- Email: [email protected]
Multidimensional independent predictors of cancer-related fatigue
604 Journal of Pain and Symptom Management
Vol. 26 No. 1 July 2003
Original Article
Multidimensional Independent Predictors of Cancer-Related Fatigue Shirley S. Hwang, RN, MS, Victor T. Chang, MD, Montse Rue, PhD, and Basil Kasimis, MD Section of Hematology/Oncology (S.S.H., V.T.C., B.K.) and Patient Care Services (S.S.H.), VA New Jersey Health Care System, East Orange, New Jersey; UMDNJ/School of Nursing (S.S.H.) and UMDNJ/New Jersey Medical School (V.T.C., B.K.), Newark, New Jersey; and Department of Biostatistical Science (M.R.), Dana-Farber Cancer Institute, Boston, Massachusetts, USA
Abstract The purpose of this study was to identify independent predictors of clinically significant fatigue based upon a multidimensional model. A total of 180 cancer patients completed the Brief Fatigue Inventory (BFI), Functional Assessment of Cancer Therapy-Fatigue (FACTF), Memorial Symptom Assessment Scale Short Form (MSAS-SF), and the Zung Self-Rating Depression Scale (SDS). Additional data included Karnofsky Performance Status (KPS) score, laboratory tests, and demographic information. The BFI usual fatigue severity ⱖ 3/10 was defined as clinically significant fatigue. Possible independent variables were identified from a biopsychosocial model of fatigue. Fisher’s exact test was used to univariately assess the association of each variable with clinically significant fatigue. Multiple logistic regression analyses were used to identify independent predictors of fatigue within each dimension, and then across dimensions. Fatigue was present in 113 (62%) patients, and 80 (44.4%) patients had usual fatigue ⱖ 3/10. The unidimensional independent predictors were use of analgesics (situation dimension); hemoglobin and serum sodium (biomedical dimension); feeling drowsy, dyspnea, pain and lack of appetite (physical symptom dimension); and feeling sad and feeling irritable (psychological symptom dimension). In a multidimensional model, dyspnea, pain, lack of appetite, feeling drowsy, feeling sad, and feeling irritable predicted fatigue independently with good calibration (Hosmer Lemeshow Chi Square ⫽ 5.73, P ⫽ 0.68) and discrimination (area under the receiver operating characteristic curve ⫽ 0.88). Physical and psychological symptoms predict fatigue independently in the multidimensional model, and superseded laboratory data. These findings support a symptom-oriented approach to assessment of cancer-related fatigue. J Pain Symptom Manage 2003;26:604–614. 쑖 2003 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.
Address reprint requests to: Shirley S. Hwang, RN, MS, Section Hematology/Oncology (111), VA New Jersey Health Care System at East Orange, 385 Tremont Avenue, East Orange, NJ 07018, USA. The views expressed herein do not necessarily reflect the views of the Department of Veterans Affairs or of the U.S. Government. Accepted for publication: November 8, 2002.
쑖 2003 U.S. Cancer Pain Relief Committee Published by Elsevier Inc. All rights reserved.
Preliminary results presented at the Annual Scientific Meeting of the American Society of Clinical Oncology at Los Angeles, California (Proc ASCO 1999;18:594a, abstract 2294).
0885-3924/03/$–see front matter doi:10.1016/S0885-3924(03)00218-5
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
605
Key Words Cancer, fatigue, predictors, analgesics, anemia, pain, feeling drowsy, dyspnea, lack of appetite, depression, veterans, sad
Introduction Fatigue is a highly prevalent symptom in cancer patients, and interest in cancer-related fatigue has increased significantly in the last few years. Fatigue can result from the prolonged stress caused by multiple factors1,2 and it may represent a final common pathway from these factors.3 Predisposing factors include demographic characteristics, underlying disease, treatment received, comorbidities, sleep disorders, immobility, and psychosocial factors.4 Depending on the study population, variables used, and fatigue measures, different independent predictors of fatigue have been reported in the literature. These include depression,5 psychological distress,6–8 physical symptoms,7,9 dyspnea,5,8 anxiety,8 poor quality of sleep,5–7,10 sex,11,12 age,11 primary cancer site,11,12 cancer treatment,10 stage of disease,11 and laboratory values such as low albumin,13,14 and anemia.13,15 The relative importance of these predictors is not known. In order to better understand the fatigue characteristics and to define clinically significant fatigue level in male veteran cancer patients, we performed a survey study using the Brief Fatigue Inventory (BFI),14 Functional Assessment of Cancer Therapy—Fatigue (FACT-F),16 the Zung Self-rating Depression Scale (SDS),17 the Memorial Symptom Assessment Scale Short Form (MSAS-SF),18 and demographic data. We also obtained the complete blood counts and chemistry profiles after the completion of the study survey. We validated both the FACT-F and BFI in our population, and found that there were significant correlations between the BFI fatigue severity, FACT-F fatigue measure, and “lack of energy” item from MSAS-SF, with correlation coefficients ranging from 0.68 to 0.88. Higher scores on the fatigue measures were associated with lower Karnofsky Performance Status (KPS)19and QOL, higher symptom distress, and greater depression scores.13 We also identified clinically relevant fatigue levels by examining the association between BFI fatigue severity and broader QOL con-
structs, such as symptom distress parameters, depression, QOL, and functional status. In summary, patients with a usual fatigue severity ⱖ 3/10 demonstrated significantly higher functional interference scores, lower QOL, and greater symptom distress.20 In this report, we performed a secondary exploratory analysis to identify the multidimensional independent predictors of clinically significant fatigue.
Methods Theoretical Model We used a biopsychosocial model to assess the patient and as a source of predictor variables for clinically significant fatigue. The predictor variables were divided into four dimensions (situational, biological, physical symptoms and psychological symptoms) and are summarized in Figure 1 and Table 1.
Patient Selection In this prospective cross-sectional design survey study, 74 outpatients and 106 inpatients diagnosed with cancer and treated at the Hematology/Oncology Section at VA New Jersey Health Care System at East Orange, NJ (VANJHCS) were recruited from September 1997 to May 1998. The VANJHCS is the sole tertiary care teaching hospital providing Hematology/Oncology services for veterans residing in the state of New Jersey. The study was approved by the Institutional Review Board, and all patients signed informed consent before participating. Each patient completed the surveys listed below and had blood drawn for a biochemical profile and complete blood count after completion of survey instruments. Demographic data were also determined.
Instruments The MSAS-SF is a validated, patient-rated instrument in which patients rate symptom distress for 28 highly prevalent physical symptoms and symptom frequency for four psychological symptoms. Each symptom is scored from
606
Hwang et al.
Vol. 26 No. 1 July 2003
Fig. 1. Proposed multidimensional conceptual model to predict fatigue.
0 to 4 ranging from “no symptom” to “very much.” For this report, we assessed the presence of nine physical symptoms (pain, difficulty concentrating, feeling drowsy, lack of appetite, weight loss, dyspnea, difficulty sleeping, constipation, and dry mouth) and the presence of
the four psychological symptoms (feeling sad, feeling nervous, feeling irritable, and worry) as predictors of clinically significant fatigue. The FACT-F16 is based on the 28- item FACT21 G —patient-rated measure of quality of life for cancer patients with any tumor type—with an
Table 1 Multidimensional Assessment to Identify Fatigue Predictors and Outcome Variables
Assessment Dimensions
Measurement/Instruments
Variables
Situational
Demographic data
Biomedical
Laboratory tests
Primary cancer sites Stage of disease Inpatient status Analgesic use 24-hour analgesic dosage Caregiver status Active cancer treatment Hours of sleep Substance abuse Age Chemistry profiles: albumin, BUN, creatinine, calcium, sodium, potassium, LDH, SGOT, SGPT, bilirubin. blood counts: WBC, Hgb Pain, difficulty concentrating, dry mouth, feeling drowsy, difficulty sleeping, dyspnea, sweats, lack of appetite, difficulty swallowing, weight loss, constipation Feeling sad, feeling nervous, worrying, feeling irritable and depression
Physical symptom
MSAS-SF
Psychological symptom
MSAS-SF Zung SDS
Significant Variables by Univariate Analysis Formed the Unidimensional Model Stage of disease Inpatient status Analgesic use 24-hour analgesic dosage
Hemoglobin level Serum sodium
Pain, difficulty concentrating, dry mouth, feeling drowsy, difficulty sleeping, dyspnea, lack of appetite, weight loss, constipation Feeling sad, feeling nervous, worrying, feeling irritable and depression
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
additional 13 items designed to measure fatigue. The FACT-F fatigue subscale is a multidimensional fatigue assessment instrument, and measures multiple fatigue characteristics and their impact on function. The BFI14 consists of two components: fatigue severity and fatigue interference. Both components are assessed using a 0–10 numeric scale. The fatigue severity assesses fatigue at its worst, usual, and right now; the fatigue interference component assesses fatigue-related functional interference in the areas of daily activity, mood, walking, work, enjoyment of life, and relationship with others. The total BFI fatigue interference is the sum of the six interference scores. The Zung SDS17 is a validated and simple instrument to assess depression. It uses a 1–4 numeric rating to assess 20 questions with a maximum possible raw score of 80. The SDS indices were derived by dividing the sum of the raw scores of the 20 items by 80 and expressing the result as a decimal. The mean value for the normal control subjects and depressive patients were established. There are three items in the Zung SDS which reflect the measurement of somatic symptoms possibly related to fatigue: “I eat as much as I used to,” “I notice that I am losing weight,” and “I get tired for no reason.” We reconstructed the depression indices by removing these three items from the calculation. The Cronbach alpha coefficient for the reconstructed 17-item depression indices was 0.84. This approach has previously been used by other fatigue researchers.8,10
Definition of Independent Variables and Modeling Rationale for Each Dimension The independent variables used in each dimension are summarized in Table 1. In the situational dimension, clinical variables related to cancer diagnosis, treatment (primary sites, stage of disease, inpatient status, treatment received, use of analgesics, 24-hour analgesics dosage, age), and environmental factors (hours of sleep, substance abuse, and caregiver status) were included. In the biomedical dimension, variables from laboratory tests were examined. In the physical symptom dimension, six frequently reported physical symptoms related to fatigue (pain, difficulty concentrating, dyspnea, lack of appetite, weight loss, and difficulty sleeping) and three other highly prevalent physical
607
symptoms (feeling drowsy, dry mouth, and constipation) from the MSAS-SF were selected. In the psychological symptom dimension, the Zung scale for depression, and four other psychological symptoms measured by MSAS-SF (feeling nervous, worrying, feeling irritable, and feeling sad) were included.
