lelllHIHIIIWHUIINTIIHIIIIIHUNHINIIIHWIHIHM

   

(1007

This is to certify that the
thesis entitled

FOCTORS ASSOCIATED WITH TIME TO RESOLUTION OF
PAIN AND FATIGUE AMONG CANCER PATIENTS
UNDERGOING CHEMOTHERAPY IN A COGNITIVE
BEHAVIORAL INTERVENTION GROUP

presented by

Sangchoon Jeon

has been accepted towards fulfillment
of the requirements for the

MS. degree in Department of Epidemiology

 

 

/"7 M flip/XML

Major Professor’s Signature

g/IO/glocé
f I
Date

MSU is an Affirmative Action/Equal Opportunity Institution

 

 

LIBRARY
Mich“; Hate
University

 

 

 

..-.-.--.-o-u-------v-t-----.-.-.-.-q—.-.-

.o-.--.-.---u-.-.-.--.-o-.-.-.-o—-c..-

 

PLACE IN RETURN BOX to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

DMZERDblg ZOE8 DATE DUE DATE DUE
'9 5 0 2 0 8

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:IClRC/DateDue.indd-p.1

FACTORS ASSOCIATED WITH TIME TO RESOLUTION OF PAIN AND FATIGUE
AMONG CANCER PATIENTS UNDERGOING CHEMOTHERAPY IN A
COGNITIVE BEHAVIORAL INTERVENTION GROUP

By

Sangchoon Jeon

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Epidemiology

2006

 

ABSTRACT
FACTORS ASSOCIATED WITH TIME TO RESOLUTION OF PAIN AND FATIGUE

AMONG CANCER PATIENTS UNDERGOING CHEMOTHERAPY IN
A COGNITIVE BEHAVIORAL INTERVENTION GROUP

By

Sangchoon Jeon

To identify factors associated with resolution of pain and fatigue, we investigated

time to resolution of these symptoms among cancer patients receiving a cognitive

behavioral intervention (CBI). Randomly assigned cancer patients undergoing a first

course of chemotherapy received a 10 contact CBI intervention delivered by trained

nurses. Patients reported onset and resolution of symptoms at every contact, and

survival analysis techniques were employed to test difference in time to resolution by

patients’ characteristics including age, sex, cancer site, stage of cancer, depression, and

co-morbid conditions. The study found that patients with a late stage of cancer were

more likely to require longer time to resolve pain. Male patients had significantly

longer fatigue resolution times than female patients. High co-morbidity and depression

were significantly associated with increased time to resolution of fatigue. We conclude

that the duration of pain is more likely to depend on the stage of disease, while the

duration of fatigue is more associated with patients’ health conditions.

ACKNOWLEDGEMENTS

I wish to thank members of my guidance committee, M. Hossein Rahbar, Ph.D.,

Charles Given, Ph.D. and Joseph Gardiner, Ph.D. who provided invaluable direction,

intellectual support, and encouragement to complete this work. My thesis advisor, Dr.

Rahbar, has spent considerable time guiding me in the preparation of this thesis,

improving my epidemiological concepts, and completing this process. I would like to

express my appreciation to Drs Charles and Barbara Given for sharing their valuable data,

intellectual support, and encouragement. I also deeply thank my wife, Hyesuk, and

daughter, Hyoeun, for their patience and support during this work.

This research was financially supported by the Data Coordinating Center

(DCC) and the Family Home Care Study in Colleges of Human Medicine and Nursing at

Michigan State University. The Institutional Review Board (IRB) of Michigan State

University has approved (IRB # 06-150) this secondary analysis using collected data

from Family Home Care for Cancer in collaboration with the Walther Cancer Institute.

Drs Barbara and Charles Given were principal investigator and co-investigator,

respectively, for this clinical intervention trial. The project was funded by the National

Cancer Institute and the National Institute of Nursing Research (CA79280), 1999 to 2002.

iii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. v
LIST OF FIGURES ......................................................................................................... vi
INTRODUCTION ............................................................................................................. 1
Pain and fatigue in chemotherapy ................................................................................... 2
Assessments of symptom resolution after chemotherapy ............................................... 5
Proportion of patients with symptom resolved ................................................................ 6
Time to symptom resolution .......................................................................................... 10
Survival analysis for assessment of time to resolution of symptoms ............................ 12
METHODS ...................................................................................................................... 16
Study design and subjects .............................................................................................. 16
Intervention .................................................................................................................... l 8
Outcome Measures ........................................................................................................ 20
Statistical Analysis ......................................................................................................... 22
RESULTS ......................................................................................................................... 26
Burdens of multiple symptoms among cancer patients ................................................. 26
Characteristics of patients in the intervention group ..................................................... 27
Factors associated with time to resolution of pain ........................................................ 31
Factors associated with time to resolution of fatigue .................................................... 33
Cox proportional hazard model for time to resolution of pain and fatigue ................... 38
DISCUSSION .................................................................................................................. 42
Association with demographic characteristics .............................................................. 43
Associations with medical conditions ........................................................................... 46
Suggestions for further study ......................................................................................... 50
Limitations ..................................................................................................................... 51
REFERENCES ................................................................................................................ 53

iv

LIST OF TABLES

Table 1. Prevalence and resolution of 11 major symptoms during 10 contacts ................ 27

Table 2. Proportion of patients who reported pain during 10 contacts by patients’
characteristics ........................................................................................................ 28

Table 3. Proportion of patients who reported fatigue during 10 contacts by patients’

characteristics ........................................................................................................ 29

Table 4. Comparison of mean number of reported symptoms during 10 contacts by

patients’ characteristics ......................................................................................... 30
Table 5. Comparison of time to resolution of pain by patients’ characteristics ................ 32
Table 6. Comparison of time to resolution of fatigue by patients’ characteristics ............ 35

Table 7. Estimated mean time to resolution and hazard ratios of resolving pain and

fatigu e after adjusting for other covariates ......................................................... 41

Figure 1.

Figure 2.

Figure 3.

Figure 4.

Figure 5.

Figure 6.

Figure 7.

Figure 8.

LIST OF FIGURES

Flow chart of recruitment and assignment to intervention ................................ 17
Graphical demonstration of onsets and resolution time of symptoms .............. 25
Graphical presentation of censored data for onset and resolution time ............. 25
Survival curves for time to resolution of pain by stage of cancer ..................... 33
Survival curves for time to resolution of fatigue by patient’s sex ..................... 36
Survival curves for time to resolution of fatigue by site of cancer ................... 36
Survival curves for time to resolution of fatigue by CES-D status ................... 37

Survival curves for time to resolution of fatigue by co-morbidity status .......... 37

vi

INTRODUCTION

According to a report from National Cancer Institute (NCI) on January 2001

approximately 9.8 million persons had a history of invasive cancer in the United States.

The National Institutes of Health (NIH) reported the overall annual cost of cancer would

be about $189.8 billion in 2004. Symptom management, which provides appropriate

methods of treatment, is a very important issue in nursing care for cancer patients(l).

Patients who are diagnosed with solid tumors and undergoing chemotherapy treatment

may experience a variety of side effects such as pain, fatigue, nausea, insomnia, vomiting,

and anxiety. These symptoms could be originated from disease itself or the side effects

of treatments, and they are a serious burden for oncologists, primary care providers, as

well as patients. Due to improvement in early detection and treatment of cancer, more

patients have been cured or have survived longer. However, concerns about health

related quality of life or costs of cancer care are still important issues for cancer patients

and caregivers. The development of symptom management may provide improved

quality of life among cancer patients and relieve the burdens of family members and care

providers. The impact of cancer-related pain and fatigue is considerable in terms of

economic issues. For instance, Curt and his colleagues reported that 75% of patients

and 40% of caregivers changed their employment status due to fatigue in their study(2, 3).

In this study patients who received a cognitive behavioral intervention were

given strategies to resolve or moderate their symptoms. Specific strategies delivered to

patients with specific symptoms, and the intervention for symptom management across

10 contact times was scheduled. Not only the efficacy of symptom managements but

also the characteristics of patients such as age, site of cancer, stage of cancer and chronic

health condition may influence symptom improvement or resolution. The identified

patients’ characteristics related to time to resolution of the symptoms would be important

considerations when symptom management strategies are designed.

Pain and fatigue in chemotherapy

Pain and fatigue are frequently reported symptoms in cancer patients, and impact

dimensions of quality of life such as physical function and depression(2). The

symptoms usually co-occur with other symptoms and are related to the underlying cancer

or therapy. Approaches to the assessment of pain and fatigue remain an issue in nursing

care for cancer patients. Given and colleagues investigated the impact of personal

characteristics on the presentation of pain and fatigue in cancer patients. They found

that a late stage of cancer, high number of co-morbid conditions, and lung cancer were

strongly associated with the presence of pain and fatigue. Surgery, radiation, and

chemotherapy were significantly associated with pain, fatigue, or their combination

within 40 days of ending treatment(4).

