Symptom clusters in patients with advanced cancer: a systematic review of observational studies.

Accepted Manuscript Symptom Clusters in Patients With Advanced Cancer: A Systematic Review of Observational Studies Skye Tian Dong, MSc Phyllis N. Butow, PhD Daniel S.J. Costa, PhD Melanie R. Lovell, MBBS, FRACP, FAChPM Meera Aga, FRACP, FAChPM, MPC, MBBS PII:

S0885-3924(14)00150-X

DOI:

10.1016/j.jpainsymman.2013.10.027

Reference:

JPS 8621

To appear in:

Journal of Pain and Symptom Management

Received Date: 27 August 2013 Revised Date:

21 October 2013

Accepted Date: 30 October 2013

Please cite this article as: Dong ST, Butow PN, Costa DSJ, Lovell MR, Aga M, Symptom Clusters in Patients With Advanced Cancer: A Systematic Review of Observational Studies, Journal of Pain and Symptom Management (2014), doi: 10.1016/j.jpainsymman.2013.10.027. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Review Article

13-00475R1

Symptom Clusters in Patients With Advanced Cancer: A Systematic Review of Observational Studies

RI PT

Skye Tian Dong, MSc, Phyllis N. Butow, PhD, Daniel S.J. Costa, PhD, Melanie R. Lovell, MBBS, FRACP, FAChPM, and Meera Aga, FRACP, FAChPM, MPC, MBBS

SC

Psycho-Oncology Co-operative Research Group (PoCoG) (S.T.D., P.N.B., D.S.J.C.), University of Sydney; HammondCare (M.R.L.), The University of Sydney Medical School; and Department

M AN U

of Palliative Care (M.A.), Braeside Hospital, HammondCare, Sydney, New South Wales, Australia

Address correspondence to:

School of Psychology Transient Building (F12)

EP

University of Sydney

TE D

Skye Tian Dong, MSc

Sydney, NSW 2006, Australia

AC C

E-mail: [email protected]

Number of Tables: 6

Number of Figures: 2

1

ACCEPTED MANUSCRIPT

Abstract Context: Advanced cancer patients typically experience multiple symptoms, which may influence patient outcomes synergistically. The composition of these symptom clusters (SCs)

RI PT

differs depending on various clinical variables, and the timing and method of their assessment. Objectives: The objectives of this systematic review were to examine the composition, longitudinal stability, and consistency across methodologies of common SCs, as well as their

SC

common predictors and outcomes.

Methods: A search of MEDLINE, CINAHL, EMBASE, Web of Science and PsycINFO

M AN U

was conducted using variants of symptom clusters, cancer, and palliative care. Results: 33 articles were identified and reviewed. Many SCs were identified, with four common groupings being anxiety-depression, nausea-vomiting, nausea-appetite loss, and fatigue-dyspnea-drowsiness-pain. SCs in most cases were not stable longitudinally. The various

TE D

statistical methods employed (most commonly principal components analysis, exploratory factor analysis and hierarchical cluster analysis) tended to reveal different SCs. Different measurement tools were employed in different studies, each containing a different array of symptoms. The

EP

predictors and outcomes of SCs were also inconsistent across studies. No studies of patient experiences of SCs were identified.

AC C

Conclusion: Although the articles reviewed revealed four groups of symptoms that tended to cluster, there is limited consistency in the way in which SCs and variables associated with them are identified. This is largely due to lack of agreement about a robust, clinically relevant definition of SCs. Future research should focus on patients’ subjective experience of SCs to inform a clinically relevant definition of SCs and how they are managed over time.

2

ACCEPTED MANUSCRIPT

Key Words: symptom clusters, advanced cancer, palliative care Running Title: Systematic Review of Symptom Clusters

AC C

EP

TE D

M AN U

SC

RI PT

Accepted for publication: October 30, 2013.

3

ACCEPTED MANUSCRIPT

Introduction It is well established that patients with incurable cancer typically experience multiple symptoms, consequently suffering a high symptom burden (1, 2). Recent literature in symptom

RI PT

management has emphasized a shift of focus from treating single symptoms to managing the dynamic nature of multiple symptom constellations (3, 4). However, currently there is no consensus on the definition of “symptom clusters” (SCs), either conceptually or

SC

methodologically, in research and clinical practice. A commonly used definition of SCs is two or more concurrent and interrelated symptoms that occur together, with a high degree of

M AN U

predictability (5, 6). Symptoms within a cluster may group together in a non-random fashion but do not require a common etiology (7, 8). Furthermore, SC composition may or may not be stable throughout the disease course, although some authors have proposed a refinement in the definition of SCs to include stability as a necessary aspect (9, 10).

TE D

There is a lack of consensus on what SCs exist in the advanced cancer population, with SCs identified by different studies varying widely (3, 5). SCs are associated with a variety of disease, demographic and biological variables, which may or may not explain a shared

EP

underlying mechanism. Some clusters may appear only in patients with a specific cancer location (for example, eating and speaking problems in patients with head and neck cancer) or after

AC C

specific treatments (such as fatigue, hair loss and nausea associated with some types of chemotherapy), whereas others may only appear in patients near to death. There have been two literature reviews to date examining multiple symptoms in advanced cancer (7, 8). Both reviews highlighted significant challenges in drawing definite conclusions about relationships between and among multiple symptoms for this population, because of significant methodological and conceptual variation among studies. These reviews did not examine the longitudinal stability of

4

ACCEPTED MANUSCRIPT

SCs in advanced cancer and to date, no reviews have explored whether SCs remain consistent across different statistical methodologies, what the predictors and outcomes of SCs are, and what the experiences of SCs are in patients with advanced cancer.

RI PT

Validation of the concept of SCs is difficult because assessing symptom inter-

relationships presents considerable methodological challenges. SC composition, consistency and stability varies widely depending on a host of measurement factors, including the optimal

SC

assessment tool (long versus short), the most clinically relevant symptom dimensions

(prevalence versus severity or distress caused), the optimal analytical method to derive the

M AN U

cluster, the optimal statistical “cut-off” points to define SCs and the optimal timing of assessment. Quantifying the nature of clusters over time requires the researcher to reach congruency between the analytical method, statistical assumptions and theoretical frameworks used to define SCs (9).

TE D

In addition, longitudinal study designs are clearly required to investigate stability of clusters over the illness trajectory and the underlying mechanisms as to why SCs differ over time (11). The temporal variability in symptom interrelationships for patients with advanced cancer

EP

limits the accuracy of SC profiles derived cross-sectionally (9, 12, 13). For example, Lasheen and colleagues (12) found that hospice patients with advanced cancer display daily fluctuations

AC C

in symptoms, with the only consistent symptom being depression, which predicted greater symptom burden and more severe symptoms. Factors such as variation in measurement timing, the number of symptoms included in an analysis and the inclusion of multiple items to measure the same construct in symptom assessment can bias the generalizability of SCs over time (9, 14). Such variability and incongruities in the measurement, composition and longitudinal stability of

5

ACCEPTED MANUSCRIPT

identified SCs has been argued to complicate interpretation, hinder comparability and undermine the clinical utility of the concept (5, 11). Co-occurring symptoms appear to have more complicated and synergistic detrimental

RI PT

associations with patients’ treatment outcomes, prognosis, functional status and quality of life than individual symptoms (15-18). Identifying the effect of SCs on patient outcomes and the impact of specific factors that predict SCs in patients with advanced cancer has been an area in

SC

need of more extensive research because of its potential to translate into targeted early

interventions (4, 7, 19). Although reviews have been undertaken to investigate the prognostic

M AN U

value of individual symptoms and multiple symptoms (16, 20, 21) in patients with advanced cancer, there is little consensus about which identified SCs predict patient outcomes such as prognosis and functional status. Furthermore, little is known about advanced cancer patients’ experiences of symptom clusters (22, 23) and to what degree individual characteristics and

TE D

coping styles mediate the impact of SCs on outcomes (24, 25). Understanding of the factors that predict the development of common SCs (5) is also limited. The lack of consensus in the current literature indicates the need for a systematic review

EP

that not only investigates what SCs have been identified in an advanced cancer population, but also how and why they differ across time, the stability of clusters over time and the impact of

AC C

different analytical methods on SC composition. Rationale and Aims

Although there have been two previous reviews of multiple symptoms in patients with advanced cancer (7, 8), no review has systematically investigated the issues raised above. In this review, we aim to investigate the following questions for the advanced cancer population: 1. What is the composition of common SCs?

6

ACCEPTED MANUSCRIPT

2. What is the longitudinal stability of SCs? 3. What is the consistency of SCs across different statistical methodologies? 4. What are the factors that predict SCs?

6. How do patients experience SCs? Methods

SC

Search Strategy and Eligibility Criteria for Identification of Studies

RI PT

5. What patient outcomes in advanced cancer are predicted by SCs?

A comprehensive literature search was performed in accordance with the guidelines of

M AN U

the AMSTAR tool to assess the methodological quality of systematic reviews and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (26, 27). The following databases were systematically searched in April 2013: MEDLINE, CINAHL, Embase, Web of Science and PsycINFO. Search results were limited to articles published from 1950 to April

TE D

2013. A total of 877 articles was retrieved from the electronic search. After duplicates were removed, 640 articles remained.

A comprehensive list of database search terms was developed. Derivatives of the terms

EP

were captured with the use of truncation symptom words appropriate to the respective databases searched. We combined three sets of key words: symptom* adj2 cluster* OR cancer OR

AC C

advanced adj2 [(cancer) OR (neoplasm)]. In order to ensure comprehensiveness in the systematic search, we also used the following derivatives for the three sets of key words above: Multiple adj2 symptom* symptom* adj2 constellation*, concurrent adj2 symptom*, Symptom* adj2 combination*, symptom* adj2 co-occurrence OR Neoplasm, oncol*, tumo$r, OR Metastas*, palliative, terminal care. . All searches were supplemented by consulting current contents, reviews and references in relevant studies.

7

ACCEPTED MANUSCRIPT

Eligibility criteria were developed to guide the selection of appropriate studies. Both quantitative (primary and secondary analysis of data sets) and qualitative studies that aimed to identify SCs or interrelationships between two or more symptoms in advanced cancer patients

RI PT

were eligible for inclusion. The reference lists of included studies also were hand-searched. Types of studies excluded were review papers, editorials, commentary/discussion papers,

theoretical papers, methodology papers, abstracts, papers focusing on one specific symptom,

SC

intervention papers, papers in languages other than English, and papers not available in full text. Participants eligible for inclusion were cancer patients with advanced disease only. There were

M AN U

no restrictions on the basis of cancer type or treatment. Articles that included children only or less than 20 advanced stage cancer patients as a subsample of a general group were excluded. Any type of medical setting, including but not limited to oncology, palliative care and general practice were eligible for inclusion. Finally, regarding symptoms, any studies that tested

TE D

statistically for a symptom cluster within a set of symptoms or reported predetermined SCs were included. Articles that did not examine relationships between and among two or more symptoms were excluded.

EP

The titles and abstracts were screened against the inclusion/exclusion criteria. Decisions regarding inclusion/exclusion were first made by one author (S.D.) and then verified by two co-

AC C

authors (D.C., P.B.). Ten percent of the titles and abstracts were independently screened by the third author (D.C.) to reduce selection bias in the search process. Two hundred eighty-four articles were excluded based on the title, 277 were excluded after reading the abstract, and a further 48 were excluded after reading the full text. Reference lists of included articles, and any studies that have cited these, also were manually searched for additional relevant articles; a further two articles were retrieved. The second author (P.B.) also read full texts of papers

8

ACCEPTED MANUSCRIPT

meeting the inclusion criteria to ensure there were no discrepancies in the inclusion of articles. A total of 33 articles were included in the review. Fig. 1 provides a flow chart of the article retrieval process, including the reasons for exclusion of articles.

