24 Journal of Pain and Symptom Management Vol. 42 No. 1 July 2011

Original Article

Symptom Clusters in Advanced CancerAna Jim�enez, PhD, Rosario Madero, PhD, Alberto Alonso, PhD,Virginia Mart�ınez-Mar�ın, MD, Yolanda Vilches, MD, Beatriz Mart�ınez, PhD,Marta Feliu, MD, Leyre D�ıaz, MD, Enrique Espinosa, PhD, and Jaime Feliu, PhDGetafe University Hospital (A.J), Getafe; and La Paz University Hospital (R.M., A.A., V.M.-M., Y.V.,

B.M., M.F., L.D., E.E., J.F.), La Paz, Spain

Abstract

Context. Patients with advanced cancer often experience multiple concurrent

symptoms. Few studies have explored symptom clusters (SCs) in this population.Objectives. The aim of the present study was to explore SCs in advanced cancer,

evaluate the characteristics associated with various clusters, and determine theirrelationship to survival.

Methods. This study included patients in the palliative care program of theHospital Universitario La Paz from 2003 to 2005. The Edmonton SymptomAssessment System and a supplement including 13 other symptoms were usedto detect symptoms. Principal component analysis was performed to determinesymptom relationships and compare SCs with associated parameters.

Results. In total, 406 patients were included, 61% men and 39% women. Themedian age was 66.4 (range 18e95). The most common primaries weregastrointestinal (35%), lung (25%), genitourinary (8%), breast (5%), and headand neck (5%) carcinomas. The following clusters were identified: confusion(cognitive impairment, agitation, urinary incontinence), neuropsychological(anxiety, depression, and insomnia), anorexia-cachexia (anorexia, weight loss,and tiredness), and gastrointestinal (nausea and vomiting). The presence of theseSCs was influenced by primary cancer site, gender, age, and performance status.Survival was related to the number of SCs present in a given patient: zero SC, 52days; one SC, 38 days; two SCs, 23 days; and three to four SCs, 19 days; P< 0.001.

Conclusion. Different SCs can be identified in patients with advanced cancer.These SCs are influenced by primary cancer site, gender, age, and EasternCooperative Oncology Group performance status, and they can have prognosticvalue. J Pain Symptom Manage 2011;42:24e31. � 2011 U.S. Cancer Pain ReliefCommittee. Published by Elsevier Inc. All rights reserved.

Key Words

Symptom cluster, advanced cancer, survival, prognostic, gender, age, performance status

Address correspondence to: Jaime Feliu, PhD, Depart-ment of Medical Oncology, La Paz University Hospi-tal, Paseo de la Castellana 261, Madrid, Spain.E-mail: jaimefeliu@hotmail.com

Accepted for publication: October 17, 2010.

� 2011 U.S. Cancer Pain Relief CommitteePublished by Elsevier Inc. All rights reserved.

IntroductionStudies of cancer-related symptoms in the

palliative care setting have traditionally beenfocused on isolated symptoms. Patients withadvanced cancer do not usually have one

0885-3924/$ - see front matterdoi:10.1016/j.jpainsymman.2010.10.266

Vol. 42 No. 1 July 2011 25Symptom Clusters in Advanced Cancer

symptom, however, which may explain whytreating one symptom may not necessarily im-prove quality of life.1

The term ‘‘symptom cluster’’ (SC) has beendefined as two or more interrelated symptomsthat present together, independent of otherSCs, and may or may not suggest a commonetiology or underlying mechanism.2,3 Investi-gation of SCs has been performed in differentscenarios: early-stage disease,4,5 specific prima-ries,6,7 specific metastatic sites,8,9 or even long-term survivors.10 However, very few studieshave been published in patients with advancedcancer,11,12 a population in which symptomsare numerous and intense.13,14 The experi-ence of multiple symptoms in this populationmay be associated with reduced quality oflife, impaired physical and social functioning,and decreased response to therapy.15e17 Asa consequence, it would be interesting toknow which symptoms are likely to cluster,whether there is an influence of SCs on thephysiology and appearance of some symptoms,and how SCs affect outcome. Better knowledgeabout SCs may help to optimize managementby designing therapies aimed at several simul-taneous symptoms.

