A randomized controlled trial to evaluate the effectiveness of a brief, behaviorally oriented intervention for cancer-related fatigue

Authors

  • Jo Armes PhD, MSc, BSc, RGN,

    Corresponding author
    1. Florence Nightingale School of Nursing and Midwifery, King's College London, Waterloo Bridge Wing, Franklin Wilkins Building, Stamford Street, London, United Kingdom
    • Florence Nightingale School of Nursing and Midwifery, Specialist and Palliative Nursing Section, King's College London, Waterloo Bridge Wing, Franklin Wilkins Building, Stamford Street, London SE1 9NH, United Kingdom
    Search for more papers by this author
    • Fax: (011) 44 (0)20-7848 3764.

  • Trudie Chalder PhD, MSc, RMN,

    1. Department of Psychological Medicine, King's College London, Weston Education Centre, London, United Kingdom
    Search for more papers by this author
  • Julia Addington-Hall PhD, BSc,

    1. School of Nursing and Midwifery, University of Southampton, Highfield, Southampton, United Kingdom
    Search for more papers by this author
  • Alison Richardson PhD, MSc, BN, RN,

    1. Florence Nightingale School of Nursing and Midwifery, King's College London, Waterloo Bridge Wing, Franklin Wilkins Building, Stamford Street, London, United Kingdom
    Search for more papers by this author
  • Matthew Hotopf PhD, MSc, MRCPsych, MBBS, BSc

    1. Department of Psychological Medicine, King's College London, Weston Education Centre, London, United Kingdom
    Search for more papers by this author

Abstract

BACKGROUND.

It has been shown that nonpharmacologic interventions are effective management techniques for cancer-related fatigue (CRF) in cancer survivors. However, few studies have investigated their effectiveness in patients who are receiving chemotherapy. In this study, the authors tested the effectiveness of a brief behaviorally oriented intervention in reducing CRF and improving physical function and associated distress in individuals who were receiving chemotherapy.

METHODS.

For this randomized controlled trial, 60 patients with cancer were recruited and received either usual care or the intervention. The intervention was delivered on an individual basis on 3 occasions over a period from 9 weeks to 12 weeks, and the objective of the intervention was to alter fatigue-related thoughts and behavior. Primary outcomes were assessed as follows: CRF using the Visual Analogue Scale-Global Fatigue; physical functioning using the European Organization for Research and Treatment of Cancer Quality-of-Life Core 30 Questionnaire, and CRF-associated distress using the Fatigue Outcome Measure. Assessments were made on 4 occasions: at baseline (T0), at the end of chemotherapy (T1), 1 month after chemotherapy (T2), and 9 months after recruitment (T3). Normally distributed data were analyzed using t tests and random-slope/random-intercept mixed models.

RESULTS.

The intervention demonstrated a trend toward improved CRF, although this effect was reduced once confounders had been controlled statistically. There was a significant improvement in physical functioning (coefficient, 10.0; 95% confidence interval, 2.5–17.5; P = .009), and this effect remained once the confounding effects of mood disturbance and comorbid disorders were controlled statistically. No decrease in fatigue-related distress was detected.

CONCLUSIONS.

The behaviorally oriented intervention brought about significant improvements in physical functioning, indicated a trend toward improved CRF, but detected no effect for fatigue-related distress. Cancer 2007. © 2007 American Cancer Society.

Cancer-related fatigue (CRF) has been defined as a “persistent, subjective sense of tiredness related to cancer or cancer treatment that interferes with usual functioning.”1 The pathophysiology of CRF is understood poorly and is believed to be complex and multifactorial.2 Among patients with cancer, it has been identified as one of the most common symptoms experienced during cytotoxic treatment, with reported prevalence rates from 65% to 99%.3 Although cancer patients report that CRF has a deleterious impact on normal functioning and quality of life, evidence indicates that CRF often is overlooked by health professionals.4, 5 When advice is given on managing CRF, the most commonly reported suggestion made is rest and relaxation.5