RI PT

Data Extraction

Information was extracted from articles meeting the inclusion criteria by three authors (S.D., P.B., D.C.). One author (S.D.) inductively assessed all included articles and recorded the

SC

design, sample characteristics, main aims, analytical method(s), measure(s), major findings and limitations, as well as specific information relevant to each of the five key research questions.

M AN U

Similar findings were grouped according to the topic area and relevant studies were included under subheadings of each of the six review aims. Deductive data extraction techniques were subsequently used to re-examine each study and the standard format of extracted data was refined to include the dimensions of symptom measured and type of medical setting as data to be

TE D

extracted from the articles. Double coding was performed for data extraction for 50% of references by two authors (D.C., P.B.) to check for accuracy. All authors engaged in iterative discussions about the organization of findings and all disagreements in data extraction were

Quality Assessment

EP

resolved through consensus.

AC C

The 33 included studies underwent quality assessment utilizing the standardized Qualsyst assessment tool (28) (Fig. 2 shows the items used to assess quality). Qualsyst comprises a quantitative scale scoring system to evaluate study quality. This assessment tool was selected because it includes an extensive manual for quality scoring with definitions and detailed instructions, and is appropriate for the study designs encountered. The first author (S.D.) assessed the quality of all studies, and the second and third authors (P.B., D.C.) independently

9

ACCEPTED MANUSCRIPT

assessed 25% of studies each. Percentage agreement between the three raters was 94%. Agreement was defined as the proportion of items where both raters gave exactly the same score (e.g., 0, 1, or 2). Any identified discrepancies were resolved through iterative discussions and

RI PT

consensus. Each study was allocated a final score, which, as outlined by Lee et al. (29), was used to define the quality of the study as: limited (80%). The quality of included studies is reported in the summary tables. For full

SC

definitions and instructions for quality scoring, see Kmet et al. (28). Results

M AN U

Description of Studies

The 33 studies included in this review are summarized in Table 1. Design. Sixteen (48%) studies were of longitudinal design (symptoms measured at several time points) and 13 (39%) were of cross-sectional design. Four (12%) measured

TE D

symptoms at one time point and outcomes at a subsequent time point (18, 30-32) (Tables 1 and 4).

Sample Characteristics (Site, Stage). Twenty-seven (82%) studies evaluated

EP

heterogeneous cancers. Of these studies, the most prevalent cancer sites were: lung (14-59%), breast (8-26%), gastrointestinal (9-43%) and prostate (4-27%). Five (15%) studies were

AC C

conducted on single primary cancer sites only, with three (9%) studies examining only lung cancer (33-35) and two (6%) studies examining only breast cancer (32, 36). One study did not specify the nature of the sample (37). The age of patients ranged from 12 years to 95 years. Only one study examined patients who were newly diagnosed with advanced cancer (38). Three (9%) studies examined patients who had terminal cancer (30, 32, 39). Five (15%) studies

10

ACCEPTED MANUSCRIPT

evaluated patients with bone metastases only (40-44) and four (12%) studies patients with brain metastases only (45-48). Treatment and Setting. A majority of the included studies (26 [79%]) specified that they

RI PT

investigated both inpatient and outpatient advanced cancer samples referred to palliative care, supportive care and/or patients receiving symptom control. Twelve (36%) studies examined a patient sample that was receiving palliative radiotherapy. Three (9%) studies examined patients

SC

receiving concurrent chemoradiation therapy only (33, 35, 44), and two (6%) studies examined patients receiving heterogeneous chemotherapy, radiation or surgery (36, 38).

M AN U

The majority of studies used samples of patients referred to one setting, such as a hospital or palliative care program, whereas seven (21%) studies included patients with advanced cancer recruited from multiple sites (13, 32, 35, 44, 49-51).

Symptom Assessment Tools. Table 2 describes the symptom assessment tools used in all

TE D

33 studies. A large majority of studies (14 [42%]) used the Edmonton Symptom Assessment Scale (ESAS). Nineteen (58%) studies used formal assessment measures (including the ESAS). Two (6%) studies (35, 38) used the Symptom Distress Scale (SDS) and two (6%) (33, 44) used a

EP

modified version of the M. D. Anderson Symptom Inventory (MDASI). One study (50) used a modified version of the Memorial Symptom Assessment Scale (MSAS). Five (15%) studies used

AC C

single symptom assessment tools to measure symptoms that were included in forming SCs. The remaining studies used a variety of author-developed, non-validated multiple symptom assessment tools (Table 2, tools 6-9), which measured on average 23 symptoms (range: 9 – 38 symptoms).

Other measures used to assess outcomes of SCs include the BPI, which was included in two (6%) studies (42, 43) to measure functional interference as a result of pain; the Spitzer

11

ACCEPTED MANUSCRIPT

Quality of Life Index (SQLI), which was used in two (6%) studies (47, 48) to measure additional variables related to quality of life; and the Center for Epidemiologic Studies Depression Scale (CES-D) was included in two (6%) studies to measure depression.

RI PT

Symptom Dimensions. Only one study measured all three domains of symptom

occurrence, severity and distress (50). The majority (28 [85%]) of the studies measured the symptom dimension of severity. Five (9%) studies measured symptom interference in addition to

SC

severity (33, 36, 42-44). Two (6%) studies evaluated symptom occurrence only (34, 51). One study evaluated the dimension of difficulty controlling the symptom (49).

M AN U

Methodology. The most common methodology used to derive SCs was Principal Components Analysis (PCA) with varimax rotation, which was used in twelve (36%) studies, followed by Cluster Analysis used in five (15%) studies. Six (18%) studies used a combination of PCA, Hierarchical Cluster Analysis (HCA) and Exploratory Factor Analysis (EFA). No

TE D

studies used EFA only to derive clusters. Two (6%) studies used regression analyses (49, 51) , two (6%) studies (34, 35) used correlation and one study (30) used paired-sample t-tests to derive SCs. Three (9%) studies analyzed the symptoms within clusters (52, 53) or associations with

EP

other factors (36) using structural equation modeling. One study assumed clusters based on previous research (18) and growth modeling was used in one other study (33). Only eight (24%)

AC C

studies specified cut-off criteria used to define a cluster. Quality of Studies. Scores on the Qualsyst quality rating scale, completed for all 33 studies, are reported in Table 6. The average quality rating was 17.74 (range: 10 – 20). Results relating to each of the five research questions are detailed below. Research Question 1: What Are the Common Symptom Clusters in Advanced Cancer Patients?

12

ACCEPTED MANUSCRIPT

Studies answering the first research question (n=32, 97%) were separated into four major classifications of SCs, defined as: Group A - associations among symptoms; Group B subgroups of individuals (i.e., groups of individuals with similar symptom profiles or

RI PT

trajectories); Group C - symptoms that change together over time; and Group D - causal relations or moderating effects (Table 3).

The majority of studies were in Group A (n = 24, 73%), where associations between

SC

symptoms were examined and derived using methods such as factor analysis, cluster analysis or correlation. A total of three (9%) studies were in Group B, where SCs were grouped by either

M AN U

cancer site (38), severity of symptom profiles (13) or occurrence of symptom profiles (51). Two (6%) studies were in Group C, where SCs were grouped by trajectories of common patterns of longitudinal stability (33, 39). Two (6%) studies were in Group D, where SCs were implied by symptom interactions (49) and causal relations (52, 53).

TE D

Four main SCs were found to be the most commonly reported in the 32 studies: anxietydepression; nausea-vomiting; nausea-appetite loss; and fatigue-dyspnea-drowsiness-pain. The symptoms of anxiety-depression were co-occurring in 15 (45%) of the 32 studies.

EP

Within these, eight studies (53%) found that this cluster existed independently (i.e., these two symptoms only), and seven (46%) reported additional symptoms co-occurring with anxiety-

AC C

depression.

The symptoms of fatigue-dyspnea-drowsiness-pain also co-occurred in 15 (45%) of the 32 studies. Within these, only one study found this cluster existed independently (45). The composition of both the above clusters differed depending on the analytical techniques used, although the core symptoms tended to remain for anxiety-depression (41, 42, 46).

13

ACCEPTED MANUSCRIPT

The cluster of nausea-vomiting was reported by nine (28%) of the 32 studies. Of these nine studies, which used a variety of measurement tools, four (44%) found this cluster occurred independently.

RI PT

The cluster of nausea-appetite loss was reported by 13 (41%) of the 32 studies, where only one study (35) found it to occur independently. Of these 13 studies, 11 (85%) used the ESAS. The occurrence and composition of nausea-appetite loss varied, with different

SC

methodologies used to derive the cluster in two studies (41, 42).

A wide array of other SCs were identified in the included studies once or twice, and are

M AN U

listed in Table 3.

Research Question 2: What Is the Longitudinal Stability of Clusters (Consistency Over Time) of Symptom Clusters in Advanced Cancer Patients?

Table 4 lists the 12 (36%) studies that answered this question. Overall, eight (67%)

TE D

studies found that SCs did not display longitudinal stability. Two studies did find “stable” clusters: breathlessness, fatigue, and anxiety (34, 55); fatigue, pain, nausea, drowsiness, dyspnea, and loss of appetite; and anxiety, depression (34, 54). One study found that the longitudinal

EP

stability of certain symptoms (pain, anxiety, nausea, shortness of breath, drowsiness, appetite) were “moderate” in the context of increasing intensities (52) and another study found both stable

AC C

(effect of anxiety on depression, drowsiness on tiredness, and appetite on well-being) and unstable (effect of pain on appetite, anxiety on well-being, and depression on well-being) associations between different symptoms over time (53). Nine (75%) longitudinal studies assessed longitudinal stability over a three-month postradiation period at four to five time points, using pre-radiation treatment as a baseline, whereas one study (stable cluster of breathlessness, fatigue, and anxiety) did this over six weeks at three

14

ACCEPTED MANUSCRIPT

time points (34). Two studies (the moderate and mixed stability studies described earlier) assessed stability one month and one week before death (52, 53). Research Question 3: Do Different Methodologies Used to Examine Symptom Clusters in

RI PT

Advanced Cancer Produce Different Results?

Five (15%) studies answered this question (Table 4), four of which also assessed the longitudinal stability of SCs. All five studies concluded that the consistencies of SCs were

SC

unstable when the different methodologies of PCA, HCA and EFA were used. Two studies found that complete consensus between HCA, EFA and PCA was never reached at any assessment time

M AN U

point and observed little correlation between the three statistical methods, despite the use of an identical dataset (41, 42). Three studies (46, 48, 55) also found that clusters were unstable across methodologies such that the clusters varied at each time point with each methodology; however, it was revealed that cluster findings using PCA and HCA correlated more strongly with each

instruments ESAS and ADSQ.

TE D

other than either did with the results of the EFA. These results were derived using the assessment

Research Question 4: What Are the Factors That Predict Symptom Clusters in Patients With

EP

Advanced Cancer?

Six (18%) studies examined the associations between predictors and SCs (Table 5).

AC C

Variables were classified as “predictors” if the SCs were used as the dependent variable. Two studies examined the association of SCs with cancer site, and both found that cancer site influenced cluster composition (38, 56). Two studies also found associations between age and gender and SC composition; one found that the confusional cluster (agitation, confusion, urinary incontinence) was more common in patients older than 70 years and that the gastrointestinal cluster (nausea, vomiting) was more common in women (31), and the other found that younger

15

ACCEPTED MANUSCRIPT

patients reported worse pain and better appetite than older patients (57) . Two studies also examined the associations between biological correlates and the SC of pain, depression and fatigue; one found that elevated neuroendocrine levels predicted pain, depression and fatigue

RI PT

(36), and the other found no evidence that plasma C-reactive protein predicted the fatigue-paindepression cluster (51).

