UNIVERSITY OF CALGARY Quality of Life After Prostate Cancer Diagnosis: A Longitudinal Prospective Cohort Study in Alberta, Canada by Megan Farris A THESIS SUBMITTED TO THE FACULTY OF GRADUATE STUDIES IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE GRADUATE PROGRAM IN COMMUNITY HEALTH SCIENCES CALGARY, ALBERTA SEPTEMBER, 2016 © Megan Farris 2016 Abstract OBJECTIVES: First, we examined the associations of post-diagnosis physical activity and change in pre-diagnosis physical activity on quality of life (QoL) in prostate cancer survivors. Then, we identified post-prostate cancer diagnosis QoL trajectories over time in the population. METHODS: 830 prostate cancer survivors were derived from a prior case-control study where information at diagnosis was collected, then survivors were re-consented into a follow-up study. Three repeated measurements of physical activity and QoL were undertaken post-diagnosis. RESULTS: We observed improvements in physical QoL in prostate cancer survivors who maintained or adopted higher levels of physical activity pre- and post-diagnosis, according to the cancer prevention physical activity guidelines compared to those who were non-exercisers. In the trajectory analysis, three physical and three mental trajectory groups were identified. CONCLUSION: With additional research, these established trajectory groups may help healthcare professionals in improving treatment and follow-up for this population of prostate cancer survivors. 2 Acknowledgements I would like to genuinely thank everyone involved in this Master’s thesis degree. Specifically, all of the staff in the Department of Cancer Epidemiology and Prevention Research at Alberta Health Services and the Department of Community Health Sciences at the University of Calgary. Foremost, I would like to thank my supervisor, Dr. Christine Friedenreich, for her constant support and mentorship throughout my Master's training. Without her careful guidance and vote of confidence, this thesis would not be the caliber it is today. I truly appreciate her time and willingness to help me achieve my goals and I would not be where I am today without her. In addition, I would like to extend a special thank you to my committee members: Dr. Karen Kopciuk for her careful guidance and eagerness to pursue new statistical methods, Dr. Kerry Courneya for his expertise, speedy reviews and generous praise and Dr. Elizabeth McGregor for her thorough remarks, epidemiological capabilities and genuine support. I would also like to recognize the contributions of my co-author, Qinggang Wang, on both manuscripts for his constructive feedback and interest in my work and data management of the study's datasets. I would like to thank Dr. Tavis Campbell, Professor, Department of Psychology at the University of Calgary for agreeing to serve as my internal-external examiner with the Department of Community Health Sciences. A special thank you is also necessary to all of the participants in the Prostate Cancer Cohort Study who generously gave their time to participate in the study making this project possible. I would like to thank all past and present study staff from the 3 Department of Cancer Epidemiology and Prevention Research who made this project possible. Thank you to my colleagues and peers who supported, mentored and kept me level-headed through this process: Eileen Shaw, Abbey Poirier, Trisha Kelly, Sarah MacLaughlin, Pamela Round and Dr. Darren Brenner. A special thanks to my peer mentor, Eileen Shaw (BB), I could not have done this without you! To my dear classmates and now friends, Kathryn Wiens, Lauren Hiersch, Amanda Barberio, Chelsea Doktorchik and my office mates Stephanie Gill and Alexis Mickle, thank you for being there for all of the ups and downs, for guidance and keeping me grounded, you truly made this experience wonderful. On a more personal note, I want to send my utmost thanks and praise to my loving parents. Without their support, willingness to listen to my problems and for providing dinner at the end of a long day, this degree would not have been finished on time. I would also like to thank my dearest boyfriend Troy Bozarth, for his patience and unconditional support of all my endeavors in every possible way; my two brothers, Mitchell Farris and Blake Farris for their support and occasional comic relief in times of stress; and my childhood friend Kathryn Saretsky for always being there for me. 4 Table of Contents Abstract ................................................................................................................................2 Acknowledgements ..............................................................................................................3 Table of Contents .................................................................................................................5 List of Tables .......................................................................................................................8 List of Figures ....................................................................................................................12 List of Symbols, Abbreviations and Nomenclature ...........................................................13 CHAPTER ONE: INTRODUCTION..................................................................... 15 1.1 Background ..............................................................................................................15 1.1.1 Burden of prostate cancer ................................................................................15 1.1.2 Importance of quality of life in prostate cancer...............................................16 1.1.2.1 Prevention of quality of life reductions through physical activity.........17 1.1.2.2 High-risk groups for low QoL in prostate cancer populations ..............18 1.2 Aims and hypotheses ...............................................................................................18 1.3 Ethics approval ........................................................................................................20 1.4 References ................................................................................................................21 CHAPTER TWO: LITERATURE REVIEW .......................................................... 25 2.1 Preamble ..................................................................................................................25 2.2 Part 1: Physical activity and prostate cancer ...........................................................25 2.2.1 Physical activity and prostate cancer risk ........................................................26 2.2.2 Physical activity and prostate cancer prognosis and survivorship ..................27 2.2.3 Physical activity and quality of life in prostate cancer survivors ....................28 2.3 Part 2: Change in quality of life after prostate cancer .............................................33 2.3.1 Long-term change in quality of life after prostate cancer: epidemiological evidence ...........................................................................................................34 2.3.2 Growth modelling and prostate cancer ............................................................37 2.4 References ................................................................................................................40 2.5 Tables .......................................................................................................................50 CHAPTER THREE: ASSOCIATION OF POST-DIAGNOSIS PHYSICAL ACTIVITY AND CHANGE IN PRE-DIAGNOSIS PHYSICAL ACTIVITY WITH QUALITY OF LIFE IN PROSTATE CANCER SURVIVORS ............. 62 3.1 Preamble ..................................................................................................................62 3.2 Abstract ....................................................................................................................64 3.3 Introduction ..............................................................................................................66 3.4 Materials and methods .............................................................................................67 3.4.1 Study design ....................................................................................................67 5 3.4.2 Data collection .................................................................................................68 3.4.3 Physical activity assessment ............................................................................69 3.4.4 Quality of life assessment ................................................................................70 3.4.5 Statistical analysis ...........................................................................................71 3.5 Results ......................................................................................................................72 3.5.1 Descriptive statistics ........................................................................................72 3.5.2 Post-diagnosis physical activity and quality of life at first follow-up .............73 3.5.3 Change in physical activity over the diagnostic period and quality of life .....74 3.6 Discussion ................................................................................................................75 3.7 Conclusion ...............................................................................................................79 3.8 Manuscript acknowledgements................................................................................80 3.9 References ................................................................................................................81 3.10 Tables and figures ..................................................................................................87 3.11 Additional analyses and results: Lasso and elastic-net regularized generalized linear model (GLMNET package) covariate selection methods............................96 CHAPTER FOUR: IDENTIFICATION AND PREDICTION OF QUALITY OF LIFE TRAJECTORIES AFTER A PROSTATE CANCER DIAGNOSIS ..... 103 4.1 Preamble ................................................................................................................103 4.2 Abstract ..................................................................................................................105 4.3 Introduction ............................................................................................................107 4.4 Materials and methods ...........................................................................................108 4.4.1 Study sample .................................................................................................108 4.4.2 Data collection ...............................................................................................109 4.4.3 Statistical analysis .........................................................................................110 4.4.3.1 Identifying trajectory groups ...............................................................110 4.4.3.2 Characteristics of trajectory groups .....................................................112 4.4.3.3 Sensitivity analyses ..............................................................................113 4.5 Results ....................................................................................................................113 4.5.1 Sample ...........................................................................................................113 4.5.2 Trajectory group selection and evaluation ....................................................114 4.5.3 Characteristics of trajectory groups ...............................................................116 4.5.4 Dropout within trajectory groups ..................................................................118 4.5.5 Sensitivity analyses .......................................................................................119 4.6 Discussion ..............................................................................................................120 4.7 Conclusion .............................................................................................................124 4.8 Manuscript acknowledgements..............................................................................124 4.9 References ..............................................................................................................125 4.10 Tables and Figures ...............................................................................................129 4.11 Additional analyses and results: Sensitivity analysis tables and figures .............143 4.11.1 Dropout within trajectory groups ................................................................143 4.11.2 Sensitivity analyses: Complete case analysis for physical and mental QoL trajectories .............................................................................................150 4.11.3 Sensitivity analyses: Two complete assessments analysis for physical and mental QoL trajectories ...........................................................................158 4.11.4 Sensitivity analysis: Not modelling dropout on covariate significance analysis ...........................................................................................................166 6 4.11.5 Sensitivity analyses: Dropout switchers between trajectory groups with and without modelling dropout ......................................................................170 4.11.6 Sensitivity analysis: Time lagged model for previous quality of life score analysis ...........................................................................................................173 CHAPTER FIVE: DISCUSSION........................................................................ 177 5.1 Summary of findings .............................................................................................177 5.2 Limitations .............................................................................................................178 5.2.1 Internal validity .............................................................................................179 5.2.1.1 Survivorship bias due to missing data and dropout .............................179 5.2.1.2 Measurement error with self-administered questionnaires ..................181 5.2.1.3 Confounding ........................................................................................183 5.2.1.4 Precision and issues with multiple statistical tests...............................185 5.2.1.5 Causal inference ...................................................................................187 5.2.2 External validity ............................................................................................187 5.3 Public health implications ......................................................................................188 5.4 Future directions ....................................................................................................190 5.5 Conclusion .............................................................................................................191 5.6 References ..............................................................................................................193 Appendices A: Lifetime Physical Activity Questionnaire ...............................................195 Appendices B: Past-Year Physical Activity Questionnaire .............................................204 Appendices C: Well-being Questionnaire .......................................................................213 7 List of Tables Table 2.1. Physical activity and SF-36 quality of life studies in prostate cancer populations. ............................................................................................................... 50 Table 2.2. Prostate cancer cohort studies examining changes in QoL post-diagnosis using the SF-36 published between 2011-2016. ....................................................... 57 Table 3.1. Characteristics for prostate cancer survivors (n=817), in the Prostate Cancer Cohort Study, Alberta, Canada, 1997-2014.................................................. 87 Table 3.2. Associations between post-diagnosis first follow-up physical activity and post-diagnosis first follow-up quality of life in prostate cancer survivors in Alberta, Canada in 1997-2002. ................................................................................. 90 Table 3.3. Associations between physical activity change (pre-diagnosis minus average post-diagnosis scores) over the diagnosis period and average postdiagnosis quality of life in prostate cancer survivors in Alberta, Canada in 19972007........................................................................................................................... 92 Table 3.4. Change in moderate-to-vigorous recreational physical activity based on meeting the cancer prevention guidelines (150 minutes/week) over the diagnostic period (pre- and average post-diagnosis) and average post-diagnosis quality of life in prostate cancer survivors in Alberta, Canada in 1997-2007. .......................... 94 Table 3.5. Associations between post-diagnosis physical activity on first follow-up post-diagnosis QoL using lasso and elastic-net regularized generalized linear model (GLMNET package) methods in prostate cancer survivors in Alberta, Canada in 1997-2002 (n=817). ................................................................................. 98 Table 4.1. Model selection to determine number of groups for physical quality of life post-diagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (19972007). ...................................................................................................................... 129 Table 4.2. Model selection for determining linear or quadratic structure of trajectories for physical quality of life post-diagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). ............................................................................ 130 Table 4.3. Model selection to determine number of groups for mental quality of life post-diagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (19972007). ...................................................................................................................... 131 Table 4.4. Model selection for determining linear or quadratic structure of trajectories for mental quality of life post-diagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007).................................................................................. 132 8 Table 4.5. Descriptive characteristics by physical quality of life trajectory group of prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007)..................... 133 Table 4.6. Descriptive characteristics by mental quality of life trajectory group of prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007)..................... 135 Table 4.7. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the averagemaintaining/increasing QoL group in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007).................................................................................. 137 Table 4.8. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a time-varying covariate in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). ................................................................................. 139 Table 4.9. Logistic regression models assessing the relationship between dropping out and covariates within the very low-maintaining physical QoL trajectory group. ... 144 Table 4.10. Logistic regression models assessing the relationship between dropping out and covariates within the low-declining physical QoL trajectory group. ......... 145 Table 4.11. Logistic regression models assessing the relationship between dropping out and covariates within the average-maintaining physical QoL trajectory group. ...................................................................................................................... 146 Table 4.12. Logistic regression models assessing the relationship between dropping out and covariates within the low-increasing mental QoL trajectory group. .......... 147 Table 4.13. Logistic regression models assessing the relationship between dropping out and covariates within the above average-declining mental QoL trajectory group. ...................................................................................................................... 148 Table 4.14. Logistic regression models assessing the relationship between dropping out and covariates within the average-maintaining mental QoL trajectory group. . 149 Table 4.15 Model selection to determine number of groups for physical quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (19972007). ...................................................................................................................... 151 Table 4.16. Model selection for determining linear or quadratic structure of trajectories for physical quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007).............................................................. 152 Table 4.17. Model selection to determine number of groups for mental quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (19972007). ...................................................................................................................... 153 9 Table 4.18. Model selection for determining linear or quadratic structure of trajectories for mental quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007).............................................................. 154 Table 4.19. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the averagemaintaining/increasing QoL group in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007).................................................................................. 155 Table 4.20. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a time-varying covariate in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007). ................................................................................. 157 Table 4.21. Model selection to determine number of groups for physical quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007). ............................................................................................................ 159 Table 4.22. Model selection for determining linear or quadratic structure of trajectories for physical quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007).............................................................. 160 Table 4.23. Model selection to determine number of groups for mental quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (19972007). ...................................................................................................................... 161 Table 4.24. Model selection for determining linear or quadratic structure of trajectories for mental quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007).............................................................. 162 Table 4.25. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the averagemaintaining/increasing QoL group in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007).................................................................................. 163 Table 4.26. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a time-varying covariate in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007). ................................................................................. 165 Table 4.27. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the averagemaintaining/increasing QoL group without modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). ............................................. 167 10 Table 4.28. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a time-varying covariate and not modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). ............................................. 169 Table 4.29. Participants who switched physical quality of life trajectory groups when modelling dropout vs. not modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). .................................................................... 171 Table 4.30. Participants who switched mental quality of life trajectory groups when modelling dropout vs. not modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). .................................................................... 172 Table 4.31. Time-lagged multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the average-maintaining/increasing QoL group in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). .................................................................... 174 Table 4.32. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory time-lagged models including physical activity as a time-varying covariate in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). .................................................................... 176 11 List of Figures Figure 3.1. Change in moderate-to-vigorous recreational physical activity guideline adherence groups over the diagnostic period in prostate cancer survivors in Alberta, Canada (2000-2007).................................................................................... 95 Figure 3.2. Total physical activity and PCS score output using GLMNET regularization. ........................................................................................................... 99 Figure 3.3. Total physical activity and MCS score output using GLMNET regularization. ......................................................................................................... 100 Figure 3.4. Recreational, occupational and household physical activity (mutually adjusted) and PCS score output using GLMNET regularization. ........................... 101 Figure 3.5. Recreational, occupational and household physical activity (mutually adjusted) and MCS score output using GLMNET regularization........................... 102 Figure 4.1. Repeated measurements and cohort timeline. Past two years of physical activity were recalled and past four weeks of QoL (N = complete QoL measurements at each time point). Abbreviations: PA = physical activity, QoL = quality of life, Dx = diagnosis................................................................................. 140 Figure 4.2. Final adjusted model selected for physical QoL trajectory groups (black lines are trajectory groups, grey lines are 95% confidence and dot symbols are observed group means at each time-point) in cohort of prostate cancer survivors in Alberta, Canada (n=817). Abbreviations: QoL=quality of life, DP=dropout probability. .............................................................................................................. 141 Figure 4.3. Final adjusted model selected for mental QoL trajectory groups (black lines are trajectory groups, grey lines are 95% confidence intervals and dot symbols are observed group means at each time-point) in cohort of prostate cancer survivors in Alberta, Canada (n=817). Abbreviations: QoL=quality of life, DP=dropout probability. .................................................................................. 142 12 List of Symbols, Abbreviations and Nomenclature Symbol > < Definition Greater than Less than ≥ ≤ ± Δ ACSM ADT BMI CaPSURE Greater than or equal to Less than or equal to Plus or minus Change Beta coefficient American College of Sports Medicine Androgen Deprivation Therapy Body Mass Index Cancer of the Prostate Strategic Urologic Research Endeavour Conjoint Health Research Ethics Board Confidence Interval European Organization for Research and Treatment of Cancer Core Quality of Life Questionnaire European Organization for Research and Treatment of Cancer Quality of Life Questionnaire-Prostate 25 Expanded Prostate Index Composite Functional Assessment of Cancer TherapyProstate/Breast/General High-Intensity Focused Ultrasound Hazard Ratio hour International Physical Activity Questionnaire International Prostate Symptom Score Incidence Rate Ratio Lifetime Physical Activity Questionnaire Mental Component Summary Medical Subject Heading Metabolic Equivalent Nerve Sparing Radical Prostatectomy Physical Component Summary Patient-Oriented Prostate Utility Scale Prostate Specific Antigen Past Year Total Physical Activity Questionnaire Quality of Life Relative Risk Standard Deviation Short Form-36 Specific, Measurable, Achievable, Realistic and Timebound CHREB CI EORTC QLQ-C30 EORTC QLQ-PR25 EPIC FACT-P/B/G HIFU HR hr IPAQ IPSS IRR LTPAQ MCS MeSH MET NSRP PCS PORPUS PSA PYTPAQ QoL RR SD SF-36 SMART 13 SMD UCLA-PCI USA Standard Mean Difference University of California, Los Angeles-Prostate Cancer Index United States of America 14 Chapter One: INTRODUCTION 1.1 Background 1.1.1 Burden of prostate cancer Prostate cancer is one of the most prevalent cancers among Canadian men (1, 2). In 2015 an estimated 24,000 new cases and 4,100 deaths from prostate cancer occurred in Canada accounting for 23.9% cancer diagnoses and 10.1% cancer deaths in men (3). Early detection and improved treatment have resulted in a decrease in mortality rates and greater numbers of prostate cancer survivors (4). In 2015 there were 176,365 prevalent prostate cancer cases in Canada and five-year relative survival rates were approximately 93% (3). One major contributing factor to the increased prevalence of prostate cancer in developed countries was the introduction of the Prostate Specific Antigen (PSA) testing in the early 1990s (5, 6). PSA is a protein produced in the prostate gland and the PSA test measures the level of PSA in blood (7). Until recently, if elevated levels of PSA were detected (>4.0 ng/mL), it was recommended to monitor and potentially biopsy for prostate cancer (8). However, benign medical conditions such as prostatitis or urinary tract infection may also result in increased PSA levels (9). Further, recent studies have also found that prostate cancer patients can have low PSA levels while others might have high PSA levels and therefore, putting into question the sensitivity and specificity of the test (10). While early detection of prostate cancer can lead to decreased mortality, overdetection of low grade non-threatening prostate cancers can negatively affect the health of individuals, as well as the healthcare system. This increase of detection of low-grade prostate cancers has resulted in over-treatment with a consequent increase in the side effects related to prostate cancer treatments. These side effects may include reduced 15 physical functioning, a compromised mental state and overall reduced quality of life (QoL) (11-15). As a result, PSA testing are under scrutiny to provide more accurate thresholds and recommendations for better detection methods of prostate cancer (16). In addition, there has been an increased awareness of the need for “watchful waiting” rather than active treatment for some early stage prostate cancers (17). 1.1.2 Importance of quality of life in prostate cancer High priorities in prostate cancer control involve improving treatment and care of prostate cancer and reducing the burdens of living beyond prostate cancer which includes improving QoL and addressing the specific needs of prostate cancer survivors (18). Soon after prostate cancer diagnosis, patients undergo different treatment regimens that include surgery, hormone therapy and radiation therapy that have been shown to improve prostate cancer survival (19). However, living longer with prostate cancer may be associated with factors that impact QoL and survivorship including: treatment side effects, co-morbidities (including related causes of death, such as coronary heart disease), age, race and lifestyle behavioural factors (20). These factors may be more detrimental than the diagnosis itself and also lead to a decreased QoL post-diagnosis. QoL assessments in cancer patients measure physical, mental, social, and spiritual well-being. Physical well-being generally refers to functional ability in daily living and is well-recognized from a public health perspective. It is measured through objective tests like muscular strength, endurance, and independence (21). Specifically for prostate cancer, sufficient function of the lower extremities is very important since the cancer may produce disease-specific deficiencies such as sexual dysfunction, impaired bowel function and urinary incontinence (22-25). Prostate cancer patients suffer psychological and sexual problems especially post16 treatment, that have major impacts on QoL (26). Alternatively, patient feedback can measure aspects of disease and treatment including: various symptoms, symptom severity, symptom frequency, nature of disability that are specific to the patient, impacts on daily life, and perceptions or feelings of the patient towards their conditions (27). The patients' concerns are important to examine since not all physical, physiological or biological exams are able to capture this type of information on patients' overall wellbeing. Therefore, patients may need their treatment regimens individualized to account for individual patient differences and to deliver optimal care after prostate cancer diagnosis. 1.1.2.1 Prevention of quality of life reductions through physical activity While different treatments negatively affect QoL after prostate cancer diagnosis (28, 29), long-term effects of diagnosis and treatment remain poorly documented and understood (15). Therefore, healthcare systems are now focusing on preventing reductions of QoL after prostate cancer diagnosis through different mechanisms. Prognostic factors such as age, stage of cancer, PSA levels, and co-morbidities are nonmodifiable (4, 30), whereas dietary intake and physical activity are modifiable lifestyle factors that may be of importance for maintaining QoL post-diagnosis (22, 31). There is a wealth of literature focusing on physical activity interventions and the impacts of treatments related to prostate cancer, including androgen deprivation therapy (ADT), which indicate improvements in QoL in prostate cancer populations (32-37). However, most studies focus on exercise interventions directly after prostate cancer diagnosis, therefore, long term QoL has not been assessed. There is a need for more research 17 focusing on different types of physical activity, as well as documented change in physical activity behaviours across the diagnostic period and the impacts on QoL. 1.1.2.2 High-risk groups for low QoL in prostate cancer populations Another target of prevention and/or intervention is to identify subgroups in the population that are at higher risk for prolonged suffering after prostate cancer diagnosis (38). In this case, higher risk would be defined as those with low or declining QoL after prostate cancer diagnosis. Identifying those who experience reduced QoL after diagnosis of prostate cancer is crucial to improving care. Change in QoL has generally been examined as a single or average pattern after diagnosis of prostate cancer. There is no substantial evidence behind why there may only be one pattern of QoL or if there are indeed differences between prostate cancer survivors’ QoL after diagnosis. There may be subtle differences between prostate cancer survivors that are missed by modelling the average QoL over time after diagnosis. Therefore, a need exists to test whether or not there are different trajectory patterns of QoL long term after prostate cancer diagnosis among prostate cancer survivors. Health professions need to be informed of different trajectory patterns in order to target specific QoL prostate cancer survivors to prevent any declines in health, QoL and survival. 