Cancer Symptom Clusters: Clinical and Research Methodology

Abstract

Introduction:

Patients with cancer experience multiple symptoms that frequently appear in groups or clusters. We conducted a comprehensive clinical review of cancer symptom cluster studies to identify common symptom clusters (SC), explore their clinical relevance, and examine their research importance.

Methods:

Published studies and review articles on cancer SC were obtained through a literature search. We identified 65 reports. These varied in assessment instruments, outcomes, design, population characteristics, and study methods.

Results:

Two main approaches to symptom cluster identification were found: clinical and statistical. Clinically determined SC were based upon observations of symptom co-occurrence, associations, or interrelations. These included fatigue-pain, fatigue-insomnia, fatigue-insomnia-pain, depression-fatigue, and depression-pain. They were analyzed by multivariate analysis. They had low to moderate statistical correlations. Disease- or treatment-related SC were influenced by primary cancer site, disease stage, or antitumor treatment. SC determined by statistical analysis were identified by factor and cluster analysis through nonrandom symptom distribution. Nausea-vomiting, anxiety-depression, fatigue-drowsiness, and pain-constipation consistently clustered by either or both of these statistical methods. The individual symptoms of pain, insomnia, and fatigue often appeared in different clusters. A consensus about standard criteria and methodological techniques for cluster analysis should be established.

Conclusions:

Several important cancer SC have been identified. Nausea-vomiting, anxiety-depression, and dyspnea-cough clusters were consistently reported. The techniques of symptom cluster identification remain a research tool, but one with considerable potential clinical importance. Further research should validate our analytical techniques, and expand our knowledge about SC and their clinical importance.

Introduction

Patients with cancer have multiple disease or treatment related symptoms (in one study a median of 6; range, 3–13).¹ Better assessment and insight into symptom mechanisms might improve treatment and quality of life (QOL).² A National Institutes of Health Conference concluded that the apparent relationship between pain, depression, and fatigue needed investigation.³ This was based on simultaneous occurrences of common burdensome symptoms, which were proposed to form a symptom cluster.⁴

Psychology, psychiatry, and general medicine have examined the co-occurrence of symptoms in clusters for many years. Such studies provided important clinical and research findings.^5,6 Cancer cluster type and composition vary across studies.^5,7 These differences may in part be explained by different assessment instruments, design, statistical methods, and population characteristics.⁷ Furthermore, there is a lack of common criteria to identify symptom clusters (SC). In this state of the art review, we will comment on symptom interrelationships, summarize important features of cancer symptom cluster studies, and discuss methodological problems.

Definitions

One definition of a symptom cluster has been that of three or more concurrent symptoms related to each other that group together.^5,8 The definition was later revised so that two symptoms could form a cluster.⁹ Currently, there is no consensus on clusters either conceptually or methodologically. Clusters imply symptoms which group together in a nonrandom fashion.^9,10 This may imply that SC have a common mechanism, and symptoms within a cluster might respond a single treatment.¹¹ SC may be stable throughout the disease course¹² within the context of a particular cancer or population and have clinical outcomes that are different from individual symptoms.^5,7,8,13,14

The number of symptoms reported within a cluster varies. Symptoms change in intensity during the disease course. It is possible that cluster composition will be influenced by symptom severity or other domain measured (prevalence or distress). The assessment instrument, timeframe, and statistical methodology will potentially alter cluster composition. An individual symptom can occur in different clusters,^15–17 which suggests multiple mechanisms for a symptom. This observation suggests that individual symptoms arise from multiple mechanisms.^11,17 Relationships between two symptoms may be indirect, perhaps mediated by a third symptom¹⁸ or co-occurrence influenced by performance status (PS).¹⁸ Complex neuroimmunologic pathways, differences in statistical methodology or symptom assessment instruments, and varied symptom cluster definitions account for the differences in the reported cluster compositions.¹⁷ Whether SC should be defined based solely on a statistical method, clinical experience, symptom domain, or by a common mechanism is unclear.^4,5,14

Methods

Cancer symptom cluster studies in adults were evaluated. A MEDLINE database was searched using OVID MEDLINE (Ovid Technologies Inc, New York, NY). “Symptom cluster” was not indexed as a MESH term. Therefore, we used the key words: symptom cluster, multiple symptoms, symptom constellation, or concurrent symptoms and neoplasm. A hand search of references was also performed. Searches were limited to original reports in English (1948–2010). The search methodology is listed in the Appendix. A total of 238 abstracts were identified for initial review (Fig. 1). All studies were independently assessed by J.K. and A.A.

FIG. 1.

Flow chart of study identification and selection. The flow chart shows the inclusion and the exclusion of the publications used in the study.

We then selected studies based on specific inclusion and exclusion criteria. Studies were included if they:

1. Aimed to identify SC or interrelationships between two or more concurrent symptoms.

2. Reported associations of clustered symptoms and patient outcomes (e.g., QOL, functional status, survival) using validated instruments or nonvalidated symptom checklists.

3. Reported at least one moderate to high correlation (r≥0.5) between cluster symptoms.

Studies were excluded if they:

1. Used cluster analysis to identify different groups of patients based on common symptom experiences.

2. Reported SC as a secondary outcome through validation of a symptom assessment instrument or multidimensional assessment scale.

Results

Sixty-five studies that met inclusion/exclusion criteria were identified (Fig. 1). Two main approaches to symptom cluster analysis were found: clinical and statistical.⁷

Clinical SC (to include cancer- and treatment-specific clusters) were identified from studies that focused on a small number of preselected symptoms. The rationales were predominantly based on clinically observed associations of symptoms that were likely to cluster such as pain, fatigue, and depression. The quality of studies widely varied. However, they usually had low to moderate statistical correlations (r<0.5). Among the 18 reports that identified clinical SC, most used correlational methods; 1 study used patient-based cluster analysis and equation model,¹⁹ 2 studies used either equation model²⁰ or another model,²¹ and 3 studies used mediation model^18,22–23 to identify symptom interrelationships. All were descriptive, except 1 interventional.²⁴ Five underwent secondary data analyses^{15,21,22,25–27}; 2 used data from randomized controlled trials (RCT).^21,22 Five studies were longitudinal^{8,19,21,24,28} and the others^{13,15,18,20,22,23,25–27,29–32} cross-sectional.

Statistically defined SC were derived from either a multisymptom assessment tool (e.g., Edmonton Symptom Assessment Scale) or a symptom checklist that included larger number of symptoms. Most used factor or cluster analysis to identify clusters. Among 47 reports, 7^33–39 assessed at least 10 symptoms, but used correlation analyses. Most studies were descriptive, except for three RCT.^39–41 One study used both qualitative and quantitative methods.³⁸ Sixteen reports were post hoc.^{9,10,12,16,33,42–52} Twenty-one reports were longitudinal^{33,36,37,39–41,43,49,53–65}; 26 had a cross-sectional design.^{9,10,12,16,34,35,38,42,44–48,50–52,66–75} Eleven used factor analysis,^{9,43–45,48,49,51,59–61,69} 1 a confirmatory factor analysis (CFA),⁴⁷ and 12 principal component analysis (PCA).^{37,39,40,50,53,55,56,58 63,66–68} Eight used cluster analysis,^{10,12,16,42,46,65,70,72} 9 more than one analytical technique,^{41,50,54,55,59,71,73–75} 2 used structural equation model,^57,62 and 3 used other models.^16,52,64

Clinical Clusters

Clinical observation

Clinically defined SC often focused on a few symptoms (usually 2 or 3) empirically considered clinically important.⁷ They were selected after multisymptom patient assessment, or formed a preselected group of symptoms named as a cluster (Table 1). This approach is based on:

1. Clinical studies or prevalence that suggest co-occurrence.²⁹

2. A proposed common pathophysiologic mechanism.¹⁵

3. Symptoms that adversely influenced function, or treatment tolerance.^8,76

Table 1.

