Predictors of Fatigue 2009 14ci

8/2/2019 Predictors of Fatigue 2009 14ci

1/11

38 Vol.36,No.1,January2009OncologyNursingForum

Breast cancer is the most common type ocancer in women in the United States (Jemalet al., 2008). Women treated with adjuvantchemotherapy or breast cancer o all stageshave an 89% ve-year survival rate (Jemal et

al). Taxane chemotherapy increasingly is used to improvesurvival (Ferguson, Wilcken, Vagg, Ghersi, & Nowak,2007); however, chemotherapy can result in distressul,debilitating atigue in approximately a third o survivorsthat persists or years ater completing treatment (Boweret al., 2006; National Comprehensive Cancer Network[NCCN], 2008). As the number o breast cancer survivorswho have received adjuvant chemotherapy increases,improving atigue and other symptoms becomes increas-ingly important to optimize quality o lie (Andrykowski,Schmidt, Salsman, Beacham, & Jacobsen, 2005; Janz et al.,2007; Lee, Cho, Miaskowski, & Dodd, 2004).

Fatigueis dened by NCCN (2008) as a distressing,persistent, subjective sense o physical, emotional, and/or cognitive tiredness, or exhaustion related to cancer orcancer treatment that is not proportional to recent activityand intereres with usual unctioning (p. FT1). Cancer-related atigue has been associated with disrupted circadi-an rhythms, disturbed sleep-wake, and activity rest (An-coli-Israel et al., 2006; Berger & Farr, 1999). Gaps exist inknowledge regarding the predictability o these rhythmsand patterns prior to and during the initial chemotherapytreatment as they aect atigue ater treatment.

The current study attempted to answer a clinically sig-nicant question: At the initiation o chemotherapy, canclinicians predict who will experience greater atigue30 days ater completing chemotherapy? This knowl-edge is important or clinicians to identiy women withbreast cancer who are most in need o interventions toprevent persistent, debilitating atigue. Currently, thestrongest evidence-based intervention that has beenrecommended or practice by the Oncology NursingSocietys (ONSs) Putting Evidence into Practice (PEP)card (Mitchell, Beck, Hood, Moore & Tanner, 2006) andNCCN (2008) to treat atigue is exercise and activity en-

PredictorsofFatigue30DaysAfterCompletingAnthracyclinePlusTaxaneAdjuvantChemotherapyforBreastCancer

Kimberly K. Wielgus, PhD, APRN, BC, Ann M. Berger, PhD, RN, AOCN, FAAN,and Melody Hertzog, PhD

hancement. Interventions to promote healthy circadianrhythms o activity, sleep-wake, activity-rest patternspotentially may reduce atigues dramatic eect on

Article

Purposes/Objectives: To identify the predictors of fatigue30 days after completing adjuvant chemotherapy for breastcancer and whether differences are observed between abehavioral sleep intervention and a healthy-eating attentioncontrol group in predicting fatigue.

Design: Descriptive, exploratory, secondary analysis of arandomized clinical trial.

Setting: Outpatient oncology patients in a midwesternU.S. city.

Sample: 96 women, ages 2983 years, 72% married, 95%white, diagnosed with stage IIIIA breast cancer, receivingadjuvant anthracycline and taxane chemotherapy.

Methods: Participants were randomized to a behavioralsleep intervention group or an attention control group.Participants completed data collection prior to and duringthe peak and rebound days of the initial chemotherapytreatment cycle and after the last treatment.

MainResearchVariables: Fatigue, circadian rhythms ofactivity, objective and subjective sleep-wake, and objectiveand subjective activity-rest.

Findings: Predictors of fatigue were less total sleep timeprior to treatment, higher fatigue prior to treatment and atthe peak, and less energy upon awakening on rebound days.In the control group, predictors of higher fatigue were higherfatigue prior to treatment, higher body mass index, highernumber of positive lymph nodes, and less daytime dysfunc-tion. For the intervention group, lower peak activity at thepeak of initial treatment differentially predicted fatigue.

Conclusions: Results suggest the sleep intervention groupparticipants who maintained activity balanced with sleep

at the peak of the initial treatment benefited most from theintervention.

ImplicationsforNursing:Nurses should screen for fatigueprior to initial chemotherapy treatment and at regular in-tervals, further assess for poor sleep in patients who reportfatigue of 4 or higher (on a 010 scale), and use evidence-based guidelines to select appropriate interventions.

This material is protected by U.S. copyright law. Unauthorized reproduction is prohibited. To purchase quantity reprints,

please e-mail [email protected] or to request permission to reproduce multiple copies, please e-mail [email protected]


2/11

OncologyNursingForumVol.36,No.1,January2009 39

womens physical and social unctioning, quality o lie,and disability (NCCN; Prue, Rankin, Allen, Gracey, &Cramp, 2006). This randomized controlled trial (RCT)was developed ater pilot testing the easibility andoutcomes and revision o a behavioral sleep intervention(Berger et al., 2002, 2003).

ConceptualFramework

The conceptual ramework or the current study (seeFigure 1) was derived rom the Integrated Fatigue Model(IFM) (Piper et al., 1998; Piper, Lindsey, & Dodd, 1987). Thecomponents rom the IFM that were proposed to infu-ence atigue were regulation and transmission (circadianrhythms o activity), sleep-wake, and activity-rest patterns.

The purposes o the study were to identiy the pre-dictors o atigue 30 days ater completing anthracy-cline plus taxane adjuvant chemotherapy or breastcancer using atigue, circadian rhythms o activity,sleep-wake, activity-rest, and demographic and medi-cal variables measured prior to and during the peakand rebound o the initial chemotherapy treatmentand to determine whether dierences exist betweena behavioral sleep intervention group and a healthyeating attention control group in predicting atigue30 days ater completing chemotherapy treatments.