Selection of a Fatigue Outcome Variable To choose a fatigue outcome variable, we performed a confirmatory factor analysis with the 13-fatigue items of FACT-F, lack of energy from MSAS-SF, and BFI usual fatigue severity. All the items loaded on one factor which accounted for 91% of the total variance. Given these results, we selected the BFI usual fatigue severity because it is easy to use in the clinical setting and a clinically relevant level has been defined and reported by our group.19 In this study, we define the patient with usual fatigue ⱖ 3/10 as the patient with clinically significant fatigue, and refer to this as fatigue in the rest of the report.
Statistical Analysis To determine predictors of clinically significant fatigue, screening tests and exploratory analyses were performed to identify potentially important predictors. The laboratory values and continuous variables (Zung depression variable, 24 hours analgesic dose, and hours of sleep) were converted to binary variables by dichotomizing each variable around its median value. Fisher’s exact tests were used to assess the association of each variable with fatigue in univariate analyses. Variables with significant association (P ⬍ 0.05) were then selected to form unidimensional multiple stepwise logistic regression models. Predictors of significant fatigue within each dimension were then combined in multiple logistic regression analyses to identify multidimensional independent predictors of fatigue. A forward stepwise procedure was used to select the independent predictors of clinically significant fatigue (P ⬍ 0.05). Calibration and discrimination of the models was assessed.22 Calibration evaluates the degree of correspondence between the probabilities of having significant fatigue and the actual fatigue experience. Calibration was assessed using the HosmerLemeshow goodness-of-fit test (Chi Square statistic). Discrimination was analyzed using the area under the receiver operating characteristic
608
Hwang et al.
(ROC) curve, which represents for all possible pairs of patients, the proportion at which the patient who experienced significant fatigue had a higher probability of having significant fatigue than the patient who did not experience it. Analyses were performed with the STATA program v 6.0 (College Station, TX).
Vol. 26 No. 1 July 2003
mean usual fatigue severity was 2.6 (range 0–10) and 80 patients (44.4%) had BFI usual fatigue severity ⭓ 3/10.
Univariate Analysis and Unidimensional Fatigue Predictors Results of the univariate analysis are summarized in Table 3 and the unidimensional multiple stepwise logistic regression analyses are summarized in Table 4.
Results Demographics Patient characteristics have been reported in detail elsewhere13 and are summarized in Table 2. There were 74 outpatients and 106 inpatients. The median age was 68 years (30–89 years) and the median educational level was 12th grade (range, grades 4–16). At the time of interview, 79 patients (44%) were receiving analgesics for pain control, with a median dose of 60 (range 10–3,840 mg) oral morphine equivalent mg daily. The median hours of sleep was 7 (range 0–20 hours). Primary cancer sites were genitourinary (65, 36%) (prostate [62], bladder [1], renal cell [1], testicular [1]), lung (40, 22%), hematologic (35, 19%), gastrointestinal (30, 17%) (colorectal [27], pancreatic [2], esophagus [1]), and head and neck (10, 6%).
Situational Dimension. For the situational dimension, the inpatient status, analgesic use, 24 hours analgesic dose, and stage of disease correlated significantly with clinically significant fatigue level. In a multiple regression model with the above four variables, analgesic use was the only independent predictor (odds ratio 3.39 P ⬍ 0.001). Biomedical Dimension. The important laboratory correlates of fatigue included hemoglobin and sodium by univariate analysis. In a multiple regression model, both serum sodium (odds ratio 2.29, P ⫽ 0.008) and hemoglobin (odds ratio 2.04, P ⫽ 0.02) predicted clinically significant fatigue independently.
Fatigue Measures Physical Symptom Dimension. The presence of all of the selected nine physical symptoms
Fatigue was present in 113 patients (63%), as measured by BFI usual fatigue severity. The
Table 2 Patient Characteristics (Total n ⫽ 180)
All Patients n ⫽ 180 Patient Characteristics Inpatient status Inpatient Outpatient Live-in caregivers Active/history of alcohol use Active/history of drug use Stage of disease No evidence of disease Localized disease Regionally advanced disease Metastatic disease Cancer treatment Chemotherapy Radiotherapy Combined radiotherapy/ Chemotherapy Hormonal therapy No active treatment
Patients with Clinically Significant Fatigue (BFI usual fatigue ⱖ 3/10) n ⫽ 80
Patients without Clinically Significant Fatigue (BFI usual fatigue ⬍ 3/10) n ⫽ 100
n
%
n
%
n
%
74 106 118 82 21
41 59 66 46 12
39 41 52 37 11
49 51 66 46 14
35 65 66 45 10
35 65 66 45 10
13 9 38 120
7 5 21 67
6 2 9 63
7 3 11 79
7 7 29 57
7 7 29 57
31 18 7
17 10 4
13 9 2
16 11 3
18 9 5
18 9 5
40 84
22 47
16 38
20 48
24 46
24 46
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
609
Table 3 Univariate Analysis for Clinically Significant Fatigue Level (BFI usual fatigue ⱖ 3/10) in 180 Cancer Patients
Variables within Each Dimension Situational dimension Analgesic use Inpatient status Active cancer treatment Live-in caregiver 24 hour analgesic dose ⬎ 35 mg Hours of sleep ⬍ 7 Age ⬎ 68 years Stage of disease (See Table 2) Biomedical dimension Hemoglobin ⱕ 12. 5 gm/dl Sodium ⱕ 137 meq/dl Albumin ⱕ 3.5 g/dl Creatinine ⬎ 1.0 mg/dl Calcium ⬎ 8.9 mg/dl Potassium ⬍ 4.4 meq/l LDH ⬎ 1888 IU/l SGOT ⬎ 25 U/l SGPT ⬎ 25 Bilirubin ⬎ 0.5 mg/dl WBC ⬎ 6.7 K/cmm Physical symptom dimension Pain Dry mouth Dyspnea Feeling drowsy Weight loss Lack of appetite Difficulty sleeping Constipation Difficulty concentrating Psychological symptom dimension Depression (Zung SDS indices) ⬎ 0.45 Worrying Feeling irritable Feeling sad Feeling nervous
All the Patients n ⫽ 180
Patients with Clinically Significant Fatigue (BFT usual fatigue ⱖ 3/10) n ⫽ 80
Patients without Clinically Significant Fatigue (BFT usual fatigue ⬍ 3/10) n ⫽ 100
Fisher’s Exact Test P-value
n
%
n
%
n
%
79 74 96 118 62 75 90
44 41 53 66 34 42 50
48 39 42 52 38 32 39
61 49 52 66 48 40 49
31 35 54 66 24 43 51
31 35 54 66 24 43 51
⬍0.001 0.04 0.04 1.00 0.001 0.76 0.88 0.007
90 86 50 80 87 76 88 82 83 86 85
50 48 28 44 48 42 49 46 46 48 47
48 48 26 38 35 32 38 37 33 39 39
60 60 33 48 44 40 48 46 41 49 49
42 38 25 42 53 44 50 45 50 47 46
42 38 25 42 53 44 50 45 50 47 46
0.01 0.004 0.39 0.36 0.23 0.76 0.88 0.88 0.29 0.76 0.65
113 93 88 80 77 67 66 59 50
63 51 49 44 43 37 36 33 28
67 55 55 57 43 48 41 37 32
84 69 69 71 54 60 51 46 40
46 38 33 23 34 19 25 22 18
46 38 33 23 34 19 25 22 18
⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 0.001 0.01 ⬍0.0001 0.001 0.001
87 61 49 47 50
48 34 37 26 28
57 38 36 35 34
71 48 45 44 43
30 23 13 12 16
30 23 13 12 16
⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001
Continuous variables were dichotomized using their values as cutoff.
