Citation
Impact of androgen deprivation therapy on quality of life in men with stage IV prostrate cancer

Material Information

Title:
Impact of androgen deprivation therapy on quality of life in men with stage IV prostrate cancer
Creator:
Withrow, Susanne ( author )
Language:
English
Physical Description:
1 electronic file (69 pages). : ;

Subjects

Subjects / Keywords:
Prostate -- Cancer ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Abstract:
Patients with late-stage prostate cancer often face a variety of treatment decisions. Interventions like androgen deprivation therapy (ADT) are not only associated with physiological side effects, but are also accompanied by effects on Quality of Life (QOL). This experimental study recruited 73 men aged 50 and older with stage IV prostate cancer assigned to either continuous or intermittent ADT to evaluate whether these treatment modalities, which reduce biologically available levels of testosterone (BioT), result in differential impacts with regard to QOL. Participants in this study were followed prospectively and assessed at three timepoints over the course of 24 to 32 months. A MANCOVA using data from timepoint 3 on nine dependent variables associated with QOL in prostate cancer patients undergoing ADT included: Physical Functioning, Role Limitations Due to Physical Health, Pain, General Health, Emotional Wellbeing, Role Limitations Due to Emotional Health, Energy, Social Functioning, and Depression. Adjustments were made for eight covariates: age, education, presence of diabetes, body mass index, time since diagnosis, processing speed, verbal learning, and verbal fluency. Group membership predicted differences in overall QOL (F(9) = 4.61, p < .01). Specifically, individuals in the continuous ADT treatment group experienced lower levels of Physical Functioning (F(1) = 12.08, p < .01) and more Role Limitations Due to Physical Health (F(1) = 6.78, p < .05). Significant multilevel correlations between levels of BioT with Physical Functioning and depressive symptoms were found after adjusting for multiple observations among participants. Other trends for non-significant associations and limitations of this study are discussed. Understanding the impact of intermittent and continuous ADT on a variety of domains associated with QOL could inform health care decisions in the future and may help to devise strategies for support and supplemental intervention approaches to improve overall treatment outcomes in late-stage prostate cancer patients undergoing ADT.
Thesis:
Thesis (M.A.)--University of Colorado Denver.
Bibliography:
Includes bibliographic references
System Details:
System requirements: Adobe Reader.
General Note:
Department of Psychology
Statement of Responsibility:
by Susanne Withrow.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
951986533 ( OCLC )
ocn951986533

Downloads

This item has the following downloads:


Full Text
IMPACT OF ANDROGEN DEPRIVATION THERAPY ON QUALITY OF LIFE
IN MEN WITH STAGE IV PROSTATE CANCER
By
SUSANNE WITHROW
B.S., University of Colorado Denver, 2013
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Master of Arts
Clinical Health Psychology Program
2016


2016
SUSANNE WITHROW
ALL RIGHTS RESERVED


This thesis for the Master of Arts degree by
Susanne Withrow
Has been approved for the
Clinical Health Psychology Program
By
James Grigsby, Chair
Krista Ranby
Kristin Kilbourn


Withrow, Susanne (M.A., Clinical Health Psychology)
Impact of Androgen Deprivation Therapy on Quality of Life in Men with Stage IV
Prostate Cancer
Thesis directed by Professor James Grigsby.
ABSTRACT
Patients with late-stage prostate cancer often face a variety of treatment decisions.
Interventions like androgen deprivation therapy (ADT) are not only associated with
physiological side effects, but are also accompanied by effects on Quality of Life (QOL).
This experimental study recruited 73 men aged 50 and older with stage IV prostate cancer
assigned to either continuous or intermittent ADT to evaluate whether these treatment
modalities, which reduce biologically available levels of testosterone (BioT), result in
differential impacts with regard to QOL. Participants in this study were followed
prospectively and assessed at three timepoints over the course of 24 to 32 months. A
MANCOVA using data from timepoint 3 on nine dependent variables associated with
QOL in prostate cancer patients undergoing ADT included: Physical Functioning, Role
Limitations Due to Physical Health, Pain, General Health, Emotional Wellbeing, Role
Limitations Due to Emotional Health, Energy, Social Functioning, and Depression.
Adjustments were made for eight covariates: age, education, presence of diabetes, body
mass index, time since diagnosis, processing speed, verbal learning, and verbal fluency.
Group membership predicted differences in overall QOL (F(9) = 4.61 ,p<.01).
Specifically, individuals in the continuous ADT treatment group experienced lower levels
of Physical Functioning (F(i) = 12.08, p < .01) and more Role Limitations Due to
Physical Health (F(i) = 6.78, p < .05). Significant multilevel correlations between levels
of BioT with Physical Functioning and depressive symptoms were found after adjusting
m


for multiple observations among participants. Other trends for non-significant
associations and limitations of this study are discussed. Understanding the impact of
intermittent and continuous ADT on a variety of domains associated with QOL could
inform health care decisions in the future and may help to devise strategies for support
and supplemental intervention approaches to improve overall treatment outcomes in late-
stage prostate cancer patients undergoing ADT.
The form and content of this abstract are approved. I recommend its publication.
Approved: Jim Grigsby
IV


ACKNOWLEDGEMENTS
I would like to acknowledge any and all individuals, including Angela Brega,
Ph.D. and Marsha Paulich, M.A., who collected and recorded data for this project at the
urologic oncology clinics of Michael Glode, MD, at the University of Colorado Health
Sciences Center, and Lawrence Karsh, MD, of Western Urological Associates, which
eventually led to the comprehensive database from which I completed this thesis. To my
advisor and thesis committee: Drs. Jim Grigsby, Krista Ranby, and Kristin Kilbourn, your
advice has helped me to grow as a researcher and clinical health psychologist. I greatly
appreciate your support and guidance throughout this process. The statistical support and
consultations I have received from Dr. Krista Ranby and Dr. Sung-Joon (Max) Min have
greatly helped to shape the nature of this project and my skill set as a researcher. I also
want to acknowledge my laboratory colleagues: Anson Kairys and Tim Brunelle. This
project has greatly benefited from your advice and support.
v


TABLE OF CONTENTS
CHAPTER
I. BACKGROUND
Introduction and Significance.........................................1
Association Between Testosterone Levels and Quality of Life Domains Related to
Physical Health.......................................................3
Association Between Testosterone Levels and Quality of Life Domains Related to
Emotional Health......................................................6
Comparison of Continuous and Intermittent ADT with Respect to Primary and
Secondary Treatment Outcomes..........................................9
Impact of Other ADT Side Effects on QOL..............................10
Objectives...........................................................11
Hypotheses...........................................................12
II. METHOD
Procedures ..........................................................13
Sample ..............................................................16
Instruments .........................................................18
Analysis.............................................................25
Statistical Power....................................................29
Study Design.........................................................30
III. RESULTS
Preliminary Data Analyses ...........................................32
Hypothesis 1.........................................................40
Hypothesis 2.........................................................42
Hypothesis 3.........................................................43
vi


IV. DISCUSSION
Summary of Results..............................................46
Limitations and Future Directions...............................48
Conclusion......................................................51
REFERENCES..........................................................54
vii


LIST OF TABLES
TABLE
1. Clinical Trial Inclusion and Exclusion Criteria............................14
2. Descriptive Statistics of Independent and Dependent Variables at Baseline...33
3. Descriptive Statistics of Independent and Dependent Variables After ADT.....34
4. Descriptive Statistics of BioT..............................................35
5. Descriptive Statistics of Covariates........................................36
6. Bivariate Correlations Among Independent Variable, Dependent Variables, and
Covariates..................................................................39
7. Group Differences in Quality of Life after Continuous or Intermittent Androgen
Deprivation Therapy.........................................................41
8. Multilevel Correlations Between Bioavailable Testosterone Levels and
Quality of Life.............................................................43
viii


LIST OF FIGURES
FIGURE
1. Study Flow Chart and Attrition Over Time..............................17
2. MANCOVA Testing Hypotheses 1 and 2....................................27
3. Scatter Plot For The Association Between Bioavailable Levels of Testosterone
and Physical Functioning.............................................45
IX


LIST OF ABBREVIATIONS
ADT
ANCOVA
BioT
BMI
CES-D
COWAT
DV
FT
IV
MANCOVA
PASAT
QOL
RAVLT
SDMT
SF-36
SHBG
Total T
WAIS-III
Androgen Deprivation Therapy
Analysis of Covariance
Bioavailable Levels of Testosterone
Body Mass Index
Center for Epidemiologic Studies Depression Scale
Controlled Oral Word Association Test
Dependent Variable
Free Levels of Testosterone
Independent Variable
Multivariate Analysis of Covariance
Paced Auditory Serial Addition Test
Quality of Life
Rey Auditory Verbal Learning Test
Symbol Digit Modalities Test
SF-36 Health Survey
Sex Hormone Binding Globulin
Total Levels of Testosterone
Wechsler Adult Intelligence Scale Third Edition
x


CHAPTER I
BACKGROUND
Introduction and Significance
In the United States, prostate cancer is one of the most common types of cancers
among men. In 2010 alone, 196,038 men in the United States received a new diagnosis of
prostate cancer. On average, 1 in 6 men will receive this diagnosis during their lifetime
(American Cancer Society, 2014). Prostate cancer is also one of the leading causes of
cancer death among men (CDC, 2013). Even though early stages of this cancer have very
good survival rates due to the slow growth of the associated tumors, the cancer cells can
invade adjacent as well as distant structures and cause severe health and treatment
complications (Nelson, Carter, DeWeese, Antonarakis, & Eisenberger, 2013). The 5-year
relative survival rate among men with advanced, stage IV metastatic prostate cancers in
which cancer cells have spread to distant lymph nodes, bones, or other organs is 28%
(American Cancer Society, 2014). However, life expectancies in late-stage prostate
cancers can be extended with several types of management strategies.
However, before choosing a management or treatment strategy, patients and
physicians need to consider a variety of factors. Some of them include the possible side
effects of available treatment options, the impact of comorbid conditions on the severity
of side effects, the risk for complications, limitations of the treatment, the cancers
characteristics, the anticipated progression of the cancer if left untreated, and the patients
preferences and characteristics such as age and genetic vulnerability. In high-risk prostate
cancers, such as metastasized cancers or those classified as stage IV cancers, routinely
recommended treatments include surgical interventions (e.g. radical prostatectomy),
1


radiation therapy, androgen deprivation therapy (ADT), or a combination of these three
treatments (Carter, 2011).
ADT is a type of hormone therapy that is based on the goal of reducing levels of
androgens in the body to prevent these hormones from reaching prostate cancer cells.
About 90% of prostate cancers are attributed to adenocarcinoma, an abnormal tissue mass
or growth (neoplasia) of glandular epithelial tissue. A feature of these neoplastic prostate
epithelial cells, as compared to normal secretory cells, is that the neoplastic cells appear
to use androgen receptor signaling not only for differentiation, but also for proliferation.
Therefore, most prostate cancer cells depend on androgens for maintenance of growth
and survival. Testosterone is the major androgenic hormone encountered in the prostate
and is therefore the focus of many outcome studies. Under conditions of androgen
suppression, cancer cells can stop growth or induce programmed cell death, which can
lead to a decrease in tumor size. However, in many cases long-term outcomes are
unchanged, because prostate cancer tumors tend to become androgen-independent. At
that progression stage, androgen suppression induced by ADT has little therapeutic effect
(Gregory, et al., 2001).
ADT interventions can be administered following a continuous or intermittent
schedule. Both continuous and intermittent ADT suppress the production of testosterone
and its associated hormones. However, intermittent ADT includes phases of suppressed
and rising testosterone levels instead of prolonged androgen suppression. The complex
impact on the bodys hormone balance contributes to the range of side effects that are
typically seen in patients undergoing ADT. In addition to impacts on Quality of Life
(QOL) and depression, side effects seen in different types of ADT interventions include
2


sexual dysfunction, hot flashes, loss of bone and muscle mass, liver toxicity,
gynecomastia, breast pain, shrinking of testicles and penis, anemia, weight gain,
increased cholesterol levels, as well as changes in cognition (Heidenreich, et al., 2013;
Schroder, Crawford, Axcrona, Payne, & Keane, 2012; American Cancer Society, 2014).
Recognizing the nature of these possible side effects seen in both continuous and
intermittent ADT, each patient and physician needs to decide whether therapeutic
benefits and increased life expectancies outweigh the risks associated with ADT
interventions. One step toward helping patients make informed health care decisions is to
understand how intermittent and continuous ADT potentially differ in causing changes in
QOL and whether these changes could be also in part due to certain patient
characteristics. Once the nature of the association between different treatment methods
using pharmacological androgen suppression and QOL is better understood, secondary
interventions can be devised to potentially reduce the impact of ADT on QOL. This study
will examine the effect of ADT treatment type on QOL while controlling for specific
participant characteristics that may also be associated with reductions in QOL. The
specific association between testosterone levels and a variety of QOL domains, regardless
of treatment type, will also be investigated.
Association Between Testosterone Levels and Quality of Life Domains Related to
Physical Health
The association between reduced levels of testosterone and worse QOL is
frequently encountered in males who produce low levels of testosterone naturally, or
through medical conditions and interventions. Moncada (2006) reports that low
testosterone levels in hypogonadal men affect seven key areas relevant to QOL: energy,
3


emotional functioning, social function, social-emotional functioning, mental functioning,
physical functioning, and sexual functioning. Prostate cancer patients undergoing ADT
also experience low levels of testosterone and may therefore show similar impacts in
areas of QOL.
Most of the literature on the association between ADT and QOL doesnt explicitly
state whether the treatment regimen studied was continuous or intermittent. In the
following review of the literature, the specific type of ADT regiment will be specified if
it was included in the original source. Some of the side effects of ADT include weight
gain, altered fat distribution, decreases in muscle mass, and a disruption of sleep patterns
(Casey, Corcoran, & Goldenberg, 2012). The presence of these symptoms could explain
why many studies have reported declines in measures of physical health among prostate
cancer patients being treated with ADT. For example, Alibhai et al. (2010) investigated
how continuous ADT influences subjective and objective measures of physical health,
compared to prostate cancer patients who did not receive ADT and healthy controls.
Overall, prostate cancer patients treated with ADT declined in the Physical Functioning,
Role Limitations Due to Physical Health, Pain, and Energy domains of the SF-36 Health
Survey QOL self-report questionnaire. Moreover, continuous ADT seems to be related to
reductions in grip strength over time. Walking distance remained stable for prostate
cancer patients treated with ADT, while this ability improved for prostate cancer without
ADT and healthy control participants (Alibhai, et al., 2010). Surprisingly, this study
found that most declines on objective and subjective measures of physical health related
QOL occurred within 3 month of ADT treatment initiation and remained relatively stable
4


thereafter. Therefore the time of QOL assessment may be crucial to examine the impact
of ADT.
However, the impact on physical health seen after treatment with ADT may also
be explained by several other variables. Dacal, Seraika, and Greenspan (2006) found that
ADT was associated with the physical component scale of the SF-36, which includes the
constructs of Physical Functioning, General Health, Pain, and Role Limitations Due to
Physical Health. However, this association was better explained by the presence of
comorbid conditions and total testosterone levels. After controlling for age, comorbidity,
and total testosterone levels, the association between ADT and physical health impacts
was no longer significant (Dacal, Sereika, & Greenspan, 2006). However, it is not clear
which comorbid conditions were implicated in the findings reported by these authors, and
whether they could be side effects of the ADT treatment itself.
Several studies have also shown that it is important to consider the impact of
comorbid conditions when evaluating disease specific QOL. For example, it has been
found that men with prostate cancer who also have prevalent type 2 diabetes mellitus had
the poorest self-reported general health related QOL, urinary control and sexual function.
The association between comorbid diabetes and QOL in this sample was independent of
treatment regimen (Thong, et al., 2011). Conditions like type 2 diabetes mellitus are
associated with symptoms also seen after ADT treatment, such as weight gain, fatigue,
and changes in urinary function. Therefore it would be important to rule out the influence
of comorbid conditions when describing the association between ADT and QOL. Among
other health and lifestyle factors that play a role in the etiology of many comorbid health
conditions, obesity (BMI >30) has been found to be associated with worse physical
5


health and vitality among prostate cancer patients undergoing ADT (Dieperink, Hansen,
Wagner, Johansen, Andersen, & Hansen, 2012). Given that weight gain is one of the
common side effects of ADT, it is possible that increased weight could influence the
association between ADT treatment and lowered physical health.
Among a sample of prostate cancer survivors who have undergone ADT, those
engaging in more physical activity had higher levels of QOL on average than those
classified as insufficiently active (Keogh, et al., 2010). Overall, only half of their sample
was classified as physically active. Given the reported declines in physical health among
prostate cancer patients undergoing ADT, several studies have aimed to improve physical
functioning through exercise interventions. In the randomized controlled trial by Bourke
and colleagues (2013), a 12-week gradual exercise and dietary advice intervention
showed improvements in disease-specific QOL, fatigue, exercise tolerance, and exercise
behavior as compared to a usual care control group. However, no changes were seen in
blood pressure, and improvements in disease-specific QOL were not maintained at the 6-
month follow-up (Bourke, et al., 2013).
Association Between Testosterone Levels and Quality of Life Domains Related to
Emotional Health
One of the earliest studies on QOL associated with ADT treatment by Herr and
OSullivan (2000) found that not only physical health is affected, but also aspects of
mental health. Specifically, men who received ADT through pharmacological means as
compared to surgical interventions showed greater cognitive-emotional distress from
intrusive thoughts and images related to cancer stress (Herr & O'Sullivan, 2000). Other
studies, such as the one by Cary and colleagues (2013), examined the impact of ADT in
6


prostate cancer patients on a wider variety of QOL domains associated with emotional
well-being. Their findings were drawn from the national CaPSURE (Cancer of the
Prostate Strategic Urologic Research Endeavor) registry and compared prostate cancer
patients who received ADT as a primary treatment with participants who received ADT
with a combination of other treatments. They reported that Role Limitations Due to
Emotional Health and Energy were consistently affected at different timepoints in
individuals who received ADT as a primary treatment. However, no overall group
differences were seen (Cary, Singla, Cowan, Carroll, & Cooperberg, 2013). Strikingly,
clinically significant reductions in Energy scores were observed in 27% of primary ADT
patients after 24 months of treatment. Reductions in Mental Health, Role Limitations Due
to Emotional Health, and Social Functioning were seen in 13%, 20%, and 23% of
primary ADT participants, respectively, after completing 24 months of the therapy
regimen.
Several other studies have taken a closer look at energy and fatigue aspects of
emotional well-being. Storey and colleagues (2012) found that approximately 43% of
men undergoing ADT experience clinically meaningful levels of fatigue. The experience
of fatigue was independently associated with the presence of pain and depression.
However, neither age, disease burden, nor treatment duration was associated with the
presence of fatigue in this sample (Storey, et al., 2012). This study shows that physical
and emotional aspects of QOL may interact to contribute to what some people have
termed the androgen deprivation syndrome.
The findings reported by Kato and colleagues (2007) show that an individuals
cancer state may be another factor that should be considered when examining the impact
7


of prostate cancer and ADT on QOL domains like energy and fatigue. Participants in
earlier stages of prostate cancer progression showed a steady decline in Energy over the
course of one year. However, patients with late-stage prostate cancer showed low initial
levels of Energy and Pain as assessed through the SF-36 QOL questionnaire and slowly
recovered over the course of one year (Kato, et al., 2007). These authors also reported
greater impacts on Role Limitations Due to Emotional Health and low levels of
Emotional Wellbeing. However, this effect was only observed among prostate cancer
patients with advanced stages of disease progression at the initiation of treatment.
Therefore the time course of prostate cancer may be another important factor in
explaining the differential impact of ADT on QOL seen in prostate cancer patients.
Saini et al. (2013) further considered the impact of ADT on depression and
anxiety in addition to evaluating other aspects of QOL in men with prostate cancer.
While this study also found the QOL domains of physical, social/family, and functional
well-being affected by ADT, they also reported that depression was greater in the ADT
group than in prostate cancer patients who were not receiving ADT (Saini, et al., 2013).
There was also a strong correlation between QOL and depression, indicating that these
two variables may measure similar constructs concerning psychological well-being. This
study did not find any group differences with regard to anxiety.
Even though many studies have reported impacts on aspects of mental health like
energy and depression, not all studies have been able to replicate the effect of ADT on
other components of emotional well-being (Dacal, Sereika, & Greenspan, 2006).
8


Comparison of Continuous and Intermittent ADT with Respect to Primary and
Secondary Treatment Outcomes
Intermittent ADT was first proposed to improve treatment-specific symptoms
because individuals were thought to have the ability to recover from side effects during
the off-treatment periods. It was also thought that intermittent ADT could slow disease
progression and delay the onset of the androgen-independent tumor phase, because
individuals spend less time in an androgen-suppressed state (Crook, Szumacher, Malone,
Huan, & Segal, 1999). However, in many instances patients are unable to regain the
ability to produce testosterone for several years after the discontinuation of intermittent
ADT treatment. In one study by Bong et al. (2008) 53% of participants who had received
four or more years of intermittent ADT for prostate cancer retained suppressed
testosterone levels for up to 2.5 years after the end of the intervention. Older participants
who started ADT after age 70 were significantly less likely to recover any amount of
testosterone production after the completion of therapy (Bong, Clarke, Hancock, &
Keane, 2008). The authors of another study of intermittent ADT found that the median
time to testosterone recovery was 100 days (14.29 weeks) after the first cycle, and 115
days (16.43 weeks) after the second intermittent ADT cycle (Tunn, Canepa,
Kochanowsky, & Kienle, 2012). The authors did not report the specific range or
variability seen in recovered testosterone levels for this study. Additionally, other studies
have reported that intermittent ADT is not consistently superior to continuous ADT with
regard to overall and disease-free survival in patients with recurrent cancer emergence
(Kratiras, Konstantinidis, & Skriapas, 2014).
9


