SEARCH

SEARCH BY CITATION

Keywords:

  • cancer;
  • fatigue;
  • veterans;
  • symptom;
  • functional interference;
  • depression;
  • quality of life;
  • survival

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

BACKGROUND

The correlation of fatigue levels with functional interference, symptom distress, and quality of life may help determine clinically significant fatigue levels.

METHODS

One hundred eighty consecutive patients with cancer completed the Functional Assessment of Cancer Therapy (FACT) General and Fatigue subscales (FACT-G and FACT-F, respectively), the Memorial Symptom Assessment Scale-Short Form (MSAS-SF), the Depression Scale (Zung), and the Brief Fatigue Inventory (BFI). The Karnofsky performance status (KPS) was determined for each patient. Multivariate analyses of variance were performed to compare fatigue models with different cut-off points to categorize fatigue levels. Cox proportional hazards analysis was performed to assess the association between fatigue severity and survival.

RESULTS

Increased fatigue levels were associated with greater symptom distress and decreased quality of life. A model with usual fatigue cut-off points of 0 (no fatigue), 1–2 (mild fatigue), 3–6 (moderate fatigue), and 7–10 (severe fatigue) was optimal in relation to functional interference items (Wilks λ, 0.36; F = 11.61; P < 0.0001), symptom distress scores (Wilks λ, 0.52; F = 10.41; P < 0.0001), and quality-of-life scores (Wilks λ, 0.50; F = 0.50; P < 0.0001). Fatigue severity predicted survival in univariate analysis (chi-square test, 25.42; P < 0.0001). The KPS, stage of disease, and number of symptoms independently predicted survival in patients with fatigue.

CONCLUSIONS

Clinically relevant fatigue levels are correlated with symptom and quality-of-life measurements. Patients with a usual fatigue severity > 3 or a worst fatigue severity > 4 on a 1–10 scale may require further assessment. Cancer 2002;94:2481–9. © 2002 American Cancer Society.

DOI 10.1002/cncr.10507

Fatigue is a highly prevalent symptom in patients with cancer. It has a significant impact on patients' functional level and quality of life.1 In a multicenter survey, fatigue was reported by 78% of patients: patients felt that fatigue affected their daily lives adversely more than cancer-related pain in working, physical and emotional well being, and the ability to enjoy of life. However, only 50% of patients with fatigue discussed fatigue with their oncologist, and 27% received treatment recommendations. The perception of how fatigue interfered in their daily lives differed between oncologists and patients, pointing to different ways of assessing and communicating about fatigue.2 We reported previously that, in patients with cancer who were seen at our Veterans Administration (VA) Medical Center, the lack of energy item from the Memorial Symptom Assessment Scale (MSAS)3 was the most prevalent and highly distressing symptom. Lack of energy was associated with the presence of other significant symptoms, such as dyspnea, nausea, lack of appetite, pain, difficulty sleeping, and difficulty swallowing.4

To better assess cancer-related fatigue, newly validated fatigue assessment tools, such as the Brief Fatigue Inventory5 (BFI) and the Functional Assessment Therapy of Cancer Therapy (FACT) Fatigue subscale (FACT-F),1 have been introduced. Mendoza et al.5 studied responses to the BFI in 309 patients with malignant disease. Those authors categorized BFI worst fatigue severity into mild, moderate, and worst fatigue groups based on correlation with BFI functional interference items, and they suggested that the cut-off point between moderate fatigue and severe fatigue should be between 6 and 7. However, the cut-off point between mild to moderate fatigue could not be determined. Of the three fatigue items (worst, usual, and now), the worst fatigue was chosen in their study because it had the largest correlation with the BFI interference items.

In this article, we report the results of categorizing fatigue severity into four different levels—none, mild, moderate, and severe fatigue—based on their correlation with broader quality-of-life (QOL) constructs as well as BFI interference items. We hypothesized that the classification of fatigue levels by 0–10 fatigue severity should be similar across different measures and different study settings. The relation between fatigue and survival also was examined.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Patient Selection

In this prospective, cross-sectional design survey, 74 inpatients and 106 outpatients with a diagnosis of cancer who were seen consecutively in the Hematology/Oncology Section at VA New Jersey Health Care System at East Orange, New Jersey, were recruited from September 1997 to May 1998. The study was approved by the Institutional Review Board, and all patients signed informed consent before participating. Each patient completed the panel of instruments listed below. Demographic data and Karnofsky performance status (KPS)6 also were determined.