Methods

The current descriptive, exploratory study was a sec-ondary analysis o data rom an RCT that tested a behav-ioral sleep intervention compared to a healthy eating at-

tention control in women with stages IIIIA breast cancerat initiation o chemotherapy. For the secondary analysis,

participants completed data collection prior to (days 2to 1) and during the peak (days 24) and rebound (days57) o the initial chemotherapy treatment and or 7 days30 days ater the last chemotherapy treatment.

SettingandSample

The RCT recruited participants (N = 219) romoutpatient oncology clinics in a large midwestern city.Inclusion criteria or the RCT were women ages 19

and older who were diagnosed or the rst time withstage I, II, or IIIA breast cancer; were post-breast cancersurgery (lumpectomy or mastectomy with or withoutreconstruction); were scheduled to begin adjuvant an-thracycline-based IV chemotherapy or breast cancer;were English-speaking; and had a Karnosky Peror-mance Scale (KPS) score o greater than or equal to 60.Exclusion criteria were unstable comorbidities and anerratic sleep schedule (rotating shit worker).

Only indings rom the participants who receivedanthracyclines plus taxanes are reported in the second-

ary analysis. O the 103 eligible participants, 96 whoprovided data prior to and at peak and rebound dayso the initial chemotherapy treatments, are included inthe analysis.

As an exploratory study, choice o predictors wasdriven by the data as well as by theory. Probabilityvalues (and signicance tests) are not accurate in thissituation, but power still was estimated to help identiya reasonable model size. A model with 28 predictors(up to 14 predictors and their product variables) thatexplains 40% o the variance will have power o 0.80to test individual predictors that uniquely explain at

least 3% o the variance i using a liberal alpha levelo 0.20.

Figure1.ConceptualModelofFactorsInfluencingFatigueExperiencedbyWomenWithStageI,II,orIIIABreastCancerNote. Copyright 1987 by Barbara F. Piper. Adapted with permission.

Prior to and During the Initial Chemotherapy Treatment

Regulation and transmission patternsCircadian rhythms of activity

Sleep-wake patterns

Sleep-wakeSubjective sleep-wake

Activity and rest patternsActivity and restSubjective activity and rest

30 Days After the Last Chemotherapy Treatment

Regulation and transmission patternsCircadian rhythms of activity

Activity and rest patternsActivity and restSubjective activity and rest

Fatigue Manifestationsand Dimensions

Sensory Cognitive/mental Affectivemeaning Behavioral/severity

Sleep-wake patterns

Sleep-wakeSubjective sleep-wake


3/11


MeasurementofVariables

Measurements and timing o the key variables usedin the current study are shown in Table 1.

Fatigue: Fatigue was measured using the 22-item mul-tidimensional revised Piper Fatigue Scale (PFS) (Piperet al., 1998). The 22 numerically scaled (010) itemswith word anchors measure our dimensions o subjec-tive atigue (behavioral/severity, sensory, cognitive/

mood, and aective meaning). Excellent validity andreliability have been reported (Fu, LeMone, McDaniel,& Bausler, 2001; Piper et al., 1998). High estimates ointernal consistency reliability (Cronbach alpha) havebeen reported or the subscales and total PFS scores inwomen with early-stage breast cancer (Berger, 1998;Berger et al., 2002). Cronbach alpha in the study or thetotal PFS score ranged rom 0.970.98.

Circadian rhythms of activity and objective sleep-wake and activity-rest:Octagonal MotionloggerTMactigraphs (Ambulatory Monitoring, Inc.) were used toquantiy the movements o the wrist to yield circadian

rhythms o activity, sleep-wake, and activity-rest indica-tors (Mormont et al., 2000). They determine sleep versus

activity based on the assumption that individuals moveless when asleep and more when awake (AmbulatoryMonitoring, Inc., 2005; Ancoli-Israel et al., 2003). Thedevice is the size o a mans wristwatch and is wornon the nondominant wrist. The actigraphs recordedactivity counts in one-minute epochs (intervals) perrecent guidelines (Littner et al., 2003). Actigraphs havebeen reported to have 88% accuracy at distinguishingwakeulness rom sleep when compared to the gold

standard o polysomnography (Cole, Kripke, Gruen,Mullaney, & Gillin, 1992; Mullaney, Kripke, & Messin,1980). The denitions o specic actigraphy variablesare shown in Figure 2.

Subjective sleep-wake: An item rom the daily di-ary,measuring energy upon awakening, was used tomeasure subjective sleep-wake. The item asked the par-ticipants to rate rom 15 When I got up this morning,I elt___ (1 = exhausted, 5 = rereshed). Each morningo data collection, inormation about the prior nightsbed time and morning wake time also were recorded inthe daily diary, and the times were used to distinguish

day and night intervals in actigraphy les (Littner etal., 2003).

Table1.DataCollectionTimetable

VariableGroup Measurement Priora Peakb Reboundc Completiond

Fatigue PFS Day 2 Day 3 Day 1

Circadian rhythms of activityPeak activity24-hour autocorrelation

Actigraphye 48 hours 72 hours 72 hours 168 hours

Objective sleep-wakeTotal sleep time


Subjective sleep-wakeEnergy upon awakening

Daily dysfunction

Daily diary

PSQI

Two days fromdaily diaryf

Day 2

Three days fromdaily diaryf

Three days fromdaily diaryf

Seven days fromdiaryf

Day 1

Objective activity-restSleep percent (day)


Subjective activity-restPCS PCS Day 2 Day 1

Demographic or medical characteristics

Body mass indexHemoglobinLymph node statusKPS

Medical recordMedical recordMedical recordKPS

Day 0Day 0Day 0Day 2

a Two-day measurement period prior to receiving the initial chemotherapy treatmentb Three-day measurement period during days 24 of the initial chemotherapy treatment; PFS measured on day 3c Three-day measurement period during days 57 of the initial chemotherapy treatmentd Seven-day measurement period 30 days following completion of the last chemotherapy treatmente Wrist actigraphyworn continuouslyfMean of daily values

KPSKarnofsky Performance Scale; PCSPhysical Component Summary of the SF-36, version 2; PFSPiper Fatigue Scale; PSQIPittsburgh Sleep Quality Index


4/11


Subjective sleep-wake during the prior month wasmeasured using the Pittsburgh Sleep Quality Index(PSQI) (Buysse, Reynolds, Monk, Berman, & Kuper,1989). A 19-item questionnaire, the PSQI yields a global

score and seven component scores, one o which isdaytime dysunction. Each item is rated on a 03 scalewith global scores ranging rom 021. Higher scoresindicate poorer sleep quality. Excellent internal reli-ability consistency (Cronbach alpha range 0.730.94)and construct validity (convergent, discriminant, anddivergent) have been reported or a sample that in-cluded women with early-stage breast cancer (Bergeret al., 2002; Carpenter & Andrykowski, 1998).

