Support Care Cancer (2005) 13:246–254 DOI 10.1007/s00520-004-0734-y
Simon Van Belle Robert Paridaens Georges Evers Joseph Kerger Dominique Bron Jan Foubert Gerrit Ponnet Didier Vander Steichel Christine Heremans Dominique Rosillon
Received: 27 July 2004 Accepted: 6 October 2004 Published online: 12 November 2004 Springer-Verlag 2004 Professor Evers passed away suddenly. The authors will always remember his enthusiastic collaboration, wise counsel, and powerful vision. This work was supported by an unrestricted grant from Janssen-Cilag Belgium. S. Van Belle ()) Medical Oncology, University Hospital Ghent, De Pintelaan 185, 9000 Gent, Belgium e-mail:
[email protected] Tel.: +32-924-02692 Fax: +32-924-03868 R. Paridaens · G. Evers University Hospital Gasthuisberg, Leuven, Belgium J. Kerger Cliniques Universitaires Mont-Godinne, U.C.L., Brussels, Belgium D. Bron · J. Foubert University Hospital Jules Bordet, Brussels, Belgium G. Ponnet University Hospital VUB, Brussels, Belgium D. Vander Steichel Fderation Belge contre le Cancer, Brussels, Belgium C. Heremans Vlaamse Liga tegen Kanker, Brussels, Belgium D. Rosillon Biopharma, Wavre, Belgium
ORIGINAL ARTICLE
Comparison of proposed diagnostic criteria with FACT-F and VAS for cancer-related fatigue: proposal for use as a screening tool
Abstract The objective was to validate the use of the proposed International Statistical Classification of Diseases and Related Health Problems (10th revision) (ICD-10) criteria for fatigue (P-ICD10) through comparison with the Functional Assessment of Cancer Therapy Fatigue (FACT-F) subscale and three visual analogue scale (VAS) qualities in cancer patients thought to be fatigued. Fatigue was assessed in 834 cancer patients at three clinical centres in Belgium, using P-ICD10, FACT-F, and VAS to assess: level of energy (VAS1), quality of life (VAS2), and ability to perform daily activities (VAS3). Of the 834 interviewed cancer patients, 54% were classified as fatigued by the P-ICD10 criteria. Internal consistency of PICD10 was very good (alpha coefficient 0.82). The principal component analysis corroborated good internal consistency with all variables included in the first component; a second component was used to identify psychological fatigue (concentration and short-term memory disabilities). An abridged set of screening tools based on the first three general symptoms of the P-ICD10 is proposed with 100% specificity and 86% specificity, respectively. There was a marked decrease in FACT-F and VAS1 scores in patients diagnosed as fatigued by the P-ICD10 (mean€SD, FACT-F 20€9 vs 39€8, VAS1 34€21 vs 61€21). A logistic regression
model between P-ICD10 criteria diagnosis and FACT-F (VAS1) identified a score of 34 (61) on the FACT-F scale as a proposed cut-off point for the diagnosis of fatigue. The ICD-10 criteria can be recommended as a diagnostic tool, whereas the FACT-F scale and the level of energy 100-mm VAS assess the intensity of fatigue, and are more suitable for follow-up of cancer-related fatigue. Keywords Fatigue · Assessment · Diagnosis · Cancer
247
Introduction
Methods
Fatigue is one of the most common complaints of people with cancer [1]. It exists in 78% to 96% of the cancer population, particularly in individuals actively undergoing treatment [2, 3, 4, 5, 6, 7, 8, 9, 10]. Fatigue is to be taken as a chronic form of tiredness, which is perceived by the patient as being unusual or abnormal, absolutely disproportionate with respect to the amount of exercise or activity he/she has carried out, and not alleviated by resting or sleeping [11]. It often persists after treatment is concluded [12]. Fatigue caused by cancer or cancer therapy was recently proposed for recognition by the World Health Organization (WHO) and for publication in the International Statistical Classification of Diseases and Related Health Problems (10th revision) (ICD-10) [13]. The “Practical Dossier: Fatigue in Cancer Patients” was developed by the Workgroup Against Cancer Fatigue (WACF), a Belgian national advisory board consisting of key opinion leaders including physicians, nurses active in the most important oncology organizations, and representatives of the two largest official patient organizations in Belgium. It contains an adaptation of the United States National Cancer Institute treatment guidelines [14] that comply with Belgian standards, evaluation and follow-up scales, and an agenda and brochure with advice on how to cope with cancer fatigue. The assessment of fatigue is multidimensional in nature [15]. Ambiguous literature and a previous lack of specific tools to measure fatigue created difficulties in establishing assessment and management guidelines. To organize a structured measurement and follow-up of cancer fatigue in individual patients, three scales [ICD-10, Functional Assessment of Cancer Therapy Fatigue (FACT-F) subscale, and a visual analogue scale (VAS)] were chosen based on a thorough study of the literature. In order to use these in a standardized way, validation was necessary. Fatigue diagnosis criteria had already been developed by Cella and proposed for inclusion in the ICD-10. In order to use the scales as standardized instruments, validation is needed. In addition, no cut-off values have been proposed for the diagnosis of fatigue. Cancer-related fatigue criteria [16] have been proposed as a diagnosis for inclusion in ICD-10–Clinical Modification (ICD-10-CM) under the preferred term neoplastic (malignant)-related fatigue (code R.53.0 in the predraft release, June 2003 [17]). The WACF scientific advisory board has reviewed this initiative and will continue to provide further guidance to safeguard the scientific and methodological set-up.
Patients Data about cancer patients suspected to be fatigued were collected in three clinical centres in Belgium. During treatment visits, fatigue was assessed using the proposed ICD-10 Criteria for Cancer-Related Fatigue for inclusion in the ICD-10-CM [18, 19] (referred to in the rest of the paper as the P-ICD10), the FACT-F [20], and three VAS scores assessing the past week’s VAS1, VAS2 and VAS3. Other data collected were demographics (sex, age, weight), history of disease (type of tumour), treatment history and current treatment, and associated aggravating factors (e.g. pain, sleep disorders, cognitive and emotional problems, etc.); hematologic and metabolic parameters (including iron and electrolytes) were also reported when available. Assessment of fatigue The ICD-10 proposed diagnosis of cancer-related fatigue questionnaire was used to identify fatigued patients. French and Dutch translations were developed for the purpose of this study and, with the original, are presented in the Appendix. The ICD-10 questionnaire was restructured into two distinct parts. The first item of the original questionnaire was separated into three symptoms while the ten other complaints were kept as originally presented. Fatigue was diagnosed if at least one of the following symptoms—significant fatigue, diminished energy, increased need to rest—were present every day or nearly every day during the same 2-week period in the past month. One objective of restructuring the questionnaire was to explore the properties of the first part as a rapid screening tool. The FACT-F is a well-known 13-item fatigue subscale utilizing a five-point Likert self-report scale ranging from 0 (not at all) to 4 (very much so). The total score varies from 0 (worst condition) to 52 (best condition) and is calculated as 52 (sum of all items). The score is determined after re-parameterization of items 7 (I have energy) and 8 (I am able to do my usual activities) where 0 is worst condition and 4 is best condition, which have an inverse relationship to the other 11 subscale items. The three VAS consist of 100-mm horizontal lines where VAS1 indicates energy level (0 = exhausted, 100 = have energy), VAS2 indicates quality of life (0 = very bad, 100 = very good), and VAS3 indicates ability to perform daily activities (0 = very bad, 100 = very good). Statistical analysis Descriptive statistics were computed per patient group classified as P-ICD10-positive or P-ICD10-negative. Internal consistency of the P-ICD10 and FACT-F was assessed by computing Cronbach’s alpha coefficient [21]. The P-ICD10 individual items were also analyzed using principal component analysis (PCA). PCA is a multivariate statistical factor analysis technique. The aim is to describe the variation of observations (i.e. patients) in a set of linear combinations of the original variables (i.e. various P-ICD10 items). The first principal component or first axis is the combination of items that explains the greatest amount of variation. The second principal component defines the next largest amount of variation and is independent of the first principal component. Correlation between ICD-10 individual symptoms and FACT-F individual items was assessed with Pearson correlation coefficients. The P-ICD10 for fatigue was compared with the FACT-F and the three VAS using logistic regression models where the probability of being P-ICD10-positive was modelled versus FACT-F and VAS scores. Logistic equations were used to calculate sensitivity and
248
specificity for various cut-off points of FACT-F and VAS versus the ICD-10. Discriminatory accuracy was summarized by a receiver operating characteristic (ROC) curve. Patients were also classified as fatigued or not by each VAS using various cut-off points (50, 55, 60, 65, and 70 mm). The sensitivity, specificity, and positive and negative likelihood ratios of the P-ICD10 were calculated for each VAS at each cut-off point value.
Results Patients Data on 834 patients (43% female) aged 14 to 91 years (median 56 years) were collected between 8 January 2001 and 9 April 2002 in three Belgian clinical centres (Brussels, n=127; Leuven, n=208; and Gent, n=499). Of the 834 patients, 454 (54%) were classified as fatiguepositive according to P-ICD10 and all patients under active therapy were diagnosed as fatigue-positive. Demographic characteristics were similar for ICD-10-positive and ICD-10-negative patients (Table 1). However, all patients except one treated with adjuvant chemotherapy were ICD-10-positive. Internal consistency of P-ICD10 The first three P-ICD10 symptoms of fatigue (significant fatigue, diminished energy and increased need to rest disproportionate to any recent change in activity level) were present in most of the P-ICD10-positive patients; 84% reported all three symptoms and 98% experienced “diminished energy” (Fig. 1). This symptom was also reported by 59% of P-ICD10-negative patients while “significant fatigue” was reported in only 20%, indicating the discriminatory property of this general symptom. Of the P-ICD10-positive patients, 84% experienced all three symptoms. Among the P-ICD10 negative patients, 31%
Table 1 Patient characteristics
Age (years) Median Range Sex (F/M) Patient treated with, n (%) Chemotherapy Hormonotherapy Immunotherapy Radiotherapy Other therapy No therapy No data a
a
Fig. 1 Occurrence of each P-ICD10 fatigue symptom in all patients (n=834) and patients classified as positive (n=454) and negative (n=380) according to the P-ICD10. Definition of items: 1, significant fatigue; 2, diminished energy; 3, increased need to rest disproportionate to any recent change in activity level; 4, complaints of general weakness or limb heaviness; 5, diminished concentration or attention; 6, decreased motivation or interest in engaging in usual activities; 7, insomnia or hypersomnia; 8, experience of sleep as unrefreshing or nonrestorative; 9, perceived need to struggle to overcome inactivity; 10, marked emotional reactivity (e.g. sadness, frustration, or irritability) to feeling fatigued; 11, difficulty completing daily tasks attributed to feeling fatigued; 12, perceived problems with short-term memory; 13, postexertional malaise lasting 2 h
experienced none of the three symptoms and only 13% reported all three symptoms. Among the ten other complaints, the most frequently reported by both P-ICD10-positive and P-ICD10-negative patients were “difficulty completing daily tasks attributed to feeling fatigued”, “insomnia or hypersomnia”, “perceived need to struggle to overcome inactivity”, and “diminished concentration or attention”. The most discriminative complaints were related to physical fatigue such as “perceived need to struggle to overcome inacP-ICD10-positive
P-ICD10-negative
(n=454)
(n=380)
56 17–91 174/280
57 14–87 182/198
170 49 6 25 68 175 48
1 (0.3) 0 0 0 0 237 (62) 142 (37)
(37) (11) (1) (6) (15) (39) (11)
Patients may be treated with more than one therapy and percentages may be higher than 100%
249
Fig. 2 Scatter plots of all patients on the first two axes of the principal component analysis
tivity” (P-ICD10-positive 61%, negative 12%), “difficulty completing daily tasks attributed to feeling fatigued” (79% vs 24%) and “post-exertional malaise lasting several hours” (48% vs 8%), and complaints concerning motivation including “decreased motivation or interest in engaging in usual activities” (65% vs 24%). The lessdiscriminatory complaints concerned cognitive impairment “perceived problems with short-term memory” (46% vs 19%). Cronbach’s alpha coefficient reached 0.82. Reliability of each individual item was assessed by computing the alpha coefficient after removing individual items. Removing non-reliable items would increase internal consistency. The alpha coefficient did not increase after removing any individual item except perceived problems with short-term memory” where alpha slightly increased from 0.818 to 0.822. Removing any of the other items decreased the internal consistency, the most reliable items being the first three general symptoms. The first component of the PCA of the 13 individual items accounted for 33% of the total variability of the 13 individual items, and the first two components for 43%. The first component can be interpreted as an axis of fatigue as indicated by a good discrimination of P-ICD10positive and P-ICD10-negative patients (Fig. 2). Component 1-positive coordinates (Eigen value) for each individual item confirmed the internal consistency of the IDC-10 criteria with the three general symptoms (items 1–3) showing the best correlation with the first component (Fig. 3). The second component allows for discrimination between patients with cognitive complaints (items 12 and 5; positively correlated) and those with physical complaints (items 4, 9, 11 and 13; negatively correlated).
Fig. 3 Principal component analysis: load (Eigen values) of the 13 original P-ICD10 items on the first two axes. See Fig. 1 for definition of items
Fig. 4 Mean score (0–4) of each FACT-F item for all patients (n =470) and patients classified as positive ( n =326) and negative (n =140) according to the P-ICD10. Definition of FACT items: 1, fatigued; 2, weak all over; 3, listless (washed out); 4, tired; 5, trouble starting tasks; 6, trouble finishing tasks; 7, have energy; 8, able to do usual tasks; 9, need to sleep during day; 10, too tired to eat; 11, need help doing usual activities; 12, frustration because of fatigue; 13, limitation of social activities
Internal consistency of FACT-F FACT-F was assessed in 470 out of the 834 patients (56%). Cronbach’s alpha coefficient of FACT-F reached 0.94, slightly higher than the coefficient obtained for the P-ICD10 items. Each of the 13 individual items correlated well with the total score of the other 12 items (correlation
250
Table 2 FACT-F and VAS scores in P-ICD10-positive and P-ICD10-negative patients. Values are means€SD (min; median; max)
P-ICD10
FACT-F VAS1 VAS2 VAS3
Number of top three symptoms experienced
Positive
Negative
Three
One or two
None
(n=326)
(n=144)
(n=333)
(n=102)
(n=35)
20.3€9.4 (0; 20; 45) 33.5€21.4 (0; 34; 100) 37.2€24.2 (0; 40; 100) 39.3€24.2 (0; 40; 100)
39.4€8.3 (16; 41; 52) 61.1€20.9 (8; 62; 100) 64.9€20.9 (10; 66; 100) 64.3€23.1 (0; 68; 100)
20.4€9.5 (0; 20; 48) 33.2€20.8 (0; 34; 95) 37.1€23.9 (0; 40; 100) 38.5€23.4 (0; 40; 100)
38.2€6.7 (21; 39; 50) 60.2€20.0 (8; 62; 100) 64.1€20.2 (21; 64; 100) 65.2€21.8 (3; 69; 100)
45.7€6.5 (23; 47; 52) 72.2€19.0 (12; 74; 100) 73.5€17.8 (35; 74; 100) 74.3€21.4 (16; 82; 100)
Table 3 Correlation coefficient between P-ICD10 items and FACT-F items ( bold type good correlation, 0.6; italic type average correlation, 0.4 r
