, Towsley, Dudley, & Barsevick,
(b) determine to what extent pain intensity affects sleep dis-
2004). In this sample, Cronbach's alpha was 0.81 for the Global
turbance and fatigue, and (c) test whether sleep disturbance
Sleep Quality Scale and 0.69 for the Sleep Disturbance Scale.
mediates the effect of pain on fatigue.
Fatigue: The Profile of Mood States (POMS) (McNair,
Lorr, & Droppleman, 1992) is a 30-item adjective checklist
Methods
with subscales that measure six dimensions of mood: tension-
Design
anxiety, depression-dejection, anger-hostility, vigor, fatigue-
This cross-sectional study was designed to evaluate the
inertia, and confusion-bewilderment. Each item is rated from
relationships among common symptoms experienced by
1­5. In this research, the five items on the fatigue scale were
individuals with cancer and used a prospective, consecutive
used as a measure of fatigue. Six independent-factor analytic
sampling approach. The study was approved by an institu-
studies have been conducted in the development and valida-
tional review board. Completion of a symptom questionnaire
tion of the POMS (McNair et al.). Meek et al.'s (2000) find-
ings for the fatigue subscale with patients with cancer showed
implied voluntary consent to participate.
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E50
The fourth step provides a statistical test that the indirect
Adjusted R2 = 0.210
Adjusted R2 = 0.277
path is greater than zero. It sometimes is referred to as the
Sleep disturbances
F = 23.10
F = 16.53
"Sobel Test" and is computed as a t test (see Figure 2).
(Pittsburgh Sleep
p < 0.001
p < 0.001
This four-step process was conducted using SPSS pro-
Quality Index)
gramming developed by Dudley, Benuzillo, and Carrico
(2004). This programming tests the mediation effect and the
a
b
proportion of variance that can be attributed to the mediation
effect.
c
Pain intensity
Fatigue (Profile of
(Brief Pain Inventory)
Mood States)
Findings
0.07a (0.01)b
Adjusted R2 = 0.201
Sample
F = 21.33
Of those who were eligible, 214 patients in the following
p < 0.001
areas consented to participate: radiation therapy (n = 81),
Raw coefficient (b)
a
outpatient oncology clinic (n = 86), and inpatient oncology
Standard error of raw coefficient (b)
b
Figure 1. Test of the Mediation Model
Table 1. Demographic Characteristics of the Sample and
Subgroup With Pain
a Cronbach's alpha of 0.91. In the current study's sample, the
Cronbach's alpha was 0.925.
Patients With Pain
Total Sample
(N = 84)
(N = 214)
Analysis
Data were analyzed with SPSS® version 11.0 (SPSS Inc.,
Characteristic
n
%
n
%
Chicago, IL). Analysis included summary statistics to exam-
Gender
ine the range, distribution, mean, and standard deviation for
Male
107
51
39
47
each subscale. Relations among symptoms were evaluated
Female
103
49
44
53
with bivariate correlations. To evaluate the effect of pain on
Missing
004
­
01
­
fatigue and sleep with analysis of variance (ANOVA), the
Race or ethnicity
mean pain intensity score was recoded using the following
African American
001
01
01
01
categories: 1­ 4 = mild pain (1), 5­6 = moderate pain (2),
Asian/Pacific Islander
003
01
­
­
and 7­10 = severe pain (3).
Native American
005
02
02
02
Hispanic
007
03
04
05
The mediation pathway from pain to fatigue through sleep
Caucasian
193
92
76
92
disturbance was tested as recommended by MacKinnon (1994).
Missing
005
­
01
­
MacKinnon's model is an extension of the three-step model
Marital status
proposed by Baron and Kenny (1986) and Judd and Kenny
Single
028
13
11
13
(1981). This multistage linear regression model of testing me-
Married or living with a partner
142
68
55
67
diation (see Figure 1) includes four tests.
Separated or divorced
021
10
10
12
1. Pain is associated with fatigue (path c).
Widowed
018
09
06
07
2. Pain is associated with sleep disturbance (path a).
Missing
005
­
02
­
3. When controlling for pain, sleep disturbance is associated
Education
with fatigue (path b), and the association between pain and
0 ­11 years
019
09
09
11
High school graduate
057
27
23
28
fatigue (path c) is either no longer statistically significant
Some college or technical school
072
35
30
37
(full mediation) or reduced significantly (partial media-
College graduate
028
14
07
09
tion).
