Alternative analytical methods for the identification of cancer related symptom clusters

KEY WORDS Symptom clusters; symptoms; cancer; symptom experience; symptom management strategies; literature review; empirical methods; multivariate methods; exploratory factor analysis;

Queensland University of Technology

School of Nursing and Midwifery

Faculty of Health Institute of Health and Biomedical Innovation

Alternative Analytical Methods for the Identification of

Cancer-Related Symptom Clusters

Helen Mary Skerman DipTeach, BSc, GradDipCompEd, MSocSc (App)

This thesis is submitted

to fulfil the requirements for the Award of Doctor of Philosophy

MAY 2010

KEY WORDS

Symptom clusters; symptoms; cancer; symptom experience; symptom management strategies; literature review; empirical methods; multivariate methods; exploratory factor analysis; common factor analysis; cluster analysis; principal axis factoring; stability; longitudinal analysis; chemotherapy; outpatients; nursing research; oncology; Theory of Unpleasant Symptoms

ABSTRACT

Advances in symptom management strategies through a better understanding

of cancer symptom clusters depend on the identification of symptom clusters that are valid and reliable The purpose of this exploratory research was to investigate alternative analytical approaches to identify symptom clusters for patients with cancer, using readily accessible statistical methods, and to justify which methods of identification may be appropriate for this context Three studies were undertaken: (1)

a systematic review of the literature, to identify analytical methods commonly used for symptom cluster identification for cancer patients; (2) a secondary data analysis

to identify symptom clusters and compare alternative methods, as a guide to best

practice approaches in cross-sectional studies; and (3) a secondary data analysis to

investigate the stability of symptom clusters over time

The systematic literature review identified, in 10 years prior to March 2007,

13 cross-sectional studies implementing multivariate methods to identify cancer related symptom clusters The methods commonly used to group symptoms were exploratory factor analysis, hierarchical cluster analysis and principal components analysis Common factor analysis methods were recommended as the best practice cross-sectional methods for cancer symptom cluster identification

A comparison of alternative common factor analysis methods was conducted,

in a secondary analysis of a sample of 219 ambulatory cancer patients with mixed diagnoses, assessed within one month of commencing chemotherapy treatment Principal axis factoring, unweighted least squares and image factor analysis

identified five consistent symptom clusters, based on patient self-reported distress ratings of 42 physical symptoms Extraction of an additional cluster was necessary when using alpha factor analysis to determine clinically relevant symptom clusters The recommended approaches for symptom cluster identification using non-multivariate normal data were: principal axis factoring or unweighted least squares for factor extraction, followed by oblique rotation; and use of the scree plot and Minimum Average Partial procedure to determine the number of factors

In contrast to other studies which typically interpret pattern coefficients alone, in these studies symptom clusters were determined on the basis of structure coefficients This approach was adopted for the stability of the results as structure coefficients are correlations between factors and symptoms unaffected by the correlations between factors Symptoms could be associated with multiple clusters as

a foundation for investigating potential interventions

The stability of these five symptom clusters was investigated in separate common factor analyses, 6 and 12 months after chemotherapy commenced Five qualitatively consistent symptom clusters were identified over time

(Musculoskeletal-discomforts/lethargy, Oral-discomforts, discomforts, Vasomotor-symptoms, Gastrointestinal-toxicities), but at 12 months two

Gastrointestinal-additional clusters were determined (Lethargy and Gastrointestinal/digestive

symptoms) Future studies should include physical, psychological, and cognitive

symptoms Further investigation of the identified symptom clusters is required for validation, to examine causality, and potentially to suggest interventions for symptom management Future studies should use longitudinal analyses to investigate

change in symptom clusters, the influence of patient related factors, and the impact

on outcomes (e.g., daily functioning) over time

TABLE OF CONTENTS

Keywords i

Abstract ii

Table of Contents v

List of Tables x

List of Figures xi

Declaration of Authorship xii Glossary of Acronyms and Terms xiii

Publications from the Research Program xv

Statement of Contribution of Co-authors xvi

Funding for the Research Program xvii

Acknowledgements xviii

Chapter 1: Introduction

1.1 Introduction 1

1.2 The Burden of Cancer 2

1.3 Rationale and Significance of the Research 3

1.4 Research Purpose and Objectives 7

1.5 Research Questions 8

1.6 Thesis Outline 9

Chapter 2: Background

2.2.2 Change in the Symptom Experience Over Time 15

2.4 The Clinical Relevance of Symptom Clusters 19 2.5 The Symptom Experience and Symptom Management Models 21

2.5.1 The Theory of Unpleasant Symptoms (TOUS) 22

Chapter 3: Symptom Cluster Identification

3.3 Approaches to Symptom Cluster Identification 31

3.3.2 The Empirical Identification of Symptom Clusters 32

Chapter 4: A Systematic Literature Review

4.1 Introduction 41

4.2 Method 42

4.3 Multivariate Methods to Identify Cancer-Related Symptom Clusters 43

4.4 Summary 83

Chapter 5: Methods 5.1 Introduction 85 5.2 The Conceptual Framework 86 5.3 Study Design 90 5.3.1 The Parent Study 90