Subjective activity-rest:The Physical ComponentSummary (PCS) of the SF-36, version 2 (Ware, Kosinski,& Dewey, 2000) was used to measure subjective activity-rest. Internal consistency (Cronbach alpha) was calculatedor each o the SF-36 subscales because it is inappropri-ate to calculate internal consistency on the PCS with itspositive and negative weighting. In the sample, Cronbachalpha or the subscales ranged rom 0.690.93.

Demographic and medical characteristics: This ormincluded the demographic and medical characteristics othe sample. Medical charts were reviewed by researchnurses to obtain height, weight, and hemoglobin.

ProcedureforDataCollection

The study was approved by a university institutionalreview board. Participants were randomized to thebehavioral sleep intervention group or the healthy-eating, attention control group. Group randomizationwas stratied by number o planned chemotherapytreatments (4 or > 4) and sel-identied type o sleeper(good or poor), using item number 6 rom the PSQI.Data collection procedures were identical or thetwo groups with the exception o the orms used ordelivery o the intervention. The intervention groupreceived the behavioral sleep intervention rom atrained research nurse. The behavioral sleep interven-

tion involved developing an Individualized Sleep Pro-motion Plan (ISPP) that consisted o sleep restriction,stimulus control, relaxation therapy, and sleep hygienecounseling. The intervention was implemented at therst visit, reinorced on day 8 o each treatment, andrevised 2 days prior to each subsequent treatment andat 30, 60, and 90 days ater completing chemotherapy.Additional intervention details are described in thepilot study reports (Berger et al., 2002; 2003). Partici-

pants in the control group were given equal time andattention and new inormation about healthy eatingand general conversation at each visit.

StatisticalAnalysis

SPSS 13.0 or Microsot Windows was used to per-orm the statistical analysis. To address the rst pur-pose o the study, zero-order correlations were used toevaluate predictors or inclusion. Hierarchical multipleregression was used or exploratory model building topredict the levels o atigue 30 days ater completingchemotherapy treatment rom predictors chosen romthe study variables and sample characteristics.

To address the second purpose o the study, theregression model was extended by adding calculatedproduct variables or each o the continuous predic-tors. To reduce the multicollinearity likely to occurwith inclusion o product variables, predictors werecentered by subtracting the mean o a variable romthe variables value (Cohen, Cohen, West, & Aiken,2002) prior to multiplication by the dummy-codedgrouping variable.

Results

The 96 women included in the sample ranged in agerom 2983. The typical participant was married, livingwith someone who was not a dependent, and White non-Hispanic. All received anthracycline-based (doxorubicin)adjuvant chemotherapy regimens or breast cancer plusa taxane; some received standard dosing (every threeweeks) and others received dose dense (every two weeks)chemotherapy. Demographic characteristics o the sampleare shown in Table 2. The two groups did not signicantly

dier prior to starting chemotherapy treatment.The correlation o each study variable with total PFS

score at treatment completion was evaluated to selectpredictors. A signicance level o p < 0.20 or the correla-tion was used as a cuto or inclusion, ollowing a mod-el-building strategy outlined by Hosmer and Lemeshow(2000). Attention was given to selecting variables romeach variable group included in the study. As indicatedin the power analysis, the sample size o the secondaryanalysis constrained the number o predictors that couldbe included in the model. Multicollinearity among avail-able variables presented an additional limitation. When

24-Hour AutocorrelationThe extent to which activity levels within 24 hours correlate withlevels of another 24 hours (Lentz, 1990); this serves as an indexof robustness of the circadian rhythm (Ambulatory Monitory, Inc.,2005; Rich et al., 2005).

Peak ActivityMean activity level plus the highest activity level of an individualscircadian rhythm of activity (Lentz, 1990)

Sleep Percent (Day)The total number of minutes of sleep during the day divided by

the total wake time (Young-McCaughan et al., 2003).Total Sleep TimeThe number of minutes asleep while in bed (Berger et al., 2005)

Figure2.DefinitionsofVariablesObtainedbyActigraphy


5/11


Table2.DemographicandMedicalCharacteristicsofSample

Characteristic

Intervention(n=53) Control(n=43) Total(N=96)

X SD Range

X SD Range

X SD Range

Age (years) 49.3 9.1 2971 52.4 11.4 3583 50.7 10.3 2983Body mass index 27.9 6.9 1649 29.4 15.0 2039 28.6 16.2 1649Hemoglobin day 1a 13.0 1.2 10.315.2 12.9 11.2 10.315.1 12.9 11.2 10.315.2

Characteristic n % n % n %

RaceBlack 12 14 13 17 15 15White 51 96 40 93 91 95

EthnicityHispanic 11 12 12 15 13 13Non-Hispanic 52 98 41 95 93 97

Marital statusSingle or never married 16 11 14 19 10 10Married 37 70 32 74 69 72Separated or divorced 19 17 14 19 13 14Widowed 11 12 13 17 14 14

EmploymentProfessional 25 47 16 37 41 43Service 13 25 16 37 29 30

Homemaker 18 15 12 15 10 10Retired 15 19 17 16 12 13Unemployed 12 14 11 12 13 13Student 11 12 11 11