Two systematic reviews have confirmed that there are no proven pharmacologic treatments for CRF; however, research testing the effectiveness ofpsychosocial/psychoeducational interventions in helping patients adapt to having cancer has demonstrated concomitant improvements in fatigue.3, 6 Controlled clinical trials of exercise interventions have reported improvements in CRF.3, 6 However, although the results from a recent meta-analysis of exercise for women who were receiving adjuvant therapy for breast cancer indicated a statistically significant improvement in cardiorespiratory fitness, no improvement was observed for CRF.7 Only 2 randomized controlled trials (RCTs) have been conducted of psychoeducational interventions specifically for CRF in patients who were receiving cytotoxic treatment.8, 9 Both of those interventions were reported as bringing about improvements in CRF. The intervention developed by Barsevick et al.8 was based on a model of energy conservation, whereas Ream et al.9 developed a supportive approach. The results reported by Barsevick et al. reflected their model, in that significant improvements in CRF were observed, but there were no differences between the experimental group and the control group in terms of physical functioning.8 The improvement reported by Ream et al.9 was based on a mean score derived from 4 fatigue visual analogue scales (VAS). Although no differences were observed between the control and experimental groups in the severity of fatigue, significant improvements were noted for distress caused by fatigue and disruption to pastimes/hobbies by fatigue. In both studies, patients were eligible to participate whether or not they had fatigue at the time of recruitment; thus, it is unclear whether the objective of the interventions was prevention or treatment of CRF. Although these studies provide some evidence that such interventions are effective, important gaps remain in our current understanding of the best way to manage CRF.

Several studies in individuals with chronic fatigue syndrome have reported that cognitive behavioral therapy (CBT) brings about significant improvements not only in fatigue but also in physical functioning.10–12 The interventions that were used in those studies were based broadly on a fear-avoidance model of understanding responses to fatigue. On the basis of this evidence, we investigated whether a brief, CRF-specific, behaviorally oriented intervention would be of benefit to patients with cancer who were receiving cytotoxic treatment and were experiencing significant CRF. The fear-avoidance model of symptom management13–16 guided the development of the intervention, and the objective of the intervention was to alter fatigue-related behavior. We hypothesized that, compared with usual care, 1) the behaviorally oriented intervention would be superior in reducing CRF, and 2) the intervention would bring about greater improvements in physical function and CRF-associated distress.

MATERIALS AND METHODS

We obtained approval for the study from the following local research ethics committees: Guys and St. Thomas' Hospital National Health Service (NHS) Trust and Bromley Hospitals NHS Trust. All participants provided written informed consent.

Design

A parallel-groups, RCT methodology was used.

Patients

Consecutive patients who were attending for chemotherapy treatment at 2 cancer centers/units in South London between October 15, 2001 and March 24, 2003 were screened for eligibility. We excluded patients who 1) were not aware of their cancer diagnosis; 2) were aged < 18 years; 3) did not have histologically proven cancer; 4) were receiving the last half of the planned course of cytotoxic treatment; 5) were unable to speak and understand English; 6) did not report significant fatigue, diminished energy, or increased need to rest disproportionate to any recent change in activity level17; 7) had a poor Eastern Cooperative Oncology Group performance status (>3)18;8) had a previous history of psychotic disorder; 9) had evidence of cognitive impairment (as judged by the health professionals who were caring for them) or central nervous system metastases; 10) were receiving psychotherapy or CBT at the time of recruitment; 11) were receiving cytokine treatment at the time of recruitment; or 12) had an uncontrolled infection at the time of recruitment. Consecutive potential participants were approached in the oncology day unit by the research fellow (J.A.), who explained the study and provided a written information sheet.

Randomization

Patients were randomized once they had provided written informed consent. Originally, we chose minimization as the method of treatment allocation. Thus, patients were allocated on the basis of age, sex, site of cancer, stage, and Hospital Anxiety and Depression Scale (HADS) scores with the objective of minimizing differences in these factors between the 2 groups.19 This method of allocation was suspended after 10 patients were allocated, because the research fellow could predict the group to which participants would be allocated according to sex. To ensure allocation concealment, simple random, permuted, block randomization was implemented. The block lengths were 4 and 6 and varied randomly. A statistician who was unconnected to the study generated the randomization, provided a central telephone service for patient allocation, and kept a copy of the randomization codes. Thus, allocation concealment was attained for the remainder of the study, and the allocation sequence was not revealed until completion of the trial.