Instruments

The MSAS short form (MSAS-SF) is a validated, patient-rated instrument in which patients rate symptom distress or frequency for 32 highly prevalent physical and psychological symptoms. Each symptom is scored from 0 to 4 ranging, from no symptom to very much. MSAS-SF subscales include the Global Distress Index (GDI), the Physical Symptom Distress Score (PHYS), and the Psychological Symptom Distress Score (PSYCH). The number of symptoms is derived from screening for the presence of 32 symptoms at each interview.7 New summary scores were generated and used in this study by removing the lack of energy item from the subscales calculation to avoid interpretation bias. In this study, we used the revised PHYS, which includes 11 prevalent symptoms (pain, lack of appetite, feeling drowsy, constipation, dry mouth, nausea, emesis, change in taste, weight loss, feeling bloated, and dizziness); and the revised GDI, which includes 4 psychological symptoms (feeling sad, worrying, feeling irritable, and feeling nervous) and 5 physical symptoms (pain, lack of appetite, feeling drowsy, constipation, and dry mouth). The Cronbach α coefficient was 0.81 for the revised PHYS and 0.77 for the revised GDI.

The Functional Assessment of Cancer Therapy (FACT-G)8 (version 3) is a validated, 28-item, general, patient-rated measure of QOL for cancer patients with any tumor type. Each item is scored from 0 to 4, anchored from not at all to very much. There are five subscales: Functional Well-Being (FWB; seven items), Physical Well-Being (PWB; seven items), Social/Family Well Being (SFWB; seven items), Relationship with physician (RMD; two items), and Emotional Well-Being (EWB; five items). The SUMQOL is the sum of all the QOL subscale. The revised PWB and revised SUMQOLwere generated and used in this study by removing the lack of energy item from the questionnaires. The Cronbach α coefficient was 0.70 for the revised PWB and 0.87 for the revised SUMQOL.

The FACT-F subscale1 is a multidimensional fatigue assessment instrument with 13 items and Likert responses. It assesses multiple fatigue characteristics and their impact on function.

The Zung Self-Rating Depression Scale (Zung) is a validated instrument for assessing depression. Patients answer 20 questions using a 0–4 numeric scale with a maximum possible raw score of 80. The depression indices were derived by dividing the sum of the raw scores on the 20 items by 80 and expressing the result as a decimal. The mean values for normal control participants and depressive patients have been established.9 There are three items in the Zung that may reflect somatic symptoms 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 α coefficient for the reconstructed, 17-item depression indices was 0.84.

The KPS is an 11-point rating scale ranging from 0 to 100 (0 = dead, and 100 = normal function). It is used to assess the patient's physical functional level related to cancer and its treatment.

The BFI5 has two components: fatigue severity and fatigue interference. The fatigue severity uses a 0–10 numeric scale to assess fatigue severity at its worst, usual, and right now; the fatigue interference uses a 0–10 scale to assess functional interference caused by fatigue in the areas of daily activity, mood, walking, work, enjoyment of life, and relations with others. The total BFI fatigue interference is the sum of the 6 interference scores. The BFI global fatigue score is the average of the 9 items (3 fatigue severity items and 6 fatigue interference items).

Fatigue Models

The boundary model used to define pain levels for the Brief Pain Inventory10 and fatigue levels for the BFI5 was used in this study. With this method, different fatigue cut-off points are defined, and each set of cut-off points is called a model. Each model is then tested against criterion variables by multiple analyses of variance (MANOVA) to determine which model has the best fit against the criterion variables, as measured by the Wilks λ and the F statistic. Fatigue severity was categorized into four fatigue groups: none, mild, moderate, and severe fatigue. The usual fatigue and worst fatigue severity items were used to establish different fatigue models for comparison. The usual fatigue severity was chosen because, in our population, it demonstrated the highest correlation with BFI interference score, MSAS-SF symptom distress subscales, the Zung depression indicies, FACT-G QOL parameters, and the FACT-F.11 The worst fatigue severity was used because worst fatigue was found informative in the original BFI validation literature.5 Based on the visual association between fatigue severity and the six BFI interference items, the FACT-F subscale, the FACT-G QOL parameters, the MSAS-SF symptom distress subscales, and the Zung depression indicies (see Figs. 1, 2), we grouped usual fatigue severity into two models and worst fatigue severity into two models. The cut-off points for each model are listed in Table 1.

thumbnail image

Figure 1. The association between usual fatigue severity and Brief Fatigue Inventory (BFI) interference scores and Functional Assessment Therapy of Cancer Therapy (FACT)-Fatigue subscale (FACT-F), FACT General SUM QOL score (FACT-G SUMQOL), and Memorial Symptom Assessment Scale-Short Form (MSAS-SF) physical symptom distress measurements.