1.2 Aims and hypotheses The aim of this thesis was to examine different aspects of prostate cancer survivors’ QoL after diagnosis. This was accomplished by examining, first, how different types of post-diagnosis physical activity as well as changes from pre- to post-diagnosis physical activity influenced QoL post-diagnosis among a cohort of prostate cancer 18 survivors. Secondly, we investigated if there were different QoL trajectories within the study population and characterized these trajectory groups systematically. Specifically, the study objectives were as follows: 1) To investigate the association between post-diagnosis total, recreational, occupational and household physical activity and QoL post-diagnosis in a cohort of prostate cancer survivors in Alberta. We hypothesize that a per unit increase in post-diagnosis physical activity will be associated with positive QoL scores in a cohort of prostate cancer survivors. 2) To investigate the association between the change in pre- and post-diagnosis total, recreational, occupational and household physical activity and QoL postdiagnosis in a cohort of prostate cancer survivors. Further, to investigate the association between adherence to cancer prevention physical activity guidelines pre- and post-diagnosis and QoL post-diagnosis in prostate cancer survivors in Alberta. We hypothesize that a per unit increase of physical activity change scores will be positively associated with QoL scores in a cohort of prostate cancer survivors. Secondly, we hypothesized that meeting cancer prevention physical activity guidelines pre- and post-diagnosis would result in higher postdiagnosis QoL scores in prostate cancer survivors compared to those who did not meet guidelines. 3) To examine post-diagnosis QoL trajectory groups in a cohort of prostate cancer survivors in Alberta, during the follow-up period. We hypothesize that there 19 would be more than one and up to four different trajectories of QoL in prostate cancer survivors over time. 1.3 Ethics approval The University of Calgary Conjoint Health Research Ethics Board (CHREB) has approved the protocol for the “Cohort Study of Physical Activity and Prostate Cancer Survival” (16369) project and approved the addition of me as a study investigator to undertake the work for this thesis project, September 25, 2015. 20 1.4 References 1. Siegel R, DeSantis C, Virgo K, Stein K, Mariotto A, Smith T, et al. Cancer treatment and survivorship statistics, 2012. CA Cancer J Clin. 2012;62(4):220-41. 2. Torre LA, Siegel RL, Ward EM, Jemal A. Global Cancer Incidence and Mortality Rates and Trends-An Update. Cancer Epidemiol Biomarkers Prev. 2016;25(1):16-27. 3. Canadian Cancer Society's Advisory Committee on Cancer S. Canadian Cancer Statistics 2015. Toronto, ON: Canadian Cancer Society, 2015. 4. Aziz NM. Cancer survivorship research: challenge and opportunity. J Nutr. 2002;132:3494S-503S. 5. Ferlay J SI, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman D, Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence and Mortality Worldwide: IARC CancerBase No. 11 Lyon, France: International Agency for Research on Cancer 2013. 6. Hoffman RM. Clinical practice. Screening for prostate cancer. N Engl J Med. 2011;365(21):2013-9. 7. Stenman U-H, Leinonen J, Zhang W-M, Finne P, editors. Prostate-specific antigen. Seminars in cancer biology; 1999: Elsevier. 8. Catalona WJ, Hudson MA, Scardino PT, Richie JP, Ahmann FR, Flanigan RC, et al. Selection of optimal prostate specific antigen cutoffs for early detection of prostate cancer: receiver operating characteristic curves. J Urol. 1994;152(6 Pt 1):2037-42. 9. Oesterling JE. Prostate-specific antigen and diagnosing early malignancies of the prostate. J Cell Biochem Suppl. 1992;16h:31-43. 10. Duffy MJ. PSA in screening for prostate cancer: more good than harm or more harm than good? Adv Clin Chem. 2014;66:1-23. 11. Wilt TJ, Brawer MK, Jones KM, Barry MJ, Aronson WJ, Fox S, et al. Radical Prostatectomy versus Observation for Localized Prostate Cancer. N Engl J Med. 2012;367(3):203-13. 12. Drummond FJ, Kinnear H, O'Leary E, Donnelly, Gavin A, Sharp L. Long-term health-related quality of life of prostate cancer survivors varies by primary treatment. Results from the PiCTure (Prostate Cancer Treatment, your experience) study. J Cancer Surviv. 2015;9(2):361-72. 13. Carlsson S, Drevin L, Loeb S, Widmark A, Lissbrant IF, Robinson D, et al. Population-based study of long-term functional outcomes after prostate cancer treatment. BJU Int. 2016;117(6b):E36-E45. 21 14. Sun F, Oyesanmi O, Fontanarosa J, Reston J, Guzzo T, Schoelles K. Therapies for Clinically Localized Prostate Cancer: Update of a 2008 Systematic Review. Comparative Effectiveness Reviews No. 146. Report No. 15-EHC004-EF. Rockville (MD): Agency for Healthcare Research and Quality, 2014. 15. Davis KM, Kelly SP, Luta G, Tomko C, Miller AB, Taylor KL. The association of long-term treatment-related side effects with cancer-specific and general quality of life among prostate cancer survivors. Urology. 2014;84(2):300-6. 16. Ilic D, Neuberger MM, Djulbegovic M, Dahm P. Screening for prostate cancer. Cochrane Database Syst Rev. 2013(1):Cd004720. 17. Chung MS, Lee SH. Current status of active surveillance in prostate cancer. Investig Clin Urol. 2016;57(1):14-20. 18. Richards M, Corner J, Maher J. The National Cancer Survivorship Initiative: new and emerging evidence on the ongoing needs of cancer survivors. Br J Cancer. 2011;105 Suppl 1:S1-4. 19. Mannuel HD, Hussain A. Evolving role of surgery, radiation, hormone therapy, and chemotherapy in high-risk locally advanced prostate cancer. Clin Genitourin Cancer. 2006;5(1):43-9. 20. Thorsen L, Courneya KS, Stevinson C, Fossa SD. A systematic review of physical activity in prostate cancer survivors: outcomes, prevalence, and determinants. Support Care Cancer. 2008;16(9):987-97. 21. Bowling A. Measuring Health: A Review of Quality of Life Measurement Scales. Buckingham: Open University Press; 2005. 22. Galvao DA, Spry N, Denham J, Taaffe DR, Cormie P, Joseph D, et al. A multicentre year-long randomised controlled trial of exercise training targeting physical functioning in men with prostate cancer previously treated with androgen suppression and radiation from TROG 03.04 RADAR. Eur Urol. 2014;65(5):856-64. 23. Elliott S, Latini DM, Walker LM, Wassersug R, Robinson JW. Androgen deprivation therapy for prostate cancer: recommendations to improve patient and partner quality of life. J Sex Med. 2010;7(9):2996-3010. 24. Galvao DA, Taaffe DR, Spry N, Joseph D, Newton RU. Cardiovascular and metabolic complications during androgen deprivation: exercise as a potential countermeasure. Prostate Cancer Prostatic Dis. 2009;12(3):233-40. 25. Sung JF, Lin RS, Pu YS, Chen YC, Chang HC, Lai MK. Risk factors for prostate carcinoma in Taiwan: a case-control study in a Chinese population. Cancer. 1999;86(3):484-91. 22 26. MacKenzie KR, Aning JJ. GPs could play key role in prostate cancer survivorship programmes. Practitioner. 2014;258(1776):27-31, 3. 27. Deshpande PR, Rajan S, Sudeepthi BL, Abdul Nazir CP. Patient-reported outcomes: A new era in clinical research. Perspect Clin Res. 2011;2(4):137-44. 28. Ball AJ, Gambill B, Fabrizio MD, Davis JW, Given RW, Lynch DF, et al. Prospective longitudinal comparative study of early health-related quality-of-life outcomes in patients undergoing surgical treatment for localized prostate cancer: a shortterm evaluation of five approaches from a single institution. J Endourol. 2006;20(10):723-31. 29. Soderdahl DW, Davis JW, Schellhammer PF, Given RW, Lynch DF, Shaves M, et al. Prospective longitudinal comparative study of health-related quality of life in patients undergoing invasive treatments for localized prostate cancer. J Endourol. 2005;19(3):318-26. 30. Rhee H, Gunter JH, Heathcote P, Ho K, Stricker P, Corcoran NM, et al. Adverse effects of androgen-deprivation therapy in prostate cancer and their management. BJU Int. 2015;115 Suppl 5:3-13. 31. Ottenbacher AJ, Day RS, Taylor WC, Sharma SV, Sloane R, Snyder DC, et al. Long-term physical activity outcomes of home-based lifestyle interventions among breast and prostate cancer survivors. Support Care Cancer. 2012;20:2483-9. 32. Chipperfield K, Brooker J, Fletcher J, Burney S. The impact of physical activity on psychosocial outcomes in men receiving androgen deprivation therapy for prostate cancer: a systematic review. Health Psychol. 2014;33(11):1288-97. 33. Gardner JR, Livingston PM, Fraser SF. Effects of exercise on treatment-related adverse effects for patients with prostate cancer receiving androgen-deprivation therapy: a systematic review. J Clin Oncol. 2014;32(4):335-46. 34. Winters-Stone KM, Beer TM. Review of exercise studies in prostate cancer survivors receiving androgen deprivation therapy calls for an aggressive research agenda to generate high-quality evidence and guidance for exercise as standard of care. J Clin Oncol. 2014;32(23):2518-9. 35. Keogh JW, MacLeod RD. Body composition, physical fitness, functional performance, quality of life, and fatigue benefits of exercise for prostate cancer patients: a systematic review. J Pain Symptom Manage. 2012;43(1):96-110. 36. Hasenoehrl T, Keilani M, Sedghi Komanadj T, Mickel M, Margreiter M, Marhold M, et al. The effects of resistance exercise on physical performance and health-related quality of life in prostate cancer patients: a systematic review. Support Care Cancer. 2015;[Epub ahead of print]. 23 37. Teleni L, Chan RJ, Chan A, Isenring EA, Vela I, Inder WJ, et al. Exercise improves quality of life in androgen deprivation therapy-treated prostate cancer: systematic review of randomised controlled trials. Endocr Relat Cancer. 2016;23(2):10112. 38. Dirksen SR, Belyea MJ, Wong W, Epstein DR. Transitions in Symptom Cluster Subgroups Among Men Undergoing Prostate Cancer Radiation Therapy. Cancer Nurs. 2016;39(1):3-11. 24 Chapter Two: LITERATURE REVIEW 2.1 Preamble This chapter reviews two separate bodies of literature: 1) the scientific evidence regarding the association between physical activity and prostate cancer risk, mortality and QoL, and; 2) the evidence regarding how QoL changes after prostate cancer diagnosis. 2.2 Part 1: Physical activity and prostate cancer There is accumulating epidemiologic evidence that physical activity is associated with both cancer risk and recently also with cancer survival. To summarize this literature, two separate searches of the MEDLINE database using PubMed, up to June 3rd, 2016, were conducted on the association between physical activity and prostate cancer risk and survival. Keywords and medical subject heading (MeSH) terms (exploded) were used. For prostate cancer risk, the search terms were: ((“physical activity” OR “motor activity” OR exercise) AND (“prostate cancer” OR “prostatic cancer”) AND (risk OR “risk factors” OR “risk factor”)), including Boolean operating terms. This search rendered 465 results, of which, the most recent select studies and review paper are reported to briefly review this literature. For prostate cancer survival, the search terms were: ((“physical activity” OR “motor activity” OR exercise) AND (“prostate cancer” OR “prostatic cancer”) AND (survivor* OR survivorship OR patients)), including Boolean operating terms. This second search rendered 467 results and no language, date, or geographical restrictions were applied. Abstracts and unpublished results were not included. Overall, only four studies fit the criteria examining how physical activity impacts survival in a prostate cancer population. All other search results were irrelevant in regards to study population, exposure assessment, outcome assessment or otherwise 25 off topic. The four relevant publications including observational studies examining physical activity as an exposure and prostate cancer as an outcome, or prostate cancer as the population of interest with mortality or progression as the outcome, are discussed below. 2.2.1 Physical activity and prostate cancer risk Physical activity has been identified as a possible protective behaviour for several cancer sites (1). There is plausible evidence for an etiologic role of physical activity in breast (2), colorectal (3), and endometrial cancers (4). However, there is inconsistent evidence that physical activity is associated with prostate cancer risk (1, 5, 6). To date, more than 55 retrospective and prospective epidemiological studies have examined the association between physical activity and prostate cancer risk (7, 8). Of these studies, most demonstrate a slight protective effect (7, 9-11), while others support a null hypothesis (12, 13) and increased risk (14, 15). These differences in the literature may be attributable to physical activity measurement error, long latency periods and slow tumour growth for prostate cancer (7). Investigations into this association often experience difficulties comparing prostate cancer cases and healthy controls, given the high probability of undetected latent prostate cancer in the general population. Further, there may be unmeasured competing risks or confounding variables potentially generating differences in results between studies. Although the evidence is inconsistent for a role of physical activity in prostate cancer risk overall, there is some evidence of a protective effect on the risk of advanced prostate cancer (6, 16). The Health Professionals Follow-up Study (17) identified no association with observed total prostate cancer cases and vigorous physical activity 26 (relative risk (RR) = 1.09, 95% confidence interval (CI): 0.94-1.28) but supported a reduced risk in the highest intensity of vigorous activity and advanced or fatal prostate cancer cases in men older than 65 years of age (RR = 0.33, 95% CI: 0.17-0.62). Furthermore, the American Cancer Society Cancer Prevention Study II Nutrition Cohort (18) collected information on recreational physical activity, reporting no association in overall prostate cancer risk (RR = 0.90, 95% CI: 0.78-1.04), but a potential association of reduced risk of aggressive prostate cancer (RR = 0.69, 95% CI: 0.52-0.92). On the contrary, the European Prospective Investigation into Cancer Cohort Study (19) reported no association between vigorous recreational physical activity and advanced prostate cancer risk (incidence rate ratio (IRR) = 1.23, 95% CI: 0.91-1.65), but an inverse association between occupational vigorous physical activity and advanced prostate cancer risk (IRR = 0.75, 95% CI: 0.53-1.05). These inconsistencies may be due to measurement error associated with the assessment of physical activity. Across studies, there is not a uniform definition of the highest levels of physical activity. Nonetheless, there remains a potential for physical activity being an etiologically relevant lifestyle factor for prostate cancer, particularly given its long latency period (20). 2.2.2 Physical activity and prostate cancer prognosis and survivorship Prostate cancer has a promising prognosis (21), with the five-year survival rate of approximately 93% in Canada (22). There is now emerging evidence regarding a potentially promising role for physical activity as a protective agent to reduce side effects of cancer, other comorbid conditions and causes of death. To date, few prospective epidemiologic studies have examined the role of increased regular physical activity in improving prostate cancer survival (23-26). Three of these studies examined the 27 association between post-diagnosis physical activity on all-cause and cancer-specific mortality (23, 24, 26). The first study examining the association between total physical activity and prostate cancer prognosis was published in 2011 (23), with 33-35% risk reductions observed in both overall and prostate cancer-specific mortality. Two other studies (24, 26) found risk reductions similar in magnitude, however, their lowest risk reductions were attributed to different amounts of recreational physical activity rather than total activity. Richman and colleagues (25), found a reduced risk of prostate cancer progression with increased walking pace regardless of walking duration. It has been hypothesized that men may reduce their physical activity as their health decreases after prostate cancer. Therefore, the focus of this study on the progression of disease rather than mortality reduces any influence of reverse causation due to declined health. There is some suggestion of consistency in the literature that physical activity improves prostate cancer survival (23-26), however, the evidence is limited to few studies. 2.2.3 Physical activity and quality of life in prostate cancer survivors A separate search of the literature of the MEDLINE database using the PubMed up to August 16th, 2016 was conducted to review the literature regarding physical activity and QoL in prostate cancer survivors. The search included keywords and MeSH terms (exploded) as follows: ((“physical activity” OR “motor activity” OR exercise) AND (“prostate cancer” OR “prostatic cancer”) AND (“quality of life” OR QoL) AND (survivor* OR survivorship OR patients)) including Boolean operating terms. No date, language, or geographical restrictions were applied. Abstracts and unpublished results were not included in this review. This search rendered 192 results, of which 39 original 28 research articles and five review articles were found to be relevant and discussed below. Other studies were excluded due to irrelevant study population, physical activity or an exercise intervention was not the exposure or QoL was not an outcome. To date, the association between physical activity and QoL in prostate cancer survivors has been limited to few long-term epidemiological studies and exercise interventions. Five systematic review articles have attempted to summarize these associations in different capacities and have a common message; physical activity improves QoL (27-31). Two systematic reviews (28, 30) limited their inclusion criteria to prostate cancer survivors who underwent an exercise intervention and received ADT. Others restricted their inclusion criteria to exercise interventions at least four weeks long (27) or resistance exercise interventions (29). Two reviews (30, 31) were able to perform a meta-analysis. Teleni et al. (2015) included five intervention studies with common measurements of QoL and found statistically significant pooled standard mean differences (SMD) of 0.29 (95% CI: 0.10-0.49) related to health-related QoL and 0.36 (95% CI: 0.11-0.61) related to disease-specific QoL. While only a fraction of exercise interventions were included in this meta-analysis, heterogeneity was minimal (I2% = 0). Alternatively, Bourke et al., 2016 included seven exercise randomized controlled trials in their meta-analysis and found no significant improvements in cancer-specific QoL with SMD = 0.13 (95% CI: -0.08-0.34). However, in a sensitivity analysis excluding studies they deemed to be low quality (predominantly due to low adherence rates), they did find a statistically significant improvement in QoL (SMD = 0.33, 95% CI: 0.08-0.58), based on three studies (32-34). 29 A majority of the epidemiological studies that have assessed the association between physical activity and QoL involve measures of QoL at one point in time postdiagnosis. Ten cross-sectional studies (35-44) were identified from the literature, of which, seven (35, 37-39, 41, 42, 44) found some statistically significant associations between physical activity and QoL. However, cross-sectional studies may be susceptible to potential measurement error in physical activity exposure measurement. Measurement of physical activity is difficult to capture accurately through self-reported questionnaires at a single time-point. Seven studies measured physical activity through validated questionnaires (38-44), such as the International Physical Activity Questionnaire (IPAQ), and utilized cut-offs according to pre-defined guidelines (40, 42-44). One study only assessed vigorous physical activity with a single question (37), which may lead to measurement error. Methods used to collect information on physical activity and QoL in these studies was very heterogeneous including: different questionnaires, QoL outcomes and time periods of exposure and outcome assessment. For these reasons, it may be difficult to glean meaningful conclusions from these cross-sectional studies in addition to the recognized methodologic limitations for cross-sectional study evidence, including lack of temporality. Two prospective cohort studies (45, 46) and one retrospective cohort study (47) have been conducted on prostate cancer survivors, addressing the association between physical activity and QoL. All three cohort studies found some significant associations between physical activity and QoL. However, improvements in QoL were not consistent across studies. For example, from the Health Professionals Follow-up Study (45), the only significant improvements of QoL identified were found in the vitality domain (p < 30 0.001) with increased total physical activity (>10 hours/week compared to <1 hour/week). Further, with increased weekly walking time in hours, there appeared to be significant interactions with time since treatment in years (p = 0.03), Gleason score (p = 0.04) and dichotomized presence of comorbidities (p = 0.02). On a separate note, two studies (46, 47) obtained repeated measurements of QoL from participants. Both repeated measures studies measured health-related QoL using the Patient-Oriented Prostate Utility Scale (PORPUS) and found significant associations with meeting cancer prevention physical activity guidelines pre- and post-operatively (47) and with an increase in different types of past-year physical activity (46). Overall, there is clear evidence that physical activity can improve QoL in prostate cancer survivors consistently. There remain gaps in the evidence, however, from long-term, large scale follow-up studies on how QoL is influenced by different types of physical activity, different doses/volumes of activity, specifically, if meeting cancer prevention physical activity guidelines is associated with QoL and if these associations are influenced by any important effect modifiers. A large number of experimental exercise intervention studies have evaluated the relation between physical activity and different aspects of QoL in prostate cancer survivors (32-34, 48-71). Most exercise intervention studies examine treatment-related outcomes as a main objective and QoL as a secondary analysis, since the detrimental side effects of prostate cancer treatments are of concern. Regardless, most studies (32, 34, 48, 50, 54, 56, 59, 61, 62, 65-69) found statistically significant improvements in some aspect of QoL during or after the exercise interventions. Some possible explanations for these inconsistencies between studies include: duration of exercise interventions, diversity of 31 different exercise interventions undertaken and adequacy of study sample size to examine subgroup effects. Exercise interventions range in duration between eight weeks (54) to as long as 15 months (69). Prostate cancer survivors, on average, survive much longer than 15 months, especially those with low-grade or slow progressing disease. Therefore, these interventions do not address the effects of exercise on QoL long-term after diagnosis of prostate cancer. Further, 17 studies (33, 48, 49, 51, 53-55, 59-62, 64, 67-71) had less than 100 prostate cancer survivors included in their exercise interventions, and therefore, these studies may have had limited power to detect statistically significant differences in QoL and type II error may be present. This hypothesis is plausible since six studies with larger sample sizes (32, 34, 50, 56, 65, 66), did find significant results in QoL outcomes. Moreover, there is a chance these exercise interventions are too heterogeneous, making direct comparisons across studies difficult. There were exercise interventions that involved resistance exercise only interventions (50, 55, 61, 62, 70), aerobic exercise only interventions (32, 48, 54, 60, 71), a mixture of resistance and aerobic exercise interventions (33, 34, 59, 67, 69), home-based/telephone exercise interventions (52, 53, 56, 58, 63-65, 68), supervised and home-based exercise interventions (51, 66) or partnerbased resistance training (49). With this variety in exercise interventions examining QoL in prostate cancer survivors, caution in the interpretation of this literature as a whole is warranted. As Bourke et al. (2016) noted, most exercise randomized controlled trials included in their meta-analysis had a high risk of bias, due to poor intervention adherence (31). This issue is particularly important to recognize when examining QoL because there is a possibility that lack of exercise adherence may be due to reduced QoL and thereby biasing the results towards the null. 32 Twelve studies examined the relationship between physical activity or an exercise intervention and QoL in prostate cancer survivors using the Short Form-36 (SF-36), presented in Table 2.1. All of these studies were completed in developed countries including: the United States of American (USA) (37, 40, 49, 51, 52, 56, 65), Australia (59, 67), Korea (61) and Canada (63, 70). Ten of the 12 studies were randomized controlled trials (49, 51, 52, 56, 59, 61, 63, 65, 67, 70), while only half of the studies (37, 56, 59, 61, 65, 67) found statistically significant associations between physical activity or an exercise intervention and QoL. Therefore, the level of evidence surrounding studies utilizing the SF-36 as a measurement of QoL are less consistent and research is needed in this area to confirm if physical activity is associated with QoL in prostate cancer survivors. In summary, there is evidence that physical activity interventions benefit prostate cancer survivors’ QoL after diagnosis. However, long-term effects of QoL are not well documented and this research is needed to understand fully the role of physical activity and QoL in prostate cancer. 2.3 Part 2: Change in quality of life after prostate cancer The importance of QoL after prostate cancer diagnosis is evident with a growing body of literature that has elucidated the change of QoL after diagnosis in epidemiologic cohort studies. Here we summarize this literature by systematically searching the MEDLINE database up to August 16th, 2016 using the following search strategy: (("Quality of life" OR "QoL") AND ("prostate cancer" OR "prostatic cancer" OR "prostate neoplasm") AND ("change" OR "over time" OR "long term" OR development OR trajector*)). This search was limited to articles written in English, adult humans with 33 no date or country restrictions. This search returned 648 hits with 70 articles fitting the criteria of examining repeated measurements of QoL in a prostate cancer study population over time. Other studies were excluded due to only one measurement of QoL or irrelevant study population. The following relevant articles are reviewed below. 2.3.1 Long-term change in quality of life after prostate cancer: epidemiological evidence A plethora of research has been published on changes in QoL after prostate cancer diagnosis. Due to the vast extant literature to review, this section is focused on the most recent five years (2011-present) of research in the field. The search of the literature identified 33 cohort studies (72-104) examining QoL changes after prostate cancer diagnosis. Twenty of these studies (73-79, 86, 87, 89-97, 101, 104) examined a particular group of prostate cancer survivors on the same treatment regimen and their QoL over time. These treatment regimens included anything from patients who were treated with a radical prostatectomy (78, 89, 90, 96), or specifically patients treated with helical tomotherapy in a hypo-fractionated radiation schedule with long-term androgen suppressions (101). Given these studies were in different populations assessing QoL before and after a specific treatment regiments, it may be difficult to determine a hierarchy of treatments most beneficial for patients diagnosed with prostate cancer. Other studies have appraised the impacts of different prostate cancer treatments on QoL (81, 84, 85, 88, 98-100, 102), of which, most studies compared QoL before and after different treatment regimens with up to 10 years of follow-up (81). Included studies assessed differences between types of surgeries (88, 99, 102), radical prostatectomy versus hormone/radiation therapies (84, 98), doses/presence of ADT (85, 100) or all 34 treatments (including nerve-sparing radical prostatectomy, non-nerve-sparing radical prostatectomy, brachytherapy, external beam radiation therapy, ADT or active surveillance) (81) and QoL after diagnosis. Generally, the more invasive the treatments, the lower the score and the greater the decline in QoL observed. A few additional studies examined pre-defined strata of prostate cancer survivors and changes in QoL. Specifically, two studies (82, 83) considered age groups, while one of the same studies (82) also stratified on the number of co-morbidities to examine the impact on changes in QoL. Further, one study stratified prostate cancer survivors by incident diabetes, prevalent diabetes or no diabetes (103) to determine QoL differences; while others assessed predictors including illness uncertainty, anxiety and fear of progression (72) and patient-physician communication (80) on changes in QoL after diagnosis of prostate cancer. These studies investigated more than one average pattern of QoL after prostate cancer diagnosis on pre-specified criteria. It may be difficult to hypothesize potential effect modifiers through which prostate cancer survivors can be stratified by a priori to determine different QoL patterns after diagnosis. Therefore, this research of studying different patterns of QoL based on strata determined a priori is limited to prior knowledge of these potential subgroups, which is lacking in the literature. Repeated measurements of QoL ranged from two assessments (73, 79, 82, 83, 88, 91, 104) to 11 assessments (93) over different post-prostate cancer diagnosis periods. Eight studies examined disease-specific QoL (88, 90, 92-94, 99, 101, 102) compared to four studies that measured general QoL (74, 76, 78-80, 82, 85) or 15 studies that examined a combination of disease-specific and general QoL outcomes (72, 73, 75, 77, 81, 83, 84, 86, 87, 89, 91, 95-98, 100, 103, 104). Overall, there is abundant literature 35 examining QoL after prostate cancer. Several of the aforementioned studies, have different study populations, different objectives and different follow-up time frames, making direct comparisons across studies difficult. In general, these studies have found non-statistically significant changes in QoL (72, 73, 82, 89, 94), improvements in QoL over time (74, 102), declines in QoL over time (75-78, 83, 88, 91, 92, 96, 98, 100, 103, 104), initial declines with no recovery or partial recovery of QoL (81, 84, 86, 87, 90, 93, 99, 101), increases and decreases of QoL throughout the follow-up (80, 97) or a mixture of decreases and increases between the different domains of QoL examined (79, 85, 95). One of the most comprehensive studies examining QoL changes after prostate cancer diagnosis by Punnen et al. (2015) publishing results from the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE), which included a large sample of prostate cancer survivors (n=3,274) and is one of most definitive studies on this topic. The CaPSURE study population underwent different kinds of treatments and were followed for disease-specific and general QoL outcomes up to 10 years after diagnosis. Most QoL measures declined early with partial recovery 1-2 years after diagnosis, followed by a plateau in QoL. Surgery had the largest impact on sexual function/bother and urinary function. Radiation affected bowel function, while ADT had the largest impacts on physical function. Their results provided insights on long-term QoL from a general prostate cancer study population and are, therefore, of particular relevance for this thesis project. There were ten studies (76-79, 81, 83, 85, 89, 98, 103) examining changes in QoL after prostate cancer diagnosis using the SF-36 questionnaire from 2011-2016 (Table 2.2). Similar to the physical activity and QoL studies, all 10 studies were conducted in 36 developed countries including the USA (79, 81, 83, 98, 103), Japan (78, 89), The Netherlands (77) and Canada (76, 85). Change in general QoL (as measured by the SF36) after prostate cancer diagnosis either declined (77-79, 85, 98), remained consistent (76, 83, 89, 103) or partially recovered (81) in these studies. Although the SF-36 is a valid and widely used questionnaire in healthy and diseased populations (105), there are several factors that may contribute to differences in results seen among the 10 studies. Some of these potentially contributing factors involve the recruitment of different prostate cancer populations due to inclusion criteria, selection and participation rates, the length of follow-up and number of repeated measurements of QoL after prostate cancer diagnosis in each study. Literature is lacking on whether or not examining average QoL for a study population actually reflects the entire populations QoL experience after prostate cancer diagnosis. Furthermore, a majority of these studies, unlike Punnen et al. (81), examined QoL for short periods of time after prostate cancer diagnosis, given the relative survival rates for prostate cancer are well over 90% (22). There is a need to examine QoL patterns for long periods of time after diagnosis to capture the entire survival experience of prostate cancer survivors to promote wellbeing beyond cancer. 