Treatments

The research fellow was trained to deliver the intervention as described in a manual that was written for the study by J.A. and T.C. The intervention consisted of 3 individual, face-to-face, 60-minute sessions at 3 to 4 weekly intervals (coinciding with administration of chemotherapy). Because the intervention was designed to be brief and to ensure some standardization, a skeleton plan for each session was devised that could be modified to meet individual needs. An outline of the proposed strategies to be used in each session and associated intended outcomes is provided in Table 1. Those patients who were allocated to the control group received standard care. CRF was not assessed routinely at either center, and the provision of advice regarding its management was delivered in an ad hoc manner.

Table 1. Core Components of the Intervention
Approach/StrategyIntended outcomes
  1. CRF indicates cancer-related fatigue.

Session 1
 Cognitive approach
  Clarify meaning of CRFIdentify core thoughts and feelings about CRF
  Clarify aim and effectiveness of current CRF coping strategiesProvide a rationale for adopting new coping methods
 Behavioral approach
  Self-monitoring of CRF and sleep disturbance using diaryIdentification of patterns to aid the development of goals and so enhance approach coping
 General approach
  Education on CRF and provision of written informationNormalize CRF; development of alternative illness perceptions on CRF to enhance approach coping
  Praise and encouragementEnhance optimism and self-confidence and encourage persistence in trying to attain goals
Session 2
 Cognitive approach
  Goal settingIncreased self-efficacy, enhanced feelings of being in control, and positive mood
 Behavioral approach
  Activity schedulingPrioritization and planning of activities to improve performance through approach coping
  Graded task managementGoals divided into manageable, small steps to encourage task performance and goal persistence and so approach coping; enhancement of self-efficacy and positive mood
  DistractionEngagement in valued activities to increase enjoyment and challenge unhelpful thoughts
 General
  Provision of written information on management CRFReinforcement of behavioral strategies for managing CRF
  Praise and encouragementEnhance optimism and self-confidence and encourage persistence in trying to attain goals
Session 3
 Cognitive approach
  Clarify aim and effectiveness of current CRF coping strategiesEncourage self-monitoring of CRF and modification of goals, activity scheduling, and graded task management
 Behavioral approach
  Modification of goal setting, activity scheduling, and graded task managementGradual increase in activity level in small and manageable increments to achieve stated goals
  Cognitive restructuringIdentification of unhelpful thoughts about CRF and the impact they make on behavioral responses; enhancement of approach coping through the development of alternative thoughts
 General approach
  Provision of written information on management CRFReinforcement of cognitive strategies for managing CRF
  Praise and encouragementEnhance optimism and self-confidence and encourage persistence in trying to attain goals

Sessions were audiotaped for supervisory purposes and to check the integrity of the intervention. Individual supervision was conducted throughout the trial, depending on need and to ensure that treatment was conducted according to the manual. In addition, 2 independent raters evaluated the integrity of the intervention on a random selection of the taped sessions using the Primary Care Rating Scale.20 This measure assesses both the extent of the therapeutic alliance between therapist and patient and evaluates the extent to which CBT and counseling techniques are used.

Outcome Measures

Three primary outcomes were selected for the study: 1) reduction in fatigue, 2) improvement in physical functioning, and 3) reduction in distress associated with fatigue. Fatigue was assessed using a VAS of global fatigue (VAS-F). Patients rated their overall fatigue on a 100-mm line anchored at each end by the following statements: ‘not at all tired’ or ‘extremely tired.’ In a review of patient-based outcome measures for clinical trials, Fitzpatrick et al.21 reported that such global measures are valid, reliable, and sensitive to change, and they recommend their use when investigators want to assess theoverall value of health care interventions. The Physical Functioning subscale on the European Organization for Research and Treatment of Cancer Quality-of-Life Questionnaire Core 30, version 322 (EORTC-QLQc30) was used as the measure of physical functioning. The EORTC-QLQc30 is a 30-item measure that consists of both multi-item and single-item subscales. These include 5 Functional scales (physical, role, emotional, cognitive, and social), 9 Symptom-Severity scales (fatigue, pain, nausea and vomiting, dyspnea, insomnia, appetite loss, constipation, diarrhea, and financial difficulties), and a Global Quality-of-Life (QOL) scale. Scores are transformed onto a scale from 0 to 100. High scores on the Functioning and QOL scales indicate high-level functioning or QOL. However, high symptom-severity scores represent greater severity. Fatigue-related distress was assessed using a Fatigue Outcome Measure (FOM) that was designed specifically for the study. The FOM consisted of 7 graphic rating scales with scores from 0 to 100. These rated not only the level of fatigue, distress caused by it, and ability to cope with CRF but also the extent to which fatigue was overwhelming, uncontrollable, unpredictable, and abnormal.

