1186

The Rapid Assessment of Fatigue Severity in Cancer

Patients
Use of the Brief Fatigue Inventory

Tito R. Mendoza, M.S.1 BACKGROUND. Fatigue is a major disease and treatment burden for cancer patients.
X. Shelley Wang, M.D.1 Several scales have been created to measure fatigue, but many are long and
Charles S. Cleeland, Ph.D.1 difficult for very ill patients to complete, or they are not easy to translate for
Marilyn Morrissey, M.P.H.1 non-English speaking patients. The Brief Fatigue Inventory was developed for the
Beth A. Johnson, R.N.1 rapid assessment of fatigue severity for use in both clinical screening and clinical
Judy K. Wendt, M.P.H.1 trials.
Stephen L. Huber, M.S.2 METHODS. The study enrolled 305 consecutive, consenting adult inpatients and
outpatients with cancer who could understand and complete the self-report mea-
1
Pain Research Group, The University of Texas sures used in the study. The same instruments also were administered to 290
M. D. Anderson Cancer Center, Houston, Texas. community-dwelling adults to obtain a comparison sample. Research staff com-
2
Pharmacoeconomics Research Group, The Uni- pleted a form that indicated the primary site and stage of the cancer, rated the
versity of Texas M. D. Anderson Cancer Center, Eastern Cooperative Oncology Group performance status of the patient, described
Houston, Texas. the characteristics of the pain, and described the current pain treatment being
provided to the patients.
RESULTS. The BFI was shown to be an internally stable (reliable) measure that
tapped a single dimension, best interpreted as severity of fatigue. It correlated
highly with similar fatigue measures. Greater than 98% of patients were able to
complete it. A range of scores defining severe fatigue was identified.
CONCLUSIONS. The BFI is a reliable instrument that allows for the rapid assessment
of fatigue level in cancer patients and identifies those patients with severe fatigue.
Cancer 1999;85:1186 –96. © 1999 American Cancer Society.

KEYWORDS: fatigue, assessment of fatigue, fatigue severity, functional interference.

F atigue is a commonly reported symptom by individuals suffering

from various diseases, including multiple sclerosis, rheumatoid
arthritis, and cancer. Fatigue is the most frequently reported symp-
This study was supported by award SIG-21 from
tom of cancer patients.1 Because of its prevalence, it is often reported
the American Cancer Society and by an unre-
stricted educational grant from Ortho Biotech Inc. as the symptom that is the most distressing and causes the greatest
amount of interference with daily life.2 In a study of breast and lung
Acknowledgments: The authors thank Hong Guo carcinoma patients, Blesch et al.3 reported that 99% of patients in
and Melissa Menzies for data entry and Martha their sample experienced fatigue. In a study of chemotherapy and
Engstrom for editing.
radiotherapy patients, Irvine et al.4 found that 61% experienced fa-
Stephen L. Huber’s current affiliation: President, tigue after treatment. The occurrence of fatigue across diagnostic and
Cortex Communications, Tampa, FL. treatment categories at all phases of the life span underscores the
need for empirically based interventions.5
Address for reprints: Charles S. Cleeland, Ph.D., Fatigue in cancer patients is associated with psychological dis-
Pain Research Group, Box 221, 1100 Holcombe
turbance, symptom distress, and decreases in functional status.4 Fa-
Blvd., Houston, TX 77030.
tigue in cancer may be caused by the disease itself, or it may be
Received July 28, 1998; accepted October 7, caused by various treatments, such as chemotherapy, radiotherapy,
1998. biological therapy, or surgery. In addition, other mechanisms that