Different SCs have been identified in patientswith advanced cancer. One study described twoSCs: 1) fatigue, drowsiness, nausea, decreasedappetite, and dyspnea, and 2) anxiety and de-pression.12 Another study identified up to sevenSCs: 1) fatigue (anorexia-cachexia), 2) neuro-psychological, 3) upper gastrointestinal, 4) nau-sea and vomiting, 5) aerodigestive, 6) debility,and 7) pain.11 The prognostic value of SCs onsurvival has not been investigated so far.

Our objective was to identify the presenceand composition of SCs in a population of pa-tients with advanced cancer, determine pa-tients’ characteristics associated with SCs, andassess the prognostic value of SCs.

Patients and MethodsPatient Selection

Between February 2003 and December 2005,437 consecutive patients were included in thepalliative care program of the Hospital Univer-sitario La Paz. Eligible patients had 1) evi-dence of progressive malignancy, 2) absenceof specific therapy that could prolong survival,

and 3) a life expectancy of less than sixmonths. They were older than 18 years andable to provide answers in an interview. Bothinpatients and outpatients were included.Nineteen patients did not consent to collect-ing their data, eight moved somewhere elsein the first week and were lost to follow-up,and four were not able to answer the interview.Finally, 406 patients were included.

InstrumentsTwenty-two symptoms were evaluated. The

Edmonton Symptom Assessment System(ESAS)18 was supplemented by questions aboutthe following: drowsiness, sweating, fever,weight loss, cognitive impairment, agitation,constipation, insomnia, dysphagia, hiccups,vomiting, and urinary or stool incontinence.These 13 symptoms added information to thatprovided by the ESAS and are systematically in-cluded in ourmedical records. A difference wasestablished between nausea (sensation of un-ease and discomfort in the upper stomachwith an urge to vomit) and vomiting (forcefulexpulsion of the contents of one’s stomach)and also between insomnia (persistent difficultyfalling asleep and/or difficulty staying asleep)and drowsiness (state of near-sleep, a strong de-sire for sleep). These 13 symptoms were scoredon a Likert scale: not at all¼ 1, a little¼ 2,moderate¼ 3, severe¼ 4, and extremelystressing¼ 5. Significant symptoms were thoserated$1 on ESAS or$2 on Likert scale scores.Performance status was assessed by the EasternCooperative Oncology Group (ECOG) scale.

Data were retrieved the first time the patientwas evaluated by the palliative medicine team.Patients gave informed consent for the inter-view and data collection. The study designwas in accordance with local ethics committeerules.

Statistical AnalysisThe statistical analysis was performed using

SPSS version 12.0 (SPSS Inc., Chicago, IL),and R 2.2 with the Design software packageversion 2.0e12. Linear transformation from1e5 to 0e10 was used to compare symptomsdetected with scales for drowsiness, diapho-resis, fever, weight loss, cognitive impairment,agitation, constipation, insomnia, dysphagia,hiccups, vomiting, and urinary or stool incon-tinence. Descriptive statistics were used to

Table 1Baseline Patient Characteristics

Characteristics n (%)

Age (years), median (range) 66 (18e95)

Gender (%)Male 247 (61)Female 159 (39)

Number of symptoms 9 (2e18)

ECOG performance status (%)0 01 26 (6)2 95 (23)3 193 (48)4 93 (23)

Primary cancer site (%)Lung 102 (25)Colorectal 77 (19)Other gastrointestinal tumors 63 (16)Breast 21 (5)Head and neck 18 (4)Hematologic 12 (3)Urinary 22 (5)Female reproductive 11 (3)Others 80 (20)

Number of metastatic sites (%)0 23 (6)1 156 (38)>1 227 (66)

26 Vol. 42 No. 1 July 2011Jim�enez et al.

summarize baseline patient demographics, dis-ease characteristics, and symptom scores.