Postgraduate
032
15
13
16
4. A statistically significant indirect path exists between pain
Missing
006
­
02
­
and fatigue through sleep disturbance.
Employment status
In practice, healthcare providers typically find evidence
Full-time
053
25
16
20
of partial mediation--that is to say, that in step 3 above, the
Part-time
011
05
04
05
association between pain and fatigue remains statistically
On sick leave or disability
036
17
21
26
significant but is reduced by a statistically significant amount.
Retired
070
33
26
32
MacKinnon's (1994) expanded model allowed the authors to
Not employed outside the home
040
19
15
18
Missing
004
­
02
­
test the mediation effect in the face of partial mediation.
Net family income ($)
Less than 14,999
042
23
21
28
15,000­34,999
046
25
20
27
a*b
35,000­69,999
059
32
22
29
t a*b =
seab
More than 70,000
036
20
12
16
Missing
031
­
09
­
Where
Age (years)
­
X
53.15
54.49
 (a *seb ) + (b *sea )
sea*b =
2
2
2
2
SD
16.57
15.48
Figure 2. Sobel Test Equation
Note. Because of rounding, not all percentages total 100.
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E51
Pain Intensity Effects on Sleep and Fatigue
Table 2. Clinical Characteristics of the Sample and
Subgroup With Pain
Two separate one-way ANOVA were performed with catego-
rized pain (mild, moderate, or severe) as the independent vari-
Total Sample
Patients With Pain
able and fatigue and sleep disturbance as dependent variables
(N = 214)
(N = 84)
(see Figures 3 and 4 and Table 5). Pain level had a significant
Characteristic
n
%
n
%
effect on fatigue (F[2, 79] = 6.17, p < 0.001) and on sleep (F[2,
81] = 9.72, p < 0.001). Scheffé tests (p = 0.05) showed that
Type of cancer
those with severe pain had significantly higher fatigue and sleep
Breast
41
19
13
16
disturbances than those with mild pain.
Lymphoma or leukemia
41
19
12
14
Prostate, testicular, or bladder
28
13
06
06
Sleep Disturbance Mediation of the Effect of Pain
Lung
14
07
05
06
on Fatigue
Gastrointestinal (colorectal,
14
07
05
06
The results are provided in Figure 1. The results indicate
esophagus, pancreas)
Head and neck
10
05
07
08
that in step 1, pain was significantly associated with fatigue
Melanoma
11
05
06
07
(path c): F(1, 80) = 21.33, p < 0.001. About 20% (adjusted
Cervical or ovarian
09
04
07
08
R2 = 0.201) of the variation in fatigue is explained by pain.
Other
46
22
23
27
In step 2, pain was significantly associated with sleep dis-
Extent of disease
turbance (path a): F(1, 82) = 23.10, (R2 = 0.210), p < 0.001.
Local
27
14
10
13
In step 3, when controlling for pain, sleep disturbance was
Regional
72
39
23
30
significantly associated with fatigue (path b): F(2, 79) =
Advanced
88
47
44
57
16.53, (R2 = 0.277), p < 0.001. Finally, step 4, the Sobel test,
Missing
27
­
07
­
was significant (p = 0.011), indicating that sleep disturbances
partially mediated the relationship between pain and fatigue.
Note. Because of rounding, not all percentages total 100.
Thirty-five percent of the variance in fatigue explained by
unit (n = 47). The study demographic characteristics are sum-
pain was accounted for by the mediation pathway. This role
marized in Table 1. Participants were 49% female with ages
of sleep disturbance as a mediator between pain and fatigue
ranging from 14­88 years and a mean age of 53. Racial and
indicates that one reason pain is related to fatigue is because
ethnic diversity was limited; 92% were Caucasian, which
sleep disturbance partially mediates the relationship between
was reflective of the state's population at that time. All were
pain and fatigue. In other words, patients who experience pain
English speaking. Of these 214 participants, 84 (39%) were
also experience fatigue; however, this relationship between
experiencing pain and completed the pain instrument included
pain and fatigue was because, at least partially, pain led to
in the questionnaire. The sample of patients with pain was
sleep disturbances that, in turn, led to fatigue.
