Validity and Reliability of a New Instrument to Measure Cancer-Related Fatigue in Adolescents

Abstract

Adolescents undergoing treatment for cancer rate fatigue as their most prevalent and intense cancer- and treatment-related effect. Parents and staff rate it similarly. Despite its reported prevalence, intensity, and distressing effects, cancer-related fatigue in adolescents is not routinely assessed during or after cancer treatment. We contend that the insufficient clinical attention is primarily due to the lack of a reliable and valid self-report instrument with which adolescent cancer-related fatigue can be measured. Our aim was to determine the reliability and construct validity of a new instrument and its ability to measure change in fatigue over time. Initial testing involved 64 adolescents undergoing curative treatment of cancer who completed the Fatigue Scale-Adolescent (FS-A) at two to four key points in treatment in one of four studies. Internal consistency estimates ranged from 0.67 to 0.95. Validity estimates involving the FS-A with the parent version ranged from 0.13 to 0.76; estimates involving the staff version and the Reynolds Depression Scale were 0.27 and 0.87 respectively. Additional validity findings included significant fatigue differences between anemic and non-anemic patients (P = 0.042) and the emergence of four factors in an exploratory factor analysis. Findings further indicate that the FS-A can be used to measure change over time (t = 2.55, P <0.01). In summary, the FS-A has moderate to strong reliability and impressive validity coefficients for a new research instrument.

Introduction

Fatigue has been identified by adolescents as the most distressing of 10 cancer-related effects at four sequential time points during the first six months of cancer treatment,^1-3 as the most frequent of 29 acute cancer-related symptoms,⁴ and as persisting symptom from months to years after successful completion of curative therapy.^5,6 Adolescents with a central nervous system neoplasm identified it as their most common daytime sleep-related complaint.⁷ Parents of pediatric oncology patients at 22 cancer centers in the United Kingdom also identified cancer-related fatigue as concerning them “quite a bit” or “very much,” with more than half reporting that their child or adolescent experienced this kind of fatigue “at least once a week,” “on most days,” or “every day.”⁸ Parents whose child or adolescent died of cancer or its treatment effects identified fatigue as the most frequent symptom during their child’s end of life (final 30 days of life).⁹ Despite its reported prevalence, intensity, and distressing effects for adolescent survivors of childhood cancer and their parents, cancer-related fatigue in adolescents is not routinely assessed during or after cancer treatment, as recommended in the evidence-based fatigue guidelines from the National Comprehensive Cancer Network.¹⁰ We contend that lack of clinical attention is primarily due to the lack of a reliable and valid instrument to measure adolescent cancer-related fatigue.

Accurate, sensitive, and quick measures of this form of fatigue might prompt much-needed routine assessments. Also, having a developmentally appropriate instrument would help clinicians and researchers implement and evaluate interventions to prevent or reduce the intensity of this troubling symptom. Here, we report the development and testing of a new instrument, The Fatigue Scale-Adolescent (FS-A), which was specifically created to comprehensively measure cancer-related fatigue in adolescents who are receiving cancer treatment. More specifically, we sought to estimate the internal consistency and construct validity of the FS-A, as well as its ability to measure change in fatigue over time. The four studies involved in the testing of the FS-A were approved by the institutional review boards of St. Jude Children’s Research Hospital and Texas Children’s Cancer Center. Written consent from parents and assent from adolescents were documented in each of the four studies

Cancer-related fatigue in children and adolescents is currently being measured by using single-item, two-item, or multi-item approaches and four different multi-symptom, self-report instruments: the Memorial Symptom Assessment Scale (MSAS), the Symptom Distress Scale (SDS), the Symptom Diary, and the Multidimensional Fatigue Scale. In the MSAS, cancer-related fatigue is labeled “lack of energy.” Patients aged 10 to 18 years who indicate a lack of energy are prompted to respond to items regarding the frequency, severity, and distress related to the lack of energy.⁴ The cancer-related fatigue item in the 10-item SDS reads: “Please put a circle around the number that most closely measures how tired you are feeling today,” on a scale of 1 (“I am not tired at all.”) to 5 (“I could not feel more tired.”).¹ Two of the Symptom Diary items measure morning and evening tiredness of children and adolescents on a scale of 0 (not tired) to 4 (most tired).^12-14 The 18-item Multidimensional Fatigue Scale, a 5-point Likert-type instrument, has matching child and parent versions ¹⁵. None of the four measurement approaches was linked to an over-arching conceptual model or framework. The exact percentage of adolescents that made up study samples was not reported. Also not reported were findings specific to the presence and intensity of fatigue in adolescents or the ability of the instruments to measure change in fatigue over time.