To detect an SC and examine whether rela-tionships existed among symptoms, principalcomponent analysis (PCA) with varimax rota-tion was conducted on the intensity scores ofthe 22 symptoms. This was followed by ascreen plot analysis of eigenvalues againstthe number of factors that accounted formost variables. Communalities were used toobserve the level of shared variance betweenitems. The Kaiser-Meyer-Oklin measure wascalculated to assess sampling adequacy(scores >0.60 indicate adequate sample sizefor the analysis). The internal consistenceand reliability of the derived SC were assessedwith the Cronbach’s alpha coefficient. De-rived factors were discussed and interpretedamong the members of the research team,and final factor structure included factorswith clinically contextual relevance.

We used analysis of variance to test the diff-erences among patient groups in subjectivesymptoms and associated clinical parameters.Survivalwasmeasured fromthedateof inclusionuntil death or last follow-up. The Kaplan-Meiermethod was applied to calculate the influenceof categorical variables on survival.

ResultsFour-hundred six patients were included: 247

men and 159 women. The mean age was 66.4years (range 18e95). The most frequent tumorwas lung cancer (25%), followed by colorectal(24%), other digestive tumors (14%), andbreast cancer (8%). The median survival was26.5 days (range 1e547). Ninety-five percentof patients died within six months. Seventy-one percent of patients had an ECOG perfor-mance status of 3 or 4 (Table 1). The mediannumber of symptoms was 9 (2e18) and themost common symptoms were asthenia, an-orexia, weight loss, and pain (Table 2).

Four SCs were identified at baseline: neuro-psychological cluster (sleep problems, depres-sion, and anxiety), anorexia-cachexia cluster(anorexia, tiredness, and weight loss), confu-sional cluster (confusion, agitation, and urinaryincontinence), and gastrointestinal symptomcluster (nausea and vomiting) (Table 3). Theseclusters were present in 239 (59%), 390 (96%),

126 (31%), and 93 patients (23%), respectively.One hundred fifty-four patients had one SC(38%), 152 had two (37%), 53 had three(13%), and five had four SCs (1%). Forty-twopatients (10%) did not have any SC.An exploratory analysis was performed to

assess whether the composition of SCs variedaccording to the kind of primary tumor.This analysis included tumors with at least100 patients, that is, lung and gastrointestinaltumors. In patients with lung cancer, the na-ture and composition of SCs did not differfrom those in the whole series, with the excep-tion of the anorexia-cachexia cluster, whichincluded anorexia and weight loss but nottiredness. However, a new SC with dyspneaand diaphoresis was identified in patientswith lung cancer. In patients with digestive tu-mors, the four general SCs were confirmedand two others emerged: one with fever anddiaphoresis and the other with hiccups anddysphagia. However, these two new SCs ac-counted for 5% of the variance, so their clin-ical relevance is questionable.We also analyzed factors related to the pres-

ence of one SC in particular. For instance, the

Table 2Frequency and Severity of Symptoms on

Admission

Variable ofSymptom

Number ofPatients Frequency

Severity (0e10)(median� SD)

Tiredness 391 97 7.65� 2.08Anorexia 384 95 7.06� 2.28Well-being 371 91 5.01� 2.71Weight loss 357 88 6.14� 2.61Pain 327 80 5.37� 2.32Depression 323 80 5.98� 2.81Dry mouth 294 73 4.84� 2.19Insomnia 263 65 4.64� 2.18Anxiety 257 63 6.08� 2.75Dyspnea 228 56 4.33� 2.61Constipation 227 56 3.70� 1.93Drowsiness 191 47 3.73� 2.07Cognitive

impairment167 41 3.67� 2.47

Nausea 157 39 3.54� 2.20Urinary

incontinence116 29 3.84� 3.13

Sweating 104 26 2.88� 1.66Agitation 102 25 3.10� 2.14Vomiting 95 23 2.77� 1.55Dysphagia 88 22 3.01� 2.14Fecal

incontinence74 18 3.38� 2.59

Hiccups 59 15 2.44� 1.17Fever 51 13 2.42� 1.22

Vol. 42 No. 1 July 2011 27Symptom Clusters in Advanced Cancer

confusional cluster was more common in pa-tients $70 years old (54% vs. 45%; P< 0.001)and patients with an ECOG performance sta-tus of 3e4 vs. 0e2 (60% vs. 23%; P< 0.001)(Table 4). The neuropsychological clusterwas more common in patients younger than70 (81% vs. 77%), although this differencewas not statistically significant (P¼ 0.06). Thegastrointestinal SC appeared more commonlyin women (51% vs. 42%; P< 0.05), in patients