The prevalence of pain in cancer patients varies from 36% to 75%, and patients

with advanced cancer experience substantial pain(5). Patients have severer pain as they

approach death. Cancer-related pain can be effectively managed in most patients even if

it cannot always be completely resolved. Pain usually occurs with other symptoms in

patients with advanced cancer. According to a study of 110 patients with advanced

cancer, patients may have a greater interference with activities of daily living when they

believe their pain to be caused by cancer rather than other etiologies(6). The NCI

recommends that the management for pain can be flexible and individualized by stage of

disease, personal preferences, and responses to pain and intervention. Cognitive-

behavioral interventions are an important management tool for pain. Some recent

studies have suggested that behavioral interventions for specific symptoms like pain and

fatigue are effective in reducing symptom burden and improving the quality of 1ife(7).

Fatigue is a common symptom among individuals with cancer who have

undergone chemotherapy. The National Comprehensive Cancer Network (NCCN)

defined cancer-related fatigue as “a persistent, subjective, sense of tiredness related to

cancer or cancer treatment that interferes with usual functioning”. A recent report from

the NCCN highlights that fatigue affects 70-100% of cancer patients(8). In a study

conducted by Greene et al, 82% of breast cancer patients who received regimens

including cyclophosphamide, methotrexate, and fluorouracil (CMF) reported fatigue after

the first course of chemotherapy(9). Other studies on perceptions of cancer-related

fatigue report a 75% agreement between patients and oncologists regarding the presence

of significant fatigue(10). A prospective cohort study with 157 breast cancer patients

receiving CMF reported that fatigue was highly prevalent during as well as after

chemotherapy treatment, and the proportion of patients suffering from fatigue started

decreasing 4 weeks after the last cycle of treatments(11). Compared to fatigue caused

by other diseases, cancer-related fatigue can be very rapid in onset and intense in severity.

Fatigue may be more distressing and interrupting to patients’ performances of daily

activities, relationships with people, and compliance with treatment than the pain. Curt

et. al. found that fatigue was ranked first (60%) as the symptom most affecting quality of

life among cancer patients(Z).

Despite the prevalence of fatigue among cancer patients, it has been difficult to

consistently identify factors associated with fatigue due to poor understanding of the

biochemical, physiological, and behavioral mechanisms of this complex symptom.

However, several risk factors associated with cancer-related fatigue have been suggested.

There are evidences that anemia, which is a common side effect of chemotherapy or

radiation therapy in cancer patients, is a major factor causing fatigue(10, 12). The

impact of anemia on fatigue may be different depending on onset time, patient age, and

comorbidity. Psychological factors such as depression and anxiety may contribute to

the development of chronic fatigue among patients with cancer before and after

chemotherapy(l3). Distress after diagnosis of cancer may cause initial fatigue and other

side effects of distress like insomnia may also increase levels of fatigue in patients

undergoing chemotherapy. As the result of distress, less daytime activity or more night-

time awakenings may contribute to cancer-related fatigue. Although several factors

contribute to the increased risk of cancer-related fatigue, the assessment of the effect of

each factor remains complex due to the multiple symptoms, the multiple etiologies,

varying severity, duration, and co-occurrence of symptoms.

Assessments of symptom resolution after chemotherapy

The cognitive behavioral intervention was developed to solve multi-component

problems including pain and fatigue among cancer patients undergoing chemotherapy

treatment. We tried to investigate which characteristics or health conditions helped or

impeded resolution of cancer-related pain and fatigue among cancer patients who

received the cognitive behavioral interventions.

Unlike symptoms from other diseases, symptoms experienced by cancer

patients undergoing chemotherapy have complex properties. First, symptoms are

temporal; the onsets and duration of symptoms may vary depending on patients’ chronic

conditions or the presence of other symptoms. Given the temporality of symptoms and

variation in their duration, follow-up time will differ depending on the time of onset for

each symptom. Second, many cancer patients undergoing chemotherapy may have

multiple-problems which have complex associations with other symptoms. Since the

presence of other symptoms may seriously affect the resolution or deterioration of pain or

fatigue, it is necessary to consider how other symptoms are associated with resolution of

pain and fatigue. Third, unlike symptoms in a normal population, symptoms in cancer

patients may have different origins. For example, pain can occur from tumor

progression, physical activity, or cancer treatment. Thus, the durations of symptoms

may differ according to their causes and sites of cancer.

Proportion of patients with symptom resolved

Many studies have examined the effect of treatments on managing symptoms or

identified the factors associated with symptom resolution. Different analytical
methodologies have been utilized in several studies to investigate the efficacy of
treatments or drugs in relieving or resolving symptoms. One approach is to compare
proportions of patients resolving a symptom among those who have a symptom at
specific time points or the end of the follow-up period(14—19). Proportion of symptom
resolution is an expression of the number of patients who resolve their symptoms among
patients experiencing the symptom in a follow-up time. In many conventional clinical
trial studies, the investigators monitored the improvement of symptoms until the end of
the research period and compared the proportions of patients who resolved or improved
symptoms between treatments or specific groups during follow-up. This method is
acceptable among studies for symptoms that are easily resolvable and have short
symptom durations, but not for cancer symptoms.

Nevertheless, in practical clinical trials there are a lot of limitations to measuring
the actual count of symptom resolutions within varying time intervals. Some symptoms
such as the side effects of diseases or treatments may be diagnosed not only in the initial
time of the follow-up but also in the middle, later in the follow-up, or around the end of
the follow-up interval. Since the proportion of patients with symptoms resolved is

observed at a fixed time point, there may be patients who do not have sufficient exposure

to the intervention due to late onset of the symptom. For example, patients who had a

symptom at the beginning of the follow-up have more chance to resolve symptom, while

patients who had a symptom around the end of the follow-up could not have enough time

to resolve the symptom. Rao and Cunningham conducted a similar study on symptoms

including; pain, anorexia, weight loss, nausea, and lethargy among patients with

advanced biliary cancer( 14). The researchers compared the proportions of resolved

symptoms between two treatment groups. Due to the nature of symptoms monitored in

the study, several patients might have had different onset times during the follow-up

period. Therefore, the proportions at the end of the follow-up could be influenced by

the onset time as well as the duration of symptoms.

Another limitation of using proportion of patients with resolved symptoms is that

the proportions may be different by timing of observations of symptom states. If

symptom resolutions are measured too early or late in the follow-up, the proportions of

symptom resolution may not capture the actual association between patients’ factors and

symptom resolution. In most cases, symptoms may not be resolved at an early follow-

up but may take more time to be resolved. Timing of capturing the proportion of

symptom resolution should be carefully decided based on the average duration of

symptoms. In some studies, the proportions of symptoms resolved were measured at

repeated times until the end of the follow—up. The prospective observational study

conducted by Curran and Kaefer presented proportions of symptom resolution rates at

every year during 5 years(15). They described the proportion of symptoms resolved for

each age group every year, but statistical tests were not employed to examine whether

there was a significant difference in the proportions of symptoms resolved between two

age groups across times. In general, the Chi-square test is used to examine the

difference in symptoms resolution rates of each group at each time. However, the Chi-

square test is not able to examine overall change of the symptom resolution rates across

time controlling for other covariates.

In many clinical trials or longitudinal studies, significant numbers of participants

are lost to follow-up. If participants are lost to follow-up due to death, severe

deterioration of symptoms, refusal to further participate, or other reasons, the final

symptom resolution of those attrited participants are not available to estimate the

proportion of symptoms resolved during the follow-up. The proportion of participants

with symptoms resolved can be overestimated if large numbers are lost to follow-up or

are not included in analyses. The loss of large numbers of participants could lead to a

smaller denominator of the proportion. In a randomized clinical trial to study symptoms

of gastro-esophageal reflux disease (GERD) which was conducted by Talley and Moore

had approximately 40% of patients lost to follow-up(18). The proportion of patients

with resolved symptoms was possibly overestimated due to the patients who dropped out

without knowing whether their symptom resolved.

Time to symptom resolution

Another approach to assessing symptom resolution is to measure time to

symptom resolution instead of proportion of patients with symptom resolved. Time to

symptom resolution, which is the length of time from symptom onset to symptom

resolution, is more informative than a binary outcome for proportion of symptom

resolution. Multiple linear regression model or the t-test are ofien employed in many

clinical trial studies to analyze time to resolution of a symptom(20-22). However, these

models assume that duration of symptom resolution has a normal distribution.

Therefore, the distribution of duration of a symptom resolution should be tested for

normality test before performing the analysis with the regression model or the t-test.

Among three identified studies using linear regression models or t-tests for analyses of

time to resolution, only one study used the Shapiro-Wilk W test to check whether time to

resolution was normally distributed(22). When time to resolution doesn’t have a normal

distribution, nonparametric methods such as the Wilcoxon Signed-rank test and Rank

10

sum test are available to compare differences in the distribution of time to resolution

between two groups.