53% female and 92% Caucasian and ranged in age from
17­82 with a mean of 54. The patients had multiple types
Discussion
of cancer primary diagnosis (see Table 2). Of the patients
with pain, 57% had advanced disease as compared to 40% of
Three symptoms that are reported commonly during the
those without pain. A chi-square test of association indicated
cancer experience were selected for this analysis: pain, sleep
that this was borderline significant (p = 0.07). No significant
disturbance, and fatigue. The first aim was to examine the
differences existed between the patients with pain and the
bivariate relationships among the three symptoms. These rela-
remaining sample on any demographic characteristics.
tionships were positive, statistically significant, and moderate
Summary statistics for the three symptom measures are
in strength. Thus, an increase in any one symptom was associ-
included in Table 3. The range of scores was acceptable, and
ated with an increase in the others. These results add evidence
adequate variance existed in the responses.
to the positive relationships reported by others (Broeckel et
al., 1998; Dodd et al., 2001; Glover et al., 1995; Hann et al.,
Relationship Among Pain, Sleep Disturbance,
1997; Jacobsen et al., 1999; Longman et al., 1999).
and Fatigue
The second aim was to determine to what extent pain in-
Bivariate Pearson correlations were used to evaluate rela-
tensity affects sleep disturbance and fatigue. Using a model in
tionships among symptom variables. A positive correlation
which pain was viewed as an independent variable, significant
existed among pain and fatigue, pain and sleep disturbance,
increases existed in fatigue and sleep disturbance as pain inten-
and fatigue and sleep disturbance. The correlations all were
sity categorically increased from mild to moderate to severe.
statistically significant and moderate in strength, ranging from
This specific approach to conceptualizing and analyzing the ef-
0.46­0.47 (see Table 4).
fect of pain is new and the linear relationship is impressive. The
Table 3. Symptom Scale Scores
­
Scale
N
Minimum
Maximum
Possible Range
X
SD
Sum of Brief Pain Inventory pain intensity
84
1
33
0­40
16.23
7.17
Sum of fatigue (Profile of Mood States)
82
5
25
5­25
16.50
5.45
Sum of the Pittsburgh Sleep Quality Index
84
0
23
0­30
13.07
5.57
sleep disturbances
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E52
Table 4. Bivariate Relationships Among Pain, Fatigue,
16
·
and Sleep Disturbance
·
Scale
Pearson Correlation
p (Two-Tailed)
N
12
·
Sum of Brief Pain Inven-
0.459
0.000
82
8
tory (BPI) pain with sum
fatigue (Profile of Mood
States [POMS])
4
Sum of BPI pain with sum
0.469
0.000
84
sleep disturbances
0
Sum fatigue POMS with
0.474
0.000
82
Mild pain
Moderate pain
Severe pain
sum sleep disturbances
Level of Pain Severity (Brief Pain Inventory)
Figure 4. Plot of Means: Effect of Pain on Sleep Disturbance
mechanism by which pain contributes to fatigue is not known,
but the influence of pain on sleep is much more explainable
given that it would serve as a counter-stimulus and significant
discomfort might keep someone from falling asleep or wake
The study is limited by the cross-sectional nature of its
them from their sleep. The clinical significance of these find-
design. Future research that is longitudinal with attention to
ings is that relieving pain is an important strategy to improving
understanding the temporal relationships among these symp-
sleep and reducing fatigue. These findings lead directly to the
toms is indicated. For example, the effect of pain on sleep
third aim of the study--to examine whether the effect of pain
disturbance may be fairly immediate, whereas the effect on
on fatigue is mediated through its effect on sleep.
fatigue may be concurrent or may lag. Research that includes
The results of the mediation analysis indicate that pain
frequent measures of each symptom is recommended. The
directly influences fatigue and indirectly influences fatigue
study sample was very heterogeneous, including many diag-
by its effect on sleep. In a full mediation model, the direct
noses, three settings, and patients at varying stages in the treat-
effect of pain on fatigue would have become insignificant
ment trajectory. Subgroup samples were too small to include
when the role of sleep was added to the model. The model
in the analysis. This analysis should be repeated in samples
demonstrates only partial mediation because the path between
that are more homogeneous in regard to diagnosis and treat-
pain and fatigue (path c) stays significant in the full model.
ments because symptom profiles may be quite different based
Thus, although some of the effect of pain on fatigue clearly
on these variables. Self-report measures that are specific to
passes through sleep, pain has a direct effect on fatigue as
each symptom were used. The use of measures more specific
well. Although the relationship can be explained only partially
to cancer populations, such as the Schwartz Cancer Fatigue
by the commonsense point of view that people who are in
Scale, and other types of sleep measures such as actigraphy
pain lose sleep and naturally report more fatigue, this finding
are recommended.
is important and leads to a potential intervention opportunity.