(pain, lack of appetite, dyspnea, feeling drowsy, weight loss, dry mouth, constipation, difficulty concentrating, and difficulty sleeping) was significantly associated with clinically significant fatigue by Fisher’s exact test (see Table 3). In a multiple regression model with these nine physical symptoms, the presence of feeling drowsy, pain, lack of appetite, and dyspnea independently predicted fatigue (odds ratio 4.25, 3.05, 2.94, 2.20; P ⬍ 0.001 and P ⫽ 0.009, 0.007, 0.04, respectively). Psychological Symptom Dimension. Depression, worrying, feeling irritable, feeling sad, and feeling nervous all correlated significantly with
fatigue by Fisher’s exact test. In a multiple logistic regression model with the above five psychological symptoms, feeling sad and feeling irritable (odds ratio 6.36, 2.63; P ⫽ 0.002, 0.03 respectively) were independent predictors of clinically significant fatigue (Table 4).
Multidimensional Multiple Logistic Regression Analysis By combining all the important variables identified by unidimensional analysis, the final multidimensional logistic regression model included nine variables. These variables were: using analgesics, hemoglobin, serum sodium,
610
Hwang et al.
Table 4 Independent Predictors by Unidimensional Multiple Logistic Regression Analyses Odds Ratio
P-value
Situational dimension Analgesic use 3.39 ⬍0.001 Biomedical dimension Sodium (ⱕ137 vs. 2.29 0.008 ⬎137 meq/dL) Hemoglobin (ⱕ12.5 2.04 0.02 vs. ⬎12.5 g/dL) Physical symptom dimension Feeling drowsy 4.25 ⬍0.001 Pain 3.05 0.009 Lack of appetite 2.94 0.007 Dyspnea 2.20 0.04 Psychological symptom dimension Feeling sad 6.36 0.002 Feeling irritable 2.63 0.03
95% CI 1.80–6.18 1.24–4.24 1.10–3.78 1.99–9.11 1.32–7.07 1.34–6.47 1.00–4.73 1.96–20.7 1.06–6.32
95% CI ⫽ 95% confidence interval. Continuous variables were dichotomized using their median values as cutoff.
pain, feeling drowsy, dyspnea, lack of appetite, feeling sad, and feeling irritable. The results of the multiple logistic regression analysis are summarized in Table 5. Multidimensional independent predictors of fatigue were feeling sad, feeling drowsy, pain, lack of appetite, feeling irritable, and dyspnea (odds ratio 3.96, 3.65, 3.48, 3.19, 2.91, 2.28; P ⫽ 0.006, 0.002, 0.007, 0.008, 0.02, 0.04 respectively). The HosmerLemeshow Chi-square statistic showed a good calibration of the model, Chi-square ⫽ 5.73, P ⫽ 0.68. Discrimination was high with area under the ROC equal to 0.88 (Table 5).
Discussion In this article, we used multiple logistic regression to identify the independent predictors of clinically significant fatigue, defined as usual fatigue ⱖ 3/10, in male cancer patients seen at a VA Medical Center. Many of the independent predictors of clinically significant fatigue identified within each dimension have been previously reported, such as analgesic use in the situational dimension;23 anemia in the biomedical dimension; feeling sad and irritable in the psychological symptom dimension; and pain, lack of appetite, and dyspnea in the physical symptom dimension. The finding that only physical and psychological symptoms can predict fatigue independently in the multidimensional model provides a new perspective on
Vol. 26 No. 1 July 2003
fatigue assessment and highlights the importance of systematic symptom assessment and management. The results from the unidimensional and multidimensional logistic analysis are illustrated in Figure 2. The inpatient status, stage of disease, use of analgesics, and 24-hour analgesic dosage all demonstrated significant association with fatigue univariately in the situational dimension. However, of these variables, only the use of analgesics predicted fatigue independently. Stone et al.23 reported that analgesic use, along with QOL, depression, dyspnea, and weight loss/ anorexia, independently predicted FACT-F score. Blesch et al.24 reported that for patients with cancer pain, the requirement of analgesics for pain control is an important contributor to fatigue. Klepstad et al.25 observed that patients with moderate to severe pain who were receiving morphine therapy reported more fatigue. Some have interpreted these results to imply that analgesics cause fatigue. Our results suggest pain is an independent predictor in both the physical symptom dimension and multidimensional models, and that pain supersedes analgesic use as a predictor of fatigue. We conclude that analgesic use can be considered a marker for the presence of pain. In addition, we have also reported that cancer patients with moderate intensity of pain had a 2.3 times higher relative risk of fatigue than patients without moderately intense pain.26 This may explain why the use of analgesic is no longer important Table 5 Multidimensional Multiple Stepwise Logistic Regression Analysis Predicting the Presence of Fatigue Defined as Usual Fatigue Severity ⱖ 3/10 Hosmer-Lemeshow Chi-square ⫽ 5.73, P ⫽ 0.68 Area under receiver operated characteristic curve (ROC): 0.88 Variables Feeling sad Feeling drowsy Pain Lack of appetite Feeling irritable Dyspnea
Odds Ratio
P-value
95% CI
3.96 3.65 3.48 3.19 2.91 2.28
0.006 0.002 0.007 0.008 0.02 0.04
1.48–10.58 1.61–8.25 1.39–8.69 1.36–7.50 1.16–7.33 1.02–5.08
95% CI ⫽ 95% confidence interval. Nine variables were included in the final multidimensional model: hemoglobin, sodium, using analgesics, feeling sad, feeling irritable, pain, dyspnea, feeling drowsy and lack of appetite. Forward stepwise approach was used. Continuous variables were dichotomized using their median values as cutoff.
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
in the multidimensional model, and shows the importance of multidimensional analyses. In the biomedical dimension, the importance of anemia to fatigue is well known and probably is the factor most amenable to treatment. It has been proposed that a hemoglobin of 12 gm/dl or greater is necessary to minimize anemia-related fatigue and optimize QOL in patients receiving chemotherapy.27,28 However, there is lack of strong evidence in patients with advanced cancer to support the correlation between anemia and fatigue.29,30 Gleeson and Spencer31 reported a small but significant improvement in weakness among patients who received blood transfusion in the palliative care setting. Stone et al.8 suggested that fatigue might be related more to the changes in hemoglobin level than to the hemoglobin level itself. Our results suggest that anemia’s role in fatigue in our population is limited, and evident in
611
the biomedical dimension only. Further studies are needed to explore the role of blood transfusion or erythropoeitin in palliating cancer-related fatigue. The serum sodium level is another independent predictor of clinically significant fatigue in the situational dimension. This finding has not been reported previously and requires further examination. The most profound findings in our study are that the physical symptoms (pain, lack of appetite, feeling drowsy, and dyspnea) and psychological symptoms (feeling sad and feeling irritable) not only predict clinically significant fatigue unidimensionally, but are the only independent predictors in the multidimensional model. The results are consistent with earlier reports that fatigue is related to various number of symptoms5–10 and support a symptom oriented approach to conceptualize the assessment and management of cancer-related
Fig. 2. Multidimensional independent predictors of fatigue.
612
Hwang et al.
fatigue. Interestingly, feeling drowsy is the only symptom that was not reported previously; it might be simply because the item feeling drowsy has not been included in previous studies. We were surprised that the items of feeling sad and feeling irritable, rather than the Zung depression score, were identified as independent predictors of clinically significant fatigue in the psychological symptom dimension. It has been suggested that by asking a single depression question such as “Are you depressed?” is a valid approach to screen for patients with depression.32 The significant association between psychological symptoms, and depression and fatigue is also well recognized.5,33,34 A few studies have examined the effect of symptom management on fatigue. In an ongoing cancer pain study conducted by our group, 117 patients with worst cancer pain ⱖ 4/10 were interviewed at baseline and 1 week after cancer pain management. There was a modest correlation between changes in worst pain severity and changes in fatigue severity, and patients who had ⱖ 2 points decrease in worst pain severity had significantly less fatigue.35 Appetite stimulants such as megestrol acetate36,37 or corticosteroids38 also reduce fatigue. Dyspnea is another significant fatigue predictor5,8 but studies to show the effect of dyspnea interventions on fatigue are lacking. As depression is strongly associated with fatigue, the effect of antidepressants on fatigue are of interest. Inconsistent results on the effectiveness of depression management for cancer treatment-related fatigue have been reported. Weitzner et al. reported that antidepressants can be helpful in the treatment of hot flashes and associated fatigue, sleep disturbance, and depression in women with breast cancer treated with chemotherapy.39 However, in a doubleblinded placebo controlled randomized trial of 738 patients who reported fatigue with chemotherapy and were randomized to receive either 20 mg of the selective serotonin re-uptake inhibitor (SSRI) paroxetine or placebo, the paroxetine group had significantly reduced depression during chemotherapy but did not experience a significant effect on fatigue.40 The limited effectiveness of symptom interventions to date on fatigue in advanced cancer patients could result from a variety of different reasons. Patients with cancer have multiple symptoms and as fatigue is a common
Vol. 26 No. 1 July 2003
result of any given severe symptom, single symptom interventions may not be effective. Second, most of the symptom interventions are only modestly effective and may not have a large impact on fatigue. Third, if a common pathophysiology underlies symptoms, fatigue, and disease progression, symptom intervention alone may be inadequate. Finally, different symptoms may cause fatigue by different mechanisms. Fatigue related to dyspnea may be different from fatigue related to pain. How these possibilities apply to a patient with fatigue may depend on the clinical circumstances and more research is needed to study the relationship between symptoms and fatigue. There are some limitations in our study. The study was conducted at a VA Medical Center and the results may not be generalizable to all advanced cancer patients. This was a secondary analysis, and further studies including both sexes in other settings are needed to confirm these findings. In summary, we have identified the multidimensional independent predictors of fatigue and suggested the importance of symptom status to fatigue. The presence of fatigue should trigger a wide-ranging review of symptoms to further evaluate the patient’s particular complaint.10,41 The information obtained from this study can be applied easily in clinical settings to help clinicians to identify patients at risk for fatigue within each dimension and to direct questions relevant to fatigue with their patients. Although there are no direct treatments for fatigue, these results support identifying the factors or symptoms amenable to treatment as a step towards reducing fatigue.