Most studies investigating the association between ADT and QOL have been
carried out using continuous administration of the anti-androgenic therapy. Very few
studies have reported results of intermittent ADT regimens, and even fewer studies have
compared the two methods of administering ADT with regard to their specific impacts on
QOL. Salonen et al. (2012) investigated whether there are differences with regard to the
impact on QOL in men undergoing continuous versus intermittent ADT. Specifically,
these authors focused on the QOL areas of pain, activity limitation, bed disability,
physical capacity, sexual functioning, social functioning, emotional well-being, vitality,
and overall health. These 10 domains were assessed with a 30-item Health-Related QOL
self-report questionnaire. Their results indicate that intermittent ADT is associated with
better outcomes in the domains of activity limitation, physical capacity, and sexual
functioning (Salonen, Taari, Ala-Opas, Viitanen, Lundstedt, & Tammela, 2012).
However, this study did not control for potentially important covariates, including
variables such as patients age, BMI, comorbidity, or other factors that may also have an
impact on QOL. Therefore it is not entirely clear whether the dosing schedule of ADT or
other variables are implicated in the mechanism leading to reductions in QOL.
Impact of Other ADT Side Effects on QOL
In the previous sections, several studies were mentioned that found that factors
like BMI, disease stage, comorbidities, time since treatment initiation, and age may have
an effect on the association between ADT and QOL. Another factor that could play a role
in decreasing QOL in prostate cancer patients undergoing ADT is cognition.
A review of cognition among men with prostate cancer who are on ADT reports that,
only a few relatively small studies have investigated the impact of this treatment on
10


cognitive functioning, and some of those studies reported contradictory results (Nelson,
Lee, Gamboa, & Roth, 2008, p. 1099). Nelson and associates reported that cognitive
changes appear to be subtle and affect specific domains. Overall, 47 to 69% of study
participants showed a decline in at least one task area. Approximately 14% of ADT
patients experienced declines in two or more cognitive domains. Visuospatial abilities,
working memory, and verbal fluency were frequently affected in ADT patients.
Additionally, declines in executive functioning were also seen in patients undergoing
ADT treatment. The review found several contradictory results reported for verbal
memory ability (Nelson, Lee, Gamboa, & Roth, 2008).
Cognitive impairment also has frequently been associated with poorer QOL
(Moore et al., 2000). Therefore, it would be important to know whether the association
between lowered testosterone levels due to ADT and decreases in QOL persists after
controlling for impacts of ADT on cognition and other patient characteristics. No study
identified in this review of the literature has examined whether continuous and
intermittent ADT have differential impacts on QOL, and whether this association is
independent of the influence of various covariates on QOL. Moreover, the nature of the
association between testosterone levels and QOL in this population, regardless of
treatment type, remains unclear.
Objectives
1) Assessment of the differential impact of continuous (CON) and intermittent (INT)
ADT on QOL among prostate cancer patients.
2) Assessment of covariates influencing the association between ADT regimen type
and QOL.
11


3) Evaluation of relationships between biologically available testosterone levels and
QOL at all treatment timepoints.
Hypotheses
1. ADT regimen type will be associated with differences in levels of overall
QOL, controlling for participant characteristics.
a. Participants undergoing continuous ADT will show lower levels of overall
QOL as compared to intermittent ADT after adjusting for the following
covariates: age, education, BMI, diagnosed diabetes mellitus, time since
diagnosis, and cognition.
2. ADT treatment regimen (CON vs. INT) will have a differential impact on
specific aspects of QOL.
a. Greater impact on specific areas of QOL will be seen in participants
receiving continuous ADT as compared to intermittent ADT.
3. Levels of biologically available testosterone (BioT) will be positively
correlated with levels of QOL regardless of ADT regimen, while controlling
for multiple observations within participants, such that low levels of
testosterone will be associated with lower QOL.
12


CHAPTER II
METHOD
Procedures
This proposed study, investigating the effect of ADT on QOL in males with
prostate cancer, was funded by the Department of Defense {Effects of androgen blockade
on cognitive function and quality of life in men with prostate cancer. Department of
Defense Prostate Cancer Research Program (PCRP), U.S. Army Medical Research and
Materiel Command, Office of Congressionally Directed Medical Research Programs
(CDMRP), PC010257). The project was a supplement to an NCI-funded project, Phase
III randomized study of intermittent versus constant combined androgen deprivation
(Bicalutamide and Goserelin) in patients with Stage IVprostate cancer responsive to
such therapy, SWOG protocol number 9346. The DoD-funded study recruited patients
who were participants in the clinical trial through the urologic oncology clinic of Michael
Glode, MD, at the University of Colorado Health Sciences Center, and through Lawrence
Karsh, MD, of Western Urological Associates. Persons eligible for this study were males
aged 50 and older who met the inclusion and exclusion criteria for the clinical trial. The
Colorado Multiple Institutional Review Board (COMIRB) approved the study. The
following inclusion and exclusion criteria were set for the clinical trial:
13


Table 1. Clinical Trial Inclusion and Exclusion Criteria
Inclusion Exclusion
Adenocarcinoma of the prostate, with or without metastases to bone, brain, liver, or lung Concurrent biological response modifier therapy or chemotherapy
Elevated PSA (5 ng/mL or greater) Concurrent hormonal therapy and at least one year since any prior neoadjuvant or adjuvant hormonal therapy
Age > 50 Any prior finasteride treatment (drug for the treatment of benign prostatic hyperplasia and male pattern baldness) Concurrent radiotherapy other than palliation of painful bone metastases Prior bilateral orchiectomy Active medical illness precluding treatment or limiting survival Second malignancy within five years except adequately treated nonmelanomatous skin cancer, in situ bladder cancer or other superficial cancer
History of neurologic disorder, head trauma with loss of consciousness, learning disability, mental retardation, history of alcoholism, or psychosis
Recruitment criteria set forth by the parent clinical trial apply to this project.
The parent study included patients with and without metastasized prostate
cancers. For inclusion, the density of prostate-specific antigen collected through blood
samples had to be greater than 5 ng/mL, which can be suggestive of prostate cancer. In
order to reduce the effect of confounding variables, participants were excluded if they
were undergoing chemotherapy (except for the reduction of pain for bone metastases) or
any other type of hormone therapy concurrently with this study, or showing the effects of
recent radiation or hormone therapy. Participants were also ineligible if they had a prior
bilateral orchiectomy, a surgical intervention that removes both testes to eliminate the
primary source of testosterone production. This study also excluded participants with
14


significant comorbid chronic medical illnesses, to reduce the risk of attrition due to
increased mortality and time demand. In order to focus on prostate cancer and its
associated malignancies, patients were also excluded if they had any other untreated
malignancy within the past five years. Confounding influences on QOL and cognitive
performance were minimized by excluding participants with a history of neurologic
disorders, head trauma with loss of consciousness, learning disability, intellectual
disability, history of alcoholism, or psychosis. During baseline assessment it became
evident that several recruited participants met at least one of the exclusion criteria. These
participants have been excluded from the data analysis.
The primary data collection sites were the University of Colorado Health Sciences
Center (UCHSC) in Aurora, Colorado, and Western Urological Associates, a private
practice in the Denver area. A clinical research associate at UCHSC who was thoroughly
familiar with the clinical trial protocol obtained participant consent. After signing an
informed consent form, the participants were randomly assigned to continuous or
intermittent ADT and were given a copy of the consent form and the original document
was filed in appropriate clinic charts.
At the time this study was designed, it was widely thought that suppressed
testosterone levels would return to baseline between 8 and 26 weeks (Akakura et al.,
1993; Nejat et al., 2000), but it was subsequently found that the range was frequently
much greater than this (e.g., Bong et al., 2008). At the outset of the study, patients on
continuous ADT were examined at three timepoints: 1) at baseline (before ADT) or
shortly after beginning ADT; 2) after 12-16 weeks of continuous ADT; and 3) after
another 12-16 weeks of continuous ADT. For those subjects on the intermittent regimen,
15


the analogous timepoints were: 1) at baseline, if possible, or shortly after beginning ADT;
2) 12 to 16 weeks following resumption of androgen blockade; and 3) 12 to 16 weeks
after the next discontinuation of androgen blockade. It was not possible to have access to
all participants before they began ADT treatment, as the medication was usually
administered in the physicians office, following a decision by patient and physician,
shortly before referral to the study (generally 3-10 days). Given what was known of the
pharmacodynamics of ADT, it was assumed that the baseline level of testosterone, and its
effects on QOL, would most likely still be observable. However, assessments of QOL
and cognition may be slightly less pronounced in participants who do not have a true
baseline assessment at timepoint 1.
Subjects underwent a set of cognitive tests and questionnaires during each of the
three assessment timepoints, requiring approximately 90 minutes. Therefore, each subject
participated for a total of 4 to 5 hours. To minimize fatigue, there were breaks midway
through administration, and more frequently if needed. Data collectors were trained to
evaluate fatigue and were asked to reschedule testing if fatigue appeared to affect
performance. The order of test administration was randomized across subjects to prevent
unanticipated order effects in the cognitive assessment. To minimize circadian variability
in performance, subjects were scheduled for assessments at the time of day that they
preferred and these assignments remained as constant as possible for each subject across
the three timepoints. Data were double-entered.
Sample
This study recruited 73 males, aged 50 to 81, with stage IV prostate cancer.
Participants were diagnosed with prostate cancer between 1989 and 2005. Data were
16


collected between 2003 and 2005. Approximately 97% of individuals from the sample
had past chemotherapy treatment; however, patients with current effects of chemotherapy
were excluded. Participants were predominately white. One individual identified himself
as African American, and four as multiracial. Participants had on average completed 15.3
years of education. Attrition of participants due to death or other reasons occurred at the
second and third timepoints as depicted in figure 1.
Study Flow Chart
FIGURE 1. Study Flow Chart and Attrition over Time
17


Instruments
Quality of Life
Quality of life (QOL) describes multiple domains of individual (or sometimes
group) well-being. Such domains can be assessed subjectively or objectively and often
include psychological, social, and physical aspects of functioning (Felce & Perry, 1995).
SF-36 Health Survey
QOL was assessed using the SF-36 Health Survey, a 36-item general health status
instrument. On this measure, the impact of health on normal activities is analyzed using
eight subscales: Physical Functioning, Pain, General Health, Role Limitations Due to
Physical Health, Role Limitations Due to Emotional Health, Energy, Emotional
Wellbeing, and Social Functioning. Each item contributes to only one subscale, on which
scores range from 0 to 100, with higher scores indicating better QOL. The information
obtained from the SF-36 can be broken down into three levels: (1) items; (2) subscales;
and, (3) summary measures, which combine subscales. The two summary measures can
be used to assess two broader aspects of QOL. The Physical Health summary measure
combines the Physical Functioning, Role Limitations Due to Physical Health, Pain, and
General Health subscales, and the Mental Health measure combines the Role Limitations
Due to Emotional Health, Emotional Wellbeing, Social Functioning, and Energy
subscales (Ware, 2002). Certain subscales, such as Emotional Wellbeingindicating
moodprovide unique information needed to examine the effect of ADT. Therefore
analyses will incorporate the eight main scales to obtain a detailed analysis of the
association between ADT regimen and testosterone on QOL.
18


Internal consistency for most subscales is above 0.8, except for the Social
Functioning subscale (0.76), which might in part be due to the small number of items
comprising that subscale. The average Cronbach alpha values for all eight subscales in
people with poor health status was 0.82 (Jenkinson, 1994). The reliability of the Physical
Health summary score is 0.92, and for the Mental Health summary score it is 0.88 (Ware,
2002). Support for the predictive validity of the two summary measures has also been
reported. For example, individuals with very high scores on the Mental Health summary
measure (e.g., 60-74) were approximately one third less likely than those individual with
lower summary scores (55-59 and 50-54) to receive a diagnosis of depression, and were
about one-third more likely to self-report higher levels of life satisfaction. The Physical
Health summary measure has also been found to predict risk of mortality. For example, a
40% increase in mortality rates is seen in individuals with scores in the 8-24 range as
compared with test-takers in the 25-34 range on the Physical Health summary measure
(Ware & Kosinskia, 2001).
Depression
Depression and its associated emotions and cognitions can affect an individuals
perception of QOL, and is therefore a closely related construct. A measure of depression
will therefore be included in this study to supplement the assessment of mental health
aspects of QOL.
Center for Epidemiologic Studies Depression Scale (CES-D)
Depression was evaluated using the Center for Epidemiologic Studies Depression
Scale (CES-D, Radi off, 1977). This 20-item self-report questionnaire contains fewer
items that might reflect physical disability than do other depression scales. The CES-D
19


has been found to be a reliable and valid measure of depressive symptomatology in
cancer populations. Among a group of breast cancer patients, the internal consistency of
the CES-D showed alpha coefficients above 0.85. It also correlated in expected directions
with related concepts such as fatigue, anxiety, and global mental health functioning,
indicating construct validity for the measure (Hann, Winter, & Jacobsen, 1999). Scores
for the CES-D range from 0 to 60, with higher scores indicating the presence of more
depressive symptoms.
Speed and Capacity of Information Processing
Information processing involves the integration and transformation of information
from the environment with the help of several cognitive systems, such as attention,
working memory, and long-term declarative memory (McLeod, 2008).
Symbol Digit Modalities Test (SDMT)
This test requires participants to pair abstract symbols with specific numbers as
quickly as possible for 90 seconds. In addition to measuring the construct of speed of
information processing, the SDMT requires attention and integration abilities. A total
score, which will be used for the analyses of this study, is calculated by adding the items
(0-110) that were correctly coded in 90 seconds and subtracting the number of incorrectly
coded items from that sum.
The construct validity of the SDMT is acceptable to good. One study investigating
cognitive impairment after athletic concussions found that the SDMT correlated 0.62 to
0.91 with the Wechsler Digit Symbol subtest (Hinton-Bayre, Geffen, Geffen, McFarland,
& Friis, 1999). In patients with multiple sclerosis, the SDMT correlated with the Paced
20


Auditory Serial Addition Test (PASAT) (r= 0.54), which also assesses the capacity and
rate of information processing. The same study observed a test-retest reliability of 0.74
(Hojnacki, Munschauer, Morrow, Weinstock-Guttman, Drake, & Benedict, 2010).
Individuals with multiple sclerosis often experience a wide range of symptoms that could
include physical, mental, and potential psychiatric problems. These symptom
combinations are somewhat different from what can be seen in prostate cancer patients.
The reported psychometric properties will therefore also be interpreted with caution.
Declarative Verbal Learning and Memory
Declarative memory refers to information and events stored and consciously
retrieved from long-term memory systems (Mastin, 2010).
Rev Auditory Verbal Learning Test (RAVLT)
This word-learning test consists of two lists, each with 15 unrelated concrete nouns.
The first list is repeated over five trials, with recall requested after each presentation of
the words. After the fifth trial, the second list of words is read to the participant, with
immediate recall of that list (the interference trial). After a 20-minute delay, recall is
tested for the first list without any further presentation of either list. This study will
analyze the delayed recall score, which is the total number of words from the first list that
are recalled correctly after a 20-minute delay.
De Sousa Magalhaes et al. (2012) examined test-retest correlations for male and
female college students aged 17 through 40. The strongest test-retest correlations were
found for the sum of scores obtained over the first five learning trials (0.96). The test-
retest correlation for words recalled after 20 minutes was 0.76. This score will be used in
21


my study as an indicator of delayed verbal memory. The Cronbach alpha reliability of the
RAVLT was 0.84, which indicates good internal consistency among the items of this test.
The RAVLT demonstrated divergent validity with the Trail Making Test, which
evaluates attention and executive function rather than verbal learning and memory.
Evidence for convergent validity is seen in the correlation between the RAVLT and the
Benton Visual Retention Test, which evaluates immediate and delayed visual memory
(De Sousa Magalhaes, Fernandes Malloy-Diniz, & Cavalheiro Hamdan, 2012).
Verbal Fluency
Verbal fluency encompasses the retrieval of verbal information from long-term
memory systems. Recall of this type of information requires executive control over
several cognitive processes, such as selective attention, inhibition, and self-monitoring
(Lezak, Howieson Loring, Hannay, & Fischer, 2004).
Controlled Oral Word Association Test (COWAT)
This measure of verbal fluency asks test-takers to name as many words as possible
that begin with the letter F in 60 seconds. The procedure is then repeated twice, but for
words that begin with the letters A and S. The number of unique words is summed across
the three trials. My analyses will use an age-adjusted scaled score (range: 2-18), which is
obtained by comparing the raw score to data obtained from a normative sample table.
The test-retest reliability for this test of verbal fluency is 0.74. On average, test-
takers gain 3 words upon being retested, which could indicate a small practice effect.
Education can significantly affect COW AT scores, and may account for up to 8% of the
total observed variance (Ruff, Light, Parker, & Levin, 1996). The total number of novel
22


words generated by test-takers in the COW AT correlates more strongly with scores on
the vocabulary subtest of the Wechsler Adult Intelligence Scale Third Edition (WAIS-
III) (0.22) than with performance on the Tower of Hanoi (0.03), a measure of procedural
learning and executive ability (Ross, Calhoun, Cox, Wenner, Kono, & Pleasant, 2007).
All of these estimates were obtained in samples of healthy adults and did not consider the
impact of aging or health impairment. This studys sample has completed an above
average number of years of education, which could increase overall testing performance
on this and other tests in this study. The absolute scores found in this study and
psychometric properties reported in the literature will therefore be interpreted with
caution.
Hormone Assays
Hormone levels were measured from 3 ml blood samples drawn from participants
who had fasted for at least 1-2 hours before the blood draw. Because the free hormone
level is the physiologically active component, total levels of testosterone, sex hormone
binding globulin (SHBG), and albumin were also obtained, and from these, free sex
hormone levels were calculated.
On average, 50% of circulating testosterone is nonspecifically or weakly bound to
albumin, and 44% is specifically bound to SHBG. Another 3.5% of testosterone
contained in plasma is associated with cortisol binding globulin. This leaves 2-3% of total
testosterone levels free or in an unbound state (De Ronde, et al., 2006). Unbound, or free
testosterone (FT) is often regarded as the portion accessible for androgenic effects.
Another estimate of bioavailable testosterone levels is BioT, which is the total amount of
FT and testosterone that is weakly bound to the protein albumin. Even though direct
23


measures of FT and BioT are often preferred due to their accuracy, the position statement
of The Endocrine Society recommends the calculation of FT and BioT as a useful and
cost-effective alternative to indicate available testosterone levels (Rosner, Auchus, Azziz,
Sluss, & Raff, 2007).
However, the calculation of FT and BioT is done using different approaches and
assumptions, and can therefore yield significantly different results when used as an
independent variable. For example, the Vermeulen and Sondergard approach is based on
the law of mass action. These algorithms use values for total testosterone, SHBG,
albumin, as well as association constants for testosterone binding to proteins such as
albumin and SHBG to generate FT values. Other approaches, such as those developed by
Morris, Ly, or Emadi-Konjin, base their algorithms on numerous direct laboratory
measurements of FT. These results were used to create an algorithm to predict future FT
(De Ronde, et al., 2006). However, these empirical equations may lack accuracy and
comparability among different labs. Algorithms based on the law of mass action are
highly dependent on accurate measurement of total testosterone and SHBG (Rosner,
Auchus, Azziz, Sluss, & Raff, 2007).
In this study, BioT was calculated according to the Vermeulen (Vermeulen,
Verdonck, & Kaufman, 1999) method. This approach was used because the algorithm is
less likely to over- or underestimate BioT concentrations in older men, who tend to have
higher SHBG concentrations. It has also been suggested that the inclusion of measured
albumin concentrations, rather than the use of standard reference values of this protein,
becomes important at lower BioT concentrations, which are expected in this study (De
Ronde, et al., 2006).
24


The Endocrine Society has proposed the following guidelines for testosterone: total
testosterone (Total T) levels greater than 320 ng/dl (11.1 nmol/L) are considered normal.
Total T values less than 200 ng/dl (6.9 nmol/L) are indicative of hypogonadism, but Total
T results between 200-320 ng/dl (6.9 -11.1 nmol/L) are considered borderline. A free
testosterone (FT) value of 6.5 ng/dl (0.23 nmol/L) is considered the lower limit of the
normal range (Rosner, Auchus, Azziz, Sluss, & Raff, 2007). These reference values can
be used to determine whether ADT interventions have been effective at reducing Total T
and FT levels.
Analysis
Preliminary Data Analysis. Frequency distributions and descriptive statistics
were obtained to evaluate data integrity and to characterize the intervention groups.
Simple correlations between group membership, measures of BioT, QOL scales and
depression, as well as cognition were calculated to assess patterns of associations among
these variables.
Hypothesis Testing.
Hypothesis 1
ADT regimen type will be associated with differences in levels of overall
QOL, controlling for participant characteristics. Participants undergoing
continuous ADT will show lower levels of overall QOL in adjusted analyses.
Using multivariate analysis of covariance (MANCOVA), this first hypothesis
examined the main effect of treatment group membership (independent variable IV) on
a composite QOL variable (dependent variable DV). The newly created composite DV
within the MANCOVA framework represents a linear combination of all included DVs,
25


and acts to maximize group differences. The test for main effects assesses whether mean
differences in the composite DV among groups at different levels of the IV are due to
chance while controlling for the influence of covariates. The DV will consist of the eight
SF-36 QOL subscales and CES-D depression questionnaire. The following covariates
were included due to their potential influence on QOL: age, education, BMI, diagnosed
diabetes mellitus (DM), time since diagnosis, and cognition. Covariates assessed at
baseline included age, years of education, diagnosis of diabetes mellitus, BMI, and time
since cancer diagnosis. Measures of cognition (i.e. processing speed, verbal learning, and
verbal fluency) at timepoint 3 were included to capture the impact of androgen
suppression on cognitive functioning.
In order to assess the influence of initial levels of QOL, an additional
MANCOVA will be computed for the same outcome variables obtained during baseline
assessments. A non-significant overall effect would indicate that the treatment groups did
not differ with regard to their QOL while controlling for covariates at the onset of the
study.
26