Another limitation of this three-symptom model is that it
Strategies to improve sleep by improved pain management
did not use other variables that may account for fatigue, such
may contribute to decreased fatigue. Research to test such
as the effects of chemotherapy, which might jointly affect
intervention strategies in a prospective way is recommended.
pain and fatigue. Other models could be proposed that include
In addition, research to evaluate the alternative mechanisms
variables such as the contribution of age, comorbidities, or
by which pain directly causes fatigue is needed. Finally, con-
length of hospital stay as well.
siderable variance in fatigue exists that is not explained in a
three-symptom cluster model. Testing of alternative models
Summary
and more complex approaches that include additional vari-
ables will increase future knowledge of the complex human
Symptoms in individuals with cancer may have complex
experience of multiple symptoms associated with cancer and
relationships with one another that need to be considered in
its treatment.
symptom management research and practice. The findings
from this study provide additional support for the concept
4
Table 5. Results of Analysis of Variance of Effect of Pain
·
(Recoded) on Fatigue and Sleep Disturbance*
·
3
·
­
­
X Fatigue
X Sleep Disturbances
2
Groups' Pain
(POMS 1 = Low
(PSQI 0 = Low
Intensity Recoded
to 5 = High)
to 30 = High)
1
Mild pain
2.44
09.34
0
Moderate pain
3.08
12.81
Mild pain
Moderate pain
Severe pain
Severe pain
3.80
15.06
Level of Pain Severity (Brief Pain Inventory)
* p < 0.001
Figure 3. Plot of Means: Effect of Pain on Fatigue
POMS--Profile of Mood States; PSQI--Pittsburgh Sleep Quality Index
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E53
of a symptom cluster in which three symptoms--pain, sleep
direct and indirect (via sleep) effect of pain on fatigue. All
disturbance, and fatigue--are interrelated (Dodd et al., 2001).
three symptoms need to be addressed in the symptom man-
agement plan. If attention is given to fatigue and pain without
The study also provides an example for how to examine a
addressing sleep disturbance, symptom management will be
symptom cluster using mediation analysis. This conceptual
less optimal. Multifocused symptom management likely will
and analytic approach can serve to evaluate other symptom
have the greatest beneficial effect.
clusters in which one symptom may mediate the relationship
between two other symptoms. Such studies can contribute to
The authors thank Jose Benuzillo, MA, in the College of Nursing at the
the development of more specific, microlevel theories related
University of Utah for his assistance with the analysis.
to symptom clusters.
If an initial symptom is directly or indirectly related to other
Author Contact: Susan L. Beck, PhD, APRN, FAAN, can be
symptoms, all need to be accounted for to conduct effective
reached at susan.beck@nurs.utah.edu, with copy to editor at rose_
symptom management. In this case, evidence supports the
mary@earthlink.net.
References
Gaston-Johansson, F., Fall-Dickson, J.M., Bakos, A.B., & Kennedy, M.J.
American Sleep Disorders Association. (1990). International classification of
(1999). Fatigue, pain, and depression in pre-autotransplant breast cancer
sleep disorders. Diagnostic and coding manual. Rochester, MN: Author.
patients. Cancer Practice, 7, 240 ­247.
Armstrong, T.S. (2003). Symptoms experience: A concept analysis. Oncology
Geisser, M.E., Perna, R., Kirsch, N.L., & Bachman, J.E. (1998). Classification
Nursing Forum, 30, 601­ 606.
of chronic pain patients with the Brief Symptom Inventory: Patient charac-
Baron, R.M., & Kenny, D.A. (1986). The moderator-mediator variable
teristics of cluster profiles. Rehabilitation Psychology, 43, 313­326.
distinction in social psychological research: Conceptual, strategic, and
Gentili, A., Weiner, D.K., Kuchibhatla, M., & Edinger, J.D. (1995). Test-retest
statistical considerations. Journal of Personality and Social Psychology,
reliability of the Pittsburgh Sleep Quality Index in nursing home residents.
51, 1173­1182.