Clinically Defined Cancer Symptom Clusters

Author, year	Symptom clusters	Sample size, primary cancer site	Symptom assessment tool	Major findings
Barsevick et al., 2006²²	Fatigue, depression	n=295 Heterogeneous (antitumor treatment v control)	FPI, General Fatigue Scale	Control group functional status after antitumor treatment mediated the relationship between fatigue and depressive symptoms
Beck et al., 2005¹⁸	Fatigue, pain, sleep disturbances	n=84 Heterogeneous (radiotherapy)	BPI-SF, PSQI, POMS	Pain influences fatigue directly as well as indirectly by its effect on sleep Pain-fatigue (r=0.5), pain-sleep disturbances (r=0.5), fatigue-sleep disturbance (r=0.5) 35% fatigue variance from mediation pathway
Chan et al., 2005²⁸	Breathlessness, fatigue, anxiety	n=27 Lung (radiotherapy)	VAS	Correlations among three symptoms varied over time: r=0.3–0.4 (T0), 0.5–0.7 (T1), 0.5–0.7 (T2) α=0.7 (T0), 0.8 (T1), 0.8 (T2)
Dodd et al., 2001⁸	1. Pain, fatigue 2. Pain, sleep insufficiency 3. Fatigue, sleep insufficiency	n=93 Heterogeneous (chemotherapy)	QOL-CA	Fatigue-pain (r=0.2), fatigue-insomnia (r=−0.1), pain-insomnia (r=−0.1) Three symptoms correlated to KPS (r=0.6) Cluster explained 48% functional status variance
Fox and Lyon, 2006²⁵	1. Depression, fatigue 2. Fatigue, pain	n=51 Survivors (lung)	SF-36 Health Status Survey, FSQOLS	Depression-fatigue (r=0.4), fatigue-pain (r=0.4) Depression (r=0.5) and fatigue (r=0.4) correlated with QOL Pain not significantly correlated with QOL Depression-fatigue cluster explained 29% QOL variance
Fox and Lyon, 2007²⁶	Depression, fatigue	n=76 Survivors (ovarian)	SF-36 Health Status Survey, FSQOLS	Depression-fatigue (r=0.5), fatigue-pain (r=0.3) Depression-fatigue cluster explained 41% QOL variance
Fox et al., 2007³¹	1. Depression, fatigue, sleep disturbance, cognitive impairment 2. Depression, fatigue, sleep disturbance, cognitive impairment, pain	n=73 Survivors (high-grade glioma)	COGMOS, HADS, BPI, BFI, GSDS, FACT-BR, FSQOLS	Cluster 1 explained 29% QOL variance Cluster 2 explained 62% functional status variance
Francoeur et al., 2005¹⁵	1. Fever, pain, depression, insomnia 2. Fever, nausea/vomiting 3. Fatigue, pain, depression 4. Fatigue, pain, weight loss 5. Fatigue, SOB, insomnia 6. SOB, anorexia, insomnia 7. Hypertension, fatigue, SOB, insomnia	n=268 Heterogeneous (radiotherapy for bone metastasis)	CESD, VRS	Depressive affect measured Significant consistent interactions between cluster symptoms
Gaston-Johansson et al., 1999³⁰	Fatigue, pain, depression	n=127 Breast	Piper Fatigue Scale, Fatigue VAS, Gaston-Johansson POM, BDI, MOS-SF General Health Survey	Pain, depression, fatigue correlated: Fatigue-pain (r=0.3), fatigue-depression (r=0.6), pain-depression (r=0.2) Depression-pain 64%, fatigue-depression 42% variance in total health status
Hoffman et al., 2007²¹	Pain, fatigue, insomnia	n=80 Lung (>40 years, newly diagnosed, receiving chemotherapy)	Cancer symptom Experience Inventory, Comorbidity Questionnaire	Significant 3-way interactions among three symptoms in model based analysis
Huang et al., 2009²³	Sleep disturbance, fatigue, depression	n=77 Primary liver	BFI-Taiwanese, PSQI, HADS, ECOG PS	Sleep disturbance-fatigue (r=0.3), sleep disturbance-depression (r=0.5), depression-fatigue (r=0.5) Depression completely mediated the effects of sleep disturbance on fatigue
Lyon et al., 2010²⁴	Depression, anxiety, fatigue, sleep disturbances	n Total=36 Breast (chemotherapy) n 1=CES device n 2=The sham CES device n 3=Control	HADS, BPI, BFI, GSDS	All symptoms correlated with each other, and most correlated with baseline CRP Depression increased over time but trended toward less in the CES group (not significant)
Miaskowski et al., 1999¹³	Fatigue, pain, sleep disturbances	n=24 Heterogeneous (palliative radiotherapy for bone metastases)	Pain NRS, Lee Fatigue Scale, CESD	Fatigue and sleep disturbances correlated with each other Depression-morning fatigue (r=0.5), depression-evening fatigue (r=0.5) Morning fatigue improved with sleep efficiency (r=0.4) and duration (r=0.5) Morning fatigue related to pain before (r=0.4) and after sleep (r=0.4)
Myers et al., 2010²⁷	Fatigue, mood swings, numbness/tingling, sleep disturbance	n 1=638 ovarian (received chemotherapy) n 2=68 ovarian (did not receive chemotherapy)	SSI, disease and treatment history survey	Symptoms of fatigue, mood swings, numbness or tingling, and sleep disturbance explained 37% memory problems variance 4 symptoms correlated with memory problems (r=0.3–0.5)
Reyes-Gibby et al. 2006²⁹	1. Pain, fatigue 2. Pain, depression 3. Fatigue, depression 4. Pain, fatigue, depression	n=2161 Survivors (heterogeneous) n=15,049 Healthy controls	CESD, single question for pain and fatigue	Cancer history: higher risk for fatigue; OR=1.4 (95% CI=1.3–1.6), depression OR=1.2 (95% CI=1.1–1.4), pain OR=1.1 (95% CI=1.0–1.3) Pain-depression-fatigue cluster most frequent, then by pain-fatigue, pain-depression, fatigue-depression
So et al., 2009²⁰	Fatigue, pain, anxiety, depression	n=215 Breast (antitumor treatment)	BFI-C, BPI-C, HADS, FACT-Breast (Chinese)	Pain-anxiety (r=0.3), pain-depression (r=0.3), fatigue-pain (r=0.5), fatigue-anxiety (r=0.5), fatigue-depression (r=0.5), anxiety-depression (r=0.6) Cluster symptoms (together with the covariates) explained 66% QOL variance
Steel et al., 2010¹⁹	Pain, fatigue, depression	n=206 Hepatobiliary	FACT-Hep EWB	High levels of pain, fatigue, and depression associated with elevated eosinophil levels Clusters did not mediate relationship between eosinophils and survival Low eosinophil levels associated with shorter survival
Thornton et al., 2010³²	Pain, depression, fatigue	n=104 Breast (newly diagnosed; advanced)	BPI, CES-D, FSI	Pain-depression-fatigue correlated to each other (r=0.6–0.7) and proposed biomarkers (r=0.3–0.8)

Note: The studies are listed in chronological order.

r, correlation coefficient; α=Cronbach's α coefficient; CI, confidence interval; GI, gastrointestinal; SOB, shortness of breath; QOL, quality of life; PTSD, post-traumatic stress disorder; KPS, Karnofsky Performance Status, ECOG PS, Eastern Cooperative Oncology Group Performance Score; RCT, randomized clinical trial; VAS, visual analogue scale; VRS, verbal rating scale, BDI, Beck Depression Inventory; BFI, Brief Fatigue Inventory; BPI, Brief Pain Inventory; HADS, Hospital Anxiety and Depression Scale; SSI, Symptom Severity Index; GSDS, General Sleep Disturbance Scale; POM, Painometer; MOS-SF, Medical Outcomes Study-Short Form, CES, Cranial Electrical Stimulation; CES-D, Center for Epidemiological Studies Depression; QOL-CA, Quality of Life-Cancer; FPI, Functional Performance Inventory; SF-36, Short-Form 36; FSQOLS, Fox Simple QOL Scale; COGMOS, Cognitive Functioning Subscale of the Medical Outcomes Scale; FACT-BR, Functional and Brain Subscales of the Functional Assessment of Cancer Therapy, PSQI: Pittsburgh Sleep Quality Index, MFSI-SF: Multidimensional Fatigue Symptom Inventory Short Form; SDS, Symptom Distress Scale; SDI, Symptom Distress Inventory; HCA, hierarchical cluster analysis; CFA, confirmatory factor analysis; EFA, explanatory factor analysis; NCI CTC, National Cancer Institute Common Toxicity Criteria; POMS, Profile of Mood States; FACT, Functional Assessment of Cancer Therapy; PSET, Physical Symptom Experience Tool; MSAS, Memorial Symptom Assessment Scale; MDASI, M.D. Anderson Symptom Inventory; ESAS, Edmonton Symptom Assessment Scale; C-PET, Checklist for Patients with Endocrine Therapy; IBCSG, International Breast Cancer Study; EORTC QLQ-C30, European Organization for Research and Treatment of Cancer Quality of Life Questionnaire; ULL, upper limb lymhedema; WCLS, Wesley Clinic Lymhedema Scale; FACT-Hep EWB, Functional Assessment of Cancer Therapy-Hepatobiliary Emotional Well-Being; SQLI, Spitzer Quality of Life Index; FSI: Fatigue Symptom Inventory.

Post hoc statistical analysis explored the relationships between these target symptoms to support symptom interrelationships. This was done by correlation (Pearson product moment correlation or Spearman's ρ) or by multiple regression analysis, with the dependent variable the target symptom of interest.^{8,13,20,21,25,26,28,30,31} The degree of association is reported by a correlation coefficient (0=no correlation, 1=excellent correlation), or interaction (significant or not) from regression models. Common symptoms like fatigue-depression-pain; fatigue-insomnia-pain; nausea-vomiting-poor appetite have been clustered in this fashion. It is noteworthy that such correlations were done on symptom pairs only.^4,8,13,76

Fatigue, sleep disturbance, depression, and pain have been studied together in relation to Karnofsky Performance Status (KPS) and QOL through clustering patients, not symptoms.⁷⁷ This particular study identified patient subgroups based on fatigue and pain severity by cluster analysis: Those who reported low levels of all symptoms had the best KPS and QOL.

Antitumor treatment

Animal and human studies suggest symptoms may be mediated by a common neuroimmunologic mechanism. Sickness behavior theory links symptoms like depression, cognitive impairment, fatigue, and pain to proinflammatory cytokines.^15,78 The therapeutic use of interferon-α (IFN-α) and interleukin-2 (IL-2) for renal cancer and melanoma produces neuropsychological symptoms, which are similar to the sickness syndrome of advanced cancer. Anxiety, asthenia, depression, fatigue, nausea, and pain are evident. They group together dependent on whether IFN-α or IL-2 were used alone or together.⁷⁹ Single cytokine immunotherapy generates different SC than do cytokine combinations.^80,81 Paroxetine improves cytokine-induced depression, anxiety, and cognitive deficits but not anorexia or fatigue. Paroxetine during chemotherapy improves depression, but not fatigue.^80,82 Genetic make-up may also play a role in cytokine generated symptoms.⁸³ The depression-fatigue-pain cluster is associated with serum CRP levels in women receiving treatment for breast cancer.²⁴ Certain clusters are likely to be influenced by antitumor treatments.