2.3.2 Growth modelling and prostate cancer Modelling change in QoL over time is important from a healthcare delivery perspective and for patients and caregivers to understand how QoL can be either maintained or improved after prostate cancer diagnosis. To date, previous research (as reviewed in section 2.3.1), examined only single patterns of QoL for the entire study population or predefined subgroups based on potential effect modifiers. It is unknown if 37 applying a single QoL trajectory for an entire population is appropriate or if the subgroups defined a priori are sensitive enough to capture QoL differences. There are inherent assumptions made when examining an average trajectory of QoL, including that every participant in a study may be represented by the average change or maintenance of QoL overtime post-diagnosis. Since prostate cancer does not affect a homogeneous group of individuals, it may be inaccurate to assume everyone follows an average trajectory of QoL. Therefore, attempting to distinguish separate QoL patterns into distinct subgroup trajectories is a knowledge gap that needs to be addressed. The modelling technique used to fulfill the third objective of my thesis is Groupbased Trajectory Modelling (GBTM), which has been used in several areas of literature including more recently medicine (106). Most studies that use this technique come from the other areas in the literature, but recently GBTM has gained attention in specific medical populations, such as cancer (106). To our knowledge, only two studies (107, 108) have considered these methods when determining different aspects of health-related QoL in a prostate cancer population. The first study (108), published in 2004, determined different recovery patterns in sexual function after radical prostatectomy. This study found four different trajectories over a five-year period. There were some limitations to this study, specifically in regards to generalizability, due to inclusion criteria and missing data. The second study (107), examined urinary function up to 12 months postprostatectomy and found three distinct subgroups in a prostate cancer study population. This study attempted to systematically characterize the subgroups by age, comorbidities, past or current smoking, PSA level and prostate volume. In both studies, there were low and high groups of QoL observed. Therefore, the rationale for applying these methods to 38 general physical and mental QoL has never been applied and is warranted in this population. 39 2.4 References 1. Young-McCaughan S. Potential for prostate cancer prevention through physical activity. World J Urol. 2012;30(2):167-79. 2. Namiranian N, Moradi-Lakeh M, Razavi-Ratki SK, Doayie M, Nojomi M. Risk factors of breast cancer in the Eastern Mediterranean Region: a systematic review and meta-analysis. Asian Pac J Cancer Prev. 2014;15(21):9535-41. 3. Mehta M, Shike M. Diet and physical activity in the prevention of colorectal cancer. J Natl Compr Cancer Netw. 2014;12(12):1721-6. 4. Keum N, Ju W, Lee DH, Ding EL, Hsieh CC, Goodman JE, et al. Leisure-time physical activity and endometrial cancer risk: dose-response meta-analysis of epidemiological studies. Int J Cancer. 2014;135(3):682-94. 5. Friedenreich CM, Thune I. A review of physical activity and prostate cancer risk. Cancer Causes Control. 2001;12(5):461-75. 6. Wilson KM, Giovannucci EL, Mucci LA. Lifestyle and dietary factors in the prevention of lethal prostate cancer. Asian J Androl. 2012;14(3):365-74. 7. Kruk J, Czerniak U. Physical activity and its relation to cancer risk: updating the evidence. Asian Pac J Cancer Prev. 2013;14(7):3993-4003. 8. Liu Y, Hu F, Li D, Wang F, Zhu L, Chen W, et al. Does physical activity reduce the risk of prostate cancer? A systematic review and meta-analysis. Eur Urol. 2011;60(5):1029-44. 9. Park SY, Haiman CA, Cheng I, Park SL, Wilkens LR, Kolonel LN, et al. Racial/ethnic differences in lifestyle-related factors and prostate cancer risk: the Multiethnic Cohort Study. Cancer Causes Control. 2015;26(10):1507-15. 10. Krishnadasan A, Kennedy N, Zhao Y, Morgenstern H, Ritz B. Nested casecontrol study of occupational physical activity and prostate cancer among workers using a job exposure matrix. Cancer Causes Control. 2008;19(1):107-14. 11. Parent ME, Rousseau MC, El-Zein M, Latreille B, Desy M, Siemiatycki J. Occupational and recreational physical activity during adult life and the risk of cancer among men. Cancer Epidemiol. 2011;35(2):151-9. 12. Grotta A, Bottai M, Adami HO, Adams SA, Akre O, Blair SN, et al. Physical activity and body mass index as predictors of prostate cancer risk. World J Urol. 2015. 13. Tao ZQ, Shi AM, Wang KX, Zhang WD. Epidemiology of prostate cancer: current status. Eur Rev Med Pharmacol Sci. 2015;19(5):805-12. 40 14. Doolan GW, Benke G, Giles GG, Severi G, Kauppinen T. A case control study investigating the effects of levels of physical activity at work as a risk factor for prostate cancer. Environ Health. 2014;13:64-069X-13-64. 15. Hallmarker U, James S, Michaelsson K, Arnlov J, Sandin F, Holmberg L. Cancer incidence in participants in a long-distance ski race (Vasaloppet, Sweden) compared to the background population. Eur J Cancer. 2015;51(4):558-68. 16. Heitkamp HC, Jelas I. Epidemiological evidence for preventing prostate cancer by physical activity. Wien Med Wochenschr. 2012;162(23-24):533-40. 17. Giovannucci EL, Liu Y, Leitzmann MF, Stampfer MJ, Willett WC. A prospective study of physical activity and incident and fatal prostate cancer. Arch Intern Med. 2005;165(9):1005-10. 18. Patel AV, Rodriguez C, Jacobs EJ, Solomon L, Thun MJ, Calle EE. Recreational physical activity and risk of prostate cancer in a large cohort of U.S. men. Cancer Epidemiol Biomarkers Prev. 2005;14(1):275-9. 19. Johnsen NF, Tjonneland A, Thomsen BL, Christensen J, Loft S, Friedenreich C, et al. Physical activity and risk of prostate cancer in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. Int J Cancer. 2009;125(4):902-8. 20. Czyz J, Szpak K, Madeja Z. The role of connexins in prostate cancer promotion and progression. Nat Rev Urol. 2012;9(5):274-82. 21. Filella X, Foj L. Emerging biomarkers in the detection and prognosis of prostate cancer. Clin Chem Lab Med. 2015;[Epub ahead of print]. 22. Canadian Cancer Society's Advisory Committee on Cancer S. Canadian Cancer Statistics 2015. Toronto, ON: Canadian Cancer Society, 2015. 23. Kenfield SA, Stampfer MJ, Giovannucci E, Chan JM. Physical activity and survival after prostate cancer diagnosis in the health professionals follow-up study. J Clin Oncol. 2011;29(6):726-32. 24. Friedenreich CM, Wang Q, Neilson HK, Kopciuk KA, McGregor SE, Courneya KS. Physical Activity and Survival After Prostate Cancer. Eur Urol. 2016;S03022838(15):01241-5. 25. Richman EL, Kenfield SA, Stampfer MJ, Paciorek A, Carroll PR, Chan JM. Physical activity after diagnosis and risk of prostate cancer progression: data from the cancer of the prostate strategic urologic research endeavor. Cancer Res. 2011;71(11):3889-95. 26. Bonn SE, Sjolander A, Lagerros YT, Wiklund F, Stattin P, Holmberg E, et al. Physical Activity and Survival among Men Diagnosed with Prostate Cancer. Cancer Epidemiol Biomarkers Prev. 2015;24(1):57-64. 41 27. Keogh JW, MacLeod RD. Body composition, physical fitness, functional performance, quality of life, and fatigue benefits of exercise for prostate cancer patients: a systematic review. J Pain Symptom Manage. 2012;43(1):96-110. 28. Chipperfield K, Brooker J, Fletcher J, Burney S. The impact of physical activity on psychosocial outcomes in men receiving androgen deprivation therapy for prostate cancer: a systematic review. Health Psychol. 2014;33(11):1288-97. 29. Hasenoehrl T, Keilani M, Sedghi Komanadj T, Mickel M, Margreiter M, Marhold M, et al. The effects of resistance exercise on physical performance and health-related quality of life in prostate cancer patients: a systematic review. Support Care Cancer. 2015;[Epub ahead of print]. 30. Teleni L, Chan RJ, Chan A, Isenring EA, Vela I, Inder WJ, et al. Exercise improves quality of life in androgen deprivation therapy-treated prostate cancer: systematic review of randomised controlled trials. Endocr Relat Cancer. 2016;23(2):10112. 31. Bourke L, Smith D, Steed L, Hooper R, Carter A, Catto J, et al. Exercise for Men with Prostate Cancer: A Systematic Review and Meta-analysis. Eur Urol. 2016;69(4):693-703. 32. Bourke L, Gilbert S, Hooper R, Steed LA, Joshi M, Catto JW, et al. Lifestyle changes for improving disease-specific quality of life in sedentary men on long-term androgen-deprivation therapy for advanced prostate cancer: a randomised controlled trial. Eur Urol. 2014;65(5):865-72. 33. Galvao DA, Taaffe DR, Spry N, Joseph D, Newton RU. Combined resistance and aerobic exercise program reverses muscle loss in men undergoing androgen suppression therapy for prostate cancer without bone metastases: a randomized controlled trial. J Clin Oncol. 2010;28(2):340-7. 34. Segal RJ, Reid RD, Courneya KS, Sigal RJ, Kenny GP, Prud'Homme DG, et al. Randomized controlled trial of resistance or aerobic exercise in men receiving radiation therapy for prostate cancer. J Clin Oncol. 2009;27(3):344-51. 35. Boisen S, Krageloh C, Shepherd D, Ryan C, Masters J, Osborne S, et al. A CrossSectional Comparison of Quality of Life between Physically Active and Under-active Older Men with Prostate Cancer. J Aging Phys Act. 2016;[Epub ahead of print]. 36. Gaskin CJ, Craike M, Mohebbi M, Salmon J, Courneya KS, Broadbent S, et al. Associations of objectively measured moderate-to-vigorous physical activity and sedentary behavior with quality of life and psychological well-being in prostate cancer survivors. Cancer Causes Control. 2016;27(9):1093-103. 37. Demark-Wahnefried W, Clipp EC, Morey MC, Pieper CF, Sloane R, Snyder DC, et al. Physical function and associations with diet and exercise: Results of a cross42 sectional survey among elders with breast or prostate cancer. Int J Behav Nutr Phys Act. 2004;1(1):16. 38. Dahn JR, Penedo FJ, Molton I, Lopez L, Schneiderman N, Antoni MH. Physical activity and sexual functioning after radiotherapy for prostate cancer: beneficial effects for patients undergoing external beam radiotherapy. Urology. 2005;65(5):953-8. 39. Humpel N, Iverson DC. Depression and quality of life in cancer survivors: is there a relationship with physical activity? Int J Behav Nutr Phys Act. 2007;4:65. 40. Blanchard CM, Courneya KS, Stein K. Cancer survivors' adherence to lifestyle behavior recommendations and associations with health-related quality of life: results from the American Cancer Society's SCS-II. J Clin Oncol. 2008;26(13):2198-204. 41. Keogh JW, Shepherd D, Krageloh CU, Ryan C, Masters J, Shepherd G, et al. Predictors of physical activity and quality of life in New Zealand prostate cancer survivors undergoing androgen-deprivation therapy. N Z Med J. 2010;123(1325):20-9. 42. Chipperfield K, Fletcher J, Millar J, Brooker J, Smith R, Frydenberg M, et al. Predictors of depression, anxiety and quality of life in patients with prostate cancer receiving androgen deprivation therapy. Psychooncology. 2013;22(10):2169-76. 43. Chipperfield K, Fletcher J, Millar J, Brooker J, Smith R, Frydenberg M, et al. Factors associated with adherence to physical activity guidelines in patients with prostate cancer. Psychooncology. 2013;22(11):2478-86. 44. Galvao DA, Newton RU, Gardiner RA, Girgis A, Lepore SJ, Stiller A, et al. Compliance to exercise-oncology guidelines in prostate cancer survivors and associations with psychological distress, unmet supportive care needs, and quality of life. Psychooncology. 2015;[Epub ahead of print]. 45. Phillips SM, Stampfer MJ, Chan JM, Giovannucci EL, Kenfield SA. Physical activity, sedentary behavior, and health-related quality of life in prostate cancer survivors in the health professionals follow-up study. J Cancer Surviv. 2015;[Epub ahead of print]. 46. Santa Mina D, Matthew AG, Trachtenberg J, Tomlinson G, Guglietti CL, Alibhai SM, et al. Physical activity and quality of life after radical prostatectomy. Can Urol Assoc J. 2010;4(3):180-6. 47. Santa Mina D, Guglietti C, Alibhai S, Matthew A, Kalnin R, Ahmad N, et al. The effect of meeting physical activity guidelines for cancer survivors on quality of life following radical prostatectomy for prostate cancer. J Cancer Surviv. 2014(1932-2267 (Electronic)). 48. Hojan K, Kwiatkowska-Borowczyk E, Leporowska E, Gorecki M, OzgaMajchrzak O, Milecki T, et al. Physical exercise for functional capacity, blood immune function, fatigue and quality of life in high-risk prostate cancer patients during 43 radiotherapy. A prospective, randomised clinical study. Eur J Phys Rehabil Med. 2016;[Epub ahead of print]. 49. Winters-Stone KM, Lyons KS, Dobek J, Dieckmann NF, Bennett JA, Nail L, et al. Benefits of partnered strength training for prostate cancer survivors and spouses: results from a randomized controlled trial of the Exercising Together project. J Cancer Surviv. 2016;10(4):633-44. 50. Segal RJ, Reid RD, Courneya KS, Malone SC, Parliament MB, Scott CG, et al. Resistance exercise in men receiving androgen deprivation therapy for prostate cancer. J Clin Oncol. 2003;21(9):1653-9. 51. Daubenmier JJ, Weidner G, Marlin R, Crutchfield L, Dunn-Emke S, Chi C, et al. Lifestyle and health-related quality of life of men with prostate cancer managed with active surveillance. Urology. 2006;67(1):125-30. 52. Demark-Wahnefried W, Clipp EC, Morey MC, Pieper CF, Sloane R, Snyder DC, et al. Lifestyle intervention development study to improve physical function in older adults with cancer: outcomes from Project LEAD. J Clin Oncol. 2006;24(21):3465-73. 53. Culos-Reed SN, Robinson JL, Lau H, O'Connor K, Keats MR. Benefits of a physical activity intervention for men with prostate cancer. J Sport Exerc Psychol. 2007;29(1):118-27. 54. Monga U, Garber SL, Thornby J, Vallbona C, Kerrigan AJ, Monga TN, et al. Exercise prevents fatigue and improves quality of life in prostate cancer patients undergoing radiotherapy. Arch Phys Med Rehabil. 2007;88(11):1416-22. 55. Hansen PA, Dechet CB, Porucznik CA, LaStayo PC. Comparing eccentric resistance exercise in prostate cancer survivors on and off hormone therapy: a pilot study. Pm r. 2009;1(11):1019-24. 56. Mosher CE, Sloane R, Morey MC, Snyder DC, Cohen HJ, Miller PE, et al. Associations between lifestyle factors and quality of life among older long-term breast, prostate, and colorectal cancer survivors. Cancer. 2009;115(17):4001-9. 57. Morey MC, Snyder DC, Sloane R, Cohen HJ, Peterson B, Hartman TJ, et al. Effects of home-based diet and exercise on functional outcomes among older, overweight long-term cancer survivors: RENEW: a randomized controlled trial. JAMA. 2009;301(18):1883-91. 58. Culos-Reed SN, Robinson JW, Lau H, Stephenson L, Keats M, Norris S, et al. Physical activity for men receiving androgen deprivation therapy for prostate cancer: benefits from a 16-week intervention. Support Care Cancer. 2010;18(5):591-9. 59. Galvao DA, Taaffe DR, Spry N, Joseph D, Newton RU. Acute versus chronic exposure to androgen suppression for prostate cancer: impact on the exercise response. J Urol. 2011;186(4):1291-7. 44 60. Truong PT, Gaul CA, McDonald RE, Petersen RB, Jones SO, Alexander AS, et al. Prospective evaluation of a 12-week walking exercise program and its effect on fatigue in prostate cancer patients undergoing radical external beam radiotherapy. Am J Clin Oncol. 2011;34(4):350-5. 61. Park SW, Kim TN, Nam JK, Ha HK, Shin DG, Lee W, et al. Recovery of overall exercise ability, quality of life, and continence after 12-week combined exercise intervention in elderly patients who underwent radical prostatectomy: a randomized controlled study. Urology. 2012;80(2):299-305. 62. Hanson ED, Sheaff AK, Sood S, Ma L, Francis JD, Goldberg AP, et al. Strength training induces muscle hypertrophy and functional gains in black prostate cancer patients despite androgen deprivation therapy. J Gerontol A Biol Sci Med Sci. 2013;68(4):490-8. 63. McGowan EL, North S, Courneya KS. Randomized controlled trial of a behavior change intervention to increase physical activity and quality of life in prostate cancer survivors. Ann Behav Med. 2013;46(3):382-93. 64. Santa Mina D, Alibhai SM, Matthew AG, Guglietti CL, Pirbaglou M, Trachtenberg J, et al. A randomized trial of aerobic versus resistance exercise in prostate cancer survivors. J Aging Phys Act. 2013;21(4):455-78. 65. Winger JG, Mosher CE, Rand KL, Morey MC, Snyder DC, Demark-Wahnefried W. Diet and exercise intervention adherence and health-related outcomes among older long-term breast, prostate, and colorectal cancer survivors. Ann Behav Med. 2014;48(2):235-45. 66. Buffart LM, Newton RU, Chinapaw MJ, Taaffe DR, Spry NA, Denham JW, et al. The effect, moderators, and mediators of resistance and aerobic exercise on health-related quality of life in older long-term survivors of prostate cancer. Cancer. 2015;[Epub ahead of print]. 67. Cormie P, Galvao DA, Spry N, Joseph D, Chee R, Taaffe DR, et al. Can supervised exercise prevent treatment toxicity in patients with prostate cancer initiating androgen-deprivation therapy: a randomised controlled trial. BJU Int. 2015;115(2):25666. 68. O'Neill RF, Haseen F, Murray LJ, O'Sullivan JM, Cantwell MM. A randomised controlled trial to evaluate the efficacy of a 6-month dietary and physical activity intervention for patients receiving androgen deprivation therapy for prostate cancer. J Cancer Surviv. 2015;9(3):431-40. 69. Zopf EM, Bloch W, Machtens S, Zumbe J, Rubben H, Marschner S, et al. Effects of a 15-Month Supervised Exercise Program on Physical and Psychological Outcomes in Prostate Cancer Patients Following Prostatectomy: The ProRehab Study. Integr Cancer Ther. 2015;[Epub ahead of print]. 45 70. Norris MK, Bell GJ, North S, Courneya KS. Effects of resistance training frequency on physical functioning and quality of life in prostate cancer survivors: a pilot randomized controlled trial. Prostate Cancer Prostatic Dis. 2015;18(3):281-7. 71. Jones LW, Hornsby WE, Freedland SJ, Lane A, West MJ, Moul JW, et al. Effects of nonlinear aerobic training on erectile dysfunction and cardiovascular function following radical prostatectomy for clinically localized prostate cancer. Eur Urol. 2014;65(5):852-5. 72. Parker PA, Davis JW, Latini DM, Baum G, Wang X, Ward JF, et al. Relationship between illness uncertainty, anxiety, fear of progression and quality of life in men with favourable-risk prostate cancer undergoing active surveillance. BJU Int. 2016;117(3):469-77. 73. Patabendi Bandarage VR, Billah B, Millar JL, Evans S. Prospective evaluation of patient-reported quality of life outcomes after external beam radiation treatment for prostate cancer in Victoria: A cohort study by the Victorian Prostate Cancer Registry. J Med Imaging Radiat Oncol. 2016;60(3):420-7. 74. Rucinska M, Kieszkowska-Grudny A, Nawrocki S. SHARP hypofractionated stereotactic radiotherapy is well tolerated in prostate cancer : Toxicity and quality of life assessment. Strahlenther Onkol. 2016;192(7):449-57. 75. Dowrick AS, Wootten AC, Howard N, Peters JS, Murphy DG. A prospective study of the short-term quality-of-life outcomes of patients undergoing transperineal prostate biopsy. BJU Int. 2016;118(1):60-7. 76. Siddiqui KM, Billia M, Arifin A, Li F, Violette P, Chin JL. Pathologic, Oncologic and Functional Outcomes of a Prospective Registry of Salvage High Intensity Focused Ultrasound Ablation for Radio-Recurrent Prostate. J Urol. 2016. 77. Peters M, Maenhout M, van der Voort van Zyp JR, Moerland MA, Moman MR, Steuten LM, et al. Focal salvage iodine-125 brachytherapy for prostate cancer recurrences after primary radiotherapy: a retrospective study regarding toxicity, biochemical outcome and quality of life. Radiother Oncol. 2014;112(1):77-82. 78. Miyake M, Tanaka N, Asakawa I, Tatsumi Y, Nakai Y, Anai S, et al. Changes in lower urinary tract symptoms and quality of life after salvage radiotherapy for biochemical recurrence of prostate cancer. Radiother Oncol. 2015;115(3):321-6. 79. Holliday EB, Dieckmann NF, McDonald TL, Hung AY, Thomas CR, Jr., Wood LJ. Relationship between fatigue, sleep quality and inflammatory cytokines during external beam radiation therapy for prostate cancer: A prospective study. Radiother Oncol. 2016;118(1):105-11. 80. Ernstmann N, Weissbach L, Herden J, Winter N, Ansmann L. Patient-physiciancommunication and health related quality of life of localized prostate cancer patients 46 undergoing radical prostatectomy - a longitudinal multilevel analysis. BJU Int. 2016; [Epub ahead of print]. 81. Punnen S, Cowan JE, Chan JM, Carroll PR, Cooperberg MR. Long-term healthrelated quality of life after primary treatment for localized prostate cancer: results from the CaPSURE registry. Eur Urol. 2015;68(4):600-8. 82. Herkommer K, Schmid SC, Schulwitz H, Dinkel A, Klorek T, Hofer C, et al. Health-related quality of life after radical prostatectomy depends on patient's age but not on comorbidities. Urol Oncol. 2015;33(6):266.e1-7. 83. Hampson LA, Cowan JE, Zhao S, Carroll PR, Cooperberg MR. Impact of age on quality-of-life outcomes after treatment for localized prostate cancer. Eur Urol. 2015;68(3):480-6. 84. Eisemann N, Nolte S, Schnoor M, Katalinic A, Rohde V, Waldmann A. The ProCaSP study: quality of life outcomes of prostate cancer patients after radiotherapy or radical prostatectomy in a cohort study. BMC Urol. 2015;15:28. 85. Alibhai SM, Breunis H, Timilshina N, Naglie G, Tannock I, Krahn M, et al. Long-term impact of androgen-deprivation therapy on physical function and quality of life. Cancer. 2015;121(14):2350-7. 86. Yamamoto S, Fujii Y, Masuda H, Urakami S, Saito K, Kozuka T, et al. Longitudinal change in health-related quality of life after intensity-modulated radiation monotherapy for clinically localized prostate cancer. Qual LIfe Res. 2014;23(5):1641-50. 87. Selli C, Bjartell A, Burgos J, Somerville M, Palacios JM, Benjamin L, et al. Burden of Illness in Prostate Cancer Patients with a Low-to-Moderate Risk of Progression: A One-Year, Pan-European Observational Study. Prostate Cancer. 2014;2014:472949. 88. Pinkawa M, Klotz J, Djukic V, Petz D, Holy R, Eble MJ. Transurethral resection of the prostate after radiotherapy for prostate cancer: impact on quality of life. Int J Urol. 2014;21(9):899-903. 89. Namiki S, Kaiho Y, Mitsuzuka K, Saito H, Yamada S, Nakagawa H, et al. Longterm quality of life after radical prostatectomy: 8-year longitudinal study in Japan. Int J Urol. 2014;21(12):1220-6. 90. Matthew AG, Alibhai SM, Davidson T, Currie KL, Jiang H, Krahn M, et al. Health-related quality of life following radical prostatectomy: long-term outcomes. Qual LIfe Res. 2014;23(8):2309-17. 91. Majewski W, Tabor K, Prokop E, Kulik R. Quality of life in patients with prostate cancer treated with radical image-guided radiotherapy. Contemp Oncol (Pozn). 2014;18(4):285-9. 47 92. Luo HC, Cheng LP, Cheng HH, Fu ZC, Liao SG, Li DS, et al. Long-term quality of life outcomes in patients with locally advanced prostate cancer after intensitymodulated radiotherapy combined with androgen deprivation. Med Oncol. 2014;31(6):991. 93. Katz AJ, Kang J. Quality of Life and Toxicity after SBRT for Organ-Confined Prostate Cancer, a 7-Year Study. Front Oncol. 2014;4:301. 94. van Gysen KL, Kneebone AB, Guo L, Vaux KJ, Lazzaro EM, Eade TN. Healthrelated quality of life using intensity-modulated radiation therapy for post-prostatectomy radiotherapy. J Med Imaging Radiat Oncol. 2013;57(1):89-96. 95. Jereczek-Fossa BA, Santoro L, Zerini D, Fodor C, Vischioni B, Dispinzieri M, et al. Image guided hypofractionated radiotherapy and quality of life for localized prostate cancer: prospective longitudinal study in 337 patients. J Urol. 2013;189(6):2099-103. 96. Jakobsson L, Fransson P. Patient Reported Outcome Measure (PROM) of Quality of Life After Prostatectomy - Results from a 5-Year Study. Open Nurs J. 2013;7:165-73. 97. Aluwini S, van Rooij P, Hoogeman M, Kirkels W, Kolkman-Deurloo IK, Bangma C. Stereotactic body radiotherapy with a focal boost to the MRI-visible tumor as monotherapy for low- and intermediate-risk prostate cancer: early results. Radiat Oncol. 2013;8:84. 98. Brassell SA, Elsamanoudi SI, Cullen J, Williams ME, McLeod DG. Healthrelated quality of life for men with prostate cancer--an evaluation of outcomes 12-24 months after treatment. Urol Oncol. 2013;31(8):1504-10. 99. Willis DL, Gonzalgo ML, Brotzman M, Feng Z, Trock B, Su LM. Comparison of outcomes between pure laparoscopic vs robot-assisted laparoscopic radical prostatectomy: a study of comparative effectiveness based upon validated quality of life outcomes. BJU Int. 2012;109(6):898-905. 100. Sevilla C, Maliski SL, Kwan L, Connor SE, Litwin MS. Long-term quality of life in disadvantaged men with prostate cancer on androgen-deprivation therapy. Prostate Cancer Prostatic Dis. 2012;15(3):237-43. 101. Pervez N, Krauze AV, Yee D, Parliament M, Mihai A, Ghosh S, et al. Quality-oflife outcomes in high-risk prostate cancer patients treated with helical tomotherapy in a hypofractionated radiation schedule with long-term androgen suppression. Curr Oncol. 2012;19(3):e201-10. 102. Lin YH, Yu TJ, Lin VC, Yang MS, Kao CC. Changes in quality of life among prostate cancer patients after surgery. Cancer Nurs. 2012;35(6):476-82. 103. Thong MS, van de Poll-Franse L, Hoffman RM, Albertsen PC, Hamilton AS, Stanford JL, et al. Diabetes mellitus and health-related quality of life in prostate cancer: 5-year results from the Prostate Cancer Outcomes Study. BJU Int. 2011;107(8):1223-31. 48 104. Berge V, Baco E, Dahl AA, Karlsen SJ. Health-related quality of life after salvage high-intensity focused ultrasound (HIFU) treatment for locally radiorecurrent prostate cancer. Int J Urol. 2011;18(9):646-51. 105. Hamoen EH, De Rooij M, Witjes JA, Barentsz JO, Rovers MM. Measuring health-related quality of life in men with prostate cancer: A systematic review of the most used questionnaires and their validity. Urol Oncol. 2015;33(2):69.e19-28. 106. Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol. 2010;6:109-38. 107. Anderson CB, Kaufman MR, Dietrich MS, Barocas DA, Chang SS, Cookson MS, et al. Recovery of urinary function after radical prostatectomy: identification of trajectory cluster groups. J Urol. 2012;187(4):1346-51. 108. Legler JM, Davis WW, Potosky AL, Hoffman RM. Latent variable modelling of recovery trajectories: sexual function following radical prostatectomy. Stat Med. 2004;23(18):2875-93. 49 2.5 Tables Table 2.1. Physical activity and SF-36 quality of life studies in prostate cancer populations. Author, year, country Physical activity assessment methods QoL assessment methods 688 One item to assess vigorous physical activity (average continuous vigorous exercise for at least 20 minutes, three or more times per week). SF-36 physical function subscale modified to only include vigorous activity not moderate. Positive association between vigorous exercise and physical function score for breast and prostate cancer patients (p<0.0001) Age, race, gender, time since diagnosis, and concurrent health behaviours 93 12-month intervention included a plantbased vegan diet, three hours per week of moderate exercise and one hour a day of stress management. SF-36: eight domains, PCS and MCS scores. Intervention vs controls on PCS scores: (time effect p=0.45, group effect p=0.18, time x group p=0.28) MCS scores: (time effect p=0.82, time effect p=0.01, time x group p=0.50) - Study Design and population Sample size DemarkWahnefried et al., 2004, USA (1) Cross-sectional: participants were 60+ years old, no more than 18 months beyond diagnosis of early stage (I-II) breast and prostate cancer identified through cancer registries, private practices and self-referrals. Daubenmier et al., 2006, USA (2) Randomized controlled trial: prostate cancer patients undergoing active surveillance, with PSA levels of 4-10 ng/mL and Gleason scores <7 randomly assigned to a lifestyle intervention or usual care control group. 50 Adjustment for confounding Analytic methods/Results DemarkWahnefried et al., 2006, USA (3) Randomized controlled trial: prostate cancer and breast cancer survivors aged >65 years old and within 18 months of diagnosis, ascertained from 13 hospitals in North Carolina, able to participate in intervention and without limiting comorbidities. 182 Treatment arm received telephone counseling and tailored print materials aimed at increased exercise and improved overall diet, as well as 12 bimonthly 20-30 min sessions over a six-month period. Functional Assessment of Cancer TherapyProstate/Breas t/ General (FACTP/B/G) and SF-36: physical function subscale. Intervention vs control on physical function score at six months: p=0.23 and 12 months: p=0.48 Blanchard et al., 2008, USA (4) Cross-sectional: participants were from the American Cancer Society’s Study of Cancer Survivors-II, aged 18+ years, diagnosed with local, regional or distant cancer, a resident of the state at diagnosis, and diagnosed between 2-10 years before sampling. 36,372 Godin LeisureTime Exercise Questionnaire, dichotomized as did or did not meet American College of Sports Medicine (ACSM) recommendations. SF-36: eight domain scores and a global health composite score. Met physical activity guidelines: mean=53.6, SD=10 Did not meet PA guidelines: mean=49.8, SD=10.9 (d=0.3) Race, stage, marital status, education, age and total number of comorbidities Mosher et al., 2009, USA (5) Randomized controlled trial: prostate, breast and colorectal cancer survivors, >65 years, >5 753 The intervention consisted of two 45-60 min telephone surveys SF-36: eight domains scores, PCS and MCS Correlations between moderate-to-vigorous physical activity min/week and: Age, level of education, number of comorbidities 51 Baseline value of the outcome, baseline martial status, smoking status, age at diagnosis, sex, Intervention vs control race, education on FACT-G QOL at attainment, and six months: p=0.38 at social 12 months: p=0.97 desirability score. years post-diagnosis, approval from physician, able to speak and write in English, no medical conditions, residence within the community, overweight but not obese, and not adhering to physical activity guidelines randomized to a homebased diet and exercise intervention of tailored mailed materials and telephone counseling. Galvao et al., Randomized controlled 2011, trial: prostate cancer Australia (6) patients aged 55-84 years, undergoing androgen suppression therapy for non-bone metastatic prostate cancer completed a progressive resistance and cardiovascular exercise program for 12 weeks. 50 (two days to three scores. weeks apart) based on the Community Healthy Activities Models Program and 24-hour diet recalls adapted for telephone administered use. Physical QoL: (p<0.0001) Mental QoL: (p=0.94) Pain: (p=0.02) Health perceptions: (p=0.01) Physical functioning: (p<0.0001) Role-physical: (p=0.053) Vitality: (p=0.0001) Mental health: (p=0.19) Social functioning: (p=0.025) Role-emotional: (p=0.65) Participants performed combined progressive resistance (last approx. 40 mins) and aerobic training (15-20 mins) twice weekly for 12 weeks. In chronic androgen suppression therapy patients, there was an increased QoL in physical function (p=0.031), general health (p=0.012) and vitality (p=0.027) and the physical health composite score (p=0.016) 52 SF-36: eight domains scores. Age and baseline values (non-specified) Park et al., 2012, Korea (7) Randomized controlled trial: prostate cancer patients undergoing laparoscopic radical prostatectomy, aged >65 years, localized disease, Eastern Cooperative Oncology Group performance score of 0 or 1, written informed consent and randomized to intervention or control groups. 66 Intervention included resistance, pelvic and Kegel exercises performed two times/week for 60 mins, conducted three weeks postoperatively for 12 weeks. SF-36: PCS and MCS scores. Pre to post- exercise intervention vs control groups and PCS score: (p<0.05) MCS score: (p<0.05) - McGowan et al., 2013, Canada (8) Randomized controlled trial: random sample of prostate cancer survivors 18+ years old, diagnosed between 2005-2009, were randomized to standard physical activity recommendations, selfadministered implementation intention, or a telephone-assisted implementation intention. 423 All intervention groups received a fact sheet on physical activity guidelines for Americans as well as information created by the researchers, the implementation intention group also included Specific, Measureable, Achievable, Realistic and FACT-P and SF-36: PCS and MCS scores. No significant improvements in FACT-P QoL, generic QoL PCS or MCS in any intervention groups. Baseline value of the outcome, age, disease stage, surgery, radiation, hormone therapy, chemotherapy, and disease status. 53 Time-bound (SMART) goal setting principles and additionally telephone calls once a week in the third group. Winger et al., 2014, USA (9) Randomized controlled trial: prostate, breast and colorectal cancer survivors, >65 years, >5 years post-diagnosis, approval from physician, able to speak and write in English, no medical conditions, residence within the community, overweight but not obese, and not adhering to physical activity guidelines randomized to a homebased diet and exercise intervention of tailored mailed materials and telephone counseling. 641 Participants reported exercise and dietary behaviors during sessions 2-15 of the 15 telephone counseling sessions, a selfreported behavior log (reporting frequency and duration of exercise in past week), during a year-long intervention period. SF-36: MCS scores. Correlations indicated Strength exercise was associated with the MCS (p<0.01) - Cormie et al., 2015, Australia Randomized controlled trial: prostate cancer patients scheduled to 63 Intervention was three months, with twice weekly European Organization for Research Mean adjusted difference between exercise group and Baseline values 54 (10) receive ADT were randomly assigned to a supervised exercise program involving aerobic and resistance exercise sessions commenced within 10 days of their first ADT injection or usual care. Norris et al., 2015, Canada (11) Randomized controlled trial: patients with nonmetastatic prostate cancer, between the ages of 18-80 years of age and were not currently performing resistance training >= 2 days a week. 30 supervised exercise sessions lasting approximately one hour. There were moderate-vigorous aerobic and resistance exercise components with appropriate warm up and cool down periods. and Treatment of Cancer Quality of Life Questionnaire -Prostate 25 (EORTC QLQ-PR25) and SF-36: eight domains scores. usual care revealed social functioning (3.8, 1.1-6.5, p=0.015), mental health (3.8, 1.1-6.5, p=0.006), mental health composite (3.6, 0.5-6.6, p=0.022), sexual function (15.2, 1.9-28.4, p=0.028) were significantly different Pilot- two-arm trial comparing three versus two days/week of resistance training for 12 weeks. SF-36: PCS and MCS scores. Trend favoring three days over two days (p<0.1) and not statistically significant. Trends toward improvements of PCS scores with three days of resistance training and MCS scores for two days of resistance training. 