Research Question 5: What Patient Outcomes in Advanced Cancer Are Predicted by Symptom

SC

Clusters?

Seven (21%) studies examined the associations between SCs and outcomes (Table 5).

M AN U

Variables were classified as “outcomes” if the SCs were used to predict them. Three studies found a negative association between SCs and prognosis (18, 30, 31). The SC of pain, depression and fatigue was found to be associated with lower functional performance status in one study (36) and lower physical functioning in another (51). Two other

TE D

studies found negative associations between SCs and functional performance status (13, 30). Finally, Francoeur (49) used depression as an outcome and found that synergistic interactions of pain, fatigue, fever and weight loss predicted depressive affect, such that the association between

EP

pain and depressive affect was moderated by the other symptoms. Research Question 6: How Do Patients Experience Symptom Clusters?

AC C

No study has explored patient perceptions of symptom clusters. Discussion

The current review has demonstrated that clusters vary considerably in composition and stability, both longitudinally and depending on the methodology and assessment tool used. Variability in authors’ definitions of the term “symptom cluster” was a central hindrance in drawing general conclusions from the included studies. Cut-offs used to determine whether

16

ACCEPTED MANUSCRIPT

symptoms had sufficient prevalence to be included in SCs were inconsistent, and whether other items such as functional status constitute SCs or are an outcome of SCs was often unclear. Methodological Issues

RI PT

Longitudinal Stability. Most of the studies that examined the longitudinal stability of SCs in this review concluded that SCs found at baseline were unstable or of mixed stability over time, regardless of the type of statistical analysis employed. Conceptually, Kirkova and colleagues

SC

(24) considered a cluster “longitudinally stable” if it contained 75% of symptoms within the initial cluster across time points, including the “sentinel” symptom, defined as the most

M AN U

prevalent, or “core” or “defining” symptom, which acts as a marker for the presence of a SC. However, prevalence alone may not be the best way to define longitudinal stability or to examine the associations between sentinel and non-sentinel symptoms, as other dimensions (e.g., severity or distress) might alter cluster composition. In addition, Kirkova and colleagues (37) did not find

symptoms in the cluster.

TE D

that the presence or absence of a sentinel symptom predicted the presence or absence of other

Two studies (34, 58) found “longitudinally stable” SCs in this review, but one study only

EP

assessed the three symptoms of breathlessness, fatigue and anxiety and employed a small sample (N = 27) with a high proportion of males (23 vs. 4) in a specific population (lung cancer patients

AC C

undergoing radiation) (34). Yennurajalingam and colleagues (58) found that physical and psychological symptom clusters were differentiated at initial visits and consistent across time at follow-up. However, the significantly higher baseline mean intensity of physical SC scores compared with psychological SC scores suggests either that psychological symptoms were less severe than physical symptoms in their sample, or that the two-item “psychological” SC measure of anxiety and depression did not comprehensively capture the full range of psychological

17

ACCEPTED MANUSCRIPT

distress experienced by advanced cancer patients. Two studies took a novel approach to defining SCs, by forming them based on how symptoms changed longitudinally, which may be more clinically relevant. Tsai and colleagues (39) found six distinct longitudinal trajectories in

RI PT

symptom intensity for 77 terminal cancer patients in a palliative care setting. Two symptom trajectories did not improve over time with palliative care. These trajectories were: a “staticincrease” trajectory indicative of the anorexia-cachexia syndrome, which included symptoms of

SC

fatigue, weakness, nausea/vomiting, taste alteration, dysphagia, diarrhea, dry mouth, and night sweats; and a “decrease-increase” trajectory that included symptoms of anorexia and dyspnea.

M AN U

Another study, which investigated 64 incurable non-small-cell lung cancer patients, found four distinct longitudinal patterns of change in physical and psychological symptoms that were present before, during, and after chemoradiation therapy (33). Both studies highlight the need to be vigilant to symptom patterns that have common longitudinal changes in clinical practice.

TE D

Research into the developmental time course of symptom clusters has the potential to better inform patients as to when to expect the greatest impact from their multiple symptoms. Furthermore, these studies provide greater understanding of the temporal pattern of symptom

EP

clusters, which can lead to the development of biological mechanism-driven symptom management and better SC prevention and control.

AC C

Analytical Methods. Findings of all SC studies included in this review must be interpreted with caution, as data across studies may not be comparable when consistency in SCs cannot be reached in the same dataset. All five studies that examined whether the implementation of varying analytical methods derive different clusters suggested that cluster composition varied according to the statistical method employed.

18

ACCEPTED MANUSCRIPT

Interestingly, three studies (46, 48, 55) revealed that cluster findings using PCA and HCA were more similar to each other than to the results of EFA, yet the authors suggest that EFA and HCA are more appropriate methods than PCA, as PCA generates clusters based on total

RI PT

variance, rather than first removing error variance. Although this is a start in defining the

“optimal” method, the disparity in methodology and lack of consensus on which statistical method should be employed to extract symptom clusters, are both limitations in this field. This

SC

highlights the need for a different approach to defining SCs, such as qualitative approaches to look beyond symptom associations, multiple methods of analysis within the same dataset (59) or

M AN U

a mixed-methods approach to defining SCs. Recent awareness of the importance of qualitative approaches in SC research (5, 22) has highlighted the notion that exploring the perspective of the patient by examining their SC experience and the meaning that SCs hold for them can form a stronger conceptual basis for SCs. Although the patient experience of individual symptoms such

TE D

as fatigue (60) and breathlessness (61) have been explored qualitatively, there has only been a handful of studies exploring the multiple symptom experience qualitatively, specific to gynecological (23), gastrointestinal (62), prostate (63) and lung cancers (64, 65). No qualitative

EP

studies to date have been conducted in advanced cancer to examine the experience, meaning and impact of SCs for patients.

AC C

Limited Symptom Assessment. A vital limitation of studies included in the review was that 42% of them employed the nine-item ESAS, which does not assess a comprehensive range of symptoms, including sleep and bowel problems, which are known to be important to patients. Others assessed a more comprehensive range of symptoms using author-developed checklists, which did not have established psychometric properties. In most studies, symptom dimensions other than frequency and severity, such as distress or controllability, were not measured.

19

ACCEPTED MANUSCRIPT

Whether clusters derived using only one dimension are clinically relevant or consistent with clusters derived using other dimensions is currently unknown. In addition, the prevalence thresholds used to determine whether symptoms will be included in the statistical determination

RI PT

of clusters ranged from >15% to 75% for studies included in this review. Often the value was neither explicitly set in advance nor specifically listed for individual symptoms, leading to the possibility of post hoc definitions of clusters with various factor loading scores (45, 66).

SC

Common Symptom Clusters

Taking into consideration the current limitations of the studies included in this review, it

M AN U

appears there are four distinct combinations of symptoms that form a base from which clusters can be described. The anxiety-depression combination appears to be the most robust across studies, time and methodologies. It constituted a cluster in almost half of the studies and is consistent across cancer subtype (56), age, and gender (67). It is important to note that all 15

TE D

studies used single-item measures of anxiety and depression, with 11 of them using the same assessment tool, the ESAS. Thus, “anxiety” and “depression” may represent more general psychological distress rather than clinically significant depression or anxiety, and further

EP

psychological symptoms may be under-represented. In some studies but not others anxiety and depression were associated with well-being and sleep disturbance.

AC C

The nausea-vomiting symptom combination appeared to be more independent (i.e., occurring without additional symptoms) and robust than nausea-appetite loss across studies. Interestingly, all three co-occurred in only one study, which was rated as low quality (44). Of the ten studies that used an assessment tool that included nausea and vomiting, nine found it clustered. Nonetheless, caution must be taken in interpreting these findings, as patterns across studies included in this review may largely be a consequence of the assessment tool chosen. For

20

ACCEPTED MANUSCRIPT

example, the ESAS contains nausea and appetite loss but not vomiting, whereas authordeveloped symptom checklists and reporting forms (30, 68) contained nausea and vomiting, but not appetite loss. In contrast to nausea-vomiting, nausea-appetite appeared to cluster

RI PT

independently in only one study (35) and co-occurred alongside a range of other symptoms in 12 studies. The composition of nausea-appetite loss varied with the different methodologies used to derive the cluster (41, 42).

SC

Fifteen studies included in our review contained clusters based on two or more of the symptoms fatigue, dyspnea, drowsiness and pain, although only one study found this to exist as

M AN U

an independent cluster (45). Thus it appears that fatigue, dyspnea, drowsiness and pain often occur in combination, sometimes alongside other symptoms, depending on cancer site, symptoms assessed and methodology used. In two studies, this cluster combined with nauseaappetite (54), but without pain in one study (56). Cheung and colleagues (56) suggested that

TE D

symptoms in this combined cluster of fatigue-dyspnea-drowsiness-nausea-appetite are implicated in the anorexia–cachexia syndrome and are also all potential side effects of chemotherapy. Proinflammatory cytokines may play a fundamental role in causing cancer-related anorexia-cachexia

EP

and may, therefore, be involved in the etiology of this group of symptoms. Nonetheless, they found that primary cancer site changed the composition of this cluster dramatically. Therefore,

AC C

the precise manner in which nausea-appetite loss interacts with fatigue-dyspnea-drowsiness-pain and the exact sequence in which they occur is unclear, suggesting that further prospective studies are warranted.

Cheung and colleagues (56) found that pain and drowsiness clustered together for cancers involving the central nervous system and head and neck, which could indicate that pain is particularly difficult to control for these sites and leads to increased opioid-induced drowsiness.

21

ACCEPTED MANUSCRIPT

A limitation of this collection of symptoms, which formed a base from which clusters can be described, was the inconsistency of terminology used to describe “fatigue,” which was sometimes used interchangeably or in combination with “lack of energy,” “drowsiness” and

RI PT

“sleep disturbance.” Predictors and Outcomes

Cancer site, age and gender predicted SC composition, highlighted by four studies that

SC

directly explored the influence of these variables (31, 38, 56, 67). However, it is difficult to draw conclusions because of the limited number of studies that have investigated each predictor.

M AN U

Several biologic markers have been suggested in the pathophysiology of cancer-related symptoms and two studies included in this review have investigated the biological markers associated with the SC of pain, depression and fatigue in patients with advanced cancer (36, 51). C-reactive protein has previously been a marker for systemic inflammation, which is known to

TE D

be associated with cancer-related pain (69). Furthermore, there is evidence to suggest that the release of treatment-induced pro- and anti-inflammatory cytokines such as interleukin [IL]-6 are implicated as mechanisms underlying the development of multiple symptoms (70, 71). Wang and

EP

colleagues found that overexpressed pro-inflammatory cytokines resulted in the significant worsening of the severity of symptoms in non-small cell lung patients undergoing concurrent

AC C

chemoradiation therapy.

Francoeur (49) aimed to explain the complex relationship between pain, depressive affect, fatigue and other symptoms, finding that depressive affect was an outcome of the synergistic interactions of pain, fatigue, fever and weight loss, such that the association between pain and depressive affect was moderated by the other symptoms. Two other research groups also have attempted to examine causal relations between symptoms (52, 53). More research into

22

ACCEPTED MANUSCRIPT

the causal mechanisms of SCs and their relationship to biological markers, as well as other predictors, is required. Although there is limited evidence of the prognostic role of SCs on cancer survival, three

RI PT

studies support the notion that symptom burden is associated with cancer survival (18, 30, 31). Another outcome which seemed to be influenced by SCs was functional performance status, although the question of whether functional status acts as a predictor, mediator, outcome or a set

SC

of symptoms to be included in a SC was inconsistent across studies in this review. Limitations of Studies Reviewed

M AN U

Besides the limitations already discussed, general limitations of the group of studies included in this review include inadequate sample sizes, exclusion of patients with cognitive limitations and the over-representation of cancer patients who were undergoing radiotherapy. Furthermore, psychological symptoms were generally not well measured, with various items

TE D

included from well-being to a single-item measure of depression. Studies were restricted by attrition and missing data, along with recruitment from one sample and one setting. For longitudinal studies, varying subgroup membership rather than the same sample over time was a

Conclusion

EP

limitation.