Multivariate Analysis of Covariance
(MANCOVA)
£>
r n ..... .......
Covariates ..........
(Age, Education, BMI, Diabetes, .......
Time Since Diagnosis,
Cognition)
FIGURE 2. MANCOVA Testing Hypotheses 1 and 2
Hypothesis 2
ADT regimen (continuous versus intermittent) will have differential effects
on specific aspects of QOL, such that greater impact will be seen in participants
receiving continuous ADT as compared to intermittent ADT.
Analysis of covariance (ANCOVA) is used to assess the relative contribution of
the included dependent variables while controlling for the covariates described above.
These tests of between-subjects effects assess whether ADT treatment group membership
is associated with differences in each of the tested domains of QOL, while controlling for
the influence of other participant characteristics. These analyses are computed
concurrently with the MANCOVA main effect in the IBM SPSS statistical software
package.
MANCOVA is a statistical technique used to assess the impact of group
differences on several DVs while controlling for the influence of covariate variables. This
technique avoids inflation in Type I error rate because the number of comparisons to be
analyzed is reduced. This technique may also show differences among variables that may
Independent Variable
Group Membership
Continuous or Intermittent ADT
Dependent Variables
Quality of Life
at Timepoint 3
(Physical Functioning, Role
Limitations due to Physical
Health, Pain. General Health,
Energy, Role Limitations due to
Emotional Health, Emotional
Wellbeing, Social Functioning,
Depression)
27


not become apparent when using several ANOVA tests for the same research question.
The MANCOVA technique is based on several assumptions that need to be met to ensure
robustness of the statistical test. First, the DVs and all linear combinations of them should
be normally distributed. Univariate and multivariate outliers can cause distributions to be
skewed and may need to be removed. Variables with outliers or non-normally distributed
variables may also need to be transformed in order to meet the assumption for normality.
Second, it is assumed that all associations between pairs of DVs, all pairs of covariates,
and all DV-covariate pairs are linear. Non-linear associations between variables reduce
the power of the test. Lastly, all DVs should exhibit equal levels of variance across the
different levels of the IV, covariates, and with other DVs.
Bivariate Correlation:
Hypothesis 3
Levels of biologically available testosterone (BioT) will be positively
correlated with levels of QOL regardless of ADT regimen while controlling for
multiple observations within participants, such that as testosterone declines, so will
different domains of QOL.
Multilevel correlations were obtained to determine the direction and strength of
the association between levels of testosterone and different domains of QOL regardless of
ADT regimen, while controlling for multiple observations within people. The eight scales
of the SF-36 and CES-D questionnaire were included to assess physical and mental
health aspects of QOL.
28


This study used IBM SPSS version 21 to compute the MANCOVAs testing
hypotheses 1 and 2. Hypothesis 3 will be computed using MPlus version 3.7. An alpha
level of 0.05 will be used to determine statistical significance.
Statistical Power
This section reviews the effect sizes of prior findings in order to determine
whether this study has adequate power. Effect sizes (Cohens d and f) for the association
between testosterone levels after non-hormone-dependent cancer treatment and QOL,
using the SF-36 subscales, have been shown to range from d = 0.22 to 0.67, depending on
the subscale being analyzed (Greenfield, et al., 2010). Based on these effect sizes, I
anticipated a medium effect (d= 0.40, f2= 0.15, or p= 0.3). MANCOVA was used in order
to test hypotheses 1 and 2. The necessary sample given the expected effect size can be
obtained using gPower (version 7.3), which does not directly estimate sample size for
MANCOVA. However, according to Dattalo (2008) the estimation strategies for
MANCOVA can be adapted in order to obtain such estimates. The sample size needed to
detect a medium effect (f2 = 0.15) with a power of 0.8 at a = 0.05 using MANCOVA,
with 2 groups, 8 covariates, and 9 response variables is 44. There are currently no agreed-
upon guidelines for determining power for multilevel correlation analyses. Therefore,
statistical power for hypothesis 3 will be determined using guidelines for bivariate
correlations, which will yield a more conservative estimate, as this statistical technique
does not adjust for shared variance across multiple observations. A sample size of 67 is
needed in order to detect a medium effect (p= 0.3) with power of 0.8 and a = 0.05 for the
correlations tested in hypothesis 3. Data were collected from 73, 70, and 60 participants
at timepoints 1, 2, and 3, respectively. Therefore, this study should have adequate power
29


to detect significant effects for most hypothesis tests. Effect sizes will be reported in
addition to the results from significance testing.
Study Design
This study represents a experimental research design in which two groups of
patients received either continuous or intermittent ADT. The sample was followed
prospectively over three assessment timepoints. Outcomes at each timepoint were
evaluated at approximately 12 to 16 week intervals. This study considers the last QOL
assessment timepoint as the outcome measure, and controls for certain participant
characteristics that may also be associated with QOL in examining hypothesis 1 and 2.
All timepoints were considered in testing hypothesis 3. Participants were selected based
on their exposure to the antecedent conditions of prostate cancer and eligibility for
treatment with ADT. The parent clinical trial randomly assigned participants to the
continuous or intermittent treatment administration groups.
This design is capable of establishing a temporal sequence for ADT and decline in
QOL. Because no control group has been included in the design, this study cannot
determine how prostate cancer alone affects QOL. Additionally, the study groups were
recruited from another clinical trial involving an urban university hospital setting and a
large urban urology private practice. Therefore this projects sample may not represent
the entire population of stage IV prostate cancer patients undergoing ADT. The analyses
for this study include covariates that potentially differ among groups and could have a
separate influence on QOL. Assessing group differences for baseline levels of QOL could
provide further evidence for the association between different ADT regimens and QOL
that is independent of initial levels of OQL.
30


The prospective design involved some attrition of participants due to death or
other reasons (n = 13). Participants who only completed the first two assessments were
not included in the analyses for hypothesis 1 and 2, but had scores that could be analyzed
to test hypothesis 3.
31


CHAPTER III
RESULTS
Preliminary Data Analyses
Descriptive statistics and bivariate correlations were obtained in order to
characterize the structure of the dataset and identify patterns of associations among the
variables. Table 2 summarizes the distribution of the dependent variables, including the
eight QOL subscales of the SF-36 and CES-D depression questionnaire, at the time of
baseline assessment. The SF-36 Physical Functioning, Role Limitations Due to Physical
Health, Pain, and General Health subscales belong to the QOL construct of physical
health. The SF-36 Emotional Wellbeing, Role Limitations Due to Emotional Health,
Energy, and Social Functioning subscales as well as the CES-D depression scale are
thought to belong to the mental health construct of QOL. Higher scores on the SF-36
subscales indicate higher levels of QOL whereas higher scores on the CES-D indicate the
presence of more depressive symptoms. All variables are reported in their original metric
unless otherwise specified. The values of skewness and kurtosis indicate that all
dependent variables assessed at baseline were adequately normally distributed and were
therefore suited for further analyses within the general linear model. No univariate
outliers were detected during the initial data screening.
32


Table 2. Descriptive Statistics of Independent and Dependent Variables at Baseline
N Mean Standard Deviation Range Skewness Kurtosis
Physical Intermittent 33 79.24 21.07 70 -0.09 -0.41
Functioning Continuous 39 71.03 21.22 90 -0.98 0.47
(SF-36) Total 72 75.14 21.46 90 -0.88 0.04
Role
Limitations Due Intermittent 34 67.65 38.70 100 -0.83 -0.86
to Physical Continuous 38 58.55 35.96 100 -0.36 -1.17
Health (SF-36) Total 72 63.36 37.29 100 -0.57 -1.10
Pain Intermittent 34 74.49 19.52 78 -0.54 -0.05
(SF-36) Continuous 39 74.10 20.91 78 -0.87 0.33
Total 73 74.63 20.22 78 -0.73 0.10
General Health Intermittent 34 48.53 9.96 40 -0.04 -0.40
(SF-36) Continuous 39 44.23 10.92 45 -0.34 -0.51
Total 73 46.49 10.78 50 -0.21 -0.33
Emotional Wellbeing Intermittent 33 83.27 11.77 44 -0.81 -0.05
Continuous 39 80.41 14.12 56 -0.83 0.07
(Sr -36)
Total 72 81.64 13.01 56 -0.85 0.16
Role
Limitations Due Intermittent 34 87.25 23.23 67 -1.59 1.12
to Emotional Continuous 39 72.65 32.33 100 -1.11 0.38
Health (SF-36) Total 73 79.73 29.10 100 -1.37 1.06
Energy Intermittent 34 60.88 17.56 65 -0.80 0.28
(SF-36) Continuous 39 58.72 17.08 70 -0.01 -0.55
Total 73 60.00 17.26 75 -0.39 -0.34
Snr.ial
Functioning (SF-36) Intermittent 34 51.10 3.60 13 3.04 7.69
Continuous 39 49.68 7.30 25 0.00 0.17
Total 73 50.34 5.84 25 0.97 1.81
Depression Intermittent 30 10.27 7.35 24 0.38 -0.95
(CES-D) Continuous 35 13.11 6.05 32 0.88 2.38
Total 65 11.83 6.73 33 0.40 0.33
Table 3 summarizes the descriptive statistics obtained from the nine dependent
variables at the third assessment timepoint, which was used in this study to assess
treatment outcomes. The values for skewness and kurtosis indicate that all dependent
variables were also normally distributed and therefore suited for further analyses within
the general linear model. No univariate outliers were detected during the initial data
33


screening. Among the SF-36 subscales, scores on the Social Functioning domain showed
the smallest range and could limit the amount of variability that could be analyzed during
hypotheses testing.
Table 3. Descriptive Statistics of Independent and Dependent Variables After
Androgen Deprivation Therapy (ADT)
N Mean Standard Deviation Range Skewness Kurtosis
Physical Intermittent 32 81.25 18.49 70 -1.41 1.61
Continuous 28 70.89 20.28 70 -0.70 -0.43
Functioning (SF-36) Total 60 76.72 19.85 70 -0.99 0.15
Role Intermittent 31 66.13 38.49 100 -0.82 -0.82
Limitations Due Continuous 28 55.36 41.03 100 -0.26 -1.64
to Physical Total 59 61.67 39.72 100 -0.55 -1.30
Health (SF-36)
Intermittent 32 75.23 22.23 90 -1.10 1.50
Pain Continuous 27 72.13 24.14 87 -0.65 -0.15
(SF-36) Total 59 73.88 22.78 90 -0.87 0.51
Intermittent 32 49.38 10.14 45 -0.56 0.35
General Health Continuous 28 46.96 12.12 55 0.44 0.86
(SF-36) Total 60 48.44 11.09 55 -0.06 0.34
Emotional Wellbeing Intermittent 32 84.88 13.35 56 -1.52 2.36
Continuous 28 83.14 11.07 40 -0.31 -0.89
(SF-36) Total 60 84.00 12.18 56 -1.05 1.13
Role Intermittent 32 80.21 27.90 100 -1.27 0.81
Limitations Due Continuous 28 77.38 32.78 100 -1.22 0.29
to Emotional
Health (SF-36) Total 60 79.24 29.92 100 -1.26 0.55
Intermittent 32 66.56 16.87 60 -0.28 -1.05
Energy Continuous 28 58.93 18.73 70 -0.64 -0.25
(SF-36) Total 60 63.2 17.94 70 -0.52 -0.33
Social Intermittent 31 51.21 6.74 25 0.10 0.70
Continuous 28 51.79 7.39 37.5 1.14 3.27
Functioning
(SF-36) Total 59 51.46 6.94 37.5 0.67 2.07
Intermittent 30 7.6 6.77 31 1.71 3.77
Depression Continuous 27 11.22 6.01 24 0.46 0.12
CCES-D'I
Total 57 9.55 6.80 31 0.92 0.74
Descriptive statistics at timepoint 3 approximately 24 to 32 weeks after initiation of
continuous or intermittent ADT
34


The majority of mean scores on most SF-36 subscales were lower and the CES-D
mean score was higher in the continuous ADT treatment group at baseline as well as at
timepoint 3. This could indicate lower levels of QOL in the continuous ADT group.
Statistical tests assessing hypotheses 1 and 2 will be able to confirm the existence and
clarify the nature of potential group differences.
The distributions of BioT levels at all assessment timepoints are summarized in
table 4. Scores on this variable among all participants appear to be normally distributed.
However, when the two intervention groups are considered separately, the distribution of
BioT levels among participants of the continuous ADT group at all assessment timepoints
may be too peaked (i.e. leptokurtic) and somewhat positively skewed. Participants in the
continuous ADT intervention group had overall lower mean levels of BioT and a smaller
range at timepoint 2 and 3 overall as a group. The intermittent group by itself has a
greater range of scores and is normally distributed.
Table 4. Descriptive Statistics of Bioavailable Levels of Testosterone (BioT)
N Mean Standard Deviation Range Skewness Kurtosis
Intermittent 34 6.48 3.42 14.51 0.12 -0.15
Baseline Continuous 38 1.3 2.2 10.35 2.74 7.93
Total 72 3.85 3.91 14.51 0.77 -0.53
Intermittent 32 6.40 2.98 14.00 -0.14 0.89
Time Point 2 Continuous 38 0.76 1.27 6.00 3.46 11.62
Total 70 3.43 3.64 14.00 0.76 -0.57
Intermittent 32 6.75 2.89 11.69 -0.67 0.24
Time Point 3 Continuous 28 1.34 1.89 6.60 2.02 3.09
Total 60 4.34 3.75 11.69 0.30 -1.34
Descriptive statistics of bioavailable testosterone level at all assessment timepoints
Levels of BioT measured in nmol/L
Levels of BioT include levels of free testosterone (FT) and the portion of
testosterone bound to the protein albumin. Average levels of FT among participants in the
continuous ADT group of this study ranged from 0.03 to 0.06 nmol/L and between 0.29
35


and 0.40 nmol/L in the intermittent group (data not included in table 4). Given that FT
values below 0.23 nmol/L are considered outside of the normal range, participants in the
continuous group were suppressed well below normal levels of FT, whereas participants
testosterone levels in the intermittent intervention group were suppressed to levels within
the lower boundaries of the normal range.
Descriptive statistics of the covariates included in statistical tests assessing
hypotheses 1 and 2 are summarized in table 5. The mean age of participants in this study
was 72.16, and they had completed an average of 15.27 years of education. The mean
BMI score of 29.08 is situated at the upper boundary of the overweight range. The
majority of participants did not report having a diagnosis of type-2 diabetes mellitus. The
diabetes variable, being dichotomous, is the only covariate that is not normally
distributed.
Table 5. Descriptive Statistics of Covariates
N Mean Standard Deviation Range Skewness Kurtosis
Age 74 72.16 5.32 26 -0.5 0.55
Education 74 15.27 3.02 13 0.17 -0.84
BMI 69 29.08 5.13 25.57 0.52 0.66
Presence of Diabetes 69 1.91 0.28 1 -2.99 7.19
Year of Diagnosis 64 1998 4.58 16 -0.32 -0.95
Processing Speed (SDMT) 60 46.93 9.06 48 -0.61 0.82
Verbal Learning (RAVLT) 61 0.84 0.21 1.3 -1.64 4.69
Verbal Fluency (COWAT) 61 38.25 13.15 67 1.27 2.29
Descriptive statistics assessed at baseline (demographic data) and at timepoint 3 (cognition)
36


Table 6 includes all pairwise correlations between the independent variable,
dependent variables, and covariates. All QOL dependent variables as well as the three
aspects of cognition (covariates) were assessed at timepoint 3. All remaining covariates
were assessed at baseline. Having received continuous ADT was associated with lower
levels of Physical Functioning (r = .454, p < .01) as well as greater depressive
symptomatology (r = -.267, p < .05). Other notable significant correlations include the
associations between the Physical Functioning subscale and Role Limitations Due to
Physical Health (r = .454, p < .01), Pain (r = .511, p < .01), General Health (r = .391, p <
.01), Emotional Wellbeing (r = 294, p < .05), Role Limitations Due to Emotional Health
(r = 314, p < .05), and Energy subscales (r = .537, p < .01). The Role Limitations due to
Physical Health subscale was significantly correlated with the Pain (r = .398, p < .01),
General Health (r = .255, p < .05), Role Limitations Due to Emotional Health (r = .494, p
< .01), and Energy subscales (r = .421, p < .01). The Pain subscale of the SF-36 QOL
questionnaire was additionally correlated with the General Health (r = .308, p < .05),
Emotional Wellbeing (r = .467, p < .01), Role Limitations Due to Emotional Health (r =
.470, p < .01), Energy (r = .580, p < .01), Social Functioning (r = -.257, p < .05), and
CES-D depression scales (r = -.429, p < .01). The Pain subscale was also positively
correlated with the Education variable (r = .277, p < .05). The General Health subscale
was associated with the Energy subscale (r = .317, p < .05). The Emotional Wellbeing
subscale was also associated with the Role Limitations due to Emotional Health (r = .573,
p < .01), Energy (r = .556, p < .01), and CES-D depression scales (r = -.700, p < .01). The
Emotional Wellbeing subscale was additionally associated with performance on the
COW AT (r = .265, p < .05). The Role Limitations Due to Emotional Health subscale was
37


correlated with the Energy (r = .509, p < .01) and CES-D depression scales (r = -.451 ,P<
.01). Lastly, the Energy subscale was significantly associated with the CES-D depression
scale (r = -.621 ,P< .01). The Energy subscale was associated with the CES-D depression
scale (r = -.569, p < .01). For the majority of SF-36 QOL domains, higher levels on one
scale are correlated with higher scores on the other, except for the association between
Social Functioning and Pain subscales on the SF-36. All associations between SF-36
subscales and CES-D scale indicate that higher levels of QOL are associated with fewer
depressive symptoms. Additionally, higher levels of education are positively correlated
with performance on tests of processing speed (r = .363, p < .05) and verbal fluency (r =
.320, p < .05).
38


On
m
Table 6. Bivariate Correlations Among the Independent Variable, Dependent Variables, and Covariates
Group QOL PF QOL RLPH QOL P QOL GH QOL EW QOL RLEH QOL E QOL SF Depression Age Education BMI Diabetes Diagnosis SDMT RVALT
OOL Pearson Correlation .262*
PF Sig. (2-tailed) .043
N 60
QOL RLPH Pearson Correlation .137 .454**
Sig. (2-tailed) .303 .000
N 59 60
QOL Pearson Correlation .068 .511** .398**
Sig. (2-tailed) .609 .000 .002
N 59 60 59
QOL GH Pearson Correlation .109 .391** .255* .308*
Sig. (2-tailed) .405 .002 .049 .017
N 60 61 60 60
QOL EW Pearson Correlation .071 .294* .134 .467** .207
Sig. (2-tailed) .590 .022 .309 .000 .109
N 60 61 60 60 61
QOL RLEH Pearson Correlation .047 JI4* .494** .470** .219 .573**
Sig. (2-tailed) .719 .014 .000 .000 .090 .000
N 60 61 60 60 61 61
QOL Pearson Correlation .213 .537** .421** .580** .317* .556** .509**
Sig. (2-tailed) .102 .000 .001 .000 .013 .000 .000
N 60 61 60 60 61 61 61
QOL SF Pearson Correlation -.041 -.123 -.231 -.257* .044 .012 -.023 -.021
Sig. (2-tailed) .755 348 .079 .049 .739 .926 .861 .876
N 59 60 59 59 60 60 60 60
Pearson Correlation -.276* -.102 -.183 -.429** -.112 -.700** -.451** -.569** -.040
Depression Sig. (2-tailed) .038 .446 .172 .001 .401 .000 .000 .000 .763
N 57 58 57 57 58 58 58 58 58
Pearson Correlation .116 -.211 -.168 -.005 -.174 .123 .111 -.033 .153 -.039
Age Sig. (2-tailed) .330 102 .199 .969 .180 .347 .392 .803 .242 .774
N 73 61 60 60 61 61 61 61 60 58
Pearson Correlation .090 .219 -.073 .277* .129 .153 -.118 .231 .057 -.163 .004
Education Sig. (2-tailed) .427 .090 .581 .032 .320 .240 .365 .073 .667 .222 .972
N 80 61 60 60 61 61 61 61 60 58 74
Pearson Correlation -.111 -.007 .014 -.143 -.006 .158 .173 -.067 .092 -.007 -.110 -.141
BMI Sig. (2-tailed) .367 .962 .917 .297 .967 .246 .201 .626 .503 .961 .368 .249
N 68 56 55 55 56 56 56 56 55 53 69 69
Pearson Correlation -.022 -.010 -.052 .007 .077 .219 .022 -.034 .066 .210 .095 .099 -.124
Diabetes Sig. (2-tailed) .855 942 .704 .962 .574 .105 .873 .801 .633 .130 .439 .418 .326
N 69 56 55 55 56 56 56 56 55 53 69 69 65
Pearson Correlation .188 .002 -.101 .056 -.053 .261 -.026 .053 .027 -.253 -.092 -.015 -.003 .099
Diagnosis Sig. (2-tailed) .138 .991 487 .699 .713 .064 .854 .713 .855 .083 .468 .905 .980 .448
N 64 51 50 50 51 51 51 51 50 48 64 64 59 61
Pearson Correlation .084 .239 .178 .238 .132 .068 .128 .235 .227 -.214 -.205 .363* .125 -.054 .120
SDMT Sig. (2-tailed) .529 .066 .178 .070 .314 .605 .331 .071 .084 .110 .116 .004 .362 .695 .408
N 59 60 59 59 60 60 60 60 59 57 60 60 55 55 50
Pearson Correlation .084 .115 .025 .016 .220 .084 .060 .070 .074 -.064 -.167 .107 -.052 .008 -.228 .112
RVLT Sig. (2-tailed) .522 .379 .849 .904 .089 .521 .645 .590 .575 .631 .198 .411 .702 .954 .108 .393
N 60 61 60 60 61 61 61 61 60 58 61 61 56 56 51 60
Pearson Correlation .107 .170 -.130 .204 .270 .265* .084 .202 .039 -.056 .016 .320* -.005 .147 -.113 .104 .208
COWAT Sig. (2-tailed) .415 .189 .322 .119 .036 .039 .521 .119 .770 .675 .905 .012 .970 .280 .429 .428 .108
N 60 61 60 60 61 61 61 61 60 58 61 61 56 56 51 60 61
QOL PF Physical Functioning, QOL RLPH Role Limitations Due to Physical Health, QOL P Pain, QOL GH General Health, QOL EW Emotional Wellbeing, QOL RLEH Role Limitations due to Emotional Health, QOL E Energy,
QOL SF Social Functioning
Variables collected at baseline: Group Membership, Age, Education, BMI, Diabetes, and Diagnosis. All other variables collected at timepoint 3.
** Correlation significant at the 0.01 level (2-tailed)
* Correlation significant at the 0.0S level (2-tailed)
Pairwise