When symptom resolution is not achieved for a given patient by the end of the

follow-up, the linear regression model and the t-test cannot use time to resolution in these

patients for analysis. The linear regression model and the t-test use only participants

who complete the follow-up study and where the symptom is resolved by the end of the

follow-up period. A prospective observational study with patients who had hemifacial

spasm conducted by Shin and Chung had 17% of participants who were lost to follow-up

or had incomplete resolution at the end of the follow-up period(21). Shin and Chung

used a linear regression model to identify factors associated with time to resolution of

hemifacial spasm symptom. In this study, 39 (17%) patients did not resolve symptoms

during the follow—up period or were lost to follow-up. Unfortunately, 39 patients

without complete time to resolution of symptom were not involved in the final model,

and the result of the treatment effect on resolution of hemifacial spasm symptoms could

be biased. Furthermore, in clinical trials when a treatment has significantly higher

proportions of lost to follow-up or incompletion of measuring time to resolution

compared with a placebo group, the linear regression model is likely to overestimate the

efficacy of treatment compared to the effect of placebo. The difference in the

11

distributions of patients with incomplete time to resolution, as well as the proportions of
the lost to follow-up between the two groups could lead biased results when linear
regression models or t-tests are used to evaluate the effect of treatment on time to

resolution of symptoms.

Survival analysis for assessment of time to resolution of symptoms

Incomplete measurement due to lost to follow-up or no resolution of a symptom
by the end of follow-up is known as a right-censored data. The main advantage of
survival analysis compared to the other methods is that survival analysis incorporates a
censored data. When right-censored data are observed due to lost to follow-up or
incomplete resolution by the end of the follow-up period, estimating proportions or use of
ordinary linear regression methods would not be appropriate. However, survival
analysis incorporates censored observations thereby reducing errors in evaluating the
distribution of time to symptom resolution. In most longitudinal clinical trials, there are
several participants who fail to complete the follow-up due to death, sickness, severe
symptoms, or loss of interest in the study. These issues impede complete assessment of
time to a symptom resolution, and increase right-censored data.

To examine time-to-event with censored data, several survival analysis

12

techniques have been developed. Kaplan and Meier (1958) introduced nonparametric

technique with survival curves. The product-limit estimator, which is called Kaplan

Meier estimator, uses both censored and non-censored observations to estimate survival

curve. Mantel (1966) developed the log-rank statistics to compare survival distributions

between two groups. Cox proportional hazard method (1972) was proposed for semi-

parametric method. This method does not require some particular probability

distribution, but assumes the same proportional hazard between groups over time.

Greenhouse and his colleagues addressed to use survival analysis method for evaluating

efficacy of a new treatment in clinical trials(23). Survival analysis method has been

widely used to analyze data with censoring, and enormous demands of statistical

methodology in clinical trials gives motivation to development of new survival analysis

techniques.

Time to symptom resolution assessed by survival analysis techniques such as

Kaplan Meier estimators or Cox proportional hazard models have the advantage of

incorporating right-censored data such as participants who fail to complete the follow-up

or to resolve their symptoms during the period. Even though these methods are

considered as nonparametric or semi-parametric, there are considered more suitable in the

presence of randomly right-censored data as compared to parametric methods. Some

13

recent studies used survival analysis techniques for assessing time to resolution of

symptoms to decrease bias from right censoring(24-27).

Product Limit estimator (Kaplan and Meier 1958) is a popular survival analysis

method for estimating the distribution of time to an event based on randomly right-

censored data. This function provides an estimator for the cumulative survival

distribution, which in our study represents the proportion of participants who have not

resolve their symptom up to time t . This estimate is a step function, which jumps only

at a time when a participant resolves his/her symptom. At the censored observations

there are no jumps except a situation where the last observation is censored in which case

the estimator drops to zero. It is important to note that the Kaplan Meier method

assumes independence between the time to an event and the time to censoring.

The Cox proportional hazard models are widely used when investigating

association between time to event such as time to symptom resolution and several other

independent variables simultaneously. The Cox-proportional hazard model provides the

estimates for the hazard ratio, which is the ratio of the hazard rates of symptom resolution

in a group compared to a referent group. In the context of symptom resolution, the

hazard rate at a particular time point, t, is the conditional probability that a participant

will resolve symptom in a short period of time after time, t, given that the symptom was

14

not resolved until time, t. The Cox proportional hazard model can control for the

effects of multiple covariates. In order to assess time to resolution of pain or fatigue,

many factors such as depression, comorbidity, and patient’s age would be considered in

the model. In this study, we will use Kaplan Meier and Cox Proportional hazard

techniques to identify significant predictors of time to resolution for pain and fatigue

among patients undergoing chemotherapy in a cognitive behavioral intervention delivered

by specially trained nurses.

15

METHODS

Study design and subjects

This study is part of a larger clinical intervention trial of Family Home Care for

cancer, which was funded by the National Cancer Institute (NCI) and the National

Institute of Nursing Research (CA79280), 1999 to 2002. This sub-study focused on

only symptom resolution among patients randomized at the beginning of the trial to the

intervention group, since no data were available for the control group that received

conventional care alone. Patients were accrued from two comprehensive cancer centers

and four community-oncology settings nurse. Recruiters were trained according to the

study protocol, and recruited 609 eligible patients who satisfied enrollment criteria for

entry (A first course of chemotherapy, greater than 20 years old, and being able to read

and speak English and cognitively able to respond to interviews). Also, patients should

not have experienced previous chemotherapy or radiation prior to the time of entry.

Consent from both patient and caregiver to participate in the study was required, and both

patients and caregivers had to be able to speak and read English, and both had to be

cognitively intact.

16

Figure 1. Flow chart of recruitment and assignment to intervention

 

609 eligible patients were

recruited from cancer centers

 

 

 

346 patients refused participation
- Lack of interest (N=115)
- Having no caregiver (N=59)
- Being overwhelmed by
disease or treatment (N=55)
- Too busy (N =48)
- Other (N=87)

 

 

 

 

 

263 patients signed the

consent form

 

 

 

 

................ p 26 patients didn’t complete

interview at baseline

 

 

 

 

237 patients and their caregivers

 

 

completed the interview at baseline

/ \

118 patients assigned to 119 patients assigned to

 

 

 

experimental group control group

 

 

 

 

 

 

 

110 patients schedule for

10 contact interventions

 

Two hundred and sixty three patients agreed to sign the consent forms, and 346

patients refused to participate due to lack of interest (N=155), having no caregivers

17

(N=59), being overwhelmed by disease and treatment (N=55), and being too busy (N=48).

In addition, 26 patients did not complete the intake interview due to inability to contact,

being too ill, or discontinuation of chemotherapy. This left 237 patients and their family

caregivers who completed the intake interview. They were randomized into either the

10-contact experimental intervention or conventional care. One hundred eighteen

patients and family caregivers were assigned to the experimental and 119 to the control

group(1). Among 118 patients in the experimental group, 8 patients were excluded from

analyses due to missing information.

Among 118 patients assigned to the experimental intervention, 110 patients were

scheduled for the 10-contact intervention to address their symptoms. Specially trained

nurses provided intervention strategies to the patients for the management of their

symptoms. In this study, subgroup analysis was performed for patients who reported

pain or fatigue during the follow-up period. Fifty of the 110 participants reported

experiencing pain and 91 reported experiencing fatigue during the follow-up period.

Intervention

The purpose of cognitive-behavioral intervention was to help patients to develop

self-management knowledge, skills, and behaviors that would enable them to manage

18

their symptoms. The intervention was based on cognitive-behavioral theory describing

adaptive strategies for patient’s addressing problems(1). After the intervention was

introduced the nurse evaluated each patient’s ability to the undertake requisite cognitive-

behavioral strategies designed to reduce the symptom severity, the impact on emotional

distress, and physical functioning(28, 29). The nurse provided specific strategies to

patients who were assigned to the experimental arm in order to assist them to manage 11

symptoms as well as their role and physical fimctioning problems, and emotional distress.

The nurses contacted patients in the intervention group to schedule the date and

times for up to lO-contacts and assessed the severity of patient’s symptoms among 11

master problems (fatigue, alopecia, Gastrointestinal tract problem, anxiety, constipation,

pain, insomnia, diarrhea, skin problem, respiratory problem, Mucositis). If the severity

of the symptom rated by patients on an llpoint scale ranging from zero (not noticeable),

to a ten (worst possible) was a five or higher, then the symptom was transferred to the

plan of care. Up to four intervention strategies could be delivered to patients for each

symptom at each contact. The strategies comprising the intervention content were

summarized using the following terms; assessment, counseling, prescribing, teaching,

skill, referring, evaluation, and consulting. At each contact patients reported the onset

and severity of each symptom and rated development of the symptoms in response to

19

interventions such as resolved, improved, no change, or deteriorated.