Table 3Symptom Cluster Structur

Items n (%) of Patients Inter

Confusional cluster 126 (31)AgitationConfusionUrinary incontinence

Neuropsychological cluster 239 (59)DepressionAnxietySleep problems

Anorexia-cachexia cluster 390 (96)AnorexiaWeight lossTiredness

Gastrointestinal symptom cluster 93 (23)NauseaVomiting

with an ECOG performance status of 3e4(42% vs. 33%; P< 0.05), and in patients withdigestive tumors as compared with lung cancer(52% vs. 25%; P< 0.001). We did not find anyvariable related to the presence of the SCanorexia-cachexia, probably because of itshigh incidence in our population. There wasno relation between the location or numberof metastases and SC. Differences betweenmen and women were only detected in the gas-trointestinal SC (Table 4).

The influence of SCs on survival also was esti-mated (Fig. 1): a reduced survival was observedfor patients included in the neuropsychological(35 vs. 21 days; P< 0.05) or gastrointestinalSC (62 vs. 36 days; P< 0.001). No survival diff-erences according to the number of symptomswere detected in the confusional or theanorexia-cachexia SC. Finally, the number ofSCs present in a given patient was related tome-dian survival: it was 52 days for patients with noSC; 38 days if there was one SC; 23 days if therewere two; and 19 days if three SCs were present(P< 0. 0001) (Fig. 2).

DiscussionPatients with advanced cancer often experi-

ence multiple coexisting symptoms caused bythe disease itself, side effects of treatment, orconcomitant conditions. As death approaches,specific symptoms of terminal disease mayappear.17 Knowledge of SCs may be importantfor symptom management, as treatment of onesymptom may be affected by other compo-nents in the same cluster.3

e and Frequencies

factor Correlations Cronbach’s a % of Variance

a¼ 0.83 17.80.7580.7600.819

a¼ 0.75 130.7180.8160.787

a¼ 0.70 7.90.8180.8010.664

a¼ 0.87 7.10.9140.926

Table 4Parameters Associated with Symptom Clusters

Parameters

CFC EC ACC GSC

No Yes No Yes No Yes No Yes

n (%) n (%) n (%) n (%)

Age (years)$70 85 (44) 100 (54)a 42 (23) 143 (77)b 1 (1) 184 (99) 113 (61) 72 (44)<70 121 (55) 100 (45) 42 (19) 129 (81) 2 (1) 119 (99) 133 (60) 88 (40)

GenderMale 127 (51) 120 (49) 54 (22) 193 (78) 2 (1) 245 (99) 168 (68) 79 (42)a

Female 79 (48) 80 (52) 30 (19) 129 (81) 1 (1) 158 (99) 78 (49) 81 (51)

ECOG PS0e2 92 (77) 27 (23)a 42 (35) 77 (65) 2 (2) 117 (98) 80 (67) 39 (33)a

3e4 114 (40) 173 (60) 42 (15) 245 (85) 1 (1) 286 (99) 166 (58) 121 (42)

Primary cancerLung cancer 62 (61) 40 (39) 24 (24) 78 (70) 1 (1) 101 (99) 76 (75) 26 (25)a

Digestive cancer 81 (58) 59 (42) 30 (21) 110 (79) 0 (0) 140 (58) 67 (48) 73 (52)

CFC¼ confusional cluster; EC¼ neuropsychological cluster, ACC¼ anorexia-cachexia cluster, GSC¼ gastrointestinal symptom cluster.aP< 0.05.bP¼ 0.06.