5.4 Measures 94

5.4.1 Measures in Parent Study - Ambulatory Care Project 94 5.4.2 Selected Measures in Current Study 96

5.5 Statistical Analysis 99

5.5.1 Data Quality 99

5.5.2 Missing Data 100

5.5.3 Complete Data for Exploratory Factor Analysis 102

5.5.4 Study 2: Identification of Symptom Clusters 102

5.5.5 Number of Factors to Retain 104

5.5.6 Simple Structure and Symptom Cluster Identification 106

5.5.7 Alternative Methods of Common Factor Extraction 107

5.5.8 Study 3: Identification of Symptom Clusters over Time 110

5.6 Sample size 112

5.7 Limitations 112

Chapter 6: The Empirical Identification of Symptom Clusters

6.1 Introduction 115

6.2 Identification of Cancer-Related Symptom Clusters: An Empirical 116

Comparison of Exploratory Factor Analysis Methods

6.3 Summary 145

Chapter 7: The Empirical Identification of Symptom Clusters over Time

7.1 Introduction 147

7.2 Cancer-Related Symptom Clusters, 6 and 12 Months after Commencing Chemotherapy: An Empirical Investigation 148

Chapter 8: Final Discussion and Conclusions 177

8.1 Key Findings 178

8.1.1 Multivariate Methods for Symptom Cluster Identification 180

8.1.2 EFA Decisions for Symptom Cluster Identification 181

8.1.3 Stability of Symptom Clusters Identified at Different Times 185

8.2 Strengths and Limitations 187

8.3 Implications of the Findings 195

8.3.1 Conceptual Implications 195

8.3.2 Analytical Implications 199

8.3.3 Implications for Clinical Practice 203

8.3.4 Implications for Future Research 204

8.4 Conclusion 206

Appendices

Appendix 1: Systematic Literature Review Summary Sheets 209

Appendix 2: Modified Rotterdam Symptom Checklist 213

Appendix 3: Pattern Matrix, Principal Axis Factoring at T1 217

Appendix 4: Structure Matrix, Principal Axis Factoring, at T2 219

Appendix 5: Structure Matrix, Principal Axis Factoring, at T3 221

References 225

LIST OF TABLES

Table 4.1 Symptom clusters identified by common factor analysis methods 70

Table 4.2 Symptom clusters identified by principal components analysis 73

Table 4.3 Symptom clusters identified by hierarchical cluster analysis 75 Table 5.1 Distribution of missing data at each time point 101

Table 6.1 Patients’ clinical characteristics 134

Table 6.2 Symptom Clusters identified by Principal Axis Factoring 135

Table 6.3 Symptom Clusters by alternative extraction methods 136

Table 7.1 Patients’ clinical characteristics 156

Table 7.2 Prevalence of patients’ symptom distress over time 158

Table 7.3 Vasomotor-symptoms and Oral-discomforts symptom clusters 160

identified within 1, at 6, and 12 months after commencing chemotherapy

Table 7.4 Gastrointestinal-related symptom clusters identified within 1, 161

at 6, and 12 months after commencing chemotherapy

Table 7.5 Musculoskeletal discomforts/lethargy symptom clusters identified 162

within 1, at 6, and 12 months after commencing chemotherapy

Table 7.6 Symptom clusters identified ONLY at 12 months after commencing 163 chemotherapy

Table A3.1 Factors identified within one month of commencing chemotherapy 218

Table A4.1 Factors identified 6 months after commencing chemotherapy 220

Table A5.1 Factors identified 12 months after commencing chemotherapy 222

LIST OF FIGURES

Figure 5.1 Conceptual Framework of the Study Adapted from the Theory of

Unpleasant Symptoms 89 Figure 5.2 Flow of Participants in Parent Study 93

QUEENSLAND UNIVERSITY OF TECHNOLOGY

DECLARATION OF AUTHORSHIP

The work contained in this thesis has not been previously submitted to meet requirements for an award at this or any other higher education institution To the best of my knowledge and belief, the thesis contains no material previously published or written by another person except where due reference is made