Secondary outcomes and potential confounding variables were assessed using the Multidimensional Fatigue Inventory20, 21 (MFI), which is a multidimensional measure of CRF; the EORTC-QLQc3023, 24; and the HADS.25 Relevant personal and disease-related data also were collected. One month after completing cytotoxic treatment, all participants were asked to describe the advice and information received on managing CRF and to rate its usefulness and their satisfaction with it.

Patient assessments were made on 4 occasions over 36 weeks or less if the patient did not complete the cytotoxic treatment protocol. These assessments were as follows: Cycle 3 of chemotherapy (baseline assessment [T0]), end of cytotoxic treatment (T1), 4 weeks after the end of cytotoxic treatment (T2), and 9 months after recruitment to the study (T3). A priori, T2 was identified as the main outcome for the study. The T3 assessment was added shortly after data collection commenced and after the publication of a systematic review of psychosocial intervention studies in cancer patients reporting that the strongest treatment effect often was several months after the completion of an intervention.26 Questionnaires either were posted or were given to patients in the chemotherapy clinic by J.A.

Power Calculation

An a priori power calculation based on a study by Spiegel et al.27 assumed that, at T2, there would be a 5-unit decrease (0–100 scale) in mean fatigue scores for the experimental group and no change for the control group. This indicated that 39 patients would be required in each group to give the study 80% power at the 5% significance level. Accounting for drop-outs, we estimated that we required 100 patients. However, we recognized that this difference was unlikely to be clinically significant28; thus, the hypotheses were worded so that a mean difference of 10 units was sought. Nine months after the commencement of data collection, recruitment to the trial was lower than anticipated (n = 23 patients). Consequently a reassessment of power was undertaken by a statistician who was unconnected to the study. This was achieved by simulating data collected from patients who were recruited to the study and who had completed more than 1 assessment (n = 13 patients) without breaking the blindness of group allocation. The distribution of group allocation, outcome, and missing data was modeled; thus, any discrepancies in the original data, such as unequal sample size, were replicated in the modeled data. The optimal threshold sample size, at a 5% significance level, yielded a total sample size of 52 patients. Consequently and because of time limitations, we reduced the sample size to be recruited to 60 patients.

Statistical Analysis

We analyzed the data using SPSS software (version 11) and Stata software (version 9) on an intention-to-treat basis. Measures of central tendency were generated for the primary (VAS-F, FOM Distress, EORTC Physical Functioning) and secondary outcome variables related to fatigue (MFI General Fatigue, MFI Physical Fatigue, MFI Mental Fatigue, MFI Reduced Motivation, and EORTC QLQc30 Fatigue), distress (HADS Anxiety and Depression), and physical functioning (MFI Reduced Activity). Comparisons of group means at T1, T2, and T3 were performed for normally distributed primary and secondary outcomes using unpaired t tests. When data were not distributed normally, group differences were assessed using the Mann-Whitney test. Although this gave an indication of the treatment effect, as judged by between-group differences, it did not take into account the probable correlation between repeated measurements on the same individuals, and it did not accommodate missing data. Thus, we analyzed normally distributed data using either random-intercept or random-slope/random-intercept mixed models using a generalized linear latent and mixed model.29 The baseline score on the dependent variable was included in all models, and we tested for interactions between treatment group and time. If they were present, then these were retained; if not, then we present results of the main effect of treatment group on the dependent variable. In these models, we included potential confounders, which were defined as variables that were related to outcome (P < .25) in univariate regression analysis.30

RESULTS

Recruitment took place over 18 months. Figure 1 shows the flow of the patients through the study. We screened 531 consecutive patients for eligibility, and 283 patients were excluded. The remaining 249 patients were invited to participate. One hundred forty-seven did not consider themselves tired enough to participate, and 42 patients refused. Thus, 60 patients were recruited and consented to the study. We excluded 5 patients who did not return the baseline questionnaires from the analysis, because only clinical data were available for them. The final sample size was 55 patients. The number of questionnaires that were returned varied across the assessments, with the greatest nonresponse rate being at T1, when patients were completing chemotherapy treatment. At the main outcome (T2), 1 patient withdrew, 7 questionnaires were not returned, and 4 patients had died. By T3, the number of patients who died increased to 14. The majority of patients (n = 22) attended all 3 intervention sessions.