55 Baseline value of the outcome, current hormone therapy and previous treatments WintersStone et al., 2016, USA (12) Randomized controlled trial: patients who had received treatment for prostate cancer, not currently undergoing radiation or chemotherapy, 60+ years of age, residing with a spouse willing to participate, not currently exercising two or more times/week and physician clearance. 64 couples Intervention compared Exercising Together to usual care. Exercising Together includes one hour resistance training sessions modified to implement spousal support and motivation done twice a week for six months. 56 SF-36: physical function, vitality subscales, PCS and MCS scores. For prostate cancer survivors, adjusted group difference in mean slope of PCS (p=0.99), MCS (p=0.39), physical function (0.72) or vitality (p=0.98) did not statistically significantly differ over the six months. Age, time since diagnosis Table 2.2. Prostate cancer cohort studies examining changes in QoL post-diagnosis using the SF-36 published between 20112016. Author, year, country Study characteristics/ Inclusion criteria & Number of participants Thong et al., 2011, USA (13) 1,811 prostate cancer patients with nonmetastatic cancer enrolled in Prostate Cancer Outcomes Study. Diabetes Brassell et al., 2013, USA (14) 595 prostate cancer patients enrolled from 2003-2010 before prostate biopsy and followed for QoL outcomes. - Determinants/ exposures Number of QoL measure(s) and tool Results Adjustment variables Four measurements at: six (baseline), 12, 24 and 60 months after initial diagnosis using SF-36 and modified questionnaire for urinary, sexual and bowel function. Men with prevalent diabetes had the poorest QoL scores and non-diabetic best scores, independent of treatment. Men with prevalent diabetes had the lowest urinary control and sexual function scores over time while men without diabetes had the highest scores. Men with incident diabetes had intermediate scores. Pre-treatment urinary sexual and bowel functioning, age, marital status, educational status, annual income, employment status, cancer stage, baseline PSA, baseline Gleason score. Eight measurements at: baseline, 3, 6, 9, 12, 18, 24 and 30 months using Expanded Prostate Cancer Index Composite (EPIC) and SF- Radical prostatectomy led to the greatest decline in urinary function. Bowel function significantly worsened with addition of hormone therapy to external beam radiotherapy. Sexual bother and function declined in all active treatment options. SF-36 domains were not affected by radical Age, race, smoking history, education, marital status, income, PSA at diagnosis, T stage, and Gleason score. 57 36 questionnaires. prostatectomy, but radiotherapy and radiotherapy plus hormone therapy declined SF-36 domains. Radiotherapy showed worsened results by the 12-24 time-period for all domains. Namiki et al., 2014, Japan (15) 91 prostate cancer patients undergoing radical prostatectomy followed for QoL outcomes in prospective cohort study conducted between 2002-2005. - Seven measurements at: baseline, 1, 2, 3, 4, 5 and 10 years’ postsurgery using SF-36 and University of California, Los AngelesProstate Cancer Index (UCLAPCI). Mental and role composite summary scores remained stable throughout the follow-up period and other domains remained constant after the first year period where sexual function declined. Peters et al., 2014, The Netherlands (16) 20 patients treated with focal salvage brachytherapy with biochemical failure greater than two years after primary treatment, unilateral biopsy proven recurrence after systematic transrectal biopsies of both prostate - Four measurements at: baseline, 1, 6 months and 3 years using SF36, European Organization for Research and Treatment of Cancer Core After three years, physical functioning was statistically significantly different (decreased), while no other SF36 scores showed any changes. Global health, general physical discomfort, fatigue and urinary symptoms significantly changed at different timepoints during follow-up. 58 - lobes, no extra-capsular extension or seminal vesicle involvement and no ADT at time of salvage were recruited between 2009-2012. Quality of Life Questionnaire (EORTC QLQC30), and EORTC QLQPR25. Alibhai et 87 patients on ADT, 87 al., 2015, prostate cancer controls, Canada (17) and 87 healthy controls (matched on age, education and baseline function). ADT Eight measurements at: baseline, 3, 6,12, 18, 24, 30 and 36 months using the SF-36. PCS score declined in ADT group (p<0.001), but remained stable. MCS scores remained stable over time in all three cohorts with no significant differences between groups. Hampson et al., 2015, USA (18) 5,362 prostate cancer patients who underwent treatments related to prostate cancer from the CaPSURE prospective cohort study. Age Two measurements at: baseline and 2 years using SF-36 and UCLA-PCI. Older men had lower mean QoL scores than younger men in all domains. Sexual function, sexual bother and urinary function showed declines over time. More men <60 years vs >70 years, experienced declines in urinary function and sexual bother at two years. Miyake et al., 2015, Japan (19) 81 patients who underwent radical prostatectomy and salvage radiotherapy for biochemical recurrence for prostate cancer. - Seven measurements: prior to, during and immediately after salvage radiotherapy and 1, 3, 6, 12 PCS, role physical problems, role emotional problems and general health domains scores were statistically significantly different (lower) immediately after completion of treatment compared to baseline (p<0.05). 59 Age, education, body mass index, and Charlson comorbidity index score. - months post treatment using the SF-36. Punnen et al., 2015, USA (20) 3,294 underwent different forms of prostate cancer treatment were captured from the CaPSURE registry prospective cohort study. Treatments: nerve-sparing radical prostatectomy (NSRP), nonNSRP, brachytherapy, external beam radiotherapy, ADT, active surveillance. Four measurements at: baseline, 2, 5 and 10 years using SF-36 and UCLA-PCI. Most QoL measures resulted in early declines with partial recovery the first 1-2 years then a plateau. Surgery had largest impact on sexual function and bother, and urinary function. Radiation had the largest impacts on bowel function, ADT on physical function. Age, year of treatment, number of comorbidities, medical insurance status, cancer progression risk at diagnosis, receipt of any secondary treatments Holliday et al., 2016, USA (21) 28 men undergoing radiotherapy for early stage prostate cancer. Fatigue Two measurements: prior to treatment and at the end of treatment using SF-36. Vitality and social functioning significantly decreased over treatment. - 60 Siddiqui et 81 patients accrued from al., 2016, the phase II study, with Canada (22) stage cT1c-T2 cancer, who received radionucleotide bone scan and computed tomography of the abdomen and pelvis to confirm cancer and high-intensity focused ultrasound (HIFU) therapy. - Four measurements at: pre-HIFU, 45, 90 and 180 months postHIFU using SF36. 61 Global SF-36 score remained constant across all time-points. - Chapter Three: ASSOCIATION OF POST-DIAGNOSIS PHYSICAL ACTIVITY AND CHANGE IN PRE-DIAGNOSIS PHYSICAL ACTIVITY WITH QUALITY OF LIFE IN PROSTATE CANCER SURVIVORS 3.1 Preamble This manuscript addresses objectives 1 and 2 of the thesis through cross-sectional analyses of different types of physical activity, change in physical activity and QoL. It was submitted to a peer-review journal, Cancer Epidemiology, Biomarkers and Prevention on June 1, 2016. This work was also accepted as a poster presentation at the following conference: Farris MS, Courneya KS, Kopciuk KA, McGregor SE, Friedenreich CM. The impact and change in lifetime and post-diagnosis physical activity on quality of life in prostate cancer survivors. Charbonneau 2nd Annual Cancer Research Symposium, Calgary, AB; February 5, 2016. Running title: Physical activity and QoL in prostate cancer survivors Keywords: physical activity, quality of life, prostate cancer, guideline adherence, cancer survivorship Financial support: Dr. Friedenreich held career awards from Alberta Innovates, Health Solutions and Alberta Cancer Foundation, Dr. Courneya was supported by the Canada 62 Research Chairs Program and Dr. McGregor by the Alberta Heritage Foundation for Medical Research. 63 3.2 Abstract Introduction: This prospective study examined the associations between post-diagnosis physical activity and change from pre-diagnosis physical activity with quality of life (QoL) in prostate cancer survivors. Materials and Methods: Prostate cancer survivors (n=830) who participated in a case-control study with invasive stage ≥II disease were followed up to 2007 to capture QoL outcomes. At baseline and three time points post-diagnosis (2000-2007), interviews/questionnaires were used to collect data on physical activity, general QoL measured by the SF-36 and other treatment/lifestyle factors. Multivariable linear regression was used to test the relation between post-diagnosis physical activity and QoL as well as the change in physical activity over the diagnostic period and QoL. Results: Both total and recreational physical activities were positively associated with physical QoL. Further, when comparing changes in physical activity levels from pre- to post-diagnosis, men who consistently met physical activity guidelines had significantly higher physical (β = 6.01, 95% CI: 4.15-7.86) and mental (β = 2.32, 95% CI: 0.29-4.34) QoL scores compared to those who did not meet guidelines pre- or post-diagnosis. Further, those who adopted and met guidelines had increased QoL, while those who relapsed experienced decreased QoL. Conclusions: Post-diagnosis recreational physical activity is associated with better physical QoL in prostate cancer survivors. Moreover, prostate cancer survivors who maintain or adopt physical 64 activity after diagnosis report substantially higher QoL than men who never exercised or stopped exercising after diagnosis. Impact: Future intervention studies should focus on achieving and maintaining adherence to physical activity guidelines post-diagnosis in prostate cancer survivors. 65 3.3 Introduction Prostate cancer is the second most common cancer in men worldwide (1, 2). Early detection and improved treatment options have increased prevalence and decreased the mortality rate associated with prostate cancer (3). Due to the increase in five-year survival rates, currently approximately 93% in Canada (4), men diagnosed with low-grade (stage I-II) prostate cancer die, more likely than not from other causes, particularly after age 75 (5). In addition, Prostate Specific Antigen (PSA) testing often over detect low grade cancers, which for the most part, are indolent for long periods of time (6). In turn, overtreatment of potentially non-life threatening disease has burdened prostate cancer survivors and the healthcare system (7, 8). Consequently, priorities in prostate cancer control involve reducing the burdens of living beyond cancer diagnosis, improving health, overall functioning, care, quality of life (QoL) (9) and addressing the specific needs of prostate cancer survivors (10). Prostate cancer survivors who undergo surgery (11, 12), receive treatments such as androgen deprivation therapy (13, 14) or radiation therapy (11, 15, 16) often experience worse QoL. Physical activity is an inexpensive, modifiable behaviour that has been shown to improve QoL after diagnosis and may be associated with survival (17-20). Observational epidemiologic research (21-32) and randomized controlled exercise intervention trials (33-40) have shown that physical activity can improve QoL in prostate cancer survivors. No study to date, however, has examined the associations between changes in physical activity across the diagnostic period and QoL in prostate cancer survivors. Further, only one study (27) explicitly examined different types of physical activity including household and occupational physical activity and QoL. Moreover, previous observational research has had methodologic limitations associated with 66 inadequate control for confounding and an inability to examine effect modification because of the small sample sizes. This study was primarily designed to examine the associations of physical activity and survival in prostate cancer survivors (17). In this current analysis, we examined the role of both total post-diagnosis physical activity and each type of activity including recreational, occupational and household activity and changes in physical activity over the diagnostic period on physical and mental QoL in prostate cancer survivors. This study is the first with repeated measurements of physical activity and QoL with a large sample size that can explore multiple subgroup analyses. 3.4 Materials and methods 3.4.1 Study design A prospective cohort of prostate cancer survivors that originated from a population-based case-control study in Alberta, Canada, was conducted between November 1997 and December 2000 (41). All cases from the original case-control study were re-contacted and followed up for post-diagnosis measurements and mortality outcomes to 2014. Details regarding this study have been published previously (17). Briefly, prostate cancer cases were histologically confirmed, invasive cases of stage II or greater prostate cancer and identified from the Alberta Cancer Registry, a population-based cancer registry. The prostate cancer survivors were <80 years of age, English speaking with no previous cancer diagnosis except for non-melanoma skin cancer. Permission to contact prostate cancer cases for in-person interviews was obtained through the referring urologist. All surviving cases from the case-control study were re-contacted by telephone for voluntary recruitment and provided written informed consent for the prostate 67 cancer cohort follow-up study (2% rate of refusal). The cohort study began in 2000, to permit follow-up and further data collection from the cases. In addition, referring urologists were recontacted for additional missing medical data. This study received ethics approval from the Alberta Cancer Research Ethics Board and the Conjoint Health Research Ethics Board at the University of Calgary. 3.4.2 Data collection Through in-person interviews, conducted in the original case-control study, participants reported on their personal health history, prostate cancer screening history, prostate conditions, surgery history, family history of cancer, lifetime physical activity patterns prior to diagnosis, dietary intake during the reference year (year prior to cancer diagnosis), lifetime alcohol consumption history, smoking habits, current social support and motivation for lifestyle change, demographic characteristics and adult height and weight at each decade from age 20-60 years. Anthropometric measurements were measured directly at the interview. Additionally, as a part of the cohort follow-up study, survivors were interviewed to assess their physical activity behaviour within the first two years after diagnosis. At the cohort follow-up interview, self-reported questionnaires on current QoL were completed by participants. At a second and third follow-ups, approximately two years apart, self-administered questionnaires asking participants about their past two years of physical activity behaviour as well as their current QoL were mailed to participants to complete. Further, medical chart abstractions and vital status assessments were completed throughout the follow-up to capture updated medical information, treatments and outcomes. 68 3.4.3 Physical activity assessment Physical activity was assessed using the Lifetime Total Physical Activity Questionnaire (LTPAQ) and the Past Year Total Physical Activity Questionnaire (PYTPAQ) previously tested for reliability and validity (42, 43). Both the LTPAQ and the PYTPAQ record all types of physical activity (occupational, household and recreational) as well as all parameters of activity (frequency, duration, intensity) from childhood to the time of diagnosis. The LTPAQ was interview-administered as part of the baseline case-control study and the results from that study have been published (41). The participants were re-interviewed between 2000-2002 using the LTPAQ restricted to their activity done since the first interview which had occurred between 1997 and 2000. Two additional assessments of physical activity were obtained through the selfadministered PYTPAQ between 2002-2004 and 2004-2007. Hence, continuous levels of physical activity throughout the participant’s life up until 2007 (when active data collection ceased) were obtained. All physical activities (pre- and post-diagnosis) were assigned a Metabolic Equivalent (MET) value defined as a ratio of the associated metabolic rate for a specific activity as compared with the resting metabolic rate. These MET values were based on the Compendium of Physical Activities (44) and used with the reported frequency and duration of activity to derive the MET-hours/week/year for each type of activity. Moderate-to-vigorous recreational activity was defined as any activity >3 METs according to the Compendium of Physical Activities categorization (44). Moderate-to-vigorous total and recreational physical activity (hour/week/year) were derived to examine activity duration similar to general cancer prevention physical activity guidelines (45, 46). 69 To assess the change in physical activity over the diagnosis period, a change score was derived using the difference between pre-diagnosis lifetime and average post-diagnosis physical activity. The average post-diagnosis physical activity variables contained up to three postdiagnosis time-points. For those with missing assessments, because of non-response or death, the average physical activity levels were derived from available complete assessments. Additionally, adherence to physical activity guidelines for cancer prevention and survivorship (150 minutes of moderate or 75 minutes of vigorous recreational physical activity) (45, 46) over the diagnosis period was assessed and behaviours were categorized as follows: consistently not meeting preand post-diagnosis (non-exercisers); not meeting pre-diagnosis to meeting guidelines postdiagnosis (adopters); meeting pre-diagnosis to not meeting guidelines post-diagnosis (relapsers); or consistently meeting guidelines pre- and post-diagnosis (maintainers). Active data collection ended in 2007; therefore we utilized the physical activity cancer prevention and survivorship guidelines specific to aerobic physical activity. 3.4.4 Quality of life assessment QoL assessments at up to three follow-ups were obtained through a single questionnaire, the RAND 36-Item Short Form Health Survey (SF-36) version 1.0. The SF-36 has been used in several settings, is one of the most widely used QoL questionnaires (47), and was validated in colorectal cancer survivors (48). The SF-36 addresses eight domains of health status that have been summarized into a physical component summary (PCS) score and a mental component summary (MCS) score. Both scores range from 0-100; higher scores indicate better QoL. Clinically meaningful QoL score differences reflect >5 point changes (49). 70 3.4.5 Statistical analysis Descriptive statistics were performed on all variables to report study characteristics. Next, multivariable least squares linear regression was used in all analyses (all statistical linear regression assumptions, including distributional, homogeneity of variances, were verified). A priori variables deemed to be predictors of QoL were forced into the models: prostatectomy surgery (yes, no), post-diagnosis Charlson co-morbidity score (50) (0, 1, 2, ≥3), aggressive (Gleason score >8 or stage >II cancer (51)) or non-aggressive disease, type of treatment (none, hormone, radiation therapy, both hormone and radiation therapy), and age at diagnosis (years). The aforementioned variables were first tested for interaction. Further, we tested the following covariates for potential confounding first, with univariate analyses (p<0.2) and secondly through backward elimination: pre-diagnosis lifetime physical activity (MET-hour/week/year) to adjust for baseline differences, nonlinearity of age at diagnosis, family history of cancer (yes, no), PSA score at diagnosis (≤4, >4 and ≤10, >10 and ≤20, >20), smoking status at diagnosis (never, former, current), total alcohol consumption at diagnosis (grams/year), region of residence (urban, rural), ethnicity (Caucasian, other), education (less than high school, high school, trade school, other non-university degree, university degree), marital status (married, other), weight at diagnosis (kg), Body Mass Index (BMI) (kg/m2), daily caloric intake at diagnosis (kcal/day), lifestyle change post-diagnosis (yes, no), family support (yes, no), friend support (yes, no) and ever joined a support group (yes, no). In addition, recreational, occupational and household physical activity types and other physical activity intensities were mutually adjusted to control for the effects of one type/intensity on another. All quantitative variables were centered at the median for ease of interpretation. 71 To investigate the associations between post-diagnosis physical activity and QoL, the first follow-up data (collected 2000-2002) for both physical activity and QoL were used. To determine changes in pre- and post-diagnosis physical activity and the influence on QoL, lifetime pre-diagnosis physical activity and average (over all available time-points) post-diagnosis physical activity data were used to derive change scores and adherence groups. Additionally, average (over all available time-points) post-diagnosis QoL scores were derived for these analyses. Sensitivity analyses were performed by removing survivors diagnosed with stage III/IV and IV metastatic cancer to account for disease severity differences. In addition, cases who died prior to completing all three post-diagnosis follow-ups were excluded in the physical activity change over the diagnostic period analysis. Estimated rates of change of QoL scores, per unit of physical activity values and their corresponding 95% confidence intervals (CI) were reported in multivariable adjusted models. A priori hypotheses were assessed with a significance level of p<0.05, using the statistical package Stata v.13 (College Town, Texas). 3.5 Results 3.5.1 Descriptive statistics There were initially 830 men who survived up until the first follow-up time-point (20002002), eight of those participants were missing first follow-up QoL data, five participants were missing first follow-up lifetime physical activity data and therefore, the analytic sample was 817 for these analyses. The majority of the study participants were diagnosed with stage II cancer, married and had at least one co-morbidity post-diagnosis (Table 3.1). The mean age of the participants at start of follow-up was 67.3 years. The sample had a post-diagnosis first follow-up 72 total physical activity median of 78.7 MET-hour/week/year, however, the recreational physical activity median was only 11.1 MET-hour/week/year. Further, their first follow-up mean PCS score was 40.8 (Standard Deviation (SD)=12.4) and MCS score was 50.8 (SD=11.8), which were approximately symmetrically distributed. 3.5.2 Post-diagnosis physical activity and quality of life at first follow-up Final multivariable adjusted models found statistically significant associations between three post-diagnosis types of physical activity in relation to first follow-up PCS score (Table 3.2). Only one type of physical activity was significantly associated with MCS score. Recreational physical activity revealed the strongest associations (relative to the other types) with an estimated increase of 0.16 (95% CI: 0.12-0.20, p<0.001) in PCS score and 0.05 (95% CI: 0.01-0.09, p=0.024) for MCS score per MET hour/week/year increase in physical activity volume. While other physical activity types were not significantly associated with MCS score, associations appeared to be consistent in direction with estimated PCS score, except household physical activity, which was slightly negative. Further, sensitivity analyses excluding stage III/IV and IV cancers (n=130) did not appear to substantially influence these results. Analyses restricted to duration (hour/week/year) of moderate-to-vigorous total and recreational physical activity are presented in Table 3.2. Recreational moderate-to-vigorous activity was positively associated with both PCS and MCS scores with estimated increases of 0.68 (95% CI: 0.50-0.85, p<0.001) and 0.22 (95% CI: 0.03-0.40, p=0.021) per hour of recreational activity respectively. Total moderate-to-vigorous physical activity was also positively associated with PCS score with an estimated increase of 0.09 (95% CI: 0.05-0.13, p<0.001) but not with MCS score ( = -0.01 (95% CI: -0.05-0.04, p=0.76)). Interestingly, the 73 sensitivity analysis excluding stage III/IV and IV cancers increased the recreational moderate-tovigorous physical activity associations, but not those for total activity. 3.5.3 Change in physical activity over the diagnostic period and quality of life The analysis evaluating changes of physical activity volume (MET hour/week/year) between lifetime (pre-diagnosis) and post-diagnosis activity on average post-diagnosis QoL for up to three follow-ups, revealed similar results to the post-diagnosis physical activity analysis (Table 3.3). While recreational physical activity change was consistently associated with PCS score ( = 0.15, 95% CI: 0.11-0.19, p<0.001 and MCS score = 0.06, 95% CI: 0.02-0.11, p=0.006), total and occupational activity change were only associated with PCS score. Similarly, household activity did not impact the PCS or MCS scores significantly. Excluding stage III/IV and IV cancers or those who died before completing all three follow-ups, did not materially change results. Table 3.3 also presents physical activity change focused on moderate-to-vigorous total and recreational activity, which indicated associations with recreational activity and both PCS and MCS QoL scores with estimated increases of 0.72 (95% CI: 0.54-0.90, p<0.001) and 0.35 (95% CI: 0.15-0.55, p=0.001) per hour of activity performed. Total physical activity was once again statistically significantly associated with PCS score ( = 0.11, 95% CI, 0.07-0.16, p<0.001) but not MCS score ( = 0.04, 95% CI: -0.01-0.09, p=0.15). Nonetheless, in the sensitivity analysis excluding participants diagnosed with stage III/IV and IV cancers, recreational activity and PCS QoL score associations were slightly heightened, however, this finding was not seen in the total activity analysis, nor in the analysis excluding those who did not complete all three follow-ups. 74 Finally, the change in behaviour over the diagnostic period according to the cancer prevention physical activity guidelines (46) relative to PCS and MCS scores is presented. This analysis provides more definitive physical activity change according to cancer prevention guidelines in existence at the time (2007). First, the median pre-diagnosis and post-diagnosis physical activity levels for each cancer prevention guideline adherence group were examined (Figure 3.1). Maintainers and adopters increased physical activity levels, while relapsers and non-exercisers reduced physical activity levels from pre-diagnosis to post-diagnosis. In multivariable analyses (Table 3.4), prostate cancer survivors who were characterized as adopters, experienced an increase of PCS score by 4.80 (95% CI: 2.82-6.78, p<0.001) and MCS score by 2.26 (95% CI: 0.09-4.43, p=0.041) relative to non-exercisers. Further, maintainers showed the strongest results with increased PCS score of 6.01 (95% CI: 4.15-7.86, p<0.001) and MCS score of 2.32 (95% CI: 0.29-4.34, p=0.025) relative to non-exercisers. Interestingly, relapsers had non-statistically significant worse scores. Slight attenuations were shown in the analysis excluding stage III/IV and IV cancers and participants that did not complete all followups due to death, specifically in MCS scores for adopters and maintainers, however, point estimates and CI’s remained fairly unchanged. Evidence of statistical interactions were not present in these analyses, therefore, estimated PCS and MCS scores for the entire study population were reported. 3.6 Discussion This study found that post-diagnosis physical activity was associated with PCS scores in prostate cancer survivors. Specifically, the positive association between recreational physical activity and PCS score was consistent across analyses, however, the magnitude of the association 75 of physical activity on PCS score were relatively small (49). Total, occupational and household activities were not associated with QoL at the first follow-up. It is also interesting to note that mean first follow-up PCS scores were 10 points lower than MCS scores. Adherence to cancer prevention physical activity guidelines resulted in five and six-point increases in average PCS scores for prostate cancer survivors who were adopters and maintainers relative to nonexercisers. These results were clinically relevant. Both relapsers and non-exercisers had reduced physical activity levels post-diagnosis (Figure 3.1). Future interventions should consider strategies to improve adherence to current exercise and cancer guidelines prior to and after prostate cancer diagnosis. Our results illustrated a positive association between physical activity and QoL in prostate cancer survivors, similar to other cohort studies (26, 27). However, the studies differed in their methods for physical activity and QoL assessment therefore, comparisons between studies are somewhat difficult. For example, in the Health Professionals Follow-up Study (24), prostate cancer-specific QoL was measured and physical activity was significantly associated with vitality/hormonal functioning scores. On the other hand, four exercise intervention studies (52-55) measured generic QoL using PCS and MCS scores. One of these interventions (55) found clinically relevant effects with PCS score (similar to our study), however, three studies (52, 53, 55) found significant associations with MCS score. On another note, to our knowledge, our study is the first to measure change in lifetime physical activity prior to and after prostate cancer diagnosis in relation to QoL. The lack of association in total activity may be attributed to the increased duration of occupational and household physical activities in the average day for those who were employed post-diagnosis. Retired men who had an occupational score of zero (Tables 3.2 & 3.3), would 76 not have contributed to these associations and therefore, may have spent more time performing other types of activities. We also considered the possibility that cancer survivors with progressing disease or diminishing health would experience a decrease in their QoL and in turn, exercise less. To minimize this bias, we performed sensitivity analyses excluding men with stage III/IV and IV cancer at diagnosis, as well as those who died before completing all post-diagnosis assessments. Regardless, we continued to observe positive associations between recreational physical activity and PCS scores. Reverse causation was of less concern with the change in physical activity analysis. Men who were maintainers and adopters had the highest QoL scores. Further, slightly stronger effects were observed for analyses restricted to moderate-vigorous recreational physical activity. Nonetheless, given that a majority of our sample had stage II prostate cancer at diagnosis and only a small portion of the sample were confirmed to be stage IV (n=55), we cannot be certain if disease severity did not have an impact on the physical activity and QoL relationship. There are multiple proposed mechanisms by which physical activity has been linked to physical and mental QoL. First, in prostate cancer exercise intervention studies, fatigue (56, 57), cardiorespiratory fitness (57), walking speed (56), upper body strength (56) and lower body functional performance (58) has been identified as possible mediators between exercise intervention and different aspects of QoL. Further, marital status, time since diagnosis and use of bisphosphonates (58) have been described as moderators. In epidemiological studies, physical activity has been shown to reduce adiposity (27, 59) and inflammation (60), which can account for increased physical functioning. Further, physical activity increases secretion of insulin (61), enhances sex hormone regulation (59) and in turn, decreases insulin resistance and QoL (62). Specifically, improved sex hormone regulation may lead to decreased erectile dysfunction and 77 improved martial relations (62-64). Therefore, modifiable behaviours like physical activity can reduce these burdens. With respect to mental QoL, physical activity has been associated with increased regulation of neurotransmitters associated with mood (65), improved self-esteem, selfefficacy and reduced anxiety (66); all aspects of psychological functioning. More work is needed to confirm these mechanisms specific to changes in recreational physical activity over the course of the diagnostic period in prostate cancer survivors. There are inherent limitations that need to be considered when interpreting these results. First, this was an observational study and therefore, subject to bias. Measurement error in physical activity assessment may exist, which could attenuate the results. However, these data were collected first, by in-person interviews and additionally by self-reported questionnaires, therefore, measurement bias was minimized. Also, measurement and adjustment for sedentary behaviour in all analyses would strengthen these results by further isolating physical activity levels. Further, in the physical activity change analysis, we used a change score and are limited in our generalizability of interpretations to individual participants. Change scores are not related to individual baseline values, although we did adjust for pre-diagnosis physical activity in general, we cannot differentiate between participant specific changes in physical activity levels. However, the analysis examining adherence to physical activity cancer prevention guidelines addresses these concerns by evaluating physical activity against a standard. Another limitation arises from missing data due to mortality. However, sensitivity analyses excluding those who died before the third follow-up did not materially change the results. Finally, we restricted our study sample to ≥ stage II cancers aged < 80 years of age, therefore we can only generalize our results to these prostate cancer populations. 