GLOSSARY OF ACRONYMS AND TERMS

AFA Alpha Factor Analysis

CARES-SF Cancer Rehabilitation Evaluation System, Short Form

CFA/FA Common Factor Analysis

ECOG European Cooperative Oncology Group

EFA Exploratory Factor Analysis

GEE Generalized Estimating Equations

GLS Generalized Least Squares

HCA Hierarchical Cluster Analysis

LSS Life Satisfaction Scale

MAP Minimum Average Partial

MCAR Missing Completely At Random

MDASI M.D Anderson Symptom Inventory

ML, MLFA Maximum Likelihood Factor Analysis

MSAS Memorial Symptom Assessment Scale

NIH National Institutes of Health

PAF Principal Axis Factoring

PCA Principal Components Analysis

QUT Queensland University of Technology

RMSEA Root Mean Square Error of Approximation

RSCL Rotterdam Symptom Checklist

SEM Structural Equation Modeling

SMC Squared Multiple Correlation

SSQT Social Support Questionnaire for Transactions

TOUS Theory of Unpleasant Symptoms

UCSF University of California, San Francisco

ULS Unweighted Least Squares

Clinically relevant Important to the patient’s experience and has some practical

consequence for symptom management and patient outcomes Communality Common variance of a variable, shared with other variables Multivariable Relationships between independent and dependent variables Multivariate Relationships among multiple dependent variables

Stability Consistent/replicated for a group at a point in time or over

time, or in individuals over time, and for different patient populations (subgroups)

PUBLICATIONS FROM THE RESEARCH PROGRAM

Skerman, H M., Yates, P M., & Battistutta, D (2009) Multivariate methods to

identify cancer-related symptom clusters Research in Nursing & Health,

32(3), 345-360

(This manuscript is presented in Chapter 4)

Skerman, H., Yates, P., & Battistutta, D (2007) A path analysis modeling the

symptom experience of cancer patients commencing adjuvant treatment in

ambulatory clinics Oncol Nurs Forum, 34(1), 214

Skerman, H M., Yates, P M., & Battistutta, D (2009) Identification of Cancer-

Related Symptom Clusters: An Empirical Comparison of Common Factor

Analysis Methods J Pain Symptom Manage (Under review)

(The submitted manuscript is presented in Chapter 6)

Skerman, H M., Yates, P M., & Battistutta, D (2009) Stability of Cancer-Related

Symptom Clusters, 6 and 12 Months after Commencing Chemotherapy

Support Care Cancer (Under review)

(The submitted manuscript is presented in Chapter 7)

FUNDING ATTRACTED BY RESEARCH PROGRAM

Queensland University of Technology Postgraduate Research Award

(APRA) Funding received from 2005 to 2008

Institute of Health and Biomedical Innovation

Funding received for a presentation at the ANZ Society of Palliative Medicine 2008 conference

I wish to express my sincere thanks to my Chief Supervisor, Professor Patsy Yates, for her belief in this topic, and her support, encouragement and patience during the course of this thesis I wish to acknowledge my Associate Supervisor, Associate Professor Diana Battistutta, for her willingness to discuss all aspects of this project, and in particular, her mentoring on the statistical perspectives developed

In the early years of this project, Dr Cameron Hurst provided informative statistical discussions on factor and cluster analysis I would like to express my thanks to Dr Anne Walsh for her friendship and support in the early years I would also like to acknowledge the continued support and encouragement of my research colleagues in the School of Nursing and IHBI, particularly the Palliative Care Group

I would like to acknowledge the financial support for this research, received from the Queensland University of Technology, the School of Nursing, and the Institute of Health and Biomedical Innovation

Finally, I wish to acknowledge the endless support of my husband, Rob, in allowing

me to achieve this goal My family has been a tower of strength and distraction, helping to maintain some reality I thank my friends for their support and patience during this time

CHAPTER 1 INTRODUCTION

1.1 The Research Problem

An individual’s symptom experience and ability to function in everyday life are increasingly recognised as important health outcomes for individuals with a chronic disease such as cancer (Lipscomb, Gotay, & Snyder, 2005; Sullivan, 2003) Patients with cancer experience multiple symptoms, so the investigation of individual symptoms to understand their symptom experience provides a limited perspective Dodd, Janson et al (2001) challenged oncology researchers to consider symptom clusters (i.e., a grouping of related, concurrent symptoms), in order to broaden the current perspectives on possible mechanisms underlying cancer related symptoms, and to suggest strategies that may advance cancer symptom management Conceptually, Dodd et al proposed that if a key symptom in a group of commonly occurring symptoms could be treated, the associated symptoms may also be relieved This assumes that either the key symptom is etiologically related or subject to common treatment approaches with other symptoms in the cluster The potential benefits of an approach which considers multiple symptoms in symptom management may therefore be: reduced or optimized polypharmacy, better outcomes for patients, increased clinician satisfaction, and reduced health care costs