Figure 1.

Flow of the participants through the study. CNS indicates central nervous system; ECOG, Eastern Cooperative Oncology Group (performance status); T0, baseline assessment; T1, assessment at the end of chemotherapy; T2, assessment 1 month after chemotherapy; T3, assessment 9 months after recruitment.

Baseline Patient Characteristics

The baseline characteristics of participants are shown in Table 2. The mean age of the 55 patients who agreed to participate was 59.1 years (standard deviation, 11.5 years), the majority of whom (n = 33 patients) were women and white British (n = 46 patients). Of these, 27 patients had a diagnosis of colorectal cancer, 44 patients had late-stage disease (stage III or IV), and 42 patients had at least 1 metastasis. Most patients had at least 1 comorbid disorder (n = 42 patients), and the most common comorbidities were cardiac disease (n = 21 patients) and musculoskeletal disease (n = 10 patients). All patients were receiving chemotherapy and were recruited at either the second or third cycle of their course of treatment. There were few obvious differences between the 2 groups. However, the control group was slightly older and included more men and individuals who lived alone. Furthermore, there were more patients in the control group with nonsmall cell lung cancer, stage IV disease, liver metastases, and comorbid cardiac disorders. This suggests that the physical condition of the control group was poorer at the baseline assessment.

Table 2. Personal and Clinical Characteristics at Baseline
CharacteristicNo. of patients (%)
Experimental group (n = 28)Control group (n = 27)Total (n = 55)
  • SD indicates standard deviation.

  • *

    Defined according to the International Union Against Cancer TNM classification.

Mean age ± SD, y57 ± 12.161.3 ± 10.659.1 ± 11.5
Sex
 Men10 (36)12 (44)22 (40)
 Women18 (64)15 (56)33 (60)
Ethnicity
 White UK23 (82)23 (85)46 (84)
 Black UK2 (7)2 (4)
 Black Caribbean2 (7)2 (4)
 Black African2 (7)2 (7)4 (7)
 Asian1 (4)1 (2)
Home occupancy
 Lives alone3 (11)5 (18.5)8 (14.5)
 Lives with someone25 (89)22 (81.5)47 (85.5)
Diagnosis
 Colon10 (36)8 (30)18 (33)
 Rectum4 (14)5 (18.5)9 (16)
 Esophagus1 (4)1 (4)2 (4)
 Ovary4 (14)4 (15)8 (14.5)
 Breast4 (14)2 (7)6 (11)
 Nonsmall cell lung cancer1 (4)4 (15)5 (9)
 Bladder1 (4)1 (4)2 (4)
 Pancreas1 (4)2 (7)3 (5.5)
 Testicular2 (7)2 (4)
Stage*
 I2 (7)2 (4)
 II5 (18)4 (15)9 (16)
 III15 (54)13 (48)28 (51)
 IV6 (21)10 (37)16 (29)
Sites of metastasis
 Local*16 (52)15 (46)31 (49)
 Liver7 (22.5)11 (33)18 (28)
 Lung7 (22.5)6 (18)13 (20)
 Bone1 (3)1 (3)2 (3)
No. of sites of metastasis
 07 (25)6 (22)13 (23)
 114 (50)14 (52)28 (51)
 24 (14)3 (11)7 (13)
 33 (11)3 (9)6 (11)
 41 (4)1 (2)
Comorbid disease
 Cardiac9 (32)12 (44)21 (38)
 Endocrine2 (7)2 (7)4 (7)
 Musculoskeletal6 (21)4 (15)10 (18)
 Respiratory1 (4)3 (11)4 (7)
 Renal2 (7)2 (4)
 Genitourinary3 (11)1 (4)4 (7)
 Blood and lymphatic3 (11)2 (7)5 (9)
 Other9 (32)8 (30)17 (31)
No. of comorbid disorders
 05 (18)8 (30)13 (24)
 112 (43)9 (33)21 (38)
 28 (29)8 (30)16 (29
 32 (7)1 (4)3 (5.5)
 41 (4)1 (4)2 (4)

Main Outcomes

The mean difference between the experimental and control groups for the normally distributed primary outcomes (global fatigue and physical functioning) at T1, T2, and T3 was assessed using 2-tailed t tests for independent groups, and the results are shown in Table 3. Statistically significant differences in favor of the experimental group were observed for VAS Global Fatigue at T2 and EORTC Physical Functioning at T2 and T3. Scores on the FOM Distress scale were not distributed normally; thus, group differences in ranked scores were assessed using nonparametric tests (Mann-Whitney), and the results are shown in Table 4. No significant difference was observed between the 2 groups at any time point.