Hypothesis 1
ADT regimen type will be associated with differences in levels of overall
QOL, controlling for participant characteristics. Participants undergoing
continuous ADT will show lower levels of overall QOL in adjusted analyses.
A between-subjects MANCOVA was performed at timepoint 3 on nine dependent
variables associated with QOL in prostate cancer patients undergoing ADT: Physical
Functioning, Role Limitations Due to Physical Health, Pain, General Health, Emotional
Wellbeing, Role Limitations Due to Emotional Health, Energy, Social Functioning, and
Depression. The CES-D was recoded to be in the same direction as all other QOL
domains for the MANCOVA (i.e., lower scores reflect greater depression) to facilitate
interpretation of the resulting composite dependent variable. Adjustments were made for
eight covariates: age, education, diabetes, time since diagnosis, processing speed, verbal
learning, and verbal fluency. The three areas of cognition were assessed at the same time
as the other dependent variables, whereas all other covariates were assessed at baseline.
The total sample of 60 participants at timepoint 3 was reduced to 38 due to exclusion of
cases with missing values on one of the included dependent variables or covariates. The
structure of the dataset satisfied the assumptions of normality, homogeneity of variance-
covariance matrices, and linearity.
As indicated in table 7, the combined dependent variable was significantly
affected by the independent variable, such that differences in group membership were
associated with differences in the composite QOL outcome variable (F(9) = 4.61 ,p<.01).
Group membership accounts for 67.5% of the variance seen in the composite QOL
dependent variable after adjusting for the influence of covariates (partial r\2 = .675). The
40


composite QOL variable included aspects of physical and mental health that contribute to
wellbeing. The specific effect of group membership on QOL is investigated under
hypothesis 2.
A MANCOVA with the same independent variable, covariates, and baseline
dependent variables was computed in order to evaluate whether group differences in
QOL were present at the outset of the study. The baseline MANCOVA indicated that
there was no effect of group membership on the composite QOL variable adjusting for
the same participant characteristics (F(9) = 0.564, p = 0.816). The obtained partial r\2
indicated that only 13.3% of the variance in QOL was explained by group membership,
after controlling for covariates at baseline.
Table 7. Group Differences in Quality of Life (QOL) After Continuous or Intermittent Androgen Deprivation
Therapy (APT)______________________________________________________________________________________________
Variable
Continuous ADT Intermittent ADT
adjusted Mean adjusted Mean
(Stand. Error)- (Stand. Error)-
F
df p value
Partial ip
Observed
Power
N = 18 N = 20
Quality of Lifeb 4.61 9 .002 .675 .979
Physical Functioning 65.23 (4.13) 86.30 (3.89) 12.08 1 .002 .301 .918
Role Limitations Due To Physical Health 44.11 (8.49) 76.55 (7.99) 6.78 1 .015' .195 .710
Pain 75.00 (5.92) 74.38 (5.57) 0.01 1 .943 .000 .051
General Health 48.18(3.21) 48.39 (3.03) 0.00 1 .966 .000 .05
Emotional Wellbeing 85.79 (2.97) 80.99 (2.80) 1.21 1 .280 .042 .186
Role Limitations Due To Emotional Health 79.54 (6.70) 83.41 (6.31) 0.16 1 .697 .005 .067
Energy 57.72 (4.84) 66.05 (4.56) 1.38 1 .251 .047 .205
Social Functioning 53.04(1.55) 50.39 (1.46) 1.36 1 .254 .046 .203
Depression 10.03(1.91) 10.92 (1.80) 0.10 1 .753 .004 .061
a Adjusting for age, education, BMI, diabetes, time since diagnosis, and cognition
b MANCOVA overall effect of group membership on composite dependent variable
All variables scored in same direction with higher scores indicating better QOL or less depressive
symptoms
** Statistically significant at the 0.01 level
' Statistically significant at the 0.05 level (2-tailed)
41


Hypothesis 2
ADT regimen (continuous versus intermittent) will have differential effects
on specific aspects of QOL, such that greater impact will be seen in physical aspects
of QOL (e.g. Physical Functioning, General Health, Pain, and Role Limitations due
to Physical Health) than on mental health aspects of QOL (e.g. Depression,
Emotional Wellbeing, Role Limitations Due to Emotional Health, Energy, and
Social Functioning).
Group differences among the nine QOL domains are summarized in table 7. The
tests for between-subjects effects within the MANCOVA showed significant group
differences in the SF-36 Physical Functioning and Role Limitations Due to Physical
Health subscales. In both instances, participants in the continuous ADT intervention had
lower scores on these QOL subscales than participants in the intermittent ADT treatment
condition. Lower scores on these SF-36 subscales indicate lower QOL. Group
membership in this study accounted for 30.1% of the variance observed in Physical
Functioning and 19.5% in Role Limitations Due to Physical Health.
The observed power among the other seven subscales ranged from .205 to .050,
which was well below the power observed for the SF-36 Physical Functioning and Role
Limitations Due to Physical Health subscales. Adjusted means for the SF-36 Role
Limitations Due to Emotional Health and Energy subscales showed a trend toward lower
average scores in the continuous ADT group, whereas adjusted means for the Emotional
Wellbeing and Social Functioning subscales were somewhat lower in the intermittent
ADT intervention group. However, group differences in these four subscales did not
reach statistical significance, and partial r\2 were all < 0.047.
42


Hypothesis 3
Levels of biologically available testosterone (BioT) will be positively
correlated with levels of QOL regardless of ADT regimen, while controlling for
multiple observations within participants, such that as testosterone declines, so will
different domains of QOL.
Multilevel correlations were obtained in order to determine the direction and
strength of associations between levels of BioT and various domains of QOL. The
correlations in table 8 include all assessment timepoints combined across treatment
regimens. The participant identification number was defined as the cluster variable to
control for multiple observations among individuals, and therefore accounting for the
nested nature of the dataset.
Table 8. Multilevel Correlations Between Bioavailable Testosterone
Levels (BioT) and Quality of Life (QOL)__________________________
Quality of Life Domain Correlation with BioT (2-tailed p value)
Physical Functioning (SF-36) .280 (.000**)
Role Limitations Due to Physical Health (SF-36) -.053 (.420)
Pain (SF-36) -.136 (.172)
General Health (SF-36) .080 (.172)
Emotional Wellbeing (SF-36) -.024 (.256)
Role Limitations Due to Emotional Health (SF-36) .051 (.411)
Energy (SF-36) .024 (.812)
Social Functioning (SF-36) .013 (.823)
Depression (CES-D) -.183 (.000**)
** Correlation significant at the .001 level (2-tailed)
All variables are in their original metric
Levels of BioT measured in nmol/L
Overall, lower levels of BioT were associated with lower levels of Physical
Functioning (r = .280, p < .001). Additionally, there was a significant association
43


between BioT and depression (r = -.183, p < .001). This negative correlation indicates
that lower levels of BioT were associated with greater depressive symptomatology. No
other correlations approached significance. No uniform direction was observed among
the non-significant correlations.
Scatter plot graphs were visually inspected and an exploratory trend analysis was
conducted to evaluate whether the associations between levels of BioT and QOL were
better explained by non-linear relationships. Figure 3 shows the association between
BioT and Physical Functioning, which was the association that visually showed the
greatest promise for a potential quadratic trend. The loess line, which was fitted to 50%
of the data points, indicates a trend towards uniform low levels of BioT (e.g. below 3.00
nmol/L) when participants also reported low and medium levels of Physical Functioning
(e.g. SF-36 Physical Functioning range 0-70). The slope of the trend line increases at
higher levels of BioT and Physical Functioning.
44


Scatter Plot
FIGURE 3. Scatter Plot For The Association Between Bioavailable Levels of
Testosterone and Physical Functioning
The exploratory trend analyses were conducted in MPlus with data aggregated
across all assessment timepoints and adjusted for multiple observations among
individuals. No significant quadratic trend was found among the association between
BioT and the nine domains of QOL. Analyses of other higher order trends for these
associations were not pursued.
45


CHAPTER IV
DISCUSSION
Summary of Results
The current study proposed and tested the nature of the association between
androgen suppression and QOL by examining group differences among the two ADT
regimens and levels of testosterone. Correlations obtained during preliminary data
analysis indicated an association between various aspects of QOL. Most notably, the SF-
36 Pain subscale was positively correlated with all remaining domains of QOL, except
for the significant negative correlations with the SF-36 Social Functioning subscale and
the CES-D. Energy was significantly correlated with seven out of eight other aspects of
QOL. Correlations were not necessarily stronger within the physical health or mental
health domains of QOL, and many significant associations were seen between the two
overarching constructs of QOL included in this study. Only two significant correlations
emerged between intervention group membership and Physical Functioning as well as
depressive symptomatology, which could indicate that the adjustment for covariates may
be necessary to clarify the nature of group differences. However, very few of the included
covariates were significantly associated with any of the dependent or independent
variables.
Levels of BioT and FT differed by regimen, indicating that participants in the
continuously suppressed group experienced a greater reduction of testosterone levels at
all timepoints. The timing of assessments was based on how much time had passed since
enrollment and the prior assessment timepoint, rather than testosterone level. Given that
individuals receiving intermittent ADT undergo off and on treatment periods,
46


testosterone levels may be higher in this group, as a proportion of participants may be
partially returning to higher levels of androgen hormones. Further analyses assessing the
presence and nature of group differences may therefore not clarify with certainty whether
any observed effect on QOL can be attributed to the type of intervention or particular
levels of BioT.
Hypothesis 1 assessed whether treatment group membership predicted differences
in overall QOL after adjusting for several participant characteristics. The MANCOVA
supported this hypothesis, with group membership accounting for 67.5% of the variance
seen in the composite QOL dependent variable after adjusting for covariates. This large
effect is comparable to the results of the Greenfield et al. study (2010), which
investigated the association between low testosterone levels after non-hormone
suppressing cancer treatment in young male cancer patients, and QOL using the SF-36.
Effect sizes in that study ranged from d = 0.22 to 0.67, with greater effects seen among
physical health aspects of QOL. A similar trend was observed in this study.
The specific nature of differences in QOL among the two intervention groups was
tested under hypothesis 2. The tests for between-subjects effects, using MANCOVA,
adjusted for the same covariates. However, only two of the nine included aspects of QOL
showed statistical significant differences. The Physical Functioning and Role Limitations
Due to Physical Health subscales of the SF-36 were significantly lower among
participants in the continuous ADT group. Those two aspects of QOL were also the only
two subscales that had adequate power to detect a statistically significant effect. The
review of effect sizes in prior studies also showed that a smaller effect could be expected
for the association between androgen suppression and mental health QOL. Therefore, the
47


exact nature of potential group differences in the remaining aspects of QOL is unclear,
given that this study was underpowered to detect those effects. However, adjusted group
means for the SF-36 Role Limitations Due to Emotional Health and Energy subscales
showed a trend towards lower average scores in the continuous ADT group, whereas
average adjusted means for the Emotional Wellbeing and Social Functioning subscales
were somewhat lower in the intermittent group.
Hypothesis 3 investigated the association between androgen suppression and
QOL from the angle of BioT levels, instead of assessing group differences between the
two treatment regimens. Two significant multilevel correlations emerged between levels
of BioT and Physical Functioning, as well as depression, after adjusting for multiple
observations among participants. Lower testosterone levels were therefore associated
with lower levels of physical functioning and higher levels of depression. Other non-
significant correlations were not always in the anticipated direction, but interpretability of
these results is limited due to the small magnitude of the remaining correlations. No
significant higher-order trend associations emerged after the visual inspection of scatter
plot graphs and exploratory trend analyses.
Limitations and Future Directions
The results obtained upon testing hypotheses 1 and 2 indicate that a greater effect
of androgen suppression is seen among patients receiving continuous ADT with regard to
two aspects of physical health QOL. However, the design of this study limits the
interpretation of these results. The addition of a true control group of untreated patients
could help to determine whether the effect of androgen suppression on QOL is only
48


present in the continuous administration group or whether both groups experience lower
levels of QOL, with greater impacts seen in patients receiving continuous ADT.
The specific timing of assessment timepoints in this study may have contributed
to the overall higher levels of BioT in the intermittent ADT intervention group. The
observed group differences seen under tests for hypothesis 1 and 2 may therefore be due
to actual effects of the intervention method, testosterone levels at the time of assessment,
or a combination of both. At this time it is not clear how long it takes for effects on QOL
after the initiation of ADT to become observable. Therefore when testosterone and QOL
levels are measured concurrently, the effect could be interpreted based on the influence of
current testosterone levels or the lingering influence of prior changes in hormone levels.
Future studies on different approaches to ADT might consider how fluctuations in
testosterone affects QOL over time, by incorporating more frequent assessments of
outcome variables.
This study lacked the power to detect many of the anticipated group differences in
specific aspects of QOL. A larger sample size may have helped to provide additional
support for the hypothesis that greater impacts are seen in physical health in comparison
to mental health aspects of QOL. However, a smaller effect seen in measures of
emotional health may have gone unnoticed in this study due the small sample size. Also,
the internal consistency of the SF-36 Social Functioning scale was lower than that of
other subscales, which makes it a less reliable measure of that aspect of QOL. This could
introduce more error variability and make it harder to detect an actual effect. However, it
would be important to further assess this potential association of ADT and testosterone
49


suppression on various aspects of QOL in future studies as even subtle impacts on
emotional health could have consequences for an individuals daily functioning.
An alternative explanation to the observed group difference may lay in the
differential treatment of patients receiving continuous and intermittent ADT. The nature
of the intermittent approach is more individualized, as patients receive treatment
medications after the partial or complete recovery of their testosterone levels. The
intermittent approach may therefore require monitoring of testosterone levels that is more
evident to the patient receiving this type of treatment. Patients receiving continuous ADT,
on the other hand, are likely to receive the same dose of medication at regular intervals,
and patients may not be as aware of any monitoring procedures. The intermittent
approach may therefore be perceived as more personalized, and patients may experience
more attention from health care professionals while they undergo treatment for their
prostate cancer. The perceived additional attention from health providers may even
promote patient-provider communication. Prior research has shown that good patient-
physician communication is associated with improved patient health outcomes, ranging
from patients emotional status and physical health, to blood pressure and blood glucose
levels (Steward, 1995). Future investigators may therefore want to investigate whether
there exist differences in patient-provider relationships associated with the continuous
and intermittent regimens, and whether such differences might influence QOL.
The association between levels of BioT and QOL was assessed using multilevel
correlations. However, only two of the nine QOL domains were significantly correlated
with levels of testosterone aggregated over all study timepoints. These analyses, which
were aggregated across timepoints, included approximately 200 observations. It is
50


therefore less likely that the non-significant multilevel correlations were due to a lack of
statistical power. Further, the estimates of correlations were small and several in the
unanticipated direction. It is possible that a meaningful, non-linear relationship exists
between these variables even though the linear relationships were not present. Within the
multilevel structure, quadratic associations were estimated between BioT and each of the
nine QOL domains. However, none of the potential quadratic associations approached
significance. Trend analyses require large sample sizes in order to detect an effect and the
exploratory analyses for hypothesis 3 may therefore have been underpowered. Future
studies may therefore want to continue investigating whether higher order trends are
more appropriate to describe the association between levels of BioT and QOL.
Furthermore, as seen in Figure 3, the association between BioT and Physical
Functioning became more pronounced at higher levels of BioT. This trend was also
observed for several other aspects of QOL. Due to the nature of this study, only men with
suppressed levels of testosterone were included. Having access to the entire spectrum of
testosterone levels may become necessary to fully understand the nature of the
association between testosterone and QOL for future investigations of this area.
Conclusion
This study found evidence for an effect of androgen suppression on QOL, looking
at group differences between participants treated with continuous and intermittent ADT,
and investigating testosterone levels separately. The results indicate that patients
receiving continuous ADT may experience lower levels of certain aspects of physical
health QOL (i.e. Physical Functioning and Role Limitations Due to Physical Health).
Lower levels of BioT were associated with lower levels of Physical Functioning and
51


greater depression. Reduction in Physical Functioning was consistently implicated in this
study across different hypothesis testing procedures, and may be one of the most readily
perceived side effects, given that outcomes in this study were assessed by self-report.
Those implementing supplemental interventions for patients undergoing ADT may
therefore want to focus on restoring and maintaining overall physical health. Intervening
in one area of QOL may also have a positive influence on other aspects of QOL, as many
of the assessed domains of QOL in this study were highly intercorrelated with each other.
However, whether a causal link exists among these domains of QOL remains to be
investigated.
The association between lower BioT and greater depression was only seen while
testing the effect of testosterone levels alone, and not while assessing treatment group
differences. It could therefore be anticipated that patients receiving either ADT regimen
may experience symptoms of depression at different times during treatment, based on
their current testosterone level. The fluctuating levels of testosterone in patients receiving
intermittent ADT may even contribute to a constantly changing mood profile. Shifts in
mood can influence an individual patients daily functioning and may contribute to
challenges for his family members and caregivers. Awareness of these potential changes
in mood could help to anticipate and prevent the development of depressive disorders in
this population, by providing early interventions for support that may also promote other
ADT treatment outcomes.
Further research is needed to examine the short- and long-term effects of ADT
and changes in testosterone levels on QOL. This study provides support for the existence
52


of such an association, but further research could clarify questions that remain
unanswered.
53


REFERENCES
Akakura, K., Bruchovsky, N., Goldenberg, S. L., Rennie, P. S., Buckley, A. R., &
Sullivan, L. D. (1993). Effects of intermittent androgen suppression on androgen-
dependent tumors. Apoptosis and serum prostate-specific antigen. Cancer 71 (9),
2782 2790.
Alibhai, S., Breunis, H., Timilshina, N., Johnston, C., Tomlinson, G., Tannock, I., et al.
(2010). Impact of Androgen-Deprivation Therapy on Physical Function and
Quality of Life in Men With Nonmetastatic Prostate Cancer. Journal of Clinical
Oncology, 28 (34), 5038 5045.
American Cancer Society. (2014, February 3). Hormone (androgen deprivation) therapy
for prostate cancer. Retrieved February 15, 2014, from
http://www.cancer.org/cancer/prostatecancer/detailedguide/prostate-cancer-
treating-hormone-therapy
American Cancer Society. (2014, March 12). Survival rates for prostate cancer.
Retrieved April 5, 2014, from
http://www.cancer.org/cancer/prostatecancer/detailedguide/prostate-cancer-
survival-rates
Bong, G. W., Clarke, J. FL, Hancock, W. C., & Keane, T. E. (2008). Serum testosterone
recovery after cessation of long-term luteinizing hormone-releasing hormone
agonist in patients with prostate cancer. Urology, 71 (6), 1177-1180.
Bourke, L., Gilbert, S., Hooper, R., Steed, L. A., Joshi, M., Catto, J. W., et al. (2013).
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. European Urology, 65 (5), 865 873.
Carter, H. B. (2011). Management of low (favourable)-risk prostate cancer. BJU
International, 1684-1695.
Cary, K. C., Singla, N., Cowan, J. E., Carroll, P. R., & Cooperberg, M. R. (2013). Impact
of Androgen Deprivation Therapy on Mental and Emotional Well-Being in Men
with Prostate Cancer: Analysis from the CaPSURE (Cancer of the Prostate
Strategic Urologic Research Endeavor) registry. The Journal of Urology, 191 (4),
964-971.
Casey, R. G., Corcoran, N. M., & Goldenberg, L. (2012). Quality of life issues in men
undergoing androgen deprivation therapy: a review. Asian Journal of Andrology ,
226-231.
54