78 There are implicit strengths that should also be noted. First, this study included a large population-based sample of prostate cancer survivors with a follow-up of 10 years for QoL outcomes. Four repeated physical activity and three QoL measurements were collected. The assessment of physical activity was comprehensive since all parameters of physical activity (i.e., frequency, duration and intensity) of three separate physical activity types (i.e., recreational, occupational and household) were captured. Further, using the PCS and MCS scores, rather than individual QoL domains reduced the issue of multiple testing while providing summary scores of physical and mental QoL. Moreover, the pre- and post-diagnosis physical activity measurements enabled us to examine associations with change in physical activity over the diagnostic period in prostate cancer survivors. Finally, several potential effect modifiers and confounding variables were considered for this analysis to reduce the chance of spurious results and to identify potential sub-groups within the population. 3.7 Conclusion This study provided insights regarding the types of physical activity that impact QoL in prostate cancer survivors. Objective measurements of physical activity by accelerometry or fitness trackers are warranted in this population to reduce any suspicions of measurement error and further confirm the benefits of recreational physical activity on QoL. Health professionals should emphasize the importance of recreational physical activity to their patients, as a modifiable and a possible behavioural change for most prostate cancer survivors. Interpretations may need to be tailored to pre-diagnosis physical activity level to ensure prostate cancer survivors can achieve recommended physical activity thresholds. Future research should focus 79 on sedentary behaviour to determine efficacy of increased physical activity post-diagnosis in a phase III trial or health promotion program. 3.8 Manuscript acknowledgements Study coordination was done by Aleata Ryhorchuk and Sana Fakih. Interviewers for this study were Jodi Parrotta, Linda Davison, Pearl Cooke, Nicole Slot, Carol-Anne Zawalykut, Catherine Munro. Data entry was done by Carla Quesnel and chart abstraction by Yvonne LeBlanc and Sarah MacLaughlin. Dr. Steven J. Angyalfi provided clinical expertise regarding prostate cancer treatment and care. 80 3.9 References 1. Siegel R, DeSantis C, Virgo K, Stein K, Mariotto A, Smith T, et al. Cancer treatment and survivorship statistics, 2012. CA Cancer J Clin. 2012;62(4):220-41. 2. Torre LA, Siegel RL, Ward EM, Jemal A. Global Cancer Incidence and Mortality Rates and Trends-An Update. Cancer Epidemiol Biomarkers Prev. 2016;25(1):16-27. 3. Aziz NM. Cancer survivorship research: challenge and opportunity. J Nutr. 2002;132:3494S-503S. 4. Canadian Cancer Society's Advisory Committee on Cancer S. Canadian Cancer Statistics 2015. Toronto, ON: Canadian Cancer Society, 2015. 5. Moyer VA, Force USPST. Screening for prostate cancer: U.S. Preventive Services Task Force recommendation statement. Ann Intern Med. 2012;157(2):120-34. 6. Draisma G, Boer R, Otto SJ, van der Cruijsen IW, Damhuis RAM, Schröder FH, et al. Lead Times and Overdetection Due to Prostate-Specific Antigen Screening: Estimates From the European Randomized Study of Screening for Prostate Cancer. J Natl Cancer Inst. 2003;95(12):868-78. 7. Chodak GW, Thisted RA, Gerber GS, Johansson JE, Adolfsson J, Jones GW, et al. Results of conservative management of clinically localized prostate cancer. N Engl J Med. 1994;330(4):242-8. 8. Albertsen PC, Fryback DG, Storer BE, Kolon TF, Fine J. Long-term survival among men with conservatively treated localized prostate cancer. JAMA. 1995;274(8):626-31. 9. Hall IJ, Lee Smith J. Evolution of a CDC Public Health Research Agenda for Low-Risk Prostate Cancer. Am J Prev Med. 2015;49(6 Suppl 5):S483-8. 10. Richards M, Corner J, Maher J. The National Cancer Survivorship Initiative: new and emerging evidence on the ongoing needs of cancer survivors. Br J Cancer. 2011;105 Suppl 1:S14. 11. Punnen S, Cowan JE, Chan JM, Carroll PR, Cooperberg MR. Long-term health-related quality of life after primary treatment for localized prostate cancer: results from the CaPSURE registry. Eur Urol. 2015;68(4):600-8. 12. Eisemann N, Nolte S, Schnoor M, Katalinic A, Rohde V, Waldmann A. The ProCaSP study: quality of life outcomes of prostate cancer patients after radiotherapy or radical prostatectomy in a cohort study. BMC Urol. 2015;15:28. 13. Dacal K, Sereika SM, Greenspan SL. Quality of life in prostate cancer patients taking androgen deprivation therapy. J Am Geriatr Soc. 2006;54(1):85-90. 81 14. Alibhai SM, Breunis H, Timilshina N, Naglie G, Tannock I, Krahn M, et al. Long-term impact of androgen-deprivation therapy on physical function and quality of life. Cancer. 2015;121(14):2350-7. 15. Brassell SA, Elsamanoudi SI, Cullen J, Williams ME, McLeod DG. Health-related quality of life for men with prostate cancer--an evaluation of outcomes 12-24 months after treatment. Urol Oncol. 2013;31(8):1504-10. 16. Yamamoto S, Fujii Y, Masuda H, Urakami S, Saito K, Kozuka T, et al. Longitudinal change in health-related quality of life after intensity-modulated radiation monotherapy for clinically localized prostate cancer. Qual LIfe Res. 2014;23(5):1641-50. 17. Friedenreich CM, Wang Q, Neilson HK, Kopciuk KA, McGregor SE, Courneya KS. Physical Activity and Survival After Prostate Cancer. Eur Urol. 2016;S0302-2838(15):01241-5. 18. Richman EL, Kenfield SA, Stampfer MJ, Paciorek A, Carroll PR, Chan JM. Physical activity after diagnosis and risk of prostate cancer progression: data from the cancer of the prostate strategic urologic research endeavor. Cancer Res. 2011;71(11):3889-95. 19. Kenfield SA, Stampfer MJ, Giovannucci E, Chan JM. Physical activity and survival after prostate cancer diagnosis in the health professionals follow-up study. J Clin Oncol. 2011;29(6):726-32. 20. Bonn SE, Sjolander A, Lagerros YT, Wiklund F, Stattin P, Holmberg E, et al. Physical Activity and Survival among Men Diagnosed with Prostate Cancer. Cancer Epidemiol Biomarkers Prev. 2015;24(1):57-64. 21. Galvao DA, Newton RU, Gardiner RA, Girgis A, Lepore SJ, Stiller A, et al. Compliance to exercise-oncology guidelines in prostate cancer survivors and associations with psychological distress, unmet supportive care needs, and quality of life. Psychooncology. 2015;[Epub ahead of print]. 22. Blanchard CM, Courneya KS, Stein K. Cancer survivors' adherence to lifestyle behavior recommendations and associations with health-related quality of life: results from the American Cancer Society's SCS-II. J Clin Oncol. 2008;26(13):2198-204. 23. Mosher CE, Sloane R, Morey MC, Snyder DC, Cohen HJ, Miller PE, et al. Associations between lifestyle factors and quality of life among older long-term breast, prostate, and colorectal cancer survivors. Cancer. 2009;115(17):4001-9. 24. Phillips SM, Stampfer MJ, Chan JM, Giovannucci EL, Kenfield SA. Physical activity, sedentary behavior, and health-related quality of life in prostate cancer survivors in the health professionals follow-up study. J Cancer Surviv. 2015;[Epub ahead of print]. 25. Humpel N, Iverson DC. Depression and quality of life in cancer survivors: is there a relationship with physical activity? Int J Behav Nutr Phys Act. 2007;4:65. 82 26. Santa Mina D, Guglietti C, Alibhai S, Matthew A, Kalnin R, Ahmad N, et al. The effect of meeting physical activity guidelines for cancer survivors on quality of life following radical prostatectomy for prostate cancer. J Cancer Surviv. 2014(1932-2267 (Electronic)). 27. Santa Mina D, Matthew AG, Trachtenberg J, Tomlinson G, Guglietti CL, Alibhai SM, et al. Physical activity and quality of life after radical prostatectomy. Can Urol Assoc J. 2010;4(3):180-6. 28. Demark-Wahnefried W, Clipp EC, Morey MC, Pieper CF, Sloane R, Snyder DC, et al. Physical function and associations with diet and exercise: Results of a cross-sectional survey among elders with breast or prostate cancer. Int J Behav Nutr Phys Act. 2004;1(1):16. 29. Dahn JR, Penedo FJ, Molton I, Lopez L, Schneiderman N, Antoni MH. Physical activity and sexual functioning after radiotherapy for prostate cancer: beneficial effects for patients undergoing external beam radiotherapy. Urology. 2005;65(5):953-8. 30. Keogh JW, Shepherd D, Krageloh CU, Ryan C, Masters J, Shepherd G, et al. Predictors of physical activity and quality of life in New Zealand prostate cancer survivors undergoing androgen-deprivation therapy. N Z Med J. 2010;123(1325):20-9. 31. Chipperfield K, Fletcher J, Millar J, Brooker J, Smith R, Frydenberg M, et al. Predictors of depression, anxiety and quality of life in patients with prostate cancer receiving androgen deprivation therapy. Psychooncology. 2013;22(10):2169-76. 32. Chipperfield K, Fletcher J, Millar J, Brooker J, Smith R, Frydenberg M, et al. Factors associated with adherence to physical activity guidelines in patients with prostate cancer. Psychooncology. 2013;22(11):2478-86. 33. Teleni L, Chan RJ, Chan A, Isenring EA, Vela I, Inder WJ, et al. Exercise improves quality of life in androgen deprivation therapy-treated prostate cancer: systematic review of randomised controlled trials. Endocr Relat Cancer. 2016;23(2):101-12. 34. Segal RJ, Reid RD, Courneya KS, Malone SC, Parliament MB, Scott CG, et al. Resistance exercise in men receiving androgen deprivation therapy for prostate cancer. J Clin Oncol. 2003;21(9):1653-9. 35. Segal RJ, Reid RD, Courneya KS, Sigal RJ, Kenny GP, Prud'Homme DG, et al. Randomized controlled trial of resistance or aerobic exercise in men receiving radiation therapy for prostate cancer. J Clin Oncol. 2009;27(3):344-51. 36. Galvao DA, Taaffe DR, Spry N, Joseph D, Newton RU. Acute versus chronic exposure to androgen suppression for prostate cancer: impact on the exercise response. J Urol. 2011;186(4):1291-7. 37. Galvao DA, Spry N, Denham J, Taaffe DR, Cormie P, Joseph D, et al. A multicentre year-long randomised controlled trial of exercise training targeting physical functioning in men 83 with prostate cancer previously treated with androgen suppression and radiation from TROG 03.04 RADAR. Eur Urol. 2014;65(5):856-64. 38. Hasenoehrl T, Keilani M, Sedghi Komanadj T, Mickel M, Margreiter M, Marhold M, et al. The effects of resistance exercise on physical performance and health-related quality of life in prostate cancer patients: a systematic review. Support Care Cancer. 2015;[Epub ahead of print]. 39. Chipperfield K, Brooker J, Fletcher J, Burney S. The impact of physical activity on psychosocial outcomes in men receiving androgen deprivation therapy for prostate cancer: a systematic review. Health Psychol. 2014;33(11):1288-97. 40. Keogh JW, MacLeod RD. Body composition, physical fitness, functional performance, quality of life, and fatigue benefits of exercise for prostate cancer patients: a systematic review. J Pain Symptom Manage. 2012;43(1):96-110. 41. Friedenreich CM, McGregor SE, Courneya KS, Angyalfi SJ, Elliott FG. Case-control study of lifetime total physical activity and prostate cancer risk. Am J Epidemiol. 2004;159(8):740-9. 42. Friedenreich CM, Courneya KS, Bryant HE. The lifetime total physical activity questionnaire: development and reliability. Med Sci Sports Exerc. 1998;30(2):266-74. 43. Friedenreich CM, Courneya KS, Neilson HK, Matthews CE, Willis G, Irwin M, et al. Reliability and validity of the Past Year Total Physical Activity Questionnaire. Am J Epidemiol. 2006;163(10):959-70. 44. Ainsworth BE, Haskell WL, Leon AS, Jacobs DR, Jr., Montoye HJ, Sallis JF, et al. Compendium of physical activities: classification of energy costs of human physical activities. Med Sci Sports Exerc. 1993;25(1):71-80. 45. Kushi LH, Doyle C, McCullough M, Rock CL, Demark-Wahnefried W, Bandera EV, et al. American Cancer Society Guidelines on nutrition and physical activity for cancer prevention: reducing the risk of cancer with healthy food choices and physical activity. CA Cancer J Clin. 2012;62(1):30-67. 46. World Health O. Recommended population levels of physical activity for health. Geneva: World Health Organization; 2010. 47. Pucci GC, Rech CR, Fermino RC, Reis RS. Association between physical activity and quality of life in adults. R I Med J. 2012;46(1):166-79. 48. Antonescu I, Carli F, Mayo NE, Feldman LS. Validation of the SF-36 as a measure of postoperative recovery after colorectal surgery. Surg Endos. 2014;28(11):3168-78. 49. Jaeschke R, Singer J, Guyatt GH. Measurement of health status. Ascertaining the minimal clinically important difference. Control Clin Trials. 1989;10(4):407-15. 84 50. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-83. 51. McGregor SE, Courneya KS, Kopciuk KA, Tosevski C, Friedenreich CM. Case-control study of lifetime alcohol intake and prostate cancer risk. Cancer Causes Control. 2013;24(3):45161. 52. Daubenmier JJ, Weidner G, Marlin R, Crutchfield L, Dunn-Emke S, Chi C, et al. Lifestyle and health-related quality of life of men with prostate cancer managed with active surveillance. Urology. 2006;67(1):125-30. 53. Winger JG, Mosher CE, Rand KL, Morey MC, Snyder DC, Demark-Wahnefried W. Diet and exercise intervention adherence and health-related outcomes among older long-term breast, prostate, and colorectal cancer survivors. Ann Behav Med. 2014;48(2):235-45. 54. McGowan EL, North S, Courneya KS. Randomized controlled trial of a behavior change intervention to increase physical activity and quality of life in prostate cancer survivors. Ann Behav Med. 2013;46(3):382-93. 55. Park SW, Kim TN, Nam JK, Ha HK, Shin DG, Lee W, et al. Recovery of overall exercise ability, quality of life, and continence after 12-week combined exercise intervention in elderly patients who underwent radical prostatectomy: a randomized controlled study. Urology. 2012;80(2):299-305. 56. Buffart LM, Galvao DA, Chinapaw MJ, Brug J, Taaffe DR, Spry N, et al. Mediators of the resistance and aerobic exercise intervention effect on physical and general health in men undergoing androgen deprivation therapy for prostate cancer. Cancer. 2014;120(2):294-301. 57. Kalter J, Kampshoff CS, Chinapaw MJ, van Mechelen W, Galindo-Garre F, Schep G, et al. Mediators of Exercise Effects on HRQoL in Cancer Survivors after Chemotherapy. Med Sci Sports Exerc. 2016. 58. Buffart LM, Newton RU, Chinapaw MJ, Taaffe DR, Spry NA, Denham JW, et al. The effect, moderators, and mediators of resistance and aerobic exercise on health-related quality of life in older long-term survivors of prostate cancer. Cancer. 2015;[Epub ahead of print]. 59. McTiernan A. Mechanisms linking physical activity with cancer. Nat Rev Cancer. 2008;8(3):205-11. 60. Petersen AM, Pedersen BK. The anti-inflammatory effect of exercise. J Appl Physiol. 2005;98(4):1154-62. 61. Werner H, Bruchim I. The insulin-like growth factor-I receptor as an oncogene. Arch Physiol Biochem. 2009;115(2):58-71. 62. Khera M. Male hormones and men's quality of life. Curr Opin Urol. 2016;26(2):152-7. 85 63. Pereira RF, Daibs YS, Tobias-Machado M, Pompeo AC. Quality of life, behavioral problems, and marital adjustment in the first year after radical prostatectomy. Clin Genitourin Cancer. 2011;9(1):53-8. 64. Ramsey SD, Zeliadt SB, Blough DK, Moinpour CM, Hall IJ, Smith JL, et al. Impact of prostate cancer on sexual relationships: a longitudinal perspective on intimate partners' experiences. J Sex Med. 2013;10(12):3135-43. 65. Erickson KI, Miller DL, Roecklein KA. The aging hippocampus: interactions between exercise, depression, and BDNF. Neuroscientist. 2012;18(1):82-97. 66. Phillips SM, McAuley E. Physical activity and quality of life in breast cancer survivors: the role of self-efficacy and health status. Psychooncology. 2014;23(1):27-34. 86 3.10 Tables and figures Table 3.1. Characteristics for prostate cancer survivors (n=817), in the Prostate Cancer Cohort Study, Alberta, Canada, 1997-2014. Study Characteristics a Gleason score at diagnosis ≤7 >7 PSA value at diagnosis <4 4-10 >10-20 >20 Stage of cancer II (T1/T2, N0, M0) III (T3, N0, M0) III/IV (T3, NX, MX) IV b Primary Treatment c Prostatectomy Hormone therapy d Radiation therapy Relationship status Married/ common law Other Education level University degree Other non-university degree Trade degree High school diploma Less than High school diploma Race Caucasian Other Region of residence Urban Rural Post-diagnosis Charlson comorbidity score 0 1 2 3+ First degree family history of prostate cancer Yes n (%) 312 (38.2%) 505 (61.8%) 75 (9.2%) 297 (36.4%) 193 (23.6%) 252 (30.8%) 630 (77.1%) 57 (7.0%) 75 (9.2%) 55 (6.7%) 240 (29.4%) 517 (63.3%) 359 (43.9%) 689 (84.3%) 128 (15.7%) 140 (17.2%) 170 (20.8%) 166 (20.3%) 86 (10.5%) 255 (31.2%) 780 (95.5%) 37 (4.5%) 482 (59.0%) 335 (41.0%) 71 (8.7%) 218 (26.7%) 214 (26.2%) 314 (38.4%) 166 (20.3%) 87 No Smoking status Current smoker Former smoker Never smoker 651 (79.7%) 114 (14.0%) 465 (56.9%) 238 (29.1%) Median (Q1, Q3) e Total lifetime alcohol intake (g/year) 1909.2 (588.4, 4507.3) Dietary caloric intake (kcal/day) 2008.3 (1644.4, 2474.5) Pre-diagnosis lifetime total physical activity f 143.7 (98.1, 198.3) f Pre-diagnosis lifetime recreational physical activity 12.4 (6.7, 21.3) Pre-diagnosis lifetime occupational physical activity f 107.9 (58.0, 160.4) Pre-diagnosis lifetime household physical activity f 17.8 (9.2, 29.9) f First follow-up total physical activity 78.7 (44.2, 126.9) First follow-up recreational physical activity f 11.1 (3.0, 25.1) f First follow-up occupational physical activity 2.9 (0.0, 50.0) First follow-up household physical activity f 33.1 (12.5, 56.0) Second follow-up total physical activity f 71.6 (35.1, 132.9) Second follow-up recreational physical activity f 13.5 (2.0, 32.6) f Second follow-up occupational physical activity 0.0 (0.0, 41.8) Second follow-up household physical activity f 25.9 (7.6, 49.7) Third follow-up total physical activity f 62.0 (28.3, 106.7) Third follow-up recreational physical activity f 13.5 (0.0, 32.9) Third follow-up occupational physical activity f 0.0 (0.0, 12.9) f Third follow-up household physical activity 21.9 (5.8, 47.5) Mean (SD) Age at diagnosis (years) 67.3 (7.4) Body mass index (kg/m2) 28.0 (3.8) First follow-up Physical component summary score 40.8 (12.4) First follow-up Mental component summary score 50.8 (11.8) Second follow-up Physical component summary score 40.8 (12.2) Second follow-up Mental component summary score 50.5 (11.7) Third follow-up Physical component summary score 40.1 (12.0) Third follow-up Mental component summary score 51.0 (11.6) a Baseline (captured at diagnosis) characteristics unless otherwise stated. b IV included: (T4, N0, M0), (T4 N0/NX, M0/MX), (any T, N1, M0/MX), (any T, any N, M1), (any T, NX, MX). c Not mutually exclusive (could have more than one treatment). d Included bilateral orchiectomy, luteinizing hormone-releasing hormone agonists, nonsteroidal antiandrogens. 88 e 66 never-drinkers. f Units: MET-hour/week/year. 89 Table 3.2. Associations between post-diagnosis first follow-up physical activity and post-diagnosis first follow-up quality of life in prostate cancer survivors in Alberta, Canada in 1997-2002. Type of physical activity Physical activity median (min, max) Physical component summary score a Mental component summary score b β 95% CI P-value β 95% CI P-value 0.04 0.16 0.02 0.01 0.02, 0.05 0.12, 0.20 0.01, 0.03 -0.01, 0.03 <0.0001 <0.0001 0.001 0.14 0.01 0.05 0.01 -0.01 0, 0.02 0.01, 0.09 -0.01, 0.02 -0.03, 0.01 0.19 0.024 0.22 0.26 0.03 0.16 0.02 0.01 0.02, 0.04 0.12, 0.20 0, 0.03 -0.01, 0.03 <0.0001 <0.0001 0.023 0.20 0.01 0.05 0.01 -0.01 -0.01, 0.02 0.01, 0.10 -0.01, 0.02 -0.04, 0.01 0.20 0.024 0.24 0.33 0.09 0.68 0.05, 0.13 0.50, 0.85 <0.0001 <0.0001 -0.01 0.22 -0.05, 0.04 0.03, 0.40 0.76 0.021 0.08 0.71 0.04, 0.13 0.53, 0.88 <0.0001 <0.0001 -0.01 0.25 -0.06, 0.04 0.05, 0.45 0.78 0.015 All activity (MET-hour/week/year) Total sample (n=817) Total 79 (0, 491) Recreational c 11 (0, 120) c Occupational 3 (0, 445) Household c 33 (0, 314) Excluding stage III/IV and IV (n=687) d Total 79 (0, 491) Recreational c 11 (0, 120) c Occupational 3 (0, 445) Household c 33 (0, 314) Moderate-to-vigorous activity (hour/week/year) Total sample (n=817) Total 28 (0, 113) c Recreational 2 (0, 28) Excluding stage III/IV and IV (n=687) d Total 28 (0, 113) c Recreational 2 (0, 28) 90 All models adjusted for: age at diagnosis, aggressive versus non-aggressive prostate cancer, prostate cancer treatment (none, hormone only, radiation therapy only or both hormone and radiation therapy), prostatectomy, post-diagnosis Charlson co-morbidity score, level of education and pre-diagnosis lifetime physical activity. a Additionally adjusted for: smoking status. b Additionally adjusted for: whether or not they joined a support group post-diagnosis and whether they had family support or not. c Mutually adjusted for other physical activity types and intensities (for moderate-vigorous activity analysis). d Stage IV cancers: 55; stage III/IV cancers: 75. 91 Table 3.3. Associations between physical activity change (pre-diagnosis minus average post-diagnosis scores) over the diagnosis period and average post-diagnosis quality of life in prostate cancer survivors in Alberta, Canada in 1997-2007. Type of physical activity Physical activity change score: Median (min, max) All activity (MET hour/week/year) Total sample (n=817) Total -60 (-425, 246) Recreational b 0 (-83, 120) b Occupational -76 (-390, 258) Household b 10 (-87, 279) Excluding stage III/IV and IV (n=687) c Total -60 (-426, 246) b Recreational 0 (-83, 120) Occupational b -76 (-390, 258) Household b 10 (-87, 279) Excluding those who died before 3rd follow-up (n=559) d Total -56 (-426, 246) Recreational b 0 (-83, 120) b Occupational -72 (-390, 258) Household b 10 (-66, 222) Physical component summary score a Mental component summary score β 95% CI P-value β 95% CI P-value 0.04 0.15 0.04 0.02 0.03, 0.05 0.11, 0.19 0.02, 0.05 0, 0.04 <0.0001 <0.0001 <0.0001 0.08 0.01 0.06 0.02 0 0, 0.03 0.02, 0.11 0, 0.03 -0.03, 0.02 0.07 0.006 0.05 0.51 0.04 0.16 0.03 0.02 0.02, 0.05 0.12, 0.20 0.01, 0.04 0, 0.05 <0.0001 <0.0001 0.001 0.05 0.01 0.09 0.02 -0.02 0, 0.03 0.04, 0.14 0, 0.04 -0.05, 0.01 0.15 <0.0001 0.09 0.16 0.04 0.14 0.03 0.03 0.02, 0.05 0.10, 0.18 0.01, 0.04 0, 0.05 <0.001 <0.001 0.001 0.027 0.02 0.08 0.02 0 0, 0.04 0.03, 0.13 0, 0.04 -0.03, 0.03 0.019 0.001 0.07 0.94 0.11 0.72 0.07, 0.16 0.54, 0.90 <0.0001 <0.0001 0.04 0.35 -0.01, 0.09 0.15, 0.55 0.15 0.001 0.10 0.81 0.05, 0.15 0.62, 1.00 <0.0001 <0.0001 0.02 0.45 -0.03, 0.08 0.23, 0.68 0.38 <0.0001 Moderate-to-vigorous activity (hour/week/year) Total sample (n=817) Total -17 (-83, 60) Recreational b 0 (-20, 21) Excluding stage III/IV and IV (n=687) c Total -17 (-83, 60) Recreational b 0 (-20, 21) 92 Excluding those who died before 3rd follow-up (n=559) d Total -16 (-83, 60) 0.11 0.06, 0.16 <0.0001 0.05 -0.01, 0.11 0.08 b Recreational 0 (-20, 21) 0.70 0.51, 0.89 <0.0001 0.43 0.21, 0.66 <0.0001 All models adjusted for: age at diagnosis, aggressive versus non-aggressive prostate cancer, prostate cancer treatment, prostatectomy, PSA value at diagnosis, post-diagnosis Charlson co-morbidity score and pre-diagnosis lifetime physical activity. a Additionally adjusted for BMI and smoking status. b Mutually adjusted for other physical activity types and intensities (for moderate-vigorous activity analysis). c Stage IV cancers: 55; stage III/IV cancers: 75. d 152 participants died before second follow-up; 106 participants died between second and third follow-up. 93 Table 3.4. Change in moderate-to-vigorous recreational physical activity based on meeting the cancer prevention guidelines (150 minutes/week) over the diagnostic period (pre- and average post-diagnosis) and average post-diagnosis quality of life in prostate cancer survivors in Alberta, Canada in 1997-2007. Moderate-to-vigorous recreational physical activity guideline adherence (hour/week/year) No. in group Physical component summary score β 95% CI P-value Mental component summary score β 95% CI P-value Total sample (n=817) Non-exercisers 272 Referent Referent Adopters 173 4.80 2.82, 6.78 <0.0001 2.26 0.09, 4.43 0.041 Relapsers 149 -0.04 -2.10, 2.01 0.97 -0.35 -2.60, 1.89 0.76 Maintainers 223 6.01 4.15, 7.86 <0.0001 2.32 0.29, 4.34 0.025 Excluding stage III/IV and IV cases (n=687) a Non-exercisers 224 Referent Referent Adopters 142 4.71 2.60, 6.81 <0.0001 2.12 -0.30, 4.54 0.09 Relapsers 125 -0.28 -2.46, 1.89 0.80 -0.64 -3.13, 1.86 0.62 Maintainers 196 6.03 4.09, 7.96 <0.0001 2.34 0.12, 4.56 0.039 Excluding those who died before 3rd follow-up (n=559) c Non-exercisers 151 Referent Referent Adopters 128 4.75 2.54, 7.00 <0.0001 1.89 -0.66, 4.46 0.15 Relapsers 99 0 -2.37, 2.36 0.99 -0.41 -3.15, 2.33 0.77 Maintainers 181 5.76 3.74, 7.78 <0.0001 2.19 -0.15, 4.52 0.07 All models adjusted for: age at diagnosis, aggressive versus non-aggressive prostate cancer, prostate cancer treatment, prostatectomy, PSA value at diagnosis and post-diagnosis Charlson co-morbidity score. a Stage IV cancers: 55; stage III/IV cancers: 75. c 152 participants died before second follow-up; 106 participants died between second and third follow-up. 94 Figure 3.1. Change in moderate-to-vigorous recreational physical activity guideline adherence groups over the diagnostic period in prostate cancer survivors in Alberta, Canada (2000-2007). 95 3.11 Additional analyses and results: Lasso and elastic-net regularized generalized linear model (GLMNET package) covariate selection methods The lasso and elastic-net regularized generalized linear model (GLMNET) package in R, is an efficient method to test all possible interactions between multiple covariates in a single step. GLMNET produces output and a figure indicating influential covariates whose coefficients deviate from zero. From there, a cut-off of the lambda value was determined. As the lambda value (a measure between 0-1) increases, the model is more conservative on covariate inclusion in the model. For this analysis, our selection was very lenient and we tested all covariates deviating from the zero line. The objective of these analyses was to determine if the GLMNET method and the backwards elimination statistical modelling covariate selection methods came to the same conclusions when examining the association between post-diagnosis physical activity and physical and mental QoL scores. Models fit with all terms and plotted visually using the GLMNET approach found few important covariates for further testing (Figures 3.2-3.5). Specifically, for total physical activity and PSC score, only one covariate deviated from zero. However, all covariates were fit in multivariable liner regression models and did appear to substantially confound the relationship between post-diagnosis physical activity and physical and mental QoL. Final models were adjusted only for a priori covariates (age at diagnosis, prostate cancer treatment, Gleason score, Charlson post-diagnosis co-morbidity index score, prostatectomy and stage of cancer) and results were very similar to the traditional backwards elimination method results for the association between post-diagnosis physical activity and physical and mental QoL (Table 3.5). Statistical significance and magnitude in the physical activity and QoL relationships did not notably differ between the GLMNET approach and the traditional approach. However, the final models were adjusted for additional variables in the traditional approach. 96 A key strength of the GLMNET approach for epidemiological research is the elimination of the issue of multiple testing. Most epidemiological studies seek to test multiple covariates in analyses to determine the best model fit and biological relevance of their data. The art of covariate selection requires multiple tests, elaborate planning and transparency and might not always be feasible with the traditional methods of backwards elimination. GLMNET is able to take into account all combinations of included covariates and may be more sensitive to subtle changes in covariate interactions. Implications of comparisons between the two model covariate selection methods that need to be considered include that these analyses were descriptive and very preliminary. A more in-depth analysis may be undertaken in future studies to determine specific advantages and disadvantages in choosing which method may be more beneficial when modelling epidemiological associations. 97 Table 3.5. Associations between post-diagnosis physical activity on first follow-up post-diagnosis QoL using lasso and elasticnet regularized generalized linear model (GLMNET package) methods in prostate cancer survivors in Alberta, Canada in 1997-2002 (n=817). Physical Component Score a Type of physical activity β 95% CI (MET-hour/week/year) Total -0.02 -0.03, -0.008 Recreational* 0.08 0.02, 0.14 Occupational* -0.02 -0.03, -0.009 Household* -0.03 -0.08, 0.02 *Mutually adjusted for other physical activity types. a Mental Component Score a P-value β 95% CI P-value 0.001 0.009 <0.001 0.192 -0.004 0.05 -0.003 -0.03 -0.02, 0.008 -0.02, 0.12 -0.02, 0.009 -0.08, 0.03 0.536 0.1156 0.580 0.360 All models adjusted for: age at diagnosis, Gleason score, stage of prostate cancer, prostate cancer treatment, prostatectomy and post- diagnosis Charlson co-morbidity score. 98 Figure 3.2. Total physical activity and PCS score output using GLMNET regularization. 99 Figure 3.3. Total physical activity and MCS score output using GLMNET regularization. 100 Figure 3.4. Recreational, occupational and household physical activity (mutually adjusted) and PCS score output using GLMNET regularization. 101 Figure 3.5. Recreational, occupational and household physical activity (mutually adjusted) and MCS score output using GLMNET regularization. 102 Chapter Four: IDENTIFICATION AND PREDICTION OF QUALITY OF LIFE TRAJECTORIES AFTER A PROSTATE CANCER DIAGNOSIS 4.1 Preamble This manuscript addresses objective three of this thesis through a group-based trajectory modelling longitudinal analysis determining trajectory groups of QoL. The manuscript was submitted for publication on June 24, 2016 and is currently under review. This work was also accepted as an oral/poster presentation at the following conferences: Farris MS, Kopciuk KA, Courneya KS, McGregor SE, Friedenreich CM. Quality of life trajectories after prostate cancer diagnosis: the role of physical activity and prognostic factors. Canadian Society of Epidemiology and Biostatistics (CSEB) Student Conference, Winnipeg, MA June 8-10, 2016. (Oral presentation) Farris MS, Kopciuk KA, Courneya KS, McGregor SE, Friedenreich CM. Quality of life trajectories after prostate cancer diagnosis: the role of physical activity and prognostic factors. 2016 Epidemiology Congress of the Americas, Miami, Florida June 21-24, 2016. (Poster presentation) Running title: Trajectories of QoL in prostate cancer survivors Keywords: group-based trajectory, quality of life, prostate cancer, mixed model 103 Financial support: Research relating to this manuscript was supported by the Canadian Institute of Health Research [2004-2007; grant number MOP-67217]; the National Cancer Institute of Canada with funds from the Canadian Cancer Society [2000-2004; grant number 011004]; and the Alberta Cancer Board-Research Initiative Program [20042007; grant number 4570)]. Christine M. Friedenreich held career awards from Alberta Innovates-Health Solutions and the Alberta Cancer Foundation, Kerry S. Courneya was supported by the Canada Research Chairs Program and S. Elizabeth McGregor by the Alberta Heritage Foundation for Medical Research. 104 4.2 Abstract Introduction: The aim of this study was to identify physical and mental quality of life (QoL) trajectories after a prostate cancer diagnosis and systematically characterize these trajectories by behaviours and prognostic factors. Materials and methods: Prostate cancer survivors diagnosed between 1997-2000 were recruited between 2000-2002 into a prospective cohort study with repeated measurements. Behavioural/prognostic data were collected through in-person interviews and questionnaires. QoL was collected at three post-diagnosis time-points, approximately two years apart using the Short Form (SF)-36 validated questionnaire. To identify physical and mental QoL trajectories, group-based trajectory modelling was undertaken. Differences between groups were evaluated by assessing potentially influential dropouts (through mortality or poor health), behavioural/prognostic factors at diagnosis or during the follow-up. Results: Three trajectories of physical QoL were identified including: averagemaintaining QoL (32.2%), low-declining QoL (40.5%) and very low-maintaining QoL (27.3%). In addition, three trajectories for mental QoL were identified: averagemaintaining QoL (66.5%), above average-declining QoL (19.7%) and low-increasing QoL (13.8%). In both physical and mental QoL, dropout (due to morality/loss to followup) was different between trajectories, thus confirming QoL and mortality were closely related. Furthermore, increased Charlson comorbidity index score was consistently associated with physical and mental QoL group membership relative to average 105 maintaining groups, while behaviours such as time-varying physical activity was only associated with physical QoL trajectories, not mental QoL trajectories. Conclusion: It was possible to define three trajectories of prostate cancer survivors related to physical and mental QoL. These data provide insights regarding means for identifying subgroups of prostate cancer survivors with lower or declining QoL after diagnosis who could be targeted for further interventions aimed at improving QoL. 106 4.3 Introduction Globally the burden of prostate cancer is on the rise, with an estimated 1.1 million new cases of prostate cancer (1) and an estimated 3.9 million prevalent cases of prostate cancer worldwide estimated in 2012 (2). Screening programs largely driven by the Prostate-Specific Antigen (PSA) testing introduced in the early 1990s have increased detection of prostate cancer, specifically in developed countries (1, 3). Early detection was believed to reduce prostate cancer mortality (4). However, early detection may also result in over-diagnosis and over-treatment of potentially non-life threatening prostate cancer (5). The consequences of treating men with non-life threatening prostate cancer are serious since long-term residual side effects, reduced functioning, a compromised mental state and overall reduced Quality of Life (QoL) (6-10) often occur in this population. While QoL outcomes have been well documented after diagnosis of prostate cancer, most research studies focus on impacts of different treatments (11) or supportive interventions (12) on QoL over time. Moreover, studies have examined QoL changes with repeated measurements long-term after diagnosis of prostate cancer (13-17). However, to our knowledge, studies that examined physical and mental QoL changes in prostate cancer survivors have made assumptions that the whole study population can be represented by a single pattern of QoL, by measuring average changes in QoL. Nonetheless, it is unknown if there are distinct subgroups of prostate cancer populations with different QoL patterns following diagnosis. Subtle differences may be missed when examining average patterns of QoL over time. Group-based trajectory modelling (GBTM) is an innovative analytic method that allows researchers to determine longitudinal subgroups within a population without making distribution assumptions (18). 