© 1999 American Cancer Society

 Rapid Assessment of Fatigue/Mendoza et al. 1187

may be responsible for fatigue include sleep distur- tients rate whether they feel “better than,” “the same
bances, environmental conditions, level of activity, as,” or “worse than” for each item. The scale is brief,
nutritional status, and diverse inherent factors.6 Based exhibits good internal consistency (alpha 0.82– 0.97)
on the available data, it is unclear how each of these and test-retest reliability, and has the ability to dis-
factors may affect the cancer patient’s experience of criminate between patients and healthy volunteers.
fatigue. However, this instrument utilizes a three-point scale
Until very recently, fatigue has not been measured that may not provide enough variability for the assess-
routinely in many diseases, such as cancer, despite its ment of different levels of fatigue. Also, some of the
impact on patients and its effect on quality of life.7 items on this checklist may be difficult for some pa-
Most commonly, fatigue is found as one item on a tients to comprehend or too idiomatic for easy trans-
scale measuring functional status or evaluating mood lation.
or as an item in a toxicity report. In recent literature Two subscales of the POMS10 have been used to
reviews, Irvine et al.8 and Richardson2 both argued for assess fatigue in cancer patients. These subscales are
better assessment tools and greater methodological the POMS-Fatigue and the POMS-Vigor. The POMS-
rigor for future research on fatigue in cancer patients. Fatigue consists of the following items: “worn-out,”
They also noted that assessment scales have rarely “listless,” “fatigued,” “exhausted,” “sluggish,” “weary,”
been subjected to adequate reliability and validity and “bushed.” The POMS-Vigor includes these items:
testing. Richardson2 encouraged the creation of more “lively,” “active,” “energetic,” “cheerful,” “alert,” “full
assessment tools designed specifically for use with the of pep,” “carefree,” and “vigorous.” On both sub-
cancer population. scales, each item is rated using a five-point scale.
Recommendations for assessing fatigue vary ac- These scales are brief enough to assess fatigue in can-
cording to the purpose of the assessment. The general cer patients but, like the Pearson–Byars, may be diffi-
approach to assessing fatigue is based on the concep- cult to translate.
tualization of fatigue as a subjective experience. The FACT-F11 was designed to measure the fa-
Therefore, it can be compared with the symptom of tigue symptoms of cancer patients with anemia. It
pain: It is what the person experiencing it says it is. consists of the 28 items of the FACT-General (FACT-G)
Most researchers have used self-report instruments to to assess health-related quality of life and an addi-
measure fatigue. In clinical practice, the focus is on tional 13 items to assess fatigue. Examples of these
efficiently obtaining information that is needed for fatigue items are “feel weak all over,” “have energy,”
patient care. Further research is needed to compare and “I am frustrated by being too tired to do the things
the various approaches to measuring fatigue and to I want to do.” Each item was rated on a five-point
clearly separate the experience of fatigue from the Likert scale ranging from “0” (not at all) to “4” (very
responses to fatigue. much so). The FACT-F demonstrates good internal
Recently, several instruments have been em- consistency and test-retest reliability. The main disad-
ployed for the assessment of fatigue. These include: vantage of the FACT-F is its length. It may be too long
the Pearson–Byars Fatigue Feeling Checklist,9 the Pro- to be given to cancer patients in a clinical setting. The
file of Mood States (POMS) Fatigue and Vigor sub- 13-item fatigue subscale, on the other hand, does
scales,10 the Functional Assessment of Cancer Thera- meet the requirement of a rapidly-administered scale,
py-Fatigue (FACT-F),11 the Piper Fatigue Self-Report but some items may be difficult to understand or to
Scale,12 the Fatigue Assessment Instrument (FAI),7 the translate into other languages.
Multidimensional Fatigue Inventory (MFI),13 the Mul- Other scales were designed to portray more than
tidimensional Fatigue Symptom Inventory (MFSI),14 one dimension of fatigue, including the Piper Fatigue
and the Fatigue Symptom Inventory (FSI).15 Some of Self-Report Scale (PFS),12 the MFI,13 and the MFSI.14
these measures are too long for very ill patients to Many of these scales have been developed carefully
complete. Others depend on English-based expres- from a theoretical standpoint, taking into account the
sions or idioms that make them difficult to translate. A multifactorial nature of fatigue, and meet accepted
significant number of cancer patients in the United standards of validity and reliability. However, their
States do not speak English or speak English as a length makes them difficult for cancer patients to
second language, making ease of scale translation a complete, especially for those who are very tired. The
goal of clinical and research assessment measures. time required to complete them makes them difficult
The Pearson–Byars Fatigue Feeling Checklist9 to use for clinical screening, or for outcome measures
consists of 13 items that are descriptive of energy in clinical trials.
levels. Examples of items include “slightly tired,” “fair- We developed a new fatigue measure, the Brief
ly well pooped,” “very lively,” and “quite fresh.” Pa- Fatigue Inventory (BFI), to address some of the con-
1188 CANCER March 1, 1999 / Volume 85 / Number 5

cerns with existing instruments. We based the BFI on data to eliminate items that did not contribute to
the Brief Pain Inventory (BPI),16 which has been used discrimination between patients and controls and
successfully to assess the severity and impact of can- items that demonstrated little variability or sensitivity
cer pain in the United States17 and in European18 and with the patient sample for the final BFI.
Asian countries.19 The simple wording of the BPI The Wisconsin sample was composed of 249 pa-
makes it easy to understand for educationally disad- tients and normal subjects. Approximately 29% of the
vantaged patients as well as easy to translate. The BPI total sample were psychiatric patients receiving treat-
measures the severity of pain and interference caused ment for depression, whereas 27% were cancer pa-
by pain using 0 –10 scales. In addition, the level of pain tients. One-third of the cancer patients were receiving
assessed by the BPI can be divided into categories of interferon treatment. Thirty-three percent of the sam-
“mild” (1– 4), “moderate” (5– 6), and “severe” (7–10) ple was comprised of normal volunteers who were
pain based on the amount of pain-related interference administered the questionnaire at a benefits fair. The
with function.20 We hoped to use the same method to remaining 11% was comprised of graduate students
establish ranges of fatigue severity. who were given the questionnaire after a night of
studying during finals week.
Study Design and Overview In contrast to the commonly held notion that
The study consisted of several phases. First, we exam- sleep affects level of fatigue, preliminary analyses of
ined fatigue by using responses to a fatigue question- the Wisconsin data suggested that items describing
naire developed at the University of Wisconsin-Mad- how people sleep, such as “feeling rested and re-
ison. The data were collected from both healthy freshed upon awakening,” “difficulty falling asleep,”
controls and patients in the Madison area. Second, by and “waking up during the night,” were not signifi-
using preliminary results from the Wisconsin data, we cantly related to reported fatigue in the patient group,
refined the items and administered a revised fatigue nor was report of the number of hours slept in the
questionnaire to patients at The University of Texas previous 24 hours. Based on this preliminary exami-
M. D. Anderson Cancer Center and to a sample of nation of fatigue and related questions in a very di-
community-dwelling adults from the Houston area verse sample, a multidisciplinary working group was
with a similar age distribution. Third, we proceeded to formed to review the fatigue literature and to develop
a formal validation of the final version of the BFI by a strategy for the simplification and validation of
demonstrating the instrument’s psychometric proper- the BFI.
ties. We showed evidence of the BFI’s validity with
construct, concurrent, and discriminant validity. We Validation Strategy Plan
also demonstrated its reliability. Fourth, we developed The validation plan involved establishing validity in
a categorization of severity of fatigue based on how three ways: through construct, concurrent, and dis-
much it interfered with function. criminant validity. Construct validity was examined by
using a data reduction technique known as factor
Development of the Brief Fatigue Inventory: Data from analysis. This procedure allowed us to discern the
the Wisconsin Fatigue Study underlying factors or constructs the BFI is supposed to
The Pain Research Group collected data on fatigue measure. Concurrent validity analysis was performed
from a fatigue questionnaire that had been adminis- by correlating the BFI with the Fatigue subscale of the
tered previously to both patients and normal subjects FACT and the Vigor and Fatigue subscales of the
at the Wisconsin Comprehensive Cancer Center at the POMS. Although there is no “gold standard” for the
University of Wisconsin-Madison. This questionnaire measurement of fatigue, it is important to evaluate
included demographic variables and fatigue-related how well the BFI compares with previously validated
items. These fatigue-related items assessed the sever- fatigue instruments. Discriminant validity was estab-
ity of fatigue, the amount of interference with function lished by comparing the mean BFI scores of patient
caused by fatigue, and the presence of factors that groups who were expected to have different levels of
worsen fatigue, such as pain and medications. Other fatigue based on Eastern Cooperative Oncology Group
items focused on how factors, such as illness, medical (ECOG) performance status. To show that the BFI
treatments, depression, worrying, “keeping up with items are reliable, Cronbach’s coefficient alpha, a
daily activities” (such as work and school work), too measure of internal consistency, was calculated. Fi-
much idle time, and difficulty sleeping, contribute to nally, we explored the possibility of defining ranges of
subjects’ fatigue. The questionnaire also included fatigue severity scores following the procedure that we
items asking about the quality of sleep. A mood mea- had used to define ranges of pain severity based on
sure, the POMS, also was administered. We used these the BPI.
 Rapid Assessment of Fatigue/Mendoza et al. 1189