Karnofsky Performance Scale score60 11 12 11 1170 12 14 12 15 14 1480 15 19 15 12 10 1090 17 32 15 35 32 33100 28 53 21 49 49 51

Cancer stageI 13 16 15 12 18 18II 39 74 27 63 66 69IIIA 11 21 11 26 22 23

Lymph node statusNegative 12 23 13 30 25 26Positive 13 32 60 20 47 52 54Positive 49 19 17 10 23 19 20

Surgical procedureLumpectomy 20 38 18 42 38 40Modified mastectomy 13 25 12 28 25 26Modified mastectomy with reconstruction 20 38 13 30 33 34

Chemotherapy protocolAdriamycin and cytoxan times four followed

by taxane times four every two weeks33 62 26 61 59 62

Adriamycin and cytoxan times four fo l-lowed by taxane times four every threeweeks

13 25 19 21 22 23

Adriamycin and cytoxan times four every

three weeks followed by taxane everyweek times 12

14 18 14 19 18 18

Adriamycin and cytoxan times four everytwo weeks followed by taxane every weektimes 12

12 14 13 17 15 15

Adriamycin and cytoxan times six followedby taxane times four every two weeks

11 12 11 11

Adriamycin and cytoxan times four every twoweeks followed by taxane times four everythree weeks

11 12 11 11

a N = 82 because of missing data.

Note. Because of rounding, not all percentages total 100.

Note. No significant differences between groups on any variable at alpha = 0.05


6/11


predictors within a variable group and measured at thesame time were highly correlated with each other, theone with the strongest correlation with atigue 30 daysater completing treatment was selected.

A total o 13 predictors were included in the predic-tion model along with a dummy-coded variable repre-senting the RCT group assignment (0 = intervention).Predictor variables measured prior to chemotherapywere atigue (PFS), body mass index (BMI), hemoglobin,

lymph node status, daytime dysunction (PSQI), PCS,KPS score, and total sleep time at night (actigraph).Vari-ables measured at the peak were atigue (PFS) and peakactivity (actigraph). Variables measured at the reboundwere 24-hour autocorrelation (actigraph), energy uponawakening (daily diary), and sleep percent during theday (actigraph). Table 3 details the descriptive statisticso the 13 predictors as well as the zero-order correlationso the predictors with the outcome o atigue 30 daysater completing treatment.

O the 96 participants included in the secondaryanalysis, 60 had complete data; 35 cases were missing

a ew values, usually on 1 or 2 variables. One case hadmissing values on multiple variables and, thereore, wasexcluded. Those with missing data diered (p < 0.20)rom those with complete data on day 1 hemoglobin(p = 0.03), PCS prior to treatment (p = 0.13), and amounto energy upon awakening at rebound (p = 0.11).Missing values were imputed using the expectation

maximization in SPSS Missing Values Analysis 13.0module, which assumes data are missing at random.To strengthen the imputation model, a small numbero additional variables correlated with the predictorshaving incomplete data were included in calculationo missing values (Collins, Schaer, & Kam, 2001). Topreserve possible interactions o the intervention groupwith other predictors, imputation was conducted oreach group independently and then combined into a

single dataset.The list o predictors was organized temporally intoour groups or entry into the hierarchical regressionmodel, with the intervention group added at a ithstep. This allowed evaluation o whether actors occur-ring at later points added signicantly to what could beexplained by previously existing characteristics.

MultipleRegressionResultsofFatigueRegressedonSelectedPredictors:Model1

Because o the exploratory nature o the analysis, a lib-eral signicance level o p < 0.20 was used to evaluate in-

dividual regression coecients as well as changes in R2 assets o variables were added. Table 4 presents the modelsummary or each step o the model. Change in R2 as setso variables were added was signicant only at step 1(p < 0.001), in which variables measured prior to treat-ment were entered, and at step 3 (p = 0.05), in whichatigue and peak activity at the peak (days 24) were

Table3.DescriptiveStatisticsandPairwiseCorrelationsof13PredictorsWithFatigueMeasured30DaysFollowingCompletionofChemotherapyTreatmentbyGroups

MeasurementTime

KeyVariablesandMeasurements

StudyGroups

Totala Interventionb Controlc

X SD r

X SD r

X SD r

Prior Fatigue (PFS total score) 112.73 11.99 0.39*** 112.89 12.01 0.35* 112.53 11.98 0.46**Body mass index 128.59 16.16 0.15** 127.92 16.92 0.08 129.42 15.01 0.27Day 1 hemoglobin 112.94 11.18 0.19** 112.96 11.21 0.21 112.93 11.15 0.15Lymph node status 0.10** 0.04 0.23Daytime (PSQI)

dysfunction110.58 11.02 0.23* 110.90 10.63 0.36* 110.98 10.74 0.10

Physical componentsummary

143.58 19.58 0.28* 143.56 19.73 0.22 143.60 19.51 0.35*

KPS score 0.21* 0.23 0.20

Total sleep time 421.39 19.03 0.24* 430.26 83.66 0.17 410.30 95.22 0.30

Peak Fatigue (PFS total score) 114.85 12.23 0.33** 115.00 12.24 0.39** 114.68 12.22 0.27Peak activity 176.46 46.17 0.13 176.94 49.34 0.31* 175.90 42.86 0.08

Rebound 24-hour autocorrelation 110.43 10.16 0.24* 110.43 10.17 0.32* 110.41 10.15 0.16Energy upon awakening 113.41 10.92 0.35*** 113.31 10.93 0.44** 113.54 10.91 0.26Sleep percent (day) 111.47 18.77 0.18 112.24 10.01 0.39** 110.58 17.08 0.11

a Total sample (N = 96) range of sample sizes for correlations was 7789.b Intervention group (n = 53) range of sample sizes for correlations was 4048.c Control group (n = 43) range of sample sizes for correlations was 3841.