Scores from the single- or two-item measures indicate that cancer-related fatigue is more prevalent than the other cancer-related symptoms measured by the same instruments; therefore, the single or two-item approach may be able to serve as an initial clinical screening for the presence of fatigue. However, assessing the reliability and validity of a single item is difficult, and only limited information can be obtained by using standard psychometric assessments. Additionally, conceptual analyses of adolescent cancer-related fatigue indicate that it has multiple conceptual domains.^15-18 Davies et al.¹⁶ labeled their inductively derived dimensions as typical tiredness (normal ebb and flow of energy that can be low but replenished readily), treatment fatigue (energy loss that is greater than its replenishment but the energy can be conserved), and shutdown fatigue (profound and sustained loss of energy). Varni et al.¹⁵ referred to general fatigue, sleep or rest fatigue, and cognitive fatigue, although they did not define each dimension of child or adolescent fatigue. A single item cannot be used to capture multiple conceptual domains or to produce a comprehensive measured outcome; therefore, the resulting fatigue score may not be adequate to guide care interventions.

Methods and Materials

Phase I: Development of the Fatigue Scale-Adolescent Instrument

We developed the FS-A in two phases. During the qualitative phase (Phase I), we conducted individual and focus group interviews to learn first-hand about the fatigue experienced by adolescents undergoing cancer treatment. To establish content validity of the instrument, we relied on expert panel review (health care professionals and adolescents with cancer). Then, during the preliminary testing phase (discussed later), we estimated reliability (internal consistency) and construct validity.

Defining Cancer-Related Fatigue in Adolescents

Invitations to participate in focus group discussions regarding cancer-related fatigue were extended to all 13- to 18-year-old patients receiving treatment as an inpatient or outpatient at one of our pediatric cancer centers during the seven-day data-collection period. Fifteen adolescents (10 girls, 67%; nine whites, 60%; eight with a diagnosis of leukemia or lymphoma (53%) participated in one of several focus groups. Two additional eligible participants declined to take part. Time since diagnosis ranged from one month to 59 months. Nine questions were used to solicit the adolescents’ perspectives regarding the characteristics and contributory and alleviating factors of their cancer-related fatigue.¹⁹ We asked these questions because the underlying conceptual model, which was qualitatively induced from interview data, indicated that cancer-related fatigue could result from disease or its treatment, and that it could be modified by patient, family, or health care provider care actions. Subsequently, 15 different adolescents receiving treatment during one randomly selected day at one treatment center were asked the same nine questions, individually so that the study team could determine whether the characteristics of cancer-related fatigue reported by adolescents differed by method (focus group versus individual interview). Interview data were analyzed using the pragmatic and semantic content analyses described by Krippendorff.²⁰ (The pragmatic approach classifies data for probable cause or effect; the content analysis method classifies data for meaning.) The sampling unit was the adolescent’s response to the specific question posed by the focus group facilitator or the individual interviewer. The recording unit was the sentence in each sampling unit. The pre-established agreement level of 90% among four independent raters for each of the individual interviews and focus group data was achieved consistently. (The agreement level was calculated as the number of coding agreements divided by the total number of agreements and disagreements.)

The definition of adolescent cancer-related fatigue qualitatively induced from the focus group and individual interview data was as follows: a complex changing state of exhaustion that at times seems to be a physical condition, at other times a mental state, and at other times a combination of physical, emotional and mental tiredness. This fatigue can be acute, episodic, or chronic.¹⁷ Importantly, the same characteristics of cancer-related fatigue were identified by both research methods, but one difference was noted: Codes related to anger and other mental changes emerged from the focus group interviews, whereas codes related to sadness rather than anger were more prevalent in the individual interviews.¹⁸

The conceptual definition of adolescent cancer-related fatigue and the coded interview data were used as the basis for items on the quantitative instrument. The FS-A was designed to measure fatigue such that one to two items for each identified characteristic were included in the instrument. The content validity index²¹ was used to rate the content relevance of the developed items on a 5-point ordinal scale (1=irrelevant and five=extremely relevant). Two panels rated the items: a panel of five pediatric oncology nurses and a panel of four adolescents. Items with a score of 3 or higher were included in the final 17-item instrument. As a result, three items were excluded. The final set of items is in Figure 1. Higher scores indicate higher fatigue. Adolescents complete the FS-A on their individually within 3 to 4 minutes.