FIGURE 1. The Brief Fatigue Inventory.

1190 CANCER March 1, 1999 / Volume 85 / Number 5

The Brief Fatigue Inventory TABLE 1

The number of items and the item format of the final Description of the Sample of Patients and Control Subjects
BFI (see Fig. 1) closely follow those of the BPI,16 a
Patients Controls
pain-assessment instrument. The one-page BFI has (n 5 305) (n 5 290)
only nine items, with the items measured on 0 –10
numeric rating scales. The advantages of using nu- Characteristic % No. % No.
meric rating scales over other types of rating scales are
Gender
discussed elsewhere.21 Three items ask patients to rate
Female 49 150 30 88
the severity of their fatigue at its “worst,” “usual,” and Male 51 154 70 201
“now” during normal waking hours, with 0 being “no Educational level
fatigue” and 10 being “fatigue as bad as you can imag- Elementary school 2 5 1 4
ine.” Six items assess the amount that fatigue has High school 42 127 11 31
College 35 104 46 128
interfered with different aspects of the patient’s life
Graduate school 21 64 42 118
during the past 24 hours. Depending on the purposes Race
of measurement, this time interval can be changed to White 81 244 85 240
the past week. The interference items include general Other 19 57 15 44
activity, mood, walking ability, normal work (includes Employment status
Employed full time 21 40 60 165
both work outside the home and housework), rela-
Homemaker 11 20 7 20
tions with other people, and enjoyment of life. The Retired 24 45 23 61
interference items are measured on a 0 –10 scale, with Disabled due to illness 34 63 1 2
0 being “does not interfere” and 10 being “completely Other 10 19 9 29
interferes.” Disease type
Hematologic
malignancies
METHOD Leukemia (acute) 17 51
Leukemia (chronic) 16 48
Subjects
Lymphoma 43 130
Community-dwelling controls Solid tumors
The control subjects (n 5 290) were members of ser- Breast 10 29
vice groups around the Houston area. Each organiza- Gastrointestinal 6 19
tion was contacted in advance, and permission was Gynecologic 2 5
Genitourinary 1 2
granted to introduce the study and recruit participants
Other 5 18
at a future organization meeting. Any person attend- ECOG performance status
ing the organization meeting on the scheduled date (0) Fully active 18 55
was eligible to participate, and participation implied (1) Restricted but
informed consent. There were no restrictions placed ambulatory 29 87
(2) Ambulatory, capable
on eligibility to participate, including any current or
of self care 43 128
past cancer diagnosis. Specific service organizations (3) Capable of only
were chosen in an attempt to gather data from a limited self care ,1 2
diverse study sample in terms of gender, ethnicity, and (4) Completely disabled
approximate age of cancer patients. Data were col- Median age in yrs (range) 55 (18–88) 57 (17–100)
lected over a 3-month period.
ECOG: Eastern Cooperative Oncology Group.