*p < 0.05; **p < 0.01; ***p < 0.001

KPSKarnofsky Performance Scale; PFSPiper Fatigue Scale; PSQIPittsburgh Sleep Quality Index


7/11


entered. In some cases, individual predictors were sig-nicant even when the test o the set o variables addedwas not.

Coecients or Model 1 with all 14 predictors are pre-sented in Table 5. Four variables were signicant predic-tors o atigue at treatment completion: atigue prior totreatment (p = 0.007), total sleep time prior to treatment(p= 0.13), atigue at the peak (p = 0.12), and amount o

energy upon awakening at rebound (p = 0.09). Each othese variables also had been a signicant predictor atthe step in which it entered the model and remainedso as other variables were added. The remaining vari-ables were not signicant at any step, with the excep-tion o BMI. BMI was not signicant in the nal model(p = 0.59), but was signicant (p < 0.12) when it enteredthe model in the rst step. It no longer made a uniquecontribution once step 2 variables were added.

MultipleRegressionResultsofFatigueRegressedonSelectedPredictorsandProduct

Variables:Model2

To evaluate whether relationships o the predictorsand atigue 30 days ater completing chemotherapyvaried by group, product variables were added in asixth step. This resulted in large variance infation ac-tors (VIF), indicating high multicollinearity. Remov-ing product variables with p values greater than 0.20reduced VIF values with little eect on the regressioncoecients o the remaining predictors. The reducedmodel is presented as Model 2 in Table 6.

Adding the ive product variables signiicantly in-creased the R2 by 0.10 (p= 0.03) to 0.440 (adjusted R2 =0.299, p < 0.001), suggesting signicant dierences be-tween the groups in the relationships o some o the pre-dictors with atigue 30 days ater completing treatment,controlling or other predictors in the model.

The only variable in the reduced model rom the earliersteps that was signicant independently was the amounto energy upon awakening at rebound (p < 0.02). Thosewith less energy upon awakening at rebound (days 57)had higher atigue at treatment completion with no di-erence between groups. Although peak activity at the

peak (days 24) was signiicant (p < 0.07), it was notinterpreted because o its signicant interaction with thegroup.

To interpret the group dierences in prediction,separate regression coecients or each predictor werecalculated. The unstandardized coecient or the origi-nal variable represents the relationship o that variableto atigue 30 days ater completing treatment or theintervention group (coded 0). The regression coecient

or the associated product variable indicates the amountthat the control groups coecient diers rom that othe intervention group. Adding the two values yields thecoecient or the control group (coded 1). By running themodel again with the coding reversed, a p value or thecoecient or the control group was obtained.

Higher atigue prior to treatment was related signi-cantly to higher atigue 30 days ater completing treat-ment in the control group (0.484, p < 0.001) but not in theintervention group (0.123, p = 0.46). A similar pattern wasobserved or BMI prior to treatment, with a signicantpositive relationship or the control group (0.097, p=

0.19) but not or the intervention group (0.016, p = 0.72).More positive lymph nodes predicted higher atigue attreatment completion in the control group (0.935, p =0.04), but not in the intervention group (0.441, p = 0.34).Lower daytime dysunction (PSQI) prior to treatmentpredicted higher atigue in the control group (0.744, p =0.11), but not in the intervention group (0.492, p = 0.39).Peak activity at the peak o treatment in the interventiongroup had a coecient o 0.012 (p = 0.07), indicatingthat in the group, lower peak activity during the peak othe initial treatment predicted higher atigue at treatmentcompletion. The coecient in the control group was notsignicant (0.008, p = 0.31).

Discussion

This secondary analysis was able to identiy thepredictors o atigue 30 days ater completing anthra-cycline-based plus taxane adjuvant chemotherapyor breast cancer. Four predictors o higher atigue 30days ater completing treatment were identied. Thesepredictors were less total sleep time prior to treatment,higher atigue prior to and at the peak o treatment, and

less energy upon awakening at rebound rom the initialtreatment. Fatigue prior to treatment was the strongestpredictor o later atigue.

This is the irst report that identiies predictors oatigue 30 days ater completing chemotherapy rommeasurements collected prior to and during the initialweek o treatment.

Previous descriptive studies have reported asso-ciations between atigue and other variables selected orinclusion in the secondary analysis. Ancoli-Israel et al.(2006) described patterns o atigue, sleep, and circadianrhythms in 85 women with breast cancer prior to the

Table4.Model1:SummaryofMultipleRegressionResultsforFatigueRegressedonSelectedPredictorsatTreatmentCompletion

Step R 2 AdjustedR2 R2Change p(R2change)

1 0.206 0.171 0.206 < 0.0012 0.256 0.187 0.050 < 0.2273 0.306 0.224 0.050 < 0.0534 0.335 0.228 0.029 < 0.3235 0.339 0.223 0.004 < 0.510

N = 95


8/11


initial chemotherapy treatment. Although no relation-ships were ound between atigue and circadian rhythmso activity and objective sleep-wake prior to initialtreatment, signicant relationships exist between atigueand subjective sleep-wake and activity-rest measures.Berger (1998) described patterns o lower circadianrhythms o activity that were associated with higheratigue during the rst our days ater the initial treat-ment. The current study is important because it adds tothe understanding o the variables related to atigue priorto and during initial chemotherapy that are predictiveo atigue ater completion o chemotherapy treatment.This is an important area o study because receiving ad-juvant chemotherapy, compared to radiation, has been

shown to signicantly predict higher atigue at treatmentcompletion (Andrykowski et al., 2005), and atigue is themost common symptom reported by survivors (Janz etal., 2007).

The secondary analysis also determined the dier-ences between a behavioral sleep intervention and ahealthy-eating, attention control group in predictingatigue 30 days ater completing chemotherapy treat-ments. When group assignment was added to the keyvariables, ve o the variables predicted higher atiguebased on the group. For control group participants,values prior to treatment o higher atigue and daytime

dysunction, higher BMI, and more positive lymphnodes predicted higher atigue 30 days ater complet-ing treatment. Individual actors prior to initiation ochemotherapy, such as atigue levels, BMI, and lymphnode status, appear to play a role above and beyond thatcontributed by the chemotherapy regimen in womensperceptions o atigue ater completing treatment.