28 Vol. 42 No. 1 July 2011Jim�enez et al.

We investigated SCs in a population of pa-tients with advanced cancer. The most com-mon cluster, anorexia-cachexia, also has beenfound in other studies.11,12,19 It can be causedby cytokine release, such as interleukin-1, in-terleukin-6, and tumor necrosis factor-alpha.20

Some studies suggest that the combination ofmedroxyprogesterone or megestrol acetatewith oral supplementation with eicosapentae-noic acid, L-carnitine, and thalidomide couldimprove nutritional status, fatigue, and appe-tite in these patients.21

The neuropsychological SC was present in59% of our patients. This cluster has been iden-tified in patients with advanced disease,11,12,19

brain metastases,8 bone metastases,22 or breastcancer.23 If we only consider the three studiesperformed in advanced disease, one of themincluded three symptoms in the SC,11 anotherincluded only anxiety and depression,12 andthe third one grouped together insomnia, an-orexia, taste alteration, dysphagia, constipation,and dry mouth/thirst.19 These differences areattributable to the tools used because insomniawas not assessed in one study,12 whereas anxietyand depression were not included in anotherone.19 The presence of the neuropsychologicalSC suggests that anxiety, depression, and insom-nia should be treated concomitantly. In this re-gard, drugs such as serotonin selective uptakeinhibitors can improve both depression andanxiety.12

The gastrointestinal SC was present in 23%of our patients, similar to other series ofpatients with advanced cancer,11,12 or evenearly-stage tumors.4,5 In contrast, we identifieda new SC that includes cognitive impairment,agitation, and urinary incontinence, whichwas present in 31% of our patients. Its patho-physiology is unknown, although it could becaused by a brain disorder secondary to the tu-mor itself, metabolic disturbances or drugs.The reason why this SC has not been previ-ously described may lie in the unique featuresof our patients or, more likely, in the absenceof items to detect confusion, agitation, andsphincter control in questionnaires used byother investigators. In fact, Walsh andRybicki11 found some degree of confusion in21% of their patients and described an SC in-cluding confusion and edema, which theynamed ‘‘debility cluster.’’ We do not knowwhy urinary incontinence clusters with cogni-tive impairment and agitation, but it couldbe because of a loss of basic neurological func-tions in the context of neuronal damage.We found that age, sex, ECOG performance

status, and the primary type of tumor in-fluenced SC. The confusional SC was morecommon in elderly patients, whereas the neu-ropsychological SC predominated in thoseless than 70 years old. Many patients with anEGOG performance status of 3e4 had theconfusional and gastrointestinal SCs, which

600550500450400350300250200150100500

lavivruSmu

C

lavivruSmu

C

1.0

.9

.8

.7

.6

.5

.4

.3

.2

.1

0.0600550500450400350300250200150100500

1.0

.9

.8

.7

.6

.5

.4

.3

.2

.1

0.0

600550500450400350300250200150100500

lavivruSmu

C

lavivruSmu

C

1.0

.9

.8

.7

.6

.5

.4

.3

.2

.1

0.0600550500450400350300250200150100500

1.0

.9

.8

.7

.6

.5

.4

.3

.2

.1

0.0

Confusional Cluster

0-2 symptoms 3 symptoms

P=NS

Gastrointestinal Symptom Cluster

Neuropsycological Cluster

0-2 symptoms

3 symptoms

P< 0.05

0-1 symtom 2 symptoms

P< 0.001 0-2 symptoms 3 symptoms

P=NS

Anorexia-Cachexia Cluster

Fig. 1. Kaplan-Meier curves according to the number of symptoms included in the same cluster.

Vol. 42 No. 1 July 2011 29Symptom Clusters in Advanced Cancer

suggests that these SCs may contribute to dete-riorated physical functioning in patients withadvanced cancer. The gastrointestinal SC alsowas common in patients with digestive tumorsand in women. A recent study described similarfindings in women.24 Other investigators havereported that ECOG performance status, bonemetastases, fluid accumulation, and psycholog-ical distress are related to an SC.19 However,they did not find any relation with age, gender,or primary origin. Maybe factors influencingthe development of an SC vary with disease evo-lution and, in fact, a series of patients with early-stage disease indicates that factors influencingan SC are disease status, use of chemotherapy,and psychological distress.25

In our series, survival decreased as the num-ber of SCs increased. Inside a cluster, survivalalso was affected by the number of symptoms.