Outcome Measures

The primary outcome measure considered in this study is time to resolution of
pain and fatigue. In each contact scheduled, on average 2 weeks apart, the trained
nurses asked patients if they had each symptom during the past 2 weeks and how severe
the symptoms were at the time of visit. Although the interventions were scheduled for 2
weeks apart on average at the beginning of intervention, in most cases patient had next
contact later than 2 weeks. We found that significant number of patients completed their
10-contact intervention longer than even 30 weeks. The status of a symptom with respect
to the interventions employed for its managements were rated in four categories resolved,
improved, no change, and deteriorated.

Since patients were asked if they experienced each symptom during the past two
weeks, symptom onset was considered to be two weeks prior to the contact time when the
symptom was first reported. At subsequent contacts, patients reported the impact of the
intervention strategies on each symptom by rating it as having deteriorated, no change,
improved, and resolved. A symptom was considered resolved if a patient reported that

his/her symptom was resolved at the last contact reporting the symptom status. Time to

20

resolution was measured as the length of time from reporting symptom onset to reporting

symptom resolution. However, some patients withdrew or completed the last contact

without resolution of a symptom due to having individual health problems or reporting

onset of symptoms in the late period of study. These situations created censored

information. The length of time from onset of symptom to the last contact reporting the

symptom without resolution is considered as censoring time. Patient characteristics such

as depression and co-morbidity were measured at the baseline observation prior to

randomization. Patient’s depression level was measured using the CES-D scale

developed by the Center for Epidemiologic Studies (Radloff, 1977). The CES-D score

is a widely used reliable measure with established cutoff of 16 or greater indicating

potential for clinical depression(30). The CES-D score is measured based on 20-items

with 4-point Liert-type scale (almost all of the time, most of the time, some of the time,

and rarely/none of the time). Co-morbidity was measured by the total number of

chronic conditions including high blood pressure, diabetes, other cancer, chronic

bronchitis emphysema, heart problem, stroke, emotional problems, arthritis/Rheumatism,

fractured hip, liver disease, incontinence, and other major health problems at the first

interview. Site of cancer was collected from an audit of patient’s medical records and

categorized as breast, lung, and a number of other cancer sites. Stage of cancer was

21

categorized according to the tumor-node-metastasis staging criteria of the American Joint

Committee on Cancer. Based on this scale, stage of cancer was collapsed into early

stage (in situ or zero, and stage I and II) and late stage (stages III and IV)(1).

Information on patients’ age and sex were obtained from their first interviews.

Statistical Analysis

In this study, patients reported onset of symptom at different contact time during

the follow-up period. Patients reported resolution of symptoms within the follow-up

period, but some patients were lost to follow-up or did not resolve their symptom by the

last contact (Figure 2). Time to resolution, X ,,X 2,~-,X k for k patients, is defined

as the duration in weeks from symptom onset to symptom resolution. The censoring

variable, Y,, Y2,-~, Y, for k patients, is defined as the duration in weeks from symptom

onset to last contact time reporting symptom status when the symptom is not resolved.

Censored data the form (Z,,6,) where Z, =min(X,,Y,) , the minimum time

between time to resolution and the censoring time, and 5,. is an indicator function taking

a value equal to one if a patient reported resolution of symptom within the follow-up

period, and zero otherwise. In other words, if a patient reported resolution of symptom

within the follow-up period, the minimum time between time to resolution and the

22

censoring time Z, =X, and 6,. =1. Otherwise, Z, =Y, and 6, =0 (Figure 3).

Survival function of Z , is defined as S (t) = p{X >1}, 1 > O , which represents
the probability that time to resolution of symptom is more than I. At the time of onset
of symptom, i.e., when t= 0, the survival function S(0) =1 indicates that none of the
participants who reported the presence of symptoms have resolved it att = O. The
hazard function defined as h(t) = Lin; p{X St+At | X 2t}/ At represents the
conditional probability of resolving a symptom during a small time interval, t+At,
given that the symptom was not resolved before time t. The cumulative hazard
function is defined by, H(t) = E h (u )du .

In order to identify factors associated with time to resolution of pain and fatigue,
we used the Cox proportional hazard model (Cox 1972), which is a semi-nonparametric
method for survival analysis. Considering p covariates Z =(Z,,Z,,~--,ZP) , the Cox
proportional hazard model is defined as the following equation;
h(r|Z)=ho(r)exp(fl,Z, + ,6223 +---+ BPZP) , where h0(t) is a baseline hazard function
corresponding to no covariates, when all Z, = 0. The hazard ratio for a binary factor ZI
is defined by, h(t | zI =1)/h(t | zl = 0)= exp(,6l), which compares the rate of symptom
resolution among the patients who have factor Zl =1 with that of the patients with Zl = 0.

In our analysis, we investigated associations between patients’ characteristics

23

including sex, age, site and stage of cancer, co-morbidity, and CES-D at baseline and our

main outcome variables time to resolution of pain and fatigue separately. We estimated

median time to resolution of symptom instead of mean time. Median time to resolution

represents time that half of patients still have not resolved their symptom. Median time

was preferred over the mean in this study because the distribution of time to resolution of

symptoms is skewed to the right, with a long tail. Since a significant proportion of

patients still had not resolved their pain or fatigue at the end of the follow-up in this study,

we believe mean resolution time might provide biased estimates as compared to the

median resolution time which is less influenced by the outliers. However, the estimated

median time could have larger variability. Another limitation of the median time to

resolution is that when less than 50% of patients have resolved their symptoms within the

follow-up period, then the estimate for median resolution time is not available.

In our univariate analysis for assessing associations between time to resolution of

symptoms and each of patients’ characteristics, we used the log rank test. For

multivariate analysis, the Cox proportional hazard model was used to test associations

between patient’s factors and time to resolution after adjusting for other covariates. All

statistical tests were performed at 5% level of significance and 95% confidence intervals

were estimated based on the final model.

 

Figure 2. Graphical demonstration of onsets and resolution time of symptoms

 

 

 

 

 

 

 

 

A
Patient A
. Resolve
Patient B
. Resolve
Patient C
Not Resolved
Patient D
Not Resolved
1St Contact 10‘h Contact

Figure 3. Graphical presentation of cenSored data for onset and resolution time

A
Patient A

Patient B

Patient C

Patient D

 

Resolve

 

. Z=X,6=1

Resolve

 

. Z=X,6=1

Not Resolved

 

Z=X,5=0

Not Resolved

 

 

Z=X,6=0

 

‘
7

Time to onset symptom

Tim to symptom Resolution

25

RESULTS

Burdens of multiple symptoms among cancer patients

Patients in the cognitive behavioral intervention group were asked to report the

status of eleven symptoms at each contact time such as pain, fatigue, alopecia,

gastrointestinal problem, insomnia, mucositis, anxiety, constipation, diarrhea, respiratory

problem, and skin problem. Seventy-five percent of patients reported at least one

symptom within the first four visits. On average, patients reported 4.5 symptoms during

the follow-up. Fatigue, gastrointestinal problems, anxiety, pain, and alopecia had a

relatively higher prevalence than the other symptoms, and median times to resolution of

these five symptoms were between 15 and 20 weeks. Fatigue was the most prevalent

symptom reported (82.7%) among patients undergoing chemotherapy and median time to

resolution of fatigue was 20 weeks. Pain was reported by 45.5 % of patients, and

median time to resolution of pain was also about 20 weeks (Table 1).

Respiratory problems, diarrhea, skin problems, and mucositis had relatively

lower prevalence of reported symptoms, all less than 20%. Furthermore, the median

time to resolution of diarrhea and mucositis were 6 and 8 weeks respectively.

Prevalence and resolution of 11 major reported symptoms within 10 contacts along with

26

the median time to resolution of these symptoms are presented in Table 1.

Table 1. Prevalence and resolution of 11 major symptoms during 10 contacts

 

Symptoms reported within the

Median Time to resolution

 

 

 

Symptoms follow-up period (in weeks)

N % Median 95% CI
Fatigue 91 82.7 20 (16, 13)
Gastrointestinal problem 63 57.3 15 (13, 20)
Anxiety 62 56.4 20 (14, *)
Pain 50 45.5 20 (17, 28)
Alopecia 66 60.0 17 (11, 21)
Insomnia 48 43.6 15 (9, 19)
Respiratory problem 15 13.6 * ( *, * )
Constipation 51 46.4 10 ( 9, 19)
Diarrhea 21 19.1 6 ( 2, 7 )
Skin problem 20 18.2 ( 6, 23)
Mucositis 14 12.7 5 ( 4, 12)

 

 

 

 

 

*: cannot be reliably estimated due to availability of limited data

Characteristics of patients in the intervention group

Among the patients assigned to the intervention group, eighty-two patients (74%)

had completed 10 contacts while 26% of patients left the study before the tenth contact.