107 To date, only two studies have utilized GBTM to evaluate surgical outcomes following prostatectomy in prostate cancer survivors including urinary function (19) and sexual function (20). To our knowledge, GBTM has never been applied to physical and mental QoL in a prostate cancer population. The primary objective of this prospective cohort study was to determine QoL trajectory patterns in prostate survivors. We hypothesized that there would be more than one and up to four trajectory groups of physical and mental QoL in our study population up to 10 years following prostate cancer diagnosis. A secondary objective and exploratory analysis was to characterize trajectory groups systematically by analyzing common prognostic and behavioral factors associated with prostate cancer. We hypothesized that factors associated with negative prognosis would be associated with low QoL groups relative to any high QoL group. 4.4 Materials and methods 4.4.1 Study sample This prostate cancer cohort study in Alberta, Canada has been previously published in detail elsewhere (21). Briefly, initial recruitment of prostate cancer cases (through the Alberta Cancer Registry) was carried out in a population-based case-control study (1997-2000) and participants were re-contacted for consent to enroll in the prospective cohort study between the years 2000 and 2002. Telephone interviews were completed at the first follow-up to obtain physical activity data and a questionnaire was mailed to participants to obtain QoL data. Subsequently, at second and third follow-ups, self-administered questionnaires were sent through the mail to obtain updated physical activity and QoL data. Final mailed questionnaire assessments were completed in 2007 108 and participants were followed for cancer outcomes until 2014 by medical chart abstractions, vital statistics updates and medical record checks. Eligibility criteria for prostate cancer cases included: histologically confirmed invasive stage II prostate cancer or greater, English speaking, <80 years of age and no previous diagnosis of cancer (excluding non-melanoma skin cancer). Ethics approval was obtained through the Alberta Cancer Research Ethics Board and the Conjoint Health Research Ethics Board at the University of Calgary. 4.4.2 Data collection The Short Form (SF)-36 , previously tested for reliability and validity in several domains (22), was used to evaluate the prostate cancer survivors QoL at three separate time-points. Composite summary scores of physical and mental QoL were calculated. Each participant needed to have all eight individual SF-36 scale scores to be considered eligible to calculate summary scores. Higher composite summary scores related to higher QoL and possible scores ranged from 0-100. Clinically meaningful differences between QoL scores were ≥5 points (23). Vital status updates of the survivors was completed on a monthly basis, and further updated at each follow-up time-point by study staff. We assumed that participants who could not be contacted because of death, refusal or loss to follow-up were likely in poor health and which consequently influenced their ability to participate in the cohort follow-ups that occurred at three separate time points between 2000 and 2007. Medical data and cancer progressions, recurrences and new primary cancer outcomes were obtained through clinical chart abstractions done by qualified health record technicians working with the Alberta Cancer Registry. 109 Other behavioural and prognostic information was obtained from the prior casecontrol study (24) through in-person interviews. Variables obtained at prostate cancer diagnosis and used in this current study included: average total energy intake (grams/day), average alcohol intake (grams/year), smoking status, Body Mass Index (BMI) (kg/m2), age in years, aggressive versus non-aggressive disease (> stage II or Gleason score ≥8 (25)), whether they had a prostatectomy, radiation therapy, hormone therapy and number of co-morbidities based on the Charlson co-morbidity index score (26). In addition, past year physical activity (MET-hours/week/year) was also collected post-diagnosis at three separate follow-ups (21). 4.4.3 Statistical analysis 4.4.3.1 Identifying trajectory groups The analysis utilized GBTM (Traj for Stata®) (27), which uses finite mixture modelling to approximate unknown distributions of trajectories across a study population. Censored normal or tobit models were used to estimate trajectories of physical and mental QoL over the post-diagnosis period (2000-2007 for active data collection) in this cohort of prostate cancer survivors. First follow-up data were collected between 20002002, second follow-up data in 2002-2004 and third follow-up in 2004-2007 with approximately two years between measurements for each participant (Figure 4.1). In the model selection process, the Bayesian Information Criterion (BIC) was utilized to determine the best model underlying the group selection and functional form (18). The BIC was used for model selection within a finite set of models and is based on the likelihood function plus a penalty term for the number of parameters in the model (28). BIC values balance model fit with model complexity and the closer the negative BIC 110 value is to zero, the better is the fit of the model. Supplemental non-statistical criteria were also implemented for model selection including: (1) all trajectory groups encompassed >10% of the total sample, (2) all trajectories were distinct from one another (visual assessment of trajectory figures looking for non-overlapping confidence intervals (CI)) and (3) posterior probabilities of group membership were >0.7 (18). Posterior probabilities represent the average probability that the trajectory group each participant was assigned to was the most appropriate group selection and model fit. Model selection is a multi-step procedure. First, we determined the most appropriate number of trajectories for the physical and mental QoL data by fitting models with up to four trajectory groups. A priori, we hypothesized that four groups would be the maximum number of groups tested for model fit according to the sample size and to avoid overfitting the data. Second, the functional form of the trajectory groups (linear or quadratic) was evaluated to obtain a base model for physical and mental QoL trajectory groups separately. Once the model trajectory groups were established, testing for influential observations and determining characteristics of trajectory groups followed (18). There was a substantial decrease in the sample size during the follow-up period that extended up to 10 years since diagnosis because of mortality, poor health or loss to follow-up, so differential dropout assessment between trajectory groups was performed. Dropout probabilities for each trajectory group were estimated to substantiate differences between groups (29). Differential dropout was evaluated based on arbitrary differences between any two trajectory groups of >10%. If there were substantial differences (>10%), dropout was considered influential and this process included/maintained in the final models. 111 4.4.3.2 Characteristics of trajectory groups Descriptive characteristics for each physical and mental QoL trajectory group were produced and assessed for initial differences between groups. Then, as an exploratory analysis, we assessed behavioural and prognostic factors to characterize physical and mental QoL trajectory group membership using multivariable multinomial logistic regression. Multivariable models included the following prognostic and behavioural risk factors: age at diagnosis (years), aggressive versus non-aggressive disease at diagnosis (Stage >II or Gleason score ≥8 (25)), prostatectomy, radiation therapy, hormone therapy, Charlson co-morbidity index score (26) post-diagnosis, approximate daily energy intake (kcal/day), average alcohol intake (grams/year), smoking status (current versus former/never (30)) and BMI at diagnosis (kg/m2). Timevarying physical activity was measured up to three time-points (MET-hours/week/year) and evaluated for associations on the groups trajectories. The trajectory group with the highest QoL was used as the reference category. Relative risk ratios (RRRs) and 95% CIs were produced based on the final adjusted models to represent the multinomial logistic regression comparisons. In addition, models produced estimated mean QoL coefficients (intercepts), slope values and physical activity as a time-varying covariate for each trajectory group. Additionally, we assessed if there were any within trajectory group dropout characteristic differences. We included a logistic regression model examining the odds of dropping out within each trajectory group that could depend on the two previous responses in the final adjusted models. 112 4.4.3.3 Sensitivity analyses Sensitivity analyses were performed on the complete case sample (all three follow-ups completed) and two complete assessments sample (two of the three followups completed), to determine differences in model selection, dropout probabilities and prognostic/behavioural risk factor associations with trajectory group membership. Further, to determine the impact of modelling dropout, multivariable multinomial logistic regression excluding dropout were performed. Then we examined participants who switched trajectory groups with and without modelling dropout to substantiate any shift in trajectory group membership. Finally, to explore the possibility of reverse causality, time-lagged models were fit based on previous physical or mental QoL scores. Final trajectory group figures were produced post-model selection. All statistical tests were two-sided and evaluated at an alpha level of 0.05, using the statistical package Stata v.13 (College Town, Texas). 4.5 Results 4.5.1 Sample There were 830 prostate cancer survivors who provided consent to participate in this cohort study. Trajectory models for both physical and mental QoL included 817 cancer survivors who had complete first follow-up data. Of these, 636 and 454 participants had complete data at the second and the third follow-up time-points, respectively. 113 4.5.2 Trajectory group selection and evaluation Three physical QoL trajectory groups were identified from the GBTM analysis over the post-diagnosis follow-up (Table 4.1). Moreover, when assessing functional form, the BIC was closest to zero when all groups were linear, thus the linear function in all groups was considered the most advantageous model fit (Table 4.2). All groups included at least 100 survivors (which would equate to 12.2% of the total sample (n = 817)). After fitting the base model, we fit models to determine if dropout, behavioural and prognostic factors were associated with model fit and group membership. First, the model including dropout did not improve model fit according to the BIC values. The base model BIC = -7228 and the dropout model BIC = -7558, in which case, the base model BIC value was closer to zero, thus a better fit. However, when we estimated dropout probabilities for each trajectory group, they were substantially different. The probability of dropping out was more than double (0.42) in the very low-maintaining QoL group, compared to 0.15 in the low-declining group and 0.21 in the average-maintaining QoL group. Therefore, the attrition process or probability of “dropout” that was based on the two previous outcomes was maintained in the model to allow for this variation across groups. Next, models were fit with all behavioural and prognostic factors in a multivariable multinomial logistic regression analysis. The BIC improved substantially with inclusion of these factors (dropout model without factors=-7558 versus with factors -7453). A plot/figure of the fully adjusted model was produced (Figure 4.2). Based on the visual assessment of this plot, there was a “very low-maintaining” group with physical QoL mean scores consistently at 24 throughout the follow-up, a “low-declining” group 114 with physical QoL scores starting at approximately 42 and declining to 36 by the third follow-up and a “average-maintaining” group, with physical QoL mean scores consistently at approximately 50. Posterior group membership probabilities were also sufficiently high with: 0.86 for the very low-maintaining group, 0.79 for the lowdeclining group and 0.86 for the average-maintaining group. In addition, groups were well separated with clear differences of >5 points between the groups. Three mental QoL trajectory groups also emerged from the GBTM analysis. The BIC value was the closest to zero in the model with three groups and became unstable with four groups (Table 4.3). Linear functions were most appropriate for each trajectory group model fit (Table 4.4). Further, all groups included more than 100 prostate cancer survivors. The BIC associated with the model including dropout for mental QoL trajectories did not achieve a better fit (base model = -7089 versus dropout model = 7442). However, like physical QoL, the low-increasing mental QoL group had statistically significantly higher dropout probabilities (0.38) compared to above averagedeclining (0.27) and average-maintaining (0.20) groups and therefore, the attrition process or probability of “dropout” that was also based on the two previous outcomes was maintained in the model. Similar to physical QoL again, the BIC value in the mental QoL multinomial logistic model including prognostic/behavioural factors substantially improved when they were included (dropout model=-7442 versus adjusted model=7425). A final figure from the fully adjusted model was produced, see Figure 4.3. According to the figure, there was a “low-increasing” QoL group that started at 26 and increased in QoL score to 36, a “above average-declining” QoL group starting at 57 and 115 declined substantially to mean scores of 38 over the follow-up and a “averagemaintaining” QoL group starting at 55 and slightly increasing to 57 in mental QoL score. Posterior group membership probabilities were 0.91, 0.80 and 0.87 for low-increasing, above average-declining and average-maintaining QoL groups, respectively. There was separation (generally >5 points) between the trajectory groups with none of the groups overlapping. 4.5.3 Characteristics of trajectory groups Descriptive characteristics of physical QoL trajectory groups are found in Table 4.5. The low-declining QoL group appeared to have the highest number of survivors who had aggressive disease (n=70) and radiation therapy (n=168). The average-maintaining QoL group had the largest proportion of survivors who did not receive hormone therapy (50%), did not have a prostatectomy (46%) and were former/never smokers (90%). Median age at diagnosis, Charlson co-morbidity index score and BMI increased across average-maintaining, low-declining and very low-maintaining QoL groups. Alternatively, median physical activity at every time-point decreased across QoL groups with the lowest levels in the very low-maintaining QoL group and the highest levels in the averagemaintaining QoL group. Table 4.6 displays descriptive characteristics of mental QoL trajectory groups. A large proportion of the participants were in the average-maintaining QoL group (568 out of 817), therefore, it is difficult to glean meaningful differences between the QoL groups. The above average-declining QoL group had the highest proportion of participants who received hormone therapy (79%) and had a prostatectomy (91%). Interestingly, the lowincreasing group had the lowest proportion of aggressive disease (11%). The above 116 average-declining QoL group had the highest median age at diagnosis (72 years), highest Charlson co-morbidity index score and the lowest physical activity levels at the first follow-up (68 MET-hour/week/year). BMI, average alcohol consumption and total energy intake did not substantially differ between trajectory groups. When examining the multinomial logistic regression models, a higher Charlson co-morbidity index score was statistically significantly associated with both very lowmaintaining physical QoL (RRR = 2.04, 95% CI:1.66-2.51) and low-declining physical QoL (RRR = 1.59, 95% CI: 1.29-1.95) group membership, relative to the averagemaintaining QoL group (Table 4.7). Similarly, BMI was statistically significantly associated with very low-maintaining and low-declining physical QoL group membership with an increase in risk per kg/m2 (very low-maintaining: RRR = 1.18, 95% CI: 1.091.27; low-declining: RRR = 1.09, 95% CI: 1.02-1.17), relative to the average-maintaining physical QoL group. Other statistically significant associations were found with hormone therapy with very low-maintaining and low-declining physical QoL group membership while prostatectomy, smoking status and age at diagnosis were only associated with very low-maintaining physical QoL group membership. Results from mental QoL group membership multinomial logistic regression were slightly attenuated compared to physical QoL trajectory group results (Table 4.7, lower). Charlson comorbidity index score was statistically significantly associated with lowincreasing (RRR = 1.18, 95% CI: 1.03-1.36) and above average-declining (RRR = 1.32, 95% CI: 1.12-1.56) mental QoL group membership, relative to the average-maintaining mental QoL group. Interestingly, aggressiveness of disease (RRR = 0.40, 95% CI: 0.190.86) and receiving radiation therapy (RRR = 0.50, 95% CI: 0.28-0.89) were statistically 117 significantly associated with only the low-increasing mental QoL group membership relative to the average-maintaining mental QoL group. Having a prostatectomy was statistically significantly associated with a three-fold risk of above average-declining QoL group membership (RRR = 3.32, 95% CI: 1.23-8.94) but was not significantly associated with low-increasing QoL group membership (RRR = 1.73, 95% CI: 0.893.36). Table 4.8 displays the mean coefficients for values from the fully adjusted models in physical and mental QoL trajectory groups. In physical QoL groups, baseline intercept coefficient QoL scores were estimated as follows: average-maintaining (49.64), low-declining (41.68) and very low-maintaining (24.37). Total physical activity measured in MET-hours/week/year was significantly and positively associated with all trajectory groups (estimated slopes of 0.02 to 0.03). The mean baseline intercept coefficients for the mental QoL trajectory group scores were: average-maintaining (54.59), above averagedeclining (57.32) and low-increasing (25.99). The above average-declining group had a negative estimated slope of -6.98, also evident in Figure 4.3. However, physical activity over time was not associated with any mental QoL trajectory groups. 4.5.4 Dropout within trajectory groups An exploratory analysis in physical QoL trajectory groups suggested that BMI was statistically significantly protective against dropping out in the very low-maintaining QoL group (OR = 0.91, 95% CI: 0.85-0.98) but not the low-declining or averagemaintaining QoL groups (data not shown). Having radiation therapy was statistically significantly protective against dropping out in the low-declining (OR = 0.52, 95% CI: 0.30-0.91) QoL group while, prostatectomy was adversely associated with dropping out 118 in the low-declining QoL group (OR = 2.30, 95% CI: 1.16-4.57) and average-maintaining QoL group (OR = 2.48, 95% CI: 1.11-5.55). In mental QoL trajectory groups, radiation therapy was statistically significantly protective against dropping out in the average-maintaining QoL group (OR = 0.50, 95% CI: 0.32-0.78). Having a prostatectomy was adversely significantly associated with dropping out in the average-maintaining QoL group (OR = 2.12, 95% CI: 1.26-3.58). In addition, current smokers had a three-fold significant increase in dropping out in the lowincreasing QoL group (OR = 3.48, 95% CI: 1.11-10.88) and almost two-fold significant increase in the average-maintaining QoL group (OR = 1.91, 95% CI: 1.09-3.35). No other significant associations were found and results were interpreted with caution due to the small sample sizes. 4.5.5 Sensitivity analyses In sensitivity analyses comparing survivors with complete follow-up data and those with only two complete follow-up assessments, the model selection and evaluation process did not differ for physical or mental trajectory groups established using all the data and for the final adjusted model fit. Multinomial logistic regression results were very similar for physical and mental QoL trajectories (data not shown). Further, when modelling dropout versus not, the multinomial logistic regression comparisons relative to the high QoL group were not different for physical or mental QoL trajectory groups. Additionally, we compared participants who switched trajectory groups when modelling dropout versus not and there were only 53 “switchers” in the physical QoL analysis and 58 “switchers” in the mental QoL analysis. Only one survivor in the mental QoL trajectory group analysis moved across two groups (low-increasing QoL group when 119 modelling dropout then switched to the average-maintaining QoL group without modelling dropout). Finally, we fit time-lagged models and found that modelling previous QoL score had a more advantageous fit with BIC values of -7168 versus -7453 for physical QoL trajectories and -7106 versus -7425 for mental QoL trajectories. However, multinomial logistic regression results did not substantially differ between the final adjusted models and the time-lagged models (data not shown). 4.6 Discussion When utilizing GBTM we identified three physical QoL trajectories that were meaningfully different according to clinically relevant thresholds (≥5 points) (23) up to 10 years following prostate cancer diagnosis. The average-maintaining and very lowmaintaining physical QoL group scores were relatively consistent across time-points. On the other hand, the low-declining physical QoL group scores declined by the third timepoint by approximately five QoL points. Prevalence in each of these groups was fairly equal, with the highest prevalence in the low-declining group (40.5%). We also discovered three mental QoL trajectories in our study population of prostate cancer survivors. The above average-declining mental QoL trajectory had substantial declines over the course of the study follow-up, whereas the low-increasing and averagemaintaining QoL trajectories experienced increases in QoL scores by the third time-point. Prevalence of the mental QoL trajectories was highly concentrated in the averagemaintaining groups with 66.5% of the population falling into this group relative to the above average-declining (19.7%) and low-increasing (13.8%) groups. In addition, we discovered that the probability of dropout was substantially different in the low physical 120 and mental QoL groups compared to the high and medium groups. Further, Charlson comorbidity index score was consistently associated with low and medium QoL group membership relative to the high QoL groups in both physical and mental trajectories. Other studies have focused on disease specific QoL in prostate cancer survivors. One study examining urinary function in prostate cancer survivors after radical prostatectomy found three trajectories of QoL post-surgery (19). Although this study focused on recovery patterns with a follow-up up to 12-months post-prostatectomy, age and number of co-morbidities were also associated with lower QoL group outcomes, similar to our study. The Prostate Cancer Outcomes Study, examined sexual function following radical prostatectomy (20). Interestingly, their GBTM analysis fit four distinct sexual function trajectories, with age, ethnicity, baseline sexual function and having had nerve-sparing surgery associated with the low QoL group. While these studies investigated different QoL outcomes, there are some similarities with regard to identifying high QoL and low QoL sub-populations, similar to our study. Other previous research a priori stratified prostate cancer cohorts based on age groups (31) and number of comorbidities (32). Age groups characterize differences in QoL similar to this study, yet, unlike our results, QoL scores post-diagnosis did not differ by number of co-morbidities. Further, despite treatment differences observed in this study between trajectory groups, mechanisms may be influenced by age of participants and length of follow-up, that cannot be teased apart in observational studies (33). Additionally, we found BMI, smoking and physical activity to be associated with medium and low QoL groups to different degrees. For example, time-varying physical activity and BMI were associated with physical QoL group trajectories but not mental 121 QoL trajectories. Moreover, smoking was associated with low groups in both physical and mental QoL trajectories but not medium QoL groups, additionally characterizing these QoL groups from one another. The ability to target prostate cancer survivors based on these characteristics may lead to enhanced QoL and survival outcomes. Our analysis that modelled dropout produced significantly higher dropout probabilities in both the physical and mental low QoL groups relative to the other groups. This finding supported the mechanism that QoL may be a predictor of morbidity and mortality (34). Clinical implications of these findings support interventions for these specific groups of prostate cancer survivors. Specifically, for the very low-maintaining physical QoL group, scores were consistently below 30, which is well below the mean score of 50 (35). The very low-maintaining physical QoL group was older, more likely to be on hormone therapy, have undergone a prostatectomy, have co-morbidities, higher BMI and more likely to be current smokers compared to the average-maintaining QoL group. These differences characterize the very low-maintaining QoL group and can make it easier for physicians to target this population for further intervention. There are limitations that should be noted in this study. First, this study was observational, and therefore, we cannot rule out the possibility of alternative explanations of our results. Confounding by indication can occur in observational studies where other factors, in this case co-morbid conditions, general health and age, may alter survivors’ baseline risk, thus explaining their QoL group membership, rather than the diagnosis of prostate cancer itself. For example, older prostate cancer survivors are more likely to have co-morbid conditions and less overall general health, resulting in a lower baseline for QoL scores after prostate cancer diagnosis. We had sufficient power to identify 122 trajectory groups using GBTM methods, however, when examining prognostic and behavioural associations between groups, precision was lacking and therefore, chance of type II error is present. This situation may be especially relevant in the low-increasing (13.8%) and above average-declining (19.7%) mental QoL trajectory groups, representing <35% of the study population. Moreover, our study only had three repeated measurements over the post-diagnosis period of two and up to 10 years post-diagnosis. We made the assumption that QoL remained consistent between follow-up time-points of up to two years post-diagnosis. In between follow-up time-points participants may experience response shift as perceptions and priorities of QoL change over time (36, 37). However, time-lagged exploratory models found that previous QoL score did predict future QoL scores and sensitivity analyses did not materially change results, therefore, these concerns are attenuated. Finally, our results may only apply to prostate cancer survivors who survived at least two and up to three years’ post-diagnosis and those with stage T2 or higher disease. Although this issue limits our generalizability, it was necessary to define a homogeneous population. While it is important to take into account limitations, our study strengths need also be recognized when interpreting these results. We utilized an innovative analysis technique which in theory does make one distributional assumption based on the ceiling and floor effects of the tobit distribution of QoL scores over time. This study is the first to identify separate groups of prostate cancer survivors following different post-diagnosis physical and mental QoL trajectories. In addition, we addressed biases inherent in longterm prospective cohort studies by identifying the role of dropping out or mortality during active data collection. Dropout proved to be important in this study population. 123 While modelling dropout, we attempted to systematically characterize the trajectory groups by several behavioural and prognostic factors measured at diagnosis or over the diagnosis period. Future research will be able to capitalize on these findings in order to target at-risk groups of prostate cancer survivors with low and declining QoL trajectories. 4.7 Conclusion Based on these analyses there were three distinct trajectories of physical and mental QoL up to 10 years post-prostate cancer diagnosis. Confirmation of these findings is warranted in this population to determine which prostate cancer survivors are truly experiencing low and/or declining QoL relative to average-maintaining QoL. Active surveillance may be an option to minimize aggressive treatment regiments. Characteristics of QoL trajectories will lead to a better understanding of differences between groups and how health professionals and researchers can use this information to plan future interventions to specifically target low/medium and/or declining QoL populations. 4.8 Manuscript acknowledgements Study coordination was done by Aleata Ryhorchuk and Sana Fakih. Interviewers for this study were Jodi Parrotta, Linda Davison, Pearl Cooke, Nicole Slot, Carol-Anne Zawalykut, Catherine Munro. Data entry was done by Carla Quesnel and chart abstraction by Yvonne LeBlanc and Sarah MacLaughlin. Dr. Steven J. Angyalfi provided clinical expertise regarding prostate cancer treatment and care. 124 4.9 References 1. Ferlay J SI, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman D, Bray, F. GLOBOCAN 2012 v1.0, Cancer Incidence and Mortality Worldwide: IARC CancerBase No. 11 Lyon, France: International Agency for Research on Cancer 2013. 2. Bray F RJ, Masuyer E, Ferlay J. Estimates of global cancer prevalence for 27 sites in the adult population in 2008. Int J Cancer. 2013;132(5):1133-45. 3. Hoffman RM. Clinical practice. Screening for prostate cancer. N Engl J Med. 2011;365(21):2013-9. 4. Brawer MK. Screening and early detection of prostate cancer will decrease morbidity and mortality from prostate cancer: the argument for. Eur Urol. 1996;29 Suppl 2:19-23. 5. Hall RR. Screening and early detection of prostate cancer will decrease morbidity and mortality from prostate cancer: the argument against. Eur Urol. 1996;29 Suppl 2:246. 6. Wilt TJ, Brawer MK, Jones KM, Barry MJ, Aronson WJ, Fox S, et al. Radical Prostatectomy versus Observation for Localized Prostate Cancer. N Engl J Med. 2012;367(3):203-13. 7. Drummond FJ, Kinnear H, O'Leary E, Donnelly, Gavin A, Sharp L. Long-term health-related quality of life of prostate cancer survivors varies by primary treatment. Results from the PiCTure (Prostate Cancer Treatment, your experience) study. J Cancer Surviv. 2015;9(2):361-72. 8. Carlsson S, Drevin L, Loeb S, Widmark A, Lissbrant IF, Robinson D, et al. Population-based study of long-term functional outcomes after prostate cancer treatment. BJU Int. 2016;117(6b):E36-E45. 9. Sun F, Oyesanmi O, Fontanarosa J, Reston J, Guzzo T, Schoelles K. Therapies for Clinically Localized Prostate Cancer: Update of a 2008 Systematic Review. Comparative Effectiveness Reviews No. 146. Report No. 15-EHC004-EF. Rockville (MD): Agency for Healthcare Research and Quality, 2014. 10. Davis KM, Kelly SP, Luta G, Tomko C, Miller AB, Taylor KL. The association of long-term treatment-related side effects with cancer-specific and general quality of life among prostate cancer survivors. Urology. 2014;84(2):300-6. 11. Abrahamsson PA. Potential benefits of intermittent androgen suppression therapy in the treatment of prostate cancer: a systematic review of the literature. Eur Urol. 2010;57(1):49-59. 125 12. Bourke L, Boorjian SA, Briganti A, Klotz L, Mucci L, Resnick MJ, et al. Survivorship and Improving Quality of Life in Men with Prostate Cancer. Eur Urol. 2015;68(3):374-83. 13. Parker PA, Davis JW, Latini DM, Baum G, Wang X, Ward JF, et al. Relationship between illness uncertainty, anxiety, fear of progression and quality of life in men with favourable-risk prostate cancer undergoing active surveillance. BJU Int. 2016;117(3):469-77. 14. Matthew AG, Alibhai SM, Davidson T, Currie KL, Jiang H, Krahn M, et al. Health-related quality of life following radical prostatectomy: long-term outcomes. Qual LIfe Res. 2014;23(8):2309-17. 