Table 3. Mean Difference (T Tests) in Primary Outcomes at Baseline (T0), T1, T2, and T3
Time*No. of patientsMean score (SD)Mean difference95% CIP
  • SD indicates standard deviation; 95% CI, 95% confidence interval; MFI, Multidimensional Fatigue Inventory; NA, not applicable; EORTC QLQc30, European Organization for Research and Treatment of Cancer Core 30 Quality-of-Life Questionnaire.

  • *

    T0, baseline assessment; T1, assessment at the end of chemotherapy; T2, assessment 1 month after chemotherapy; T3, assessment 9 months after recruitment.

VAS global fatigue
 T0
  Control group2764.3 (20.3)NANANA
  Experimental group2763.9 (20.9)NANANA
 T1
  Control group1663.1 (17.7)−7.4−7.1 to −21.9.31
  Experimental group2055.8 (23)   
 T2
  Control group2256.3 (21.6)−15.92–29.8.03
  Experimental group2240.4 (24)   
 T3
  Control group1951.1 (25.1)−17.1−1.5–35.7.07
  Experimental group1734 (29.2)   
EORTC QLQc30 Physical Functioning
 T0
  Control group2755.4 (16.2)NANANA
  Experimental group2662.7 (19.5)NANANA
 T1
  Control group1660.8 (14.6)8.8−19.7 to −2.11
  Experimental group2069.7 (16.8)   
 T2
  Control group2257.3 (19.8)19.7−30.2 to −9.2.001
  Experimental group2277 (14.4)   
 T3
  Control group1962.1 (24.8)17.1−31.7 to −2.5.02
  Experimental group1779.2 (17.3)   
Table 4. Ranked Differences (Mann-Whitney) in Fatigue Outcome Measure Distress Associated With Cancer-related Fatigue
Time*No. of patientsMedian scoreU test scoreExact P
  • NA indicates not applicable.

  • *

    T0, baseline assessment; T1, assessment at the end of chemotherapy; T2, assessment 1 month after chemotherapy; T3, assessment 9 months after recruitment.

T0
 Control group2755NANA
 Experimental group2748NANA
T1
 Control group1650124.36
 Experimental group1934  
T2
 Control group2229.5183.25
 Experimental group2124  
T3
 Control group1825149.90
 Experimental group1727  

Data were then analyzed using a random-slope/random-intercept mixed model. For fatigue, we detected no interaction between group allocation and time. There was a trend toward a significant difference for treatment group favoring the experimental arm (coefficient, 10.1; 95% confidence interval [95% CI], −0.6–20.8; P = .07), as shown in Figure 2. This became smaller after controlling for comorbid medical conditions and HADS score at baseline (coefficient, 8.9: 95% CI, −1.6–19.4; P = .095). For physical functioning, we detected no interaction between group and time. There was a statistically significant difference between treatment groups (coefficient, 10; 95% CI, 2.5–17.5; P = .009), as demonstrated in Figure 3. This reduced after controlling for baseline HADS score and comorbid illness (coefficient, 8.3; 95% CI, 0.6–16.1; P = .04).

Figure 2.

Mean Fatigue scores on the Visual Analogue Scale. 95% CI indicates 95% confidence interval.

Figure 3.

Mean Physical Functioning scores on the European Organization for Research and Treatment of Cancer Quality-of-Life Core 30 Questionnaire. 95% CI indicates 95% confidence interval.