AC C

Overall, our review demonstrated the strong need for consistency in defining what, why and how clusters of symptoms co-exist in patients with advanced cancer. It is difficult to reach definitive conclusions from the studies reviewed because of a host of methodological inconsistencies apparent in this field. Greater clarity on this topic can only be achieved when some agreement is reached on a common clinically relevant definition of symptom clusters, optimal analytical method, symptom dimensions to be included and assessment tools to be used.

23

ACCEPTED MANUSCRIPT

Qualitative studies in this field may add to a substantial gap in the literature about which clusters are clinically relevant and what the patient is actually experiencing when they have a cluster of symptoms. No studies to date have examined advanced cancer patients’ subjective experiences

systematic review to address inconsistencies and gaps in the literature. Disclosures and Acknowledgments

RI PT

of symptom clusters and how they are managed over time. Future research can be guided by this

SC

No funding was received for this study and the authors have no conflicts of interest to declare. AU: CONFIRM THAT NO FUNDING WAS RECEIVED.

1.

M AN U

References

Teunissen SCCM, de Graeff A, Voest EE, de Haes JCJM. Are anxiety and depressed

mood related to physical symptom burden? A study in hospitalized advanced cancer patients. Palliat Med 2007; 21:341-346.

Davis MP, Kirkova J. Lifting symptom burden - how far off the ground are we? Support

TE D

2.

Care Cancer 2008;16:757-761. 3.

Xiao C. The state of science in the study of cancer symptom clusters. Eur J Oncol Nurs

4.

Esper P. Symptom clusters in individuals living with advanced cancer. Semin Oncol Nurs

AC C

2010;26:168-174. 5.

EP

2010;14:417-34.

Aktas A. Cancer symptom clusters: current concepts and controversies. Curr Opin

Support Palliat Care 2013;7:38-44. 6.

Miaskowski, Dodd M, Lee K. Symptom clusters: the new frontier in symptom

management research. J Natl Cancer Inst Monogr 2004(32):17-21.

24

ACCEPTED MANUSCRIPT

7.

Gilbertson-White S, Aouizerat BE, Jahan T, Miaskowski C. A review of the literature on

multiple symptoms, their predictors, and associated outcomes in patients with advanced cancer. Palliat Support Care 2011;9:81-102. Bedi H, Hird A, Campos S, Chow E. Symptom clusters in metastatic cancer: a critical

appraisal. J Pain Manag 2010;3:17-29. 9.

Kim HJ, Abraham I, Malone P. Analytical methods and issues for symptom cluster

Kirkova J, Walsh D, Aktas A, Davis MP. Cancer symptom clusters: old concept but new

M AN U

data. Am J Hosp Palliat Care 2010;27:282-288. 11.

SC

research in oncology. Curr Opin Support Palliat Care 2013;7:45-53. 10.

RI PT

8.

Kirkova J, Aktas A, Walsh D, Davis MP. Cancer symptom clusters: clinical and research

methodology. J Palliat Med 2011;14:1149-1166. 12.

Lasheen W, Walsh D, Hauser K, Gutgsell T, Karafa MT. Symptom variability during

TE D

repeated measurement among hospice patients with advanced cancer. Am J Hosp Palliat Med 2009;26:368-375. 13.

Husain A, Myers J, Selby D, Thomson B, Chow E. Subgroups of advanced cancer

201;14:1246-1253.

Aktas A, Kirkova J, Walsh D, et al. The psychometric properties of cancer multi-

AC C

14.

EP

patients clustered by their symptom profiles: quality-of-life outcomes. J Palliat Med

symptom assessment instruments: a comprehensive review. J Pain Symptom Manage 2012;43:334-335. 15.

Kim JE, Dodd MJ, Aouizerat BE, Jahan T, Miaskowski C. A review of the prevalence

and impact of multiple symptoms in oncology patients. J Pain Symptom Manage 2009; 37:715736.

25

ACCEPTED MANUSCRIPT

16.

Trajkovic-Vidakovic M, de Graeff A, Voest EE, Teunissen SCCM. Symptoms tell it all:

a systematic review of the value of symptom assessment to predict survival in advanced cancer patients. Crit Rev Oncol Hematol 2012;84:130-148. Fan G, Filipczak L, Chow E. Symptom clusters in cancer patients: a review of the

RI PT

17.

literature. Curr Oncol 2007;14:173-179. 18.

Aktas A, Walsh D, Rybicki L. Symptom clusters and prognosis in advanced cancer.

19.

SC

Support Care Cancer 2012;20:2837-2843.

Berger AM, Yennub S, Millionc R. Update on interventions focused on symptom

M AN U

clusters: what has been tried and what have we learned? Curr Opin Support Palliat Care 2013;7:54-59. 20.

Teunissen SC, Graeff Ad, de Haes HC, Voest EE. Prognostic significance of symptoms

of hospitalised advanced cancer patients. Eur J Cancer 2006;42:2510-2516. Wilkie DJ, Ezenwa MO. Pain and symptom management in palliative care and at end of

TE D

21.

life. Nurs Outlook 2012;60:357-364. 22.

Bennion A, Molassiotis A. Qualitative research into the symptom experiences of adult

2013;21:9-25.

Lopez V, Copp G, Brunton L, Molassiotis A. Symptom experience in patients with

AC C

23.

EP

cancer patients after treatments: a systematic review and meta-synthesis. Support Care Cancer

gynecological cancers: the development of symptom clusters through patient narratives. J Support Oncol 2011;9:64-71. 24.

Kirkova J, Walsh D, Rybicki L, et al. Symptom severity and distress in advanced cancer.

Palliat Med 2010;24:330-339.

26

ACCEPTED MANUSCRIPT

25.

Lobb E, Kearsley J, Lacey J, Liauw W, Thomas D. A study to explore factors which

contribute to coping with uncertainty in patients with advanced cancer. Psychooncology 2011;20:66-67. Shea BJ, Grimshaw JM, Wells GA, et al. Development of AMSTAR: a measurement tool

RI PT

26.

to assess the methodological quality of systematic reviews. BMC Med Res Methodol 2007;7:10. 27.

Moher D, Liberati A, Tetzlaff J, Altman DG. Preferred reporting items for systematic

28.

SC

reviews and meta-analyses: the PRISMA statement. Ann Intern Med 2009;151:264-269.

Kmet LM, Lee RC, Cook LS. Standard quality assessment criteria for evaluating primary

M AN U

research papers from a variety of fields. Edmonton, Albert: Alberta Heritage Foundation for Medical Research, 2004. 29.

Lee L, Packer TL, Tang SH, Girdler S. Self-management education programs for age-

related macular degeneration: a systematic review. Australas J Ageing 2008;27:170-176. Tsai JS, Wu CH, Chiu TY, Chen CY. Significance of symptom clustering in palliative

TE D

30.

care of advanced cancer patients. J Pain Symptom Manage 2010;39:655-662. 31.

Jimenez A, Madero R, Alonso A, et al. Symptom clusters in advanced cancer. J Pain

32.

EP

Symptom Manage 2011;42:24-31.

Bender CM, Ergun FS, Rosenzweig MQ, Cohen SM, Sereika SM. Symptom clusters in

33.

AC C

breast cancer across 3 phases of the disease. Cancer Nurs 2005;28:219-225. Wang XS, Fairclough DL, Liao Z, 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.

27

ACCEPTED MANUSCRIPT

34.

Chan CWH, Richardson A, Richardson J. 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. Sarna L, Brecht M. Dimensions of symptom distress in women with advanced lung

RI PT

35.

cancer: a factor analysis. Heart Lung 1997;26:23-30. 36.

Thornton LM, Andersen BL, Blakely WP. The pain, depression, and fatigue symptom

SC

cluster in advanced breast cancer: covariation with the hypothalamic-pituitary-adrenal axis and the sympathetic nervous system. Health Psychol 2010;29:333-337.

Kirkova J, Aktas A, Walsh D, Rybicki L, Davis MP. Consistency of symptom clusters in

M AN U

37.

advanced cancer. Am J Hosp Palliat Med 2010;27:342-346. 38.

Fodeh SJ, Lazenby M, Bai M, et al. Functional impairments as symptoms in the symptom

cluster analysis of patients newly diagnosed with advanced cancer. J Pain Symptom Manage

39.

TE D

2013; 46;500-510.

Tsai JS, Wu CH, Chiu TY, Hu WY, Chen CY. Symptom patterns of advanced cancer

patients in a palliative care unit. Palliat Med 2006;20:617-622. Chow E, Fan G, Hadi S, Filipczak L. Symptom clusters in cancer patients with bone

EP

40.

metastases. Support Care Cancer 2007;15:1035-1043. Khan L, Cramarossa G, Chen E, et al. Symptom clusters using the edmonton symptom

AC C

41.

assessment system in patients with bone metastases: a reanalysis comparing different statistical methods. World J Oncol 2012;3:23-32. 42.

Chen E, Khan L, Zhang L, et al. Symptom clusters in patients with bone metastases - a

reanalysis comparing different statistical methods. Support Care Cancer 2012;20:2811-2820.

28

ACCEPTED MANUSCRIPT

43.

Hadi S, Fan G, Hird AE, et al. Symptom clusters in patients with cancer with metastatic

bone pain. J Palliat Med 2008;11:591-600. 44.

Shi Z, Qiu H, Yu S. The investigation of symptoms burden and treatment status in

45.

RI PT

patients with bone metastasis. Chinese-German J Clin Oncol 2010;9:63-67.

Chow E, Fan G, Hadi S, et al. symptom clusters in cancer patients with brain metastases.

Clin Oncol 2008;20:76-82.

Chen E, Khan L, Zhang L, et al. Symptom clusters in patients with brain metastases--a

SC

46.

reanalysis comparing different statistical methods. J Radiat Oncol 2013;2:95-102. Hird A, Wong J, Zhang LY, et al. Exploration of symptoms clusters within cancer

M AN U

47.

patients with brain metastases using the Spitzer Quality of Life Index. Support Care Cancer 2010;18:335-342 48.

Khan L, Cramarossa G, Lemke M, et al. Symptom clusters using the Spitzer Quality of

TE D

Life Index in patients with brain metastases - a reanalysis comparing different statistical methods. Support Care Cancer 2013;21:467-473. 49.

Francoeur RB. The relationship of cancer symptom clusters to depressive affect in the

50.

EP

initial phase of palliative radiation. J Pain Symptom Manage 2005;29:130-155. Chaiviboontham S, Viwatwongkasem C, Hanucharurnkul S, McCorkle R. Symptom

51.

AC C

clusters in Thais with advanced cancer. Pacific Rim Int J Nurs Res 2011;15:265-277. Laird BJ, Scott AC, Colvin LA, et al. Pain, depression, and fatigue as a symptom cluster

in advanced cancer. J Pain Symptom Manage 2011;42:1-11 52.

Hayduk L, Olson K, Quan H, Cree M, Cui Y. Temporal changes in the causal

foundations of palliative care symptoms. Qual Life Res 2010;19:299-306.

29

ACCEPTED MANUSCRIPT

53.

Olson K, Hayduk L, Cree M, 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. Yennurajalingam S, Kwon JH, Urbauer DL, et al. Consistency of symptom clusters

RI PT

54.

among advanced cancer patients seen at an outpatient supportive care clinic in a tertiary cancer centre. Palliat Support Care 2013;7:1-8.