Seventy-four percent of patients were female and 55% were older than 60 years of age.

Thirty-eight percent of patients had breast cancer, 38.2 % had lung cancer, and 27% had

other types of cancer.

CES-D score greater or equal to 16, suggesting clinical depression.

Overall, 67% had late stage cancer.

27

About 25% of patients had

 

Forty four percent

of patients reported at least 3 co-morbid conditions.

Prevalence (%) of reported pain by patients’ characteristics is presented in Table

2. Overall, 45.5% of patients reported pain. Prevalence of reported pain among

patients with CES-D of at least 16 was 57%, the highest among other subgroups defined

in Table 2. Prevalence of reported pain among male patients was 55%. Prevalence of

reported pain among early stage cancer patients was the lowest, 36%.

Table 2. Proportion of patients who reported pain during 10 contacts by patients’

 

 

 

 

 

 

 

 

 

characteristics
No. of patients Prevalence (%) of
Factor N , ,
reported pain reported pain

Patient’s Sex Male 29 16 55.2

Female 81 34 42.0
Patient’s Age < 60 years old 49 25 51.0

2 60 years old i 60 25 41.7
Site of Cancer Breast Cancer 42 20 47.6

Lung Cancer 38 19 50.0

Other Cancers 30 11 36.7
Stage of Cancer Early Stage 36 13 36.1

Late Stage 74 37 50.0
Patient’s CES-D < 16 70 29 41.4

2 16 28 16 57.1
Co-morbidity 0 — 2 60 28 46.7

3 + 48 21 43.8
Overall 110 50 45.5

 

 

 

 

 

28

Prevalence of reported fatigue by patients’ characteristics is presented in Table 3.

Overall, 82.7% of the patients reported fatigue. Prevalence of reported fatigue among

breast cancer patients was 90.5%, the highest among other subgroups defined in Table 3.

Prevalence of reported fatigue among patients with CES-D of at least 16 was 89.3%.

Prevalence of reported fatigue among patients with other types of cancer was the lowest,

70%.

Table 3. Proportion of patients who reported fatigue during 10 contacts by patients’

 

 

 

 

 

 

 

 

 

characteristics
No. of patients Prevalence (%) of
Factor N . .
reported fatigue reported fatigue
Patient’s Sex Male 29 21 72.4
Female 81 70 86.4
Patient’s Age < 60 years old 49 43 87.8
2 60 years old 60 47 78.3
Site of Cancer Breast Cancer 42 38 90.5
Lung Cancer 38 32 84.2
Other Cancer 30 21 70.0
Stage of Cancer Early Stage 36 32 88.9
Late Stage 74 59 79.7
Patient’s CES-D < 16 70 58 82.9
2 16 28 25 89.3
Co-morbidity 0 — 2 6O 47 78.3
3 + 48 42 87.5
Overall 110 91 82.7

 

 

 

 

 

29

Patients assigned to the intervention group reported multiple symptoms at each
contact and the total number of symptoms that patients experienced could affect the
resolution of any given symptom. We compared the mean number of symptoms by
patients’ characteristics using t-test or ANOVA. Patient characteristics such as sex, age,
site of cancer, stage of cancer, and co—morbidity were not significantly associated with the
numbers of symptoms reported. However, patients with CES-D scores of at least 16 had
a mean of 5.4 symptoms while patients with CES-D less than 16 had a mean of 4.4
symptoms. This difference is considered marginally significant, p-value=0.0867.
Table 4 provides the mean number of reported symptoms during 10 contacts by patients’

characteristics along with the corresponding p-values

Table 4. Comparison of mean number of reported symptoms during 10 contacts by
patients’ characteristics

 

 

 

 

 

 

 

 

 

 

Total number of reported
Factor N symptoms P-value
Mean Std
Patient’s Sex Male 29 4.10 2.23 0.1889
Female 81 4.72 2.11
Patient’s Age < 60 years old 49 4.63 2.26 0.7511
2 60 years old 60 4.50 2.09
Site of Cancer Breast Cancer 42 4.90 2.07
Lung Cancer 38 4.45 2.18 0.3662
Other Cancer 30 4.20 2.22

 

 

30

 

 

 

 

 

 

 

 

 

Continues Table 4.
Total number of reported
Factor N symptoms P-value
Mean Std
Stage of Cancer Early Stage 36 4.47 2.17 0.7807
Late Stage 74 4.59 2.15
Patient’s CES-D Low: < 16 70 4.43 1.84
High: 2 16 28 5.39 2.66 0'0867
Co-morbidit 0 — 2 60 4.40 2.11
y 3 + 48 4.79 2.23 0'6773
Overall 110 4.52 2.17

 

 

 

 

 

Factors associated with time to resolution of pain

Median time to resolution of pain was estimated by Kaplan Meier method, and

the log-rank test was employed to test association between time to resolution of pain and

patients’ characteristics as shown in Table 5. Patients with early stage cancer had a

median pain resolution time of 10 weeks as compared with that of late stage cancer

patients who had a median pain resolution time of 23 weeks. This difference indicates a

significant association between stage of cancer and time to resolution of pain, p-

value=0.0113. We did not find any significant associations between time to resolution

of pain and other patient characteristics shown in Table 5. Figure 4 provides a graphical

presentation of significant association between time to resolution of pain and stage of

cancer.

31

Table 5. Comparison of time to resolution of pain by patients’ characteristics

 

 

 

 

 

 

 

 

 

Time to resolution of
Factor N pain in weeks P-value 8
Median 95% Cl
Patient’s Sex Male 16 19 (9, *)
0.9981
Female 34 20 (17, 28)
Patient’s Age < 60 years old 25 19 (13, 25)
0.5449
2 60 years old 25 23 (13, 28)
Site of Cancer Lung Cancer l9 19 (13, 28)
0.6497
Other Cancers 31 22 (13, 15)
Stage of cancer Early Stage 13 10 (7, 22)
0.0113
Late Stage 37 23 (19, 28)
Patient’s CES-D Low: < 16 29 20 (10, 23)
0.3115
High: 2 16 16 25 (13, 28)
Co-morbidity 0 — 2 28 22 (13, 25)
0.9434
3 + 21 19 (13, 28)
Overall 50 20 (17, 28)

 

 

 

 

 

 

 

*: cannot be reliably estimated due to limited data
a: p-value for testing the equality of time to resolution of pain by patients’ characteristics
using log-rank test

32

Figure 4. Survival curves for time to resolution of pain by stage of cancer

 

1.00"

    
 

 

Late Stage

 

0
~
0'!

 

 

 

 

 

 

Proportion of not resolved symptom

 

 

 

 

0.50 ‘ .
Early Stage
0.25
0.00
I I I l 1 TI
0 5 to 15 20 25 30

"EEK

Factors associated with time to resolution of fatigue

Comparison of median time to resolution of fatigue by patient’s sex, site of

cancer, CES-D status, and co-morbidity revealed several interesting associations.

Patients with high co-morbidity (at least 3) had a median fatigue resolution time of 23

weeks as compared with that of low co-morbidity who had a median fatigue resolution

time of 14 weeks. This difference indicates a significant association between co-

morbidity status and time to resolution of fatigue (p-value=0.0088). Patients with lung

cancer had a median fatigue resolution time of 29 weeks as compared with that of the

33

patients with other types of cancer who had a median fatigue resolution time of 19 weeks.

This difference suggests a significant association between site of cancer and time to

resolution of fatigue (p-value=0.0194). Patients with high CES-D (at least 16) had a

median fatigue resolution time of 23 weeks as compared with that of low CES-D who

had a median fatigue resolution time of 16 weeks. This difference was also statistically

significant with a p-value of 0.0290. We also found a statistically significant association

between the sex of the patient and time to resolution of fatigue, p-value=0.0143.

Female patients had a median fatigue resolution time of 19 weeks which was lower than

the median time to resolution for male patients. Since the survival curve for time to

resolution of fatigue for male patients remained above 0.5, as shown in Figure 5, we

could not reliably estimate the exact median time to resolution of fatigue for male

patients.

Median times to resolution of fatigue by other patients’ characteristics are shown

in Table 6. Figures 5-8 provide graphical presentations of significant associations

between time to resolution of fatigue and patients’ characteristics, sex, cancer site, CES-D

and co-morbidity status.