Secondary Outcomes

Two-tailed t tests for independent groups were used to assess mean differences at T1, T2, and T3 between the 2 groups for normally distributed secondary outcomes (MFI General Fatigue, MFI Physical Fatigue, MFI Mental Fatigue, MFI Reduced Motivation, and EORTC QLQc30 Fatigue), and the results are shown in Table 5. Statistically significant differences in favor of the experimental group were observed for MFI Physical Fatigue at T2 and T3 and for EORTC Fatigue at T2. Scores on the MFI Reduced Activity scale were not distributed normally; thus, group differences in ranked scores were assessed using nonparametric tests (Mann-Whitney), and the results are shown in Table 6. There was a statistically significant difference at T2, and this reduced to borderline significance at T3. No statistically significant differences were observed for the remaining MFI scales. The mean difference in HADS Anxiety and Depression scores at T1, T2, and T3 were assessed using 2-tailed t tests for independent groups. No statistically significant differences were observed between the 2 groups at any time point.

Table 5. Mean Difference (t Tests) in Secondary Outcomes at Baseline (T0), T1, T2, and T3
Time*No. of patientsMean score (SD)Mean difference95% CIP
  • SD indicates standard deviation; 95% CI, 95% confidence interval; MFI, Multidimensional Fatigue Inventory; NA, not available; EORTC QLQc30, European Organization for Research and Treatment of Cancer Core 30 Quality-of-Life Questionnaire.

  • *

    T0, baseline assessment; T1, assessment at the end of chemotherapy; T2, assessment 1 month after chemotherapy; T3, assessment 9 months after recruitment.

MFI General Fatigue
 T0
  Control group2714.2 (3.1)NANANA
  Experimental group2715.1 (3.2)NANANA
 T1
  Control group1614.3 (3.6)0.6−3.1–1.9.35
  Experimental group2014.9 (3.6)   
 T2
  Control group2213.4 (3)−0.6−3.1, 1.9.63
  Experimental group2212.8 (4.8)   
 T3
  Control group1912.4 (4)−1.3−1.7–4.5.38
  Experimental group1711.1 (5.1)   
MFI Physical Fatigue
 T0
  Control group2715.2 (3.7)NANANA
  Experimental group2715.4 (3.9)NANANA
 T1
  Control group1614.6 (3.6)−7.4−2.5–3.3.31
  Experimental group2014.3 (4.6)   
 T2
  Control group2214.7 (3.3)−2.40–4.8.05
  Experimental group2212.3 (4.5)   
 T3
  Control group1913.5 (4.2)−3.30.4–6.3.03
  Experimental group1710.2 (4.6)   
MFI Mental Fatigue
 T0
  Control group2711.6 (4.6)NANANA
  Experimental group2712 (5.4)NANANA
 T1
  Control group1610.7 (4.1)1.1−4.5–2.2.51
  Experimental group2011.8 (5.5)   
 T2
  Control group229.3 (4.1)0.2−3–2.6.90
  Experimental group229.5 (5.1)   
 T3
  Control group199 (3.9)−0.5−2.4–3.5.72
  Experimental group178.5 (4.7)   
MFI Reduced Motivation
 T0
  Control group2711.6 (4.2)NANANA
  Experimental group2711.5 (4)NANANA
 T1
  Control group169.6 (5.5)−1.1−3.9–1.8.44
  Experimental group2010.7 (4.6)   
 T2
  Control group229.3 (3)−0.5−1.6–2.8.59
  Experimental group228.7 (2.9)   
 T3
  Control group199.2 (36.3)−1.5−0.6–3.7.15
  Experimental group178.7 (2.9)   
EORTC QLQc30 Fatigue
 T0
  Control group2759.7 (22.3)NANANA
  Experimental group2654.3 (26.9)NANANA
 T1
  Control group1650.4 (21.8)−1.3−15.4–17.9.88
  Experimental group2049.1 (24.9)   
 T2
  Control group2257.1 (26)−15.2−0.3–30.6.05
  Experimental group2241.9 (25.4)   
 T3
  Control group1942 (21.4)−12.6−3.2–28.4.12
  Experimental group1729.4 (24.5)   
Table 6. Ranked Differences (Mann Whitney) in Multidimensional Fatigue Inventory Reduced Activity
Assessment*No. of patientsMedian MFI scoreU test scoreExact P
  • MFI indicates Multidimensional Fatigue Inventory; NA, not applicable.

  • *

    T0, baseline assessment; T1, assessment at the end of chemotherapy; T2, assessment 1 month after chemotherapy; T3, assessment 9 months after recruitment.