For sleep intervention participants, lowerpeak ac-tivity (less activity during the day and more at nightas measured by actigraphy) was the only predictor ohigher atigue 30 days ater completing treatment. Theresult suggests that those in the sleep intervention groupwho were able to maintain day activity balanced withnight sleep at the peak o the initial treatment beneted

most rom the intervention. In addition, these resultssuggest that the sleep intervention was able to decreasethe eects o the actors that predicted higher atiguein the control group at treatment completion. No otherstudies have used a behavioral sleep intervention toreduce atigue in women with breast cancer; thereore,the ndings cannot be compared. A limited number osmall studies have used a behavioral sleep interventionto relieve atigue in patients with cancer (Berger et al.,2002, 2003; Mitchell et al., 2007).

The results are useul or hypothesis generation andguiding uture study. However, interpretation o the

Table5.Model1a:MultipleRegressionResultsofFatigueRegressedonSelectedPredictors30DaysAfterCompletingChemotherapyTreatment

Step Variable

Coefficient

B SE Beta p

Priorb Fatigue (PFS Total) 0.331 0.120 0.307 0.007**Body mass index 0.021 0.038 0.058 0.587**Hemoglobin 0.030 0.184 0.016 0.870**Lymph node status 0.183 0.324 0.057 0.574**

Priorb Daytime dysfunction (PSQI) 0.120 0.353 0.037 0.735**PCS 0.002 0.028 0.007 0.951**Karnofsky Performance Scale 0.080 0.278 0.033 0.773**Total Sleep Time 0.004 0.003 0.171 0.128**

Peakc Fatigue (PFS Total) 0.164 0.104 0.167 0.120**Peak activity 0.004 0.005 0.087 0.423**

Reboundd 24-hour autocorrelation 1.009 1.852 0.071 0.587**Amount of energy upon awakening 0.454 0.262 0.191 0.087**Sleep percent (day) 0.007 0.034 0.028 0.832**

Grouping variable Grouping variable 0.272 0.412 0.062 0.510**

N = 95

*p < 0.20; **p < 0.05a R2

(14,80)= 0.339, p = 0.001 (adjusted R2 = 0.223)

b Two-day measurement period prior to receiving the initial chemotherapy treatmentc Three-day measurement period during days 24 of the initial chemotherapy treatmentd Three-day measurement period during day 57 of the initial chemotherapy treatment

PCSPhysical Component Summary of the SF-36, version 2; PFSPiper Fatigue Scale; PSQIPittsburgh Sleep Quality Index

Note. Table presents results for all predictors in model simultaneously in step 5.


9/11


model was limited by small sample size and multicol-linearity. Another limitation is the lack o racial or ethnicdiversity in the sample. A strength o the secondaryanalysis is that although missing data are inherentin studies o human subjects and present challenges,methods were used to maximize the use o availabledata or analysis.

RecommendationsforResearch

Using the entire data set o 219 participants, theprediction model could be reanalyzed with increasedpower to determine stability o the predictors with ad-ditional participants who received only anthracycline-based chemotherapy treatments. Also, data analysisrom one or all chemotherapy treatments and at latertimes (60 and 90 days ater treatment completion andone year ater the initial treatment) may urther explainthe relationships between circadian rhythms o activ-

ity, sleep-wake, activity-rest, demographic or medicalcharacteristics, and atigue.

ImplicationsforPractice

NCCN (2008) cancer-related atigue guidelines recom-mended education and counseling regarding atigue

patterns as central to eective management. Resultso the current study provided inormation regardingvariables prior to treatment that predicted higher atigue30 days ater completing treatment. The nding assistswith identiying those most in need o early and inten-sive teaching on strategies to manage atigue. The stron-gest predictor was the atigue level o the individualprior to receiving the initial chemotherapy treatment.This supports the simple, and yet widely underusedrecommendation o the NCCN cancer-related atigueguidelines to screen all patients or atigue prior to treat-ment and at regular intervals.

Table6.Model2a:MultipleRegressionResultsofFatigueRegressedonSelectedPredictorsandProductVariables30DaysAfterCompletingChemotherapyTreatment

Step Variable

Coefficients

B SE Beta p

Priorb Fatigue (PFS Total) 0.123 0.165 0.114 0.459**Body mass index 0.016 0.044 0.044 0.724**Hemoglobin 0.142 0.186 0.075 0.448**Lymph node status 0.441 0.459 0.138 0.340**

Priorb Daytime dysfunction (PSQI) 0.492 0.566 0.153 0.388**PCS 0.006 0.027 0.024 0.835**Karnofsky Performance Scale 0.253 0.274 0.104 0.360**Total sleep time 0.003 0.003 0.111 0.327**

Peakc Fatigue (PFS total) 0.113 0.103 0.115 0.274**Peak activity 0.012 0.006 0.250 0.065**

Reboundd 24-hour autocorrelation 0.354 1.852 0.025 0.849**Amount of energy upon awakening 0.638 0.261 0.269 0.017**Sleep percent (day) 0.033 0.036 0.130 0.366**

Grouping variable Grouping variable 0.269 0.396 0.062 0.498**

Product variables Fatigue (PFS) prior times group 0.361 0.220 0.219 0.105**Body mass index times group 0.097 0.073 0.150 0.189**Lymph node status times group 1.376 0.626 0.310 0.031**Daytime dysfunction (PSQI) times group 1.236 0.749 0.280 0.103**Peak activity times group 0.019 0.010 0.254 0.051**

N = 95

*p < 0.20; **p < 0.05a R2

(19, 75)= 0.440, p < 0.001 (adjusted R2 = 0.299)

b Two-day measurement period prior to receiving the initial chemotherapy treatment.c Three-day measurement period during days 24 of the initial chemotherapy treatment.d Three-day measurement period during day 57 of the initial chemotherapy treatment.