Phase II: Testing of the FS-A Instrument

The FS-A was tested in four separate studies involving four distinct groups of adolescent oncology patients receiving treatment at St. Jude Children’s Research Hospital or Texas Children’s Cancer Center. We had no missing data. We conducted the first study, Measuring Fatigue in Childhood Cancer (MFCC), to estimate the internal consistency of the FS-A and its construct validity by testing for associations between measures of individual items on the FS-A and those of the parent and staff fatigue reports. Both the Fatigue Scale-Parent (FS-P) and the Fatigue Scale-Staff (FS-S) were designed to measure parent and staff reports of fatigue in the adolescent being treated for cancer. Scores on the FS-A were also tested for associations with measures of depression on the Reynolds Depression Scale.²² All three respondent groups completed the study instruments at two time points: the beginning (T1) and end (T2) of reinduction therapy (a period of 6 to 7 weeks) for acute lymphoblastic leukemia (ALL) (Table 1).

Table 1.

Number of Adolescents, Parents, and Staff Completing Study Instruments by Study

	Adolescents (FS-A)				Parents (FS-P)				Staff (FS-S)
Study	T1	T2	T3	T4	T1	T2	T3	T4	T1	T2	T3	T4
MFCC	21	16	-	-	20	13	-	-	18	9	-	-
SLEEP	15	14	15	15	14	14	15	15	-	-	-	-
SLEEP2	11	11	10	9	10	11	10	8	-	9	9	8
CLUSTERS	17	14	17	-	17	14	16	-	-	-	-	-
Total	64	55	42	24	61	52	41	23	18	18	9	8

MFCC = Measuring Fatigue in Childhood Cancer; SLEEP = Sleep, Fatigue and Dexamethasone in Childhood Acute Lymphocytic Leukemia; SLEEP2 = Sleep, Fatigue and Enhanced Physical Activity in Hospitalized Oncology Patients; CLUSTERS = Symptom Clusters in Pediatric Oncology; - signifies that the variable was not measured at this time point or in the particular study.

We conducted the second study, Sleep, Fatigue and Dexamethasone in Childhood Acute Lymphoblastic Leukemia (SLEEP), to determine the relationship between exposure to dexamethasone and patients’ sleep and fatigue. This was accomplished by studying two consecutive five-day periods during continuation therapy for childhood ALL. Adolescents and their parents completed the fatigue instruments (FS-A and FS-P) at four time points spanning a contiguous 10-day period during ALL maintenance therapy (T1 – Day 2; T2 – Day 5 of the 5-day period before the dexamethasone pulse; T3 – Day 2; and T4 – Day 5 of the contiguous 5-day period during the dexamethasone pulse (Table 1).

Our aim in the third study, Sleep, Fatigue and Enhanced Physical Activity in Hospitalized Pediatric Oncology Patients (SLEEP2), was to measure sleep and fatigue in adolescents hospitalized for two to four days of chemotherapy for a solid tumor or acute myeloid leukemia (AML). Adolescents, their parents, and their nurses completed their respective fatigue instruments in mid- to late afternoon of each day during the inpatient stay (Table 1). Sleep quality indicators (sleep efficiency, sleep duration, and nocturnal awakenings) were measured by actigraphy. Time points for data collection were on the day of hospital admission (T1) and on each subsequent day of hospitalization (T2-T4).

The fourth study, Symptom Clusters in Pediatric Oncology (CLUSTERS), was conducted to examine the symptom cluster (fatigue, nausea, and vomiting), and sleep disturbances experienced by adolescent oncology patients who received doxorubicin, cisplatin, and/or ifosfamide chemotherapy. Adolescents completed the FS-A at three different points in treatment during their first course of chemotherapy (T1 was the first day of chemotherapy; T2, the final day of chemotherapy in that same course; and Time 3, a week after Time 2).

In the first three studies (MFCC, SLEEP, and SLEEP2), adolescents also voiced their opinions regarding perceived causes of their fatigue by selecting from options on a 10-item list (Figure 2). The items were rated on a 5-point scale, with 1 meaning the item did “not at all” cause fatigue and 5 meaning the item caused fatigue “all the time.”