Acknowledgments Ms. Michelle Rindos helped with interviewing the patients.
References 1. Cella D, Peterman A, Passik S, et al. Progress toward guidelines for the management of fatigue. Oncology 1998;12:1–9. 2. Aistars J. Fatigue in the cancer patients: a conceptual approach to a clinical problems. Oncol Nurs Forum 1987;14:25–30.
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
613
3. Portenoy RK, Itri LM. Cancer-related fatigue: guidelines for evaluation and management. Oncologist 1999;4:1–10.
18. Chang VT, Hwang SS, Feuerman M, et al. The Memorial Symptom Assessment Scale Short Form: Validity and reliability. Cancer 2000;89:1162–1171.
4. Portenoy RK, Miaskowski C. Assessment and management of cancer-related fatigue. Berger A, Portenoy RK, Weissman DE (eds). Principles and practice of supportive oncology. Philadelphia: Lippincott-Raven, 1998:109–118.
19. Karnofsky DA, Burchenal JH. The clinical evaluation of chemotherapeutic agents in cancer. In: MacLeod CM (ed). Evaluation of chemotherapeutic agents. New York: Columbia University Press, 1949: 191–205.
5. Okuyama T, Akechi T, Kugaya A, et al. Factors correlated with fatigue in disease-free breast cancer patients: application of the Cancer Fatigue Scale. Support Care Cancer 2000;8:215–222.
20. Hwang SS, Chang VT, Cogswell C, et al. Clinical relevance of fatigue levels in cancer patients at a Veteran Administration Medical Center. Cancer 2002;94:2481–2489.
6. Broeckel JA, Jacobsen PB, Horton J, et al. Characteristics and correlates of fatigue after adjuvant chemotherapy for breast cancer. J Clin Oncol 1998;16:1689–1696.
21. Cella DF, Tulsky DS, Gray G, et al. The Functional Assessment of Cancer Therapy Scale: development and validation of the general measure. J Clin Oncol 1993;11:570–579.
7. Irvine DM, Vincent L, Graydon JE, et al. Fatigue in women with breast cancer receiving radiation therapy. Cancer Nurs 1998;21:127–135.
22. Hosmer DW, Lemeshow S. Applied logistic regression. Second edition. New York: Wiley, 2000.
8. Stone P, Richards M, A’Hern R, et al. A study to investigate the prevalence, severity and correlates of fatigue among patients with cancer in comparison with a control group of volunteers without cancer. Ann Oncol 2000;11:561–567. 9. Irvine D, Vincent L, Graydon JE, et al. The prevalence and correlates of fatigue in patients receiving treatment with chemotherapy and radiotherapy. A comparison with the fatigue experienced by healthy individuals. Cancer Nurs 1994;17:367–378. 10. Smets EM, Visser MR, Willems-Groot AF, et al. Fatigue and radiotherapy: (A) experience in patients undergoing treatment. Br J Cancer 1998;78:899–906. 11. Pater JL, Zee B, Palmer M, et al. Fatigue in patients with cancer: results with National Cancer Institute of Canada Clinical Trials Group studies employing the EORTC QLQ-C30. Support Care Cancer 1997;5:410–413. 12. Stone P, Hardy J, Broadley K, et al. Fatigue in advanced cancer: a prospective controlled crosssectional study. Br J Cancer 1999;79:1479–1486. 13. Hwang SS, Chang VT, Kasimis B. Comparison of three fatigue measures in a veterans cancer population. Cancer Invest 2003;21:355–365. 14. Mendoza TR, Wang XS, Cleeland CS, et al. The rapid assessment of fatigue severity in cancer patients. Cancer 1999;85:1186–1196. 15. Cella D. Factors influencing quality of life in cancer patients: anemia and fatigue. Semin Oncol 1998;25(Suppl 7):43–46. 16. Yellen SB, Cella SF, Webster K, et al. Measuring fatigue and other anemia-related symptoms with the functional assessment of cancer therapy (FACT) measurement system. J Pain Symptom Manage 1997;13:63–74. 17. Zung WW. A self-rating depression scale. Arch General Psychiatry 1965;12:63–70.
23. Stone P, Richardson A, Ream E, et al. Cancerrelated fatigue: inevitable, unimportant and untreatable? Results of a multi-center patient survey. Cancer Fatigue Forum. Ann Oncol 2000;11:971–975. 24. Blesch KS, Paice JA, Wickham R, et al. Correlates of fatigue in people with breast and lung cancer. Oncol Nurs Forum 1991;18:81–87. 25. Klepstad P, Borchgrevink PC, Kaasa S. Effects on cancer patients’ health-related quality of life after the start of morphine therapy. J Pain Symptom Manage 2000;20:19–26. 26. Chang VT, Hwang SS, Feuerman M, et al. Symptom and quality-of-life survey of medical oncology patients at a veterans affairs medical center: a role for symptom assessment. Cancer 2000;88:1175–1183. 27. Sabbatini P. The relationship between anemia and QOL in cancer patients. Oncologist 2000;5: 19–23. 28. Cleeland CS, Demetri GD, Glaspy J, et al. Identifying hemoglobin level for optimal QOL: results of an incremental analysis. Proc ASCO 1999;18:574a, Abstract 2215. 29. Bruera E, Brenneis C, Michaud M, et al. Association between asthenia and nutritional status, lean body mass, anemia, psychological status, and tumor mass in patients with advanced breast cancer. J Pain Symptom Manage 1989;4:59–63. 30. Morant R. Asthenia: An important symptom in cancer patients. Cancer Treat Rev 1996;22(Suppl A): 117–121. 31. Gleeson G, Spencer D. Blood transfusion and its benefit in palliative care. Palliat Med 1995;9:307–313. 32. Chochinov HM, Wilson KG, Enns M, et al. “Are you depressed?” Screening for depression in the terminally ill. Am J Psychiatry 1997;154:674–676. 33. Smets EM, Garssen B, Cull A, et al. Application of the multidimensional fatigue inventory (MFI-20)
614
Hwang et al.
Vol. 26 No. 1 July 2003
in cancer patients receiving radiotherapy. Br J Cancer 1996;73:241–245.
blind, placebo-controlled study. J Am Geriatr Soc 2000;48:485–492.
34. Dimeo F, Stieglitz RD, Novelli-Fischer U, et al. Correlation between physical performance and fatigue in cancer patients. Ann Oncol 1997;8: 1251–1255.
38. Bruera E, Roca E, Cedaro L, et al. Action of oral methylprednisolone in terminal cancer patients: a prospective randomized double-blind study. Cancer Treat Rep 1985;69:751–754.
35. Hwang S, Chang V, Alejandro Y, et al. Cancer pain management and its impact on fatigue, psychological distress and depression in male cancer patients at a VA medical center. J Pain 2002;3 (2 Suppl 1):44, Abstract 774.
39. Weitzner MA, Moncello J, Jacobsen PB, et al. A pilot trial of paroxetine for the treatment of hot flashes and associated symptoms in women with breast cancer. J Pain Symptom Manage 2002;23: 337–345.
36. Bruera E, Ernst S, Hagen N, et al. Effectiveness of megestrol acetate in patients with advanced cancer: a randomized, double-blind, crossover study. Cancer Prev Control 1998;2:74–78.
40. Morrow GR, Hickok JT, Raubertas RF, et al. Effect of an SSRI antidepressant on fatigue and depression in seven hundred thirty-eight cancer patients treated with chemotherapy: a URCC CCOP study. ASCO 2001;20: Abstract 1531.
37. Yeh SS, Wu SY, Lee TP, et al. Improvement in quality-of-life measures and stimulation of weight gain after treatment with megestrol acetate oral suspension in geriatric cachexia: results of a double-
41. Richardson A, Ream E. The experience of fatigue and other symptoms in patients receiving chemotherapy. Eur J Cancer Care 1996;5(Suppl):24–30.