Download figure to PowerPoint

thumbnail image

Figure 2. The association between worst fatigue severity and Brief Fatigue Inventory (BFI) functional interference scores and Functional Assessment Therapy of Cancer Therapy (FACT)-Fatigue subscale (FACT-F), FACT General SUM QOL (FACT-G SUMQOL), and Memorial Symptom Assessment Scale-Short Form (MSAS-SF) physical symptom distress measurements.

Download figure to PowerPoint

Table 1. Proposed Possible Fatigue Levels for Multiple Analyses of Variance
ModelNo fatigueMild fatigueModerate fatigueSevere fatigue
SeverityNo.SeverityNo.SeverityNo.SeverityNo.
Usual fatigue Model 10671–2333–6647–1016
Usual fatigue Model 20671–2333–5586–1022
Worst fatigue Model 10461–3324–6637–1038
Worst fatigue Model 20461–4405–6557–1038

Statistical Analysis

Identifying the most significant fatigue model

We tested the multivariate null hypothesis that there was no difference between no fatigue, mild fatigue, moderate fatigue, and severe fatigue levels when they were compared simultaneously on the following four criterion variable sets based on the broader QOL construct. The criterion variable sets included functional dimension: six BFI interference items (activity, mood, walk, work, relation with others, and enjoyment of life); symptom dimension: MSAS-SF symptom parameters (PHYS, PSYCH, GDI, and number of symptoms) and Zung Depression Indices; QOL dimension: FACT-G QOL parameters (PWB, FWB, and EWB); and a summary dimension: combination of summary BFI total interference scores and FACT-G SUMQOL. The Wilks λ, and the F statistics derived from MANOVA were used to compare the fatigue models and to select the model with the highest level of significance. The comparisons discussed below were then performed with the most significant fatigue model.

Comparison of MSAS-SF parameters, FACT-G parameters, FACT-F, and depression among the four different fatigue levels

A one-way analysis of variance (ANOVA) was performed to assess the sensitivity of the FACT-F, MSAS-SF subscales, FACT-G parameters, KPS, BFI total interference score, and Zung depression indices to the different fatigue levels. The Tukey multiple comparison test was performed to determine which pairs of fatigue groups were significantly different from one another (Table 3).

Table 2. Comparison of Different Fatigue Models by Multiple Analyses of Variancea
Assessment toolUsual fatigue Model 1Usual fatigue Model 2Worst fatigue Model 1Worst fatigue Model 2
Wilks coefficientF statisticWilks coefficientF statisticWilks coefficientF statisticWilks coefficientF statistic
  • BFI: Brief Fatigue Inventory; MSAS-SF: Memorial Symptom Assessment Scale-Short Form; FACT-G: Functional Assessment of Cancer Therapy-General; SUMQOL: SUM FACT-G QOL score.

  • a

    All P values were < 0.0001.

BFI (6 items)0.3611.610.3611.520.487.910.487.93
MSAS-SF subscales0.5210.410.549.820.559.350.578.77
FACT-G subscale0.5015.460.5214.470.5413.520.5513.21
BFI interference/SUMQOL0.3934.820.3934.150.4825.710.4825.36
Table 3. Summary Statistics and Results of the One-Way Analysis of Variance using Usual Fatigue Model 1
 No fatigueMild fatigueModerate fatigueSevere fatigue
Usual fatigue severity01–23–67–10
Number of patients67336416
 MeanSDMeanSDMeanSDMeanSDF statisticP value
  • SD: standard deviation; FACT-F: Functional Assessment of Cancer Therapy-Fatigue; MSAS-SF: Memorial Symptom Assessment Scale Short Form; PHYS: Physical Symptom Distress Subscale; PSYCH: Psychological Symptom Distress Subscale; GDI: Global Distress Index; FACT-G: FACT General; SUMQOL: SUM FACT-G QOL score; BFI: Brief Fatigue Inventory; KPS: Kamofsky performance status.