Patients
The data on patients (n 5 305) were collected at The psychiatric illness at the time of the study. Compliance
University of Texas M. D. Anderson Cancer Center was excellent. Only five patients refused to participate
with both inpatients and outpatients from the depart- for the following reasons: “do not want to take the
ments of Bone Marrow Transplantation, Leukemia, time,” “not up to it,” and “too tired to write.”
Lymphoma, Gastrointestinal Oncology, and Radiation Table 1 presents the descriptive characteristics of
Oncology. Patients were at least 18 years of age, were the controls and patients who participated in the
able to read and understand English, had a patholog- study. The control and patient samples were approx-
ical diagnosis of cancer, and had given verbal consent imately the same age. Most of the controls were male,
to participate. Patients were excluded if their clinical white, fully employed, and highly educated. The pa-
status was felt to be too poor to allow them to com- tient sample was evenly represented by gender and
plete the survey or if they were diagnosed with a major was mostly white. Thirty-four percent of the patients
 Rapid Assessment of Fatigue/Mendoza et al. 1191

were disabled due to illness. Although they were not as nase, alanine amino transferase, bilirubin, blood urea
highly educated as the controls, the majority of the nitrogen, creatinin, and serum albumin. The study
patients in the sample had at least a college degree. coordinator or designated study nurse completed this
The three most common primary cancer diagnoses checklist record.
were lymphoma, leukemia (acute), and leukemia
(chronic). More than half of the patients had poor Statistical Method Used to Categorize Fatigue Severity
performance status (ECOG performance rating of 2 or We explored the possibility that patients could be
greater). grouped as having mild, moderate, and severe fatigue
based on the “fatigue worst” item. We investigated
Instruments Used and Data Collected possible boundaries for the groupings by examining
In addition to the preliminary fatigue questionnaire, the correlation between patients’ self-reported inter-
the survey packet contained two previously validated ference with function and their ratings on the “fatigue
measures of fatigue and a questionnaire to gather worst” item. We chose “fatigue worst” to represent
basic demographic information. The primary disease fatigue severity for conceptual simplicity and because
site was also recorded. this item has the largest correlation with the interfer-
The previously validated measures included the ence items.
following: 1) The FACT-F and FACT-Anemia subscales By using the same boundary models we used in
of the FACT assessment system.11 This 20-item instru- establishing cut points between categories for pain
ment contains 13 items dealing with fatigue and 7 severity on the BPI,20 we tested two boundaries for
items dealing with anemia. Each question is scored on mild levels of fatigue and two boundaries for severe
a five-point rating scale from 0 (“not at all”) to 4 (“very levels of fatigue. For mild fatigue, we tested whether a
much”). 2) The POMS-Vigor and POMS-Fatigue sub- cut point between 3 and 4 or between 4 and 5 was the
scales, which measure subjective mood states.10 The optimal point for distinguishing “mild” fatigue from
15-item instrument includes 8 items describing feel- “nonmild” fatigue. For severe fatigue, we evaluated
ings of vigor (“energetic,” “full of pep”) and 7 items whether a cut point between 6 and 7 or between 7 and
describing feelings of fatigue (“listless,” “bushed”). 8 was the optimal point for distinguishing “severe”
Each question is scored on a five-point rating scale fatigue from “nonsevere” fatigue. With two possible
from 0 (“not at all”) to 4 (“extremely”). cut points for mild fatigue and two possible cut points
for severe fatigue, there are four possible combina-
Patient Data (Checklist) tions with which to distinguish the three levels of
In addition to the assessment instruments mentioned fatigue severity, with “moderate” falling between
above, information on factors that might influence “mild” and “severe.”
fatigue scores for the patient group also were col- We assumed that persons with “mild” fatigue
lected. This information was recorded on a checklist, would have significantly less fatigue-related interfer-
which assessed disease status, nutritional status, treat- ence than those with “nonmild” fatigue. We also as-
ment information, and current laboratory values sumed that persons with “severe” fatigue would have
thought to be related to fatigue. Disease status infor- significantly greater interference than those with
mation included cancer diagnosis, stage, infection ev- “nonsevere” fatigue. Six fatigue interference items
idence, ECOG performance status, and eating prob- were used in a multivariate analysis to test for cut
lems (e.g., loss of appetite, vomiting). Information on points between severity categories.
usual and current body weight also was recorded. In a multivariate analysis of variance (MANOVA),
Treatment regimen data consisted of whether the pa- there is more than one dependent variable, so there is
tient had undergone radiotherapy, chemotherapy, more than one between group and error sum of
surgical therapy, bone marrow transplantation, or squares as well as between group and error sums of
blood transfusion within the past 30 days. It was also cross products. These sums of squares and cross prod-
noted whether or not the patient was taking opioid ucts form the basis of multivariate test criteria in
analgesics and/or was receiving biotherapy (e.g., in- much the same way as the between-group and error
terleukin-2, epoetin-alpha, growth-colony stimulating mean squares do in univariate analysis of variance
factor, tumor necrosis factor). Finally, patient labora- (ANOVA). Unlike in a single-factor ANOVA in which
tory data results with a possible relation to fatigue there is only one dependent variable and, hence, one
were recorded if they were available within the last F-value associated with the between group and error
week. These included laboratory variables, such as sum of squares, there are several multivariate test
hemoglobin level, white blood cell count, platelets, criteria in MANOVA. Three of the most commonly
total and absolute neutrophil count, lactic dehydroge- used are Pillai’s trace, Wilk’s lambda, and Hotelling’s
1192 CANCER March 1, 1999 / Volume 85 / Number 5

TABLE 2
Means and Standard Deviations on Key Fatigue Outcome Measures
for Patients and Control Subjects

Range of
possible Patients Controls
Measuresa scores (n 5 303) (n 5 275)

BFI* 0–10 4.7 (2.8) 2.2 (1.8)