These results support the idea that symptomburden is associated with survival.17,19,26

A limitation of our study is that we did notuse a standard checklist of symptoms. TheESAS tool includes just nine symptoms, andwe added 13 other symptoms that are com-monly seen in patients with advanced disease.Also, ESAS measures symptom severity,whereas the scale we used for the additional13 symptoms measures distress. This differ-ence in the composition and quality of scalescould have influenced results.11,12,19 Symp-toms are multidimensional, and it has beensuggested that the most relevant dimensionbe measured for investigational purposes, byusing the most valid method.27 Another limita-tion of our study lies in the features of the pop-ulation included, with a clear predominanceof males and a low proportion of patients

600550500450400350300250200150100500

lavivruSmu

C1.0

.9

.8

.7

.6

.5

.4

.3

.2

.1

0.0

Survival time (days)

SC: 0

SC: 1

SC: 2

SC: 3-4

Fig. 2. Kaplan-Meier curves for the patients accord-ing the number of SCs (P< 0.0001). SC, symptomcluster.

30 Vol. 42 No. 1 July 2011Jim�enez et al.

with breast cancer as compared with otherseries.11,12,19 This might have influenced thefrequency and composition of the SC.

Despite these limitations, three of the fourSCs that we found also have been described byother investigators.11,12,19 A different matter iswhether prevalence and intensity of symptomschange with disease evolution. We cannot an-swer this question because our study was cross-sectional, and longitudinal studies are neededto assess the evolution of SC over time. Thiskind of study has been performed in other sce-narios, reinforcing the concept of different pat-terns of SCs at different time points.5,8,9

The optimal statistical tool to identify an SCis a matter of debate. Whereas the goal of PCAis data reduction, the goal of common factoranalysis (CFA) is to explain the correlation be-tween variables. PCA assumes symptom mea-surements that are error free, with no uniquevariance; it also inflates factor loadings and al-ters factor structure.28 However, researchershave shown that similar groupings may resultfrom CFA and PCA, provided that the commu-nalities are high (>0.70).29 As communalitieswere high in our study, we do not think thatthe use of PCA had a great influence on thecomposition of SCs.

In conclusion, we have identified four SCsin patients with advanced cancer, which areinfluenced by primary cancer site, gender,age, and ECOG performance status. Survivalwas affected by the presence of an SC. For

this reason, physicians should consider SCs tomanage several symptoms in combinationand optimize quality of life in patients with ad-vanced cancer.

Disclosures and AcknowledgmentsNo funding was received for this study, and

the authors declare no conflicts of interest.

References1. Fan G, Filipczak L, Chow E. Symptom clustersin cancer patients: a review of the literature. CurrOncol 2007;14:173e179.

2. Dodd M, Janson S, Facione N, et al. Advancingthe science of symptom management. J Adv Nurs2001;33:668e676.

3. Kim H, McGuire DB, Tulman L, Barsevick AM.Symptom clusters: concept analysis and clinical im-plications for cancer nursing. Cancer Nurs 2005;28:270e282.

4. Chen ML, Lin CC. Cancer symptom clusters:a validation study. J Pain Symptom Manage 2007;34:590e599.

5. Molassiotis A, Wengstrom Y, Kearney N. Symp-tom cluster patterns during the first year after diag-nosis with cancer. J Pain Symptom Manage 2010;39:847e858.

6. KimHJ, BarsevickAM,TulmanL,McDermott PA.Treatment-related symptom clusters in breast cancer:a secondary analysis. J Pain Symptom Manage 2008;36:468e479.

7. Gift AG, Jablonski A, Stommel M, Given CW.Symptom clusters in elderly patients with lung can-cer. Oncol Nurs Forum 2004;31:202e212.