A major limitation is that the number of symptoms selected in a cluster directly determined the cluster observed. Significant correlations might still exist between symptoms not included in a predefined cluster, or perhaps symptom relationships would differ based on the symptom used as the dependent variable. It is important to justify the theoretical foundations for the choice of symptoms for analysis.

Statistical Clusters

Both hierarchical cluster (HCA) and factor analyses place multiple symptoms into groups (clusters) without any prior assumptions about the symptom relationships (Table 2). Large data sets are analyzed to identify associations among symptoms that may define a clinically important (but previously unknown) relationship, but this is speculative. Some used one analysis to identify clusters but validate the cluster by another.⁹

Table 2.

Statistically Defined Cancer Symptom Clusters

Author, year	Symptom clusters	Sample size, primary cancer site	Symptom assessment tool	Major findings
Aprile et al., 2008⁵²	1. GI Cluster 1: Fatigue, anorexia, dehydration, nausea, vomiting 2. GI Cluster 2. Anorexia and dehydration with taste alteration, mucositis, dry mouth 3. Dermatologic: Dry skin, skin pigmentation/hand-foot syndrome reaction, rash, itching 4. Wound complication toxicity 5. Pulmonary: cough, dyspnea, infection 6. Fewer: Chills, weight loss, weakness	n=300 Colorectal (chemotherapy)	NCI CTC	Distance matrix analyses identified 6 principal chemotherapy-induced toxicity clusters. All clusters had at least 5 strong connections with other toxicities: dehydration, fatigue, anorexia, pain, weight loss
Aprile et al., 2009¹⁶	1. GI: Nausea, vomiting, chills, taste alteration, dehydration, pain, fever, constipation, diarrhea, distension/bloating 2. Constitutional: Anorexia, weight loss, fatigue 3. Dermatologic: Dry skin, skin pigmentation, rash, itching, wound complication, hemorrhage/bleeding 4. Respiratory: Cough, dyspnea, pain, palpitation, infection 5. Psychologic: Anxiety, depression, cystitis	n=300 Colorectal (chemotherapy)	NCI CTC	Five treatment-related toxicity clusters identified
Bender et al., 2005⁴⁶	Common single cluster for 3 studies Fatigue, feeling lack of energy, decreased physical strength/weakness, feeling depressed or blue, feeling anxious or nervous, loss of concentration	Study 1 Breast (before treatment) (n 1=40) Study 2 Survivors (breast) (n 2=88) Study 3 Breast (metastatic) (n 3=26)	Symptom Checklist, POMS, FACT, Kupperman Menopausal QOL Scale	Three clusters identified for three phases of breast cancer Each cluster composed of symptoms related to fatigue, perceived cognitive impairment, and mood problems r values not reported
Bender et al., 2008⁴⁸	Clusters in People with Cancer History 1. Skin rash, itching, night sweats, constipation, dizziness, abdominal pain, back pain 2. Fatigue, nausea, diarrhea, generalized pain, sleeping problems 3. Leaking urine, frequent ruination, walking problems, balance problems 4. Weight gain, overeating, SOB, chest palpitations, joint pain	n 1=149 Cancer history n 2=846 Chronic health problems without cancer	Comorbidity Questionnaire	Individuals with chronic health problems and cancer may not have unique symptom clusters compared to chronic health problems without cancer
Breen et al., 2009⁵¹	1. GI: Nausea, vomiting, pain 2. General malaise: Tiredness, feeling weak, headaches 3. Emotional: Feeling depressed, feeling anxious 4. Nutritional: Changes to appetite, weight loss or gain 5. General physical: Mouth/throat problems, SOB	n=192 Breast, GI, or lymphoma	Chemotherapy Symptom Assessment Scale, HADS	Five factors explained 37% variance in symptom distress scores
Chen and Tseng, 2006⁹	1. Sickness: Pain, fatigue, sleep disturbance, lack of appetite, drowsiness 2. GI: Nausea, vomiting 3. Emotional: Distress, sadness	n=151 Heterogeneous (half chemotherapy)	MDASI-Chinese, HADS-D	Three factors 55% variance α=0.7–0.9 Functional status negatively associated with sickness cluster
Chen et al., 2007⁴⁷	1. Sickness: Pain, fatigue, sleep disturbance, lack of appetite, drowsiness 2. GI: Nausea, vomiting 3. Emotional: Distress, sadness	n=321 Heterogeneous (Most either chemo- or radiation-therapy)	MDAS-Taiwanese	r=0.5–0.7 Functional status negatively associated with all three clusters
Cheung et al., 2009⁶⁶	1. Fatigue, drowsiness, nausea, decreased appetite, dyspnea 2. Anxiety, depression	n=1366 Advanced cancer	ESAS	Two clusters explained 55% total variance r=0.2–0.7 α=0.8 Symptom clusters differed by cancer site
Cheung et al., 2010⁶⁷	Various clusters for men and women, and younger and older patients	n=1358 Heterogeneous (advanced)	ESAS	Symptom clusters stratified by age and gender to find demographic influences on cluster composition
Chow et al., 2007⁵³	Baseline 1. Fatigue, pain, drowsiness, poor sense of well-being 2. Anxiety, depression 3. SOB, nausea, poor appetite Follow-up A. Variable clusters B. Consistent clusters: 1. Fatigue, drowsiness 2. Anxiety, depression	n=518 Bone metastases (radiotherapy)	ESAS	Three clusters 66% total baseline variance α=0.6–0.8 Cluster composition changed during post-radiation therapy
Chow et al., 2008³⁷	Baseline 1. Fatigue, drowsiness, SOB, pain 2. Anxiety, depression 3. Poor appetite, nausea, poor sense of well-being Follow-up Variable clusters at each time point	n=170 Brain metastasis (radiotherapy)	ESAS	Three clusters 62% total variance α=0.6–0.7
Dodd et al., 2005⁵⁴	Time 1 (after first chemotherapy cycle) Sensory, GI-related, cognitive-respiratory, pain-fatigue Time 2 (cancer treatment completed) Cognitive, pain/fatigue, GI-related Time 3 (4–6 months following treatment completion) Cognitive, pain/fatigue, lower GI-related, upper GI-related, sensory	n=80 Breast	26-item symptom checklist	Clusters at T1, T2, T3 accounted for 50%, 54%, 59% variance, respectively Pain-fatigue (T1: r=−0.4), cognitive-pain/fatigue (T2: r=−0.6, T3: r=−0.7)
Fan et al., 2007⁶⁸	1. Lack of appetite, nausea, poor sense of well-being, pain 2. Fatigue, drowsiness, SOB 3. Anxiety, depression	n=1296 Heterogeneous (radiotherapy)	ESAS	Three clusters 62% variance r=0.1–0.6 α=0.7–0.8
Fu et al., 2009⁶⁹	1.Depression: Sadness/depression, anxiety, poor sex drive, grief/loss 2.Chemotherapy: Fatigue, poor sex drive, hot flashes, headache, insomnia, poor memory 3.Hormone: Poor appetite, SOB, nausea, lymhedema, neuropathy 4.Pain-related: Insomnia, muscle aches, bone pain	n=139 Survivors (breast)	MSAS-SF	Each factor explained various level symptom variances: Factor 1 (30%), Factor 2 (12%), Factor 3 (9%), Factor 4 (7%)
Gift et al., 2003⁴³	Nausea, fatigue, vomiting, weakness, poor appetite, weight loss, altered taste	n=112 Lung (newly diagnosed)	PSET	Number of symptoms and severity correlated over time (r=0.3–0.4); α=0.7 (T1), 0.7 (T2), 0.8 (T3) Correlations between cluster symptoms varied over time (r=0.2–0.5)
Gift et al., 2004⁴⁴	Nausea, fatigue, vomiting, weakness, poor appetite, weight loss, altered taste	n=220 Lung (≥65 years, early diagnosed)	PSET, SF-36 Health Status Survey	Number and severity of cluster symptoms correlated to physical function Four factor based model 34% physical function variance α=0.7
Glaus et al., 2006⁷⁰	Menopausal symptom cluster: Hot flashes, weight gain, tiredness, decreased sexual interest, vaginal dryness	n=373 Breast (receiving hormone therapy)	C-PET, IBCSG Linear Analogue Scales	r≥0.8 Cluster symptoms dendogram
Gleason et al., 2007⁴¹	Baseline 1. Language: Can't find the right words, can't read like before, can't write like before 2. Mood: Feel sad, feel anxious, depressed mood	n=66 Brain metastasis or newly diagnosed primary brain tumor	FACT, FACT-Brain, CESD	Menopausal cluster symptoms dendogram Two clusters consistent over time
	Follow-up1. Language: Can't find the right words, can't read like before, can't write like before, trouble with eyesight2. Mood: Feeling sad and anxious, depressed mood
Hadi et al., 2008⁵⁶	1. Walking ability, general activity, normal work, enjoyment of life, worst pain2. Relations with others, mood, sleep	n=348Bone metastasis (radiotherapy)	BPI	Two baseline clusters 67% total varianceα=0.7–0.9Two clusters disintegrated at 4, 8, and 12 weeks post-RTAll symptom severity items improved over timeIn nonresponders, two clusters disappeared week 4, reemerged week 8, disintegrated week 12
Hadi et al., 2008⁵⁵	Baseline1. Activity-related: Walking ability, general activity, normal work, enjoyment of life, worst pain2. Psychology-related: Relations with others, mood, sleep	n=52Bone metastasis	BPI	Cluster 1 and Cluster 2 explained 67% and 10% total variance, respectivelyr=0.4–0.9α=0.8–0.9
	Week 4 and Week 8: Clusters varied at each time point
	Week 12: No cluster found
Hayduk et al., 2010⁵⁷	Pain, anxiety, nausea, SOB, drowsiness, appetite, tiredness, depression, well-being (57)	n=82Heterogeneous (terminal)	ESAS	Causal connections between symptoms change 1 week-1 month before death
Henoch et al., 2009⁷¹	Consistent clusters across instruments and analyses1. Pain: Pain, nausea, bowel issues, appetite loss, fatigue2. Mood: Mood, outlook, concentration, insomnia3. Respiratory: Breathing, cough	n=400Lung (newly diagnosed)	EORTC, SDS	Correlation analysis within clusters: Pain r=0.2–05; Mood r=0.3–0.5; Respiratory r=0.1–0.4α=0.6–0.7 explained 18–19% varianceCluster symptoms dendogram
Hird et al., 2010⁵⁸	Baseline1. Trouble concentrating, decreased alertness, confusion, imbalance problems, memory loss, weakness, fatigue, vision problems, problems with smell, hearing or tingling2. Nausea, vomiting, headache, dizziness3. Seizures, numbness, speech difficulty, personality change	n=129Brain metastasis (radiotherapy)	SQLI, 17-item Symptom Questionnaire	SQLI items PCA identified 2 clusters; Clusters 64% total variance; α=0.4–0.7Additional 17 symptoms PCA identified 3 clusters; Clusters explained 42% total variance; α=0.5–0.8