  • a

    No significant difference between no fatigue and mild fatigue by one-way analysis.

  • b

    No significant difference between moderate fatigue/severe fatigue by one-way analysis.

  • c

    Statistical significance between all the levels of comparison.

FACT-F fatigue subscale44.317.6841.159.6726.139.6914.889.3377.54< 0.0001a
MSAS-SF lack of energy0.591.001.181.072.561.223.350.8450.90< 0.0001a
MSAS-SF summary scales
 PHYS0.350.480.720.591.140.701.530.7125.53< 0.0001b
 PSYCH0.340.490.590.701.100.901.280.9014.78< 0.0001ab
 GDI0.410.450.720.611.320.691.670.8133.81< 0.0001ab
 No. of symptoms5.603.909.105.9013.404.8014.305.3034.45< 0.0001b
FACT-G subscales
 Physical well being21.583.4320.423.3417.084.2612.815.2329.68< 0.0001a
 Functional well being20.615.6819.306.3311.975.977.945.5736.94< 0.0001ab
 Emotional well being18.421.9017.393.2514.724.7313.445.2515.32< 0.0001ab
 SUMQOL83.939.8081.3612.4866.1312.6456.8014.4839.85< 0.0001a
Functional status
 BFI total interference score1.103.007.2710.6424.2514.4638.8713.2084.30< 0.0001c
 KPS84.5013.7076.1016.6063.6015.2060.6018.5024.76< 0.0001ab
Depression
 Zung depression indices0.380.090.440.110.550.120.600.1333.25< 0.0001ab
Comparison of differences in prevalence and mean distress scores for individual symptoms among the four different usual fatigue levels

The 12 most prevalent symptoms and the mean score for each symptom, as assessed with the MSAS-SF, were tabulated. The difference in prevalence rates for each symptom between four different fatigue levels was assessed with a chi-square test (Table 4). The differences between mean distress scores for each symptom among the four different fatigue levels were assessed with a one-way ANOVA (Table 5).

Table 4. Prevalence of Symptoms Measured by the Memorial Symptom Assessment Scale Short Form using Usual Fatigue Level—Model 1
 No fatigue N = 67 patients (%)Mild fatigue N = 33 patients (%)Moderate fatigue N = 64 patients (%)Worst fatigue N = 16 patients (%)Pearson chi-square testP value
Pain3764819434.520.000
Shortness of breath2842648827.300.000
Lack of appetite1624567534.480.000
Feeling drowsy1636726845.520.000
Dry mouth3445677518.240.000
Weight loss304253568.560.03
Difficulty sleeping1926505616.140.001
Feeling irritable1018435023.890.000
Changes in food taste927424320.730.000
Feeling sad1212444423.220.000
Worrying1930484313.210.004
Constipation1633503116.780.001
Feeling bloated131528377.550.06
Table 5. Individual Symptom Distress by Different Fatigue Levels using Usual Fatigue Model 1
SymptomNo fatigue (n = 67 patients)Mild fatigue (n = 33 patients)Moderate fatigue (n = 64 patients)Severe fatigue (n = 16 patients)F statisticP value
MeanSDMeanSDMeanSDMeanSD
  1. SD: standard deviation.

Pain0.861.341.721.571.951.383.151.1114.50< 0.0001
Shortness of breath0.531.020.831.161.651.512.851.419.00< 0.0001
Lack of appetite0.360.960.511.091.361.502.351.6115.04< 0.0001
Feeling drowsy0.230.590.651.001.551.241.951.5723.79< 0.0001
Dry mouth0.570.961.041.421.491.382.41.7011.14< 0.0001
Weight loss0.511.041.161.551.31.491.91.895.990.0007
Difficulty sleeping0.370.950.991.511.231.401.551.726.370.0004
Feeling irritable0.220.670.440.990.951.241.191.287.670.0001
Changes in food taste0.210.800.651.221.051.390.801.175.860.0008
Feeling sad0.250.750.300.881.031.261.061.487.920.0001
Worrying0.461.000.721.211.171.311.131.454.220.006
Constipation0.431.080.651.101.321.490.701.305.740.0009
Feeling bloated0.240.740.220.580.701.271.051.594.550.004
Correlation between fatigue severity and survival