Chen E, Nguyen J, Khan L, et al. Symptom clusters in patients with advanced cancer: a

SC

55.

reanalysis comparing different statistical methods. J Pain Symptom Manage 2012;44:23-32. Cheung WY, Le LW, Zimmermann C. Symptom clusters in patients with advanced

M AN U

56.

cancers. Support Care Cancer 2009;17:1223-1230. 57.

Cheung WY, Le LW, Zimmermann C. Symptom clusters in patients with advanced

cancers. Support Care Cancer 2009;17:1223-1230. DUPLICATE OF 56 Yennurajalingam S, Urbauer DL, Chacko R, et al. Impact of an outpatient

TE D

58.

interdisciplinary team (IDT) consultation on symptom clusters in advanced cancer patients seen at a supportive care outpatient clinic (OSC) in a tertiary cancer center. J Clin Oncol 2009;27.

59.

EP

Abstract e20542.

Calman L, Brunton L, Molassiotis A. Developing longitudinal qualitative designs:

AC C

lessons learned and recommendations for health services research. BMC Med Res Methodol 2013;13:14. 60.

Magnusson K, Möller A, Ekman T, Wallgren A. A qualitative study to explore the

experience of fatigue in cancer patients. Eur J Cancer Care 1999;8:224-232. 61.

Gysels M, Bausewein C, Higginson IJ. Experiences of breathlessness: a systematic

review of the qualitative literature. Palliat Support Care 2007;5:281-302.

30

ACCEPTED MANUSCRIPT

62.

Ellis J, Brearley S, Craven O, Molassiotis A. Understanding the symptom experience of

patients with gastrointestinal cancers in the first year following diagnosis: findings from a qualitative longitudinal study. J Gastrointest Cancer 2013;44:60-67. Brunton L, Booker J, Molassiotis A. Making sense of symptoms in men with prostate

RI PT

63.

cancer: a longitudinal qualitative exploration. Int J Urologic Nurs 2012;6:107-114.

Molassiotis A, Lowe M, Ellis J, et al. The experience of cough in patients diagnosed with

lung cancer. Support Care Cancer 2011;19:1997-2004. 65.

SC

64.

Molassiotis A, Lowe M, Blackhall F, Lorigan P. A qualitative exploration of a respiratory

M AN U

distress symptom cluster in lung cancer: cough, breathlessness and fatigue. Lung Cancer 2011;71:94-102. 66.

Fan G, Hadi S, Chow E. Symptom clusters in patients with advanced-stage cancer

referred for palliative radiation therapy in an outpatient setting. Support Cancer Ther

67.

TE D

2007;4:157-162.

Cheung WY, Le LW, Gagliese L, Zimmermann C. Age and gender differences in

symptom intensity and symptom clusters among patients with metastatic cancer. Support Care

68.

Walsh D, Rybicki L. Symptom clustering in advanced cancer. Support Care Cancer

AC C

2006;14:831-836. 69.

EP

Cancer 2011;19:417-423.

Dantzer R, Meagher MW, Cleeland CS. Translational approaches to treatment-induced

symptoms in cancer patients. Nat Rev Clin Oncol 2012;9:414-426. 70.

Wang XS, Shi Q, Williams LA, et al. Serum interleukin-6 predicts the development of

multiple symptoms at nadir of allogeneic hematopoietic stem cell transplantation. Cancer 2008;113:2102-2109.

31

ACCEPTED MANUSCRIPT

71.

Wang XS, Shi Q, Williams LA, et al. Inflammatory cytokines are associated with the

development of symptom burden in patients with NSCLC undergoing concurrent chemoradiation therapy. Brain Behav Immun 2010;24:968-974. Chen E, Nguyen J, Cramarossa G, et al. Symptom clusters in patients with advanced

RI PT

72.

cancer: sub-analysis of patients reporting exclusively non-zero ESAS scores. Palliat Med

AC C

EP

TE D

M AN U

SC

2011;26:826-833.

32

ACCEPTED MANUSCRIPT

Table 1: Summary table of all 33 studies Design

Sample, Cancer site, treatment, setting

Focus of study/ primary aims

Method/analytic technique(s)

Walsh and Rybicki (2006), Symptom clustering in advanced cancer (69)

Design: Crosssectional

N = 922 incurable patients Age range: 12 - 94 - Cancer Site: Heterogeneous - lung (25%), breast (8%), colorectal (8%) - Treatment: palliative care, majority not receiving antitumour treatments - Setting: inpatient and outpatients referred to one palliative care program in the United States

Identify the presence and composition of symptom clusters

Hierarchical cluster analysis. Agglomerative hierarchical method with average linkage.

Aktas, Walsh and Rybicki (2012), USA, Symptom clusters and prognosis in advanced cancer (18)

Design: Longitudinal design, crosssectional symptoms - Symptoms collected at one data point. Survival estimated at 3 time-points: 1, 3 and 6 months following referral. Design: Crosssectional

Same sample as Walsh and Rybicki (2006) N = 831 incurable patients Age range: 12 - 94 - Site: Heterogeneous - lung (25%), breast (8%), colorectal (8%) - Treatment: palliative care, majority not receiving antitumour treatments - Setting: inpatient and outpatients referred to one palliative care program in the United States N = 922 from Walsh and Rybicki (2006) Age range: not specified N = 181 incurable patients - Site: not specified - Treatment: palliative care - Setting: inpatient and outpatients referred to one palliative care program in the United States

Examine whether the seven clusters found in Walsh and Rybick (2006) were related to cancer prognosis. To examine whether symptom clusters may have prognostic value.

33

SC

Cut-off criteria: Only included 25 symptoms with >15% overall prevalence in the cluster analysis. Correlation of >0.68 between pairs defined final clusters.

Uncertain whether some symptoms e.g. weight loss, vomiting, might be more truly described as syndromes. Checklist not a validated symptom instrument. Only measured occurrence.

Quality Rating Score 20

ADSC-I [5]

Post hoc analysis. Checklist not a validated symptom assessment instrument.

20

ADSC-1 [5] in N = 922 sample;

Second dataset was a much smaller sample N = 181. Different symptom checklists used for 2 studies, 2 symptoms (weakness and lack of energy) not assessed in second study. Checklist not a validated symptom instrument. Only prevalence was used to define interrelationships between sentinel and non-sentinel symptoms, but severity or distress may alter cluster composition. N = 181 group characteristics (e.g. Cancer type) not sufficiently described.

17

M AN U

Cox proportional hazards analysis (analyses of survival with proportional hazards models). Survival at 1, 3 and 6 months following referral, estimated by the Kaplan-Meier method

Limitations

TE D

Cut-off criteria: Only included 25 symptoms with >15% overall prevalence in the cluster analysis.

EP

Assess consistency of SCs between 2 independent datasets. Evaluate if the use of a sentinel symptom may abbreviate assessment but retain acceptable accuracy in predicting the presence or absence (prevalence only) of other symptoms.

AC C

Kirkova, Aktas, Walsh, Rybiki and Davis (2010), Consistency of symptom clusters in advanced cancer (37)

Symptom assessment tool(s) ADSC-1 [5]

RI PT

Author and Title

Agglomerative hierarchical cluster analysis on both samples. Cut-off criteria: Both sample only included >15% prevalence, Correlation of >0.68 to define a cluster. Positive predictive value (true positives) and negative predictive value (true negatives) of the sentinel symptom to predict the presence of non-sentinel symptoms in each cluster were calculated separately.

ADSC-2 [6] in N = 181 sample.

ACCEPTED Focus of study/ primary aimsMANUSCRIPT Method/analytic technique(s)

Author and Title

Design


Tsai, Wu and Chiu (2006), Symptom patterns of advanced cancer patients in a palliative care unit (39)

Design: Longitudinal - Symptom severity assessed at 3 timepoints: on admission, one week after admission, two days before death.

N = 77 terminal patients Age range: 16 – 86 - Site: Heterogeneous - lung (23%), liver (16%), colorectal (13%) - Treatment: majority not receiving antitumour treatments - Setting: one palliative care unit in Taiwan

To define longitudinal symptom change patterns of advanced cancer patients.

Tsai, Wu, Chiu and Chen (2010), Significance of symptom clustering in palliative care of advanced cancer patients (30)

Design: Longitudinal design, crosssectional symptoms

N = 427 terminal patients on admission Age range: 27 - 93 (N = 394 with complete data for cluster analysis to examine associations) - Site: Heterogeneous - lung (20%), liver (18%), colorectal (11%) - Treatment: heterogeneous, majority not receiving antitumour treatments - Setting: one palliative care unit in Taiwan N = 268 incurable patients, terminally ill patients were excluded. Age range: 30 – 89 - Site: Heterogeneous breast (21%), lung (20%), head and neck (14%) - Treatment: recently initiated home-based palliative care and outpatient palliative radiation to relieve painful bone metastases. - Setting: five area hospitals in the United States

(1) conduct longitudinal evaluations of symptom management (2) Define symptom patterns (explore symptom clustering) of advanced cancer patients. (3) To investigate the underlying mechanisms that lead to aggregation of symptoms and factors associated with symptom clustering.

34

RI PT

Patients with cognitive limitations were excluded (N = 441). Clusters at a given timepoint were not examined. Clusters were only formed by how symptoms changed longitudinally. Scale had different response options. Could have done the analysis with repeated measures.


ADSRF [7]

‘Pain complex’ cluster was a single item.

14

Curvilinear and moderated regression analyses and simple slope plots

ADSS [8]

19

Cut-off criteria: Not specified.

CES-D [11]

Unclear hypotheses. .60 used to define a cluster.

ESAS only measures 10 items, not adequately comprehensive to assess the range of symptoms patients are experiencing and may result in under-identification of clusters.


Not longitudinal.

Examine age and gender differences in symptom intensity and symptom clusters among outpatients with advanced cancer. 49.8% were male and 50.2%were female. The median age was 64 (range 19 to 99): 39.6% were ≤60 and 60.4% were >60.

Assess the existence of breathlessness, fatigue and anxiety as a SC in advanced lung cancer patients undergoing palliative radiation

AC C

Same sample as Cheung et al 2009 study N = 1366 incurable patients Age range: 18.7 - 74.1 - Site: Heterogeneous gastrointestinal (27%), lung (14%), breast (11%) - Treatment: palliative care planning, pain management, symptom control - Setting: referred to one outpatient palliative radiotherapy clinic in Canada N = 27 incurable lung cancer patients Age range: 47 – 75 - Site: advanced lung cancer - Treatment: Palliative radiation - Setting: oncology outpatient unit of a Hong Kong hospital

(1) characterize the pattern of symptom clusters among advanced cancer outpatients (2) explore the impact of different cancer sites on these clusters

Limitations

RI PT


SC

Cheung, Le, Zimmermann (2009), Symptom clusters in patients with advanced cancers (56)

Symptom assessment tool(s) ESAS [2]

Principal component analysis (PCA) with varimax rotation

ESAS [2]

Inclusion of gender-specific cancers and higher proprtions of lung, gastrointestinal and genitourinary cancers in the older cohort raises the issue that the observed differences may also reflect differences in cancer sites. Not longitudinal. ESAS limitations.

20

VAS [10]

Only assessed 3 symptoms. Sample size way too small. 23 males vs. 4 females - high proportion of male patients. VAS as a measure of anxiety requires further validation.

16

Cut-off criteria: Factor loading >.60 used to define a cluster.

M AN U


TE D

Design

EP

Author and Title

Pearson product-moment correlation and Spearman coefficients. MANOVA test to examine changes of symptom intensity over time. Cut-off criteria: Not specified.