	At Month 1 and 3Numbers and constitutes of the clusters varied at each time point
Husain et al., 2010⁴²	Two common clusters pre- and post-paracentesis1. Depression, anxiety2. Fatigue, appetite, well-being, mobility	n=57Heterogeneous (with malignant ascites)	ESAS, EORTC QLQ-C30, Abdominal Distension and Mobility Questionnaire	ESAS symptoms and EORTC QLQ-C30 items clustering in a dendogram
Jarden et al., 2009⁴⁰	Treatment-Related Clusters1. GI: Nausea, vomiting, stomach pain, loss of appetite, diarrhea2. Cognitive: Diminished concentration, memory problems, fatigue3. Affective: Nervousness, anxiety, stress4. Functional/Pain: Muscle and joint aches5. Mucositis: Mouth pain, throat pain, difficulty swallowing	n=42Hematological malignancies (including intervention and control groups)	Stem Cell Transplantation Symptom Assessment Scale	Five factors explained 55% and 58%, respectively, total variance in the intervention and control groupsIn the intervention group, significant reduction in symptom intensity over time for all clusters except affective symptom cluster
Karabulu et al., 2010⁷²	1. General activity, mood, work, relations with other people, walking, enjoyment of life2. Sleep disturbance, difficulty in remembering, pain, distress, sadness, fatigue, dry mouth, appetite loss3. Nausea, vomiting, SOB, numbness, drowsiness	n=287Heterogeneous	MDAS-Turkish	Cluster symptoms dendogram
Kenefick, 2006³³	1. Concentration, fatigue2. Concentration, outlook3. Appearance, outlook	n=57Breast	SDS	Three clusters persisted over the three time pointsComplexity of symptoms interactions not well described by correlation coefficients calculation
Kenne Sarenmalm et al., 2007³⁴	Lack of energy, difficulty sleeping, pain, worrying problems with sexual interest or activity, feeling sad, dry mouth, feeling drowsy, feeling irritable, feeling nervous, constipation, lack of appetite	n=56Breast	MSAS, HADS, EORTC QLQ-C30, IBCSG QOL Core Questionnaire-Breast	Each of 12 symptoms correlated with several other symptomsWorrying-feeling nervous, worrying-feeling sad, nausea-lack of appetite had strong correlations
Kenne Sarenmalm et al., 2008³⁶	1. Fatigue, depression2. Fatigue, pain3. Pain, depression	n=56Breast	MSAS, HADS, EORTC QOL-C30	Highly significant correlations within cluster symptomsFatigue, pain, depression explained 68–72% distress varianceDistress explained 44–46% variance in QOL
Kim et al., 2008⁴⁹	Time 11. Psychoneurologic: Depressed mood, fatigue, insomnia, painTime 21. GI: Nausea, vomiting, decreased appetite2. Psychoneurologic: Depressed mood, insomnia, pain, hot flashesTime 31. GI: Nausea, vomiting, decreased appetite2. Psychoneurologic: Depressed mood, cognitive disturbance, fatigue, insomnia, pain	n=282Breast (antitumor treatment)	General Fatigue Scale, POMS-Short Form, PSQI, Side Effect Check List	α=0.6 (T1), 0.7 (T2), 0.8 (T3)Breast cancer had higher symptom severity scores than prostate cancer
Kim et al., 2009⁵⁹	1. Mood-cognitive: Difficulty concentrating, difficulty sleeping, feeling sad, worrying, itching, feeling irritable2. Sickness-behavior: Pain, lack of energy, feeling drowsy3. Treatment-related/pain	n=160Prostate and breast (radiotherapy)	MSAS	Three relatively stable symptom clusters over timeα=0.6–0.8 (T1), 0.4–0.8 (T2), 0.5–0.8 (T3)
Kim et al., 2009⁶⁰	Three similar clusters based on prevalence and severity scores1. Mood-cognitive: Difficulty concentrating, difficulty sleeping, feeling sad, worrying, itching, feeling irritable2. Sickness-behavior: Pain, lack of energy, feeling drowsy3. Treatment-related/pain cluster	n=160Prostate and breast (after radiotherapy)	MSAS	Occurrence: α=0.3–1.0 (Cluster 1), 0.6–1.0 (Cluster 2), 0.7–0.8 (Cluster 3)Severity: α=0.8 (Cluster 1), 0.7 (Cluster 2), α=0.4 (Cluster 3)The factor solution derived using severity ratings fit the data betterBreast cancer had higher symptom cluster severity scores than prostate cancer
Kirkova et al., 2010¹²	Two common clusters1. Nausea, vomiting2. Neuropsychologic: Sleep problems, depression, anxiety3. Aerodigestive: Dysphagia, dyspnea, cough, hoarseness	n 1=922Advanced cancern 2=181Advanced cancer	38-item symptom checklist	r≥0.7Cluster symptoms dendogram
Kuo et al., 2002³⁵	Various symptom clusters	n=73Lung	SDS, POMS, Coping Strategies Scale	Symptom distress correlated to physical (fatigue, lack of appetite, insomnia, increased sputum, difficulty breathing) and psychological symptoms (fatigue-inertia, confusion bewilderment, depression-dejection)Anxiety-depression (r=0.9)
Kurtz et al., 2007³⁹	A: Nausea, vomiting, diarrhea, poor appetiteD: Insomnia, pain, fatigueE: Difficulty breathing, cough	n 1=110 (treatment)n 2=112 (control)	15-item symptom checklist	Intervention reduced Cluster D severities and number of hospitalizations.High symptom severity for Cluster D and E at baseline associated with elevated levels after interventionComorbid conditions associated with higher symptom severity for Clusters A and D after intervention
Maliski et al., 2008⁵⁰	1. Fatigue: Fatigue, sexual dysfunction, bowel dysfunction, urinary dysfunction2. Emotional distress: Fatigue, sexual dysfunction, bowel dysfunction, urinary dysfunction, pain	n=402Prostate	Prostate Cancer Index Short Form, SF-36 Health Status Survey	Fatigue-emotional distress (r=0.7), fatigue-pain (r=0.4)
Molassiotis et al., 2010⁶¹	1. GI: Nausea, vomiting, bloating, lack of appetite, weight loss, diarrhea, dry mouth, mouth sores, pain2. Hand/foot: Numbness/tingling in hand/feet, swelling of arms/legs3. Body image: Hair loss, do not like myself, skin changes4. Respiratory: SOB, cough5. Nutritional6. Emotional: Dizziness, feeling drowsy, feeling sad, worrying, feeling nervous, difficulty concentrating, dizziness, difficulty sleeping	n=143Heterogeneous	MSAS	Slightly different clusters identified across time
Olson et al., 2008⁶²	Various symptom pairs and their inter-relationships elucidated (e.g., the impact of anxiety on depression, drowsiness on tiredness, and appetite on well-being)	n=82Heterogeneous (terminal)	ESAS	Model explains >70% variation in each of two symptoms (tiredness and depression) and well-beingEffects of anxiety to depression (0.7), drowsiness to tiredness (0.9)Well-being improved with appetite and decreased anxiety and drowsiness, collectively account for 72% well-being variance
Reyes-Gibby et al., 2007⁶³	Fatigue, lack of appetite	n=48Pancreas (chemoradiation)	MDAS	Fatigue and lack of appetite formed a distinct grouping over time (r values not reported)
Ridner et al., 2005³⁸	Treatment-Related Lymphedema ClusterAlteration in limb sensation, loss of confidence in body, decreased physical activity, fatigue, psychosocial distress	n 1=64 breast cancer survivors with lymhedeman 2=64 breast cancer survivors without lymhedeman total=128	Symptom Checklist, Skin/Arm Condition, CESD, POMS-SF, FACT-Breast, ULL 27, WCLS	Arm size difference correlated to total symptoms number (r=0.2), mean symptom intensity (r=0.3), mean symptom distress (r=0.2)ECF-mean symptom intensity (r=0.2)Subjects with lymhedema had poorer QOL
Ryu et al., 2010⁷³	1. Pain/appetite: Pain, nausea, loss of appetite, fever, change in taste2. Fatigue-related: Feeling fatigued, back pain, lack of energy, sadness3. GI: Jaundice, stomach pain/discomfort, stomach swelling/cramps, diarrhea4. Itching, constipation	n=180Primary liver	HADS, FACT-Hep, Symptom Checklist	α=0.8 (Cluster 1), 0.7 (Cluster 2), 0.70 (Cluster 3), 0.5 (Cluster 4)High-symptom group had significantly poorer functional status and QOL
Sarna et al., 1997⁴⁵	1. GI distress: Nausea, appetite loss2. Emotional and physical suffering: Pain, bowel dysfunction, appearance, outlook3. Respiratory distress: Cough, difficulty breathing, insomnia4. Malaise: Fatigue, inability to concentrate, outlook	n=60Lung (women)	SDI	Four factor based model had 63% varianceKPS related to distress with difficulty breathing, bowel dysfunction, altered appearance, outlook, frequent pain, cough, fatigue, frequent and severe nausea (r range −0.3 to −0.5)
Tsai et al., 2010⁷⁵	1. Loss of energy: Fatigue, weakness2. Poor intake: Anorexia, taste alteration, dysphagia, constipation3. Autonomic dysfunction: Restlessness/heat, dizziness, insomnia, night sweats4. Aerodigestive impairment: Nausea/vomiting, abdominal fullness, dyspnea	n=427Advanced cancer	Symptom checklist	Six patients group with different clusters identifiedSurvival, functional status, and bone metastasis associated with certain clusters
Walsh and Rybicki, 2006¹⁰	1. Fatigue/Anorexia-Cachexia: Easy fatigue, weakness, anorexia, lack of energy, dry mouth, early satiety, weight loss, taste change2. Neuropsychologic: Sleep problems, depression, anxiety3. Upper GI: Dizzy spells, dyspepsia, belching, bloating4. Nausea/Vomiting: Nausea, vomiting5. Aerodigestive: Dysphagia, dyspnea, cough, hoarseness6. Pain: Pain, constipation7. Debility: Edema, confusion	n=922Heterogeneous (advanced)	38-item symptom checklist	r≥0.7Cluster symptoms dendogram
Wang et al., 2006⁶⁴	1. Steady: Pain, sore throat2. Early: Nausea, vomiting3. Early/Late: Fatigue, lack of appetite, drowsiness, sleep disturbance, dry mouth, distress	n=64Lung (locally advanced)	MDASI	Mixed-effects model identified three symptom clusters over the treatment courseEarly/late increase symptom cluster had the highest predictive value for total interference
Wang et al., 2008⁷⁴	1. General: Pain, Fatigue, sleep disturbance, distress, SOB, drowsiness, dry mouth, sadness, numbness2. GI: Nausea, vomiting	n=108Lung	MDASI-Taiwanese	Two factors explained 66% total variancer=0.3–0.9α=0.9 (Cluster 1), α=0.3 (Cluster 2)Cluster symptoms dendogram
Yamagishi et al., 2009⁶⁵	1. Fatigue, somnolence2. Pain, dyspnea, numbness3. Nausea, appetite loss, constipation4. Psychological distress	n=462Heterogeneous (chemotherapy)	MDASI	Cluster symptoms dendogram