CDC. (2013, October 23). Prostate Cancer Statistics. Retrieved April 5, 2014, from
Center for Disease Control and Prevention:
http ://www. cdc .gov/CAN CER/prostate/ stati sties/
Crook, J., Szumacher, E., Malone, S., Huan, S., & Segal, R. (1999). Intermittent
androgen suppression in the management of prostate cancer. Urology, 53 (3),
530-534.
Dacal, K., Sereika S. M., & Greenspan, S. L. (2006). Quality of life in prostate cancer
patients taking androgen deprivation therapy. Journal of the American Geratrics
Society, 54 (1), 85-90.
Dattalo, P. (2008). Determining Sample Size: Balancing Power, Precision, and
Practicality (Pocket Guides to Social Work Research Methods). Oxford, United
Kingdom: Oxford University Press.
De Ronde, W., Van Der Schouw, Y. T., Polis, H. A., Gooren, L. J., Muller, M., Grobbee,
D. E., et al. (2006). Calculation of Bioavailable and Free Testosterone in Men: A
Comparison of 5 Published Algorithms. Clinical Chemistry, 52 (9), 1777-1784.
De Sousa Magalhaes, S., Fernandes Malloy-Diniz, L., & Cavalheiro Hamdan, A. (2012).
Validity convergent and reliability test-retest of the Rey Auditory Visual Learning
Test. Clinical Neuropsychiatry, 9 (3), 129-137.
Dieperink, K. B., Hansen, S., Wagner, L., Johansen, C., Andersen, K. K., & Hansen, O.
(2012). Living alone, obesity and smoking: Important factors for quality of life
after radiotherapy and androgen deprivation therapy for prostate cancer. Acta
Oncologica, 51 (6), 722 729.
Felce, D., & Perry, J. (1995). Quality of life: Its definition and measurement. Research in
Developmental Disabilities, 16 (1), 51-74.
Greenfield, D. M., Walters, S. J., Coleman, R. E., Hancock, B. W., Snowden, J. A.,
Shalet, S. M., et al. (2010). Quality of Life, Self-Esteem, Fatigue, and Sexual
Function in Young Men After Cancer. Cancer, 116 (6), 1592 1601.
Gregory, C. W., He, B., Johnson, R. T., Ford, O. H., Mohler, J. L., French, F. S., et al.
(2001). A Mechanism for Androgen Receptor-mediated Prostate Cancer
Recurrence after Androgen Deprivation Therapy. Cancer Research 61 (11),
:4315-4319.
Hann, D., Winter, K., & Jacobsen, P. (1999). Measurement of depressive symptoms in
cancer patients: evaluation of the Center for Epidemiological Studies Depression
Scale (CES-D). JPsychosom Res, 46 (5), 437 43.
55


Heidenreich, A., Bastian, P. J., Bellmunt, J., Bolla, M., Joniau, S., Mason, M. D., et al.
(2013). Guidelines on Prostate Cancer. 2013: European Association of Urology.
Herr, H. W., & O'Sullivan, M. (2000). Quality of life of asymptomatic men with
nonmetastatic prostate cancer on androgen deprivation therapy. Journal of
Urology, 136(6), 1743-1746.
Hinton-Bayre, A. D., Geffen, G. M., Geffen, L. B., McFarland, K. A., & Friis, P. (1999).
Concussion in contact sports: reliable change indices of impairment and recovery.
Journal of clinical and experimental neuropsychology, 21 (1), 70-86.
Hojnacki, D., Munschauer, F., Morrow, S., Weinstock-Guttman, B., Drake, A., &
Benedict, R. (2010). Psychometrics and normative data for the Multiple Sclerosis
Functional Composite: replacing the PASAT with the Symbol Digit Modalities
Test. Multiple Sclerosis, 16 (2), 228-237.
Iverson, G. L. (2001). Interpreting change on the WAIS-III/WMS-III in clinical samples.
Archives of Clinical Neuropsychology, 16(2), 183-191.
Jenkinson, C. (1994). Criterion Validity and Reliability of the SF-36 in a Population
Sample. Quality of Life Research ,3(1), 7-12.
Kato, T., Komiya, A., Suzuki, H., Imamoto, T., Ueda, T., & Ichikawa, T. (2007).
Effect of androgen deprivation therapy on quality of life in Japanese men with
prostate cancer. International Journal of Urology, 14 (5), 416-421.
Keogh, J. W., Shepherd, D., Krageloh, C. U., Ryan, C., Masters, J., Shepherd, G., et
al. (2010). Predictors of physical activity and quality of life in New Zealand
prostate cancer survivors undergoing androgen-deprivation therapy. The New
Zealand medical journal, 123 (1325), 20 29.
Kratiras, Z., Konstantinidis, C., & Skriapas, K. (2014). A review of continuous vs
intermittent androgen deprivation therapy: Redefining the gold standard in the
treatment of advanced prostate cancer. Myths, facts and new data on a perpetual
dispute. International braz j urol, 40 (1), 3 15.
Lezak, M. D., Howieson D. B., Loring, D. W., Hannay, H. J., & Fischer, J. S. (2004).
Neuropsychological assessment (4th Edition ed.). New York, NY: Oxford
University Press.
Mastin, L. (2010). Declarative (Explicit) & Procedural (Implicit) Memory. Retrieved
April 21, 2014, from The Human Memory: http://www.human-
memory.net/types_declarative.html
McLeod, S. (2008). Information Processing. Retrieved April 21, 2014, from Simply
Psychology: http://www.simplypsychology.org/information-processing.html
56


Moncada, I. (2006). Testosterone and men's quality of life. The Aging Male 9 (4), 189-
194.
Moore, K. A., Jones, R. M., & Burrows, G. D. (2000). Quality of Life and Cognitive
Function of Liver Transplant Patients: A Prospective Study. Liver
Transplantation 6 (5), 633-642.
Nelson, C. J., Lee, J. S., Gamboa, M. C., & Roth, A. J. (2008). Cognitive Effects of
Hormone Therapy in Men With Prostate Cancer. Cancer, 113 (5), 1097-1106.
Nelson, W. G., Carter, H. B., DeWeese, T. L., Antonarakis, E. S., & Eisenberger, M. A.
(2013). Abeloffs Clinical Oncology (5th edition ed.). Saunders.
Nejat, R. J., Rashid, H. H., Bagiella, E., Katz, A. E., & Benson, M. C. (2000). A
prospective analysis of time to normalization of serum testosterone after
withdrawal of androgen deprivation therapy. Urology, 164 (6), 1891 1894.
Radloff, L. S. (1977). The CES-D Scale: a self-report depression scale for research in the
general population. Applied Psychological Measurement, 1 (3), 385 401.
Rosner, W., Auchus, R. J., Azziz, R., Sluss, P. M., & Raff, H. (2007). Position statement:
Utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society
position statement. The Journal of clinical endocrinology and metabolism 92 (2),
405-413.
Ross, T. P., Calhoun, E., Cox, T., Wenner, C., Kono, W., & Pleasant, M. (2007). The
reliability and validity of qualitative scores for the Controlled Oral Word
Association Test. Archives of clinical neuropsychology, 22 (4), 475-488.
Ruff, R. M., Light, R. H., Parker, S. B., & Levin, H. S. (1996). Benton Controlled Oral
Word Association Test: reliability and updated norms. Archives of clinical
neuropsychology, 11 (4), 329-338.
Saini, A., Berruti, A., Cracco, C., Sguazzotti, E., Porpiglia, F., Russo, L., et al. (2013).
Psychological distress in men with prostate cancer receiving adjuvant androgen-
deprivation therapy 1,. Urologic oncology, 31 (3), 352 358.
Salonen, A. J., Taari, K., Ala-Opas, M., Viitanen, J., Lundstedt, S., & Tammela, T. L.
(2012). Advanced Prostate Cancer Treated with Intermittent or Continuous
Androgen Deprivation in the Randomised FinnProstate Study VII: Quality of Life
and Adverse Effects. European Urology, 63, 111 120.
Schroder, F., Crawford, E. D., Axcrona K., Payne, H., & Keane, T. E. (2012). Androgen
deprivation therapy: past, present and future. BJUInternational, 109, 1-12.
57


Stewart, M. A. (1995). Effective physician-patient communication and health outcomes:
a review. Canadian Medical Association Journal, 152 (9), 1423 1433.
Storey, D. J., McLaren D. B., Atkinson M. A., Butcher, I., Frew L. C., Smyth J. F.,
et al. (2012). Clinically relevant fatigue in men with hormonesensitive prostate
cancer on long-term androgen deprivation therapy. Annals of oncology, 23 (6),
1542- 1549.
Thong, M. S., van de Poll-Franse, L., Hoffman, R. M., Albertsen, P. C., Hamilton, A.
S., Stanford, J. L., et al. (2011). Diabetes mellitus and health-related quality of
life in prostate cancer: 5-year results from the Prostate Cancer Outcomes Study.
BJUInternational, 107 (8), 1223 1231.
Tunn, U. W., Canepa, G., Kochanowsky, A., & Kienle, E. (2012). Testosterone recovery
in the off-treatment time in prostate cancer patients undergoing intermittent
androgen deprivation therapy. Prostate cancer and prostatic diseases, 15 (3),
296-303.
Vermeulen, A., Verdonck, L., & Kaufman, J. M. (1999). A critical evaluation of simple
methods of the estimation of free testosterone in serum. The Journal of Clinical
Endocrinology & Metabolism 84 (10), 36666-3672.
Ware, J. E. (2002). SF-36 Health Survey Update. Spine 25 (24), 3130-3139.
Ware, J. E., & Kosinskia, M. (2001). Interpreting SF-36 Summary Health Measures: A
Response. Quality of Life Research, 10 (5), 405-413.
58


Full Text

PAGE 1

IMPACT OF ANDROGEN DEPRIVATION THERAPY ON QUALITY OF LIFE IN MEN WITH STAGE IV PROSTATE CANCER By SUSANNE WITHROW B.S ., University of Colorado Denver, 2013 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment o f the requirements for the degree of Master of Arts Clinical Health Psychology Program 2016

PAGE 2

2016 SUSANNE WITHROW ALL RIGHTS RESERVED

PAGE 3

ii This thesis for the Master of Arts degree by Susanne Withrow Has been approved for the Clinical Health Psychology Program B y James Grigsby Chair Krista Ranb y Kristin Kilbourn January 8 2016

PAGE 4

iii Withrow, Susanne (M.A., Clinical Health Psychology) Impact of Androgen Deprivation Therapy on Quality of Life i n Men with Stage IV Prostate Cancer Thesis di rected by Professor James Grigsby ABSTRACT Patients with late stage prostate cancer often face a variety of treatment decisions. Interventions like androgen deprivation therapy (ADT) are not only associated with physiological side effects, but are also accompanied by effects on Quality of L ife (QOL ). This experimental study recruited 73 men aged 50 a nd older with stage IV prostate cancer assigned to either continuous or intermittent ADT to evaluate whether these treatment modalities, which reduce biologically available levels of testosterone (BioT) result in differential impacts with regard to QOL. P articipants in this study were followed prospectively and assessed at three timepoint s over the course of 24 to 32 months. A MANCOVA using data from timepoint 3 on nine dependent variables associated with QOL in prostate cancer patients undergoing ADT incl uded: Physical Functioning, Role Limitations Due to Physical Health, Pain, General Health, Emotional Wellbeing, Role Limitations Due to Emotional Health, Energy, Social Functioning, and Depression. Adjustments were made for eight covariates: age, education presence of diabetes, body mass index, time since diagnosis, processing speed, verbal learning, and verbal fluency. Group membership predicted differences in overall QOL ( F (9) = 4.61, p < .01). Specifically, individuals in the continuous ADT treatment group experienced lower levels of Physical Functioning (F (1) = 12.08, p < .01) and more Role Limitations Due to Physical Health (F (1) = 6.78, p < .05) Significant multilevel corr elations bet ween levels of BioT with Physical Functioning and depressive symptoms were found after adjusting

PAGE 5

iv for multiple observations among participants. Other trends for non significant associations and limitations of this study are discussed. Understanding the impa ct of intermittent and continuous ADT on a variety of domains associated with QOL could inform health care decisions in the future and may help to devise strategies for support and supplemental intervention approaches to improve overall treatment outcomes in late stage prostate cancer patients undergoing ADT The form and content of this abstract are approved. I recommend its publication. Approved: Jim Grigsby

PAGE 6

v ACKNOWLEDGEMENTS I would like to acknowledge any and all individuals including Angela Brega, Ph.D. and Marsha Paulich, M.A who collected and recorded data for this project at the urologic oncology clinics of Michael GlodÂŽ, MD, at the University of Colorado Health Sciences Center, and Lawrence Karsh, MD, of Western Urological Associates w hich eventually led to the comprehensive database from which I completed this thesis To m y advisor and thesis committee: Drs. Jim Grigsby Krista Ranby and Kristin Kilbourn y our advice ha s helped me to grow as a researcher and clinical health psychologi st I greatly appreciate your support and guidance throughout this process. The statistical support and consultations I have received from Dr. Krista Ranby and Dr. Sung Joon (Max) Min have greatly helped to shape the nature of this project and my skill set as a researcher I also want to acknowledge my lab oratory colleagues : Anson Kairys and Tim Brunelle This project has greatly benefited from your advice and s upport

PAGE 7

vi TABLE OF CONTENTS CHAPTER I. BACKGROUND I ntroduction and Significance ................................ ................................ .................. 1 Association Between Testosterone Levels and Quality of Life Domains Related to Physical Health ................................ ................................ ................................ ........ 3 A ssociation Between Testosterone Levels and Quality of Life Domains Related to Emotional Health ................................ ................................ ................................ ..... 6 Comparison of Continuous and Intermitte nt ADT with Respect to Primary and Secondary Treatment Outcomes ................................ ................................ .............. 9 Impact of Other ADT Side Effects on QOL ................................ .......................... 10 Objectives ................................ ................................ ................................ .............. 11 Hypotheses ................................ ................................ ................................ ............. 12 II. METHOD P rocedures ................................ ................................ ................................ ........... 13 Sample ................................ ................................ ................................ ................. 16 Instruments ................................ ................................ ................................ .......... 18 Analysis ................................ ................................ ................................ ................ 25 Statistical Power ................................ ................................ ................................ ... 29 Study Design ................................ ................................ ................................ ........ 30 III RESULTS Preliminary Data Analyses ................................ ................................ ................... 32 Hypothesis 1 ................................ ................................ ................................ .......... 40 Hypothesis 2 ................................ ................................ ................................ .......... 42 Hypothesis 3 ................................ ................................ ................................ .......... 43

PAGE 8

vii I V. DISCUSSION Summary of Results ................................ ................................ ............................... 46 Limitations and Future Directions ................................ ................................ ......... 48 Conclusion ................................ ................................ ................................ ............. 51 REFERENCES ................................ ................................ ................................ ............. 54

PAGE 9

viii LIST OF TABLES TABLE 1 Clinical Trial Inclusion and Exclusion Criteria ................................ .................... 14 2 Descriptive Statistics of Independent and Dependent Variables at Baseline ....... 33 3 Descriptive Statistics of Independent and Dependent Variables After ADT ....... 34 4 Descriptive Statistics of BioT ................................ ................................ ............... 35 5 Descriptive Statistics of Covariates ................................ ................................ ...... 36 6 Bivariate Correlations Among Independent Variable, Dependent Variables, and Covariates ................................ ................................ ................................ .............. 39 7 Group Differences in Quality of Life after Continuous or Intermittent Androgen Deprivation Therap y ................................ ................................ ............................. 41 8 Multilevel Co rrelations Between Bioavailable Testosterone Levels and Quality of Life ................................ ................................ ................................ ....... 43

PAGE 10

ix LIST OF FIGURES FIGURE 1 Study Flow Chart and Attrition Over T ime ................................ .......................... 17 2. MANCOVA Testing H ypotheses 1 and 2 ................................ ........................... 27 3 Scatter Plot For The Association Between Bioavailable Levels of Testosterone and Physical Functioning ................................ ................................ .................... 45

PAGE 11

x LIST OF ABBREVIATIONS ADT Androgen Deprivation Therapy ANCOVA Analysis of C ovariance BioT B ioavailable Levels of T estosterone BMI Body Mass Index CES D Center for Epidemiologic Studies Depression Scale COWAT Controlled Oral Word Association Test DV Dependent Variable FT Free Levels of Testosterone IV Independent Variable MANCOVA Multivariate Analysis of C ovariance PASAT Paced Auditory Serial Addition Test QOL Quality of Life RAVLT Rey Auditory Verbal Learning Test SDMT Symbol Digit Modalities Test SF 36 SF 36 Health Survey SHBG Sex Hormone Binding G lobulin Total T Total Levels of Testosterone WAIS III We chsler Adult Intelligence Sc ale Third Edition

PAGE 12

1 CHAPTER I BACKGROUND Introduction and Significance In the United States, prostate cancer is one of the most common types of cancers among men. In 2010 alone, 196,038 men in the United States received a new diagnosis of prostate cancer. On average, 1 in 6 men will receive this diagnosis during their lifetim e (American Cancer Society, 2014). Prostate cancer is also one of the leading causes of cancer death among men (CDC, 2013). Even though early stages of this cancer have very good survival rates due to the slow growth of the associated tumors, the cancer ce lls can invade adjacent as well as distant structures and cause severe health and treatment complications (Nelson, Carter, DeWeese, Antonarakis, & Eisenberger, 2013). The 5 year relative survival rate among men with advanced, stage IV metastatic prostate c ancers in which cancer cells have spread to distant lymph nodes, bones, or other organs is 28% (American Cancer Society, 2014). How ever, life expectancies in late stage prostate cancers can be extended with several types of management strategies. However, before choosing a management or treatment strategy, patients and physicians need to consider a variety of factors. Some of them include the possible side effects of available treatment options, the impact of comorbid conditions on the severity of side eff ects, the risk for complications, limitations of the treatment, the cancer's characteristics, the anticipated progression of the cancer if left untreated, and the patient's preferences and characteristics such as age and genetic vulnerability. In high risk prostate cancers, such as metastasized cancers or those classified as stage IV cancers, routinely recommended treatments include surgical interventions (e.g. radical prostatectomy),

PAGE 13

2 radiation therapy, androgen deprivation therapy (ADT), or a combination o f these three treatments (Carter, 2011). ADT is a type of hormone therapy that is based on the goal of reducing levels of androgens in the body to prevent these hormones from reaching prostate cancer cells. About 90% of prostate cancers are attributed to adenocarcinoma, an abnormal tissue mass or growth (neoplasia) of glandular epithelial tissue. A feature of these neoplastic prostate epithelial cells, as compared to normal secretory cells, is that the neoplastic cells appear to use androgen receptor signa ling not only for differentiation, but also for proliferation. Therefore, most prostate cancer cells depend on androgens for maintenance of growth and survival. Testosterone is the major androgenic hormone encountered in the prostate and is therefore the f ocus of many outcome studies. Under conditions of androgen suppression, cancer cells can stop growth or induce programmed cell death, which can lead to a decrease in tumor size. However, in many cases long term outcomes are unchanged, because prostate canc er tumors tend to become androgen independent. At that progression stage, androgen suppression induced by ADT has little therapeutic effect (Gregory, et al., 2001). ADT interventions can be administered following a continuous or intermittent schedule. Bo th continuous and intermittent ADT suppress the production of testosterone and its associated hormones. However, intermittent ADT includes phases of suppressed and rising testosterone levels instead of prolonged androgen suppression. The complex impact on the body's hormone balance contributes to the range of side effects that are typically seen in patients undergoing ADT. In addition to impacts on Quality of Life (QOL) and depression side effects seen in different types of ADT interventions include

PAGE 14

3 sexual dysfunction, hot flashes, loss of bone and muscle mass, liver toxicity, gynecomastia, breast pain, shrinking of testicles and penis, anemia, weight gain, increased cholesterol levels, as well as changes in cognition (Heidenreich, et al., 20 13; Schršder, C rawford, Axcrona Payne, & Keane, 2012; American Cancer Society, 2014). Recognizing the nature of these possible side effects seen in both continuous and intermittent ADT, each patient and physician needs to decide whether therapeutic benefits and increas ed life expectancies outweigh the risks associated with ADT interventions. One step toward helping patients make informed health care decisions is to understand how intermittent and continuous ADT potentially differ in causing changes in QOL and whether th ese changes could be also in part due to certain patient characteristics. Once the nature of the association between different treatment methods using pharmacological androgen suppression and QOL is better understood, secondary interventions can be devised to potentially reduce the impact of ADT on QOL. This study will examine the effect of ADT treatment type on QOL while controlling for specific participant characteristics that may also be associated with reductions in QOL. The specific association between testosterone levels and a variety of QOL domains, regardless of treatment type, will also be investigated. Association Between Testosterone Levels and Quality of Life Dom ains Related to Physical Health The association between reduced levels of testoster one and worse QOL is frequently encountered in males who produce low levels of testosterone naturally, or through medical conditions and interventions. Moncada (2006) reports that low testosterone levels in hypogonadal men affect seven key areas relevant t o QOL: energy,

PAGE 15

4 emotional functioning, social function, social emotional functioning mental functioning, physical functioning, and sexual functioning. Prostate cancer patients undergoing ADT also experience low levels of testosterone and may therefore show similar impacts in areas of QOL. Most of the literature on the association between ADT and QOL doesn't explicitly state whether the treatment regimen studied was continuous or intermittent. In the following review of the literature, the specific type of ADT regiment will be specified if it was included in the original source. Some of the side effects of ADT include weight gain, altered fat distribution, decreases in muscle mass, and a disruption of sleep patterns (Casey, Corcoran, & Goldenberg, 2012). Th e presence of these symptoms could explain why many studies have reported declines in measures of physical health among prostate cancer patients being treated with ADT. For example, Alibhai et al. (2010) investigated how continuous ADT influences subjectiv e and objective measures of physical health, compared to prostate cancer patients who did not receive ADT and healthy controls. Overall, prostate cancer patients tr eated with ADT declined in the Physical F unction ing Role Limitations Due to Physical Health Pain and Energy domains of the SF 36 Health Survey QOL self report questionnaire. Moreover, continuous ADT seems to be related to reductions in grip strength over time. Walking distance remained stable for prostate cancer patients treated with ADT, whil e this ability improved for prostate cancer without ADT and healthy control participants (Alibhai, et al., 2010). Surprisingly, this study found that most declines on objective and subjective measures of physical health related QOL occurred within 3 month of ADT treatment initiation and remained relatively stable