8. Chow E, Fan G, Hadi S, et al. Symptom clusterin cancer patients with brain metastases. Clin Oncol2007;20:76e82.

9. Hadi S, Zhang L, Hird A, de Sa E, Chow E. Val-idation of symptom clusters in patients with meta-static bone pain. Curr Oncol 2008;15:211e218.

10. Finnegan L, Campbell RT, Estwing C, et al.Symptom cluster experience profiles in adult survi-vors of childhood cancers. J Pain Symptom Manage2009;38:258e269.

11. Walsh D, Rybicki L. Symptom clustering in ad-vanced cancer. Support Care Cancer 2006;14:831e836.

12. Cheung WY, Le LW, Zimmermann C. Symptomcluster in patients with advanced cancers. SupportCare Cancer 2009;17:1223e1230.

13. Chiu TY, Hu WY, Chen CY. Prevalence and se-verity of symptoms in terminal cancer patients:

Vol. 42 No. 1 July 2011 31Symptom Clusters in Advanced Cancer

a study in Taiwan. Support Care Cancer 2000;8:311e313.

14. Jim�enez A, Feliu J, Mart�ınez B, et al. Descriptiveanalysis of clinical factors affecting terminally illcancer patients. Support Care Cancer 2009;17:261e269.

15. Dodd MJ, Miaskowski C, Paul SM. Symptomclusters and their effect on the functional status ofpatients with cancer. Oncol Nurs Forum 2001;28:465e470.

16. Miaskowski C, Aouizerat BE, Dodd M,Cooper B. Conceptual issues in symptom cluster re-search and their implications for quality-of-life as-sessment in patients with cancer. J Natl CancerInst Monogr 2007;37:39e46.

17. Cheung WY, Barmala N, Zarinehbaf S, Rodin G,Le LW. The association of physical and psychologi-cal symptom burden with time to death among pal-liative cancer outpatients. J Pain Symptom Manage2008;37:297e304.

18. Bruera E, Kuehn N, Miller MJ, Selmer P,Macmillan K. The Edmonton Symptom AssessmentSystem (ESAS): a simple method for the assessmentof palliative care patients. J Palliat Care 1991;7:6e9.

19. Tsai JS, Wu CH, Chiu TY, Chen CY. Significanceof symptom clustering in palliative care of advancedcancer patients. J Pain Symptom Manage 2010;39:655e662.

20. Kayacan O, Karnak D, Beder S, et al. Impact ofTNF-alpha and IL-6 levels on development of ca-chexia in newly diagnosed NSCLC patients. Am JClin Oncol 2006;29:328e335.

21. Mantovani G, Maccio A, Madeddu C, et al. Ran-domized phase III clinical trial of five different armsof treatment in 332 patients with cancer cachexia.Oncologist 2010;15:200e211.

22. Chow E, Fan G, Hadi S, Filipczak L. Symptomclusters in cancer patients with bone metastases.Support Care Cancer 2007;15:1035e1043.

23. Bender CM, Ergyn FS, Rosenzweig MQ,Cohen SM, Sereika SM. Symptom clusters in breastcancer across 3 phases of the disease. Cancer Nurs2005;28:219e225.

24. Cheung WY, Le LW, Gagliese L, Zimmermann C.Age and gender differences in symptom intensityand symptom clusters among patients with metastaticcancer. Support Care Cancer 2011;19:417e423.

25. Chen ML, Tseng HC. Symptom clusters in can-cer patients. Support Care Cancer 2006;14:825e830.

26. Hwang SS, Scott CB, Chang VT, et al. Predictionof survival for advanced cancer patients by recursivepartitioning analysis: role of Karnofsky performancestatus, quality of life, and symptom distress. CancerInvest 2004;22:678e687.

27. Barsevick AM, Whitmer K, Nail LM, et al. Symp-tom cluster research: conceptual, design, measure-ment, and analysis issues. J Pain Symptom Manage2006;31:85e95.

28. Kim H. Common factor analysis versus principalcomponent analysis: choice for symptom cluster re-search. Asian Nurs Res 2008;2:17e24.

29. Skerman HM, Yates PM, Battistutta D. Multivar-iate methods to identify cancer-related symptomclusters. Res Nurs Health 2009;32:345e360.