Vol. 26 No. 1 July 2003
Original Article
Multidimensional Independent Predictors of Cancer-Related Fatigue Shirley S. Hwang, RN, MS, Victor T. Chang, MD, Montse Rue, PhD, and Basil Kasimis, MD Section of Hematology/Oncology (S.S.H., V.T.C., B.K.) and Patient Care Services (S.S.H.), VA New Jersey Health Care System, East Orange, New Jersey; UMDNJ/School of Nursing (S.S.H.) and UMDNJ/New Jersey Medical School (V.T.C., B.K.), Newark, New Jersey; and Department of Biostatistical Science (M.R.), Dana-Farber Cancer Institute, Boston, Massachusetts, USA
Abstract The purpose of this study was to identify independent predictors of clinically significant fatigue based upon a multidimensional model. A total of 180 cancer patients completed the Brief Fatigue Inventory (BFI), Functional Assessment of Cancer Therapy-Fatigue (FACTF), Memorial Symptom Assessment Scale Short Form (MSAS-SF), and the Zung Self-Rating Depression Scale (SDS). Additional data included Karnofsky Performance Status (KPS) score, laboratory tests, and demographic information. The BFI usual fatigue severity ⱖ 3/10 was defined as clinically significant fatigue. Possible independent variables were identified from a biopsychosocial model of fatigue. Fisher’s exact test was used to univariately assess the association of each variable with clinically significant fatigue. Multiple logistic regression analyses were used to identify independent predictors of fatigue within each dimension, and then across dimensions. Fatigue was present in 113 (62%) patients, and 80 (44.4%) patients had usual fatigue ⱖ 3/10. The unidimensional independent predictors were use of analgesics (situation dimension); hemoglobin and serum sodium (biomedical dimension); feeling drowsy, dyspnea, pain and lack of appetite (physical symptom dimension); and feeling sad and feeling irritable (psychological symptom dimension). In a multidimensional model, dyspnea, pain, lack of appetite, feeling drowsy, feeling sad, and feeling irritable predicted fatigue independently with good calibration (Hosmer Lemeshow Chi Square ⫽ 5.73, P ⫽ 0.68) and discrimination (area under the receiver operating characteristic curve ⫽ 0.88). Physical and psychological symptoms predict fatigue independently in the multidimensional model, and superseded laboratory data. These findings support a symptom-oriented approach to assessment of cancer-related fatigue. J Pain Symptom Manage 2003;26:604–614. 쑖 2003 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.
Address reprint requests to: Shirley S. Hwang, RN, MS, Section Hematology/Oncology (111), VA New Jersey Health Care System at East Orange, 385 Tremont Avenue, East Orange, NJ 07018, USA. The views expressed herein do not necessarily reflect the views of the Department of Veterans Affairs or of the U.S. Government. Accepted for publication: November 8, 2002.
쑖 2003 U.S. Cancer Pain Relief Committee Published by Elsevier Inc. All rights reserved.
Preliminary results presented at the Annual Scientific Meeting of the American Society of Clinical Oncology at Los Angeles, California (Proc ASCO 1999;18:594a, abstract 2294).
0885-3924/03/$–see front matter doi:10.1016/S0885-3924(03)00218-5
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
605
Key Words Cancer, fatigue, predictors, analgesics, anemia, pain, feeling drowsy, dyspnea, lack of appetite, depression, veterans, sad
Introduction Fatigue is a highly prevalent symptom in cancer patients, and interest in cancer-related fatigue has increased significantly in the last few years. Fatigue can result from the prolonged stress caused by multiple factors1,2 and it may represent a final common pathway from these factors.3 Predisposing factors include demographic characteristics, underlying disease, treatment received, comorbidities, sleep disorders, immobility, and psychosocial factors.4 Depending on the study population, variables used, and fatigue measures, different independent predictors of fatigue have been reported in the literature. These include depression,5 psychological distress,6–8 physical symptoms,7,9 dyspnea,5,8 anxiety,8 poor quality of sleep,5–7,10 sex,11,12 age,11 primary cancer site,11,12 cancer treatment,10 stage of disease,11 and laboratory values such as low albumin,13,14 and anemia.13,15 The relative importance of these predictors is not known. In order to better understand the fatigue characteristics and to define clinically significant fatigue level in male veteran cancer patients, we performed a survey study using the Brief Fatigue Inventory (BFI),14 Functional Assessment of Cancer Therapy—Fatigue (FACT-F),16 the Zung Self-rating Depression Scale (SDS),17 the Memorial Symptom Assessment Scale Short Form (MSAS-SF),18 and demographic data. We also obtained the complete blood counts and chemistry profiles after the completion of the study survey. We validated both the FACT-F and BFI in our population, and found that there were significant correlations between the BFI fatigue severity, FACT-F fatigue measure, and “lack of energy” item from MSAS-SF, with correlation coefficients ranging from 0.68 to 0.88. Higher scores on the fatigue measures were associated with lower Karnofsky Performance Status (KPS)19and QOL, higher symptom distress, and greater depression scores.13 We also identified clinically relevant fatigue levels by examining the association between BFI fatigue severity and broader QOL con-
structs, such as symptom distress parameters, depression, QOL, and functional status. In summary, patients with a usual fatigue severity ⱖ 3/10 demonstrated significantly higher functional interference scores, lower QOL, and greater symptom distress.20 In this report, we performed a secondary exploratory analysis to identify the multidimensional independent predictors of clinically significant fatigue.
Methods Theoretical Model We used a biopsychosocial model to assess the patient and as a source of predictor variables for clinically significant fatigue. The predictor variables were divided into four dimensions (situational, biological, physical symptoms and psychological symptoms) and are summarized in Figure 1 and Table 1.
Patient Selection In this prospective cross-sectional design survey study, 74 outpatients and 106 inpatients diagnosed with cancer and treated at the Hematology/Oncology Section at VA New Jersey Health Care System at East Orange, NJ (VANJHCS) were recruited from September 1997 to May 1998. The VANJHCS is the sole tertiary care teaching hospital providing Hematology/Oncology services for veterans residing in the state of New Jersey. The study was approved by the Institutional Review Board, and all patients signed informed consent before participating. Each patient completed the surveys listed below and had blood drawn for a biochemical profile and complete blood count after completion of survey instruments. Demographic data were also determined.
Instruments The MSAS-SF is a validated, patient-rated instrument in which patients rate symptom distress for 28 highly prevalent physical symptoms and symptom frequency for four psychological symptoms. Each symptom is scored from
606
Hwang et al.
Vol. 26 No. 1 July 2003
Fig. 1. Proposed multidimensional conceptual model to predict fatigue.
0 to 4 ranging from “no symptom” to “very much.” For this report, we assessed the presence of nine physical symptoms (pain, difficulty concentrating, feeling drowsy, lack of appetite, weight loss, dyspnea, difficulty sleeping, constipation, and dry mouth) and the presence of
the four psychological symptoms (feeling sad, feeling nervous, feeling irritable, and worry) as predictors of clinically significant fatigue. The FACT-F16 is based on the 28- item FACT21 G —patient-rated measure of quality of life for cancer patients with any tumor type—with an
Table 1 Multidimensional Assessment to Identify Fatigue Predictors and Outcome Variables
Assessment Dimensions
Measurement/Instruments
Variables
Situational
Demographic data
Biomedical
Laboratory tests
Primary cancer sites Stage of disease Inpatient status Analgesic use 24-hour analgesic dosage Caregiver status Active cancer treatment Hours of sleep Substance abuse Age Chemistry profiles: albumin, BUN, creatinine, calcium, sodium, potassium, LDH, SGOT, SGPT, bilirubin. blood counts: WBC, Hgb Pain, difficulty concentrating, dry mouth, feeling drowsy, difficulty sleeping, dyspnea, sweats, lack of appetite, difficulty swallowing, weight loss, constipation Feeling sad, feeling nervous, worrying, feeling irritable and depression
Physical symptom
MSAS-SF
Psychological symptom
MSAS-SF Zung SDS
Significant Variables by Univariate Analysis Formed the Unidimensional Model Stage of disease Inpatient status Analgesic use 24-hour analgesic dosage
Hemoglobin level Serum sodium
Pain, difficulty concentrating, dry mouth, feeling drowsy, difficulty sleeping, dyspnea, lack of appetite, weight loss, constipation Feeling sad, feeling nervous, worrying, feeling irritable and depression
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
additional 13 items designed to measure fatigue. The FACT-F fatigue subscale is a multidimensional fatigue assessment instrument, and measures multiple fatigue characteristics and their impact on function. The BFI14 consists of two components: fatigue severity and fatigue interference. Both components are assessed using a 0–10 numeric scale. The fatigue severity assesses fatigue at its worst, usual, and right now; the fatigue interference component assesses fatigue-related functional interference in the areas of daily activity, mood, walking, work, enjoyment of life, and relationship with others. The total BFI fatigue interference is the sum of the six interference scores. The Zung SDS17 is a validated and simple instrument to assess depression. It uses a 1–4 numeric rating to assess 20 questions with a maximum possible raw score of 80. The SDS indices were derived by dividing the sum of the raw scores of the 20 items by 80 and expressing the result as a decimal. The mean value for the normal control subjects and depressive patients were established. There are three items in the Zung SDS which reflect the measurement of somatic symptoms possibly related to fatigue: “I eat as much as I used to,” “I notice that I am losing weight,” and “I get tired for no reason.” We reconstructed the depression indices by removing these three items from the calculation. The Cronbach alpha coefficient for the reconstructed 17-item depression indices was 0.84. This approach has previously been used by other fatigue researchers.8,10
Definition of Independent Variables and Modeling Rationale for Each Dimension The independent variables used in each dimension are summarized in Table 1. In the situational dimension, clinical variables related to cancer diagnosis, treatment (primary sites, stage of disease, inpatient status, treatment received, use of analgesics, 24-hour analgesics dosage, age), and environmental factors (hours of sleep, substance abuse, and caregiver status) were included. In the biomedical dimension, variables from laboratory tests were examined. In the physical symptom dimension, six frequently reported physical symptoms related to fatigue (pain, difficulty concentrating, dyspnea, lack of appetite, weight loss, and difficulty sleeping) and three other highly prevalent physical
607
symptoms (feeling drowsy, dry mouth, and constipation) from the MSAS-SF were selected. In the psychological symptom dimension, the Zung scale for depression, and four other psychological symptoms measured by MSAS-SF (feeling nervous, worrying, feeling irritable, and feeling sad) were included.