Data Analysis

Cronbach’s alpha was calculated as an estimate of the internal consistency of the FS-A, FS-P, and FS-S. Individual FS-A items were examined for their contribution to the overall estimate of internal consistency. Construct validity was estimated by using the known groups method (anemic vs. non-anemic), an exploratory factor analysis using a principal components factor extraction with varimax orthogonal factor rotation (SAS Inc., Cary, NC). We also tested for associations between items on the FS-A, the FS-P, the FS-S, and the Reynolds Depression Scale. ^23,24 The ability of the FS-A to measure change over time in cancer-related fatigue was estimated by designating the baseline for each study to be Time 1 and a subsequent data collection point to be the endpoint — referred to in this change analysis as “Time 2.” In the MFCC study, Time 1 was the first day of re-induction, and Time 2 was at the end of reinduction. In the SLEEP study, Time 1 was the second day of the first 5-day period, and Time 2 was the second day of the second 5-day period. In the SLEEP2 study, Time 1 was the day of admission to the hospital, and Time 2 was the third day of hospitalization. In the CLUSTERS study, Time 1 was the first day of chemotherapy, and Time 2 was a week after the final dose of chemotherapy in that same course. Finally, a generalized linear model (PROC GLM) was used to determine whether gender, diagnosis (ALL vs. solid tumor vs. all other diagnoses), and race (white vs. black vs. all others) were associated with the FS-A scores at Time 1.

Results

Sample

A total of 64 adolescents, 61 parents, and 18 staff participated in the four studies (Table 1). Of the participating adolescents, the majority were male (56.2%), receiving treatment for ALL (60.9%), and white (67.2%). The overall mean (SD) age was 15.32 years (1.52), similar to the mean age within each of the four studies. In all four studies, the age range of participants was 13 to 18 years. Across the four studies, only one study (SLEEP2) did not have the same approximate ratio of male to female participants. Only ALL patients were enrolled on two studies (MFCC and SLEEP), while patients being treated for diverse types of cancer were enrolled on the other two studies. The numbers of those participating in each study differed slightly by race (Table 2).

Table 2.

Characteristics of Adolescents by Study

		Study Groups				Total
		MFCC	SLEEP	SLEEP2	CLUSTERS
Gender		9 (42.86)	8 (53.33)	3 (27.27)	8 (47.06)	28 (43.75)
Female	n (%)
Male	n (%)	12 (57.14)	7 (46.67)	8 (72.73)	9 (52.94)	36 (36.25)
Diagnosis		21 (100.0)	15 (100.0)	0 (0)	3 (17.65)	39 60.94)
ALL	n (%)
AML	n (%)	0 (0)	0 (0)	3 (27.27)	0 (0)	3 (4.69)
HD/Lymphoma	n (%)	0 (0)	0 (0)	0 (0)	6 (35.29)	6 (9.38)
Solid Tumor	n (%)	0 (0)	0 (0)	8 (72.73)	8 (47.06)	16 (25.00)
Race		3 (14.29)	4 (26.67)	0 (0)	3 (17.65)	10 (15.63)
Black	n (%)
Hispanic	n (%)	0 (0)	0 (0)	0 (0)	6 (35.29)	6 (9.38)
Other	n (%)	1 (4.76)	2 (13.33)	2 (18.18)	0 (0)	5 (7.81)
White	n (%)	17 (80.95)	9 (60.00)	9 (81.82)	8 (47.06)	43 (67.19)
Age
	n	21	15	11	17	64
	Mean	15.46	14.82	15.68	15.35	15.32
	SD	1.62	1.60	1.48	1.37	1.52
	Min	13.08	12.75	13.84	13.00	12.75
	Median	15.52	14.83	16.01	16.00	15.06
	Max	18.26	18.14	18.16	17.00	18.26

ALL = acute lymphoblastic leukemia; AML = acute myelogenous leukemia; HD = Hodgkin’s disease; SD = standard deviation.

AU: PLS CHECK EXPANSIONS OF ALL, AML, HD AND SD FOR ACCURACY.