Note: The studies are listed in chronological order.

r, Correlation coefficient; α, Cronbach's α coefficient; CI, confidence interval; GI, gastrointestinal; SOB, shortness of breath; QOL, quality of life; PTSD, post-traumatic stress disorder; KPS, Karnofsky Performance Status; ECOG PS, Eastern Cooperative Oncology Group Performance Score; RCT, randomized clinical trial; VAS, visual analogue scale; VRS, verbal rating scale; BDI, Beck Depression Inventory; BFI, Brief Fatigue Inventory; BPI, Brief Pain Inventory; HADS, Hospital Anxiety and Depression Scale; SSI, Symptom Severity Index; GSDS, General Sleep Disturbance Scale; POM, Painometer; MOS-SF, Medical Outcomes Study-Short Form; CES, Cranial Electrical Stimulation; CES-D, Center for Epidemiological Studies Depression; QOL-CA, Quality of Life-Cancer; FPI, Functional Performance Inventory; SF-36, Short-Form 36; FSQOLS, Fox Simple QOL Scale; COGMOS, Cognitive Functioning Subscale of the Medical Outcomes Scale; FACT-BR, Functional and Brain Subscales of the Functional Assessment of Cancer Therapy; PSQI, Pittsburgh Sleep Quality Index; MFSI-SF, Multidimensional Fatigue Symptom Inventory Short Form; SDS, Symptom Distress Scale; SDI, Symptom Distress Inventory; HCA, hierarchical cluster analysis; CFA, confirmatory factor analysis; EFA, explanatory factor analysis; NCI CTC, National Cancer Institute Common Toxicity Criteria; POMS, Profile of Mood States; FACT, Functional Assessment of Cancer Therapy; PSET, Physical Symptom Experience Tool; MSAS, Memorial Symptom Assessment Scale; MDASI, M.D. Anderson Symptom Inventory; ESAS, Edmonton Symptom Assessment Scale; C-PET, Checklist for Patients with Endocrine Therapy; IBCSG, International Breast Cancer Study; EORTC QLQ-C30, European Organization for Research and Treatment of Cancer Quality of Life Questionnaire; ULL, upper limb lymhedema; WCLS, Wesley Clinic Lymhedema Scale; FACT-Hep EWB, Functional Assessment of Cancer Therapy-Hepatobiliary Emotional Well-Being; SQLI, Spitzer Quality of Life Index; FSI, Fatigue Symptom Inventory.

Cluster analysis

Cluster analysis is used either to group subjects or symptoms. Patients are clustered into subgroups based upon particular common symptoms.⁷⁷ This is very different from clustering symptoms into groups, since individuals can experience multiple symptom clusters. HCA uses co-occurrences between pairs of symptoms to form clusters similar to gene mapping; this is usually conducted post hoc. The number of symptoms included and cluster content depends on the assessment instrument, and statistical cut-off value for significant correlation level.

Factor and principal component analysis

Factor analysis and PCA are used when validating symptom assessment instruments. This technique combines symptoms from datasets by assigning several symptoms into one group that is relatively independent in associations with other groups of symptoms. The underlying factors or components are based on the statistical strength within a correlational matrix. Correlations are not based on a single dependent symptom variable unlike multiple regression analysis used in identification of clinical SC. Internal consistency (tested by Cronbach α coefficient) is an other statistical method used to determine consistency.⁸⁴ An α value >0.7 or even 0.9 for clinical purposes supports internal consistency.⁸⁵ There are two types of factor analysis: exploratory (EFA) and confirmatory (CFA). EFA establishes the strength of relationship between symptoms, and determines the symptom factor structure, which explains the maximum amount of variance. CFA establishes the validity the factor model in the settings of convergent and divergent validity. EFA generates a theory about a cluster construct, while CFA uses a model supported by theory or previous research to validate a cluster.⁸⁶ Cluster studies that use only EFA should be interpreted as exploratory and associations should not be mistaken for causation. Therefore, any cluster also needs to be critically examined to see whether it is clinically consistent.

Controversies

Fatigue clusters: Role of primary cancer site or disease stage

Fatigue was sound to be the most frequent dependant symptom variable. Despite differences in population and design, fatigue consistently clustered with pain.⁷⁶ Fatigue-insomnia, fatigue-depression, and depression-pain clustered less so.^{8,30,76,87–93} The depression-fatigue cluster had a consistent negative effect on QOL across all studies.^25,30,31 The fatigue-pain relationship varied based on pain severity, fatigue duration, disease stage, and treatment.^94,95 Time of data collection during therapy affected symptom severity within a cluster.²⁰ Depression-fatigue-insomnia-pain cluster was associated with the lowest KPS and worst QOL when these symptoms were severe, and vice versa.⁷⁷ Correlations between fatigue and anxiety or depression were weak in advanced cancer but strong in healthy controls.^15,96 In advanced lung cancer, fatigue correlated with depression.⁹⁷ Two studies (and a clinical review) found severe fatigue clustered with dyspnea in advanced cancer.^96,97 Fatigue, pain, lack of appetite, dry mouth, insomnia, and drowsiness formed a “sickness cluster” in those with advanced cancer undergoing chemotherapy or radiation therapy.^9,47,64

Depression clusters: Symptom numbers and severity

The depression-fatigue-pain cluster during palliative radiation had different relationships which depended on symptom severity. Depression correlated with both pain and fatigue in different regression models depending on severity.¹⁵ Depression-fever-pain-trouble sleeping were one cluster, depression-fatigue-dyspnea-trouble sleeping another. Symptoms can participate in different clusters based on severity.¹⁵ Another study identified SC based on change in severity over time.⁶⁴ Both symptom prevalence and severity seem important in cluster analysis. Limited symptom numbers and different time frames for assessment “past month” and “past week” also influenced these symptom interactions, and consequent cluster composition.^15,76

Depression, fatigue and pain: Statistical methods and populations

Clusters of clinically selected symptoms involve depression, fever, fatigue, pain, dyspnea, and weight loss. Such clusters differ based on disease stage, population, symptom severity, and symptom number.¹⁵ Methodology also influences cluster composition. By HCA the relationship between depression, fatigue, and pain is inconsistent, although these symptoms cluster by factor analysis.^9,44,46,47 Differences in the symptom domains used may have influenced the cluster composition.⁷ In general, factor analysis uses symptom severity and distress; whereas HCA uses prevalence data.