At the commencement of this project in 2005, symptom cluster research was

in its early stages There were many gaps in our knowledge and understanding of symptom clusters Few studies identified symptom clusters empirically, and with no

specific guidance about analytical methods for determining valid and reliable symptom clusters, there was a risk that research in this field would not realise its potential Hence, the overall aim of this study was to investigate analytical approaches that are conceptually and contextually relevant for symptom cluster identification It was intended that the outcomes of this investigation would provide guidance for the most effective analytic methods for identifying valid and reliable symptom clusters, to support the advancement of symptom management in oncology

1.2 The Burden of Cancer

In the developed world, despite a reduction in the incidence of some cancers, the overall incidence of cancer cases continues to rise each year This trend is expected to continue, due in part to the ageing population (WHO, 2003) In Australia, cancer is a leading cause of death, and based on 2005 data, an estimated 111,000 new cases will be diagnosed in 2009 The current expectation is that 1 in 2 men, and 1 in 3 women, will be diagnosed with cancer before the age of 85 years (Cancer Council Australia, 2009) In 2005, the most common cancers, excluding non-melanoma skin cancer, were prostate (16, 349 cases), colorectal (13, 076), breast (12,265), skin melanomas (10,684), and lung cancer (9,182) Apart from sex-specific cancers, almost all cancers occur at higher rates in men than women The cost of health care services for cancer in Australia exceeds $3.8 billion

Nevertheless, the outlook for cancer patients has improved, due to earlier detection through screening, improved prevention strategies, technologies, and methods of treatment (Aziz & Rowland, 2003) In Australia, more than half the

will survive more than 5 years after diagnosis (Cancer Council Australia, 2009) Hence, with the increasing number of cancer survivors, many of whom are expected

to live for longer periods, further research is necessary, to understand the experience

of cancer patients in the short and long term, and to adequately address their needs (Haylock, 2006)

1.3 Rationale and Significance of the Research

A significant aspect of the cancer experience is the experience of disease- and treatment-related symptoms Research has typically conceptualized symptom experiences in terms of two attributes: (a) symptom occurrence (frequency and duration), and (b) symptom distress (Armstrong, 2003; Goodell & Nail, 2005; Lenz, Pugh, Milligan, Gift, & Suppe, 1997), where frequency is the number of times the symptom is experienced in a given time interval, and duration indicates how long that experience lasts The occurrence of symptoms often causes distress, which may

be physical or mental upset, anguish, or suffering (Rhodes & McDaniel, 1999) Importantly, an individual’s perceived degree of symptom distress has been identified as the main stimulus for individuals to act to relieve symptoms (Fu, Anderson, McDaniel, & Armer, 2002; Sweed, Schiech, Barsevick, Babb, & Goldberg, 2002), not simply the occurrence of the symptom For example, patients distressed by severe pain, or fatigue, are likely to be motivated to relieve, decrease,

or prevent distress, whereas patients experiencing minimal or no distress may not bother (Fu, LeMone, & McDaniel, 2004) A person’s symptom experience has thus been described as subjective, reflecting changes in an individual’s biological, psychological, cognitive, and social functioning (Dodd, Janson et al., 2001) For instance, the sensation of fatigue reflects changes in an individual’s ability to

conceptualizations emphasize the impact of individual symptoms for cancer patients, they also suggest that when multiple symptoms occur together, an individual’s distress is likely to be magnified Thus, an investigation of cancer related symptoms should reflect the patient’s perspective of their experience, and incorporate the

potential of symptom clusters

Despite advances in early diagnosis and insights into the causes and treatments for cancer, strategies to manage an individual’s experience of cancer-related symptoms and the side-effects of cancer have not progressed at the same pace In 2002, the National Institutes of Health (NIH) Symptom Management Conference Panel resolved multi-symptom research provided an opportunity to improve symptom management The NIH research directive was to focus initially,

on the most common side-effects of cancer and treatment, namely, pain, depression and fatigue (National Institutes of Health State-of-the-Science Panel, 2003) Subsequently, a number of correlation based studies investigated the relationships between cancer-related symptoms, their predictors and impact on daily living (Dodd, Miaskowski, & Paul, 2001; Francoeur, 2005) A potential benefit of a symptom cluster approach was that treating one symptom in a cluster may, directly or indirectly, resolve related symptoms, based on the assumption of a clinical, etiological relationship (e.g., treatment based) versus a statistical, least squares relationship (e.g., regression)