PCSPhysical Component Summary of the SF-36, version 2; PFSPiper Fatigue Scale; PSQIPittsburgh Sleep Quality Index

Note. Table presents results for all predictors and five product variables in model simultaneously in step 6. Variable name X group refers to

the product variable. Model results including 13 product variables were R2(27, 67) = 0.466, p = 0.006 (adjusted R2 = 0.250). Product variableswith p > 0.30 were removed. Model 2 represents the regression coefficients for the product variables retained.


10/11


NCCN (2008) suggested sel-monitoring o atiguelevels. Because atigue levels during the peak o treat-ment were ound to be another strong predictor oatigue at treatment completion, regular screening oratigue should include an assessment o the patientsatigue levels at the peak o treatment. The amounto energy upon awakening at rebound also predictedatigue at treatment completion. A sel-report diarycould be completed or seven days ater the initial

treatment, including ratings o atigue and an itemassessing the energy upon awakening, to urtherocus teaching on promotion o quality sleep duringthe next visit to clinic as recommended by the NCCNguidelines.

Less total sleep time prior to treatment also predictedhigher atigue at treatment completion. An evidence-based intervention, such as the behavior sleep inter-vention tested in the current study, can be initiated inwomen who identiy themselves as poor sleepers priorto treatment. Patient education to promote qualitysleep or all women starting adjuvant chemotherapy or

breast cancer is a rst step. The ONS PEP card or sleep-wake disturbances provides the nurse clinician withinormation that can be taught to patients to improvetheir sleep quality (Page, Berger, & Johnson, 2006).Instruction at the initiation o chemotherapy maydecrease the atigue experienced at the peak, aectthe amount o energy upon awakening at rebound,and encourage daytime activity(Mitchell et al., 2007).

Ongoing assessment o atigue and sleep quality willidentiy individuals who develop these problems a-ter the initial chemotherapy treatment. Intervening asearly as possible in the course o chemotherapy oersthe potential or lower atigue at treatment comple-tion, thereby possibly decreasing the disability andquality-o-lie issues associated with this distressulsymptom.

The authors greatfully acknowledge Barbara F. Piper, DNSc, RN,AOCN, FAAN, Lynne Farr, PhD, Nancy Waltman, PhD, RN, andAmy Kessinger, MD, for their support in this work.

Kimberly K. Wielgus, PhD, APRN, BC, is an assistant professor,Ann M. Berger, PhD, RN, AOCN, FAAN, is an associate profes-sor, and Melody Hertzog, PhD, is an assistant professor, all at theUniversity of Nebraska Medical Center in Omaha. This researchwas funded by the Breast Cancer Training Program Fellowship,the U.S. Army Breast Cancer Research Program Award (20012002; DADM17-00-1-0361), an American Cancer Society Doc-toral Scholarship in Cancer Nursing (20022004; DSCN-02-213-01-SCN), and the National Aeronautics and Space AdministrationNebraska Space Grants Scholarships (2002, 2003) awardedto Wielgus and by a National Institutes of Health and the Na-

tional Institute of Nursing Research (#5R01NR007762-05) grantawarded to Berger. No financial relationships to disclose. Men-tion of specific products and opinions related to those productsdo not indicate or imply endorsement by the Oncology NursingForum or the Oncology Nursing Society. Wielgus can be reachedat [email protected], with copy to editor at [email protected]. (Submitted December 2007. Accepted for publication April1, 2008.)

Digital Object Identifier: 10.1188/09.ONF.38-48

ReferencesAmbulatory Monitoring, Inc. (2005). Action4(c) procedure manual.

Ardsley, NY: Author.Ancoli-Israel, S., Cole, R., Alessi, C., Chambers, M., Moorcrot, W., &

Pollak, C.P. (2003). The role o actigraphy in the study o sleep andcircadian rhythms. Sleep, 26(3), 342392.

Ancoli-Israel, S., Liu, L., Marler, M.R., Parker, B.A., Jones, V., Sadler,G.R., et al. (2006). Fatigue, sleep, and circadian rhythms prior tochemotherapy or breast cancer. Supportive Care in Cancer, 14(3),201209.

Andrykowski, M.A., Schmidt, J.E., Salsman, J.M., Beacham, A.O., &Jacobsen, P.B. (2005). Use o a case-denition approach to identiycancer-related atigue in women undergoing adjuvant therapy orbreast cancer.Journal of Clinical Oncology, 23(27), 66136622.

Berger, A.M. (1998). Patterns o atigue and activity and rest duringadjuvant breast cancer chemotherapy. Oncology Nursing Forum,

25(1), 5162.Berger, A.M., & Farr, L. (1999). The infuence o daytime inactivity andnighttime restlessness on cancer-related atigue. Oncology NursingForum, 26(10), 16631671.

Berger, A.M., Parker, K.P., Young-McCaughan, S., Mallory, G.A.,Barsevick, A.M., Beck, S.L., et al. (2005). Sleep/wake disturbancesin people with cancer and their caregivers: State o the science[Online exclusive]. Oncology Nursing Forum, 32(6), E98E126.Retrieved November 25, 2008, rom http://ons.metapress.com/content/7244v4525u2j6408/ulltext.pd

Berger, A.M., Von Essen, S., Kuhn, B.R., Piper, B.F., Agrawal, S.,Lynch, J.C., et al. (2003). Adherence, sleep, and atigue outcomesater adjuvant breast cancer chemotherapy: Results o a easibilityintervention study. Oncology Nursing Forum, 30(3), 513522.