Item Analysis

Of the 14 items on the FS-A, the four with the highest mean scores were items 1 (“My body has felt tired.”), 4 (“I want more rest.”), 6 (“It’s harder to keep up with schoolwork.”) and 9 (“I am able to do my usual activities.”). Item 9 was a reverse-coded item. The lowest mean score was for item 13 (“I don’t feel like being with others.”). Individual items were also examined to determine the number of patients who endorsed each response option and the number of patients who endorsed the higher fatigue intensity ratings of 4 or 5. The five items that had the higher fatigue intensity ratings were items 3 (“I move more slowly.”), 4, 6, 7 (“I don’t feel like doing much.”), and 9. The average total scale FS-A scores increased in the three studies over the multiple data collection points. Parent ratings reflect that same increase over time (Table 3). The perceived causes with the highest mean scores were items 2 (“Staying overnight in the hospital”), 3 (“Treatment”), 8 (“Being bored”), and 10 (“Feeling sick”).

Table 3.

Descriptive Statistics for Total Fatigue Scale Scores (FS-A, FS-P, FS-S)

		MFCC		SLEEP				SLEEP2				CLUSTERS
		Time		Time				Time				Time
		1	2	1	2	3	4	1	2	3	4	1	2	3
FS-A	N	21	16	15	14	15	15	11	11	10	9	17	15	17
	Mean	27.33	26.13	26.00	23.64	30.07	32.93	24.00	30.45	32.30	34.89	31.59	35.33	39.18
	SD	8.44	9.51	9.68	8.60	9.97	13.93	5.93	9.18	11.97	13.50	6.65	7.04	13.35
	Median	27.00	23.50	23.00	21.00	28.00	30.00	23.00	31.00	32.00	36.00	32.00	38.00	41.00
	Min	14.00	14.00	14.00	14.00	17.00	17.00	15.00	15.00	15.00	15.00	20.00	24.00	15.00
	Max	48.00	53.00	49.00	44.00	48.00	58.00	33.00	43.00	58.00	60.00	43.00	49.00	63.00
FS-P	n	20	13	14	14	15	15	10	11	10	8	17	15	16
	Mean	47.60	43.69	37.71	35.07	44.47	44.40	38.70	42.27	42.90	39.63	41.82	51.00	49.44
	SD	7.57	7.26	12.96	12.66	19.14	18.24	7.13	12.85	13.49	15.26	11.09	10.80	12.71
	Median	48.50	46.00	38.50	34.50	41.00	42.00	39.50	43.00	46.00	33.50	43.00	52.00	52.50
	Min	33.00	27.00	18.00	19.00	17.00	18.00	24.00	22.00	20.00	24.00	23.00	27.00	32.00
	Max	65.00	55.00	64.00	55.00	76.00	72.00	50.00	69.00	68.00	66.00	60.00	66.00	74.00
FS-S	n	18	9	—	—	—	—	—	9	9	8	—	—	—
	Mean	14.61	12.22	—	—	—	—	—	18.22	19.00	15.88	—	—	—
	SD	4.82	5.24	—	—	—	—	—	5.26	7.30	7.95	—	—	—
	Median	14.00	10.00	—	—	—	—	—	17.00	16.00	12.50	—	—	—
	Min	9.00	7.00	—	—	—	—	—	11.00	12.00	10.00	—	—	—
	Max	25.00	24.00	—	—	—	—	—	28.00	32.00	34.00	—	—	—

SD = standard deviation; — = not measured.

Internal Consistency

The FS-A had strong coefficient alpha estimates for 11 of the 13 data collection points, as did the FS-P. The Cronbach-if-item-deleted coefficients ranged from 0.597 to 0.956, indicating that two items (9 and 10) diminished the FS-A reliability at five of the 13 collection times. Item to total-scale correlations ranged from 0.24 to 0.92 in each study with a single exception: item 10 had a low correlation at one time point (−0.088) in one study. No correlation coefficient at the level of item-to-item comparison exceeded 0.80; therefore, multicolinearity among the items did not exist.²⁵ The FS-S had strong coefficients for all five data points, and the RDS had strong coefficients for both data points (Table 4).

Table 4.

Internal Consistency Estimates (Standardized Cronbach Alpha Coefficients)

	MFCC			SLEEP				SLEEP2			CLUSTERS
Study	T1	T2	T1	T2	T3	T4	T1	T2	T3	T4	T1	T2	T3
FS-A	0.81	0.93	0.90	0.89	0.91	0.95	0.76	0.85	0.93	0.93	0.67	0.67	0.93
FS-P	0.75	0.84	0.92	0.93	0.97	0.97	0.56	0.89	0.90	0.93	0.89	0.88	0.94
FS-S	0.85	0.97	—	—	—	—	—	0.86	0.95	0.95	—	—	—
RDS	0.87	0.87	—	—	—	—	—	—	—	—	—	—	—

— = not measured.