Kaplan-Meier survival analysis was performed to assess for differences in median survival between four different fatigue levels (Fig. 3). A univariate Cox proportional hazards regression analysis was performed to assess the significance of usual fatigue severity as a predictor of survival. A multivariate Cox proportional hazards regression analysis was performed for the entire population and for patients with fatigue only. The variables in the multivariate Cox proportional hazards regression model included usual fatigue severity, inpatient status, stage of disease, KPS, number of symptoms, MSAS-SF PHYS, MSAS-SF PSYCH, and FACT-G SUMQOL. Analyses were performed with the STATA program, V6.0 (Stata Corp., College Station, TX).

thumbnail image

Figure 3. Kaplan-Meier estimated survival curves by usual fatigue severity groups.

Download figure to PowerPoint

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Demographics

Patient characteristics have been reported in detail elsewhere.11 There were 74 inpatients and 106 outpatients with a median age of 68 years (range, 30–89 years) and a median educational level of 12 grades (range, 4–16 grades). There was no evidence of disease in 13 patients (7%), localized disease in 9 patients (5%), regionally advanced disease in 38 patients (21%), and metastatic disease in 120 patients (67%). Fatigue was present in 134 patients (74%) by BFI worst fatigue, with a mean worst fatigue severity of 3.9 (range, 0.0–10.0) and in 113 patients (62%) by the BFI usual fatigue, with mean usual fatigue of 2.6 (range, 0.0–10.0).

Defining Clinically Significant Fatigue Levels

Box plots of the correlation between fatigue levels and functional interference, symptom distress, and QOL are presented for usual fatigue (Fig. 1) and worst fatigue (Fig. 2). Fatigue levels show a nonlinear correlation with all the criterion variables. By examining the association between fatigue severity and FACT-F, there were two steep slope changes for usual fatigue severity (Fig. 1) and for worst fatigue severity (Fig. 2). The slopes changed between 2 and 3 and between 6 and 7 for usual fatigue severity and changed between 3 and 4 and between 6 and 7 for worst fatigue severity. The fatigue groupings in all of the fatigue models showed significant multivariate effects (P < 0.0001) for all of the proposed criterion variable sets (Table 2).

The usual fatigue Models 1 and 2 showed more significant correlation with the six BFI interference items than the worst fatigue models. Both usual fatigue models also revealed more significant MANOVA results for the symptom criterion variables set and in the FACT-G QOL criterion variable sets. Of the two usual fatigue models, usual fatigue Model 1 showed the most significant results, with Wilks λ, 0.36 and F = 11.61 (P < 0.0001) for the six BFI interference items; Wilks λ, 0.52 and F = 10.41 (P < 0.0001) for the MSAS-SF symptom distress subscales/Zung depression indicies; Wilks λ, 0.50 and F = 15.46 (P < 0.0001) for FACT-G QOL subscales; and Wilks λ, 0.39 and F = 34.82 (P < 0.0001) for the BFI total interference score/FACT-G SUMQOL.

All fatigue models provided similar significant results for the analyses described below. In this report, we present the results from usual fatigue Model 1, because it was the most significant fatigue model in the MANOVA.

Association between Fatigue Levels and MSAS-SF Symptom Distress Subscales, FACT-G QOL Parameters, KPS, Depression Indicies, and FACT-F

Table 3 summarizes the one-way ANOVA results in assessing the variance of the following variables by different fatigue levels: FACT-F, MSAS-SF lack of energy item, MSAS-SF symptom distress subscales, FACT-G parameters, BFI total interference scores, KPS, and Zung depression indices. There were significant differences in all the variables between different fatigue levels according to the one-way ANOVA (P < 0.0001).

The BFI total interference scores showed significant differences in all the pairs of fatigue levels by Tukey multiple comparison tests (F = 84.30; P < 0.0001). The mean BFI total interference score was 1.10 for the no fatigue group, 7.27 for the mild fatigue group, 24.25 for the moderate fatigue group, and 38.87 for the severe fatigue group.

In comparisons between the no fatigue group and the mild fatigue group, differences were present for the MSAS-SF PHYS and number of symptoms. There was no significant difference between the no fatigue group and the mild fatigue group according to the Tukey multiple comparison test for the following measures: MSAS-SF PSYCH, MSAS-SF GDI, FACT-G parameters (PWB, FWB, EWB, and SUMQOL), FACT-F, KPS, and Zung depression indices. However, all of these parameters illustrated slightly worse scores in the mild fatigue group compared with the no fatigue group.