Journal of the American Geriatric Society, 43, 1317­1318.
Beck, S.L. (2004). Symptom clusters: Impediments and suggestions for
Gift, A.G., Jablonski, A., Stommel, M., & Given, C.W. (2004). Symptom
solutions. Journal of the National Cancer Institute Monographs, 32,
clusters in elderly patients with lung cancer. Oncology Nursing Forum,
137­138.
31, 202­212.
Beck, S.L., Schwartz, A.L., Towsley, G., Dudley, W., & Barsevick, A. (2004).
Given, C.W., Given, B., Azzouz, F., Kozachik, S., & Stommel, M. (2001).
Psychometric evaluation of the Pittsburgh Sleep Quality Index in cancer
Predictors of pain and fatigue in the year following diagnosis among elderly
patients. Journal of Pain and Symptom Management, 27, 140 ­148.
cancer patients. Journal of Pain and Symptom Management, 21, 456­466.
Berger, A.M., & Farr, L. (1999). The influence of daytime inactivity and
Glover, J., Dibble, S.L., Dodd, M.J., & Miaskowski, C. (1995). Mood states
nighttime restlessness on cancer-related fatigue. Oncology Nursing Forum,
of oncology outpatients: Does pain make a difference? Journal of Pain
26, 1663­1671.
and Symptom Management, 10, 120 ­128.
Berger, A.M., & Walker, S.N. (2001). An Explanatory Model of Fatigue in
Hann, D.M., Jacobsen, P.B., Martin, S.C., Kronish, L.E., Azzarello, L.M., &
women receiving adjuvant breast cancer chemotherapy. Nursing Research,
Fields, K.K. (1997). Fatigue in women treated with bone marrow trans-
50, 42­52.
plantation for breast cancer: A comparison with women with no history
Broeckel, J.A., Jacobsen, P.B., Horton, J., Balducci, L., & Lyman, G.H.
of cancer. Supportive Care in Cancer, 5, 44 ­52.
(1998). Characteristics and correlates of fatigue after adjuvant chemother-
Jacobsen, P.B., Hann, D.M., Azzarello, L.M., Horton, J., Balducci, L., &
apy for breast cancer. Journal of Clinical Oncology, 16, 1689­1696.
Lyman, G.H. (1999). Fatigue in women receiving adjuvant chemotherapy
Buysse, D.J., Reynolds, C.F., III, Monk, T.H., Berman, S.R., & Kupfer, D.J.
for breast cancer: Characteristics, course, and correlates. Journal of Pain
(1989). The Pittsburgh Sleep Quality Index: A new instrument for psychi-
and Symptom Management, 18, 233­242.
atric practice and research. Psychiatry Research, 28, 193­213.
Judd, C.M., & Kenny, D.A. (1981). Process analysis: Estimating mediation
Buysse, D.J., Reynolds, C.F., III, Monk, T.H., Hoch, C.C., Yeager, A.L., &
in treatment evaluations. Evaluation Review, 5, 602­ 619.
Kupfer, D.J. (1991). Quantification of subjective sleep quality in healthy
Lawrence, D.P., Kupelnick, B., Miller, K., Devine, D., & Lau, J. (2004). Evi-
elderly men and women using the Pittsburgh Sleep Quality Index (PSQI).
dence report on the occurrence, assessment, and treatment of fatigue in cancer
Sleep, 14, 331­338.
patients. Journal of the National Cancer Institute Monographs, 32, 40­50.
Carpenter, J.S., Elam, J.L., Ridner, S.H., Carney, P.H., Cherry, G.J., &
Lenz, E.R., & Pugh, L.C. (2003). The Theory of Unpleasant Symptoms. In M.
Cucullu, H.L. (2004). Sleep, fatigue, and depressive symptoms in breast
Smith & P.R. Liehr (Eds.), Middle range theory for nursing (pp. 69­90).
cancer survivors and matched healthy women experiencing hot flashes.
New York: Springer.
Oncology Nursing Forum, 31, 591­598.
Lin, C.C., & Ward, S.E. (1995). Patient-related barriers to cancer pain man-
Clark, J., Cunningham, M., McMillan, S., Vena, C., & Parker, K. (2004).
agement in Taiwan. Cancer Nursing, 18, 16 ­22.