POMS-fatigue* 0–28 15.1 (8.8) 6.3 (5.7)
Fatigue subscale of the FACT* 0–52 23.3 (15.1) 40.6 (8.6)
Anemia specific items of the FACT* 0–80 41.4 (19.4) 64.0 (11.5)
POMS-vigor* 0–32 11.1 (7.6) 20.6 (6.9)

a
Higher levels of fatigue are associated with lower scores on anemia specific items of the Functional
Assessment of Cancer Therapy (FACT), the fatigue subscale of the FACT, and the Profile of Mood States
(POMS)-vigor and with higher scores on the Brief Fatigue Inventory (BFI) and the POMS-fatigue
subscale.
* Statistically significant difference between patients and controls at the P , 0.01 level.
FIGURE 2. Comparison of Brief Fatigue Inventory (BFI) scores between
patients and controls. Most of the controls tend to have scores of at most 5,
whereas most patients have at least 5.
trace,22,23 which can each be transformed into statis-
tics approximated by the F distribution. If these three
criteria agree, then it would reinforce the appropriate-
presents a descriptive summary of the key fatigue
ness of the cut points.
outcome measures by the two groups: normal controls
To determine whether a “fatigue worst” cut point
and patients. A mean BFI fatigue score, which is dis-
of between 3 and 4 or between 4 and 5 provides a
cussed further below, was calculated from the nine
more optimal range to describe “mild” fatigue by the
BFI items. On all of the outcome measures, the pa-
impact of the six interference items, one can perform
tients reported significantly higher levels of fatigue
two separate multivariate analyses.22 The first analysis
scores than the controls.
would create two groups based on a “fatigue worst”
Figure 2 indicates that a large proportion of con-
rating cut point of between 3 and 4. The second anal-
trols report scores at the lower end of the distribution
ysis would construct two groups based on a “fatigue
of BFI scores, whereas the scores of cancer patients
worst” rating cut point of between 4 and 5. The larger
tend to be distributed uniformly. A greater proportion
of the two F statistics (either from Hotelling, or Wilk,
of patients report higher levels of fatigue compared
or Pillai) would correspond to the better cut point.
with controls. These findings give evidence for the
We extended the preceding idea by using two cut
sensitivity of the BFI with cancer patients.
points instead of one. Thus, instead of creating two
groups based on “mild” and “nonmild,” we created
three groups based on “mild,” “moderate,” and “se- Establishing the Validity of the Brief Fatigue Inventory
vere.” The larger F statistic from MANOVA should be The validation of the BFI was carried out with the
associated with the cut points that maximally discrim- patient data. Construct validity was shown by factor
inate fatigue severity. In this analysis, we performed analysis. Concurrent validity was demonstrated by
four separate MANOVAs using the six interference correlating the BFI with commonly used fatigue mea-
items (relationship with others, work, enjoyment of sures, such as the fatigue subscales of the POMS and
life, walk, mood, activity) as dependent variables and FACT, the POMS-Vigor, and the nonfatigue items of
the three levels of fatigue severity (mild, moderate, the FACT. Finally, discriminant validity was presented
severe) as the between-groups factor. by comparing the BFI scores of patient groups who
were expected to have differing levels of fatigue.
RESULTS
Preliminary analyses indicated that several items of Construct Validity
the fatigue questionnaire did not discriminate be- Factor analysis is a multivariate procedure which al-
tween controls and patients or did not vary within the lowed us to determine the underlying constructs mea-
patient group and were therefore dropped from the sured by the items in the BFI. This procedure identi-
final nine-item BFI. The eliminated items included fied a single underlying construct among the nine BFI
those concerning sleep quality and also the items items. The factor loadings were high, and ranged from
“least fatigue” and “ability to think clearly.” Table 2 0.81 for usual fatigue to 0.92 for activity. This pattern
 Rapid Assessment of Fatigue/Mendoza et al. 1193