Author and Title

Design


Chaiviboontham et al (2011), Thailand, Symptom clusters in Thais with advanced cancer (50)


N = 240 incurable cancer patients Age range: 18 – 86 - Site: Heterogeneous gastrointestinal (25%), breast (21%), hepato-biliary (15%) - Treatment: not specified, 'not receiving aggressive treatment' - Setting: three tertiary hospitals in Thailand

Explore SCs in Thai patients with advanced cancer

Jimenez et al (2011), Symptom clusters in advanced cancer (31)

Design: Longitudinal design, crosssectional symptoms

N = 406 advanced cancer patients with progressive malignancy. Age range: 18 – 95 - Site: Heterogeneous gastrointestinal (35%), lung (25%), genitourinary (8%) - Treatment: not specified, 'absence of specific therapy that could prolong survival'. - Setting: inpatients and outpatients referred to one palliative care program in Spain.

(1) Identify the presence and composition of SCs in advanced cancer (2) determine patients' characteristics associated with various clusters (3) determine the prognostic value of SCs - ie. Determine their relationship to survival.

N = 654 incurable patients with advanced unresectable disease and cancer cachexia Age range: not specified - Site: Heterogeneous - Lung (31%), upper GI (28%), pancreatic (28%) - Treatment: excluded patients undergoing anticancer therapies or chemotherpay, radiotherapy or surgery four weeks before study entry - Setting: patients recruited as part of two existing clinical trials in patients at multiple sites across different countries (UK based study).

To examine whether pain, depression, and fatigue exist as a symptom cluster in advanced cancer patients with cachexia andmight be related to the presence of systemic inflammation (SI).

42

RI PT

SC

Principal component analysis with varimax rotation. Kaplan-Meier curves for patients according to the number of SCs. Cut-off criteria: not specified.

M AN U

TE D

EP


Cut-off criteria: Factor loading of >.40 used to define a cluster.

AC C

Laird et al (2011), UK, Pain, depression, and fatigue as a symptom cluster in advanced cancer (51)

Principal component analysis with varimax rotation.

Symptom assessment tool(s) Memorial Symptom Assessment Scale (MSAS)

Kruskal-Wallis test was used to examine the relationship between the physical functioning subscale and CRP. Cut-off criteria: A threshold of 50th to 75th %iles was used to represent a cut-off where patients had moderate to severe intensity.

ESAS [2]* *added additional 13 symptoms.

Limitations

Cross sectional.

ESAS only measures 10 items, not adequately comprehensive to assess the range of symptoms patients are experiencing and may result in under-identification of clusters.


20

Although they increased the number of symptoms, this was not a standard checklist. Predominance of males (61% male).

EORTC QLQ-C30 [16]

Analysis not based on severity dichotomised as present or absent. Mixed setting (unclear if it was inpatient, outpatient or both, or from what countries they were recruited from - i.e. didn't give details of sample recruitment characteristics).

19


Author and Title

Design


Shi et al (2010), The investigation of symptoms burden and treatment status in patients with bone metastasis (44)


N = 120 incurable patients with bone metastases Age range: 14 – 86 - Site: Heterogeneous - lung (33%), breast (20%), nasopharyngeal carcinoma (9%) - Treatment: palliative chemoradiation - Setting: six hospitals across China

(1) investigate symptom burden and treatment status in cancer patient with bone metastasis

Spearman’s rank correlation.

Sarna and Brecht (1997), Dimensions of symptom distress in women with advanced lung cancer: a factor analysis (35)


N = 60 women with incurable lung cancer Age range: 33 – 80 - Site: Heterogeneous - bone (23%), central nervous system (17%), other lung (15%). - Treatment: chemotherapy and radiation - Setting: oncology clinics and private offices in the United States

Explore the structure of symptom distress in women with advanced lung cancer

Principal component analysis with varimax rotation.

Bender et al (2005). Symptom clusters in breast cancer across 3 phases of the disease (32)

Design: Longitudinal, but crosssectional sample examined for this review

N = 26 (Total: N = 154) women with incurable breast cancer (Stage IV) with mild anemia Age range: not specified - Treatment: majority receiving radiotherapy - Setting: two breast cancer clinics in the United States.

(1) compare the prevalence of symptoms attributable to breast cancer or its treatment (2) to identify and describe symptom clusters across 3 phases of the disease - studies I, II, III (only study III is relevant to this review).

Cut-off criteria: not specified.

RI PT

*14 items measuring severity, 5 items interference , translated into Chinese.


Method inadequate. Unclear how they measured QoL. Unclear what they treated as predictors or outcomes. Scale not validated. Unclear how they derived the clusters. One table has mean+/unstated number.


Narrow range of symptoms. Despite using the KPS, they didn't look at which 'factors' were associated with the KPS, only examined individual symptoms associated with KPS. Stability of results uncertain due to low N: 60 at the lower end of the commonly used standard number of 5 to 20 subjects per variable for multivariate analyses, the results should be interpreted conservatively.

19

POMS [17]

Measures may not have provided a complete assessment, 4 symptoms excluded In the assessment of women with late-stage breast cancer. Sample size very small for late-stage disease for the analytic techniques used.

17

SC

M AN U

TE D

EP

Hierarchical cluster analysis

Limitations

SDS [1]


AC C 43

Symptom assessment tool(s) MDASI [3]*

ACCEPTED MANUSCRIPT

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

44

Author-developed Symptom Questionnaire (ADSQ) Author-developed breathlessness-fatigueanxiety measure (ADBFA) Brief Pain Inventory (BPI)

Center for Epidemiologic Studies

Severity and interference

38 items. 0-3 point categorical scale (none, mild, moderate, severe) 25 items (weakness and lack of energy not assessed). 0- to 3-point categorical scale (none, mild, moderate, and severe). 19 items. Some scored 0-10 (none to extreme) and 0 - 3 (none to severe). Psychological symptoms of depression, anxiety and aggression on 1-5 (almost none to extreme). 9 items. Measured 'difficulty in controlling symptom' over past month on a five-point category scale (completely, a lot, some, little, none). 17 items. Measured brain metastases-specific symptoms on a 4-point scale (none - severe).

Severity

Occurrence, severity and distress Severity

Severity

Severity

3 items (breathlessness, fatigue, and anxiety) on a visual analogue scale. Horizontal 100 mm line with descriptive phrases indicating no symptom on left, max symptom on right. 7 items, 11-point numerical rating scales. Sensory (pain) and affective components. Functional interference included: general activity, normal work, walking ability, mood, sleep, relations with others, and enjoyment of life. 20 items. Measured depression over the past week on a four-category scale (rarely, some of

[14]

Memorial Delirium Assessment Scale (MDAS)

[15]

Spitzer Quality of Life Index (SQLI)

Difficulty controlling symptom Severity

Occurrence

[16]

[17] [18]

[19]

[20] [21] [22]

Severity of worst pain and functional interference. Occurrence

SC

Memorial Symptom Assessment Scale (MSAS) Author-developed Symptom Checklist-1 (ADSC-1) Author-developed Symptom Checklist-2 (ADSC-2) Author -developed Symptom Reporting Form (ADSRF) Author-developed Symptom Scale (ADSS)

Total: 19 items: 13 items measuring severity, 6 items measuring interference during the last 24 hours. Severity: 0 to 10 numerical rating scale, with 0 being “not present” and 10 being “as bad as you can imagine.” Interference: 0 to10 scale, with 0 being “does not interfere” and 10 being “completely interferes.” 32 items. 0-4 (not at all to very much) 5 point Likert scale over the past week.

[13]

M AN U

[4]

Dimension Distress

TE D

[3]

Edmonton Symptom Assessment System (ESAS) M.D. Anderson Symptom Inventory (MDASI)

Description 13 cancer-specific symptoms. Likert-type format 1-5 (normal or no distress to extensive distress). Indicate how you felt ‘lately’. 9 items. 11-point categorical scale (0-10; absence to worst possible symptom)

EP

[2]

Assessment tool Symptom Distress Scale (SDS)

AC C

[1]

Assessment tool Depression Scale (CESD) Fatigue Symptom Inventory - Disruption Index (FSI-DI)

Description the time, occasionally, most).

Dimension

14-item self-report measure assessing the severity, frequency, and daily pattern of fatigue as well as its perceived interference with quality of life. Disruption Index can be computed by summing items 5 – 11. Severity is measured on separate 11-point scales (0=not at all fatigued; 10=as fatigued as I could be) that assess most, least, and average fatigue in the past week as well as current fatigue. 10-item, 4-point clinician-rated scale (possible range 0–30, from 0-none to 3-severe) designed to quantify the severity of delirium in medically ill patients. 5 domains activity, daily living, health, support, outlook,3-point scale of 0 (worst) to 2 (best). *Only assessed pain, fatigue, and emotional functioning (as a measure of depression)

Severity, interference with QoL

RI PT

Table 2: Measures used

Subscales of the European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-C30 (EORTC QLQ-C30). Profile of Mood States (POMS) Enforced Social Dependency Scale (ESDS) Eastern Cooperative Oncology Group (ECOG) performance status scale Karnofsky Performance Status QualTime (QT) Palliative Performance Scale (PPS)

Severity

Occurrence

65 item self-report measure of mood and affective state. Likert scale 0-5. Activities of Daily Living: measures personal and social competence, 6 point scale. Good validity 6-point scale ranging from “normal” (0) to “dead” (5)

Severity

11-point scale ranging from “normal” (100%) to “dead” (0%), Combination of quality of life, as measures by QLQ-C15-PAL and time (Husain) Performance status (Husain)

N/A

N/A

N/A

N/A N/A

ACCEPTED MANUSCRIPT

Table 3: Most commonly occurring symptom combinations (Q1a – Q1d) occurring at baseline Group

Author, Title

Cluster 1: Anxiety, depression

Cluster 2: Nausea, vomiting

A

Walsh and Rybicki (2006)

Yes + sleep problems

Yes

Yes

Yes

Cluster 3: Nausea, appetite loss

Cluster 4: Fatigue, dyspnea, drowsiness, pain

•

Kirkova, Aktas, Walsh, Rybiki and Davis (2010)

RI PT

Measure: ADSC-1 [5]

A

Measure: ADSS [7]

A

Yes + depression - dyspnea, drowsiness

Thornton et al (2010)

Yes

A

Chow et al (2007)

Yes

Measure: ESAS [2] Chow et al (2008)

Yes

Measure: ESAS [2] A

Fan, Hadi, Chow (2007) Measure: ESAS [2]

45

Yes

EP

Yennurajalingam et al (2013) Measures: ESAS [2]; MDAS [14]

AC C

A

TE D

Measures: [11], [12], [13]

A

M AN U

Yes + dyspnea, abdominal fullness

Tsai, Wu, Chiu and Chen (2010)

SC

Measures: ADSC-1 [5] in N = 922 sample; ADSC-2 [6] in N = 181 sample. A

Other clusters identified:

(3) fatigue, weakness, anorexia, lack of energy, dry mouth, early satiety, weight loss, taste changes • (4) dizzy spells, dyspepsia, belching, bloating • (5) dysphagia, dyspnea, cough, hoarseness • (6) edema, confusion • (7) pain, constipation • (3) anorexia, weight loss, early satiety, taste change • (4) constipation, confusion • (5) hoarseness, dysphagia • (6) dizziness, dyspepsia • (7) belching, bloating • (8) dyspnea, cough • (2) fatigue and weakness • (3) anorexia, taste alteration, dysphagia, constipation, dry mouth/thirst • (4) restlessness/heat, dizziness, insomnia, night sweats • (5) pain only No other clusters

Yes (see Cluster 4)

Yes + appetite, nausea

No other clusters

Yes + dyspnea

Yes + well-being - dyspnea

No other clusters

Yes + well-being

Yes

No other clusters

Yes + well-being, pain

Yes - pain

No other clusters

Group

Author, Title

A

Hird, Chow et al (2010)

ACCEPTEDCluster MANUSCRIPT Cluster 2: 3: Cluster 4: Nausea, vomiting Nausea, Fatigue, dyspnea, appetite loss drowsiness, pain Yes + dizziness, headache


Measure: BPI [11] Cheung, Le, Zimmermann (2009)

Yes

Yes (See Cluster 4)

Measure: ESAS [2] A

Chan, Richardson (2005) Measure: VAS [10] Yes + dry mouth, numbness/tingling, problem of urination, diarrhoea, feeling sad, feeling irritable

Chaiviboontham et al (2011) Measure: MSAS [4]

A

Jimenez et al (2011),

Yes + insomnia

Yes

Bender et al (2005) Measure: POMS [17]

46

AC C

Sarna and Brecht (1997) Measure: SDS [1]

A

Yes + appetite loss

Shi et al (2010) Measure: MDASI [3]**

A

EP

Measure: ESAS [2]* A

TE D

A

Yes + fatigue, lack of energy, decreased physical strength, loss of concentration

SC

A

Hadi, Fan, Hird Chow, et al (2008)

Yes (see Cluster 2)

Yes

•

(2) trouble concentrating, decreased alertness, confusion, imbalance problems, memory loss, weakness, fatigue, vision problems, and problems with smell, hearing, or tingling. • (3) seizures, numbness, speech difficulty, and personality change. Note: 2 clusters were also found with SQLI[15] at baseline: • (1) activity, daily living, health • (2) support, outlook • (1) walking ability, general activity, normal work, enjoyment of life and worst pain. • (2) relations with others, mood and sleep.