Validity

As theorized, the FS-A scores were more significantly correlated with the FS-P scores than with the FS-S scores (Table 5). Four factors were retained from the exploratory factor analysis because they met the proportion, eigenvalue, measure of sampling adequacy, and interpretability criteria.^26,27 The four factors are Cognitive and Physical Weariness, Added Effort and Assistance Needed to do Usual Activities, Needing Rest and Feeling Angry, and Avoiding Social Activities (Table 6). The final communality estimate total was 6.091, and the cumulative proportion of variance explained by the four retained factors was 1.0232. All items loaded onto the four rotated factors; correlation coefficients ranged from .353 to .758. Also as theorized, the correlations between the total scale scores of the RDS and the FS-A were strong but low to moderate among the factors of both instruments. The RDS factor of anhedonia had particularly low and nonsignificant correlations with the FS-A factors (Table 7).

Table 5.

Correlation Coefficients Among the Adolescent (FS-A), Parent (FS-P), and Staff (FS-S) Fatigue Reports by Study at Baseline (T1)

		FS-A	FS-P
	FS-A	—	0.7578 (P=0.0001) n=20
MFCC	FS-P	0.7578 (P=0.0001) n=20	—
	FS-S	0.2691 (P=0.2803) n=18	0.1528 (P=0.5450) n=18
SLEEP2	FS-A	—	0.13448 (P=0.7111) n=10
SLEEP	FS-A	—	0.65159 (P=0.0116) n=14
CLUSTERS	FS-A	—	0.1963 (P=0.4502) n=17
	FS-A	—	0.4667 (P<0.0001) n=61
All Patients	FS-P	0.54283 (P<0.0001) n=61	—
	FS-S	0.2691 (P=0.2803) n=18	0.1528 (P=0.5450) n=18

— = not measured.

Table 6.

Exploratory Factor Analysis Using the FS-A Scores from Baseline (T1)

Factor	Factor Label	Items/Factor Loading
1	Cognitive and Physical Weariness	8/0.714 2/0.705 6/0.389
2	Added Effort and Assistance Needed to Do Usual Activities	3/0.609 9/0.474 12/0.699 14/0.559
3	Needing Rest and Feeling Angry	1/0.549 4/0.500 10/0.492 11/0.325
4	Avoiding Social Activities	5/0.448 7/0.648 13/0.536

Table 7.

Correlation Between the Factors of the Reynolds Depression Scale and of the FS-A in the MFCC study (T1: n=21)

			FS-A (T1, n=21)
Reynolds Depression Scale	Total Score	Factor 1	Factor 2	Factor 3	Factor 4
		Cognitive and Physical Weariness	Added Effort and Assistance Needed to do Usual Activities	Needing Rest and Feeling Angry	Avoiding Social Activities
Total score	0.7113 (P=0.0003)	0.6936 (P=0.0005)	0.5481 (P=0.0101)	0.3362 (P=0.1362)	0.6030 (P=0.0038)
Factor 1 Generalized Demoralization	0.4813 (P=0.0272)	0.4150 (P=0.0614)	0.5014 (P=0.0206)	0.1222 (P=0.5978)	0.4207 (P=0.0576)
Factor 2 Despondency and Worry	0.5752 (P=0.0064)	0.6061 (P=0.0036)	0.5102 (P=0.0181)	0.1542 (P=0.5044)	0.5109 (P=0.0179)
Factor 3 Somatic Vegetative	0.5811 (P=0.0057)	0.4840 (P=0.0262)	0.3264 (P=0.1487)	0.5249 (P=0.0146)	0.4057 (P=0.0681)
Factor 4 Anhedonia	0.5383 (P=0.0118)	0.6584 (P=0.0012)	0.2542 (P=0.2661)	0.2899 (P=0.2024)	0.5000 (P=0.0210)
			FS-A (T2, n=15)
Reynolds Depression Scale	Total Score	Factor 1	Factor 2	Factor 3	Factor 4
Total score	0.7124 (P=0.0029)	0.4396 (P=0.1011)	0.5846 (P=0.0221)	0.5824 (P=0.0227)	0.8039 (P=0.0003)
Factor 1	0.2531 (P=0.3628)	0.2715 (P=0.3277)	0.1040 (P=0.7124)	0.2332 (P=0.4029)	0.3331 (P=0.2250)
Factor 2	0.6770 (P=0.0056)	0.2797 (P=0.3127)	0.5331 (P=0.0407)	0.6255 (P=0.0126)	0.7857 (P=0.0005)
Factor 3	0.7972 (P=0.0004)	0.3919 (P=0.1485)	0.7912 (P=0.0004)	0.6580 (P=0.0077)	0.7395 (P=0.0016)
Factor 4	0.3120 (P=0.2575)	0.4020 (P=0.1375)	0.2236 (P=0.4231)	0.1181 (P=0.6751)	0.4679 (P=0.0786)