Sleep-wake disturbances in people with cancer part II: Evaluating the
Longman, A.J., Braden, C.J., & Mishel, M.H. (1999). Side-effects burden, psy-
evidence for clinical decision making. Oncology Nursing Forum, 31,
chological adjustment, and life quality in women with breast cancer: Pattern
747­771.
of association over time. Oncology Nursing Forum, 26, 909­915.
Daut, R.L., Cleeland, C.S., & Flanery, R.C. (1983). Development of the
MacKinnon, D.P. (1994). Analysis of mediating variables in prevention and
Wisconsin Brief Pain Questionnaire to assess pain in cancer and other
intervention research. NIDA Research Monograph, 139, 127­153.
diseases. Pain, 17, 197­210.
McGuire, D.B. (2004). Occurrence of cancer pain. Journal of the National
Dodd, M.J., Miaskowski, C., & Lee, K.A. (2004). Occurrence of symptom
Cancer Institute Monographs, 32, 51­56.
clusters. Journal of the National Cancer Institute Monographs, 32, 76­78.
McNair, D.M., Lorr, M., & Droppleman, L.F. (1992). Profile of Mood States
Dodd, M.J., Miaskowski, C., & Paul, S.M. (2001). Symptom clusters and
manual (2nd ed.). San Diego, CA: Educational and Industrial Testing
their effect on the functional status of patients with cancer. Oncology
Service.
Nursing Forum, 28, 465­ 470.
Meek, P.M., Nail, L.M., Barsevick, A., Schwartz, A.L., Stephen, S., Whitmer,
Dudley, W.N., Benuzillo, J.G., & Carrico, M.S. (2004). SPSS and SAS
K., et al. (2000). Psychometric testing of fatigue instruments for use with
programming for the testing of mediation models. Nursing Research,
cancer patients. Nursing Research, 49, 181­190.
53, 59­ 62.
Miaskowski, C., & Lee, K.A. (1999). Pain, fatigue, and sleep disturbances
Engstrom, C.A., Strohl, R.A., Rose, L., Lewandowski, L., & Stefanek,
in oncology outpatients receiving radiation therapy for bone metastasis: A
M.E. (1999). Sleep alterations in cancer patients. Cancer Nursing, 22,
pilot study. Journal of Pain and Symptom Management, 17, 320­332.
143­148.
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E54
Morin, C.M. (1993). Insomnia: Psychological assessment and management.
Serlin, R.C., Mendoza, T.R., Nakamura, Y., Edwards, K.R., & Cleeland,
New York: Guilford.
C.S. (1995). When is cancer pain mild, moderate, or severe? Grading pain
Nail, L.M. (2002). Fatigue in patients with cancer. Oncology Nursing Forum,
severity by its interference with function. Pain, 61, 277­284.
29, 537­546.
Stasi, R., Abriani, L., Beccaglia, P., Terzoli, E., & Amadori, S. (2003).
Patrick, D.L., Ferketich, S.L., Frame, P.S., Harris, J.J., Hendricks, C.B.,
Cancer-related fatigue: Evolving concepts in evaluation and treatment.
Levin, B., et al. (2003). National Institutes of Health state-of-the-science
Cancer, 98, 1786 ­1801.
conference statement: Symptom management in cancer: Pain, depression,
Stein, M.B., Chartier, M., & Walker, J.R. (1993). Sleep in nondepressed
and fatigue, July 15­17, 2002. Journal of the National Cancer Institute,
patients with panic disorder: I. Systematic assessment of subjective sleep
95, 1110­1117.
quality and sleep disturbance. Sleep, 16, 724 ­726.
Reynolds, C.F., III, Hoch, C.C., Buysse, D.J., Houck, P.R., Schlernitzauer, M.,
Stein, M.B., Kroft, C.D., & Walker, J.R. (1993). Sleep impairment in patients
Pasternak, R.E., et al. (1993). Sleep after spousal bereavement: A study of
with social phobia. Psychiatry Research, 49, 251­256.
recovery from stress. Biological Psychiatry, 34, 791­797.
Woods, N.F., Mitchell, E.S., & Lentz, M. (1999). Premenstrual symptoms:
Sarna, L. (1993). Correlates of symptom distress in women with lung cancer.
Delineating symptom clusters. Journal of Women's Health and Gender-
Cancer Practice, 1, 21­28.
Based Medicine, 8, 1053­1062.
ONCOLOGY NURSING FORUM ­ VOL 32, NO 3, 2005
E55