Berger, A.M., Von Essen, S., Kuhn, B.R., Piper, B.F., Farr, L., Agrawal, S.,

et al. (2002). Feasibility o a sleep intervention during adjuvant breast

cancer chemotherapy. Oncology Nursing Forum, 29(10), 14311441.Bower, J.E., Ganz, P.A., Desmond, K.A., Bernaards, C., Rowland, J.H.,

Meyerowitz, B.E., et al. (2006). Fatigue in long-term breast cancersurvivors: A longitudinal investigation. Cancer, 106(4), 751758.

Buysse, D.J., Reynolds, C.F., III, Monk, T.H., Berman, S.R., & Kuper,D.J. (1989). The Pittsburgh Sleep Quality Index: A new instrument orpsychiatric practice and research. Psychiatry Research, 28(2), 193213.

Carpenter, J.S., & Andrykowski, M.A. (1998). Psychometric evalua-tion o the Pittsburgh Sleep Quality Index.Journal of PsychosomaticResearch, 45(1), 513.

Cohen, P., Cohen, J., West, S.G., & Aiken, L.S. (2002).Applied multipleregression/correlation analysis for the behavioral sciences (3rd ed.).Mahwah, NJ: Lawrence Erlbaum Associated.

Cole, R.J., Kripke, D.F., Gruen, W., Mullaney, D.J., & Gillin, J.C. (1992).

Automatic sleep/wake identiication rom wrist activity. Sleep,15(5), 461469.Collins, L.M., Schaer, J.L., & Kam, C.M. (2001). A comparison o inclu-

sive and restrictive strategies in modern missing data procedures.Psychological Methods, 6(4), 330351.

Ferguson, T., Wilcken, N., Vagg, R., Ghersi, D., & Nowak, A.K. (2007).Taxanes or adjuvant treatment o early breast cancer. CochraneDatabase for Systematic Reviews, 17(4), CD004421.

Fu, M., LeMone, P., McDaniel, R.W., & Bausler, C. (2001). A multi-variate validation o the dening characteristics o atigue. NursingDiagnosis, 12(1), 1527.

Hosmer, D.W., & Lemeshow, S. (2000).Applied logistic regression (2nded.). New York: Wiley.

Janz, N.K. , Muhahid, M., Chung, L.K. , Lantz, P.M., Hawley, S.T.,Morrow, M., et al. (2007). Symptom experience and quality o lie


11/11


o women ollowing breast cancer treatment. Journal of WomensHealth, 16(9), 13481361.

Jemal, A., Siegel, R., Ward, E., Hao, Y., Xu, J., Murray, T., et al. (2008). Can-cer statistics, 2008. CA: A Cancer Journal for Clinicians, 58(2), 7196.

Lee, K., Cho, M., Miaskowski, C., & Dodd, M. (2004). Impaired sleepand rhythms in persons with cancer. Sleep Medicine Reviews, 8(3),199212.

Lentz, M.J. (1990). Time-series analysiscosinor analysis: A specialcase. Western Journal of Nursing Research, 12(3), 408412.

Littner, M., Kushida, C.A., Anderson, W.M., Bailey, D., Berry, R.B.,Davila, D.G., et al. (2003). Practice parameters or the role o actig-raphy in the study o sleep and circadian rhythms: An update or

2002. Sleep, 26(3), 337341.Mitchell, S.A., Beck S.L., Hood L.E., Moore, K., & Tanner, E.R. (2007).

Putting Evidence Into Practice: Evidence-based interventions oratigue during and ollowing cancer and its treatment. ClinicalJournal of Oncology Nursing, 11(1), 99113.

Mormont, M.C., Waterhouse, J., Bleuzen, P., Giacchetti, S., Jami, A.,Bogdan, A., et al. (2000). Marked 24-h rest/activity rhythms areassociated with better quality o lie, better response, and longersurvival in patients with metastatic colorectal cancer and goodperormance status. Clinical Cancer Research, 6(8), 30383045.

Mullaney, D.J., Kripke, D.F., & Messin, S. (1980). Wrist-actigraph esti-mation o sleep time. Sleep, 3(1), 8392.

National Comprehensive Cancer Network. (2008). NCCN ClinicalPractice Guidelines in OncologyTM: Cancer-related fatigue [v.1.2008].

Retrieved November 25, 2008, rom http://www.nccn.org/proessionals/physician_gls/PDF/atigue.pd

Page, M., Berger, A.M., & Johnson, L.B. (2006). Putting Evidence IntoPractice: Evidence-based interventions or sleep-wake distur-bances. Clinical Journal of Oncology Nursing, 10(6), 753767.

Piper, B.F., Dibble, S.L., Dodd, M.J., Weiss, M.C., Slaughter, R.E., &Paul, S.M. (1998). The revised Piper Fatigue Scale: Psychometricevaluation in women with breast cancer. Oncology Nursing Forum,25(4), 677684.

Piper, B.F., Lindsey, A.M., & Dodd, M.J. (1987). Fatigue mechanismsin cancer patients: Developing nursing theory. Oncology NursingForum, 14(6), 1723.

Prue, G., Rankin, J., Allen, J., Gracey, J., & Cramp, F. (2006). Cancer-relatedatigue: A critical appraisal. European Journal of Cancer, 42(7), 846863.

Rich, T., Innominato, P.F., Boerner, J., Mormont, M.C., Iacobelli, S.,Baron, B., et al. (2005). Elevated serum cytokines correlated withaltered behavior, serum cortisol rhythm, and dampened 24-hourrest-activity patterns in patients with metastatic colorectal cancer.Clinical Cancer Research, 11(5), 17571764.

Ware, J.E., Kosinski, M., & Dewey, J.E. (2000).How to score version twoof the SF-36Health Survey. Lincoln, RI: QualityMetric.

Young-McCaughan, S., Mays, M.Z., Arzola, S.M., Yoder, L.H., Drami-ga, S.A., Leclerc, K.M., et al. (2003). Change in exercise tolerance,activity and sleep patterns, and quality o lie in patients withcancer participating in a structured exercise program. OncologyNursing Forum, 30(3), 441454.

Documents

Predictors of Fatigue 2009 14ci