Selection of a Fatigue Outcome Variable To choose a fatigue outcome variable, we performed a confirmatory factor analysis with the 13-fatigue items of FACT-F, lack of energy from MSAS-SF, and BFI usual fatigue severity. All the items loaded on one factor which accounted for 91% of the total variance. Given these results, we selected the BFI usual fatigue severity because it is easy to use in the clinical setting and a clinically relevant level has been defined and reported by our group.19 In this study, we define the patient with usual fatigue ⱖ 3/10 as the patient with clinically significant fatigue, and refer to this as fatigue in the rest of the report.
Statistical Analysis To determine predictors of clinically significant fatigue, screening tests and exploratory analyses were performed to identify potentially important predictors. The laboratory values and continuous variables (Zung depression variable, 24 hours analgesic dose, and hours of sleep) were converted to binary variables by dichotomizing each variable around its median value. Fisher’s exact tests were used to assess the association of each variable with fatigue in univariate analyses. Variables with significant association (P ⬍ 0.05) were then selected to form unidimensional multiple stepwise logistic regression models. Predictors of significant fatigue within each dimension were then combined in multiple logistic regression analyses to identify multidimensional independent predictors of fatigue. A forward stepwise procedure was used to select the independent predictors of clinically significant fatigue (P ⬍ 0.05). Calibration and discrimination of the models was assessed.22 Calibration evaluates the degree of correspondence between the probabilities of having significant fatigue and the actual fatigue experience. Calibration was assessed using the HosmerLemeshow goodness-of-fit test (Chi Square statistic). Discrimination was analyzed using the area under the receiver operating characteristic
608
Hwang et al.
(ROC) curve, which represents for all possible pairs of patients, the proportion at which the patient who experienced significant fatigue had a higher probability of having significant fatigue than the patient who did not experience it. Analyses were performed with the STATA program v 6.0 (College Station, TX).
Vol. 26 No. 1 July 2003
mean usual fatigue severity was 2.6 (range 0–10) and 80 patients (44.4%) had BFI usual fatigue severity ⭓ 3/10.
Univariate Analysis and Unidimensional Fatigue Predictors Results of the univariate analysis are summarized in Table 3 and the unidimensional multiple stepwise logistic regression analyses are summarized in Table 4.
Results Demographics Patient characteristics have been reported in detail elsewhere13 and are summarized in Table 2. There were 74 outpatients and 106 inpatients. The median age was 68 years (30–89 years) and the median educational level was 12th grade (range, grades 4–16). At the time of interview, 79 patients (44%) were receiving analgesics for pain control, with a median dose of 60 (range 10–3,840 mg) oral morphine equivalent mg daily. The median hours of sleep was 7 (range 0–20 hours). Primary cancer sites were genitourinary (65, 36%) (prostate [62], bladder [1], renal cell [1], testicular [1]), lung (40, 22%), hematologic (35, 19%), gastrointestinal (30, 17%) (colorectal [27], pancreatic [2], esophagus [1]), and head and neck (10, 6%).
Situational Dimension. For the situational dimension, the inpatient status, analgesic use, 24 hours analgesic dose, and stage of disease correlated significantly with clinically significant fatigue level. In a multiple regression model with the above four variables, analgesic use was the only independent predictor (odds ratio 3.39 P ⬍ 0.001). Biomedical Dimension. The important laboratory correlates of fatigue included hemoglobin and sodium by univariate analysis. In a multiple regression model, both serum sodium (odds ratio 2.29, P ⫽ 0.008) and hemoglobin (odds ratio 2.04, P ⫽ 0.02) predicted clinically significant fatigue independently.
Fatigue Measures Physical Symptom Dimension. The presence of all of the selected nine physical symptoms
Fatigue was present in 113 patients (63%), as measured by BFI usual fatigue severity. The
Table 2 Patient Characteristics (Total n ⫽ 180)
All Patients n ⫽ 180 Patient Characteristics Inpatient status Inpatient Outpatient Live-in caregivers Active/history of alcohol use Active/history of drug use Stage of disease No evidence of disease Localized disease Regionally advanced disease Metastatic disease Cancer treatment Chemotherapy Radiotherapy Combined radiotherapy/ Chemotherapy Hormonal therapy No active treatment
Patients with Clinically Significant Fatigue (BFI usual fatigue ⱖ 3/10) n ⫽ 80
Patients without Clinically Significant Fatigue (BFI usual fatigue ⬍ 3/10) n ⫽ 100
n
%
n
%
n
%
74 106 118 82 21
41 59 66 46 12
39 41 52 37 11
49 51 66 46 14
35 65 66 45 10
35 65 66 45 10
13 9 38 120
7 5 21 67
6 2 9 63
7 3 11 79
7 7 29 57
7 7 29 57
31 18 7
17 10 4
13 9 2
16 11 3
18 9 5
18 9 5
40 84
22 47
16 38
20 48
24 46
24 46
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
609
Table 3 Univariate Analysis for Clinically Significant Fatigue Level (BFI usual fatigue ⱖ 3/10) in 180 Cancer Patients
Variables within Each Dimension Situational dimension Analgesic use Inpatient status Active cancer treatment Live-in caregiver 24 hour analgesic dose ⬎ 35 mg Hours of sleep ⬍ 7 Age ⬎ 68 years Stage of disease (See Table 2) Biomedical dimension Hemoglobin ⱕ 12. 5 gm/dl Sodium ⱕ 137 meq/dl Albumin ⱕ 3.5 g/dl Creatinine ⬎ 1.0 mg/dl Calcium ⬎ 8.9 mg/dl Potassium ⬍ 4.4 meq/l LDH ⬎ 1888 IU/l SGOT ⬎ 25 U/l SGPT ⬎ 25 Bilirubin ⬎ 0.5 mg/dl WBC ⬎ 6.7 K/cmm Physical symptom dimension Pain Dry mouth Dyspnea Feeling drowsy Weight loss Lack of appetite Difficulty sleeping Constipation Difficulty concentrating Psychological symptom dimension Depression (Zung SDS indices) ⬎ 0.45 Worrying Feeling irritable Feeling sad Feeling nervous
All the Patients n ⫽ 180
Patients with Clinically Significant Fatigue (BFT usual fatigue ⱖ 3/10) n ⫽ 80
Patients without Clinically Significant Fatigue (BFT usual fatigue ⬍ 3/10) n ⫽ 100
Fisher’s Exact Test P-value
n
%
n
%
n
%
79 74 96 118 62 75 90
44 41 53 66 34 42 50
48 39 42 52 38 32 39
61 49 52 66 48 40 49
31 35 54 66 24 43 51
31 35 54 66 24 43 51
⬍0.001 0.04 0.04 1.00 0.001 0.76 0.88 0.007
90 86 50 80 87 76 88 82 83 86 85
50 48 28 44 48 42 49 46 46 48 47
48 48 26 38 35 32 38 37 33 39 39
60 60 33 48 44 40 48 46 41 49 49
42 38 25 42 53 44 50 45 50 47 46
42 38 25 42 53 44 50 45 50 47 46
0.01 0.004 0.39 0.36 0.23 0.76 0.88 0.88 0.29 0.76 0.65
113 93 88 80 77 67 66 59 50
63 51 49 44 43 37 36 33 28
67 55 55 57 43 48 41 37 32
84 69 69 71 54 60 51 46 40
46 38 33 23 34 19 25 22 18
46 38 33 23 34 19 25 22 18
⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 0.001 0.01 ⬍0.0001 0.001 0.001
87 61 49 47 50
48 34 37 26 28
57 38 36 35 34
71 48 45 44 43
30 23 13 12 16
30 23 13 12 16
⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001 ⬍0.0001
Continuous variables were dichotomized using their values as cutoff.