Yes + nausea, appetite - pain

No other clusters

Yes + anxiety - pain, drowsiness

No other clusters

Yes + concentration, lack of energy, nervous, sleeping difficulty, bloated, worrying, lack of appetite, weight loss, appearance - drowsiness, fatigue

•

M AN U

A

RI PT

Measures: ADSQ [9]; SQLI [15]


•

•

• Yes + constipation - dyspnea, drowsiness

(3) cough, dizziness, difficulty swallowing, mouth sores, changes in food taste, hair loss, constipation (4) feeling drowsy, sweating, having sexual problem, itching, swelling, arms/leg, changes in skin (3) cognitive impairment, agitation, urinary incontinence (4) anorexia, weight loss, and tiredness

No other clusters

•

(2) pain frequency, pain severity, bowel, appearance, outlook • (3) cough, breathing, insomnia; • (4) fatigue, concentration. No other clusters

Author, Title

ACCEPTEDCluster MANUSCRIPT Cluster 2: 3: Cluster 4: Nausea, vomiting Nausea, Fatigue, dyspnea, appetite loss drowsiness, pain PCA: No PCA: Yes PCA: Yes + dyspnea + well-being - dyspnea

Cluster 1: Anxiety, depression PCA: Yes

A

EFA: Yes

EFA: No

EFA: Yes (see Cluster 4)

HCA: Yes

HCA: No

PCA: Yes

PCA: No

PCA: Yes + well-being

EFA: Yes + appetite, well-being

EFA: No

EFA: No

HCA: Yes + dyspnea

Khan, Chow et al (2012)

EFA: Yes + nausea, appetite, wellbeing

RI PT

Group

A

PCA: No

EFA: No

EFA: No

HCA: No

HCA: No

HCA: No

Chen, Khan, Chow et al (2012a) Measure: BPI [11]

Chen, Chow et al (2012b)

A

Measure: ESAS [2] Chen, Chow et al (2011), Measure: ESAS [2]

HCA: Yes

PCA: Yes + nausea, appetite loss, well-being

EFA: Yes HCA: Yes

47

EP

EFA: Yes

PCA: No

EFA: No

AC C

PCA: Yes

A

HCA: No

PCA: No

EFA: No

HCA: No other clusters

PCA: No other clusters

SC

EFA: No other clusters

EFA: Yes + nausea

HCA: Yes

HCA: Yes (see Cluster 1)

PCA: No

PCA: No

EFA: No

EFA: No

HCA: No

HCA: No

PCA: Yes + pain, wellbeing

PCA: Yes - pain


EFA: Yes + appetite, nausea, wellbeing

HCA: Yes + pain, wellbeing

EFA: No other clusters

PCA: Yes

M AN U

Measure: ESAS [2]

HCA: Yes + fatigue, drowsiness, dyspnea PCA: No

TE D

Chen, Khan, Chow et al (2013)


HCA: Yes + well-being - dyspnea

Measure: ESAS [2] A


HCA: (3): pain, sense of well-being

PCA: (1) Worst pain, general activity, walking ability, normal work, enjoyment of life (2) mood, relations with others, sleep

EFA: (1) Worst pain, general activity, walking ability, normal work, enjoyment of life HCA: (1) Worst pain, general activity, walking ability, normal work, enjoyment of life (2) mood, relations with others PCA: No other clusters EFA: No other clusters HCA: No other clusters

HCA: Yes - pain

PCA: Yes + pain, wellbeing

PCA: Yes - pain


EFA: Yes + appetite, nausea, wellbeing


EFA: No other clusters HCA: No other clusters

Cluster 1: Anxiety, depression + nausea, appetite loss, well-being

PCA: No

A

EFA: No

ACCEPTEDCluster MANUSCRIPT Cluster 2: 3: Cluster 4: Nausea, vomiting Nausea, Fatigue, dyspnea, appetite loss drowsiness, pain HCA: Yes HCA: No + pain, wellHCA: Yes being - pain

PCA: Yes + headache, dizziness

EFA: Yes + headache, speech difficulty

PCA: No

PCA: No

EFA: No

EFA: No

HCA: Yes + headache, dizziness HCA: No

HCA: No

Khan, Chen, Chow et al (2013) Measures: ADSQ [9]; SQLI [15] Old (>60y.o): Yes

A

48

AC C

EP

TE D

M AN U

HCA: No

RI PT

Author, Title

SC

Group

Old (>60y.o): No

Cheung, Le, Gagliese, Zimmermann (2011),

Young: Yes + well-being

Young: No

Measure: ESAS [2]

Male: Yes

Male: No

Old (>60y.o): Yes + pain

Young: Yes (see Cluster 4) Male: No

Old (>60y.o): No


PCA: • (2) weakness, memory loss, confusion, trouble concentrating, decreased alertness, imbalance problems, vision problems, problems with smell/hearing/tingling, fatigue • (3) seizures, speech difficulty, numbness, personality change EFA: • (2) memory loss, confusion, trouble concentrating, decreased alertness, imbalance problems, problems with smell/hearing/tingling, fatigue , dizziness, personality changes • (3) weakness, seizures, vision problems, numbness HCA: • (2) weakness, memory loss, confusion, trouble concentrating, decreased alertness, imbalance problems, problems with smell/hearing/tingling, fatigue • (3) seizures, speech difficulty, problems with smell/hearing/tingling, numbness, personality change Note: clusters were also examined using SQLI[15] at baseline: PCA: (1) activity, daily living, health (2) outlook, support EFA: No clusters found HCA: (1) activity, daily living, health (2) outlook, support Old: No other clusters Young: No other clusters Male: No other clusters

Young: Yes + appetite, nausea - dysponea

Female: No other clusters

Author, Title

Cluster 1: Anxiety, depression + pain

Female: Yes + well-being

B

ACCEPTEDCluster MANUSCRIPT Cluster 2: 3: Cluster 4: Nausea, vomiting Nausea, Fatigue, dyspnea, appetite loss drowsiness, pain Male: Yes + appetite Female: Yes - pain, drowsiness Female: No (see Cluster 4) Female: Yes + appetite, nausea - dyspnea

Fodeh et al (2013)

RI PT

Group

B

M AN U

SC

Measures: SDS [1]; ESDS [18]

Husain et al (2011)

Measure: QLQ-C30 [16] Tsai, Wu and Chiu (2006) Measure: ADSRF [7]

49

EP

C

Laird et al (2011)

AC C

B

TE D

Measures: ESAS [2]

Yes +depression -dyspnea, drowsiness


Note: SCs were organised according to cancer site: • (1) GI: outlook, insomnia, appearance, concentration, eating/feeding • (2) GI: appetite, bowel, insomnia, eating/feeding, appearance • (3) Gynaecological: nausea, insomnia, eating/feeding, concentration, pain • (4) Head and neck: dressing, eating/feeding, bathing, toileting, walking • (5) Lung: cough, walking, eating/feeding, breathing, insomnia Note: SCs were organised around cancer functional impairments: • (1) HIGH tired, MODERATE drowsy, appetite, well-being, pain, depression anxiety • (2) HIGH tired, MODERATE drowsy, appetite, well-being, LOW pain depression, anxiety • (3) LOW tired, drowsy, appetite, well-being, pain, depression, anxiety No other clusters

Note: SCs were organised around six different longitudinal trajectories in symptom intensity: • (1) continuous static: restless/heat, abdominal fullness, constipation, dizziness, insomnia • (2) static-increase: fatigue, weakness, nausea/vomiting, taste alteration, dysphagia, diarrhea, dry mouth, night sweats • (3) decrease-static: pain, depression • (4) decrease-increase: anorexia, dyspnea • (5) static-decrease: aggression • (6) gradually decrease: anxiety.

Group

Author, Title

C

Wang et al (2006)


ACCEPTEDCluster MANUSCRIPT Cluster 2: 3: Cluster 4: Nausea, vomiting Nausea, Fatigue, dyspnea, appetite loss drowsiness, pain Yes

Francoeur (2005)

SC

D

RI PT

Measure: MDASI [3]* **

D

M AN U

Measure: ADSS [8]; CES-D [11]

Yes +anxiety, shortness of breath, drowsiness, pain

Hayduk et al (2010)

D

Olson, Hayduk et al (2008)

TE D

Measure: ESAS [2]

Yes

AC C

EP

Measure: ESAS [2]


Note: Clusters were organised around different development patterns during and after treatment: • (1) Nausea-vomiting cluster grouped as: ‘early/mid-therapy increases’ • (2) Steady increase during therapy: sore throat, pain • (3) Early/late-therapy increases: lack of appetite, distress, drowsiness, sleep disturbance, dry mouth • (4) Early decrease/no change: affective symptoms (sadness), cognitive issues, difficulty remembering, shortness of breath, cough, numbness. Note: Clusters were organised around cross-over effects on depressive affect: (1) pain, fever, fatigue, sleep (2) pain, appetite, weight loss (3) nausea, fever, fatigue (4) breathlessness, fatigue, sleep (5) breathlessness, appetite, weight loss, sleep Note: Clusters were organised around stability in a causal model: • (1) Nausea-appetite loss cluster grouped as: ‘Moderately stable symptoms’ • (2) Labile/stable symptoms: tiredness, depression, well-being. Note: Clusters were organised around stability in a causal model: • (2) appetite/well-being • (3) drowsiness/tiredness The authors suggested these symptoms do not constitute symptom clusters, but are pairs of symptoms in which changes in the first symptom consistently lead to changes in the second symptom at two points in time.

Notes: ‘Shortness of breath’ was labelled as ‘dyspnea’ in all cases. ‘Lack of energy’ was not labelled as ‘fatigue’ in all cases. Symptoms in the identified SCs included functional impairments/interference and/or activities of daily living as items. * Additional 13 items: drowsiness, sweating, fever, weight loss, cognitive impairment, agitation, constipation, insomnia, dysphagia, hiccups, vomiting, urinary or stool incontinence ** 14 items measuring severity, 5 items interference, translated into Chinese. *** Two additional symptoms (cough and sore throat) were included from the MDASI lung module.

50

ACCEPTED MANUSCRIPT

Table 4: Studies assessing longitudinal and methodological stability (Question 2 and 3) Analytic technique(s)

Timeframe of longitudinal assessment

Major finding(s) regarding longitudinal stability and/or or stability across methodology

Hayduk et al (2010) Olson, Hayduk et al (2008).