Study populations also influenced the SC independent of statistical method. In newly diagnosed lung cancer, fatigue did not cluster with depression and pain by factor analysis.^43,44 The three symptoms clustered by the same method in breast cancer after completion of chemotherapy.⁴⁹ Depression, fatigue, and pain clustered by HCA at various stages of breast cancer⁴⁶ but not in advanced cancer.¹⁰ Variability in the depression-fatigue-pain cluster (analyzed by factor analysis or HCA) suggests either a multiple or complex symptom etiology, or an important influence of the statistical analysis itself.

Clusters and Body Systems

Gastrointestinal symptoms: Clusters and assessment instruments

Nausea and vomiting clustered consistently.^{9,10,44,98–100} Constipation also clustered with pain.¹⁰ Other gastrointestinal (GI) symptoms are found in various clusters: anorexia with nausea and vomiting,^44,45,49,99 dry mouth,⁷² constipation and dry mouth,³⁴ or with early satiety, weight loss, and taste changes.^10,91 In four studies anorexia clustered with fatigue, weight loss, and certain GI symptoms.^10,43,44,61 Dry mouth and early satiety are common, but are infrequently included in cancer symptom assessment instruments. This will significantly bias any cluster identified by these instruments.

Respiratory symptoms: Same symptoms, different clusters

Dyspnea consistently clusters with cough^10,45 but also with dysphagia and muscle weakness.¹⁰ This suggests that dyspnea has multiple mechanisms (e.g., muscle wasting, diaphragmatic fatigue). Dyspnea also clusters with insomnia and memory problems, perhaps from a neuropsychological etiology.^45,101

Psychologic symptoms: Are they a cluster?

Anxiety and depression cluster together consistently.^10,46 Insomnia and/or drowsiness cluster with anxiety and/or depression,^10,100 cough and difficulty breathing,⁴⁵ pain, fatigue, anorexia,⁵ and cognitive symptoms.¹⁰² Two studies found that anxiety and depression clustered before and after palliative radiotherapy.^37,53 Similar cluster patterns involving various affective symptoms are reported in other studies.^9,67,69

Cluster Validity

Methodological heterogeneity makes it difficult to draw firm conclusions about cluster validity and reliability. Neither clinical nor statistical symptom associations can be interpreted as causal.^4,14,17,103 The clinical relevance of cancer SC is still not defined. Clinically defined SC explore symptom interrelations only between two symptoms (or correlate a single symptom with function, PS, and comorbidities).^18,77 They cannot be used to validate SC that contain more than two symptoms without comprehensive data collection. In addition, low to moderate statistical correlations (r<0.5) were frequently used as significant values for such clusters; this is unconvincing evidence. Clusters during or after antitumor treatment cannot be generalized to advanced cancer.^82,102

Factor analysis or HCA objectively examines nonrandom symptom distribution without a preconceived relationship as a bias. Neither factor analysis nor HCA will correct for assessment bias; both are sensitive to outliers.¹⁰⁴ It is unclear if a single factor or a cluster of a limited symptom numbers can form a separate cluster with any reasonable internal consistency.^9,91 Clusters identified by HCA but have low to moderate correlation (r<0.70) should be treated cautiously. If clinically validated against patient outcomes, they might suggest a common underlying mechanism.^{8-10,13,43,44,46,98}

Few SC have been consistently reproduced even using similar assessment instruments and methods.^{9,43,44,47,98} Consistent symptom pairs such as nausea-vomiting, anxiety-depression were nevertheless identified as clusters by both methods, and thus support the cluster concept.^9,10,45,99 All studies involve significant subjective judgments about symptom numbers, dimensions, time frame, and correlation coefficients.

Cancer Symptom Clusters and Patient Care

The National Hospice Study endorsed a common terminal pathway in advanced cancer.¹⁰⁵ Others believe specific symptoms in terminal illness differ in prevalence, severity, and association, and depend on the disease.¹⁰⁶ Pain, fatigue, depression, and insomnia (although frequently clustered clinically) do not cluster consistently by statistical analysis. This supports the idea of complex symptom interrelationships, and also challenges the clinical utility of this particular cluster.

Presumably a common mechanism for some SC may exist, e.g., cytokine-induced symptoms.^79–81 Treatment of one symptom may positively influence the rest within the cluster. A common cluster pathophysiology would thus facilitate both assessment and optimize treatment. Symptoms within clusters may respond similarly to drug regimens. There is some evidence that this is true for gastrointestinal¹⁰⁷ and psychologic¹⁰⁸ symptoms. Specific treatment may not produce the same response for symptoms within a cluster.¹⁰⁸ Nausea and vomiting after chemotherapy is an example of a consistent treatment-related cluster.^9,47 Both improve with antiemetics. Opioids suppress cough and dyspnea¹⁰⁹; this is not true however for pain and constipation, the latter worsens with opioids.

Future Research

In this comprehensive clinical review, SC had an important clinical implication in cancer assessment, and symptom control although the clinical message of some clusters is unclear. Cluster type and composition depends on the populations, disease stage, assessment method (e.g., symptom domain, chronology, symptom wording), instrument, timeframe, and statistical method. All symptoms within a cluster should be considered and targeted for assessment and intervention; otherwise important therapeutic benefits might be overlooked. The link between individual clusters and specific pathophysiologic mechanisms is still unknown; this challenges the mechanistic cluster theory. Association does not mean causation; symptoms may cluster, but actually be indirectly related, i.e., pain, opioids, and constipation. Longitudinal cluster studies are needed to examine such relationships. Clinically and statistically identified cluster symptoms should be evaluated within treatment outcomes. A consensus around any statistical methodology employed should be established. Cluster methodology is not well established. Limited research data provides some methodological guidance. For cluster validation, symptom numbers, dimensions, timeframe, and scales are to be explicit and uniform. Clusters may change with disease stage. Different SC may also occur due to the cancer, or a result of antitumor treatment. Relationships between neuroimmunologic responses, proinflammatory cytokines, and SC need to be explored. Given a relationship between clusters and a specific biologic mechanism, a single treatment may ameliorate all symptoms within that cluster, rather than any individual symptom. Any relationship between cancer prognosis and SC also needs to be explored further.

Footnotes

Acknowledgments

Jordanka Kirkova was partially supported by a fellowship grant in end-of-life care from the Mt. Sinai Health Care Foundation. The Harry R. Horvitz Center for Palliative Medicine is a World Health Organization Demonstration Project in Palliative Medicine and an ESMO Designated Center of Integrated Oncology and Palliative Care.

Author Disclosure Statement

The authors declared no conflicts of interest with respect to authorship and/or publication of this article.

The authors received no financial support for the research and/or authorship of this article.

Appendix A.

Methodology of Literature Search

The following key words and search processes were used in this study:

1. MEDLINE (OVID) review was performed using the following key words: (neoplasms) AND (symptom cluster (words with any ending, in either order, in the article title, abstract, or indexing) OR symptom constellation OR concurrent symptom OR cluster study OR cluster studies OR multiple symptoms).

2. The study abstracts were examined by two individuals (J.K. and A.A.) independently.

3. Reference search by hand was performed using hard copies of the studies identified through MEDLINE.

4. Full texts selected for the review were independently assessed by J.K. and A.A.

References

Donnelly

, Walsh

, Rybicki

. The symptoms of advanced cancer: Identification of clinical and research priorities by assessment of prevalence and severity. J Palliat Care, 1995; 11:27–32.

Stromgren

, Goldschmidt

, Groenvold

et al. Self-assessment in cancer patients referred to palliative care: A study of feasibility and symptom epidemiology. Cancer, 2002; 94:512–520.

NIH State-of-the-Science Statement on symptom management in cancer:. Pain, depression, and fatigue. NIH Consens State Sci Statements, 2002; 19:1–29.

Miaskowski

, Dodd

, Lee

. Symptom clusters: The new frontier in symptom management research. J Natl Cancer Inst Monogr, 2004; 32:17–21.

Kim

, McGuire

, Tulman

, Barsevick

. Symptom clusters: concept analysis and clinical implications for cancer nursing. Cancer Nurs, 2005; 28:270–282; quiz 83–84.

Aktas

, Walsh

, Rybicki

. Symptom clusters: myth or reality? Palliat Med., 2010; 24:373–385.

Kirkova

, Walsh

. Cancer symptom clusters—A dynamic construct. Support Care Cancer, 2007; 15:1011–1013.

Dodd

, Miaskowski

, Paul

. Symptom clusters and their effect on the functional status of patients with cancer. Oncol Nurs Forum, 2001; 28:465–470.

Chen

, Tseng

. Symptom clusters in cancer patients. Support Care Cancer, 2006; 14:825–830.

10.

Walsh

, Rybicki

. Symptom clustering in advanced cancer. Support Care Cancer, 2006; 14:831–836.

11.