PAGE 16

5 thereafter. Therefore the time of QOL assessment may be crucial to examine the impact of ADT. However, the impact on physical health seen after treatment with ADT may also be explained by several other variables. Dacal, Seraika, and Greenspan (2006) found that ADT was associated with the physical component scale of the SF 36, which includes the constructs of Physical F unct ioning, General Health, P ain and Role Limitations Due to Physical H ealth. Ho wever, this association was better explained by the presence of comorbid conditions and total testosterone levels. After controlling for age, comorbidity, and total testosterone levels, the association between ADT and physical health impacts was no lon ger significant (Dacal, Sereika & Greenspan, 2006). However, it is not clear which comorbid conditions were implicated in the findings reported by these authors, and whether they could be side effects of the ADT treatment itself. Several studies have also sh own that it is important to consider the impact of comorbid conditions when evaluating disease specific QOL. For example, it has been found that men with prostate cancer who also have prevalent type 2 diabetes mellitus had the poorest self reported general health related QOL, urinary control and sexual function. The association between comorbid diabetes and QOL in this sample was independent of treatment regimen (Thong, et al., 2011). Conditions like type 2 diabetes mellitus are associated with symptoms als o seen after ADT treatment, such as weight gain, fatigue, and changes in urinary function. Therefore it would be important to rule out the influence of comorbid conditions when describing the association between ADT and QOL. Among other health and lifestyl e factors that play a role in the etiology of many comorbid health conditions, obesity (BMI 30) has been found to be associated with worse physical

PAGE 17

6 health and vitality among prostate cancer patients undergoing ADT (Dieperink, Hansen, Wagner, Johansen, An dersen, & Hansen, 2012). Given that weight gain is one of the common side effects of ADT, it is possible that increased weight could influence the association between ADT treatment and lowered physical health. Among a sample of prostate cancer survivors w ho have undergone ADT, those engaging in more physical activity had higher levels of QOL on average than those classified as insufficiently active (Keogh, et al., 2010). Overall, only half of their sample was classified as physically active. Given the repo rted declines in physical health among prostate cancer patients undergoing ADT, several studies have aimed to improve physical functioning through exercise interventions. In the randomized controlled trial by Bourke and colleagues (2013), a 12 week gradual exercise and dietary advice intervention showed improvements in disease specific QOL, fatigue, exercise tolerance, and exercise behavior as compared to a usual care control group. However, no changes were seen in blood pressure, and improvements in diseas e specific QOL were not maintained at the 6 month follow up (Bourke, et al., 2013). Association Between Testosterone Levels and Quality of Life Domains Related to Emotional Health One of the earliest studies on QOL associated with ADT treatment by Herr and O'Sullivan ( 2000) found that not only physical health is affected, but also aspects of mental health. Specifically, men who received ADT through pharmacological means as compared to surgical interventions showed greater cognitive emotional distress fro m intrusive thoughts and images related to cancer stress (Herr & O'Sullivan, 2000). Other studies, such as the one by Cary and colleagues (2013), examined the impact of ADT in

PAGE 18

7 prostate cancer patients on a wider variety of QOL domains associated with emoti onal well being. Their findings were drawn from the national CaPSURE (Cancer of the Prostate Strategic Urologic Research Endeavor) registry and compared prostate cancer patients who received ADT as a primary treatment with participants who received ADT wit h a combination of other treatments. They reported that Role L imitations Due to Emotional H ealth and Energy were consistently affected at different timepoints in individuals who received ADT as a primary treatment. However, no overall group differences wer e seen (Cary, Singla, Cowan, Carroll, & Cooperberg, 2013). Strikingly, clinically significant reductions in Energy scores were observed in 27% of primary ADT patients after 24 months of treatment Reductions in Mental H ealth, Role Limitations Due to Emotional Health, and Social F unctioning were seen in 13%, 20%, and 23% of primary ADT participants, respectively, after completing 24 months of the therapy regimen. Several other studies have taken a closer look at energy and fatigue aspect s of emotional well being Storey and colleagues (2012) found that approximately 43% of men undergoing ADT experience clinically meaningful levels of fatigue. The experience of fatigue was independently associated with the presence of pain and depression. However, neit her age, disease burden, nor treatment duration was associated with the presence of fatigue in this sample (Storey, et al., 2012). This study shows that physical and emotional aspects of QOL may interact to contribute to what some people have termed the an drogen deprivation syndrome. The findings reported by Kato and colleagues (2007) show that an individual's cance r state may be another factor that should be consider ed when examining the impact

PAGE 19

8 of prostate cancer and ADT on QOL domains like energy and fat igue Participants in earlier stages of prostate cancer progression showed a st eady decline in Energy over the course of one ye ar. However, patients with late stage prostate cancer showed low initial levels of Energy and Pain as assessed through the SF 36 QOL questionnaire and slowly recovered over the course of one year (Kato, et al., 2007). These authors also reported g reater impacts on Role Limitations Due to Emotional H ealth and low levels of Emotional Wellbeing However, this effect was only observed among prostate cancer patients with advanced stages of disease progression at the initiation of treatment. Therefore the time course of prostate cancer may be another important factor in explaining the differential impact of ADT on QOL seen in prostate can cer patients. Saini et al. (2013) further considered the impact of ADT on depression and anxiety in addition to evaluating other aspects of QOL in men with prostate cancer. While this study also found the QOL domains of physical, social/family, and functi onal well being affected by ADT, they also reported that depression was greater in the ADT group than in prostate cancer patients who were not receiving ADT (Saini, et al., 2013). There was also a strong correlation between QOL and depression, indicating t hat these two variables may measure similar constructs concerning psychological well being. This study did not find any group differences with regard to anxiety. Even though many studies have reported impacts on aspects of mental health like energy and de pression not all studies have been able to replicate the effect of ADT on other components of emoti onal well being (Dacal, Sereika & Greenspan, 2006).

PAGE 20

9 Comparison of Continuous and Intermittent ADT with Respect to Primary and Secondary Treatment Outcomes Intermittent ADT was first proposed to improve treatment specific symptoms because individuals were thought to have the ability to recover from side eff ects during the off treatment periods. It was also thought that intermittent ADT could slow disease progression and delay the onset of the androgen independent tumor phase, because individuals spend less time in an androgen suppressed state (Crook, Szumach er, Malone, Huan, & Segal, 1999). However, in many instances patients are unable to regain the ability to produce testosterone for several years after the discontinuation of intermittent ADT treatment. In one study by Bong et al. (2008) 53% of participants who had received four or more years of intermittent ADT for prostate cancer retained suppressed testosterone levels for up to 2.5 years after the end of the intervention. Older participants who started ADT after age 70 were significantly less likely to re cover any amount of testosterone production after the completion of therapy (Bong, Clarke, Hancock, & Keane, 2008). The authors of another study of intermittent ADT found that the median time to testosterone recovery was 100 days (14.29 weeks) after the fi rst cycle, and 115 days (16.43 weeks) after the second intermittent ADT cycle (Tunn, Canepa, Kochanowsky, & Kienle, 2012). The authors did not report the specific range or variability seen in recovered testosterone levels for this study. Additionally, othe r studies have reported that intermittent ADT is not consistently superior to continuous ADT with regard to overall and disease free survival in patients with recurr ent cancer emergence (Kratiras, Konstantinidis, & Skriapas, 2014).

PAGE 21

10 Most studies investigat ing the association between ADT and QOL have been carried out using continuous administration of the anti androgenic therapy. Very few studies have reported results of intermittent ADT regimens, and even fewer studies have compared the two methods of admin istering ADT with regard to their specific impacts on QOL. Salonen et al. (2012) investigated whether there are differences with regard to the impact on QOL in men undergoing continuous versus inte rmittent ADT. Specifically, these authors focused on the QO L areas of pain, activity limitation, bed disability, physical capacity, sexual functioning, social functioning, emotional well being, vitality, and overall health. These 10 domains were assessed with a 30 item Health Related QOL self report questionnaire. Their results indicate that intermittent ADT is associated with better outcomes in the domains of activity limitation, physical capacity, and sexual functioning (Salonen, Taari, Ala Opas, Viitanen, Lundstedt, & Tammela, 2012). However, this study did not control for potentially important covariates, including variables such as patients' age, BMI, comorbidity, or other factors that may also have an impact on QOL. Therefore it is not entirely clear whether the dosing schedule of ADT or other variables are im plicated in the mechanism leading to reductions in QOL. Impact of Other ADT Side Effects on QOL In the previous section s several studies were mentioned that found that factors like BMI, disease stage, comorbidities, time since treatment initiation, and age may have an effect on the association between ADT and QOL. Another factor that could play a role in decreasing QOL in prostate cancer patients undergoing ADT is cognition. A review of cognition among men with prostate cancer who are on ADT reports that, "only a few relatively small studies have investigated the impact of this treatment on

PAGE 22

11 cognitive functioning, and some of those studies reported contradictory results" (Nelso n, Lee, Gamboa, & Roth, 2008, p. 1099). Nelson and associates reported that cognitive changes appear to be subtle and affect specific domains. Overall, 47 to 69% of study participants showed a decline in at least one task area. Approximately 14% of ADT pat ients experienced declines in two or more cognitive domains. Visuospatial abilities, working memory, and verbal fluency were frequently affected in ADT patients. Additionally, declines in executive functioning were also seen in patients undergoing ADT trea tment. The review found several contradictory results reported for verbal memory ability (Nelson, Lee, Gamboa, & Roth, 2008). Cognitive impairment also has frequently been associated with poorer QOL (Moore et al., 2000). Therefore, it would be important t o know whether the association between lowered testosterone levels due to ADT and decreases in QOL persists after controlling for impacts of ADT on cognition and other patient characteri stics. No study identified in this review of the literature has examin ed whether continuous and intermittent ADT have differential impacts on QOL, and whether this association is independent of the influence of various covariates on QOL. Moreover, the nature of the association between testosterone levels and QOL in this popu lation, regardless of treatment type, remains unclear. Objectives 1) Assessment of the differential impact of continuous (CON) and intermittent (INT) ADT on QOL among prostate cancer patients. 2) Assessment of covariates influencing the association between ADT regimen type and QOL.

PAGE 23

12 3) Evaluation of relationships between biologically available testosterone levels and QOL at all treatment timepoints. Hypotheses 1. ADT regimen type will be associated with differences in levels of overall QOL, controlling for participant characteristics. a. Participants undergoing continuous ADT will show lower levels of overall QOL as compared to intermittent ADT after adjusting for the following covariates: age, education, BMI diagnosed diabetes mellitus, time since diagnosis and cognit ion. 2. ADT treatment regimen (CON vs. INT) will have a differential impact on specific aspects of QOL. a. Greater impact on specific areas of QOL will be seen in participants receiving continuous ADT as compared to intermittent ADT. 3. Levels of biologically available testosterone (BioT) will be positively correlated with levels of QOL regardless of ADT regimen while controlling for multiple observations within participants such that low levels of testosterone will be associated with lower QOL.

PAGE 24

13 CHAPTER II METHOD Procedures This proposed study, investigating the effect of ADT on QOL in males with prostate cancer, was funded by the Department of Defense ( Effects of androgen blockade on cognitive function and quality of life in men with prostate cancer. De partment of Defense Prostate Cancer Research Program (PCRP) U.S. Army Medical Research and Materiel Command Office of Congressionally Directed Medical Research Programs (CDMRP), PC010257 ) The project was a supplement to an NCI funded project, Phase III randomized study of intermittent versus constant combined androgen deprivation (Bicalutamide and Goserelin) in patients with Stage IV prostate cancer responsive to such therapy, SWOG protocol number 9346. The DoD funded study recruited patients who were pa rticipants in the clinical trial through the urologic oncology clinic of Michael GlodÂŽ, MD, at the University of Colorado Health Sciences Center, and through Lawrence Karsh, MD, of Western Urological Associates. Persons eligible for this study were males a ged 50 and older who m et the inclusion and exclusion criteria for the clinical trial. The Colorado Multiple Institutional Review Board (COMIRB) approved the study The following inclusion and exclusion criteria were set for the clinical trial:

PAGE 25

14 The parent study included patients with and without metastasized prostate cancers. For inclusion, the density of p rostate specific antigen collected through blood samples had to be greater than 5 ng/mL, which can be suggestive of prostate cancer. In order to reduce the effect of confounding variables, participants were excluded if they were undergoing chemotherapy (except for the reduction of pain for bone metastases) or any other type of hormone therapy concurrently with this study, or showing the effects of recent radiation or hormone therapy. Participants were also ineligible if they had a prior bilateral orchiectomy, a surgical intervention that removes both testes to eliminate the primary source of testosterone production. This study also excluded participants wi th

PAGE 26

15 significant comorbid chronic medical illnesses, to reduce the risk of attrition due to increased mortality and time demand. In order to focus on prostate cancer and its associated malignancies, patients were also excluded if they had any other untreated malignancy within the past five years Confounding influences on QOL and cognitive performance were minimized by excluding participants with a history of neurologic disorders, head trauma with loss of consciousness, learning disability, intellectual disab ility, history of alcoholism, or psychosis. During baseline assessment it became evident that several recruited participants met at least one of the exclusion criteria. These participants have been excluded from the data analysis. The primary data collect ion sites were the University of Colorado Health Sciences Center (UCHSC) in Aurora, Colorado, and Western Urological Associates, a private practice in the Denver area. A clinical research associate at UCHSC who was thoroughly familiar with the clinical tri al protocol obtained participant consent. After signing an informed c onsent form, the participants were randomly assigned to continuous or intermittent ADT and were given a copy of the consent form and the original document was filed in appropriate clinic charts. At the time this study was designed, it was widely thought that suppressed testosterone levels would return to baseline between 8 and 26 weeks (Akakura et al., 1993; Nejat et al., 2000), but it was subsequently found that the range was frequently m uch greater than this (e.g., Bong et al., 2008). At the outset of the study, patients on continuous ADT were examined at three timepoints: 1) at baseline (before ADT) or shortly after beginning ADT; 2) after 12 16 weeks of continuous ADT; and 3) after anot her 12 16 weeks of continuous ADT. For those subjects on the intermittent regimen,

PAGE 27

16 the analogous timepoints were: 1) at baseline, if possible, or shortly after beginning ADT ; 2) 12 to 16 weeks following resumption of androgen blockade; and 3) 12 to 16 week s after the next discontinuation of androgen blockade. It was not possible to have access to all participants before they began ADT treatment, as the medication was usually administered in the physician's office, following a decision by patient and physici an, shortly before referral to the study (generally 3 10 days). Given what was known of the pharmacodynamics of ADT, it was assumed that the baseline level of testosterone, and its effects on QOL, would most likely still be observable. However, assessments of QOL and cognition may be slightly less pronounced in participants who do not have a true baseline assessment at timepoint 1. Subjects underwent a set of cognitive tests and questionnaires during each of the three assessment timepoints, requiring appro ximately 90 minutes. Therefore, each subject participated for a total of 4 to 5 hours. To minimize fatigue, there were breaks midway through administration, and more frequently if needed. Data collectors were trained to evaluate fatigue and were asked to r eschedule testing if fatigue appeared to affect performance. The order of test administration was randomized across subjects to prevent unanticipated order effects in the cognitive assessment. To minimize circadian variability in performance, subjects were scheduled for assessments at the time of day that they preferred and these assignments remained as constant as possible for each subject across the three timepoints. Data were double entered. Sample This study recruited 73 m ales, aged 50 to 81, with stage IV prostate cancer. Participants were diagnosed with prostate cancer between 1989 and 2005. Data were

PAGE 28

17 collected between 2003 and 2005. Approximately 97% of individuals from the sample had past chemotherapy treatment; however, patients with current effects of chemotherapy were excluded. Participants were predominately white. One individual identified himself as African American, and four as multiracial. Participants had on average completed 15.3 years of education. Attrition of participants due to death or other reasons occurred at the second and third timepoint s as depicted in figure 1 F IGURE 1. Study Flow Chart and Attrition o ver T ime

PAGE 29

18 Instruments Quality of Life Quality of life (QOL) describes multiple domains of individual (or sometimes group) well being. Such domains can be assessed subjectively or objectively and often include psychological, social, and physical aspects of functioning (Felce & Perry, 1995) SF 36 Health Survey QOL was assessed using the SF 36 Heal th Survey a 36 item general health status instrument On this measure, the impact of health on normal activities is analyzed using eight subscales : Physical F unctioning, P ain, General H ealth Role Limitations Due to Physical Health, Role L im itations Due t o Emotional H ealth, Energy, Emotional Well being, and Social F unctioning Each item contributes to only one subscale, on which scores range from 0 to 100, with higher scores indicating better QOL. The information obtained from the SF 36 can be broken down i nto three levels: (1) items; (2) sub scales; and, (3) summary measures which combine subscales The t wo summary measures can be used to assess two broader aspects of QOL. The Physical Health summary measure combines the Physical F unc tioning, Role Limitations Due to Physical Health, Pain, and General H ealth subscales, and the Mental Health measure combines the Role Limitations Due to Emotional Health, Emotional Wellbeing, Social F unctioning, and Energy subscales (Ware, 200 2) Certain subscales, such as Emotional W ell being indicating mood provide unique information needed to examine the effect of ADT. Therefore analyses will incorporate the eight main scales to obtain a detailed analysis of the association between ADT r egimen and testosterone on QOL.

PAGE 30

19 Internal consistency for most subscale s is above 0.8, except for the Social F unctioning subscale (0.76), which might in part be due to the small number of items comprising that subscale. The average Cronbach alpha values f or all eight subscales in people with poor health status was 0.82 (Jenkinson, 1994) The reliability of the Physical H ealth summary score is 0 .92 and for the Mental H ealth summary score it is 0.88 (Wa re, 2002) Support for the predictive validity of the two summary measures has also been reported. For example, individual s with very high scores on the Mental H ealth summary measure (e.g., 60 74) were approximately one third less likely than those in dividual with lower summary scores (55 59 and 50 54) to receive a diagnosis of depression and were about one third more likely to self report higher le vels of life satisfaction. The Physical H ealth summary measure has also been found to predict risk of mo rtality. For example, a 40% increase in m ortality rates is seen in individuals with scores in the 8 24 range as compared with test ta kers in the 25 34 range on the Physical H ealth summary measure (Ware & Kosinskia, 2001) Depression Depression and its associated emotions and cognitions can affect an individual's perception of QOL, and is therefore a closely related construct. A measure of depression will therefore be included in this study to supplement the assessment of me ntal health aspects of QOL. Center for Epidemiologic Studies Depression Scale (CES D) Depression was evaluated using the Center for Epidemiologic Studies Depression Scale (CES D, Radloff, 1977). This 20 item self report questionnaire contains fewer it ems that might reflect physical disability than do other depression scales. The CES D

PAGE 31

20 has been found to be a reliable and valid measure of depressive symptomatology in cancer populations. Among a group of breast cancer patients, the internal consistency of the CES D showed alpha coefficients above 0.85 It also correlated in expected directions with related concepts such as fatigue, anxiety, and global mental health functioning indicating construct validity for the measure (Hann, Wi nter, & Jacobsen, 1999) Scores for the CES D range from 0 to 60, with higher scores indicating the presence of more depressive symptoms. Speed and Capacity of Information Processing Information processing involves the integration and transformation of information from the environment with the help of several cognitive systems, such as attention, working memory and long term declarative memory (McLeod, 2008) Symbol Digit Modalities Test (SDMT) This test requires participants to pair abstract symbols with specific numbers as quickly as possible for 90 seconds. In addition to measuring the construct of speed of information processing, th e SDMT requires attention and integration abilities. A total score which will be used for the analyses of this study, is calculated by adding the items (0 110) that were correctly coded in 90 seconds and subtracting the number of incorrectly coded items from that sum. The construct validity of the SDMT is accepta ble to good. One study investigating cognitive impairment after athletic concussions found that the SDMT correlate d 0.62 to 0.91 with the Wechsle r Digit Symbol subtest (Hinton Bayre, Geffen, Geffen, McFarland, & Friis, 1999) I n patients with multiple sclerosis, the SDMT correlated with the Paced

PAGE 32

21 Auditory Serial Addition Test (PASAT) (r= 0 .54), which also assesses the capacity and rate of information processing The same study observed a test retest reliability of 0.74 (Hojnacki, Munschauer, Morrow, Weinstock Guttman, Drake, & Benedict, 2010) Individuals with multiple sclerosis often experience a wide range of symptoms that could include physical, mental, and potential psychiatric problems These symp tom combinations are somewhat different from what can be seen in prostate cancer patients. The reported psychometric properties will therefore also be interpreted with caution. Declarative Verbal Learning and Memory Declarative memory refers to informati on and events stored and consciously retrieved from long term memory systems (Mastin, 2010) Rey Auditory Verbal Learning Test (RAVLT) This word learning test consists of two lists, each with 15 unrelated concrete nouns The first list is repeated over five trials with r ecall requested after each presentation of the words. After the fifth trial, the second list of words is read to the participant, with immediate recall of that list (the interference trial) After a 20 minute delay, recall is tested for the first list without any further presentation of either list. This study will analyze the delayed recall score, which is the total number of words from the first list that are recalled correctly after a 20 minute delay. De S ousa Magalh‹es et al. (2012) examined test retest correlations for male and female college students aged 17 through 40. The strongest test retest correlations were found for the sum of scores obtained over the first five learning trials (0.96). The test re test correlation for words recalled after 20 minutes was 0.76. This score will be used in