Structural equation modelling. Structural equation modelling

Severity assessed at two timepoints: one month before death and one week before death. Severity assessed at two timepoints: one month before death and one week before death.

Mixed longitudinal stability. Symptoms that displayed no significant stability: tiredness, depression, well-being. Moderately stable symptoms: pain, anxiety, nausea, shortness of breath, drowsiness, appetite. Symptom relationships that remained stable did so in the context of increasing intensities of all symptoms as death approached. Mixed longitudinal stability. Stable associations at both time-points: the effects of anxiety on depression, drowsiness on tiredness, appetite on well-being. Unstable associations: the effects of pain on appetite, anxiety on well-being, and depression on well-being. It is unknown whether these associations will hold over a longer period of time. Causal foundations in symptom patterns changed over time.

Yennurajaling am et al (2013)

Wilcoxon signedrank test

Severity assessed at 2 timepoints baseline (initial) and first follow-up visit (14 days median, range 14weeks).

2 clusters stable longitudinally: (1) fatigue, pain, nausea, drowsiness, dyspnea, loss of appetite; (2) anxiety, depression. Stable/constant between baseline and follow-up, but severity lessened during the interval. Physical and psychological symptom clusters were differentiated at initial visits and consistent across time at follow-up, but severity decreased longitudinally, suggesting those who experience severe symptoms can attain significant improvement by the time of the first follow-up visit after an initial palliative care consultation.

Chow et al (2007)

Spearman correlation, Chisquare test.

Severity assessed at 5 timepoints post-radiation treatment: weeks 1, 2, 4, 8, 12

Unstable longitudinally. Clusters changed over time for both responders and non-responders. Symptom clusters also changed over time post-radiation, suggesting the effect of radiation on some symptoms, especially pain, may influence other symptoms within the same cluster or influence other clusters. Pain did not remain stable with any cluster in both responders and non-responders - alleviation of pain in influenced experience of other symptoms.

Chow et al (2008)

Wilcoxon nonparametric test

Severity assessed at 5 timepoints post-radiation treatment: weeks 1, 2, 4, 8, 12

Unstable longitudinally. Three clusters identified at baseline were inconsistent over time – they changed each week. Nonetheless, a strong cluster presence still appeared after whole brain radiation at 12 weeks. These symptoms consistently clustered together: (1) fatigue and drowsiness (2) anxiety and depression. Some symptoms showed an overall increase in severity over time, whereas others were experienced by a greater proportion at week 12.

Hadi, Fan, Hird Chow, et al (2008)

Chi-square test

Chen, Chow et al (2011)

PCA, HCA, EFA

Symptom severity measured at 4 timepoints: before radiation (baseline) then at 1, 2, 3 months post-RT. Symptom severity measured at 4 timepoints: before radiation (baseline) then at 1, 2, 3 months post-radiation therapy.

Unstable longitudinally. Two clusters disappeared in the responders over time, but displayed an irregular pattern in non-responders over time. In responders to RT, the two symptom clusters disintegrated at 4, 8, and 12 weeks post-RT. All symptom severity items improved over time. In non-responders, two clusters had disappeared at week 4, reemerged at week 8, and disintegrated at week 12. Unstable longitudinally. Symptom consistency was poor. Analysis was performed on the ‘non-zero’ subgroup and the total patient sample, at each timepoint. Significant variations existed in SC number and composition extracted using the 3 different methods at each time point for both groups.

Chan, Richardson (2005)

MANOVA test to examine changes of symptom intensity over time.

51

SC

M AN U

TE D

EP

AC C

Symptom severity assessed at 3 timepoints: baseline (1 day prior to palliative radiotherapy), week 3, week 6.

Papers which assessed different methodology:

RI PT

Author

Stable longitudinally. The lack of interaction between time and symptom showed that there was no evidence that patterns of change across time among breathlessness, fatigue, and anxiety were different. Overall, symptoms at 6 weeks were higher than week 3, whereas week 3 was higher than baseline. At baseline the median intensity of symptoms was mild, becoming progressively worse at T1 and T2. The correlations between the 3 symptoms were moderately strong at T1 and T2.

Khan, Chow et al (2012)

PCA, HCA, EFA at each timepoint. ESAS [2]

Symptom severity assessed at 5 timepoints post-radiation treatment: weeks 1, 2, 4, 8, 12

Chen, Khan, Chow et al (2013)

PCA, HCA, EFA at each timepoint. ESAS [2]

Symptom severity assessed at 5 timepoints post-radiation treatment: weeks 1, 2, 4, 8, 12

Chen, Khan, Chow et al (2012a)

PCA, HCA, EFA at each timepoint. BPI [10]

Symptom severity assessed at baseline, 1, 2, and 3 months following radiation treatment.

ESAS [2]

52

Unstable across methodology. The quantity, composition, occurrence of symptom clusters varied based on which statistical method was employed, indicating a common analytical method is necessary for consistency and comparison for future research. Cluster findings using PCA and HCA correlated more strongly with each other than either did with the findings of EFA. Mixed longitudinal stability. Varying patterns of symptom cluster presentation over time were observed in the responders versus non-responders subgroups regardless of the analytical method employed. Only one HCA derived core cluster of symptoms composed of worst pain, general activity, walking ability, normal work, and enjoyment of life frequently constituted a ‘stable’ cluster.

Symptom severity measured at 4 timepoints: before radiation (baseline) then at 1, 2, 3 months post-RT.

Unstable across methodology. Absolute consensus among all three statistical methods was never reached at any assessment time point. Little correlation was observed in the symptom cluster findings of PCA, EFA, and HCA in the total patient sample. The presence and composition of symptom clusters derived varied depending on which statistical analysis method was employed. Unstable longitudinally. No strong similarities were consistently observed for 17 symptoms. Lack of complete consensus amongst clusters at all timepoints with 3 methods. Although entire clusters varied over time, some domains consistently clustered together: (1) activity and daily living from SQLI items (2) memory loss, confusion and trouble concentrating (3) nausea-vomiting from 17 additional symptoms.

Cross-sectional

Unstable across methodology. Lack of complete consensus at all timepoints with 3 methods. HCA revealed similar clusters to PCA at baseline for SQLI items. EFA did not derive any clusters from SQLI items at baseline. N/A for longitudinal stability.

TE D

PCA, HCA and EFA

Unstable across methodology. A complete consensus between HCA, EFA, PCA for the number and composition of SCs was not reached at any time point. Little correlation in clusters was found between the 3 statistical methods despite use of an identical data set. Different SCs were observed in responders and non-responders with all 3 methods and clusters varied at each time point within each subgroup. Unstable longitudinally. Inconsistency in SC composition was observed at different time intervals.

EP

Chen, Chow et al (2012b)

PCA, HCA and EFA at each timepoint. ADSQ [9]; SQLI [15]

Unstable longitudinally. Cluster composition changed each week and by different methods.

AC C

Khan, Chen, Chow et al (2013)

ACCEPTED MANUSCRIPT Major finding(s) regarding longitudinal stability and/or or stability across methodology

RI PT

Timeframe of longitudinal assessment

SC

Analytic technique(s)

M AN U

Author

Unstable across methodology. Complete consensus in all three statistical methods was never reached at any assessment time point. Increasingly diverging patterns of symptom cluster development over time were observed in the responder vs. non-responder subgroups. Symptom pairs comprising anxiety and depression or fatigue and drowsiness consistently presented in the same cluster despite the shifting of other symptoms in the cluster over time. The symptom cluster findings of HCA and PCA correlated more frequently with each other than either did with the results of EFA.

ACCEPTED MANUSCRIPT

Table 5: Studies with associated predictors and outcomes (Question 4 and 5) Predictors of SCs

Outcomes of SCs

Major finding(s)

Fodeh et al (2013)

Cancer site because they were organising clusters around the cancer site Yes - sympathetic nervous system SNS and HPA axis hormones (neuroendocrine levels) Plasma catecholamines, Plasma adrenocorticotropic hormone, plasma cortisol. Yes - cancer sites

Not evaluated

5 clusters identified, organised around cancer site, including both symptoms and functional impairments. This suggests that functional impairments behave and act as symptoms during the diagnostic phase of newly diagnosed late-stage cancer patients. Elevated hormones of the sympathetic nervous system (SNS) indicated by plasma epinephrine and norepinephrine and hormones of the hypothalamic-pituitary-adrenal (HPA) axis indicated by plasma levels of cortisol and adrenocorticotropic hormone. Elevated neuroendocrine levels (SNS and HPA axis hormones) predicted (explained significant shared variance between) pain, depression and fatigue, while controlling for disease and demographic variables. The presence of distant metastases, functional status and absence of partner were also positively associated with PDF symptoms. 2 clusters found. Primary cancer site influences cluster composition, such that (1) pain and drowsiness clustered for cancers involving the central nervous system and head&neck (2) anxiety and depression clustered for solid tumors (3) decreased appetite and poor well-being clustered for rbeast, lung, gastrointestinal, genitourinary and gynecological malignancies.

Cheung, Le, Gagliese, Zimmermann (2011) Jimenez et al (2011)

Yes - age and gender

Not evaluated

In patients with advanced cancers, symptom patterns differ according to age and gender. Younger patients reported worse pain (4.9 vs. 4.5, p=0.02) and better appetite (4.7 vs. 5.3, p=0.002) than older patients. Females reported poorer scores than males for nausea (2.6 vs. 2.2, p=0.02).

Patient characteristics: primary cancer site, gender, age, and performance status (ECOG [19])

Survival (prognosis)

The presence of the 4 SCs 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, 52 days; one SC, 38 days; two SCs, 23 days; and three to four SCs, 19 days; P

Symptom clusters and survival in Portuguese patients with advanced cancer.

Usefulness of Positron Emission Tomography in Patients with Syphilis: A Systematic Review of Observational Studies.

Statin use and mortality in cancer patients: Systematic review and meta-analysis of observational studies.

Studies of the symptom dyspnoea: a systematic review.

Identification of symptom clusters among patients with heart failure: an international observational study.

Systematic review of observational studies assessing bleeding risk in patients with atrial fibrillation not using anticoagulants.

Symptom clusters and quality of life among patients with advanced heart failure.

Anorexia nervosa and cancer: a protocol for a systematic review and meta-analysis of observational studies.

How can we help family carers manage pain medicines for patients with advanced cancer? A systematic review of intervention studies.

A systematic review: non-pharmacological interventions in treating pain in patients with advanced cancer.

Symptom Clusters in Ovarian Cancer Patients With Chemotherapy After Surgery: A Longitudinal Survey.

Symptom Clusters in Advanced Cancer Patients: An Empirical Comparison of Statistical Methods and the Impact on Quality of Life.

Prognosis of adenosquamous carcinoma compared with adenocarcinoma in uterine cervical cancer: a systematic review and meta-analysis of observational studies.

Methotrexate and liver fibrosis in people with psoriasis: a systematic review of observational studies.

Risk of coronary artery disease in patients with idiopathic inflammatory myopathies: a systematic review and meta-analysis of observational studies.

Cardiovascular disease risk factors in patients with hidradenitis suppurativa: a systematic review and meta-analysis of observational studies.

palliative care setting.

Body Mass Index and Risk of Gallbladder Cancer: Systematic Review and Meta-Analysis of Observational Studies.

Hormonal contraceptive choice for women with PCOS: a systematic review of randomized trials and observational studies.

Bleeding Risk with Long-Term Low-Dose Aspirin: A Systematic Review of Observational Studies.

Conceptualizing prognostic awareness in advanced cancer: a systematic review.

Influences of cancer symptom knowledge, beliefs and barriers on cancer symptom presentation in relation to socioeconomic deprivation: a systematic review.

Symptom Control Trials in Patients With Advanced Cancer: A Qualitative Study.

Communication in healthcare interactions in dementia: a systematic review of observational studies.