Thirty-five patients were in the anemic and non-anemic groups. Of these, 15 were considered anemic on the basis of criteria established by the Third National Health and Nutrition Examination Survey. ²⁸ As hypothesized, anemic patients had statistically higher fatigue scores than non-anemic patients by parent report (P=0.0422). Five items (1, 4–6, and 10) were endorsed 3 to 5 times more often by anemic than non-anemic patients. Based on the PROC GLM results, the FS-A total scale scores at baseline did not differ by gender (F=0.17, P=0.69), race (F = 2.39, P=0.14), or diagnosis (F= 0.19, P=0.67).

Ability to Measure Change Over Time

Across the four studies, the FS-A scores increased significantly between the two designated points, baseline (Time 1) and endpoint (Time 2) (t=2.55, P=0.01), although the FS-P scores did not (t=1.74, P= 0.09). Within the SLEEP2 and CLUSTERS studies, the FS-A measured significant change (Table 8). The FS-A change scores increased significantly for males over time. By parent report, fatigue increased significantly for females over time (Table 9).

Table 8.

Assessing Change Over Time in Adolescent Fatigue by Using the FS-A and FSP: Across All Studies and Within Each Study

		n	Mean	SD	Median	Min	Max	t	P
Difference of FSA scores between Times 2 and 1		57	4.37	12.04	2.00	−18.00	35.00	2.74	<0.01
Difference of FSP scores between Times 2 and 1		52	1.92	11.05	1.50	−20.00	37.00	1.26	0.22
		n	Mean	SD	Median	Min	Max	t	P
Difference of FSA scores between Times 2 and 1	MFCC	15	−2.40	9.97	−5.00	−18.00	23.00	−0.93	0.37
	SLEEP	15	4.07	12.00	2.00	−17.00	27.00	1.31	0.21
	SLEEP2	10	8.20	10.56	4.50	0.00	34.00	2.46	0.036
	Clusters	17	7.59	13.73	6.00	−19.00	37.00	2.28	0.0367
Difference of FSP scores between Times 2 and 1	MFCC	13	−4.62	6.87	−3.00	−18.00	2.00	−2.42	0.032
	SLEEP	14	5.86	12.90	2.00	−9.00	37.00	1.70	0.11
	SLEEP2	9	2.22	14.42	0.00	−20.00	32.00	0.46	0.66
	Clusters	16	7.50	13.87	8.00	−15.00	28.00	2.16	0.047

SD = standard deviation.

Table 9.

Assessing Change Over Time in Fatigue Scores by Using the FS-A and FS-P for All Patients by Gender

		n	Mean	SD	Median	Min	Max	t	P
Differences between FS-A scores at Time 2 and Time 1	Female	22	4.14	13.00	3.50	−19.00	34.00	1.49	0.15
	Male	35	4.14	11.99	3.00	−18.00	37.00	2.04	0.049
Differences between FS-P scores at Time 2 and Time 1	Female	20	5.85	10.09	6.50	−16.00	24.00	2.59	0.018
	Male	32	1.41	14.26	-2.50	−20.00	37.00	0.56	0.58

Discussion

The lack of a reliable and valid self-report indicator of adolescent fatigue related to cancer contributes to insufficient clinical attention and assessments of fatigue in this patient group. The overall psychometric assessments of the FS-A indicate that this instrument is able to reliably and validly measure cancer-related fatigue in 13- to 18-year-olds. Certain limitations to this work exist. The sample size, though impressive for this age of pediatric oncology patients, is just shy of the minimum recommended (five participants per item) for an exploratory factor analysis. As a result, the four-factor solution needs to be judiciously considered and tested further in a larger study of adolescents. Second, the adolescents who participated in this research were undergoing treatment, a time typically associated with highest reported fatigue⁷; therefore, findings do not represent adolescent survivors of childhood cancer who have completed treatment.