PAGE 33

22 my study as an indicator of delayed verbal memory. The Cronbach alpha reliability of the RAVLT was 0.84, which indicates good internal consistency among the items of this test. The RAVLT demonstrated divergent validity with the Trail Making Test, which evaluates attention and executive function rather than verbal learning and memory. Evidence for convergent validity is seen in the correlation between the RAVLT and the Benton Visual Retention Test, which evaluates immediate and delayed visual memory (De Sousa Magalh‹es, Fernandes Malloy Diniz, & Cavalheiro Hamdan, 2012) Verbal Fluency Verbal fluency encompasses the retrieval of verbal infor mation from long term memory systems. Recall of this type of information requires executive control over several cognitive processes such as selective at tention, inhibition, and self monitoring (Lezak, Howieson Loring, Hannay, & F ischer, 2004) Controlled Oral Word Association Test (COWAT) This measure of verbal fluency asks test takers to name as many words as possible that begin with the letter F in 60 seconds. The procedure is then repeated twice, but for words that begin with the letters A and S The number of unique words is summed across the three trials. My analyses will use an age adjusted scaled score ( range: 2 18), which is obtained by comparing the raw score to data obtained from a normative sample table. The test retest reliability for this test of verbal fluency is 0.74. On average, test takers gain 3 words upon being retested, which could indicate a small practice effect. Education can significantly affect COWAT scores, and may account for up to 8% of the total observed variance (Ruff, Light, Parker, & Levin, 1996) The total number of novel

PAGE 34

23 words generated by test takers in the COWAT correlates more strongly with scores on the vocabulary subtest of the Wechsler Adult Intelligence Scale Third Edition (WAIS III) (0.22) than with performance on the Tower of Hanoi (0.03), a measure of procedura l learning and executive ability (Ross, Calhoun, Cox, Wenner, Kono, & Pleasant, 2007) All of these estimates were obtained in samples of healthy adults and did not consider the impact of aging or health impairment. This study' s sample has completed an above average number of years of education, which could increase overall testing performance on this and other tests in this study. The absolute scores found in this study and psychometric properties reported in the literature wil l therefore be interpreted with caution. Hormone Assays Hormone levels were measured from 3 ml b lood samples drawn from participants who had fasted for at least 1 2 hours before the blood draw. Because the free hormone level is the physiologically active component, total levels of testosterone sex hormone binding globulin (SHBG), and albumin were also obtained and from these, free sex hormone levels were calculated On average, 50% of circulating testosterone is nonspecifically or weakly bound to album in, and 44% is specifically bound to SHBG. Another 3.5% of testosterone contained in plasma is associated with cortisol binding globulin. This leaves 2 3% of total testosterone levels free or in an unbound state (De Ronde, et al., 2006) Unbound, or free t estosterone (FT) is often regarded as the portion accessible for androgenic effects. Another estimate of bioavailable testosterone levels is BioT, which is the total amount of FT and testosterone that is weakly bound to the protein albumin. Even though dir ect

PAGE 35

24 measures of FT and BioT are often preferred due to their accuracy, the position statement of The Endocrine Society recommends the calculation of FT and BioT as a useful and cost effective alternative to indicate available testosterone levels (Rosner, A uchus, Azziz, Sluss, & Raff, 2007) However, the calculation of FT and BioT is done using different approaches and assumptions, and can therefore yield significantly different results when used as an independent variable For example, the Vermeulen and S ondergard approach is based on the law of mass action. These algorithms use values for total testosterone, SHBG, albumin, as well as association constants for testosterone binding to proteins such as albumin and SHBG to generate FT values. Other approaches such as those developed by Morris, Ly, or Emadi Konjin, base their algorithms on numerous direct laboratory measurements of FT. These results were used to create an algorithm to predict future FT (De Ronde, et al., 2006) However, these empirical equati ons may lack accuracy and comparability among different labs. Algorithms based on the law of mass action are highly dependent on accurate measurement of total testosterone and SHBG (Rosner, Auchus, Azziz, Sluss, & Raff, 2007) In this study, BioT was calc ulated according to the Vermeulen (Vermeulen, Verdonck, & Kaufman, 1999) method. This approach was used because the algorithm is less likely to over or underestimate BioT concentrations in older men, who tend to have higher SHBG concentrations. It has als o been suggested that the inclusion of measured albumin concentrations, rather than the use of standard reference values of this protein, becomes important at lower BioT concentrations, which are expected in this study (De Ronde, et al., 2006)

PAGE 36

25 The Endocr ine Society has proposed the following guidelines for testosterone: total testosterone (Total T) levels greater than 320 ng/dl (11.1 nmol/L) are considered normal. Total T values less than 200 ng/dl (6.9 nmol/L) are indicative of hypogonadism, but Total T results between 200 320 ng/dl (6.9 11.1 nmol/L) are considered borderline. A free testosterone (FT) value of 6.5 ng/dl (0.23 nmol/L) is considered the lower limit of the normal range (Rosner, Auchus, Azziz, Sluss, & Raff, 2007) These reference val u es can be used to determine whether ADT interventions have been effective at reducing Total T and FT levels. Analysis Preliminary Data Analysis. F requency distributions and descriptive statistics w er e obtained to evaluate data integrity and to characterize the intervention groups. Simple correlations between group membership, measures of BioT, QOL scales and depression as well as cognition were calculated to assess patterns of associations among these variables. Hypothesis Testing Hypothesis 1 ADT regimen type will be associated with differences in levels of overall QOL, controlling for participant characteristics. P articipants undergoing continuous ADT will show lower levels of overall QOL in adjusted analyses Using multivariate analysis of covariance (MANCOVA), this first hypothesis examine d the main effect of treatment group membership (independent variable IV) on a composite QOL variable (dependent variable DV). The newly created composite DV within th e MANCOVA framework represents a linear combination of all included DVs,

PAGE 37

26 and acts to maximize group differences. The test for main effects assesses whether mean differences in the composite DV among groups at different levels of the IV are due to chance wh ile controlling for the influence of covariates. The DV will consist of the eight SF 36 QOL subscales and CES D depression questionnaire. The following covariates we re included due to their potential influence on QOL: age, education, BMI, diagnosed diabete s mellitus (DM) time since diagnosis and cognition. Covariates assessed at baseline include d age, years of education, diagnosis of diabetes mellitus, BMI and time since cancer diagnosis Measures of cognition (i.e. processing speed, verbal learning, and verbal fluency) at timepoint 3 we re included to capture the impact of androgen suppression on cognitive functioning In order to assess the influence of initial levels of QOL, an additional MANCOVA will be computed for the same outcome variables obtained during baseline assessments. A non significant overall effect would indicate that the treatment groups did not differ with regard to their QOL while controlling for covariates at the onset of the study.

PAGE 38

27 FIGURE 2. MANCOVA Testing H ypotheses 1 and 2 Hypothesis 2 ADT regimen (continuous versus intermittent) will have differential effects on specific aspects of QOL, such that greater impact will be seen in participants receiving continuous ADT as compared to intermittent ADT. Analysis of covaria nce (ANCOVA) is used to assess the relative contribution of the included dependent variables while controlling for the covariates described above. These tests of between subjects effects assess whether ADT treatment group membership is associated with diff erences in each of the tested domains of QOL while controlling for the influence of ot her participant characteristics These analyses are computed concurrently with the MANCOVA main effect in the IBM SPSS statistical software package. MANCOVA is a statist ical technique used to assess the impact of group differences on several DVs while controlling for the influence of covariate variables. This technique avoids inflation in Type I error rate because the number of comparisons to be analyzed is reduced. This technique may also show differences among variables that may

PAGE 39

28 not become apparent when using several ANOVA tests for the same research question. The MANCOVA technique is based on several assumptions that need to be met to ensure robustness of the statistica l test. First, the DVs and all linear combinations of them should be normally distributed. Univariate and multivariate outliers can cause distributions to be skewed and may need to be removed. Variables with outliers or non normally distributed variables m ay also need to be transformed in order to meet the assumption for normality. Second, it is assumed that all associations between pairs of DVs, all pairs of covariates, and all DV covariate pairs are linear. Non linear associations between variables reduce the power of the test. Lastly, all DVs should exhibit equal levels of variance across the different levels of the IV, covariates, and with other DVs. Bivariate Correlation: Hypothesis 3 Levels of biologically available testosterone (BioT) will be positi vely correlated with levels of QOL regardless of ADT regimen while controlling for multiple observations within participants, such that as testosterone declines, so will different domains of QOL. Multilevel correlations w er e obtained to determine the direction and strength of the association between levels of testosterone and different domains o f QOL regardless of ADT regimen while controlling for multiple observations within people. The eight scales of the SF 36 and CES D questionnaire w er e included to ass ess physical and mental health aspects of QOL.

PAGE 40

2 9 This study use d IBM SPSS version 21 to compute the MANCOVAs testing hypotheses 1 and 2 Hypothesis 3 will be computed using MPlus version 3.7. An alpha level of 0.05 will be used to determine statistica l significance. Statistical Power This section review s the effect sizes of prior findings in order to determine whether this study has adequate power. Effect sizes (Cohen's d and f 2 ) for the association between testosterone levels after non hormone depen dent cancer treatment and QOL using the SF 36 subscales have been shown to range from d = 0.22 to 0.67 depending on the subscale being analyzed (Greenfield, et al., 2010) Based on these effect sizes, I anticipate d a medium e ffect (d= 0.40, f 2 = 0.15, or = 0.3) MANCOVA w as used in order to test hypotheses 1 and 2. The necessary sample given the expected effect size can be ob tained using gPower (version 7.3 ) which does not directly estimate sample size for MANCOVA However, according to Dattalo (2008) the estimation strategies for MAN C OVA can be adapted in order to obtain such estimates. The sample size needed to detect a medium effect (f 2 = 0.15) with a power of 0.8 at = 0.05 using MANCOVA, wi th 2 groups, 8 co variates, and 9 response variables is 4 4. There are currently no agreed upon guideline s for determining power for multilevel correlation analyses. Therefore, statistical power for hypothesi s 3 will be determined using guidelines for bivari ate correlations, which will yield a more conservative estimate as this statistical technique does not adjust for shared variance across multiple observations. A sample size of 67 is needed in order to detect a medium effect ( = 0.3) with power of 0.8 and = 0.05 for the correlations tested in hypothesis 3. Data were collected from 73, 70, and 60 participants at timepoints 1, 2, and 3, respectively. Therefore, this study should have adequate power

PAGE 41

30 to detect significant effec ts for most hypothesis tests. Effect sizes will be reported in addition to the results from significance testing. Study Design This study represents a experimental research design in which two groups of patients received either continuous or intermittent ADT. The sample was followed prospectively over three assessment timepoints. Outcomes at e ach timepoint were evaluated at approximately 12 to 16 week intervals. This study consider s the last QOL assessment timepoint as the outcome measure, and control s fo r certain participant characteristics that may also be associated with QOL in examining hypothesis 1 and 2. All timepoints w er e considered in testing hypothesis 3. Participants were selected based on their exposure to the antecedent conditions of prostate cancer and eligibility for treatment with ADT. The parent clinical trial randomly assigned participants to the continuous or intermittent treatment administration groups. This design is capable of establishing a temporal sequence for ADT and decline in Q OL. Because no control group has been included in the design, this study cannot determine how prostate cancer alone affects QOL. Additionally, the study group s were recruited from another clinical trial involving an urban university hospital setting and a large urban urology priva te practice. Therefore this project's sample may not represent the entire population of stage IV prostate cancer patients undergoing ADT. T he analyses for this study include covariates that potentially differ among group s and could have a separate influence on QOL. Assessing group differences for baseline levels of QOL could provide further evidence for the association between different ADT r egimens and QOL that is independent of initial levels of OQL

PAGE 42

31 The prospective design involved some attrition of participants due to death or other reasons ( n = 13). Participants who only completed the first two assessments w er e not included in the analyse s for hypothesis 1 and 2, but had scores that could be analyzed to test hypothesis 3.

PAGE 43

32 CHAPTER III RESULTS Preliminary Data Analyses Descriptive statistics and bivariate correlations were obtained in order to characterize the structure of the dataset and identify patterns of associations among the variables. Table 2 summarizes the distribution of the dependent variables, including the e ight QOL subscales of the SF 36 and C ES D depression questionnaire at the time of baseline assessment The SF 36 Physical Functioning, Role Limitations Due to Physical Health, Pain, and General Health subscales belong to the QOL construct of physical heal th. The SF 36 Emotional Wellbeing, Role Limitations Due to Emotional Health, Energy, and Social Functioning subscales as well as the CES D depression scale are thought to belong to the mental health construct of QOL. Higher scores on the SF 36 subscales in dicate higher levels of QOL whereas higher scores on the CES D indicate the presence of more depressive symptoms. All variables are reported in their original metric unless otherwise specified. The values of skewness and kurtosis indicate that all dependen t variables assessed at baseline we re adequately normally distributed and were therefore suited for further analyses within the general linear model. No univariate outliers were detected during the initial data screening.

PAGE 44

33 Table 3 summarizes the descriptive statistics obtained from the nine dependent variables at the third assessment timepoint which was used in this study to assess treatment outcomes T he values for skewness and kurtosis indicate that al l dependent variables we re also normally di stributed and therefore suited for further analyses within the general linear model. No univariate outliers were detected during the initial data

PAGE 45

34 screening. Am ong the SF 36 subscales, scores on the Social Functioning domain showed the smallest range and co uld limit the amount of variability that could be analyzed during hypotheses testing.

PAGE 46

35 The majority of mean scores on most SF 36 subscales we re lower and the CES D mean score wa s higher in the continuous ADT treatment group at baseline as well as at timepoint 3. This could indicate lower levels of QOL in the continuous AD T group. Statistical tests assessing hypotheses 1 and 2 will be able to confirm the existence and clarify the nature of potential group differences. The distribution s of BioT levels at all assessment timepoint s are summarized in table 4. Scores on this variable among all participants appear to be normally distributed. However, when the two intervention groups are considered separately, the distribution of BioT levels among participant s of the continuous ADT group at all assessment timepoint s may be too peaked (i.e. leptokurtic) and somewhat positively skewed. Participants in the continuous ADT intervention group had overall lower mean levels of BioT and a smaller range at timepoint 2 a nd 3 overall as a group. The intermittent group by itself has a greater range of scores and is normally distributed. Levels of BioT include levels of free testosterone (FT) and the portion of testosterone bound to the protein albumin. Average levels of FT among participants in the continuous ADT group of this study ranged from 0.03 to 0.06 nmol/L and between 0 .29

PAGE 47

36 and 0 .40 nmol/L in the intermittent group (data not included in table 4). Given that FT values below 0.23 nmol/L are considered outside of the normal ra nge, participants in the continuous group were suppressed well below normal levels of FT whereas participants testosterone levels in the intermittent intervention group were suppressed to levels within the lower boundaries of the normal range. Descriptive statistics of the covariates included in statistical tests assessing hypotheses 1 and 2 are summarized in table 5. The mean age of participants in this study was 72.16 and they ha d completed an average of 15.27 years of education. The mean BM I score of 29.08 is situated at the upper boundary of the overweight range The majority of participants did not report having a diagnosis of type 2 diabetes mellitus. The diabetes variable being dichotomous, is the only covariate that is not normally dis tributed.

PAGE 48

37 Table 6 includes all pairwise correlations between the independent variable, depend ent variables, and covariates. All QOL dependent variables as well as the three aspects of cognition (covariates) were assessed at timepoint 3. All remaining c ovariates were assessed at baseline. Having received continuous ADT was a ssociated with lower levels of Physical Functioning ( r = .454, p < .01) as well as greater depressive symptomatology (r = .267, p < .05) Other n otable significant correlations inclu de the associations between the Physical Functioning subscale and Role Limitations Due to Physical Health ( r = .454, p < .01 ), Pain ( r = .511 p < .01), General Health (r = .391, p < .01 ), Emotional Wellbeing (r = 294, p < .05), Role Limitations Due to Emotional Health (r = 314, p < .05) and Ener gy subscales (r = .537 p < .01). The Role Limitations due to Physical Health subscale was significantly correlated with the Pain (r = .398, p < .01 ), General Health (r = .255, p < .05), Rol e Limitations D ue to Emotional Health (r = .494 p < .01) and Energy subscales (r = .421 p < .01). The Pain subscale of the SF 36 QOL questionnaire was additionally correlated w ith the General Health (r = .308 p < .05 ), Emotional Wellbeing (r = .467 p < .01), Role Limitations D ue to Emotional Health (r = .470, p < .01), Energy (r = .580 p < .01), Soci al Functioning (r = .257 p < .05), and CES D depression scales (r = .429 p < .01). The Pain subscale was also positively correlated with the Education variable (r = .277, p < .05). The General Health subscale was associated with the Energy subscale (r = .317, p < .05). The Emotional Wellbeing subscale was also associated with the Role Limitations d ue to Emotional Health (r = .573, p < .01), Energy (r = 556 p < .01), and CES D depres sion scales (r = .700 p < .01). The Emotional Wellbeing subscale was additionally associated with performance on the COWAT (r = .265, p < .05). The Role Limitations Due to Emotional Health subscale was

PAGE 49

38 corr elated with the E nergy (r = .509 p < .01) and CE S D depression scales (r = .451 p < .01). Lastly, the Energy subscale was significantly associated with the CES D depression scale (r = .621, p < .01). The Energy subscale was associated with the CES D depression scale (r = .569, p < .01). For the majority of SF 36 QOL domains, higher levels on one scale are correlated with higher scores on the other, except for the association between Social Functioning and Pain subscales on the SF 36. All associations between SF 36 subs cales and CES D scale indicate that higher levels of QOL are associated with fewer depressive symptoms. Additionally, higher levels of education are positively correlated with performance on tes ts of processing speed (r = .363 p < 05) and verbal fluency (r = .320 p < .05).

PAGE 50

39

PAGE 51

40 Hypothesis 1 ADT regimen type will be associated with differences in levels of overall QOL, controlling for participant characteristics. P articipants undergoing continuous ADT will show lower levels of overall QOL in adjusted analyses. A between subjects MANCOVA was performed at timepoint 3 on nine dependent variables associated with QOL in prostate cancer patients undergoing ADT: Physical Functioning, Role Limitations Due to Physical Health, Pain, General Health, Emotional Wellbeing, Role Limitations Due to Emotional Health, Energy, Social Functioning, and Depression. The CES D was recoded to be in the same direction as all other QOL domai ns for the MANCOVA (i.e., lower scores reflect greater depression) to facilitate interpretation of the resulting composite dependent variable. Adjustments were made for eight covariates: age, education, diabetes, time since diagnosis, processing speed, ver bal learning, and verbal fluency. The three areas of cognition were assessed at the same time as the other dependent variables whereas all other covariates were assessed at baseline. The total sample of 60 participants at timepoint 3 was reduced to 38 due to exclusion of cases with missing values on one of the included dependent variables or covariates. The structure of the dataset satisfied the assumptions of normality, homogeneity of variance covariance matrices, and linearity As indicated in table 7, t he combined dependent variable was significantly a ffected by the independent variable such that differences in group membership were associated with differences in the composite QOL outcome variable (F (9) = 4.61, p < .01). Group membership accounts for 67.5% of the variance seen in the composite QOL dependent variable after adjusting for the influence of covariates (partial # 2 = .675). The

PAGE 52

41 composite QOL variable included aspects of physical and mental health that contribute to wellbeing. The specific effect of group membership on QOL is investigated under hypothesis 2. A MANCOVA with the same independent variable, covariates, and baseline dependent variables was computed in order to evaluate whet her group differences in QOL were present at the o ut set of the study. The baseline MANCOVA indicated that t here was no effect of group membership on the composite QOL variable adjusting for the sam e participant characteristics (F (9) = 0.564, p = 0.816). The obtained partial # 2 indicated that only 13.3% of the variance i n QOL was explained by group membership, after controlling for covariates at baseline.

PAGE 53

42 Hypothesis 2 ADT regimen (continuous versus intermittent) will have differential effects on specific aspects of QOL, such that greater impact will be seen in physical aspects of QOL (e.g. P hysical Functioning, General Health, Pain, and Role Limitations due to Physical H ealth) than on ment al health aspects of QOL (e.g. Depression, Emotional Wellbeing, Role Limitations Due to Emotional Health, Energy, and Social F unctioning) Group differences among the nine QOL domains are summarized in table 7. The tests for between subjects effects within th e MANCOVA showed significant group differences in the SF 36 Physical Functioning and Role Limitations Due to Physical Health subscales. In both instances, participants in the continuous ADT intervention had lower scores o n these QOL subscales than particip ants in the intermittent ADT treatment condition. Lower scores on these SF 36 subscales indicate lower QOL. Group membership in this study accounted for 30.1% of the variance observed in Physical Functioning and 19.5% in Role Limitations Due to Physical He alth. The observed power amon g the other seven subscales ranged from .205 to .050, which was well below the power observed for the SF 36 Physical Functioning and Role Limitations Due to Physical Health subscales. Adjusted means for the SF 36 Role Limitatio ns Due to Emotional Health and Energy subscales showed a trend toward lower average scores in the continuous ADT group whereas adjusted means for the Emotional Wellbeing and Social Functioning subscales were somewhat lower in the intermittent ADT interven tion group. However, group differences in these four subscales did not reach statistical significance and partial # 2 were all 0 .047.

PAGE 54

43 Hypothesis 3 Levels of biologically available testosterone (BioT) will be positively correlated with levels of QOL reg ardless of ADT regimen while controlling for multiple observations within participants, such that as testosterone declines, so will different domains of QOL. Multilevel correlations were obtained in order to determine the direction and strength of association s between levels of Bi oT and various domains of QOL. The c orrelations in table 8 include all assessment timepoint s combined across treatment re gimen s The participant identification number was defined as the cluster variable to control for multi ple o bservations among individuals and therefore account ing for the nested nature of the dataset. Overall, lower levels of BioT we re associated with lower levels of Physical Functioning (r = .280, p < .001). Additionally, there wa s a significant associ ation

PAGE 55

44 between BioT and depression (r = .183, p < .001). This negative correlation indicates that lower levels of BioT we re associated with greater depressive symptom atology No other correlations approach ed significance. No uniform direction was observed among the non significant correlations. Scatter plot graphs were visually inspected and an exploratory trend analysis was conducted to evaluate whether the association s be tween levels of BioT and QOL were better explained by non linear relationships. Figu re 3 shows the association between BioT and Physical Functioning, which was the association that visually showed the greatest promise for a potential quadratic trend. The loess line, which was fit ted to 50% of the data points, indicates a trend towards uniform low levels of BioT (e.g. below 3.00 nmol/L) when participants also reported low and medium levels of Physical Functioning (e.g. SF 36 Physical Functioning range 0 70). The slope of the trend line increases at higher levels of BioT and Physical Func tioning.