Aktas

, Walsh

. The importance of symptom clusters in cancer care. Eur J Palliat Care, 2010; 17:110–115.

12.

Kirkova

, Aktas

, Walsh

, Rybicki

, Davis

. Consistency of symptom clusters in advanced cancer. Am J Hosp Palliat Care, 2010; 27:342–346.

13.

Miaskowski

, Lee

. Pain, fatigue, and sleep disturbances in oncology outpatients receiving radiation therapy for bone metastasis: a pilot study. J Pain Symptom Manage, 1999; 17:320–332.

14.

Dodd

, Miaskowski

, Lee

. Occurrence of symptom clusters. J Natl Cancer Inst Monogr, 2004; 32:76–78.

15.

Francoeur

. The relationship of cancer symptom clusters to depressive affect in the initial phase of palliative radiation. J Pain Symptom Manage, 2005; 29:130–155.

16.

Aprile

, Ramoni

, Keefe

, Sonis

. Links between regimen-related toxicities in patients being treated for colorectal cancer. Curr Opin Support Palliat Care, 2009; 3:50–54.

17.

Barsevick

, Whitmer

, Nail

, Beck

, Dudley

. Symptom cluster research: Conceptual, design, measurement, and analysis issues. J Pain Symptom Manage, 2006; 31:85–95.

18.

Beck

, Dudley

, Barsevick

. Pain, sleep disturbance, and fatigue in patients with cancer: Using a mediation model to test a symptom cluster. Oncol Nurs Forum, 2005; 32:542.

19.

Steel

, Kim

, Dew

et al. Cancer-related symptom clusters, eosinophils, and survival in hepatobiliary cancer: An exploratory study. J Pain Symptom Manage, 2010; 39:859–871.

20.

, Marsh

, Ling

et al. The symptom cluster of fatigue, pain, anxiety, and depression and the effect on the quality of life of women receiving treatment for breast cancer: A multicenter study. Oncol Nurs Forum, 2009; 36:E205–214.

21.

Hoffman

, Given

, von Eye

, Gift

, Given

. Relationships among pain, fatigue, insomnia, and gender in persons with lung cancer. Oncol Nurs Forum, 2007; 34:785–792.

22.

Barsevick

, Dudley

, Beck

. Cancer-related fatigue, depressive symptoms, and functional status: A mediation model. Nurs Res, 2006; 55:366–372.

23.

Huang

, Lin

. The mediating effects of depression on sleep disturbance and fatigue: Symptom clusters in patients with hepatocellular carcinoma. Cancer Nurs., 2009; 32:398–403.

24.

Lyon

, Schubert

, Taylor

. Pilot study of cranial stimulation for symptom management in breast cancer. Oncol Nurs Forum, 2010; 37:476–483.

25.

Fox

, Lyon

. Symptom clusters and quality of life in survivors of lung cancer. Oncol Nurs Forum, 2006; 33:931–936.

26.

Fox

, Lyon

. Symptom clusters and quality of life in survivors of ovarian cancer. Cancer Nurs, 2007; 30:354–361.

27.

Myers

, Sousa

, Donovan

. Predictors of self-reported memory problems in patients with ovarian cancer who have received chemotherapy. Oncol Nurs Forum, 2010; 37:596–603.

28.

Chan

, Richardson

. A study to assess the existence of the symptom cluster of breathlessness, fatigue and anxiety in patients with advanced lung cancer. Eur J Oncol Nurs, 2005; 9:325–333.

29.

Reyes-Gibby

, Aday

, Anderson

, Mendoza

, Cleeland

. Pain, depression, and fatigue in community-dwelling adults with and without a history of cancer. J Pain Symptom Manage, 2006; 32:118–128.

30.

Gaston-Johansson

, Fall-Dickson

, Bakos

, Kennedy

. Fatigue, pain, and depression in pre-autotransplant breast cancer patients. Cancer Pract, 1999; 7:240–247.

31.

Fox

, Lyon

, Farace

. Symptom clusters in patients with high-grade glioma. J Nurs Scholarsh, 2007; 39:61–67.

32.

Thornton

, Andersen

, Blakely

. The pain, depression, and fatigue symptom cluster in advanced breast cancer: Covariation with the hypothalamic-pituitary-adrenal axis and the sympathetic nervous system. Health Psychol, 2010; 29:333–337.

33.

Kenefick

. Patterns of symptom distress in older women after surgical treatment for breast cancer. Oncol Nurs Forum, 2006; 33:327–335.

34.

Kenne Sarenmalm

, Ohlen

, Jonsson

, Gaston-Johansson

. Coping with recurrent breast cancer: Predictors of distressing symptoms and health-related quality of life. J Pain Symptom Manage, 2007; 34:24–39.

35.

Kuo

, Ma

. Symptom distresses and coping strategies in patients with non-small cell lung cancer. Cancer Nurs, 2002; 25:309–317.

36.

Kenne Sarenmalm

, Ohlen

, Oden

, Gaston-Johansson

. Experience and predictors of symptoms, distress and health-related quality of life over time in postmenopausal women with recurrent breast cancer. Psychooncology, 2008; 17:497–505.

37.

Chow

, Fan

, Hadi

et al. Symptom clusters in cancer patients with brain metastases. Clin Oncol (R Coll Radiol), 2008; 20:76–82.

38.

Ridner

. Quality of life and a symptom cluster associated with breast cancer treatment-related lymphedema. Support Care Cancer, 2005; 13:904–911.

39.

Kurtz

, Kurtz

, Given

. Symptom clusters among cancer patients and effects of an educational symptom control intervention. Cancer Ther, 2007; 5:105–112.

40.

Jarden

, Nelausen

, Hovgaard

, Boesen

, Adamsen

. The effect of a multimodal intervention on treatment-related symptoms in patients undergoing hematopoietic stem cell transplantation: A randomized controlled trial. J Pain Symptom Manage, 2009; 38:174–190.

41.

Gleason

Jr. , Case

, Rapp

et al. Symptom clusters in patients with newly-diagnosed brain tumors. J Support Oncol, 2007; 5:427–433, 436.

42.

Husain

, Bezjak

, Easson

. Malignant ascites symptom cluster in patients referred for paracentesis. Ann Surg Oncol, 2010; 17:461–469.

43.

Gift

, Stommel

, Jablonski

, Given

. A cluster of symptoms over time in patients with lung cancer. Nurs Res, 2003; 52:393–400.

44.

Gift

, Jablonski

, Stommel

, Given

. Symptom clusters in elderly patients with lung cancer. Oncol Nurs Forum, 2004; 31:202–212.

45.

Sarna

, Brecht

. Dimensions of symptom distress in women with advanced lung cancer: A factor analysis. Heart Lung, 1997; 26:23–30.

46.

Bender

, Ergyn

, Rosenzweig

, Cohen

, Sereika

. Symptom clusters in breast cancer across 3 phases of the disease. Cancer Nurs, 2005; 28:219–225.

47.

Chen

, Lin

. Cancer symptom clusters: a validation study. J Pain Symptom Manage, 2007; 34:590–599.

48.

Bender

, Engberg

, Donovan

et al. Symptom clusters in adults with chronic health problems and cancer as a comorbidity. Oncol Nurs Forum, 2008; 35:E1–E11.

49.

Kim

, Barsevick

, Tulman

, McDermott

. Treatment-related symptom clusters in breast cancer: A secondary analysis. J Pain Symptom Manage, 2008; 36:468–479.

50.

Maliski

, Kwan

, Elashoff

, Litwin

. Symptom clusters related to treatment for prostate cancer. Oncol Nurs Forum, 2008; 35:786–793.

51.

Breen

, Baravelli

, Schofield

et al.

Is symptom burden a predictor of anxiety and depression in patients with cancer about to commence chemotherapy?

Med J Aust, 2009; 190,7 Suppl:S99–104.

52.

Aprile

, Ramoni

, Keefe

, Sonis

. Application of distance matrices to define associations between acute toxicities in colorectal cancer patients receiving chemotherapy. Cancer, 2008; 112:284–292.

53.

Chow

, Fan

, Hadi

, Filipczak

. Symptom clusters in cancer patients with bone metastases. Support Care Cancer, 2007; 15:1035–1043.

54.

Dodd

, Cho

, Cooper

et al. Advancing our knowledge of symptom clusters. J Support Oncol, 2005; 3,6 Suppl 4:30–31.

55.

Hadi

, Zhang

, Hird

, de Sa

, Chow

. Validation of symptom clusters in patients with metastatic bone pain. Curr Oncol, 2008; 15:211–218.

56.

Hadi

, Fan

, Hird

et al. Symptom clusters in patients with cancer with metastatic bone pain. J Palliat Med, 2008; 11:591–600.

57.

Hayduk

, Olson

, Quan

, Cree

, Cui

. Temporal changes in the causal foundations of palliative care symptoms. Qual Life Res, 2010; 19:299–306.

58.

Hird

, Wong

, Zhang

et al. Exploration of symptoms clusters within cancer patients with brain metastases using the Spitzer Quality of Life Index. Support Care Cancer, 2010; 18:335–342.

59.

Kim

, Jahan

, Aouizerat

et al. Changes in symptom clusters in patients undergoing radiation therapy. Support Care Cancer, 2009; 17:1383–1391.

60.

Kim

, Jahan

, Aouizerat

et al. Differences in symptom clusters identified using occurrence rates versus symptom severity ratings in patients at the end of radiation therapy. Cancer Nurs, 2009; 32:429–436.