(pain, lack of appetite, dyspnea, feeling drowsy, weight loss, dry mouth, constipation, difficulty concentrating, and difficulty sleeping) was significantly associated with clinically significant fatigue by Fisher’s exact test (see Table 3). In a multiple regression model with these nine physical symptoms, the presence of feeling drowsy, pain, lack of appetite, and dyspnea independently predicted fatigue (odds ratio 4.25, 3.05, 2.94, 2.20; P ⬍ 0.001 and P ⫽ 0.009, 0.007, 0.04, respectively). Psychological Symptom Dimension. Depression, worrying, feeling irritable, feeling sad, and feeling nervous all correlated significantly with
fatigue by Fisher’s exact test. In a multiple logistic regression model with the above five psychological symptoms, feeling sad and feeling irritable (odds ratio 6.36, 2.63; P ⫽ 0.002, 0.03 respectively) were independent predictors of clinically significant fatigue (Table 4).
Multidimensional Multiple Logistic Regression Analysis By combining all the important variables identified by unidimensional analysis, the final multidimensional logistic regression model included nine variables. These variables were: using analgesics, hemoglobin, serum sodium,
610
Hwang et al.
Table 4 Independent Predictors by Unidimensional Multiple Logistic Regression Analyses Odds Ratio
P-value
Situational dimension Analgesic use 3.39 ⬍0.001 Biomedical dimension Sodium (ⱕ137 vs. 2.29 0.008 ⬎137 meq/dL) Hemoglobin (ⱕ12.5 2.04 0.02 vs. ⬎12.5 g/dL) Physical symptom dimension Feeling drowsy 4.25 ⬍0.001 Pain 3.05 0.009 Lack of appetite 2.94 0.007 Dyspnea 2.20 0.04 Psychological symptom dimension Feeling sad 6.36 0.002 Feeling irritable 2.63 0.03
95% CI 1.80–6.18 1.24–4.24 1.10–3.78 1.99–9.11 1.32–7.07 1.34–6.47 1.00–4.73 1.96–20.7 1.06–6.32
95% CI ⫽ 95% confidence interval. Continuous variables were dichotomized using their median values as cutoff.
pain, feeling drowsy, dyspnea, lack of appetite, feeling sad, and feeling irritable. The results of the multiple logistic regression analysis are summarized in Table 5. Multidimensional independent predictors of fatigue were feeling sad, feeling drowsy, pain, lack of appetite, feeling irritable, and dyspnea (odds ratio 3.96, 3.65, 3.48, 3.19, 2.91, 2.28; P ⫽ 0.006, 0.002, 0.007, 0.008, 0.02, 0.04 respectively). The HosmerLemeshow Chi-square statistic showed a good calibration of the model, Chi-square ⫽ 5.73, P ⫽ 0.68. Discrimination was high with area under the ROC equal to 0.88 (Table 5).
Discussion In this article, we used multiple logistic regression to identify the independent predictors of clinically significant fatigue, defined as usual fatigue ⱖ 3/10, in male cancer patients seen at a VA Medical Center. Many of the independent predictors of clinically significant fatigue identified within each dimension have been previously reported, such as analgesic use in the situational dimension;23 anemia in the biomedical dimension; feeling sad and irritable in the psychological symptom dimension; and pain, lack of appetite, and dyspnea in the physical symptom dimension. The finding that only physical and psychological symptoms can predict fatigue independently in the multidimensional model provides a new perspective on
Vol. 26 No. 1 July 2003
fatigue assessment and highlights the importance of systematic symptom assessment and management. The results from the unidimensional and multidimensional logistic analysis are illustrated in Figure 2. The inpatient status, stage of disease, use of analgesics, and 24-hour analgesic dosage all demonstrated significant association with fatigue univariately in the situational dimension. However, of these variables, only the use of analgesics predicted fatigue independently. Stone et al.23 reported that analgesic use, along with QOL, depression, dyspnea, and weight loss/ anorexia, independently predicted FACT-F score. Blesch et al.24 reported that for patients with cancer pain, the requirement of analgesics for pain control is an important contributor to fatigue. Klepstad et al.25 observed that patients with moderate to severe pain who were receiving morphine therapy reported more fatigue. Some have interpreted these results to imply that analgesics cause fatigue. Our results suggest pain is an independent predictor in both the physical symptom dimension and multidimensional models, and that pain supersedes analgesic use as a predictor of fatigue. We conclude that analgesic use can be considered a marker for the presence of pain. In addition, we have also reported that cancer patients with moderate intensity of pain had a 2.3 times higher relative risk of fatigue than patients without moderately intense pain.26 This may explain why the use of analgesic is no longer important Table 5 Multidimensional Multiple Stepwise Logistic Regression Analysis Predicting the Presence of Fatigue Defined as Usual Fatigue Severity ⱖ 3/10 Hosmer-Lemeshow Chi-square ⫽ 5.73, P ⫽ 0.68 Area under receiver operated characteristic curve (ROC): 0.88 Variables Feeling sad Feeling drowsy Pain Lack of appetite Feeling irritable Dyspnea
Odds Ratio
P-value
95% CI
3.96 3.65 3.48 3.19 2.91 2.28
0.006 0.002 0.007 0.008 0.02 0.04
1.48–10.58 1.61–8.25 1.39–8.69 1.36–7.50 1.16–7.33 1.02–5.08
95% CI ⫽ 95% confidence interval. Nine variables were included in the final multidimensional model: hemoglobin, sodium, using analgesics, feeling sad, feeling irritable, pain, dyspnea, feeling drowsy and lack of appetite. Forward stepwise approach was used. Continuous variables were dichotomized using their median values as cutoff.
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
in the multidimensional model, and shows the importance of multidimensional analyses. In the biomedical dimension, the importance of anemia to fatigue is well known and probably is the factor most amenable to treatment. It has been proposed that a hemoglobin of 12 gm/dl or greater is necessary to minimize anemia-related fatigue and optimize QOL in patients receiving chemotherapy.27,28 However, there is lack of strong evidence in patients with advanced cancer to support the correlation between anemia and fatigue.29,30 Gleeson and Spencer31 reported a small but significant improvement in weakness among patients who received blood transfusion in the palliative care setting. Stone et al.8 suggested that fatigue might be related more to the changes in hemoglobin level than to the hemoglobin level itself. Our results suggest that anemia’s role in fatigue in our population is limited, and evident in
611
the biomedical dimension only. Further studies are needed to explore the role of blood transfusion or erythropoeitin in palliating cancer-related fatigue. The serum sodium level is another independent predictor of clinically significant fatigue in the situational dimension. This finding has not been reported previously and requires further examination. The most profound findings in our study are that the physical symptoms (pain, lack of appetite, feeling drowsy, and dyspnea) and psychological symptoms (feeling sad and feeling irritable) not only predict clinically significant fatigue unidimensionally, but are the only independent predictors in the multidimensional model. The results are consistent with earlier reports that fatigue is related to various number of symptoms5–10 and support a symptom oriented approach to conceptualize the assessment and management of cancer-related
Fig. 2. Multidimensional independent predictors of fatigue.
612
Hwang et al.
fatigue. Interestingly, feeling drowsy is the only symptom that was not reported previously; it might be simply because the item feeling drowsy has not been included in previous studies. We were surprised that the items of feeling sad and feeling irritable, rather than the Zung depression score, were identified as independent predictors of clinically significant fatigue in the psychological symptom dimension. It has been suggested that by asking a single depression question such as “Are you depressed?” is a valid approach to screen for patients with depression.32 The significant association between psychological symptoms, and depression and fatigue is also well recognized.5,33,34 A few studies have examined the effect of symptom management on fatigue. In an ongoing cancer pain study conducted by our group, 117 patients with worst cancer pain ⱖ 4/10 were interviewed at baseline and 1 week after cancer pain management. There was a modest correlation between changes in worst pain severity and changes in fatigue severity, and patients who had ⱖ 2 points decrease in worst pain severity had significantly less fatigue.35 Appetite stimulants such as megestrol acetate36,37 or corticosteroids38 also reduce fatigue. Dyspnea is another significant fatigue predictor5,8 but studies to show the effect of dyspnea interventions on fatigue are lacking. As depression is strongly associated with fatigue, the effect of antidepressants on fatigue are of interest. Inconsistent results on the effectiveness of depression management for cancer treatment-related fatigue have been reported. Weitzner et al. reported that antidepressants can be helpful in the treatment of hot flashes and associated fatigue, sleep disturbance, and depression in women with breast cancer treated with chemotherapy.39 However, in a doubleblinded placebo controlled randomized trial of 738 patients who reported fatigue with chemotherapy and were randomized to receive either 20 mg of the selective serotonin re-uptake inhibitor (SSRI) paroxetine or placebo, the paroxetine group had significantly reduced depression during chemotherapy but did not experience a significant effect on fatigue.40 The limited effectiveness of symptom interventions to date on fatigue in advanced cancer patients could result from a variety of different reasons. Patients with cancer have multiple symptoms and as fatigue is a common
Vol. 26 No. 1 July 2003
result of any given severe symptom, single symptom interventions may not be effective. Second, most of the symptom interventions are only modestly effective and may not have a large impact on fatigue. Third, if a common pathophysiology underlies symptoms, fatigue, and disease progression, symptom intervention alone may be inadequate. Finally, different symptoms may cause fatigue by different mechanisms. Fatigue related to dyspnea may be different from fatigue related to pain. How these possibilities apply to a patient with fatigue may depend on the clinical circumstances and more research is needed to study the relationship between symptoms and fatigue. There are some limitations in our study. The study was conducted at a VA Medical Center and the results may not be generalizable to all advanced cancer patients. This was a secondary analysis, and further studies including both sexes in other settings are needed to confirm these findings. In summary, we have identified the multidimensional independent predictors of fatigue and suggested the importance of symptom status to fatigue. The presence of fatigue should trigger a wide-ranging review of symptoms to further evaluate the patient’s particular complaint.10,41 The information obtained from this study can be applied easily in clinical settings to help clinicians to identify patients at risk for fatigue within each dimension and to direct questions relevant to fatigue with their patients. Although there are no direct treatments for fatigue, these results support identifying the factors or symptoms amenable to treatment as a step towards reducing fatigue.