The investigation of symptom clusters is complex, given the variety of cancer diagnoses, treatments, patient characteristics and methods to collect and analyse data Conceptual frameworks to understand the symptom experience and determine

strategies to manage symptoms have been proposed (Armstrong, 2003; Dodd, Janson

et al., 2001; Lenz et al., 1997) and may be modified as new knowledge unfolds Furthermore, there is no agreed definition of the composition and fundamental characteristics of symptom clusters; there is limited understanding of how and why particular symptoms occur simultaneously; and there is a paucity of research to investigate multiple symptoms, as the disease progresses and treatments proceed To ensure symptom cluster research is of high scientific quality and adds to the knowledge in this field, a range of issues require further consideration, including the study design (e.g., heterogeneous/homogeneous samples), symptom measurements (e.g., scales, domains), the most useful timing of the symptom assessment, clinically, and appropriate analytical methods for symptom cluster identification

Currently, there is limited knowledge of the methodological decisions for symptom cluster research that may have important implications for research findings For example, a variety of symptom assessment tools exist, enabling the assessment of variation in individual symptoms, common to specific diagnoses and treatments (Beck, 2004; Miaskowski, Dodd, & Lee, 2004) As a consequence, different symptom clusters are likely, although a core set of common symptoms may form a cluster, consistently

There are two approaches to symptom cluster identification: the clinical approach and the empirical approach A clinically identified symptom cluster comprises symptoms that are observed to occur together, and may be associated Frequently co-occurring and distressing symptoms, such as pain, depression, fatigue, and insomnia have been targeted (Barsevick, 2007a, 2007b) However, in this pre-

determined symptom cluster approach, there is some uncertainty about statistical relatedness among symptoms in the cluster An empirical approach to identification has the advantage of identifying symptom clusters, based on statistically determined relationships, although the type and strength of the relationship between symptoms

in a cluster is undefined (Dodd, Miaskowski et al., 2001)

To date, investigators have been encouraged to test various conceptualizations and analytical approaches to define and identify cancer symptom clusters (Barsevick, Whitmer, Nail, Beck, & Dudley, 2006; Miaskowski, Aouizerat,

Dodd, & Cooper, 2007), as there is no best practice analytical approach However,

this variation in approaches to understanding symptom clusters is likely to contribute

to the variation in the symptom clusters identified The research and clinical implications of applying different methodologies are therefore a topic for further investigation To achieve reliable and valid results for clinical application, an agreed conceptual, methodological, and analytical basis is necessary to guide the research Specifically, the focus of the present study is to investigate analytical methods to empirically identify symptom clusters In examining the implications of using alternative analytical methods, this study considers related conceptual and methodological issues

At the commencement of this thesis, few studies empirically-identified cancer-related symptom clusters, and the stability of symptom clusters was investigated only in one study (Cooley, Short, & Moriarty, 2002; Gift, Stommel, Jablonski, & Given, 2003) Furthermore, there was no recommended approach to identify symptom clusters from the many multivariable and multivariate approaches

available Hence, further exploratory research was required, to identify and understand cancer related symptom clusters, and to justify which methods of identification may be appropriate for this context

This research study was conducted from a data analyst perspective, to apply and critique commonly used analytical methods for the identification of symptom clusters The significance of the exploratory research presented in this thesis was intended primarily to raise researcher awareness of issues associated with the selection of an analytical approach to identify symptom clusters, and to suggest analytical guidelines for future symptom cluster research

1.4 Research Purpose and Objectives

The purpose of this exploratory research was to investigate alternative approaches to identify symptom clusters for patients with cancer, using readily accessible statistical methods to assess multiple, related, co-occurring symptoms The conceptual framework for the identification of symptom clusters was the Theory

of Unpleasant Symptoms (Lenz et al., 1997), which incorporates the concept of multiple symptoms that occur simultaneously and may be interrelated The formation

of symptom clusters may be influenced by the demographic and clinical characteristics of patients, and symptom clusters may negatively affect patient outcomes Consequently, this theory conceptualises important theoretical relationships to investigate the assessment and management of symptom clusters

This thesis comprises three integrated studies designed to address the research objectives and answer the research questions outlined below The research objectives for this thesis were to: (a) investigate the appropriateness (i.e.,

conceptually and contextually) of the multivariate methods commonly used for

symptom cluster identification for cancer patients, to suggest best practice approaches (Study 1); (b) based on the best practice approaches of Study 1,

investigate alternative analytical methods to identify symptom clusters for cancer outpatients, at a point in time; (Study 2); and (c) continuing the theme of implementing multivariate methods for the empirical identification of symptom clusters and using approaches identified in Study 1 and Study 2, identify and investigate the stability of symptom clusters for cancer patients over one year (Study 3) An overall objective was to identify conceptual, methodological and analytical issues associated with the multivariate approaches implemented

(a) Do alternative analytical approaches identify similar symptom

clusters?

(b) What are the key issues associated with the empirical

identification of symptom clusters, in terms of selecting best

practice approaches (e.g., software limitations and study

design)?

Study 2:

What symptom clusters are identified by alternative, multivariate, best

practice methods determined in Study 1, for ambulatory cancer outpatients

within 1 month of commencing chemotherapy?