61.

Molassiotis

, Wengstrom

, Kearney

. Symptom cluster patterns during the first year after diagnosis with cancer. J Pain Symptom Manage, 2010; 39:847–858.

62.

Olson

, Hayduk

, Cree

et al. The changing causal foundations of cancer-related symptom clustering during the final month of palliative care: A longitudinal study. BMC Med Res Methodol., 2008; 8:36, 10.1186/1471-2288-8-36

63.

Reyes-Gibby

, Chan

, Abbruzzese

et al. Patterns of self-reported symptoms in pancreatic cancer patients receiving chemoradiation. J Pain Symptom Manage, 2007; 34:244–252.

64.

Wang

, Fairclough

, Liao

et al. Longitudinal study of the relationship between chemoradiation therapy for non-small-cell lung cancer and patient symptoms. J Clin Oncol, 2006; 24:4485–4491.

65.

Yamagishi

, Morita

, Miyashita

, Kimura

. Symptom prevalence and longitudinal follow-up in cancer outpatients receiving chemotherapy. J Pain Symptom Manage, 2009; 37:823–830.

66.

Cheung

, Le

, Zimmermann

. Symptom clusters in patients with advanced cancers. Support Care Cancer, 2009; 17:1223–1230.

67.

Cheung

, Le

, Gagliese

, Zimmermann

. Age and gender differences in symptom intensity and symptom clusters among patients with metastatic cancer. Support Care Cancer, 2011; 19:417–423.

68.

Fan

, Hadi

, Chow

. Symptom clusters in patients with advanced-stage cancer referred for palliative radiation therapy in an outpatient setting. Support Cancer Ther, 2007; 4:157–162.

69.

, Crew

, Jacobson

et al. Ethnicity and persistent symptom burden in breast cancer survivors. J Cancer Surviv, 2009; 3:241–250.

70.

Glaus

, Boehme

, Thurlimann

et al. Fatigue and menopausal symptoms in women with breast cancer undergoing hormonal cancer treatment. Ann Oncol, 2006; 17:801–806.

71.

Henoch

, Ploner

, Tishelman

. Increasing stringency in symptom cluster research: A methodological exploration of symptom clusters in patients with inoperable lung cancer. Oncol Nurs Forum, 2009; 36:E282–292.

72.

Karabulu

, Erci

, Ozer

, Ozdemir

. Symptom clusters and experiences of patients with cancer. J Adv Nurs, 2010; 66:1011–1021.

73.

Ryu

, Kim

, Cho

et al. Symptom clusters and quality of life in Korean patients with hepatocellular carcinoma. Cancer Nurs, 2010; 33:3–10.

74.

Wang

, Tsai

, Chen

, Lin

. Symptom clusters and relationships to symptom interference with daily life in Taiwanese lung cancer patients. J Pain Symptom Manage, 2008; 35:258–266.

75.

Tsai

, Wu

, Chiu

, Chen

. Significance of symptom clustering in palliative care of advanced cancer patients. J Pain Symptom Manage, 2010; 39:655–662.

76.

Armstrong

, Cohen

, Eriksen

, Hickey

. Symptom clusters in oncology patients and implications for symptom research in people with primary brain tumors. J Nurs Scholarsh, 2004; 36:197–206.

77.

Miaskowski

, Cooper

, Paul

et al. Subgroups of patients with cancer with different symptom experiences and quality-of-life outcomes: A cluster analysis. Oncol Nurs Forum, 2006; 33:E79–89.

78.

Lee

, Dantzer

, Langley

et al. A cytokine-based neuroimmunologic mechanism of cancer-related symptoms. Neuroimmunomodulation, 2004; 11:279–292.

79.

Capuron

, Ravaud

, Dantzer

. Early depressive symptoms in cancer patients receiving interleukin 2 and/or interferon alfa-2b therapy. J Clin Oncol, 2000; 18:2143–2151.

80.

Capuron

, Gumnick

, Musselman

et al. Neurobehavioral effects of interferon-alpha in cancer patients: Phenomenology and paroxetine responsiveness of symptom dimensions. Neuropsychopharmacology, 2002; 26:643–652.

81.

Illman

, Corringham

, Robinson

Jr.

et al.

Are inflammatory cytokines the common link between cancer-associated cachexia and depression?

J Support Oncol, 2005; 3:37–50.

82.

Morrow

, Hickok

, Roscoe

et al. Differential effects of paroxetine on fatigue and depression: A randomized, double-blind trial from the University of Rochester Cancer Center Community Clinical Oncology Program. J Clin Oncol, 2003; 21:4635–4641.

83.

Flores

, Mogil

. The pharmacogenetics of analgesia: Toward a genetically-based approach to pain management. Pharmacogenomics, 2001; 2:177–194.

84.

Streiner

, Norman

. Reliability in Health Measurement Scales A Practical Guide to Their Development and Use, 3rd. Oxford: Oxford University Press, 2003.

85.

Bland

, Altman

. Cronbach's alpha. BMJ., 1997; 314:572.

86.

DeCoster

. Overview of Factor Analysis. 1998. www.stat-help.com/notes.html. 2011 March 19.

87.

Blesch

, Paice

, Wickham

et al. Correlates of fatigue in people with breast or lung cancer. Oncol Nurs Forum, 1991; 18:81–87.

88.

Bower

, Ganz

, Desmond

et al. Fatigue in breast cancer survivors: Occurrence, correlates, and impact on quality of life. J Clin Oncol, 2000; 18:743–753.

89.

Broeckel

, Jacobsen

, Horton

, Balducci

, Lyman

. Characteristics and correlates of fatigue after adjuvant chemotherapy for breast cancer. J Clin Oncol, 1998; 16:1689–1696.

90.

Given

, Given

, Azzouz

, Kozachik

, Stommel

. Predictors of pain and fatigue in the year following diagnosis among elderly cancer patients. J Pain Symptom Manage, 2001; 21:456–466.

91.

Given

, Given

, Azzouz

, Stommel

. Physical functioning of elderly cancer patients prior to diagnosis and following initial treatment. Nurs Res, 2001; 50:222–232.

92.

Jacobsen

, Hann

, Azzarello

et al. Fatigue in women receiving adjuvant chemotherapy for breast cancer: Characteristics, course, and correlates. J Pain Symptom Manage, 1999; 18:233–242.

93.

Ciaramella

, Poli

. Assessment of depression among cancer patients: The role of pain, cancer type and treatment. Psychooncology, 2001; 10:156–165.

94.

Hickok

, Morrow

, McDonald

, Bellg

. Frequency and correlates of fatigue in lung cancer patients receiving radiation therapy: Implications for management. J Pain Symptom Manage, 1996; 11:370–377.

95.

Kaasa

, Loge

, Knobel

, Jordhoy

, Brenne

. Fatigue. Measures and relation to pain. Acta Anaesthesiol Scand, 1999; 43:939–947.

96.

Stone

, Hardy

, Broadley

et al. Fatigue in advanced cancer: A prospective controlled cross-sectional study. Br J Cancer, 1999; 79:1479–1486.

97.

Okuyama

, Tanaka

, Akechi

et al. Fatigue in ambulatory patients with advanced lung cancer: Prevalence, correlated factors, and screening. J Pain Symptom Manage, 2001; 22:554–564.

98.

Cleeland

, Mendoza

, Wang

et al. Assessing symptom distress in cancer patients: The M.D. Anderson Symptom Inventory. Cancer, 2000; 89:1634–1646.

99.

Okuyama

, Wang

, Akechi

et al. Japanese version of the MD Anderson Symptom Inventory: A validation study. J Pain Symptom Manage, 2003; 26:1093–1104.

100.

Wilmoth

, Coleman

, Smith

, Davis

. Fatigue, weight gain, and altered sexuality in patients with breast cancer: Exploration of a symptom cluster. Oncol Nurs Forum, 2004; 31:1069–1075.

101.

Mystakidou

, Cleeland

, Tsilika

et al. Greek M.D. Anderson Symptom Inventory: Validation and utility in cancer patients. Oncology, 2004; 67:203–210.

102.

Cordova

, Studts

, Hann

, Jacobsen

, Andrykowski

. Symptom structure of PTSD following breast cancer. J Trauma Stress, 2000; 13:301–319.

103.

Beck

. Symptom clusters: Impediments and suggestions for solutions. J Natl Cancer Inst Monogr, 2004; 137–138.

104.

Kotagal

, Luders

, Williams

, Nichols

, McPherson

. Psychomotor seizures of temporal lobe onset: Analysis of symptom clusters and sequences. Epilepsy Res, 1995; 20:49–67.

105.

Wachtel

, Allen-Masterson

, Reuben

, Goldberg

, Mor

. The end stage cancer patient: Terminal common pathway. Hosp J, 1988; 4:43–80.

106.

Klinkenberg

, Willems

, van der Wal

, Deeg

. Symptom burden in the last week of life. J Pain Symptom Manage, 2004; 27:5–13.

107.

Nelson

, Walsh

. Metoclopramide in anorexia caused by cancer-associated dyspepsia syndrome (CADS) J Palliat Care, 1993; 9:14–18.

108.

Theobald

. Cancer pain, fatigue, distress, and insomnia in cancer patients. Clin Cornerstone, 2004; 6,Suppl 1D:S15–21.

109.

Homsi

, Walsh

, Nelson

, LeGrand

, Davis

. Hydrocodone for cough in advanced cancer. Am J Hosp Palliat Care, 2000; 17:342–346.