of factor loadings is indicative of the association of the TABLE 3

nine BFI items with a single factor. Means and Standard Deviations of the Brief Fatigue Inventory,
Profile of Mood States, and Functional Assessment of Cancer Therapy
across Performance Status Categories
Single-Factor Model Fit
ECOG 5 0 ECOG 5 1 ECOG 5 2
The screen plot was used to capture the correct num- Measure (n 5 55) (n 5 86) (n 5 154)
ber of factors. Clearly, the eigenvalues of 6.9, 0.57, and
0.36 for the first three factors indicate that most of the BFI* 2.3 (2.2) 4.0 (2.5) 6.0 (2.4)
data can be explained by a single construct. In fact, the Anemia specific items of the FACT* 60.2 (15.0) 45.7 (16.5) 32.1 (16.4)
Fatigue subscale of the FACT* 38.1 (11.4) 26.4 (13.2) 16.1 (12.5)
first factor explains about 75% of the variability in the
POMS-fatigue* 6.9 (6.8) 13.4 (7.6) 18.9 (7.8)
data. In addition, this single-factor model fits well POMS-vigor* 17.8 (8.1) 11.7 (7.2) 8.2 (5.5)
according to Harman’s24 rule that the standard devi-
ation of the residuals be slightly less than or approxi- ECOG: Eastern Cooperative Oncology Group; BFI: Brief Fatigue Inventory; FACT: Functional Assess-
mately equal to the standard error of a correlation ment of Cancer Therapy; POMS: Profile of Mood States.
* Significantly different at each level of ECOG performance status (P , 0.001).
coefficient, which is the reciprocal of the square root
of the sample size. In our case, the standard deviation
of the residuals is 0.04 compared with the standard
deviation of the correlation coefficient, which is 0.06. Discriminant Validity
Because the BFI measures a single construct, the Treatment- or disease-related anemia (represented by
arithmetic mean of the nine BFI items can be used as hemoglobin level) is commonly related to fatigue in
a global BFI score. If responses to several items are cancer patients. Due to the different hemoglobin stan-
missing, then a mean BFI score adjusted for the num- dards based on gender, we calculated separate stan-
ber of missing values can be computed. Ware et al.25 dard scores for males and females. By using these
recommended that a scale score be calculated if re- adjusted hemoglobin levels, the correlation between
spondents answered at least one-half of the items. the BFI and hemoglobin was statistically significant
Thus, if an individual responds to the majority of the (r 5 20.36, P , 0.001). Hemoglobin also was signifi-
items (at least five items on the BFI), then a mean BFI cantly related to both the Fatigue subscale of the FACT
score can be computed. However, if less than five (r 5 0.38, P , 0.001) and the POMS-Fatigue (r 5
items are completed, then it is inappropriate to com- 20.34, P , 0.001).
pute a mean BFI score. Discriminant validity also was examined by com-
paring the BFI scores of groups formed based on other
variables (i.e., ECOG performance status, FACT-Fa-
Factor Analysis of the Profile of Mood States and
tigue subscale, and POMS-Fatigue) that were expected
Functional Assessment of Cancer Therapy-Fatigue Scales
to be associated with severity of fatigue. The ECOG
Separate factor analyses of the items from the fatigue
performance status rating is widely used to assess
subscale of the FACT and the items from the POMS
disease severity.26 Patients with more severe disease
fatigue scale also resulted in a single factor for each
were expected to have greater levels of fatigue. Table 3
scale. Factor analysis of the items comprising both the
shows that the mean BFI scores, in fact, were signifi-
Vigor and Fatigue subscales of the POMS yielded two
cantly different across different performance status
factors. Similarly, a factor analysis of the items in both
ratings based on the ECOG scale. The two other fa-
the fatigue subscale and the anemia-specific items of
tigue instruments, the Fatigue subscale of the FACT
the FACT resulted in two factors.
and the POMS-Fatigue, showed similar discriminant
validity.
Concurrent Validity
This type of validity was established by showing that
the BFI is closely related to existing instruments that Reliability of the Brief Fatigue Inventory
measure fatigue. Two previously validated measures, To show that the nine items comprising the BFI are
POMS-Fatigue and the Fatigue subscale of the FACT, reliable, we calculated the Cronbach’s coefficient al-
were used in this analysis. The BFI was significantly pha for these items. Coefficient alphas range from 0 to
correlated with both the FACT (r 5 20.88, P , 0.001) a maximum of 1, with higher values indicating little
and the POMS (r 5 0.84, P , 0.001) Fatigue subscales. measurement error. Individual alphas for each item (if
The FACT and POMS Fatigue subscales also were sig- deleted) were 0.95 for both activity and work and 0.96
nificantly correlated with one another (r 5 20.92, P , for the remaining items. The internal consistency co-
0.001). efficient of 0.96 supports the reliability of the BFI.
1194 CANCER March 1, 1999 / Volume 85 / Number 5

TABLE 4
F Statistics for Fatigue Severity Levels Based on Various Test Criteria in Multivariate Analyses of Variance

Possible boundary models

Model Mild Moderate Severe Pillai’s trace Wilk’s lambda Hotelling’s trace

1 1–3 4–6 7–10 19.18 25.10 31.53

2 1–4 5–6 7–10 19.04 24.66 30.73
3 1–4 5–7 8–10 17.17 21.68 26.49
4 1–3 4–7 8–10 16.78 21.01 25.51

Physiologic Predictors of Fatigue

Due to potentially different etiologies of fatigue with
different cancers, the patient group was divided into
two different diagnostic categories. The first group
was made up of those with solid tumors, such as
breast and lung carcinoma. The second group was
made up of those diagnosed with hematologic malig-
nancies, such as leukemia, lymphoma, and multiple
myeloma. This categorization was made to examine
the correlation between laboratory values and fatigue
in these two different groups.
Laboratory data collected from the checklist in-
cluded hemoglobin level, white blood cell count,
platelets, total and absolute neutrophil count, and
factors related to nutritional status, such as lactic de-
FIGURE 3. Plot of mean Brief Fatigue Inventory (BFI) interference items
hydrogenase, alanine amino transferase, bilirubin,
(composite of 6 interference items) against fatigue severity measured by
blood urea nitrogen, creatinin, and serum albumin.
“fatigue worst”. Note the slopes between 3 and 4 and between 6 and 7.
These laboratory data were used in an exploratory
regression analysis to determine which of these vari-
ables would significantly predict BFI scores. In the
Categorizing Severity of Fatigue
We wanted to determine whether there were catego-
exploratory analyses, stepwise, backward, and forward
ries of fatigue severity that might be described as
regressions arrived at the same results.
“mild,” “moderate,” or “severe.” We explored these
In the hematologic group, both albumin and
models, as described above in Methods, based on
hemoglobin were significant predictors of fatigue. Al-
combinations of numerical scale cut points that had
bumin, with a standardized coefficient of 20.4, is a
been useful previously in categorizing pain severity.
stronger predictor of fatigue than hemoglobin (stan-
Table 4 shows that the cut point categories shown in
dardized coefficient of 20.2). Hemoglobin levels, after model 1 were optimal based on the agreement of the
adjustment for the effect of gender, were converted multivariate criteria described in Methods: 1–3 for
back to raw data values for easier interpretation. A mild, 4 – 6 for moderate, and 7–10 for “severe” fatigue.
unit drop in hemoglobin (g/dL) would increase the However, when we examined the interference for the
BFI score by 0.3 unit provided albumin level is held lower boundaries, we found that these groupings did
constant. A unit drop in albumin (g/dL) would in- not discriminate as well as the groupings for the upper
crease the BFI score by 1.7 units provided that hemo- boundaries. The two models with the highest F values,
globin level is held constant. For the hematologic models 1 and 2, indicate that the cut point for “severe”
group, this two-predictor regression model (hemoglo- fatigue was consistently between 6 and 7.
bin and albumin) explained 26% of the variance. Figure 3 presents mean BFI interference score
In the solid tumor group, only albumin was a (composite of six items) graphed against “fatigue
significant predictor of fatigue. A unit drop in albumin worst.” The optimal cut points are associated with
level would increase the BFI score by as much as 2.4 large increases in interference. As fatigue severity in-
units. This regression model accounted for 20% of the creases, so does fatigue interference. However, the
variance in the solid tumor group. rate of increase (slope) changes as both of these items
 Rapid Assessment of Fatigue/Mendoza et al. 1195