(a) Are the same symptom clusters identified by each method? (b) Which approaches may be recommended for cancer symptom

(b) Are the symptom clusters identified at the commencement of

chemotherapy, stable at 6 and 12 months?

For all studies:

What are the key conceptual, methodological, and analytical issues associated with the empirical identification of symptom clusters by multivariate methods?

1.6 Thesis Outline

The research problem for this project was determined in the context of an emerging area of cancer nursing research It is important to keep in mind that symptom clusters may be stable over time, but different patients may contribute to the symptom cluster differentially, due to individual differences in treatment and disease progression The background information for this thesis is presented in

cancer symptom cluster research since 2001, when a formal definition of a symptom cluster was proposed The concepts of a patient’s symptom experience and a symptom cluster are explained Alternative approaches to symptom cluster identification are outlined

In Chapter 3, to identify issues relevant to measuring cancer symptoms for the identification of symptom clusters, a brief review of the literature is presented Alternative approaches to symptom cluster identification are discussed Commonly used empirical methods of identification for symptom clusters are described Key conceptual, methodological, and analytical issues are identified for further

consideration by researchers, to determine best practice analytical methods for

symptom cluster identification

In Chapter 4, a published article of a systematic literature review of multivariate methods to identify cancer-related symptom clusters represents Study 1

of this research project The review critiqued various multivariate approaches commonly used to identify symptom clusters in adult cancer patients, with particular emphasis on their conceptual and contextual appropriateness, and best practice approaches

In Chapter 5, the conceptual framework and the methods for Study 2 and Study 3 of this research project are outlined The study design and procedures of the original study are presented prior to specification of the study design, procedures, and selected measures for this study The data preparation and procedures to deal

with missing data are detailed, followed by the analytical approaches for Study 2 and Study 3

In Chapter 6, a manuscript, submitted for publication, is the report of Study

2 Based on the best practice analytical methods identified in Study 1, the symptom

experience of patients was investigated, in terms of symptom clusters identified at a point in time In this secondary analysis of data, symptom distress ratings were assessed, one month after patients commenced chemotherapy, in a sample of 219 ambulatory cancer patients with mixed diagnoses Compared to the standard interpretations in factor analysis, alternative interpretations were implemented Recommendations were made for analytical choices in future symptom cluster studies In Chapter 7, a manuscript, submitted for publication, is the report of Study

3 which was an investigation of the stability of the symptom clusters identified in Study 2, within one month of commencing chemotherapy, and 6 and 12 months later, using separate exploratory factor analysis methods

In Chapter 8, the overall findings are discussed, to highlight the contribution

of this thesis to symptom cluster research The strengths and limitations of the project are discussed, and recommendations are made for the application of specific analytical methods for symptom cluster identification The clinical implications of this investigation of the cancer symptom experience, using a symptom cluster approach, are presented In conclusion, suggestions for further research including more advanced analytical approaches in longitudinal studies are proposed The conceptual, methodological and analytical issues identified throughout this project are addressed where relevant

CHAPTER 2 BACKGROUND

2.2 The Cancer Symptom Experience

An important aspect of cancer and its treatment is the experience and management of symptoms Symptom management is dynamic and results from a subjective response to the symptoms experienced (Fu et al., 2004) The symptom experience is an individual’s subjective response to symptom occurrence, intensity, distress, frequency, and meaning (Armstrong, 2003) Symptom experience may be influenced by personal and medical characteristics (e.g., age, gender, culture, financial status, family role, personality, cognitive capacity, motivation, physical

capacity, type of cancer, and disease progression (Dodd, Janson et al., 2001) For example, in a study of 434 ambulatory cancer patients with mixed diagnoses (Degner

& Sloan, 1995), women were more distressed than men, older patients were less distressed than younger patients, patients with advanced disease were more distressed than those with early-stage disease, lung cancer patients experienced highest distress, and genitourinary patients were least distressed Studies of adults with cancer report more distress for patients with lung cancer than for any other type

of cancer (Cooley, 2000)

2.2.1 Change in the Symptom Experience

Typically, people with cancer experience multiple symptoms, due to the disease, treatment, and comorbid conditions In a study of 1000 palliative care patients with mixed diagnoses, receiving no cancer related treatment, the median number of symptoms per patient was 11 (range 1-27) and the prevalence of ten symptoms ranged from 50% to 84% (Walsh, Donnelly, & Rybicki, 2000) Cancer treatments may exacerbate existing disease related symptoms (e.g., pain), or result in the development of a range of new, treatment induced symptoms (Khalid et al., 2007) The majority of patients undergoing chemotherapy and/or radiotherapy report significant fatigue during the course of treatment, as well as nausea, tiredness, hair loss, sleeping difficulties, and loss of appetite (Bower et al., 2000; Glaus et al., 2006; Iop, Manfredi, & Bonura, 2004; Okuyama et al., 2001)