By comparing the MSAS-SF parameters (PHYS, PSYCH, GDI, and number of symptoms), FACT-G FWB, FACT-G EWB, KPS, and Zung depression indices between moderate and severe fatigue groups, worse scores were present in the severe fatigue group. However, the differences did not reach statistical significance.

Differences in Symptom Prevalence and Distress between Different Fatigue Levels

There was a significant difference in the mean number of symptoms in different fatigue groups by one-way ANOVA (F = 34.45; P < 0.0001). The mean number of symptoms was 5.6 symptoms for no fatigue, 9.1 symptoms for mild fatigue, 13.4 symptoms for moderate fatigue, and 14.3 symptoms for severe fatigue. There was no difference between the moderate and severe fatigue groups.

The most frequently reported symptoms in both the severe fatigue group and the moderate fatigue group included pain, shortness of breath, lack of appetite, feeling drowsy, dry mouth, weight loss, difficulty sleeping, feeling irritable, changes in food tastes, feeling sad, worrying, and constipation. There was a significant difference in the prevalence of each symptom between the different fatigue levels by chi-square statistic, with chi-square tests ranging from 8.56 (P < 0.03) for weight loss to 45.52 (P < 0.0001) for feeling drowsy (Table 4).

A one-way ANOVA was performed to compare the variance for each symptom among different fatigue levels. The results, which are shown in Table 5, indicate that individual symptom distress scores for the most prevalent 12 individual symptoms differed significantly between different fatigue levels.

Relation between Fatigue and Survival

One hundred twenty-one patients (67%) died with median survival of 14.8 months (range, 0.03–51.0 months) for the entire population as of April 1, 2001. Thirty-six patients (54%) in the no fatigue group died with median survival of 31.3 months (range, 0.1–51.1 months), 22 patients (67%) in the mild fatigue group died with median survival of 16.5 months (range, 0.7–42.6 months), 48 patients (75%) in the moderate fatigue group died with median survival of 6.8 months (range, 0.16–42.7 months), and 15 of 16 patients (94%) in the severe fatigue group died with median survival of 5.2 months (range, 0.37–41.9 months). Figure 3 shows that there was a significant difference in survival among the four different fatigue groups by Kaplan-Meier survival analysis with log-rank chi-square tests (23.19; P < 0.0001).

According to a Cox proportional hazards regression analysis, the usual fatigue severity only predicted survival in univariate analyses (hazard ratio, 1.19; 95% confidence interval, 1.12–1.28; chi-square test, 25.42; P < 0.0001). In a multivariate Cox proportional hazards regression model with usual fatigue severity and other known survival predictors (inpatient status, stage of disease, KPS, number of symptoms, PHYS, PSYCH, and FACT-G SUMQOL), the KPS, number of symptoms, stage of disease, and PHYS (hazard ratios: 0.94, 0.91, 1.92, and 1.70, respectively; P < 0.001, P < 0.005, P < 0.001, and P < 0.05, respectively) independently predicted survival (chi-square test [8], 135.23; P < 0.0001) for the entire population. For patients who presented with fatigue (usual fatigue severity > 0: n = 113 patients; chi-square test [8], 88.41; P < 0.0001), the independent survival predictors were KPS, stage of disease, and number of symptoms (hazard ratios: 0.94, 2.57, and 0.92, respectively; P < 0.001, P < 0.001, and P < 0.01 respectively).

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

For practitioners, the ability to assign a meaning to patient ratings of fatigue would aid in the assessment and management of fatigue. In this study, we used the association between fatigue severity and broader QOL constructs, which include symptom distress parameters, depression, QOL parameters, and functional status, to define clinically relevant fatigue levels. We also used the 0–10 rating of fatigue severity to categorize the fatigue levels, because it is administered easily in daily clinical practice. Four key conclusions can be drawn from this study.

First, differences in fatigue levels correspond to different levels of symptom distress and QOL as well as interference. These differences remained after possibly confounding items had been removed from the instruments for symptom distress, QOL, and depression. On this basis, fatigue severity can be categorized into four different fatigue groups. We then defined and compared fatigue models with different fatigue cut-off points by MANOVA. We were able to reproduce the results of the model proposed by Mendoza et al.5 in our patient population. However, the usual fatigue severity rating may be more informative in our study population.