Acknowledgments Ms. Michelle Rindos helped with interviewing the patients.
References 1. Cella D, Peterman A, Passik S, et al. Progress toward guidelines for the management of fatigue. Oncology 1998;12:1–9. 2. Aistars J. Fatigue in the cancer patients: a conceptual approach to a clinical problems. Oncol Nurs Forum 1987;14:25–30.
Vol. 26 No. 1 July 2003
Multidimensional Independent Predictors of Fatigue
613
3. Portenoy RK, Itri LM. Cancer-related fatigue: guidelines for evaluation and management. Oncologist 1999;4:1–10.
18. Chang VT, Hwang SS, Feuerman M, et al. The Memorial Symptom Assessment Scale Short Form: Validity and reliability. Cancer 2000;89:1162–1171.
4. Portenoy RK, Miaskowski C. Assessment and management of cancer-related fatigue. Berger A, Portenoy RK, Weissman DE (eds). Principles and practice of supportive oncology. Philadelphia: Lippincott-Raven, 1998:109–118.
19. Karnofsky DA, Burchenal JH. The clinical evaluation of chemotherapeutic agents in cancer. In: MacLeod CM (ed). Evaluation of chemotherapeutic agents. New York: Columbia University Press, 1949: 191–205.
5. Okuyama T, Akechi T, Kugaya A, et al. Factors correlated with fatigue in disease-free breast cancer patients: application of the Cancer Fatigue Scale. Support Care Cancer 2000;8:215–222.
20. Hwang SS, Chang VT, Cogswell C, et al. Clinical relevance of fatigue levels in cancer patients at a Veteran Administration Medical Center. Cancer 2002;94:2481–2489.
6. Broeckel JA, Jacobsen PB, Horton J, et al. Characteristics and correlates of fatigue after adjuvant chemotherapy for breast cancer. J Clin Oncol 1998;16:1689–1696.
21. Cella DF, Tulsky DS, Gray G, et al. The Functional Assessment of Cancer Therapy Scale: development and validation of the general measure. J Clin Oncol 1993;11:570–579.
7. Irvine DM, Vincent L, Graydon JE, et al. Fatigue in women with breast cancer receiving radiation therapy. Cancer Nurs 1998;21:127–135.
22. Hosmer DW, Lemeshow S. Applied logistic regression. Second edition. New York: Wiley, 2000.
8. Stone P, Richards M, A’Hern R, et al. A study to investigate the prevalence, severity and correlates of fatigue among patients with cancer in comparison with a control group of volunteers without cancer. Ann Oncol 2000;11:561–567. 9. Irvine D, Vincent L, Graydon JE, et al. The prevalence and correlates of fatigue in patients receiving treatment with chemotherapy and radiotherapy. A comparison with the fatigue experienced by healthy individuals. Cancer Nurs 1994;17:367–378. 10. Smets EM, Visser MR, Willems-Groot AF, et al. Fatigue and radiotherapy: (A) experience in patients undergoing treatment. Br J Cancer 1998;78:899–906. 11. Pater JL, Zee B, Palmer M, et al. Fatigue in patients with cancer: results with National Cancer Institute of Canada Clinical Trials Group studies employing the EORTC QLQ-C30. Support Care Cancer 1997;5:410–413. 12. Stone P, Hardy J, Broadley K, et al. Fatigue in advanced cancer: a prospective controlled crosssectional study. Br J Cancer 1999;79:1479–1486. 13. Hwang SS, Chang VT, Kasimis B. Comparison of three fatigue measures in a veterans cancer population. Cancer Invest 2003;21:355–365. 14. Mendoza TR, Wang XS, Cleeland CS, et al. The rapid assessment of fatigue severity in cancer patients. Cancer 1999;85:1186–1196. 15. Cella D. Factors influencing quality of life in cancer patients: anemia and fatigue. Semin Oncol 1998;25(Suppl 7):43–46. 16. Yellen SB, Cella SF, Webster K, et al. Measuring fatigue and other anemia-related symptoms with the functional assessment of cancer therapy (FACT) measurement system. J Pain Symptom Manage 1997;13:63–74. 17. Zung WW. A self-rating depression scale. Arch General Psychiatry 1965;12:63–70.
23. Stone P, Richardson A, Ream E, et al. Cancerrelated fatigue: inevitable, unimportant and untreatable? Results of a multi-center patient survey. Cancer Fatigue Forum. Ann Oncol 2000;11:971–975. 24. Blesch KS, Paice JA, Wickham R, et al. Correlates of fatigue in people with breast and lung cancer. Oncol Nurs Forum 1991;18:81–87. 25. Klepstad P, Borchgrevink PC, Kaasa S. Effects on cancer patients’ health-related quality of life after the start of morphine therapy. J Pain Symptom Manage 2000;20:19–26. 26. Chang VT, Hwang SS, Feuerman M, et al. Symptom and quality-of-life survey of medical oncology patients at a veterans affairs medical center: a role for symptom assessment. Cancer 2000;88:1175–1183. 27. Sabbatini P. The relationship between anemia and QOL in cancer patients. Oncologist 2000;5: 19–23. 28. Cleeland CS, Demetri GD, Glaspy J, et al. Identifying hemoglobin level for optimal QOL: results of an incremental analysis. Proc ASCO 1999;18:574a, Abstract 2215. 29. Bruera E, Brenneis C, Michaud M, et al. Association between asthenia and nutritional status, lean body mass, anemia, psychological status, and tumor mass in patients with advanced breast cancer. J Pain Symptom Manage 1989;4:59–63. 30. Morant R. Asthenia: An important symptom in cancer patients. Cancer Treat Rev 1996;22(Suppl A): 117–121. 31. Gleeson G, Spencer D. Blood transfusion and its benefit in palliative care. Palliat Med 1995;9:307–313. 32. Chochinov HM, Wilson KG, Enns M, et al. “Are you depressed?” Screening for depression in the terminally ill. Am J Psychiatry 1997;154:674–676. 33. Smets EM, Garssen B, Cull A, et al. Application of the multidimensional fatigue inventory (MFI-20)
614
Hwang et al.
Vol. 26 No. 1 July 2003
in cancer patients receiving radiotherapy. Br J Cancer 1996;73:241–245.
blind, placebo-controlled study. J Am Geriatr Soc 2000;48:485–492.
34. Dimeo F, Stieglitz RD, Novelli-Fischer U, et al. Correlation between physical performance and fatigue in cancer patients. Ann Oncol 1997;8: 1251–1255.
38. Bruera E, Roca E, Cedaro L, et al. Action of oral methylprednisolone in terminal cancer patients: a prospective randomized double-blind study. Cancer Treat Rep 1985;69:751–754.
35. Hwang S, Chang V, Alejandro Y, et al. Cancer pain management and its impact on fatigue, psychological distress and depression in male cancer patients at a VA medical center. J Pain 2002;3 (2 Suppl 1):44, Abstract 774.
39. Weitzner MA, Moncello J, Jacobsen PB, et al. A pilot trial of paroxetine for the treatment of hot flashes and associated symptoms in women with breast cancer. J Pain Symptom Manage 2002;23: 337–345.
36. Bruera E, Ernst S, Hagen N, et al. Effectiveness of megestrol acetate in patients with advanced cancer: a randomized, double-blind, crossover study. Cancer Prev Control 1998;2:74–78.
40. Morrow GR, Hickok JT, Raubertas RF, et al. Effect of an SSRI antidepressant on fatigue and depression in seven hundred thirty-eight cancer patients treated with chemotherapy: a URCC CCOP study. ASCO 2001;20: Abstract 1531.
37. Yeh SS, Wu SY, Lee TP, et al. Improvement in quality-of-life measures and stimulation of weight gain after treatment with megestrol acetate oral suspension in geriatric cachexia: results of a double-
41. Richardson A, Ream E. The experience of fatigue and other symptoms in patients receiving chemotherapy. Eur J Cancer Care 1996;5(Suppl):24–30.