15. Selli C, Bjartell A, Burgos J, Somerville M, Palacios JM, Benjamin L, et al. Burden of Illness in Prostate Cancer Patients with a Low-to-Moderate Risk of Progression: A One-Year, Pan-European Observational Study. Prostate Cancer. 2014;2014:472949. 16. Yamamoto S, Fujii Y, Masuda H, Urakami S, Saito K, Kozuka T, et al. Longitudinal change in health-related quality of life after intensity-modulated radiation monotherapy for clinically localized prostate cancer. Qual LIfe Res. 2014;23(5):1641-50. 17. Ernstmann N, Weissbach L, Herden J, Winter N, Ansmann L. Patient-physiciancommunication and health related quality of life of localized prostate cancer patients undergoing radical prostatectomy - a longitudinal multilevel analysis. BJU Int. 2016; [Epub ahead of print]. 18. Nagin DS, Odgers CL. Group-based trajectory modeling in clinical research. Annu Rev Clin Psychol. 2010;6:109-38. 19. Anderson CB, Kaufman MR, Dietrich MS, Barocas DA, Chang SS, Cookson MS, et al. Recovery of urinary function after radical prostatectomy: identification of trajectory cluster groups. J Urol. 2012;187(4):1346-51. 20. Legler JM, Davis WW, Potosky AL, Hoffman RM. Latent variable modelling of recovery trajectories: sexual function following radical prostatectomy. Stat Med. 2004;23(18):2875-93. 21. Friedenreich CM, Wang Q, Neilson HK, Kopciuk KA, McGregor SE, Courneya KS. Physical Activity and Survival After Prostate Cancer. Eur Urol. 2016;S03022838(15):01241-5. 22. Antonescu I, Carli F, Mayo NE, Feldman LS. Validation of the SF-36 as a measure of postoperative recovery after colorectal surgery. Surg Endos. 2014;28(11):3168-78. 23. Jaeschke R, Singer J, Guyatt GH. Measurement of health status. Ascertaining the minimal clinically important difference. Control Clin Trials. 1989;10(4):407-15. 126 24. Friedenreich CM, KS. C, Bryant HE. The lifetime total physical activity questionnaire: development and reliability. 2004;159(8):749. 25. McGregor SE, Courneya KS, Kopciuk KA, Tosevski C, Friedenreich CM. Casecontrol study of lifetime alcohol intake and prostate cancer risk. Cancer Causes Control. 2013;24(3):451-61. 26. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40(5):373-83. 27. Showcase RC, Jones BL, Nagin DS, editors. A Stata Plugin for Estimating GroupBased Trajectory Models, 2012. 28. Schwarz G. Estimating the Dimension of a Model. Ann Statist. 1978:461-4. 29. Haviland AM, Jones BL, Nagin DS. Group-based Trajectory Modeling Extended to Account for Nonrandom Participant Attrition. Socio Meth Res. 2011;40(2):367-90. 30. Dieperink KB, Hansen S, Wagner L, Johansen C, Andersen KK, Hansen O. Living alone, obesity and smoking: important factors for quality of life after radiotherapy and androgen deprivation therapy for prostate cancer. Acta Oncol. 2012;51(6):722-9. 31. Hampson LA, Cowan JE, Zhao S, Carroll PR, Cooperberg MR. Impact of age on quality-of-life outcomes after treatment for localized prostate cancer. Eur Urol. 2015;68(3):480-6. 32. Herkommer K, Schmid SC, Schulwitz H, Dinkel A, Klorek T, Hofer C, et al. Health-related quality of life after radical prostatectomy depends on patient's age but not on comorbidities. Urol Oncol. 2015;33(6):266.e1-7. 33. Kendal WS. Age Bias in Time From Diagnosis Comparisons of Prostate Cancer Treatment. Am J Clin Oncol. 2016;[Epub ahead of print]. 34. Sullivan PW, Nelson JB, Mulani PM, Sleep D. Quality of life as a potential predictor for morbidity and mortality in patients with metastatic hormone-refractory prostate cancer. Qual LIfe Res. 2006;15(8):1297-306. 35. Ware JE, Jr., Gandek B, Kosinski M, Aaronson NK, Apolone G, Brazier J, et al. The equivalence of SF-36 summary health scores estimated using standard and countryspecific algorithms in 10 countries: results from the IQOLA Project. International Quality of Life Assessment. J Clin Epidemiol. 1998;51(11):1167-70. 36. Serda IFBC, Valle AD, Marcos-Gragera R. Prostate cancer and quality of life: analysis of response shift using triangulation between methods. J Gerontol Nurs. 2014;40(6):32-41. 127 37. Gerlich C, Schuler M, Jelitte M, Neuderth S, Flentje M, Graefen M, et al. Prostate cancer patients' quality of life assessments across the primary treatment trajectory: 'True' change or response shift? Acta Oncol. 2016:1-7. 128 4.10 Tables and Figures Table 4.1. Model selection to determine number of groups for physical quality of life postdiagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Number of groups BIC Base model Estimated % in each group BIC 1 1 -7498.93 100 2 -7277.90 1 -221.03 42.09 3a -7238.69 2 -39.21 25.61 4 -7239.99 3 1.3 6.36 Abbreviation: BIC = Bayesian Information Criterion. a BIC value closest to zero. 129 2 3 4 57.91 40.19 27.17 34.20 35.95 30.51 Table 4.2. Model selection for determining linear or quadratic structure of trajectories for physical quality of life post-diagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Model BIC Base model a (1 1 1) -7228.18 (1 1 2) -7231.79 1 (1 2 1) -7231.23 2 (2 1 1) -7231.95 3 (1 2 2) -7234.93 4 (2 1 2) -7235.56 5 (2 2 1) -7235.00 6 (2 2 2) -7238.69 7 Abbreviation: BIC = Bayesian Information Criterion. a BIC value closest to zero. 130 BIC 3.61 3.05 3.05 3.77 7.38 6.82 10.51 Table 4.3. Model selection to determine number of groups for mental quality of life post-diagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (19972007). Number of groups BIC Base model BIC Estimated % in each group 1 1 -7410.63 100 2 -7123.37 1 -287.26 20.17 3a -7098.90 2 -24.47 14.51 4b -7114.01 3 15.11 14.51 Abbreviation: BIC = Bayesian Information Criterion. 2 3 4 79.83 9.69 9.69 75.80 75.80 0 a BIC value closest to zero. b Four groups warning: variance matrix is non-symmetric or highly singular. 131 Table 4.4. Model selection for determining linear or quadratic structure of trajectories for mental quality of life post-diagnosis in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Model BIC Base model a (1 1 1) -7088.81 (1 1 2) -7091.94 1 (1 2 1) -7092.03 2 (2 1 1) -7092.58 3 (1 2 2) -7095.62 4 (2 1 2) -7095.71 5 (2 2 1) -7095.21 6 (2 2 2) -7098.90 7 Abbreviation: BIC = Bayesian Information Criterion. a BIC value closest to zero. 132 BIC 3.13 3.22 3.77 6.81 6.9 6.4 10.09 Table 4.5. Descriptive characteristics by physical quality of life trajectory group of prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Characteristics Number of participants at first follow-up Number of participants at second follow-up Number of participants at third follow-up Type of Cancer Aggressive Non-aggressive Hormone therapy a Yes No/missing Radiation Therapy Yes No Prostatectomy Yes No Smoking status Current Former/never Physical Quality of Life Trajectory Groups Average-maintaining Low-declining Very low-maintaining 259 (32%) 208 (35%) 174 (35%) N (%) 330 (40%) 242 (41%) 217 (44%) 228 (28%) 145 (24%) 104 (21%) 36 (14%) 224 (86%) 70 (21%) 264 (79%) 49 (22%) 174 (78%) 131 (50%) 129 (50%) 227 (68%) 107 (32%) 159 (71%) 64 (29%) 93 (36%) 167 (64%) 168 (50%) 166 (50%) 98 (44%) 125 (56%) 140 (54%) 120 (46%) 247 (74%) 87 (26%) 190 (85%) 33 (15%) 26 (10%) 234 (90%) 53 (16%) 281 (84%) 35 (16%) 188 (84%) Median (Q1-Q3) Age at diagnosis Post-diagnosis Charlson co-morbidity score Pre-diagnosis lifetime PA b PA at first follow-up b 64 (59-70) 1 (0.5-2) 172 (121-225) 98 (58-140) 133 68 (63-73) 2 (1-3) 138 (98-193) 77 (46-124) 72 (66-75) 3 (2-4) 126 (84-171) 64 (29-107) PA at second follow-up b 89 (46-143) PA at third follow-up b 75 (45-124) BMI 27 (25-29) Average total alcohol intake (g/year) c 1747 (451-4368) Average total energy intake (kcal/day) 2033 (1623-2465) Abbreviations: BMI = body mass index; PA = physical activity. 74 (40-134) 57 (25-99) 28 (26-31) 2010 (695-4679) 1998 (1659-2475) 43 (15-91) 42 (15-90) 29 (26-31) 1758 (572-4259) 1970 (1644-2493) a Included bilateral orchiectomy, luteinizing hormone-releasing agonists, nonsteroidal antiandrogens, steroidal antiandrogens. b Physical activity measurements are: total MET-hour/week/year. c 66 never-drinkers. 134 Table 4.6. Descriptive characteristics by mental quality of life trajectory group of prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Mental QoL Trajectory Groups Characteristics Number of participants at first follow-up Number of participants at second follow-up Number of participants at third follow-up Type of Cancer Aggressive Non-aggressive Hormone therapy a Yes No/missing Radiation Therapy Yes No Prostatectomy Yes No Smoking status Current Former/never Average-maintaining Above average-declining Low-increasing 568 (70%) 411 (69%) 368 (74%) N (%) 134 (16%) 108 (18%) 75 (15%) 115 (14%) 76 (13%) 52 (11%) 112 (20%) 451 (80%) 30 (22%) 107 (88%) 13 (11%) 104 (89%) 339 (60%) 224 (40%) 108 (79%) 29 (21%) 70 (60%) 47 (40%) 253 (45%) 310 (55%) 63 (46%) 74 (54%) 43 (37%) 74 (63%) 371 (66%) 192 (34%) 125 (91%) 12 (9%) 81 (69%) 36 (31%) 64 (11%) 499 (89%) 25 (18%) 112 (82%) 25 (21%) 92 (79%) Median (Q1-Q3) Age at diagnosis Post-diagnosis Charlson co-morbidity score PA at first follow-up b Pre-diagnosis lifetime PA b 67 (61-72) 2 (1-3) 80 (47-128) 161 (111-207) 135 72 (67-76) 3 (2-4) 68 (36-112) 155 (111-210) 67 (61-73) 2 (1-3) 79 (43-130) 137 (95-190) PA at second follow-up b 77 (44-140) PA at third follow-up b 69 (37-111) BMI 28 (25-30) Average total alcohol intake (g/year) c 1835 (529-4430) Average total energy intake (kcal/day) 2009 (1644-2460) Abbreviations: BMI = body mass index; PA = physical activity. 40 (16-97) 35 (6-92) 28 (25-31) 1858 (628-4294) 1990 (1596-2374) 62 (17-111) 34 (16-78) 28 (26-31) 2105 (822-5137) 2013 (1673-2669) a Included bilateral orchiectomy, luteinizing hormone-releasing agonists, nonsteroidal antiandrogens, steroidal antiandrogens. b Physical activity measurements are: total MET-hour/week/year. c 66 never-drinkers. 136 Table 4.7. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the average-maintaining/increasing QoL group in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Prognostic/behavioural factors Low/above average-declining versus average-maintaining Quality of Life RRR a Very low/low-maintaining/increasing versus average-maintaining Quality of Life 95% CI RRR a 95% CI 1.11 1.50 0.64 1.99 2.29 2.04 1.18 2.77 1.06-1.16 0.76-2.95 0.35-1.17 1.14-3.48 1.07-4.89 1.66-2.51 1.09-1.27 1.23-6.26 1.00 0.40 0.50 1.01 1.73 1.18 1.04 2.35 0.97-1.04 0.19-0.86 0.28-0.89 0.61-1.67 0.89-3.36 1.03-1.36 0.97-1.10 1.27-4.36 Physical QoL trajectories Age at diagnosis (years) 1.04 0.99-1.08 b Aggressiveness of disease 1.41 0.72-2.79 b Radiation therapy 1.10 0.57-2.12 b Hormone therapy 1.76 1.02-3.02 b Prostatectomy 1.25 0.58-2.68 Charlson co-morbidity score 1.59 1.29-1.95 BMI (kg/m2) 1.09 1.02-1.17 b Smoking status 1.71 0.75-3.90 Mental QoL trajectories Age at diagnosis (years) 1.06 1.01-1.11 b Aggressiveness of disease 0.89 0.44-1.81 Radiation therapy b 0.57 0.30-1.10 Hormone therapy b 1.61 0.82-3.16 b Prostatectomy 3.32 1.23-8.94 Charlson co-morbidity score 1.32 1.12-1.56 2 BMI (kg/m ) 1.04 0.96-1.12 Smoking status b 1.79 0.75-4.24 Abbreviations: BMI = body mass index; RRR = relative risk ratio; CI = confidence interval. a All models were adjusted for time varying total physical activity, dropout probabilities and all other factors in table. 137 b Dichotomized variables. 138 Table 4.8. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a timevarying covariate in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Trajectory groups Physical quality of life trajectories Average-maintaining Low-declining Very low-maintaining Mental quality of life trajectories Average-maintaining Above average-declining Low-increasing a p-value < 0.001. b Estimated mean coefficients (Standard Error) Baseline intercept Slope Physical activity slope 49.64 (1.38) a 41.68 (1.16) a 24.37 (1.28) a -0.03 (0.49) -1.85 (0.46) a 0.19 (0.55) 0.02 (0.01) b 0.02 (0.01) a 0.03 (0.01) b 54.59 (0.76) a 57.32 (2.37) a 25.99 (1.97) a 0.45 (0.30) -6.98 (1.13) a 2.62 (0.84) b 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) p-value < 0.01. 139 Figure 4.1. Repeated measurements and cohort timeline. Past two years of physical activity were recalled and past four weeks of QoL (N = complete QoL measurements at each time point). Abbreviations: PA = physical activity, QoL = quality of life, Dx = diagnosis. 140 Figure 4.2. Final adjusted model selected for physical QoL trajectory groups (black lines are trajectory groups, grey lines are 95% confidence and dot symbols are observed group means at each time-point) in cohort of prostate cancer survivors in Alberta, Canada (n=817). Abbreviations: QoL=quality of life, DP=dropout probability. 141 Figure 4.3. Final adjusted model selected for mental QoL trajectory groups (black lines are trajectory groups, grey lines are 95% confidence intervals and dot symbols are observed group means at each time-point) in cohort of prostate cancer survivors in Alberta, Canada (n=817). Abbreviations: QoL=quality of life, DP=dropout probability. 142 4.11 Additional analyses and results: Sensitivity analysis tables and figures 4.11.1 Dropout within trajectory groups As an exploratory and sensitivity analysis, we examined the association between the estimated log odds of dropping out on behavioural/prognostic factors within each trajectory group. This analysis was performed to determine if certain factors were associated with dropping out in each trajectory group. These results are described in the manuscript. 143 Table 4.9. Logistic regression models assessing the relationship between dropping out and covariates within the very low-maintaining physical QoL trajectory group. Characteristics OR 95% CI Physical activity at first follow-up 1.00 0.99, 1.00 Age at diagnosis (years) 0.98 0.92, 1.02 a 1.92 0.89, 4.12 Aggressiveness of disease 0.55 0.29, 1.04 Radiation therapy a a 1.10 0.56, 2.15 Hormone therapy a 1.46 0.51, 4.17 Prostatectomy Charlson co-morbidity score 0.96 0.83, 1.12 2 BMI (kg/m ) 0.91 0.85, 0.98 3.89 1.39, 10.90 Smoking status a Note: All analyses are mutually adjusted for other variables in table. P-value 0.24 0.54 0.10 0.07 0.78 0.48 0.63 0.013 0.010 Abbreviations: BMI= body mass index; OR= odds ratio; CI= confidence interval. a Dichotomous variables. 144 Table 4.10. Logistic regression models assessing the relationship between dropping out and covariates within the low-declining physical QoL trajectory group. Characteristics OR 95% CI Physical activity at first follow-up 1.02 0.99, 1.00 Age at diagnosis (years) 1.02 0.98, 1.06 a 1.25 0.71, 2.21 Aggressiveness of disease a 0.52 0.30, 0.91 Radiation therapy 1.35 0.80, 2.28 Hormone therapy a a 2.30 1.16, 4.57 Prostatectomy Charlson co-morbidity score 0.87 0.74, 1.02 BMI (kg/m2) 0.99 0.93, 1.05 a 1.77 0.94, 3.36 Smoking status Note: All analyses are mutually adjusted for other variables in table. P-value 0.78 0.25 0.43 0.022 0.27 0.018 0.08 0.67 0.08 Abbreviations: BMI= body mass index; OR= odds ratio; CI= confidence interval. a Dichotomous variables. 145 Table 4.11. Logistic regression models assessing the relationship between dropping out and covariates within the average-maintaining physical QoL trajectory group. Characteristics OR 95% CI Physical activity at first follow-up 1.00 0.99, 1.00 Age at diagnosis (years) 1.02 0.98, 1.07 a 0.95 0.43, 2.07 Aggressiveness of disease 0.59 0.29, 1.23 Radiation therapy a a 0.91 0.50, 1.65 Hormone therapy a 2.48 1.11, 5.55 Prostatectomy Charlson co-morbidity score 0.73 0.57, 0.93 2 BMI (kg/m ) 1.05 0.96, 1.14 2.99 1.23, 7.31 Smoking status a Note: All analyses are mutually adjusted for other variables in table. P-value 0.98 0.28 0.89 0.16 0.77 0.028 0.012 0.31 0.016 Abbreviations: BMI= body mass index; OR= odds ratio; CI= confidence interval. a Dichotomous variables. 146 Table 4.12. Logistic regression models assessing the relationship between dropping out and covariates within the low-increasing mental QoL trajectory group. Characteristics OR 95% CI Physical activity at first follow-up 0.99 0.99, 1.00 Age at diagnosis (years) 1.04 0.97, 1.12 2.38 0.56, 10.06 Aggressiveness of disease a 0.86 0.30, 2.45 Radiation therapy a a 0.80 0.32, 1.97 Hormone therapy 2.36 0.64, 8.75 Prostatectomy a Charlson co-morbidity score 0.79 0.62, 1.02 BMI (kg/m2) 0.95 0.85, 1.07 a 3.48 1.11, 10.88 Smoking status Note: All analyses are mutually adjusted for other variables in table. P-value 0.18 0.25 0.24 0.78 0.62 0.20 0.07 0.38 0.032 Abbreviations: BMI= body mass index; OR= odds ratio; CI= confidence interval. a Dichotomous variables. 147 Table 4.13. Logistic regression models assessing the relationship between dropping out and covariates within the above average-declining mental QoL trajectory group. Characteristics OR 95% CI Physical activity at first follow-up 0.99 0.99, 1.00 Age at diagnosis (years) 1.02 0.95, 1.10 1.32 0.52, 3.34 Aggressiveness of disease a a 0.63 0.28, 1.40 Radiation therapy a 1.30 0.51, 3.28 Hormone therapy 2.46 0.59, 10.31 Prostatectomy a Charlson co-morbidity score 0.93 0.77, 1.12 BMI (kg/m2) 0.97 0.89, 1.06 a 2.88 0.93, 8.88 Smoking status Note: All analyses are mutually adjusted for other variables in table. P-value 0.20 0.52 0.56 0.25 0.58 0.22 0.45 0.50 0.07 Abbreviations: BMI= body mass index; OR= odds ratio; CI= confidence interval. a Dichotomous variables. 148 Table 4.14. Logistic regression models assessing the relationship between dropping out and covariates within the average-maintaining mental QoL trajectory group. Characteristics OR 95% CI Physical activity at first follow-up 1.00 0.99, 1.00 Age at diagnosis (years) 1.02 0.99, 1.05 a 1.38 0.89, 2.15 Aggressiveness of disease a 0.50 0.32, 0.78 Radiation therapy a 1.26 0.85, 1.86 Hormone therapy a 2.12 1.26, 3.58 Prostatectomy Charlson co-morbidity score 0.93 0.82, 1.05 2 BMI (kg/m ) 1.00 0.95, 1.05 1.91 1.09, 3.35 Smoking status a Note: All analyses are mutually adjusted for other variables in table. P-value 0.91 0.23 0.15 0.002 0.26 0.005 0.22 0.99 0.024 Abbreviations: BMI= body mass index; OR= odds ratio; CI= confidence interval. a Dichotomous variables. 149 4.11.2 Sensitivity analyses: Complete case analysis for physical and mental QoL trajectories Another sensitivity analysis was done to determine if only including prostate cancer survivors with complete data at all three time-points influenced the model selection, dropout probabilities and behavioural/prognostic factors. While the number and function of trajectories were similar, it is interesting to note that the function form was better fit with a quadratic function in the low-declining group of physical QoL. The quadratic term was not a significantly better fit as the change in BIC was <2, however, it is interesting to point out that there may even be more variation in that particular trajectory. Further, smoking status was no longer associated with the very lowmaintaining or low-increasing groups. This might be attributed to the fact that men who were current smokers were more likely to dropout/die during active data collection, as smokers tend to have other poor health habits that may increase their mortality. Other associations were attenuated compared to the total sample (n=817), however, no other substantial changes were noted. 150 Table 4.15 Model selection to determine number of groups for physical quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (19972007). Number of groups BIC 1 -5278.86 2 -5096.88 1 a 3 -5057.43 2 4 -5062.24 3 a BIC value closest to zero. Base model BIC -181.98 -39.45 4.81 151 Estimated % in each group 1 100 38.32 19.70 3.75 2 3 4 61.68 42.45 21.19 37.85 39.81 35.26 Table 4.16. Model selection for determining linear or quadratic structure of trajectories for physical quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007). Model (1 1 1) a (1 1 2) (1 2 1) a (2 1 1) (1 2 2) (2 1 2) (2 2 1) (2 2 2) a BIC value closest to zero. BIC -5050.82 -5054.31 -5050.64 -5054.30 -5054.24 -5057.79 -5053.83 -5057.43 Base model 1 2 3 4 5 6 7 152 BIC 3.49 -0.18 3.48 3.42 6.97 3.01 6.61 Table 4.17. Model selection to determine number of groups for mental quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (19972007). Number of groups BIC 1 -5259.52 2 -5027.17 1 a 3 -5001.75 2 4 -5016.19 3 a BIC value closest to zero. Base model BIC -232.35 -25.42 14.44 153 Estimated % in each group 1 100 18.22 8.72 8.72 2 3 4 81.78 19.11 19.12 72.17 72.17 <0.01 Table 4.18. Model selection for determining linear or quadratic structure of trajectories for mental quality of life post-diagnosis in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007). Model (1 1 1) a (1 1 2) (1 2 1) (2 1 1) (1 2 2) (2 1 2) (2 2 1) (2 2 2) a BIC value closest to zero. BIC -4991.66 -4994.83 -4995.25 -4994.98 -4998.35 -4998.15 -4998.59 -5001.75 Base model 1 2 3 4 5 6 7 154 BIC 3.17 3.59 3.32 6.69 6.49 6.93 10.09 Table 4.19. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the average-maintaining/increasing QoL group in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007). Prognostic/behavioural factors Low/above average-declining versus Average-maintaining Quality of Life RRR a 95% CI Very low/low-maintaining/increasing versus Average-maintaining Quality of Life RRR a Physical QoL trajectories Age at diagnosis (years) 1.06 1.01, 1.11 1.21 b Aggressiveness of disease 1.02 0.45, 2.33 0.62 b Radiation therapy 1.38 0.62, 3.05 0.79 Hormone therapy b 1.54 0.83, 2.85 1.54 b Prostatectomy 1.16 0.49, 2.75 2.63 Charlson co-morbidity score 1.48 1.19, 1.84 1.98 2 BMI (kg/m ) 1.14 1.04, 1.25 1.36 Smoking status b 2.01 0.74, 5.46 1.91 Mental QoL trajectories Age at diagnosis (years) 1.04 0.98, 1.10 0.97 b Aggressiveness of disease 0.68 0.28, 1.65 0.24 Radiation therapy b 0.57 0.26, 1.22 0.46 b Hormone therapy 1.40 0.68, 2.89 1.18 Prostatectomy b 2.31 0.86, 6.21 1.16 Charlson co-morbidity score 1.47 1.22, 1.77 1.32 BMI (kg/m2) 1.04 0.95, 1.13 1.06 b Smoking status 0.99 0.31, 3.19 0.98 Abbreviations: BMI = body mass index; RRR = relative risk ratio; CI = confidence interval. a 95% CI 1.12, 1.30 0.22, 1.76 0.33, 1.88 0.70, 3.35 0.89, 7.78 1.51, 2.58 1.21, 1.54 0.44, 8.24 0.92, 1.03 0.05, 1.09 0.16, 1.30 0.54, 2.59 0.39, 3.48 1.05, 1.66 0.96, 1.17 0.26, 3.71 All models were adjusted for time varying total physical activity, dropout probabilities and all other factors in table. 155 b Dichotomized variables. 156 Table 4.20. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a time-varying covariate in prostate cancer survivors (n = 454) in Alberta, Canada (1997-2007). Trajectory groups Physical quality of life trajectories Average-maintaining Low-declining Very low-maintaining Mental quality of life trajectories Average-maintaining Above average-declining Low-increasing ***p-value<0.001 Estimated mean coefficients (Standard Error) Baseline intercept Slope Physical activity slope 51.16 (1.24)*** 43.37 (1.22)*** 29.56 (1.66)*** -0.37 (0.46) -2.19 (0.48)*** -1.65 (0.66)* 0.01 (0.01)* 0.02 (0.01)* 0.02 (0.01)** 54.85 (0.83)*** 49.84 (2.39)*** 22.32 (3.20)*** 0.49 (0.31) -3.90 (1.04)*** 3.54 (1.17)** 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) **p-value<0.01 *p-value<0.05 157 4.11.3 Sensitivity analyses: Two complete assessments analysis for physical and mental QoL trajectories This second sensitivity analysis was completed for similar purposes of the complete-case sample. These results had little to no effects on the model selection, dropout probabilities nor the associations of behavioural/prognostic factors with group membership. The complete case analysis and the two complete assessments analysis speak to the robustness of the GBTM modelling strategy to effectively detect meaningful subgroups in a study population. 158 Table 4.21. Model selection to determine number of groups for physical quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (19972007). Number of groups BIC 1 -6761.16 2 -6543.64 1 a 3 -6505.65 2 4 -6506.51 3 a BIC value closest to zero. Base model BIC -217.52 -37.99 0.86 159 Estimated % in each group 1 100 40.50 22.96 4.49 2 3 4 59.50 41.30 25.90 35.74 37.53 32.09 Table 4.22. Model selection for determining linear or quadratic structure of trajectories for physical quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007). Model (1 1 1) a (1 1 2) (1 2 1) (2 1 1) (1 2 2) (2 1 2) (2 2 1) (2 2 2) a BIC value closest to zero. BIC -6495.42 -6499.08 -6498.63 -6498.91 -6502.34 -6502.57 -6501.94 -6505.65 Base model 1 2 3 4 5 6 7 160 BIC 3.66 3.21 3.49 6.92 7.15 6.52 10.23 Table 4.23. Model selection to determine number of groups for mental quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (19972007). Number of groups BIC 1 -6700.98 2 -6429.25 1 a 3 -6404.44 2 4b -6419.35 3 a BIC value closest to zero. b Base model BIC -271.73 -24.81 14.91 Estimated % in each group 1 100 19.32 11.57 11.57 2 3 4 80.67 11.86 11.89 76.57 76.57 <0.01 Four groups warning: variance matrix is non-symmetric or highly singular. 161 Table 4.24. Model selection for determining linear or quadratic structure of trajectories for mental quality of life post-diagnosis in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007). Model (1 1 1) a (1 1 2) (1 2 1) (2 1 1) (1 2 2) (2 1 2) (2 2 1) (2 2 2) a BIC value closest to zero. BIC -6394.25 -6397.35 -6397.45 -6397.97 -6400.86 -6401.07 -6400.81 -6404.44 Base model 1 2 3 4 5 6 7 162 BIC 3.1 3.2 3.72 6.61 6.82 6.56 10.19 Table 4.25. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the average-maintaining/increasing QoL group in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007). Prognostic/behavioural factors Low/above average-declining versus Average-maintaining Quality of Life RRR a 95% CI Very low/low-maintaining/increasing versus Average-maintaining Quality of Life RRR a Physical QoL trajectories Age at diagnosis (years) 1.04 0.99, 1.09 1.16 b Aggressiveness of disease 1.17 0.57, 2.41 0.84 b Radiation therapy 1.13 0.58, 2.23 0.70 Hormone therapy b 1.74 1.00, 3.03 2.03 b Prostatectomy 1.06 0.49, 2.29 2.23 Charlson co-morbidity score 1.56 1.27, 1.92 2.15 2 BMI (kg/m ) 1.11 1.03, 1.20 1.25 Smoking status b 1.97 0.81, 4.78 2.33 Mental QoL trajectories Age at diagnosis (years) 1.06 1.01, 1.11 0.98 Aggressiveness of disease b 0.93 0.45, 1.92 0.48 Radiation therapy b 0.54 0.28, 1.02 0.47 b Hormone therapy 1.56 0.80, 3.04 1.01 Prostatectomy b 3.57 1.34, 9.46 1.42 Charlson co-morbidity score 1.33 1.13, 1.57 1.28 BMI (kg/m2) 1.04 0.96, 1.12 1.03 b Smoking status 1.83 0.78, 4.33 2.07 Abbreviations: BMI = body mass index; RRR = relative risk ratio; CI = confidence interval. a 95% CI 1.09, 1.23 0.37, 1.91 0.34, 1.41 1.07, 3.88 0.89, 5.55 1.71, 2.70 1.14, 1.36 0.83, 6.52 0.94, 1.03 0.20, 1.13 0.23, 0.96 0.56, 1.81 0.64, 3.13 1.08, 1.51 0.96, 1.11 0.97, 4.43 All models were adjusted for time varying total physical activity, dropout probabilities and all other factors in table. 163 b Dichotomized variables. 164 Table 4.26. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a time-varying covariate in prostate cancer survivors (n = 636) in Alberta, Canada (1997-2007). Trajectory groups Physical quality of life trajectories Average-maintaining Low-declining Very low-maintaining Mental quality of life trajectories Average-maintaining Above average-declining Low-increasing ***p-value<0.001 Estimated mean coefficients (Standard Error) Baseline intercept Slope Physical activity slope 49.85 (1.34)*** 42.34 (1.19)*** 25.75 (1.41)*** -0.13 (0.48) -2.04 (0.47)*** -0.39 (0.60) 0.01 (0.01)** 0.02 (0.01)* 0.03 (0.01)** 54.59 (0.79)*** 57.35 (2.27)*** 24.13 (2.32)*** 0.44 (0.31) -7.12 (1.07)*** 3.20 (0.94)*** 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) **p-value<0.01 *p-value<0.05 165 4.11.4 Sensitivity analysis: Not modelling dropout on covariate significance analysis Similar to the above, this analysis was used as a tool to test the GBTM analysis for robustness by modelling behavioural/prognostic factors with and without the logistic regression dropout model that was based on previous outcome. Associations of groupmembership in regards to behavioural/prognostic factors did not change statistically or meaningfully between this model and the fully adjusted model including dropout. 166 Table 4.27. Multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the average-maintaining/increasing QoL group without modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Prognostic/behavioural factors Low/above average-declining versus Average-maintaining Quality of Life RRR a 95% CI Very low/low-maintaining/increasing versus average-maintaining Quality of Life RRR a Physical QoL trajectories Age at diagnosis (years) 1.06 1.01, 1.10 1.12 b Aggressiveness of disease 1.34 0.69, 2.60 1.47 b Radiation therapy 1.02 0.53, 1.94 0.56 b Hormone therapy 1.53 0.90, 2.60 2.01 Prostatectomy b 1.52 0.74, 3.15 2.49 Charlson co-morbidity score 1.49 1.23, 1.82 1.95 2 BMI (kg/m ) 1.08 1.01, 1.16 1.18 b Smoking status 1.99 0.91, 4.34 3.55 Mental QoL trajectories Age at diagnosis (years) 1.05 1.00, 1.10 0.99 Aggressiveness of disease b 0.92 0.44, 1.90 0.33 b Radiation therapy 0.79 0.41, 1.52 0.53 Hormone therapy b 1.40 0.70, 2.79 0.98 Prostatectomy b 2.40 0.89, 6.45 1.55 Charlson co-morbidity score 1.35 1.15, 1.59 1.18 BMI (kg/m2) 1.03 0.96, 1.12 1.04 b Smoking status 1.17 0.46, 3.01 2.18 Abbreviations: BMI = body mass index; RRR = relative risk ratio; CI = confidence interval. a All models were adjusted for time varying total physical activity and all other factors in table. 167 95% CI 1.07, 1.17 0.74, 2.90 0.30, 1.04 1.14, 3.57 1.17, 5.33 1.60, 2.39 1.09, 1.27 1.60, 7.88 0.95, 1.02 0.14, 0.79 0.29, 0.97 0.58, 1.66 0.78, 3.07 1.02, 1.36 0.97, 1.10 1.18, 4.04 b Dichotomized variables. 168 Table 4.28. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory models including physical activity as a time-varying covariate and not modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Trajectory groups Physical quality of life trajectories Average-maintaining Low-declining Very low-maintaining Mental quality of life trajectories Average-maintaining Above average-declining Low-increasing ***p-value<0.001 Estimated mean coefficients (Standard Error) Baseline intercept Slope Physical activity slope 50.59 (1.17)*** 41.63 (1.14)*** 24.32 (1.30)*** -0.37 (0.43) -1.90 (0.43)*** -0.01 (0.57) 0.01 (0.01)* 0.02 (0.01)** 0.03 (0.01)*** 54.62 (0.72)*** 52.28 (2.83)*** 24.64 (2.17)*** 0.41 (0.28) -5.39 (1.27)*** 3.04 (0.91)*** 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) **p-value<0.01 *p-value<0.05 169 4.11.5 Sensitivity analyses: Dropout switchers between trajectory groups with and without modelling dropout These sensitivity analyses were elaborated on in the manuscript and did not substantially contribute to the interpretations of these data as there was a small number of prostate cancer survivors who switched trajectory groups with and without modelling dropout. These results speak to the resiliency of the GBTM method and its utility to define trajectories in a study population. 170 Table 4.29. Participants who switched physical quality of life trajectory groups when modelling dropout vs. not modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). With dropout Very low-maintaining Low-declining Average-maintaining Total Very low-maintaining 3 0 3 Without dropout Low-declining 19 23 42 171 Total Average-maintaining 0 8 8 19 11 23 53 Table 4.30. Participants who switched mental quality of life trajectory groups when modelling dropout vs. not modelling dropout in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). With dropout Low-increasing Above average-declining Average-maintaining Total Low-increasing 0 0 0 Without dropout Above averagedeclining 13 9 22 172 Total Average-maintaining 1 35 36 14 35 9 58 4.11.6 Sensitivity analysis: Time lagged model for previous quality of life score analysis Time-lagged GBTM was used in this analysis to determine if modelling the previous physical and mental QoL score had any effect on the trajectory groups observed, the model fit and characterization of trajectory groups. Briefly, the physical and mental time-lagged GBTM analysis found three trajectory groups similar to the non-time-lagged models shown in the main body of the manuscript. As noted in the manuscript, the timelagged models had a more advantageous fit of the data. Further, differences between characterization of trajectory groups were not substantial and these differences are hypothesized to be due to chance or spurious in nature. Further, mean coefficient intercepts, slope values and physical activity as a time-varying covariate maintained similar statistical significance between time-lagged models and models presented in the manuscript. These analyses rule out the possibility of response shift, which was noted as a limitation of this study. Our results suggest that future research replicating these results should plan a priori to test time-lagged models as their main modelling strategy to fit trajectories in prostate cancer populations. 173 Table 4.31. Time-lagged multinomial logistic regression models of influential factors on physical and mental QoL trajectory group membership relative to the average-maintaining/increasing QoL group in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Prognostic/behavioural factors Low/above average-declining versus Average-maintaining Quality of Life RRR a 95% CI Very low/low-maintaining/increasing versus Average-maintaining Quality of Life RRR a Physical QoL trajectories Age at diagnosis (years) 1.06 1.01, 1.10 1.12 b Aggressiveness of disease 1.34 0.69, 2.60 1.47 b Radiation therapy 1.02 0.53, 1.94 0.56 b Hormone therapy 1.53 0.90, 2.61 2.01 Prostatectomy b 1.52 0.73, 3.16 2.49 Charlson co-morbidity score 1.49 1.23, 1.82 1.95 2 BMI (kg/m ) 1.08 1.01, 1.16 1.18 b Smoking status 1.99 0.91, 4.35 3.55 Mental QoL trajectories Age at diagnosis (years) 1.05 1.00, 1.10 0.99 Aggressiveness of disease b 0.92 0.44, 1.90 0.33 b Radiation therapy 0.79 0.41, 1.52 0.53 Hormone therapy b 1.40 0.70, 2.80 0.98 Prostatectomy b 2.40 0.89, 6.47 1.55 Charlson co-morbidity score 1.35 1.15, 1.59 1.18 BMI (kg/m2) 1.03 0.96, 1.12 1.04 b Smoking status 1.17 0.46, 3.02 2.18 Abbreviations: BMI = body mass index; RRR = relative risk ratio; CI = confidence interval. a All models were adjusted for time varying total physical activity and all other factors in table. 174 95% CI 1.07, 1.17 0.74, 2.91 0.30, 1.04 1.13, 3.57 1.16, 5.34 1.60, 2.39 1.09, 1.27 1.60, 7.91 0.95, 1.02 0.14, 0.79 0.29, 0.97 0.58, 1.66 0.78, 3.07 1.02, 1.36 0.97, 1.11 1.18, 4.05 b Dichotomized variables. 