- © 2009 by American Society of Clinical Oncology
Evaluation of Screening Instruments for Cancer-Related Fatigue Syndrome in Breast Cancer Survivors
- Corresponding author: Patrick Stone, MD, Department of Palliative Medicine, Division of Mental Health, 6th floor Hunter Wing, St George's University of London, Cranmer Terrace, London, SW17 ORE, United Kingdom; e-mail: pstone{at}sgul.ac.uk.
-
Presented in part in poster format at the 44th Annual Meeting of the American Society of Clinical Oncology, May 30–June 3, 2008, Chicago, IL.
Abstract
Purpose Cases of cancer-related fatigue syndrome (CRFS) can be reliably indentified using a diagnostic interview combined with a structured psychiatric interview. However, these interviews are time consuming to conduct, require specialist training, and are not suitable for routine clinical use. The purpose of this study was to identify whether a screening questionnaire could identify patients at high risk of clinically significant fatigue who should be considered for a suitable intervention.
Patients and Methods The diagnostic interview for CRFS and the structured clinical interview for the diagnostic and statistical manual of mental disorders were used in order to identify breast cancer survivors who fulfilled the criteria for CRFS. Two fatigue questionnaires (the Bidimensional Fatigue Scale [BFS] and the Functional Assessment of Cancer Therapy-Fatigue subscale [FACT-F]) were administered in order to determine their screening properties.
Results Two hundred women were interviewed and 60 women fulfilled the criteria for CRFS. The BFS cutoff score of 11 had a sensitivity of 92%, a specificity of 53%, a positive predictive value (PPV) of 46%, and a negative predictive value (NPV) of 94%. The FACT-F cutoff score of 36 had a sensitivity of 80%, a specificity of 71%, a PPV of 55%, and a NPV of 89%.
Conclusion The BFS and FACT-F cutoff scores can be used to identify breast cancer survivors at higher risk of clinically significant ongoing post treatment fatigue. Neither scale can be used as a diagnostic instrument for CRFS.
INTRODUCTION
Cancer-related fatigue (CRF) is a distressing symptom that can occur at all stages of treatment1,2 and longer term in disease-free survivors.3 However, the prevalence of fatigue varies depending on which population is studied and which assessment tool is used.4,5 There is no consensus about the best method to assess CRF, and numerous assessment scales have been developed.6 The reason for the abundance of tools derives in part from a lack of an agreed definition about what constitutes CRF.
CRF as a concept can be thought of on a continuum of severity. The lower end of the spectrum will be level with the population baseline.7 The upper end of the spectrum is more difficult to define in practice. Some groups have used a cutoff score either on a quality of life tool8 or on a visual analog scale.9 The limitation with both these methods is that they fail to capture the complexity of the symptom of fatigue and are not able to identify the most severely fatigued sub-groups.
In order to rigorously define the clinical phenotype of CRF, Cella and colleagues10 proposed a set of diagnostic criteria modeled on the International Classification of Diseases manual. This approach is similar to that used to diagnose chronic fatigue syndrome.11 The criteria are a set of symptoms which relate to fatigue, its functional impact, and duration. To fulfill these diagnostic criteria, the subject must have experienced six of 11 fatigue-related symptoms on most days or every day for 2 weeks in the previous month. At least one of these symptoms must have been “significant fatigue, lack of energy, or an increased need to rest.” The fatigue needs to be sufficiently severe to have had an impact on daily life. There must be evidence from the history, physical examination, or laboratory findings that the symptoms are a consequence of cancer or cancer therapy. Finally the fatigue should not be “primarily a consequence of comorbid psychiatric disorders such as major depression, somatisation disorder or delirium.” Most of these criteria can be assessed using a short diagnostic interview.10 However, in order to systematically evaluate patients with comorbid psychiatric disorders it is necessary to also perform a psychiatric assessment using an instrument such as the Structured Clinical Interview for the Diagnostic and Statistical Manual of Mental Disorders (SCID). If all criteria are met then the patient can be diagnosed as a case of CRF syndrome (CRFS).
The full diagnostic interview and the SCID usually require 60 to 90 minutes to complete. However, the fatigue interview does contain a number of skip items, such that the complete interview does not need to be administered to all subjects. In the clinical setting, these skip items reduce the over all interview burden considerably. However, in the context of a research setting, or when it is necessary to rigorously identify subjects for inclusion in a therapeutic trial, it can be important that subjects are assessed using both the diagnostic interview and the SCID. Thus, for instance, if one wished to compare fatigued subjects with nonfatigued subjects it would be important that both groups were matched for the presence or absence of comorbid psychiatric diagnoses. Outside of a research setting, the time-consuming nature of these interviews are likely to limit their routine use.
The aim of this study was to evaluate the use of two fatigue questionnaires as screening instruments for identifying cases of CRFS without comorbid psychiatric diagnoses. This group, identified on screening, can in practical terms be assumed to have clinically significant fatigue. They should also be considered for a suitable intervention to treat cancer related fatigue based on available guidelines9 or current evidence.12,13
PATIENTS AND METHODS
Patients
Approval for the study was given by Wandsworth research ethics committee and St George's NHS trust research and development office.
Disease-free breast cancer patients were recruited from a nurse led follow-up clinic. These women had undergone successful primary therapy (any combination of surgery, radiation therapy, and/or chemotherapy) and were seen between 3 months and 2 years after treatment was completed. Women had histologically proven breast cancer (stage of I/IIb) at the time of diagnosis and were clinically disease free at the time of entry to the study.
Methods
This study was part of a larger investigation of the prevalence and correlates of CRFS in breast cancer survivors. After written, informed consent had been obtained women underwent the Diagnostic Interview for CRFS (DICRFS) and a structured psychiatric interview (SCID). Women also completed two fatigue questionnaires.
DICRFS
The DICRFS10 was designed by Cella and colleagues in order to assess the four essential criteria required for a diagnosis of CRFS. Criterion A is the presence of two weeks of significant fatigue in the preceding month and the presence of at least five other fatigue-related symptoms. Criterion B is that the fatigue has had a significant effect on work or self-care. Criterion C is that there is evidence from the history or clinical examination that the fatigue symptoms are a consequence of cancer or cancer therapy. Criterion D is that the symptoms are not primarily a consequence of a comorbid psychiatric disorder.
This last criterion is very difficult to operationalise in practice. The original guidance suggested that the decision about whether or not to include such patients should be dependant on the investigator's interpretation of the impact of any comorbid psychiatric diagnosis. The authors suggested10 that certain fatigue characteristics might suggest that the symptom was more likely to be related to cancer (eg, its timing of onset in relation to treatment, the extent of diurnal variation or the relative prominence of cognitive symptoms). Previous studies have labeled subjects with comorbid psychiatric diagnoses as “cases” of CRFS if, in the investigators' opinion, the fatigue was not primarily a consequence of the psychiatric disorder.14,15 However, we were concerned that inclusion of such patients would introduce an unacceptable level of subjectivity into the classification process. For this reason we decided to exclude any such patients from the diagnosis of CRFS.
SCID
The nonpatient version of the SCID16 was used for this study. This is designed for use in subjects who are not known psychiatric patients. It includes modules that cover: mood disorders; psychotic screening; substance use disorders; anxiety disorders; somatoform disorders; eating disorders and adjustment disorders. All of the SCID interviews were conducted by the same researcher (S.A.) after training from two consultant psychiatrists.
Bidimensional Fatigue Scale
The Bidimensional Fatigue Scale (BFS)17 (also known as the Chalder fatigue scale or the Fatigue Questionnaire) is an 11-item questionnaire with seven items assessing physical fatigue and four items assessing mental fatigue. Each item is answered on a 4-point scale. Scores can range between 0 and 33 with higher scores indicating greater fatigue. It has been validated for use in cancer patients in several studies.18,19
Functional Assessment of Cancer Therapy-Fatigue Subscale
The Functional Assessment of Cancer Therapy-Fatigue (FACT-F) subscale 20 is a 13-item questionnaire that is part of the 20-item anemia (FACT-An) module of the FACIT quality of life assessment system.21 The FACT-F has been extensively used in a range of cancer populations.22,23 Each item can be answered on a five-point scale. Scores can range between 0 and 52 with lower scores indicating greater fatigue.
Statistical Methods
The effectiveness of the fatigue questionnaires as screening instruments was assessed by comparing the scores against the presence or absence CRFS as determined by the interview schedule described above. Receiver Operating Characteristic curves (ROC) were plotted for each of the questionnaires using SPSS (version 15; SPSS Institute, Cary, NC). The optimal cutoff scores for the questionnaires was determined by maximizing the sum of the sensitivity and specificity.
RESULTS
Over a 2-year period 292 women were deemed eligible for inclusion in the study and 208 consented to participate. Eight women did not proceed to interview for a variety of reasons leaving 200 women who completed the full interview process. Subjects had a mean age of 58 years (SD, 12.3; range, 29 to 89) and the majority were white (n = 163; 78%). Overall 60 (30%) of 200 participants met the criteria for CRFS. Of the 140 women who were noncases of CRFS, 36 had a psychiatric disorder and 104 had neither fatigue nor a psychiatric disorder. The 36 women had a comorbid psychiatric diagnosis that may have resulted in fatigue and are excluded from further analysis. There was no statistically significant difference in age, ethnicity, stage at diagnosis, time since diagnosis, or intensity of primary treatment regimen between the cases of CRFS and the noncases. A detailed comparison of the characteristics of the CRFS cases and the nonpsychiatric, nonfatigued women will be presented in a separate publication.
The ROC curves for the BFS and the FACT-F are shown in Figures 1 and 2 respectively. The area under each ROC curve was statistically significantly greater than baseline. The performance of each of the scales is summarized in Tables 1 and 2. The optimal cutoff for the FACT-F was found to be 35 of 36 and the optimal cutoff for the BFS was 10 of 11. The BFS had the highest sensitivity (92%) and the highest NPV (95%). The FACT-F had the highest specificity (71%) and the highest PPV (55%).
Receiver operating characteristic curve for Bidimensional Fatigue Scale (BFS).
Receiver operating characteristic curve for Functional Assessment of Cancer Therapy-Fatigue (FACT-F) subscale.
Screening Characteristics of BFS
Screening Characteristics of FACT-F
DISCUSSION
CRF is a common and distressing problem for cancer patients.24 Fatigue is also a common problem in the general population7 and for those with both physical25 and mental26 illnesses. For many years there has been confusion in the literature about what level of fatigue should be considered to be clinically significant and how to determine the prevalence of fatigue in cancer populations.2,4,27 The proposed diagnostic criteria for CRFS are a first step toward improving standardization and categorisation of this problem. All of the diagnostic interviews in this study were conducted by the same investigator (S.A.) and so there is no available data on inter-rater reliability.
The CRFS criteria have now been evaluated in several studies14,28,29 and have generally been found to be easy to apply and to have face validity as a means of identifying patients with debilitating fatigue. However, strict application of the criteria can be time-consuming and may limit the usefulness of this approach in either a research or in a clinical setting. Reliable screening tools might help to reduce the burden of clinical assessments using the SCID and the DICRFS and might make widespread use of the diagnostic criteria more acceptable as a robust method of identifying clinically significant fatigue. This study was conducted in disease-free breast cancer patients. Therefore, the generalizability of the current findings to men with cancer or women with other types of cancer remains unknown.
The ideal screening tool should have a high sensitivity (ie, it should detect most cases of the condition). It should also have a high NPV (ie, a negative test result should virtually exclude the diagnosis of the condition). Using these criteria the BFS appears to the best screening tool for detecting cases of CRFS. The low specificity and the poor PPV of both the BFS and the FACT-F mean that neither scale could be used as a case identifier for CRFS. There is currently no alternative to definitively making the diagnosis of CRFS without undertaking a clinical interview.
The benefits of using the BFS as a screening tool in a research setting are clear—the interview time is reduced by at least 50 hours for every 100 patients screened. This reduction is achieved by screening out a significant proportion of participants who will not meet the CRFS criteria (or who will have a comorbid psychiatric disorder). The benefits of using the BFS as a screening tool for significant fatigue in clinical practice are less clear. It is very useful for eliminating patients without significant fatigue (ie, those who score < 11) as the NPV is very high. However, the low PPV means a large number of patients will be identified as potentially having clinically significant fatigue. Approximately half of these patients would not meet the criteria for CRFS. One potential use of the BFS could be in identifying patients who are suitable for entry to a fatigue management program which was reserved for the more severe cases. Screening with the BFS may be a useful method of selecting patients in need of further evaluation.
The usefulness of the FACT-F as a screening tool is less clear. The sensitivity is lower than the BFS this limits its use in a research setting. It cannot be used in the same manner as the BFS to reduce the requirement of a full diagnostic interview. The use of this tool for screening would eliminate too may cases of CRFS and this would reduce the generalisability of any findings from a research study. In clinical practice the NPV and PPV make it a useful tool to identify patients with and without significant fatigue based on a cutoff score of 36. The FACT-F has also been used in a large number of intervention studies to treat CRF12 and has well-validated psychometric properties.30 Our data supports the usefulness of the FACT-F in clinical studies—it may, for instance, allow for subanalyses to determine the effect of an intervention on the most fatigued group. A limitation to its use in routine clinical practice is the relatively high number of false negatives. Our data indicates that 20 of every 100 women would be wrongly identified as not having significant fatigue.
The performance of the BFS and the FACT-F compare favorably with other screening instruments used in cancer patients. The Hospital Anxiety and Depression Scale (HADS)31 is a 14-item screening instrument used to detect anxiety and depression in patients with physical illnesses including cancer. There have been a number of studies that have investigated its use as a screening tool32–34 at various stages of cancer treatment. In these studies the HADS has a sensitivity ranging from 0.75 to 0.80 and a specificity ranging from 0.80 to 0.85 to detect a major depressive disorder on psychiatric interview. This gives a positive predictive value of 0.44 and a negative predictive value of 0.84. The specificity of the HADS is comparable to the screening properties of the BFS and FACT-F. However both fatigue instruments have a lower specificity than the HADS and will result in more false positive results. The HADS has been successfully used to identify cancer patients with undetected mood disorder suitable for treatment.35, 36 The similarity in screening properties further supports the suggestion that the BFS and FACT-F may have a useful role in screening for significant fatigue.
Before effective interventions or clinical trials can be undertaken in CRFS it will be necessary to introduce reliable methods of identifying cases of the condition. Unfortunately the diagnostic criteria for CRFS and the SCID are time-consuming and labor-intensive to administer. Our data provides a more robust method of screening for fatigue in practice than currently available methods used such as an arbitrary cutoff score on a visual analog scale. This is the first study to demonstrate that short fatigue questionnaires can be used as reliable screening instruments to identify cases of clinically significant fatigue in need of further evaluation and treatment.
AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
The author(s) indicated no potential conflicts of interest.
AUTHOR CONTRIBUTIONS
Conception and design: Susanna Alexander, Patrick C. Stone
Administrative support: Susanna Alexander
Provision of study materials or patients: Susanna Alexander, Patrick C. Stone
Collection and assembly of data: Susanna Alexander, Patrick C. Stone
Data analysis and interpretation: Susanna Alexander, Ollie Minton, Patrick C. Stone
Manuscript writing: Susanna Alexander, Ollie Minton, Patrick C. Stone
Final approval of manuscript: Susanna Alexander, Ollie Minton, Patrick C. Stone
Acknowledgment
We thank Paul Andrews, PhD, Janine Mansi, MD, Laura Assersohn, MD, Anup Sharma, MD, Dibeyshwar Banerjee, MD, Sue Lownes, RGN, Joe Diffley, RGN, Catherine Coleman, MD, and Sarah White, Bsc.
Footnotes
-
Supported by Grant No. C11075/A7143 from Cancer Research UK.
-
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
- Received July 16, 2008.
- Accepted October 3, 2008.