PAGE 56

45 FIGURE 3 Scatter Plot For The Association Between Bioavailable Levels of Testosterone and Physical Functioning The exploratory trend analyses were conducted in MPlus with data aggregated across all assessment timepoint s and adjusted for mu ltiple observations among individuals. No significant quadratic trend was found among the association between BioT and the nine domains of QOL. Analyses of other higher order trends for these associations were not pursued.

PAGE 57

46 CHAPTER IV DISCUSSION Summary of Results The current study proposed and tested the nature of the association between androgen suppression and QOL by examining group differences among the two ADT regimen s and levels of testosterone. C orrelations obtained during preliminary data analys is indicated an association between various aspects of QOL. Most notably, t he SF 36 Pain subscale was positively correlated with all remaining domains of QOL except for the significant negative correlation s with the SF 36 Social Functioning subscale and t he CES D. Energy was s ignificantly correlated with seven out of eight other aspects of QOL. Correlations were not necessarily stronger within the physical health or mental health domains of QOL, and many significant associations we re seen between the two o verarching constructs of QOL included in this study. Only two significant correlation s emerged between intervention group membership and Physical Functioning as well as depressive symptomatology which could indicate that the adjustment for covariate s may be necessary to clarify the nature of group differences. However, very few of the included covariates were significantly associated with any of the dependent or independent variable s Levels of BioT and FT differed by regimen, indicating that partic ipants in the continuous ly suppressed group experienced a greater reduction of testosterone levels at all timepoint s. The timing of assessments was based on how much time had passed sinc e enrollment and the prior assessment timepoint rather than testoster one level Given that individuals receiving intermittent ADT undergo off and on treatment periods,

PAGE 58

47 testosterone levels may be higher in this group as a proportion of participants may be partially returning to higher levels of androgen hormones. Further an alyses assessing the presence and nature of group differences may therefore not clarify with certainty whether any observed effect on QOL can be attributed to the type of intervention or particular levels of BioT. Hypothesis 1 assessed whether treatment group me mbership predicted differences i n overall QOL after adjusting for several participant characteristics. The MANCOVA supported this hypothesis with g roup membership account ing for 67.5% of the variance seen in the composite QOL dependent variable af ter adjusting for covariates This large effect is comparable to the results of the Greenfield et al study (2010), which investigated the association between low testosterone levels after non hormone suppressing cancer treatment in young male cancer patie nts, and QOL using the SF 36. Effect sizes in that study range d from d = 0.22 to 0.67, with greater effects seen among physical health aspects of QOL. A similar trend was observed in this study. The specific nature of differences in QOL among the two intervention groups was tested under hypothesis 2. The tests for between subjects effects using MANCOVA, adjusted for the same covariates. However, only two of the nine included aspects of QOL showed statistical significant differences. The Physical Funct ioning and Role Limitations Due to Physical Health subscales of the SF 36 were significantly lower among participants in the continuous ADT group. Those two aspects of QOL were also the only two subscales that had adequate power to detect a statistically s ignificant effect. The review of effect sizes in prior studies also showed that a smaller effect could be expected for the association between androgen suppression and mental health QOL. T herefor e the

PAGE 59

48 exact nature of potential group differences in the rem aining aspects of QOL is unclear, given that this study was underpowered to detect those effects. However, a djusted group means for the SF 36 Role Limitations Due to Emotional Health and Energy subscales showed a trend towards lower average scores in the c ontinuous ADT group whereas average adjusted means for the Emotional Wellbeing and Social Functioning subscales were somewhat lower in the intermittent group. Hypothesis 3 investigated the association between androgen suppression and QOL from the angle of BioT levels instead of assessing group differences between the two treatment regimens Two significant multilevel correlations emerged between levels of BioT and Physical Fun ctioning as well as depression after adjusting for multiple observations among participants. Lower testosterone levels were therefore associated with lower levels of physical functioning and higher levels of depressi on Other non significant correlations were not always in the anticipated direction, but in terpretability of these results is limited due to the small magnitude of the remaining correlations. No significant higher order trend associations emerged after the visual inspection of scatter plot graphs and exploratory trend analyses. Limitations and Future Directions The results obtained upon testing hypotheses 1 and 2 indicate that a greater effect of androgen suppression is seen among patients receiving continuous ADT with regard to two aspects of physical health QOL. However, the design of this st udy limits the interpretation of these results. The addition of a true control group of untreated patients could help to determine whether the effect of androgen suppression on QOL is only

PAGE 60

49 present in the continuous administration group or whether both grou ps experience lower levels of QOL with greater impacts seen in patients receiving continuous ADT. The specific timing of assessment timepoint s in this study may have contributed to the overall higher levels of BioT in the intermittent ADT intervention gr oup. T he observed group differences seen under tests for hypothesi s 1 and 2 may therefore be due to actual effects of the intervention method, testosterone levels at the time of assessment, or a combination of both. At this time it is not clear how long it takes for effects on QOL after the initiation of ADT to become observable Therefore when testosterone and QOL levels are measured concurrently, the effect could be interpreted based on the influence of current testosterone levels or the lingering infl uence of prior changes in hormone levels Future studies on different approaches to ADT might consider how fluctuations in testosterone a ffects QOL over time by incorporating more frequent assessments of outcome variables. This study lacked the power to detect many of the anticipated group differences in specific aspects of QOL. A larger sample size may have helped to provide additional support for the hypothesis that greater impacts are seen in physical health in comparison to mental health aspects of QO L. However, a smaller effect seen in measures of emotional health may have gone unnoticed in this study due the small sample size. Also, the internal consistency of the SF 36 Social Functioning scale was lower than that of other subscales, which makes it a less reliable measure of that aspect of QOL. This could introduce more error variability and make it harder to detect an actual effect. However, it would be important to further assess this potential association of ADT and testosterone

PAGE 61

50 suppression on vari ous aspects of QOL in future studies as even subtle impacts on emotional health could have consequences for an individual's daily functioning. An alternative explanation to the observed group difference may lay in the differential treatment of patients re ceiving continuous and intermittent ADT. The nature of the intermittent approach is more individualized, as patients receive treatment medications after the partial or complete recovery of their testosterone levels. The intermittent approach may therefore require monitoring of testosterone levels that is more evident to the patient receiving this type of treatment. Patients receiving continuous ADT on the other hand are likely to receive the same dose of medication at regular intervals and patients may not be as aware of any monitoring procedures Th e intermittent approach may therefore be perceived as more personalized and patients may experience more attention f rom health care professionals while they undergo treatment for their pros tate cancer. The perceived additional attention from health providers may even promote patient provider communication. Prior research has shown that good patient physician communication is associated with improved patient health outcomes ranging from pati ents' emotional status and physical health, to blood pressure and blood glucose levels (Steward, 1995). Future investigators may therefore want to investigate whether there exist differences in patient provider relationships associated with the continuous and intermittent regimens, and whether such difference s might influence QOL. The association between levels of BioT and QOL was assessed using multilevel correlations. However, only two of the nine QOL domains were significantly correlated with levels of testosterone aggregated over all study timepoint s. These analyses which were aggregated across timepoint s included approximately 200 observations. It is

PAGE 62

51 therefore less likely that the non significant multilevel correlations were due to a lack of statistical power. Further, the estimates of correlations were small and several in the unanticipated direction. It is possible that a meaningful, non linear relationship exists between these variables even though the linear relation ships were not present. Within the multilevel structure, quadratic assoc iations were estimated between B ioT and each of the nine QOL domains. However, none of the potential quadratic associations approached significance. Trend analyses require large sample sizes in order to detect an effect and the exploratory analyses for hypothesis 3 may therefore have been underpowered. Future studies may therefore want to continue investigating whether higher order trends are more appropriate to describe the association between levels of BioT and QOL. Furthermore, as seen in Figure 3, the association between BioT and Physical Functioning became more pronounced at higher levels of BioT. This trend was also observed for several other aspects of QOL. Due to the nature of t his study, only men with suppressed levels of testosterone were included. Having access to the entire spectrum of testosterone levels may become necessary to fully understand the nature of the association between testosterone and QOL for future investigati ons of this area. Conclusion This study found evidence for an effect of androgen suppression on QOL looking at group differences between participants treated with continuous and intermittent ADT a nd inv estigating testosterone levels separately The res ults indicate that patients receiving continuous ADT may experience lower levels of certain aspects of physical health QOL (i.e. Physical Functioning and Role Limitations Due to P hysical Health). L ower levels of BioT were associated with lower levels of Physical Functioning and

PAGE 63

52 greater depression. Reduction in Physical Functioning was consistently implicated in this study across different hypothesis testing procedures and may be one of the most rea dily perceived side effects given that outcomes in this study were assessed by self report. Those implementing s upplemental interventions for patients undergoing ADT may therefore want to focus on restoring and maintaining overall physical health. Interve ning in one area of QOL may also have a positive infl uence on other aspects of QOL as m any of the assessed domains of QOL in this study were highly inter correlated with each other. However, whether a causal link exists among these domains of QOL remains t o be investigated. The association between lower BioT and greater depression was only seen while testing the effect of testosterone levels alone and not while assessing treatment group differences. It could therefore be anticipated that patients receiving either ADT regimen may experience symptoms of depression at different times during treatment based on their current testosterone level The flu ctuating levels of testosterone in patients receiving intermittent ADT may even contribute to a constantly changing mood profile. Shifts in mood can influence an individual patient's daily function ing and may contribute to challenges for his family members and caregivers. Awareness of these potential changes in mood could help to anticipate and prevent the development of depressive disorders in this population by providing early interventions for support that may also promote other ADT treatment outcomes. Further research is needed to examine the short and long term effects of ADT and changes in testosterone levels on QOL. This study provides support for the existence

PAGE 64

53 of such an association, but further research could clarify questions that remain unanswered

PAGE 65

54 REFERENCES Akakura, K., Bruchovsky, N., Goldenberg, S. L., Rennie, P. S., Buckley, A. R., & Sullivan, L. D. (1993). Effects of intermittent androgen suppression on androgen dependent tumors. Apoptosis and serum prostate sp ecific antigen. Cancer 71 (9), 2782 2790. Alibhai, S., Breunis, H., Timilshina, N., Johnston, C., Tomlinson, G., Tannock, I., et al. (2010). Impact of Androgen Deprivation Therapy on Physical Function and Quality of Life in Men With Nonmetastatic Prost ate Cancer. Journal of Clinical Oncology 28 (34), 5038 5045. American Cancer Society. (2014, February 3). Hormone (androgen deprivation) therapy for prostate cancer Retrieved February 15, 2014, from http://www.cancer.org/cancer/prostatecancer/detailed guide/prostate cancer treating hormone therapy American Cancer Society. (2014, March 12). Survival rates for prostate cancer Retrieved April 5, 2014, from http://www.cancer.org/cancer/prostatecancer/detailedguide/prostate cancer survival rates Bong, G. W. Clarke, J. H., Hancock, W. C., & Keane, T. E. (2008). Serum testosterone recovery after cessation of long term luteinizing hormone releasing hormone agonist in patients with prostate cancer. Urology 71 (6), 1177 1180. Bourke, L., Gilbert, S., Hooper, R., Steed, L. A., Joshi, M., Catto, J. W., et al. (2013). 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. European Urology 65 (5), 865 873. Carter, H. B. (2011). Management of low (favourable) risk prostate cancer. BJU International 1684 1695. Cary, K. C., Singla, N., Cowan, J. E., Carroll, P. R., & Cooperberg, M. R. (2013). Impact of Androgen Deprivation Therapy on Mental and Emotional Well Being in Men with Prostate Cancer: Analysis from the CaPSURE (Cancer of the Prostate Strategic Urologic Research Endeavor) registry. The Journal of Urology 191 (4), 964 971. Casey, R. G., Corcoran, N. M., & Goldenberg, L. (2 012). Quality of life issues in men undergoing androgen deprivation therapy: a review. Asian Journal of Andrology 226 231.

PAGE 66

55 CDC. (2013, October 23). Prostate Cancer Statistics Retrieved April 5, 2014, from Center for Disease Control and Prevention: htt p://www.cdc.gov/CANCER/prostate/statistics/ Crook, J., Szumacher, E., Malone, S., Huan, S., & Segal, R. (1999). Intermittent androgen suppression in the management of prostate cancer. Urology 53 (3), 530 534. Dacal, K., Sereika S. M., & Greenspan, S. L. (2006). Quality of life in prostate cancer patients taking androgen deprivation therapy. Journal of the American Geratrics Society 54 (1), 85 90. Dattalo, P. (2008). Determining Sample Size: Balancing Power, Precision, and Practicality (Pocket Guides to Social Work Research Methods). Oxford, United Kingdom: Oxford University Press. De Ronde, W., Van Der Schouw, Y. T., Polis, H. A., Gooren, L. J., Muller, M., Grobbee, D. E., et al. (2006). Calculation of Bioavailable and Free Testosterone in Men: A Comparison of 5 Published Algorithms. Clinical Chemistry 52 (9), 1777 1784. De Sousa Magalh‹es, S., Fernandes Malloy Diniz, L., & Cavalheiro Hamdan, A. (2012). Validity convergent and reliability test retest of the Rey Auditory Visual Learning Test. Clinical Neuropsychiatry 9 (3), 129 137. Dieperink, K. B., Hansen, S., Wagner, L., Johansen, C., Andersen, K. K., & Hansen, O. (2012). Living alone, obesity and smoking: Important factors for quality of life after radiotherapy and androgen depriva tion therapy for prostate cancer. Acta Oncologica 51 (6), 722 729. Felce, D., & Perry, J. (1995). Quality of life: Its definition and measurement. Research in Developmental Disabilities 16 (1), 51 74. Greenfield, D. M., Walters, S. J., Coleman, R. E. Hancock, B. W., Snowden, J. A., Shalet, S. M., et al. (2010). Quality of Life, Self Esteem, Fatigue, and Sexual Function in Young Men After Cancer. Cancer 116 (6), 1592 1601. Gregory, C. W., He, B., Johnson, R. T., Ford, O. H., Mohler, J. L., French, F. S., et al. (2001). A Mechanism for Androgen Receptor mediated Prostate Cancer Recurrence after Androgen Deprivation Therapy. Cancer Research 61 (11), :4315 4319. Hann, D., Winter, K., & Jacobsen, P. (1999). Measurement of depressive symptoms in cance r patients: evaluation of the Center for Epidemiological Studies Depression Scale (CES D). J Psychosom Res 46 (5), 437 43.

PAGE 67

56 Heidenreich, A., Bastian, P. J., Bellmunt, J., Bolla, M., Joniau, S., Mason, M. D., et al. (2013). Guidelines on Prostate Cancer. 2013: European Association of Urology. Herr, H. W., & O'Sullivan, M. (2000). Quality of life of asymptomatic men with nonmetastatic prostate cancer on androgen deprivation therapy. Journal of Urology 136 (6), 1743 1746. Hinton Bayre, A. D., Geffen, G. M ., Geffen, L. B., McFarland, K. A., & Friis, P. (1999). Concussion in contact sports: reliable change indices of impairment and recovery. Journal of clinical and experimental neuropsychology 21 (1), 70 86. Hojnacki, D., Munschauer, F., Morrow, S., Weinst ock Guttman, B., Drake, A., & Benedict, R. (2010). Psychometrics and normative data for the Multiple Sclerosis Functional Composite: replacing the PASAT with the Symbol Digit Modalities Test. Multiple Sclerosis 16 (2), 228 237. Iverson, G. L. (2001). Int erpreting change on the WAIS III/WMS III in clinical samples. Archives of Clinical Neuropsychology 16 (2), 183 191. Jenkinson, C. (1994). Criterion Validity and Reliability of the SF 36 in a Population Sample. Quality of Life Research 3 (1), 7 12. Kato, T., Komiya, A., Suzuki, H., Imamoto, T., Ueda, T., & Ichikawa, T. (2007). Effect of androgen deprivation therapy on quality of life in Japanese men with prostate cancer. International Journal of Urology 14 (5), 416 421. Keogh, J. W., Shephe rd, D., KrŠgeloh, C. U., Ryan, C., Masters, J., Shepherd, G., et al. (2010). Predictors of physical activity and quality of life in New Zealand prostate cancer survivors undergoing androgen deprivation therapy. The New Zealand medical journal 123 (13 25), 20 29. Kratiras, Z., Konstantinidis, C., & Skriapas, K. (2014). A review of continuous vs intermittent androgen deprivation therapy: Redefining the gold standard in the treatment of advanced prostate cancer. Myths, facts and new data on a "perpetu al dispute". International braz j urol 40 (1), 3 15. Lezak, M. D., Howieson D. B., Loring, D. W., Hannay, H. J., & Fischer, J. S. (2004). Neuropsychological assessment (4th Edition ed.). New York, NY: Oxford University Press. Mastin, L. (2010). Decla rative (Explicit) & Procedural (Implicit) Memory Retrieved April 21, 2014, from The Human Memory: http://www.human memory.net/types_declarative.html McLeod, S. (2008). Information Processing Retrieved April 21, 2014, from Simply Psychology: http://www.si mplypsychology.org/information processing.html

PAGE 68

57 Moncada, I. (2006). Testosterone and men's quality of life. The Aging Male 9 (4), 189 194. Moore, K. A., Jones, R. M., & Burrows, G. D. (2000). Quality of Life and Cognitive Function of Liver Transplant Pati ents: A Prospective Study. Liver Transplantation 6 (5), 633 642. Nelson, C. J., Lee, J. S., Gamboa, M. C., & Roth, A. J. (2008). Cognitive Effects of Hormone Therapy in Men With Prostate Cancer. Cancer 113 (5), 1097 1106. Nelson, W. G., Carter, H. B., DeWeese, T. L., Antonarakis, E. S., & Eisenberger, M. A. (2013). Abeloff's Clinical Oncology (5th edition ed.). Saunders. Nejat, R. J., Rashid, H. H., Bagiella, E., Katz, A. E., & Benson, M. C. (2000). A prospective analysis of time to normalization of ser um testosterone after withdrawal of androgen deprivation therapy. Urology, 164 (6), 1891 1894. Radloff, L. S. (1977). The CES D Scale: a self report depression scale for research in the general population. Applied Psychological Measurement, 1 (3), 385 401. Rosner, W., Auchus, R. J., Azziz, R., Sluss, P. M., & Raff, H. (2007). Position statement: Utility, limitations, and pitfalls in measuring testosterone: an Endocrine Society position statement. The Journal of clinical endocrinology and met abolism 92 (2), 405 413. Ross, T. P., Calhoun, E., Cox, T., Wenner, C., Kono, W., & Pleasant, M. (2007). The reliability and validity of qualitative scores for the Controlled Oral Word Association Test. Archives of clinical neuropsychology 22 (4), 475 488. Ruff, R. M., Light, R. H., Parker, S. B., & Levin, H. S. (1996). Benton Controlled Oral Word Association Test: reliability and updated norms. Archives of clinical neuropsychology 11 (4), 329 338. Saini, A., Berruti, A., Cracco, C., Sguazzotti, E. Porpiglia, F., Russo, L., et al. (2013). Psychological distress in men with prostate cancer receiving adjuvant androgen deprivation therapy1,. Urologic oncology 31 (3), 352 358. Salonen, A. J., Taari, K., Ala Opas, M., Viitanen, J., Lundstedt, S., & Tammela, T. L. (2012). Advanced Prostate Cancer Treated with Intermittent or Continuous Androgen Deprivation in the Randomised FinnProstate Study VII: Quality of Life and Adverse Effects. European Urology 63 111 120. Schršder, F., Crawford, E. D., A xcrona K., Payne, H., & Keane, T. E. (2012). Androgen deprivation therapy: past, present and future. BJU International 109 1 12.

PAGE 69

58 Stewart, M. A. (1995). Effective physician patient communication and health outcomes: a review. Canadian Medical Associati on Journal 152 (9), 1423 1433. Storey, D. J., McLaren D. B., Atkinson M. A., Butcher, I., Frew L. C., Smyth J. F., et al. (2012). Clinically relevant fatigue in men with hormonesensitive prostate cancer on long term androgen deprivation therapy Annals of oncology 23 (6), 1542 1549. Thong, M. S., van de Poll Franse, L., Hoffman, R. M., Albertsen, P. C., Hamilton, A. S., Stanford, J. L., et al. (2011). Diabetes mellitus and health related quality of life in prostate cancer: 5 year result s from the Prostate Cancer Outcomes Study. BJU International 107 (8), 1223 1231. Tunn, U. W., Canepa, G., Kochanowsky, A., & Kienle, E. (2012). Testosterone recovery in the off treatment time in prostate cancer patients undergoing intermittent androgen deprivation therapy. Prostate cancer and prostatic diseases 15 (3), 296 303. Vermeulen, A., Verdonck, L., & Kaufman, J. M. (1 999). A critical evaluation of simple methods of the estimation of free testosterone in serum. The Journal of Clinical Endocrinology & Metabolism 84 (10), 36666 3672. Ware, J. E. (2002). SF 36¨ Health Survey Update. Spine 25 (24), 3130 3139. Ware, J. E., & Kosinskia, M. (2001). Interpreting SF 36 Summary Health Measures: A Response. Quality of Life Research 10 (5), 405 413.