34

Table 6. Comparison of time to resolution of fatigue by patients’ characteristics

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Time to resolution of
Factor N fatigue in weeks P-value 8
Median 95% CI
Patient’s Sex Male 21 * C“, *)
.0143
Female 69 19 (14, 21)
Patient’s Age < 60 years old 43 18 (12, 23)
.2546
2 60 years old 46 21 (16, 30)
Site of Cancer Lung Cancer 32 29 (18, *)
.0194
Other Cancers 58 19 (12, 21)
Stage of Cancer Early Stage 32 20 (14, 22)
.6400
Late Stage 58 20 (16, 29)
Patient’s CES-D Low: < 16 24 16 (14, 21)
.0290
High: 2 16 58 23 (20, 30)
Co-morbidity 0 — 2 47 14 (10, 20)
.0088
3 + 41 23 (19, 30)
Overall 90 20 (16, 23)

 

 

*: cannot be reliably estimated due to limited data
a: p-value for testing the equality of time to resolution of fatigue by patients’
characteristics using log-rank test

35

Figure 5. Survival curves for time to resolution of fatigue by patient’s sex

Proportion of not resolved symptom

Proportion of not resolved symptom

 

0.?5‘

0.50“

0.25‘I

 

0.00‘

   

9

 

I Male Patients

 

 

 

 

 

 

 

    

 

 

 

 

 

 

 

T T I I I I T
5 IO 15 20 25 30 35
"EEK
Figure 6. Survival curves for time to resolution of fatigue by site of cancer
1.00‘ An
0.75‘
Lung Cancer
Other Cancer Sites A A ..
0.50% u u u V
0.25 * T‘ U v C
0.00“
I I r I I I I l
5 10 IS 20 25 30 35

"EEK

36

 

Figure 7. Survival curves for time to resolution of fatigue by CES-D status

 

 

 

 

       
 

 

 

 

 

 

 

 

1.004 I”
19‘“ CESDZ 16

S
Emmi
Lg CESD< 16
§ 0.50
‘5
G
“-t
o
e:
_O
E
8- 0 25‘
o .
5:

0.00 I I I r l I

0 5 lo 15 20 25 30

”EEK

Figure 8. Survival curves for time to resolution of fatigue by co-morbidity status

 

 

  
  
 

 

 

 

 

r .00 . 5‘
§ 0.75 Comorbidity 22 3
*I
% Comorbidity < 3
§ 0.50 ‘
a
o
«5'.
O
r:
,o
E
o
8" 0.25 ~
H
04
0.00
I I I I I I r
0 5 to 15 20 25 30

37

Cox proportional hazard model for time to resolution of pain and fatigue

The Cox proportional hazard models were employed to examine the
association between patients’ characteristics and time to resolution of pain and fatigue
after controlling for patients’ characteristics such as co-morbidity, depression level, age,
sex, and stage of cancer.

After controlling for the covariates, only stage of cancer maintained its
significant association with time to resolution of pain. Patients with early stage cancer
had the shorter time to resolution of pain than that of late stage cancer patients. The
ratio of the rates of resolving pain over time (Hazard Ratio: HR) for patients with early
stage of cancer compared to those with late stage of cancer is 3.06 (95% confidence
interval = [1.21, 773]). Patients with early stage cancer had an estimated median pain
resolution time of 13 weeks as compared with that of late stage cancer patients who had
an estimated median pain resolution time of 23 weeks. This ratio indicates a significant
association between stage of cancer and time to resolution of pain, p-value=0.0180. We
did not find any other significant associations between time to resolution of pain and
other patient characteristics. Also various interaction terms between covariates were
tested but these did not have any significant effect on time to resolution of pain.

While only stage of cancer was associated with time to resolution of pain, co-

38

morbidity, CES-D score, and patient’s sex had significant associations with time to
resolution of fatigue after controlling for other covariates. Patients with high co-
morbidity had the longer median time to resolution of fatigue compared with that of low
co-morbidity. The ratio of the rate of resolving fatigue for patients with the larger number
of co-morbid conditions (2 3) compared to those with less that 3 (< 3) co-morbid
conditions is 0.54 (95% confidence interval = [0.30, 096]). Patients with high co-
morbidity (Z 3) had an estimated median fatigue resolution time of 29 weeks as compared
with that of low co-morbidity who had an estimated median fatigue resolution time of 22
weeks. This ratio indicates a significant association between co-morbidity and time to
resolution of fatigue, p-value=0.0374.

Patients with high CES-D had the longer time to resolution of fatigue compared
with those with low CES-D. The'ratio of the rate of resolving fatigue for patients with
high CES-D (2 16) compared to those with low CES-D (< 16) is 0.53 (95% confidence
interval = [0.53, 103]). Patients with high CES-D (Z 16) had an estimated median
fatigue resolution time of 29 weeks as compared with that of low CES—D who had a
median fatigue resolution time of 22 weeks. This ratio indicates a marginally significant
association between CES-D score and time to resolution of fatigue, p-value=0.0628.

Female patients had shorter times to resolution of fatigue compared with male

39

patients. The ratio of resolving fatigue over time for female patients compared with male

patients is 3.73 (95% confidence interval = [1.53, 910]). Female patients had an

estimated median fatigue resolution time of 19 weeks as compared with male patients

who had an estimated median time to resolution of 30 weeks. This ratio indicates a

significant association between patient’s sex and time to resolution of fatigue, p-

value=0.0038.

The other patients’ characteristics were not significantly associated with time to

resolution of fatigue after controlling for covariates. Although lung cancer patients had

longer time to resolution of fatigue compared to those with other cancer sites in

univariate analysis, the effect of cancer sites was disappeared after controlling for

patient’s sex.

40

Table 7. Estimated mean time to resolution and hazard ratios of resolving pain and
fatigue after adjusting for other covariates

 

 

 

 

 

Median
Hazard 95% Hazard Ratio
Symptom Variables Time
Ratio Confidence Limits
(week)
Pain Late Stage 23
Stage of Cancer 3.06 1.21 7.73
Early Stage 13
Fatigue < 16 22
CES-D score 0.53 0.27 1.03
2 16 29
0 —— 2 22
Co-morbidity 0.54 0.30 0.96
3 + 29
Male 30
Patient’s Sex 3.73 1.53 9.10
Female 19

 

 

 

 

 

 

 

41

DISCUSSION

We proposed to use survival analysis techniques for assessing symptom

resolution in clinical trials. The survival methods allowed using the information of

patients who were lost to follow-up or didn’t have enough chance to resolve symptoms

until the end of the follow-up.

Only stage of cancer was significantly associated with time to resolution of pain.

Generally cancer patients at the end of life suffer serious pain, and these patients are more

likely to have late stage cancer. The late stage cancer related with the end of life may

contribute to the association between the late stage of cancer and longer time to

resolution of pain. However, resolution of pain was not significantly influenced by

other patient’s baseline medical conditions such as CES-D and co-morbidity at baseline.

Stage of cancer, patient’s sex, depression, and co-morbid conditions were

identified as significant factors related to time to resolution of fatigue in both the log-rank

test and the Cox proportional hazard model. Unlike pain, time to resolution of fatigue

was longer in patients with higher CES-D and co-morbidity at baseline. The presence

of high CES-D or multiple co-morbid conditions can make patients more tired and it will

be burdens to resolve fatigue. Male patients required longer time to resolve fatigue than

42

female patients.

The identified associations in this study will help to develop a strategy of

symptom management in several ways. For example, it will help to anticipate which

patients are potentially vulnerable to pain or fatigue and predict how long patients need to

receive interventions to resolve the symptoms.

Association with demographic characteristics

Patient’s sex: The experiences of pain and fatigue were reported differently

between male and female patients. Pain was more commonly reported in male patients,

while fatigue was more common in female patients in this study. However, the

difference in prevalence of reported pain and fatigue by patient’s sex was not statistically

significant. Patient’s sex is significantly associated with time to resolution of fatigue in

both univariate and multivariate survival analysis. Female patients resolved their

fatigue in a relatively shorter period of time as compared with male patients. More than

half of male patients reported that their fatigue was not resolved until 30 weeks. A

longitudinal study with lung cancer patients undergoing radiotherapy(31) and a cross-

sectional study for symptoms in advanced cancer(32) didn’t find gender difference in

prevalence of fatigue among cancer patients. Based on the result of her two studies

43

 

(RCT and cross sectional study) Miaskowski found no gender differences in severity of

pain among cancer patients. This was confirmed by Turk and Okifuji in a retrospective

study(33). Miaskowski also reviewed the relevant studies for gender difference in

cancer-related fatigue. In her review paper, four studies indicated that female patients

had higher fatigue severity than male patients while another four studies found no

significant gender differences in the severity or intensity of fatigue. Higher severity of

fatigue among female patients does not support the result of shorter time to resolution of

fatigue in female patients as discovered in our analysis. Along with the lack of

supportive evidence suggesting longer time to resolution of fatigue among male patients,

there are several limitations regarding gender effect on time to resolution of fatigue.

These include; unequal censoring of data between groups and correlations with site of

cancer. These issues will be discussed in the limitations section. Therefore, although

the model suggests that female patients are more likely to resolve fatigue earlier than

male patients, the gender differences are not conclusive.

Patient’s age: Age of patients is a potential factor associated with pain in terms

of prevalence, severity, and duration of symptom. Patients who were older than 59 year

of age reported less pain compared to younger patients. Traditionally, elderly patients

tend to underreport their pain, because they may consider pain as a normal part of aging

44

and they may believe that their complaint of pain will disturb physician’s treatments of

their cancer(34). A cohort study conducted with cancer patients in Michigan observed

that patients older than 64 years more frequently reported both pain and fatigue(4). The

underreporting of pain among elderly patients may be associated with higher levels of

severity or intensity of pain. If elderly patients report relatively severe pain, compared

to younger patients, then underreporting among elder patients may adversely affect their

time to resolution of pain. According to the result from univariate analysis with the log-

rank test, there was no significant age effect on time to resolution of pain even if median

time to resolution among elder patients (23 weeks) was longer than among younger

patients (19 weeks). The age effect was not significant after controlling for co-

morbidity, CES-D, patient’s sex, stage of cancer. The association between age and

anxiety, which contributes to chronic fatigue in cancer patients(35), possibly has

responsible to an insignificant aging effect. Among patients reporting pain, the mean

age of patients with anxiety was 57 years old and the mean age of those without anxiety

was 61 years old. In a longitudinal study conducted with hospitalized cancer patients in

Tokyo, younger patients are more distressed and reported more anxiety than elderly

patients(36). It will be meaningful test to assess the age effect on time to resolution of

pain in patients with and without anxiety separately in future study.

45

Associations with medical conditions

Stage of cancer: Stage of cancer is an important potential factor associated with

cancer-related pain. Unlike fatigue, pain was reported more frequently among late stage

cancer patients than those diagnosed at early stages. It was observed in previous studies

that advanced cancer was more likely to be related to occurrence of pain(4, 37).

However, fatigue was not significantly more prevalent among patients with late stage.

Late stage became a risk factor for pain and also disrupted the resolution of pain.

According to the result of both univariate and multivariate analyses in this study, stage of

cancer was identified as an important factor associated with time to resolution of pain.

Patients having late stage cancer needed longer times to resolve pain than those having

early stage cancer. By comparison, the median time to resolution of pain in late stage

cancer patients was more than two times the median time in early stage of cancer. After

adjusting for other covariates, the probability of resolving pain among early stage cancer

patients was more than three times the probability in late stage cancer patients. In an

earlier survey of cancer-related pain with the Brief Pain Inventory (BPI), 86% of patients

with advanced cancer believe that their pain caused by cancer itself(6). When patients

believe their pain is caused by cancer, they have a greater interference with activities of

46

daily living(5). Patients with late stage of cancer may experience more intensive and

severe pain, and their pain will be relatively more difficult to resolve compared to pain

among those with early stage cancer. Those burdens among patients with later stages of

cancer may lead to longer times to resolution of pain.

Co-morbidity: Patients’ co-morbidities at their baseline observation have shown

to be an important causal factor for cancer-related fatigue(8). In this analysis, patients

with higher numbers of co-morbid conditions were more likely to experience fatigue

compared to those with low numbers of co-morbid conditions. However, it was

observed that co-morbidity was not associated with the prevalence of pain.

In contrast, according to NCCN practice guidelines for cancer-related fatigue,

patients’ co-morbidity and depression are known to be associated with fatigue. The

guideline recommends that more attention should be paid to co-morbidity in conjunction

with the treatment of cancer-related fatigue(8). High co-morbidity at baseline was

significantly associated with longer time to resolution of fatigue. Patients reporting co-

morbid conditions more than 2 at baseline had 23 weeks of median time to resolution

while those who had lower co-morbid conditions had less than 14 weeks median time to

resolution. According to a cohort study among cancer patients who were older than 64

years of age, high co-morbidity, late stage of cancer, and lung cancer were related to both

47

pain and fatigue(4). They suggested that co-morbid conditions had an effect on pain and

fatigue through the way that cancer interacts with these conditions. We found that co-

morbidity was associated with fatigue in terms of prevalence and time to resolution. Co-

morbidity can be an important factor to provide apprOpriate management for relieving

cancer-related fatigue.

Depression: Depression at baseline is positively associated with occurrence of

pain and fatigue reported by cancer patients. In this study, patients with high CES-D

scores (2 16) reported pain and fatigue more commonly than those with lower CES-D

scores. Further, depression accompanying a diagnosis of cancer may be associated with

greater fatigue. In general, a measures of fatigue administered to cancer patients have a

high positive correlation with a measure of depression(38). Cancer patients may

experience depressive disorders from medical factors, psychological factors, and social

factors, and their depression also may be complicated by their fatigue. Fatigue can be

the result of depression and it can be also a cause of depression. Fatigue and depression

may co-occur without any evidence of a causal relationship, because depression can have

the same pathology as fatigue(39). Depression at baseline is associated with time to

resolution of fatigue as well as prevalence of fatigue. High CES-D was significantly

associated with longer time to resolution of fatigue in both univariate and multivariate

48

analyses. Correlations between fatigue and depression have been documented in cancer

patients(40, 41). A study with Hodgkin’s disease conducted by Loge and colleagues

found that 26% of the patients had fatigue for 6 months or longer and their fatigue was

correlated with higher levels of depression(42). In a longitudinal study performed with

cancer patients undergoing radiotherapy at a medical center in Amsterdam, the

correlation between fatigue measured by the Multidimensional Fatigue Inventory (MFI-

20) and the mood component of the CES-D was assessed at the start of treatment, 2

weeks after completion of treatment, and 9 months later(39). The correlations were

significant over time, and especially the correlation with both general fatigue and

physical fatigue was higher after treatment than at the start of treatment. Therefore,

patient’s CES-D score at baseline may be a good predictor of time to resolution of fatigue.

Although it was not observed that depression at baseline was significantly associated with

time to resolution of pain, there was the pattern of increasing time to resolution of pain as

increasing CES-D at baseline. The lack of significance between depression and time to

resolution of pain may be due to the small sample size. We had only 50 patients

reporting pain while 91 patients reported fatigue. This could be a reason for identifying

a fewer significant factors associated with time to resolution of pain as compared with

several significant factors associated with time to resolution of fatigue.

49

Suggestions for further study

Several factors significantly prolonged the time to symptom resolutions and they

included; late stage of cancer, high number of co-morbidity conditions, and high CES-D,

which are associated with intense symptom severity. Symptom severity could play an

important role between patient factors and time to resolution of pain and fatigue. It will

be necessary to investigate how overall symptom severity contributes to the time to

resolution of pain and fatigue.

We also suggest that it will be necessary to distinguish fatigue by self-reported

mental fatigue (lack of motivation and mental exhaustion) and physical fatigue

(limitations on activity) in future research. Physical fatigue factors were more

pronounced than mental fatigue factors(43). These two kinds of fatigue may differ in

cause, intensity, duration, and factors associated with resolution. Cognitive behavioral

interventions may differentially influence time to resolution between mental and physical

fatigue. Future research will be able to examine how differently patient factors and

intervention strategies influence the relief of mental and physical fatigue. Patients with

mental fatigue may require different strategies than those needed to resolve physical

fatigue.

50

Limitations

This is a secondary analysis of data from a cognitive behavioral intervention

trial. The original study did not intend to use survival analysis techniques for

identifying factors associated with time to resolution of symptoms. Therefore, the

actual dates of symptom onset and resolution were not collected, rather we estimated the

time to resolution of symptom based on the dates of contacts when patient reported

symptom onset or resolution.

While patients were scheduled to contact the trained nurses every 2 weeks, the

actual time intervals varied. Although the interval between contacts was scheduled to

be on average 2 weeks, some patients had longer intervals between contacts due to the

fact that they could not be contacted at the designated times. Unequal time intervals

between contacts may lead to an unequal chance to report symptom status among patients.

For instance, a patient could have had a 2-week time interval between the first and second

contact while another patient could have had 3-week interval. Even though both

patients might have resolved their pain within 2 weeks, the time to resolution of pain for

these two patients could be recorded differently. Therefore, estimation of time to

resolution could be influenced by the different time intervals.

Consistency of self-reports of pain and fatigue is a considerable issue. Due to

51

a lack of uniformity in measurement and methodology in symptom research, the

occurrence of pain and fatigue had been estimated in large range of intervals across

studies. Reasons for the lack of consistency in terms of measurement issues include

individual difference in conceptualization of pain and fatigue and lack of consensus on

the criteria to define the resolution of symptoms(44). Since measuring symptom

resolution depends on self-reporting by patients, individual characteristic may influence

the variation of time to resolution of symptoms. The gender effects on time to

resolution of pain and fatigue were possibly biased by individual difference in definitions

of symptom resolution.