T0
 Control group2717NANA
 Experimental group2717NANA
T1
 Control group1615146.66
 Experimental group1913  
T2
 Control group2215159.05
 Experimental group2110  
T3
 Control group1812103.07
 Experimental group179  

Data were then analyzed using a random-intercept mixed model. There was an interaction between group and time for MFI Physical Fatigue, indicating that patients in the treatment group experienced greater improvement on this dimension (coefficient, 1.6; 95% CI, 0.2–3.1; P = .02), as shown in Figure 4. This effect remained after controlling for potential confounders (coefficient, 1.6; 95% CI, 0.1–3.0; P = .03), as shown in Figure 4. The intervention was not associated with any improvement in other dimensions or measures of fatigue, including EORTC-QLQc30 Fatigue, MFI General Fatigue, MFI Mental Fatigue, and MFI Reduced Motivation. Results from the assessment of the intervention integrity suggest not only that there was a good therapeutic alliance with patients but also that the approach adopted in the intervention contained more elements categorized as CBT than counseling, thus confirming that the intervention was congruent with what had been planned.

Figure 4.

Mean Physical Fatigue scores on the Multidimensional Fatigue Inventory. 95% CI indicates 95% confidence interval.

DISCUSSION

The results from this study indicated a trend toward the intervention resulting in improved CRF, although this was reduced once the most obvious confounders had been controlled statistically. This finding is supported by a statistically significant improvement in the secondary outcome: MFI Physical Fatigue. There was improvement in physical functioning at T2 and T3, and this effect remained once the confounding effect of HADS score and comorbid disorders was controlled statistically. A statistically significant improvement in MFI Reduced Activity further supports this finding, contrary to the findings of Barsevick et al.,8 who reported a statistically significant difference in fatigue for patients who received the psychoeducational intervention compared with the control group. However, those authors were unable to detect an improvement in physical functioning. This may have been because the basis of the intervention used by Barsevick et al. was energy conservation. The objectives of our intervention were to regulate and increase physical activity, which explains the sustained improvement in physical functioning. In the study by Ream et al.,9 no effect was observed on the single item ‘extent of fatigue,’ which may have been because participants without significant fatigue were recruited. In our study, 77% of the sample had metastatic disease, which may have overwhelmed the effect of the intervention on CRF.

We were unable to detect decreases in fatigue-related distress, anxiety, or depression. However, the objective of the intervention was not to ameliorate mood disturbance, as evidenced when the integrity of the intervention was rated independently. Nevertheless, anxiety and depression were statistically significant, independent predictors of fatigue and physical functioning at both time points. Similar associations were observed in cross-sectional studies of CRF prevalence in patients both on and off cytotoxic treatment.31–33 Jacobsen and Weitzner34 suggest that part of the reason for this is overlap in the way fatigue and depression are measured. Clearly, the relation between fatigue and mood disturbance requires further investigation.

Potential limitations of the current study include inadequate allocation concealment at the start of the study, lack of a control for the behaviorally oriented intervention, and possible observer bias. It is notoriously difficult to develop a placebo intervention to control for nonpharmacologic interventions; thus, we believed that it was more important to test whether the intervention was beneficial to patients who had significant fatigue. It was not possible to minimize the effect of observer bias by blinding the researcher (J.A.) who conducted the assessments regarding the group to which patients were allocated, because the researcher also was responsible for delivering the intervention. Thus, it is possible that the treatment effect may be attributable to outcome expectations of both the researcher and the patients.

The small sample size poses several limitations to the study. First, the power of the study is low, and we are unable to provide precise estimates of treatment effect. Second, despite the use of randomization, there is an increased risk of imbalances occurring between experimental and control groups, leading to confounding of results. The assessment of baseline characteristics suggests that the control group was in poorer physical health. Although data analysis controlled for the effects of such imbalances to an extent, the number of confounding variables that could be included was limited by the small sample size.

The results of this study are promising. The patients who were randomized to receive the behaviorally oriented intervention experienced improved physical functioning and a trend toward less fatigue, although these effects were reduced once potential confounding variables were controlled for statistically. However, because the sample was small and heterogeneous and because the power of the study low, we are unable to produce precise estimates of treatment. Next, a larger RCT will be required to detect the precise effect of the intervention and to demonstrate that it is applicable under clinical conditions.

Acknowledgements

We thank all of the individuals who participated, without whom this study would not have been possible.

Ancillary