175 Table 4.32. Maximum likelihood estimates for the mean coefficients and corresponding standard errors for physical and mental quality of life trajectory groups from final adjusted group-based trajectory time-lagged models including physical activity as a time-varying covariate in prostate cancer survivors (n = 817) in Alberta, Canada (1997-2007). Trajectory groups Physical quality of life trajectories Average-maintaining Low-declining Very low-maintaining Mental quality of life trajectories Average-maintaining Above average-declining Low-increasing ***p-value<0.001 Estimated mean coefficients (Standard Error) Baseline intercept Slope Physical activity slope 50.59 (1.18)*** 41.63 (1.15)*** 24.32 (1.30)*** -0.37 (0.43) -1.90 (0.43)*** -0.04 (0.57) 0.01 (0.01)* 0.02 (0.01)** 0.03 (0.01)*** 54.62 (0.72)*** 52.28 (2.84)*** 24.64 (2.18)*** 0.41 (0.28) -5.39 (1.27)*** 3.04 (0.91)*** 0.01 (0.01) 0.01 (0.01) 0.01 (0.01) **p-value<0.01 *p-value<0.05 176 Chapter Five: DISCUSSION 5.1 Summary of findings The overall aim of this thesis was to explore long-term QoL in prostate cancer survivors’ post-diagnosis. This aim was achieved by first investigating associations between physical activity and QoL in a cohort of prostate cancer survivors to determine options for tertiary prevention and future intervention studies in this population. Subsequently, long-term trajectories of QoL using GBTM were used to determine if there were distinct subgroups of prostate cancer survivors who follow different patterns of QoL after diagnosis. When examining physical activity as a modifiable behaviour that can improve QoL, we found that recreational physical activity improved physical QoL after diagnosis. Specifically, those who maintained physical activity according to cancer prevention physical activity guidelines across the diagnostic period improved physical and mental QoL to the greatest extent compared to non-exercisers. In examining trajectories of QoL, we discovered there were indeed three different subgroups of physical and mental QoL trajectories present in our study population. These trajectory groups ranged in prevalence, with the highest prevalence in the low-declining physical QoL trajectory and the average-maintaining mental QoL trajectory. This finding denotes a greater reported problem for physical QoL over mental QoL in this population. In addition, there were several different patterns of QoL scores among the trajectory groups. For physical QoL, both high and low trajectory groups maintained their QoL scores, while the medium group declined over the follow-up. However, in mental QoL trajectories, high and low groups increased in QoL scores by the third time-point while again, the medium group declined. The mental above averagedeclining QoL group declined substantially. This result may be concerning to healthcare professionals since the mean score reduction was more than 10 points which represents a 177 substantial clinically relevant decrease in QoL (1). Dropout probabilities were the highest in the very low-maintaining physical, the low-increasing and above average-declining mental groups. Finally, we systematically characterized the trajectory groups and found multiple factors associated with low and medium groups compared to the corresponding high group for physical and mental QoL. An increase in Charlson comorbidity index score was consistently associated with both low and medium physical and mental QoL group membership relative to the corresponding high group. Hormone therapy and increased BMI were related to both very lowmaintaining and low-declining physical QoL group memberships, while age at diagnosis, having a prostatectomy and being a current smoker were related to very low-maintaining physical QoL group membership. Additionally, aggressiveness of disease, radiation therapy and being a current smoker were related to low-increasing mental QoL group membership relative to the averagemaintaining group. Finally, age at diagnosis and having a prostatectomy were associated with the above average-declining mental QoL group membership. Together, these observations, provide relevant evidence to inform future research to further develop and validate these findings. 5.2 Limitations The limitations specific to each analysis have been highlighted in chapters three and four. This section will discuss, in depth, potential threats to the internal and external validity of this study and the magnitude of these concerns when communicating results to other researchers, health professionals, patients and other stakeholders. The main threats to internal and external validity that will be discussed below are: survivorship bias due to missing data and dropout, reporting bias with respect to self-administered questionnaires, potential confounding, potential 178 lack of precision based on sample size, causal inference and generalizability of results to the greater prostate cancer survivor population. 5.2.1 Internal validity Internal validity refers to the method by which a study is conducted and if all of the components accurately represented the true association present. This topic is most important when making causal inferences about study results. There are two threats to internal validity: systematic error and random error. Systematic error is a fundamental flaw in the study design, implementation, measurement, analysis or interpretation of the study that may distort the true association between the exposure and the outcome. The most common types of systematic error are selection, measurement and confounding biases. While systematic error cannot be resolved by increasing the sample size, random error is variability based on the role of chance. Random error may dilute an association of interest if there is a lack of precision due to small sample sizes (2). Each one of these biases pertaining to internal validity will be discussed below. 5.2.1.1 Survivorship bias due to missing data and dropout Survivorship bias is a type of selection bias that can occur when the selection probabilities are related to the exposure and the outcome; this bias is a main concern in prospective cohort studies with long follow-up periods (3). The prostate cancer cohort study was derived from a prior population-based case-control study. Cases were derived from the Alberta Cancer Registry which is a province-wide cancer registry with an ascertainment rate of approximately 95% (4). From the prior case-control study (5), 75.7% of eligible men with prostate cancer were included in the initial case-control study. The remaining cases from the 179 prior case-control study who were not included in the case-control sample were 265 men who refused (19.8%), 34 who had incomplete interviews at post-diagnosis (2.5%), and 24 who were unsuitable for an interview at first follow-up (1.8%). From there, men re-consented to participate in the prospective cohort study, in which there were three who refused (2% refusal rate), one man whose primary cancer turned out to not be prostate cancer, and 154 who died before the first follow-up (14% of the original case-control study sample). Consequently, survivorship bias may be present if the reason why prostate cancer survivors did not re-consent to the prospective cohort study was related to their physical activity levels and QoL scores. A hypothetical example of this mechanism for the association between physical activity and QoL may arise if prostate cancer survivors who did not consent to participate to the prospective cohort study had very low levels of physical activity and QoL scores. The prostate cancer survivors who were not included in this cohort study might have experienced several treatment side effects, may not have adjusted well to their disease, might have been ashamed of their lifestyle, could have had potentially higher cancer stages or lethal prostate cancer and died early after diagnosis and would not be included in our study. This would lead to an overestimation of the association between physical activity and QoL present in these results. When examining the stage distribution between the prior case-control study and the prospective cohort study, there were substantially more (n=39, 25%) stage IV cancers in the 157 men who did not re-consent to the cohort study compared to those who did re-consent to the cohort study (n=55, 5.7%). However, because there was <20% of the original case-control study population who did not consent to the study, these concerns are somewhat mitigated. There were also prostate cancer survivors who dropped out, were lost to follow-up or died during active data collection, which were other possible sources of survivorship bias. In our 180 analyses we tried to substantiate the magnitude of this potential bias by conducting sensitivity analyses and specifically, in the QoL GBTM analysis, by accounting for dropout in our final models. In the physical activity and QoL analyses, sensitivity analyses ruled out the possibility of advanced cancer stage (III/IV and IV) or mortality explaining these results, which reduces the likelihood of survivorship bias. In addition, sensitivity analyses were performed in the GBTM analyses to dismiss the role of missing data on model selection and characteristics of trajectory groups. While significance of some characteristics may have changed slightly in these sensitivity analyses, meaningful clinically relevant differences were not observed. On another note, in the QoL trajectory analysis, we determined that there were differential probabilities of dropping out due to mortality and being lost to follow-up between the trajectory groups. Therefore, in the very low-maintaining physical QoL group and the low-increasing mental QoL group, the patterns of maintenance or increasing QoL score may be superficial, since men with declining QoL could have dropped out of the study. It is possible that those who had started with low baseline QoL values have poorer general health. This possibility was supported in the analyses characterizing the trajectory groups. Nonetheless, from these analyses we were able to determine that dropout probabilities do differ between trajectory groups and therefore, there is a need to examine these very low/low-maintaining/increasing groups further to tease apart the mechanisms by which dropout is affecting the QoL patterns over time. 5.2.1.2 Measurement error with self-administered questionnaires Recall error is a type of measurement error that arises from erroneous recollection of information, usually associated with exposure status (2). These errors may impact study results depending on the magnitude of the misreporting of exposure information. In this study, all 181 physical activity and QoL data were obtained through either interviewer-administered or selfadministered questionnaires. The differences between interviewer-administered and self-reported data are minimal with respect to recollection of information (6). However, since interviews and self-administered questionnaires both rely on the participant to recall their exposure or outcome information, these methods of data ascertainment may be prone to recall error. Recollection of physical activity behaviours may be difficult for some participants, especially if these activities are not part of a consistent routine. In this case, participants may over- or under-estimate their physical activity levels leading to errors in physical activity exposure assessment. These errors may be especially susceptible to the longer exposure periods being collected, specifically, the lifetime physical activity measurement. Further, for older participants, a lifetime measurement of physical activity would encompass many decades of behaviour to recall, which may be more difficult than for the younger participants. To mitigate these issues of recall error, several methods were employed in this study including: collection of lifetime physical activity using a reliable questionnaire (7), cognitive interviewing methods (8), a recall calendar sent to the participants prior to their interview and different quality control mechanisms within the study design and staff. Moreover, post-prostate cancer diagnosis, another three repeated measurements of past one to two years of physical activity were collected from the participants. The past-year physical activity measurements may also be subject to the same errors. However, to reduce errors, the questionnaire used was tested for validity and reliability (9), repeated measurements were taken from the same participants at up to three time-points and similar quality control methods were implemented as with the lifetime physical activity assessment. The detailed and careful collection of physical activity behaviour in this study may alleviate concerns regarding the possibility of potential bias introduced by the aforementioned recall error. Notwithstanding, 182 objective measurements of physical activity are recommended for increased accuracy in estimating physical activity levels in future research. In respect to measurement of QoL information, these data were collected through selfadministered questionnaires. Since QoL is a patient reported outcome and very personal to each individual, it may be difficult to determine any substantial measurement errors from misreporting. However, the subjectivity of an individual’s own statement of how he/she feels may introduce measurement and recollection errors (10). While these measures can bridge the gap of quality of care between health professionals and patients, there may be additional complications in interpreting these data. In this study, the SF-36 questionnaire was used to collect QoL information. The SF-36 is the most commonly used health status measure and has been validated in cancer populations (11, 12). Therefore, systematic error due to lack of reliability and validity is unlikely from these measurements. However, as is common with most types of self-reported QoL data, the SF-36 may be subject to non-differential misclassification bias that would attenuate statistical associations (10). This bias may lead to QoL scores that are very similar without any variation. However, when utilizing the component summary scores for physical and mental QoL, the distributions approximated a normal distribution with enough variation to analyze these data. When interpreting these data, it is important to consider the magnitude of potential errors and how future research may improve methods to reduce measurement error when examining these associations. 5.2.1.3 Confounding Confounding is the intermixing of an extraneous variable on the association of interest. This extraneous variable is an independent risk factor of the outcome, associated with the 183 exposure and not on the causal pathway (2, 3). In this thesis, several potential confounding variables were considered in the analyses (chapters three and four). Multiple steps were taken to appropriately test for potential confounding by factors that were hypothesized to possibly act as confounders. In addition, for analyses presented in chapter three, two different model selection methods were used; first, backward elimination (presented in the manuscript) and second, GLMNET (presented in the additional results sections). Both of these model-generating methods found similar conclusions, with only slight differences in the final covariates that were included in the final models as confounders. While our analyses did control for several confounders, there is still a possibility of residual confounding due to unmeasured factors that could not be considered in these models. This issue was specifically of concern in chapter four, when dichotomizing smoking status and treatment variables. The GBTM methods cannot model categorical variables with more than two levels. In this case, we made assumptions when dichotomizing some variables, which may result in residual confounding. First, never and former smokers have been documented in the literature to have similar QoL (13), however, this similarity may not exist or be relevant when identifying QoL trajectories. In addition, other variables such as race/ethnicity, marital status and socioeconomic status were not included in the GBTM analyses, partially because our study population was very homogeneous with respect to these factors. There is evidence supporting the different associations between race/ethnicity and disadvantaged/low socioeconomic status and QoL after prostate cancer diagnosis (14, 15). Further, research has also indicated that having a partner may be an indicator of better QoL after prostate cancer diagnosis (16). These distinctions may need to be teased apart to clarify our results and to identify features of this population to consider for future research. 184 Confounding by indication was potentially present in the analyses presented in chapter four. The definition is as follows: when an extraneous variable is an independent risk factor for QoL in the non-exposed population, is associated with the exposure of interest as well as not on the causal pathway. In this case, in chapter four, a brief discussion of the situation when it is not the prostate cancer diagnosis that is driving the QoL patterns, but perhaps the increase in age, number of co-morbidities or lack of general health that may have created a heightened baseline risk for some survivors. In this situation, our results cannot disentangle the difference between the diagnoses of prostate cancer versus the potential decline of health that is associated with aging. It is possible that low, medium and high QoL patterns exist in healthy populations as well. Since this research area is very novel and similar analyses have not been carried out in healthy populations, it is unknown whether or not these groups (if they exist) would be comparable to the trajectory groups of physical and mental QoL found in this thesis. Likewise, it is unknown, if other diseased populations within this age range and progression would follow similar trajectory patterns. It is possible that these trajectory patterns are specific to prostate cancer survivors, however, research thus far using these methodologies has been limited and we were unable to account for this potential confounding by indication bias. Ultimately, to address this limitation, future research including prostate cancer survivors and healthy men in the same age range and demographics would need to be conducted to determine the impact of prostate cancer on QoL trajectory groups after diagnosis. 5.2.1.4 Precision and issues with multiple statistical tests Although random error is a function of the precision of the estimates produced and observed, multiple statistical tests may lead to type I error. Type I error occurs when the 185 observed significant result occurred by chance, which can occur when multiple testing is done and the likelihood of a significant result increases (2). In our physical activity and QoL study (presented in chapter three), several analyses were run. The majority of the associations were proposed a priori based on gaps in the literature, the research question and available data. These analyses were expanded to include more in-depth examination of duration (hours/week) and recreational physical activity based on cancer prevention physical activity guideline adherence. In addition, our sample size allowed us to test multiple potential effect modifiers and confounders, which further lends to the issue of multiple testing. However, meaningful results came from this testing and we were able to fully adjust the models for potential influential variables. In the GBTM analysis (chapter four), the precision of the trajectory groups found was evident as the CI surrounding the trajectory patterns (Figures 4.2 & 4.3) were quite narrow. This analysis was sufficiently powered to determine groups within the sample, as shown by our results. However, the multinomial logistic regression models were not powered to detect differences between trajectory groups. The CIs were very wide and therefore, these results should be interpreted with caution. The issue of multiple testing is less of a concern in the QoL trajectory analysis since characterizing the trajectory groups was an exploratory analysis. On another note, in both chapters three and four, we utilized the component summary scores of physical and mental QoL, which are overall measures of general QoL. Using the component summary scores, rather than the eight individual domain scores, allowed for more direction in the analyses, ease of interpretation and a reduction in statistical testing compared to analyzing all eight domains separately. While issues of precision and multiple testing were present in this thesis, nonetheless, meaningful investigation of potential modification, confounding, dropout, trajectory characteristics and sensitivity analyses will lead to more focused research in the future. 186 5.2.1.5 Causal inference Causation between an exposure and an outcome may be evaluated through the Bradford Hill Criteria (2). First of all, with the design of the study and methods, temporality of the association between physical activity and QoL and QoL trajectories was established. These participants were enrolled in the study and followed over time. In addition, the association between physical activity and QoL has been examined in several different populations including several prostate cancer study populations. The literature generally supports this association alluding to a level of consistency between these results. However, the strength of the association may be questionable. Only the adherence to the cancer prevention physical activity guidelines analysis resulted in clinically relevant QoL score differences. With respect to the QoL trajectory analysis (chapter four), there have only been two previous studies that examined post-prostate cancer diagnosis disease-specific QoL (urinary function (17) sexual function (18)) trajectories using GBTM methods. As noted in chapter four, these studies also found different trajectories of prostate cancer survivors. However, the time frame in which these studies followed participants for QoL measurements were shorter and the population restrictions were different than for this study. The associations determining QoL trajectories in prostate cancer survivors is not supported by a large body of literature and therefore, more research are needed to confirm these findings for potential future policy initiative or change in standard care to be implemented in this population. 5.2.2 External validity External validity is the extent to which study results can be generalized to other, more broad populations (2). If internal validity is not met, external validity cannot be assessed or 187 considered. Threats to external validity may involve very rigorous inclusion criteria, thus resulting in population specific estimates not applicable to other sources. While this study population came from a prior population-based case-control study and cases were collected from a population registry with an ascertainment rate of 94.5%, there are inherent limitations to the generalizability of the results. First of all, re-consent into the prospective cohort study was not considered until two to three years post-prostate cancer diagnosis. Therefore, only prostate cancer cases who survived to that point could participate and the results can only be generalized to this specific group of survivors. While this issue may be of concern, there was only a 2% refusal rate and approximately 82% of the original case-control study participants did participate in the cohort follow-up, therefore, the threats to generalizability may be modest. Furthermore, due to our inclusion criteria, these results can only be generalized to men with stage T2 or greater prostate cancer, who were less than 80 years at diagnosis. Although these inclusion criteria may limit generalizability of the results, they may also rule out other competing risks, confounding and other alternative explanations of these results, thus, strengthening the internal validity. Further research representing all prostate cancer survivors followed for QoL trajectories directly after prostate cancer diagnosis is warranted. 5.3 Public health implications This thesis has provided new empirical evidence regarding the changes in QoL after prostate cancer diagnosis and the role of physical activity in these temporal changes. This research has implications for changes to clinical practice and future public health policy initiatives. Specifically, the results observed in chapter three build stronger evidence towards recommending recreational physical activity and specifically moderate-vigorous recreational 188 physical activity to improve physical QoL after prostate cancer diagnosis regardless of age at diagnosis (<80 years), aggressiveness of cancer, treatment type (hormone or radiation therapy), prostatectomy or degree of comorbidity. In addition, we also found that among prostate cancer survivors who either maintained high levels of physical activity from pre to post-diagnosis or who increased their levels of activity to those recommended for cancer prevention, there were higher levels of QoL. The fact that both maintainers and adopters benefitted almost equally in QoL improvements, suggests that adopting a physically active lifestyle after prostate cancer diagnosis may be equally beneficial as maintaining a physically active lifestyle throughout the cancer diagnosis and survival experience. To further emphasize this point, we found that men who relapsed in their levels of activity experienced declines in QoL which might be preventable with regular physical activity. Improvements in physical QoL were found to be of clinical relevance in both analyses and therefore, health professionals should be encouraged to recommend recreational moderate-vigorous physical activity as a tertiary prevention method to prevent declines and/or achieve increases in QoL after prostate cancer diagnosis. This type of evidence is of relevance for organizations such as Exercise is Medicine®, an initiative started by the American College of Sports Medicine that seeks to provide evidence for primary care health providers that can be used to prescribe exercise as a means of treating their patient population for a wider range of chronic disease conditions. The core objectives of Exercise is Medicine® include: the initiation and collaboration of research projects to enhance implementation of physical activity in all domains of healthcare, to serve as a coordinating centre for these implementation projects and to collaborate on developing models to educate healthcare professionals in community and clinical settings to effectively promote physical activity behaviour change. Specifically, exercise-referral schemes, training for healthcare professionals in 189 physical activity prescription and integration of wearable technology in promotion of physical activity may be appropriate next research steps for prostate cancer QoL and survival research in order to reduce overall burden of disease (19). In chapter four, we identified trajectory patterns of QoL after prostate cancer diagnosis that are also of relevance for clinical practice. We identified different and potentially “at-risk” low or declining QoL subgroups within the larger population of prostate cancer survivors who would warrant closer follow-up and possibly clinical interventions. Our research will require further confirmation in future studies. If these subgroups are found in future research, tertiary prevention intervention studies could focus on at-risk groups, who would benefit the most from these interventions, to help improve or maintain QoL after prostate cancer diagnosis and overall, reduce the burden of prostate cancer after diagnosis. This research is of particular relevance for public health, given that prostate cancer patients are surviving longer and a proportion of them may be experiencing declines in their QoL long after their diagnosis. If subgroups of these individuals at risk for such QoL declines can be identified early after prostate cancer diagnosis, there is the possibility of improving their long-term QoL through interventions targeting their activity levels. 5.4 Future directions Several areas for future research on this topic can be considered given the substantial body of evidence that has clearly demonstrated improvements in QoL associated with physical activity done after prostate cancer diagnosis. Ideally, these include studies that examine different exercise interventions aimed at delineating the most effective, population-based approaches that will be feasible, sustainable and have maximal adherence for different populations of prostate 190 cancer survivors. Efficacy studies will need to investigate how these interventions can be personalized to ensure behavioural change at the individual level and effectiveness trials will need to examine various interventions and how they can be best implemented at the population level. Another topic area for future research is education for healthcare professionals to incorporate recommendations on physical activity after prostate cancer diagnosis in their interactions with this clinical population. Any cancer diagnosis is stressful and overwhelming, therefore, research that investigates the most effective timing and type of intervention with this population is needed to ensure lifestyle changes can be made to maintain or improve QoL after prostate cancer diagnosis. The research evidence to date on QoL trajectories after prostate cancer diagnosis remains limited, hence, additional research is essential to replicate our results to determine if the three subgroups of physical and mental QoL that we identified here are present in other prostate cancer populations. It may be feasible in the future, once sufficient confirmatory evidence has accumulated, to target research studies on the high risk sub-groups who have low and/or declining QoL with a focus on improving QoL in the low groups and preventing reductions in QoL over time in the declining QoL groups. 5.5 Conclusion This thesis adds to the literature providing potential methods of improving the QoL of prostate cancer survivors after diagnosis. QoL is a central component to most patients’ lives that should be addressed when considering the healthcare needs of prostate survivors after diagnosis. Our research examining how physical activity improves QoL found moderate-to-vigorous recreational physical activity was the most beneficial for physical QoL. In addition, adhering to 191 cancer prevention physical activity guidelines both pre- and post-diagnosis or just post-diagnosis was positively related to physical QoL, compared to those who did not meet guidelines at either time point. These results were clinically meaningful and should be taken into consideration when healthcare professionals are giving prostate cancer survivors advice on how to improve their QoL. In addition, our analyses identified three physical and three mental QoL trajectories up to 10 years after diagnosis of prostate cancer. These groups had very different QoL patterns providing evidence that there may indeed be different sub-populations experiencing very different QoL after prostate cancer diagnosis. Our results have provided insights for future interventions to improve and/or maintain QoL long-term in this population. 192 5.6 References 1. Jaeschke R, Singer J, Guyatt GH. Measurement of health status. Ascertaining the minimal clinically important difference. Control Clin Trials. 1989;10(4):407-15. 2. Patten S. Epidemiology for Canadian Students: Principles, Methods and Critical Appraisal: Brush Education; 2015. 3. Rothman KJ, Greenland S, Lash TL. Modern Epidemiology: Wolters Kluwer Health/Lippincott Williams & Wilkins; 2008. 4. Chen V, Wu X, Andrews P. Cancer in North America, 1991-1995. Sacramento, CA: North American Association of Central Cancer Registries; 1999. 5. Friedenreich CM, McGregor SE, Courneya KS, Angyalfi SJ, Elliott FG. Case-control study of lifetime total physical activity and prostate cancer risk. Am J Epidemiol. 2004;159(8):740-9. 6. Oleckno WA. Epidemiology: Concepts and Methods: Waveland Press; 2008. 7. Friedenreich CM, Courneya KS, Bryant HE. The lifetime total physical activity questionnaire: development and reliability. Med Sci Sports Exerc. 1998;30(2):266-74. 8. Willis G. Cognitive interviewing and questionnaire design: a training manual. Hyattsville, MD: National Center for Health Statistics; 1994. 9. Friedenreich CM, Courneya KS, Neilson HK, Matthews CE, Willis G, Irwin M, et al. Reliability and validity of the Past Year Total Physical Activity Questionnaire. Am J Epidemiol. 2006;163(10):959-70. 10. Hanestad BR. Errors of measurement affecting the reliability and validity of data acquired from self-assessed quality of life. Scand J Caring Sci. 1990;4(1):29-34. 11. Antonescu I, Carli F, Mayo NE, Feldman LS. Validation of the SF-36 as a measure of postoperative recovery after colorectal surgery. Surg Endos. 2014;28(11):3168-78. 12. Wang R, Wu C, Zhao Y, Yan X, Ma X, Wu M, et al. Health related quality of life measured by SF-36: a population-based study in Shanghai, China. BMC Public Health. 2008;8(1):1-8. 13. Dieperink KB, Hansen S, Wagner L, Johansen C, Andersen KK, Hansen O. Living alone, obesity and smoking: important factors for quality of life after radiotherapy and androgen deprivation therapy for prostate cancer. Acta Oncol. 2012;51(6):722-9. 14. Brassell SA, Elsamanoudi SI, Cullen J, Williams ME, McLeod DG. Health-related quality of life for men with prostate cancer--an evaluation of outcomes 12-24 months after treatment. Urol Oncol. 2013;31(8):1504-10. 193 15. Sevilla C, Maliski SL, Kwan L, Connor SE, Litwin MS. Long-term quality of life in disadvantaged men with prostate cancer on androgen-deprivation therapy. Prostate Cancer Prostatic Dis. 2012;15(3):237-43. 16. Bellardita L, Rancati T, Alvisi MF, Villani D, Magnani T, Marenghi C, et al. Predictors of health-related quality of life and adjustment to prostate cancer during active surveillance. Eur Urol. 2013;64(1):30-6. 17. Anderson CB, Kaufman MR, Dietrich MS, Barocas DA, Chang SS, Cookson MS, et al. Recovery of urinary function after radical prostatectomy: identification of trajectory cluster groups. J Urol. 2012;187(4):1346-51. 18. Legler JM, Davis WW, Potosky AL, Hoffman RM. Latent variable modelling of recovery trajectories: sexual function following radical prostatectomy. Stat Med. 2004;23(18):2875-93. 19. Lobelo F, Stoutenberg M, Hutber A. The Exercise is Medicine Global Health Initiative: a 2014 update. Br J Sports Med. 2014;48(22):1627-33. 194 Appendices A: Lifetime Physical Activity Questionnaire 195 196 197 198 199 200 201 202 203 Appendices B: Past-Year Physical Activity Questionnaire 204 205 206 207 208 209 210 211 212 Appendices C: Well-being Questionnaire 213 214 215 216 217 218 219 220 221 222 223 224