increase. For instance, the steepest slope was between sional measures also are probably too long to be com-
6 and 7. This suggests that, for every unit increase in pleted when fatigue is of interest in a clinical trial. The
fatigue severity, the increase in interference was larger inclusion of long assessment instruments in such tri-
between 6 and 7 than anywhere else along the “fatigue als often leads to missing data, because patients find
worst” continuum. The slope between 3 and 4 was not them difficult to complete or because it is difficult to
as steep as the slope between 6 and 7. This indicates schedule sufficient time for their administration.
that, although it can be said that 7–10 is the optimal These longer scales probably will find their application
“severe” level of fatigue, the steepness of the slope for in descriptive studies of fatigue, in which they can be
the lower boundary model is less pronounced. At this given in a one-on-one situation with study personnel
point, finding the optimal cut point for “mild” and and when enough time can be scheduled for their
“moderate” fatigue severity should be investigated completion.
further. The three fatigue assessment tools presented here
could serve to rapidly identify those patients with clin-
DISCUSSION ically-significant fatigue. In this study, we examined
The nine-item BFI taps into a single dimension that BFI scores to identify cut points that might discrimi-
can be thought of as the subjective report of fatigue nate between patients with different levels of fatigue
severity. It is a reliable (internally stable) instrument severity. We used the same approach that we had used
that is correlated with measures of performance status previously to identify “mild,” “moderate,” and “se-
(patients who are more ill report higher levels of fa- vere” levels of pain. This involved comparing fatigue
tigue) and with physiological markers of anemia (he- severity ratings with the patient’s ratings of how fa-
moglobin) and nutritional status (albumin) known to tigue interfered with common functional domains.
be associated with fatigue. With pain severity, three distinct groupings (mild,
The performance of the BFI in this study is virtu- moderate, and severe) could be formed, and the cor-
ally interchangeable with a 13-item fatigue scale that relation between functional impairment and pain se-
is a part of the FACT system of assessment as well as verity was nonlinear. The three groupings of pain se-
with the Fatigue scale of the POMS. Both of these verity clearly stood apart from one another. The
other instruments also demonstrate a single factor correlation between fatigue severity rating and func-
that can be thought of as assessing severity of fatigue. tional interference, however, is more linear at the
All three scales are easily administered, and patients lower end of the range of fatigue scores; therefore, and
can complete them quickly, suggesting that they can the development of cut points between mild and mod-
be used for clinical screening as well as for outcome erate categories of fatigue severity is less straightfor-
measures in clinical trials in which fatigue severity is ward.
an outcome of interest. We found that fatigue severity ratings could be
The BFI, however, may have some advantages thought of as forming two groups, “severe” and “non-
over the other two measures. Its use of simple, single- severe,” with those patients rating their worst fatigue
word designations of fatigue severity levels and func- at a 7 or greater as having “severe” fatigue. By using
tional domains makes it very easy to understand. this range (7–10) to indicate severe fatigue, 35% of our
Based on our experience with the BPI, translation of patient population and only 5% of our community
the BFI into other languages should be simple and control population would be identified as having se-
straightforward. Many patients find the response rat- vere fatigue. It is interesting to note that, with pain
ing system (which uses 0 –10 scales) easy to under- severity ratings, 7–10 also defines a group with severe
stand, and this type of rating fits in well with the 0 –10 pain.20 Clearly, those rating their fatigue severity at a 7
ratings of symptom severity that are often used in or greater are very tired and most probably need some
clinical practice. type of clinical intervention to improve their function.
The BFI and the other two measures used in this However, because of the linear correlation between
study do not capture the multiple dimensions that fatigue severity and functional interference found at
longer instruments were designed to represent, such the lower end of fatigue severity, we suggest that the
as the cognitive, affective, and somatic components of cut points between “mild” and “moderate” fatigue
fatigue. These multidimensional scales, however, are should be regarded as provisional in nature.
often too long for tired patients to complete. In clinical Fatigue is endemic in those with cancer. Most of
practice, one might wish to screen for patients with those working in oncology are familiar with patients
high levels of fatigue based on a measure such as the who are totally disabled by their fatigue. Because of
BFI, then do additional assessments to determine the the few current treatments available, fatigue is more
causes of fatigue in those patients. The multidimen- difficult to treat than pain. There are few clinical trials
1196 CANCER March 1, 1999 / Volume 85 / Number 5