The rapid completion time and lack of unanswered items indicates that the FS-A is readily completed by adolescents receiving cancer treatment and is not unduly burdensome. The causes of fatigue most frequently endorsed by the adolescents were treatment-related, indicating that the adolescents view cancer therapy, being hospitalized as part of that therapy, and being bored in their clinical situation as contributing to higher fatigue intensity scores. These adolescent-reported causes of fatigue are very similar to those reported by 7- to 12-year-old oncology patients.^4,29

The moderate to high internal consistency coefficients at the majority of time points indicate that the FS-A is a reliable indicator of adolescent cancer-related fatigue. The two items that contributed to a lower internal consistency coefficient did so at 5 of 13 time points, but they were not consistently problematic. The reliability of these two items will need to be further tested to determine whether they should be removed from the instrument.

The ability of the FS-A to distinguish between known groups (anemic vs. non-anemic patients) is particularly important. The possibility that certain items (in this case, five) can most sensitively distinguish between these two groups of adolescents may mean that it is possible to reduce the total number of items on the FS-A. Such a decision would have to come after further testing and be carefully considered in light of the reliability and validity findings. The scores on the FS-A did not differ by race, gender, or diagnosis at baseline, but the FS-A scores for males did increase significantly over time, whereas female’s self-report fatigue scores did not. This gender difference could be a function of having fewer female participants. It is also possible that this difference was secondary to the larger number of female participants who did not complete the second measurement point (female n=6; male n=1). Perhaps those females who did not complete the second data point differed in some important way from those who did, and the fact that their scores were not counted constituted a source of possible bias. If this kind of gender difference over time persists in future studies, the possibility of a gender-specific intervention may need to be considered. In other work with non-cancer groups of adolescents, fatigue was more prevalent in females as pubertal development increased³⁰ and was more intense in older adolescents.³¹

We anticipated that correlations between the FS-A and the FS-P would be higher than those between the FS-A and FS-S because adolescents and their parents are physically closer (at home and at the clinic and hospital) than are adolescents and the clinic or inpatient nurses. The briefer, focused interactions between adolescents and nurses may not be sufficient to allow for a sensitive rating by the nurse proxy rater. These findings suggest that seeking the nurse ratings when the nurse-adolescent interactions are abbreviated is not likely to yield ratings that will be similar enough to those of the adolescent experiencing the fatigue to be useful in clinical assessments or as the basis for clinical interventions. A similar pattern of relatedness among patient, parent, and staff fatigue ratings has been previously noted for 7- to 12-year old oncology patients.²⁹

The strong correlation between the FS-A and the RDS indicates that certain aspects of fatigue and depression are experienced jointly by adolescents with cancer; this finding is consistent with the conceptual definition of adolescent cancer-related fatigue as perceived emotional, physical, and mental aspects that are sometimes combined. A significant relation has been reported between adolescent chronic fatigue syndrome and depression in groups of patients without cancer³² and between fatigue and depression in children being treated for cancer.³¹ Certain factors of the FS-A and the RDS were not consistently associated at the two measured time points; this lack of consistency may indicate that future testing is needed to determine whether the items of those non-significantly related factors help distinguish between the adolescent who is fatigued but not depressed and the adolescent who is both fatigued and depressed.

Although the four factors retained from the exploratory factor analysis need to be interpreted with caution because of the small ratio of study participants to number of instrument items, the factors reflect the key characteristics of the conceptual definition of adolescent cancer-related fatigue. However, the factors need to be considered preliminary pending further validation.

Conclusions

The FS-A has achieved strong internal consistency and moderate to strong construct validity estimates as a new clinical research instrument. The four-factor solution is consistent with the conceptual definition underlying the construct of cancer-related fatigue in adolescents and the underlying conceptual model of fatigue as being caused by both the cancer and its treatment, but as being moderated by certain adolescent, parent, or staff activities. Particularly impressive was the ability of the FS-A to capture differences over time, because this ability is essential to using the FS-A in interventions designed to prevent or diminish cancer-related fatigue in adolescents. Future testing will be important to confirm the four-factor solution and to determine whether specific items of the 14 on the FS-A merit being removed from the instrument.

Figure 1.

The Fatigue Scale for 13- to 18-Year-Olds (FS-A)

Figure 2.

FS-A Items Designed to Measure Adolescents’ Perceived Causes of Fatigue