In this study, some characteristics such as patient’s sex and stage of cancer have

unequal proportions of censored data. Among patients reporting fatigue, 71% of male

patients and 36% of female patients were censored, and among patients reporting pain

38% of patients with early stage of cancer and 62% of those with late stage cancers were

censored. If unequal censoring occurs due to beneficial or adverse effect of gender or

stage of cancer, then the comparisons may be biased.

52

 

10.

REFERENCES

Given CW. GB, Rahbar M., Jeon S. et al. Effect of Cognitive Behavioral
Intervention on Reducing Sympto Severity during Chemotherapy. Journal of
Clinical Oncology 2004;22:507-516.

Curt GA, Breitbart W, Cella D, et a1. Impact of cancer-related fatigue on the lives
of patients: new findings from the Fatigue Coalition. Oncologist 2000;5z353-60.

Curt GA BW, Cella D, Groopman JE, Homing SJ, Itri LM, et a1. Impact of
cancer-related fatigue on the lives of patients: new findings from the fatigue
coalition. Oncologist 2000;5z353-360.

Given CW, Given B, Azzouz F, Kozachik S, Stommel M. Predictors of pain and
fatigue in the year following diagnosis among elderly cancer patients. J Pain
Symptom Manage 2001;21:456-66.

Weiss SC, Emanuel LL, Fairclough DL, Emanuel EJ. Understanding the
experience of pain in terminally ill patients. Lancet 2001;357:1311-5.

Twycross R. HJ, Bergl S. A survey of pain in patients with advanced cancer.
Journal of Pain and Symptom Management 1996;12:273-82.

Given CW. GB, Kozachik S., et a1. Pain and fatigue management: results of a
randomized clinical trial. Oncology Nurse Forum 2002;29:949-956.

Mock V AA, Barsevick A, et a1. National Comprehensive Cancer Network
oncology practice guidelines for cancer-related fatigue. Oncology 2000;14:151-
161.

Greene D NL, Fieler VK, et al. A comparison of patient-reported side effects
among three chemotherapy resimens for breast cancer. Cancer Practice 1994;2157-
62.

Vogelzang NJ BW, Cella D, Curt GA, Groopman JE, Homing SJ, et a1. Patient,
caregiver, and oncologist perceptions of cancer-related fatigue: results of a triparte
assessment survey. Semin Hematol 1997;34:4-12.

53

ll.

12.

13.

14.

15.

l6.

17.

18.

19.

20.

de Jong N, Candel MJ, Schouten HC, Abu-Saad HH, Courtens AM. Prevalence
and course of fatigue in breast cancer patients receiving adjuvant chemotherapy.
Ann Oncol 2004;15:896-905.

Glaspy J, Bukowski R, Steinberg D, Taylor C, Tchekmedyian S, Vadhan-Raj S.
Impact of therapy with epoetin alfa on clinical outcomes in patients with

nonmyeloid malignancies during cancer chemotherapy in community oncology
practice. Procrit Study Group. J Clin Oncol 1997;15:1218-34.

Reich SG The tired patient: psychological versus organic causes. Hosp Med
1986;22:142-154.

Rao S, Cunningham D, Hawkins RE, et a1. Phase III study of 5FU, etoposide and
leucovorin (FELV) compared to epirubicin, cisplatin and SF U (ECF) in previously
untreated patients with advanced biliary cancer. Br J Cancer 2005;92:1650-4.

Curran MJ, Kaefer M, Peters C, Logigian E, Bauer SB. The overactive bladder in
childhood: long-term results with conservative management. J Urol
2000;163:574-7.

Miner TJ, Brennan MF, J aques DP. A prospective, symptom related, outcomes
analysis of 1022 palliative procedures for advanced cancer. Ann Surg
2004;240:719-26; discussion 726-7.

Marrie TJ, Lau CY, Wheeler SL, Wong CJ, Feagan BG Predictors of symptom
resolution in patients with community-acquired pneumonia. Clin Infect Dis
2000;3 1 : 1362-7.

Talley NJ, Moore MG, Sprogis A, Katelaris P. Randomised controlled trial of
pantoprazole versus ranitidine for the treatment of uninvestigated heartburn in
primary care. Med J Aust 2002;177:423-7.

Le Saux N, Gaboury I, Baird M, et al. A randomized, double-blind, placebo-
controlled noninferiority trial of arnoxicillin for clinically diagnosed acute otitis
media in children 6 months to 5 years of age. Cmaj 2005;172:335-41.

Pratter MR, Bartter T, Akers S, DuBois J. An algorithmic approach to chronic
cough. Ann Intern Med 1993;119:977-83.

54

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

Shin J C, Chung UH, Kim YC, Park CI. Prospective study of microvascular

decompression in hemifacial spasm. Neurosurgery 1997;40:730-4; discussion
734-5.

Prasad AS, Fitzgerald JT, Bao B, Beck F W, Chandrasekar PH. Duration of
symptoms and plasma cytokine levels in patients with the common cold treated

with zinc acetate. A randomized, double-blind, placebo-controlled trial. Ann
Intern Med 2000;133:245-52.

Greenhouse J B, Stangl D, Bromberg J. An introduction to survival analysis:
statistical methods for analysis of clinical trial data. J Consult Clin Psychol
1989;57:536-44.

Yale SH, Liu K. Echinacea purpurea therapy for the treatment of the common
cold: a randomized, double-blind, placebo-controlled clinical trial. Arch Intern
Med 2004;164: 1237-41.

Sjostrom S, Sillen U, Bachelard M, Hansson S, Stokland E. Spontaneous
resolution of high grade infantile vesicoureteral reflux. J Urol 2004;172:694-8;
discussion 699.

Metlay JP, Atlas SJ, Borowsky LH, Singer DE. Time course of symptom
resolution in patients with community-acquired pneumonia. Respir Med
1998;92:1137-42.

Coenen S, Van Royen P, Michiels B, Denekens J. Optimizing antibiotic
prescribing for acute cough in general practice: a cluster-randomized controlled
trial. J Antimicrob Chemother 2004;54:661-72.

Oliver J W, Kravitz RL, Kaplan SH, Meyers FJ. Individualized patient education
and coaching to improve pain control among cancer outpatients. J Clin Oncol
2001;19:2206-12.

McLachlan SA, Allenby A, Matthews J, et a1. Randomized trial of coordinated
psychosocial interventions based on patient self-assessments versus standard care

to improve the psychosocial functioning of patients with cancer. J Clin Oncol
2001;19:4117-25.

Comstock GW, Helsing KJ. Symptoms of depression in two communities.

55

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

Psychol Med 1976;6z551-63.

Hickok JT, Morrow GR, McDonald S, Bellg AJ. Frequency and correlates of
fatigue in lung cancer patients receiving radiation therapy: implications for
management. J Pain Symptom Manage 1996;11:370-7.

Walsh D, Donnelly S, Rybicki L. The symptoms of advanced cancer: relationship
to age, gender, and performance status in 1,000 patients. Support Care Cancer
2000;82175-9.

Miaskowski C. Gender differences in pain, fatigue, and depression in patients
with cancer. J Natl Cancer Inst Monogr 2004:139-43.

Ward SE, Goldberg N, Miller-McCauley V, et a1. Patient-related barriers to
management of cancer pain. Pain 1993;52:319-24.

Reich SG The tired patient: psychological versus organic causes. Hosp Med
1986;22:142-154.

Matsushita T, Matsushima E, Maruyama M. Anxiety and depression of patients
with digestive cancer. Psychiatry Clin Neurosci 2005;59:576-83.

Daut RL, Cleeland CS. The prevalence and severity of pain in cancer. Cancer
1982;50:1913-8.

Jacobsen PB. Assessment of fatigue in cancer patients. J Natl Cancer Inst Monogr
2004:93-7.

Visser MR, Smets EM. Fatigue, depression and quality of life in cancer patients:
how are they related? Support Care Cancer 1998;62101-8.

Hopwood P, Stephens RJ. Depression in patients with lung cancer: prevalence and
risk factors derived from quality-of-life data. J Clin Oncol 2000;18:893-903.

Newell S, Sanson-Fisher RW, Girgis A, Ackland S. The physical and psycho-
social experiences of patients attending an outpatient medical oncology
department: a cross-sectional study. Eur J Cancer Care (Engl) 1999;8173-82.

Loge J H, Abrahamsen AF, Ekeberg, Kaasa S. Fatigue and psychiatric morbidity
among Hodgkin's disease survivors. J Pain Symptom Manage 2000;19:91-9.

56

 

43.

44.

F urst CJ AE. Dimensions of fatigue during radiotherapy: an application of the
multidimensional Fatigue Inventory. Support Care Cancer 2001;91355-360.

Patrick DL, Ferketich SL, Frame PS, et a1. National Institutes of Health State-of-
the-Science Conference Statement: Symptom Management in Cancer: Pain,
Depression, and Fatigue, July 15-17, 2002. J Natl Cancer Inst 2003;95:1110-7.

57

 

   

I[illjgrllijjljijtijjjl