The severity of cancer related symptoms may vary depending on the type of treatment and the sequence of treatments Williams et al (2001) reported chemotherapy patients (n=109) experienced greater severity than radiotherapy patients (n=161), with respect to feeling sluggish, difficulty sleeping, taste change,

fever, bruising, nausea, numbness, and hair loss On the other hand, radiotherapy patients reported greater severity than chemotherapy patients for weight loss, difficulty swallowing, sore throat, loss of interest in sex, and skin changes Furthermore, Donovan et al (2004) assessed fatigue at the beginning, middle, and end of treatment for 134 early stage breast cancer patients receiving radiotherapy and/or chemotherapy At the beginning of treatment, fatigue severity was greater for women undergoing chemotherapy than for those treated with radiotherapy, but fatigue increased for women treated with radiotherapy alone, compared to women pre-treated with chemotherapy, indicating a possible response shift (Donovan et al., 2004)

2.2.2 Change in the Symptom Experience Over Time

To date, the majority of multiple symptom studies have been cross-sectional, although after treatment, cancer related symptoms may persist For instance, in breast and ovarian patients, fatigue fluctuated during chemotherapy, and persisted for

at least 6 months (Payne, 2002), and fatigue in 763 breast cancer survivors persisted

10 years after diagnosis (Bower et al., 2006) As the disease and/or treatment progresses, symptom severity may change, such that, for 76 newly diagnosed Stage I-III breast cancer patients, the severity of worse sleep, fatigue, and depression increased during chemotherapy (Liu et al., 2009) Hence, to better understand the symptom experience, longitudinal studies are necessary, as many related factors change throughout the cancer continuum, from diagnosis to recovery or palliation

The cancer symptom experience is dynamic, as evidenced for 117 lung cancer patients over a six month period, following mixed treatments (Cooley, Short,

& Moriarty, 2003) Prevalence of the most distressing symptoms (fatigue, insomnia,

appetite) decreased from 0 to 3 months, but increased from 3 to 6 months Consistent with the findings from cross-sectional studies, many of the 13 symptoms were associated with demographic variables and treatment type These findings have been supported by more recent studies, where fatigue associated with prostate cancer was investigated before, during, and after radiation therapy, finding younger men with high fatigue levels at commencement of treatment were at risk for higher levels of fatigue, over the course of treatment (Miaskowski et al., 2008)

While symptom duration may be hours, days, months, or years, and symptom occurrence may vary across disease and treatment trajectories, timing is the one dimension of the symptom experience that has been least studied Due to the complex and dynamic nature of cancer and treatment related symptoms, there is much to understand about the day-to-day side effects of adjuvant therapy (Downie, Mar Fan, Houédé-Tchen, Yi, & Tannock, 2006), and how the symptom experience changes prior to, during, and following completion of treatment Thus, longitudinal studies are necessary to examine temporal change in symptoms

2 3 Concept of a Symptom Cluster

Despite advances in our understanding of cancer biology and improved treatment options, the negative effects of unrelieved symptoms may be: (a) interruption to treatment (Cleeland, 2000), (b) impaired functioning in everyday life (Dodd, Miaskowski et al., 2001), and (c) poorer quality of life (Chang, Hwang, Feuerman, & Kasimis, 2000) As a result, a diverse body of research exists, in an endeavor to understand the mechanisms, experiences, and effects of cancer related symptoms A new direction in cancer research considers symptom clusters

While cancer related symptoms have been conceptualized and defined in various ways, research indicates that people with cancer commonly experience multiple symptoms as a result of pathophysiological changes associated with the disease or the side-effects of treatments (Cleeland, 2000) Moreover, studies indicate that the concurrence of symptoms may adversely influence the symptom experience

An increase in the number and distress of symptoms has been associated with decreased quality of life (Chang et al., 2000; Cooley et al., 2003; Sweed et al., 2002), less effective role performance, and lower cognitive functioning (Bennett, Winters-Stone, & Nail, 2006; Sweed et al., 2002) These data have stimulated interest in research investigating multiple, co-occurring symptoms (symptom cluster research),

as a new direction for advancing cancer symptom management

Originally, a symptom cluster was conceptualized as three or more related and co-occurring symptoms (Dodd, Miaskowski et al., 2001), although co-occurring symptoms need not be related (e.g., hair loss and constipation) Furthermore, the effects of co-occurring symptoms may be synergistic and negatively impact important patient outcomes such as functional ability More recently, following a review of the concept of symptom clusters across psychology/psychiatry, general medicine and nursing, Kim, McGuire, Tulman, and Barsevick (2005) re-defined the definition of a symptom cluster as follows:

“A symptom cluster consists of two or more symptoms that are related to each other and that occur together Symptom clusters are composed of stable groups of symptoms, are relatively independent of other clusters, and may reveal specific underlying dimensions of symptoms Relationships among

symptoms within a cluster should be stronger than relationships among

symptoms across different clusters Symptoms in a cluster may or may not share a common etiology” (p 278)

One of the current issues in symptom cluster research is the need to clarify

the definition of a symptom cluster In particular, relatedness was not defined, but

may be attributed to covariance (correlation), a common mechanism, or a common etiology for symptoms (Dodd, Miaskowski et al., 2001), or when the impact associated with one symptom is different to the multiple symptom experience Miaskowski et al (2004) suggested this latter attribute may be used as a criterion for the significance of a symptom cluster Several characteristics of symptom clusters have not been specified, including: (a) the minimum number of symptoms required

to constitute a cluster, (b) the type and strength of the relationships, (c) the stability

of clusters, and (d) whether the existence of a symptom cluster suggests an underlying mechanism

Most of the attributes of a symptom cluster proposed by Kim et al (2005) reflect symptom clusters empirically derived from an exploratory factor analysis The assumption in exploratory factor analysis is that the correlations among symptoms are due to a common underlying cause The potential advantage of this approach to symptom cluster identification may be the revelation of a common etiology or underlying mechanism The alternative scenario, that symptoms in a cluster may not share the same etiology, is a basis for alternative analytical approaches to symptom cluster identification

2.4 The Clinical Relevance of Symptom Clusters

There are several reasons symptom clusters may be clinically important A growing body of evidence suggests that symptom clusters influence patient outcomes (Dodd, Miaskowski et al., 2001; Fox & Lyon, 2006; Gift, Jablonski, Stommel, & Given, 2004; Given, Given, Azzouz, & Stommel, 2001) The clinical relevance of symptom clusters thus lies in the potential they offer for new directions in the assessment and management of multiple symptoms (Miaskowski et al., 2004; Niven, 2003) That is, managing the symptom experience for individuals undergoing treatment is complex and dynamic, as symptoms may have multiple causes and influences A dilemma for cancer care is that the intervention should address the cause For example, the pathophysiology of cancer pain is relatively well understood (Cleeland et al., 2000) For advanced cancer pain, the pathophysiology is often multifactorial and complex, resulting directly from tumor involvement, the long-term effects of treatments, comorbid conditions, or a combination of these (Cleeland, 2000) Various etiologies for post-treatment pain include postsurgical pain, mucositis, lymphedema, and metastasis (Shapiro & Recht, 2001) For many other cancer-related symptoms, the pathophysiological mechanisms underlying the symptom are often unknown, so strategies must focus on controlling symptoms and minimizing their impact For example, the mechanisms underlying chemotherapy-induced diarrhoea and constipation are under investigation (Gibson & Keefe, 2006) Similarly, the precise causative factors for fatigue are often unknown, so interventions may be directed at related symptoms, such as sleep disturbances and nutritional deficiencies

Cleeland et al (2003) speculated that cancer related symptoms may have a shared biological mechanism and Armstrong (2004) proposed proinflammatory cytokines, associated with sickness behavior in animals, may play a role in symptom cluster production However, Miaskowsi and Aiouzerat (2007) suggested that although symptoms in a cluster may share a common etiology (e.g., nausea, diarrhoea, and anorexia may be associated with a gastrointestinal tumour or chemotherapy), this does not imply they share a common biologic mechanism On the other hand, symptoms in a cluster may have different etiologies (e.g., pain from bone metastasis, fatigue from treatment, depression from financial problems), but have a shared biological mechanism (Miaskowski & Aouizerat, 2007) Adding to this complexity is the growing understanding that various psychological, socio-cultural, and environmental factors may also influence how symptoms are experienced For example, depression is known to exacerbate pain and fatigue in individuals with cancer (Lawrence, Kupelnick, Miller, Devine, & Lau, 2004)

To date, researchers have not clearly shown that symptoms with different etiologies constitute a symptom cluster Nevertheless, the possibility of shared underlying causes of multiple symptoms in cancer patients may suggest physiological, psychological, behavioral, or biological mechanisms that differentiate the clusters (Armstrong, Cohen, Eriksen, & Hickey, 2004; Cleeland et al., 2003; Lee

et al., 2004; Parker et al., 2005) Understanding the causes and underlying mechanisms of symptom clusters may in turn identify opportunities for intervention That is, where common mechanisms exist, there is the theoretical possibility that treating one symptom in a cluster may relieve other related symptoms in the cluster and result in reduced therapeutic or pharmacological needs and improved patient