that assess fatigue as an outcome, and fatigue as a side Weller S. The development of an instrument to measure the
effect of therapy is often only crudely assessed. The subjective dimension of fatigue. Funk S, Tornquist E, Cham-
pagne M, Archer CL, Wiese R, editors. Key aspects of com-
use of simple, easily administered, and easily scored
fort: management of pain, fatigue and nausea. New York:
fatigue scales should open the way for epidemiologic Springer, 1989;199 –208.
studies on fatigue, should improve communication 13. Smets EM, Garssen B, Bonke B, De Haes JC. The multidi-
about fatigue between patients and those who care for mensional fatigue inventory (MFI) psychometric qualities of
them, and should facilitate clinical trials that are fo- an instrument to assess fatigue. J Psychosom Res 1995;39(5):
315–25.
cused on the development of new treatments for fa-
14. Stein KD, Martin SC, Hann DM, Jacobsen PB. A multidimen-
tigue. sional measure of fatigue for use with cancer patients. Can-
cer Practice 1998;6(3):143–52.
REFERENCES 15. Hann DM, Jacobsen PB, Martin SC, Kronish LE, Azzarello
1. Glaus A, Crow R, Hammond S. A qualitative study to explore LM, Fields KK. Fatigue in women treated with bone marrow
the concept of fatigue/tiredness in cancer patients and in transplantation for breast cancer: a comparison with
healthy individuals. Support Care Cancer 1996;4:82–96. women with no history of cancer. Support Care Cancer
2. Richardson A. Fatigue in cancer patients: a review of the 1997;5:44 –52.
literature. Eur J Cancer Care 1995;4:20 –32. 16. Cleeland CS. Measurement of pain by subjective report.
3. Blesch KS, Paice JA, Wickham R, Harte N, Schnoor D, Paul S, Chapman CR, Loeser JD, editors. Issues in pain measure-
et al. Correlates of fatigue in people with breast or lung ment. Vol. 12 of advances in pain research and therapy. New
cancer. Oncol Nurs Forum 1991;18(1):81–7. York: Raven Press, 1989;391– 403.
4. Irvine D, Vincent L, Graydon JE, Bubela N, Thompson L. The 17. Cleeland CS, Gonin R, Hatfield AK, Edmonson JH, Blum RH,
prevalence and correlates of fatigue in patients receiving Stewart JA, et al. Pain and its treatment in outpatients with
treatment with chemotherapy and radiotherapy: a compar- metastatic cancer. New Engl J Med 1994;330:592–96.
ison with the fatigue experienced by healthy individuals. 18. Larue F, Colleau SM, Brasseur L, Cleeland CS. Multicentre
Cancer Nurs 1994;17(5):367–78. study of pain and its treatment in France. BMJ 1995;310:
5. Winningham ML, Nail LM, Burke MB, Brophy L, Cimprich B, 1034 –37.
Jones LS, et al. Fatigue and the cancer experience: the state 19. Wang XS, Mendoza TR, Gao SZ, Cleeland CS. The Chinese
of the knowledge. Oncol Nurs Forum 1994;21(1):23–36. version of the Brief Pain Inventory (BPI-C): its development
6. Nail L, Winningham M. Fatigue. Groenwald SL, Froggr M, and use in a study of cancer pain. Pain 1996;61:407–16.
Goodman M, Yarbro C, editors. Cancer nursing: principles 20. Serlin RC, Mendoza TR, Nakamura Y, Edwards KR, Cleeland
and practice. 3rd ed. Boston: Jones & Bartlett, 1993;608 –19. CS. When is cancer pain mild, moderate or severe? Grading
7. Schwartz JE, Jandorf L, Krupp LB. The measurement of pain severity by its interference with function. Pain 1995;61:
fatigue: a new instrument. J Psychosom Res 1993;37(7):753– 277– 84.
62. 21. Cleeland CS. Pain assessment in cancer. Osoba D, editor.
8. Irvine DM, Vincent L, Bubela N, Thompson L, Graydon JE. A Effect of cancer on quality of life. Boca Raton, FL: CRC Press,
critical appraisal of the research literature investigating fa- Inc., 1991;293–305.
tigue in the individual with cancer. Cancer Nurs 1991;14(4): 22. Srivastava MS, Carter EM. An introduction to applied mul-
188 –99. tivariate statistics. New York: North Holland, 1983.
9. Pearson PG, Byars G. The development and validation of a 23. Marascuilo LA, Levin JR. Multivariate statistics in the social
checklist measuring subjective fatigue: report no. 56. Ran- sciences: a researcher‘s guide. Monterey, CA: Brooks/Cole
dolph AFB, TX: USAF School of Aviation, 1956;56 –115. Publishing Company, 1983.
10. McNair DM, Lorr M, Droppleman LF. EdITS manual for the 24. Harman HH. Modern factor analysis. 2nd edition. Chicago:
profile of mood states. San Diego, CA: Educational and University of Chicago Press, 1967.
Industrial Testing Service, 1971;8. 25. Ware JE, Snow KK, Kosinski M, Gandek MS. SF-36 health
11. Yellen SB, Cella DF, Webster K, Blendowski C, Kaplan E. survey manual and interpretation guide. Boston, MA: Nim-
Measuring fatigue and other anemia-related symptoms with rod Press, 1993.
the Functional Assessment of Cancer Therapy (FACT) mea- 26. Oken M, Creech RTD, Horton J, Davis T, McFadden E,
surement system. J Pain Symptom Manage 1997;13(2):63– Carbone P. Toxicity and response criteria of the Eastern
74. Cooperative Oncology Group. Am J Clin Oncol 1982;5:
12. Piper BF, Lindsey AM, Dodd MJ, Ferketich S, Paul SM, 649 –55.