Second, for the no fatigue group and the mild fatigue group, other than the BFI fatigue functional interference score, there was no statistically significant difference between these groups for the symptom distress, QOL, and depression scores. Third, there was a significant difference in symptom prevalence and symptom distress scores between the different fatigue groups. Fourth, fatigue severity predicted length of survival in univariate analyses but not in multivariate analyses.

These findings are of importance in developing a better understanding of the meaning of patient-rated fatigue levels. These findings support the hypothesis that different fatigue levels can correspond to different levels of QOL and symptom levels and across different populations.

Different classifications of fatigue levels have been proposed. Morant used the 0–100 visual analog scale to assess the fatigue severity and grouped fatigue into mild fatigue (< 30), moderate fatigue (30–60), and severe fatigue (> 60) to compare the difference of various symptoms between different fatigue groups.12 In the current study, we observed a nonlinear relation between both usual fatigue and worst fatigue severity and all of the proposed parameters (Figs. 1, 2) and suggested cut-off points to categorize the fatigue levels. The MANOVA results indicated that the usual fatigue models are slightly superior to the worst fatigue models and that Model 1 is slightly better than Model 2. Further supporting usual Model 1 is our perception that, in Figure 1, the slopes change between 2 and 3 and between 6 and 7 in relation to the FACT-F subscale.

Fatigue levels also can be grouped by worst fatigue severity with cut-off points of 0, 1–3, 4–6, and 7–10. The differences between the worst fatigue model and the usual fatigue model are relatively small.

Differences in the study populations between a National Cancer Institute (NCI)-designated cancer center and a VA Medical Center may account for some of the differences between our findings and the findings of Mendoza et al.,5 who studied fatigue in an NCI-designated cancer center. Their sample was comprised of 49% female patients and 51% male patients with a median age of 55 years. Most of the patients in their sample had at least a college degree. More than two-thirds of their study population had hematologic malignancies (n = 229 patients; 76%). We studied male veterans at a VA medical center with a median age of 68 years and median 12th grade education level. In contrast, hematologic malignancies were present in 19% of our patient population. This population reflects the national VA population, which differs from the general population in that it experiences a higher mortality rate,13 reflects the lower 10% of the socioeconomic strata,14 and has poor health status scores compared with non-VA populations.15 All of these facts may have contributed to differences in fatigue results. Despite these differences, we drew similar conclusions using the same methodology.

Our results suggest that attention to fatigue should start when patients present with moderate fatigue (usual fatigue severity ≥ 3 or worst fatigue severity ≥ 4), which accounts for 44–50% of our study population. Patients in the moderate to severe fatigue groups had a median number of 13–14 symptoms with significantly higher depression scores. Higher individual symptom distress also was observed in both the moderate fatigue group and the severe fatigue group. In other reports, fatigue was highly associated with physical symptoms and psychological symptoms, such as pain,16–18 depression,19 psychological distress,20–23 dyspnea,6, 10, 19, 22 anxiety,22 and poor quality of sleep.19–21, 24

We conclude that fatigue can be considered as a symptom that arises in response to other symptoms and should not be associated only with one given symptom, such as depression.11 This interpretation also is supported by the stress theory proposed by Aistars.23 The presence of fatigue should trigger a wide-ranging review of symptoms to further understand the patient's particular complaints.

We are unaware of previous reports on the association between fatigue and survival. Fatigue may be a marker for disease severity based on its significant association with symptom distress, QOL, KPS, and laboratory values in our patient population.11 It is not surprising that patients with higher fatigue severity have a shorter survival. In a multivariate model, well-documented predictors of survival, such as KPS, the number of symptoms, and stage of disease,25 independently predicted survival in patients with fatigue.

In summary, we were able to identify clinically relevant fatigue levels. Attention to fatigue may be indicated when patients present with usual fatigue severity ≥ 3 or worst fatigue ≥ 4. Patients with moderate to severe fatigue demonstrate significantly higher functional interference scores, lower QOL, and greater symptom distress. A potential limitation of this study is generalizability the results to a non-VA, mixed-gender population of patients with malignant disease. Further studies are needed to prospectively confirm these results.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The authors thank Dr. Charles Scott for valuable statistical comments and Ms. Michelle Rindos for interviewing the patients.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES