International Journal of Psychiatry in Clinical Practice, 2006; 10(1): 27 /32
ORIGINAL ARTICLE
Comparison between the SCOFF Questionnaire and the Eating Attitudes Test in patients with eating disorders
SHUN’ICHI NOMA1, YOSHIKATSU NAKAI2, SEIJI HAMAGAKI3, MINAKO UEHARA1, AKIKO HAYASHI1 & TAKUJI HAYASHI1 Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
1
Department of Psychiatry, Graduate School of Medicine, Kyoto University, Kyoto, Japan, 2School of Health Sciences, Faculty of Medicine, Kyoto University, Kyoto, Japan, and 3Takagi Psychiatric Clinic, Kyoto, Japan
Abstract Objective: The SCOFF was developed as a simple, five-question screening tool for eating disorders to be used in primary care. The aim of this study was to examine the appropriateness of each question in comparison with the Eating Attitudes Test-26. Methods: The SCOFF and the EAT-26 were administered to 80 patients with eating disorders who had received treatment from May through October 2003 in Japan. Results: The scores of the SCOFF and those of the EAT-26 were positively correlated (P B/0.001), and each question of the SCOFF was strongly associated with different items of the EAT26. The detection rates with the SCOFF of the patients with anorexia nervosa or bulimia nervosa and that of the patients with eating disorders not otherwise specified (EDNOS) were 96.2 and 48.1%, respectively. The scores of the SCOFF and the values of body mass index were significantly correlated (P /0.041), and the detection rate with the SCOFF of patients with low body weight and EDNOS was low (10%). Conclusion: Each question of the SCOFF has been selected appropriately for reflecting characteristics of eating disorders, although one question about body weight loss should be modified to detect patients with anorexia nervosa more accurately.
Key Words: Eating disorders, screening tool, SCOFF, Eating Attitudes Test-26, primary care
Introduction Eating disorders are prevalent among young females in Western countries and in several Asian countries. According to a nationwide epidemiological survey in Japan, the estimated prevalences of anorexia nervosa and bulimia nervosa among females aged 10 /29 years in 1998 were 0.06 and 0.03%, respectively [1]. These rates were lower than those in Western countries: among young white females the prevalence of anorexia nervosa is between 0.2 and 0.7% [2/4], and that of bulimia nervosa is between 0.5 and 1.5% [2,4,5]. The situation of eating disorders in Japan, however, is expected to become similar to that in Western countries in the near future because the prevalence of eating disorders in Japan has increased markedly in recent years: the estimated prevalences of anorexia nervosa and bulimia nervosa in females aged 10 /29 years in 1992 were only 0.02 and 0.007% [6]. Doctors in Eastern and Western countries have opportunities to encounter patients with eating disorders in daily primary care.
Morgan et al. [7] introduced the SCOFF questionnaire as a simple screening tool for eating disorders. The SCOFF consists of only five questions, and its high sensitivity and specificity have been confirmed [8]. Because the validity of the SCOFF administered as a written questionnaire as well as an oral interview has also been established [9], this scale can be used in various clinical situations. However, Morgan et al. did not explain in their original paper why they chose these questions as items of the SCOFF, and subsequent studies have not verified the appropriateness of each question. Furthermore, detecting patients who do not meet the full criteria for anorexia nervosa or bulimia nervosa may be more difficult and less pertinent for this screening tool [8]. Although this supposition is understandable, eating disorder not otherwise specified (EDNOS) is now the most common diagnostic category of eating disorders [10]. The characteristics of patients with EDNOS who cannot be detected with the SCOFF may reveal weak points of this tool, and screening for eating disorders
Correspondence: Shun’ichi Noma, M.D., Department of Psychiatry, Graduate School of Medicine, Kyoto University, 54 Shogoin-Kawaharacho, Sakyo, Kyoto, 606-8507 Japan. Tel: /81 75 751 3373. Fax: /81 75 751 3246. E-mail: [email protected]
(Received 21 September 2004; accepted 18 July 2005) ISSN 1365-1501 print/ISSN 1471-1788 online # 2006 Taylor & Francis DOI: 10.1080/13651500500305275
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
28
S. Noma et al.
would be more effectively if these weak points were eliminated. The purpose of this study was to evaluate each question of the SCOFF by administering it to patients with eating disorders, including EDNOS. For that purpose, the Eating Attitudes Test-26 (EAT-26) was administered at the same time. The EAT-40, and then the EAT-26, were originally developed by Garner et al. [11,12] as a rating scale that accurately assesses the eating patterns of patients with anorexia nervosa, but it can be used to screen for eating disorders because there is little overlap in the frequency distributions of patients with eating disorders and healthy subjects [11,12]. In this paper, the results of the SCOFF were compared with those of the EAT-26 in patients with eating disorders, and the effectiveness of the SCOFF was verified.
Method Subjects The subjects were 80 patients with eating disorders who attended Kyoto University Hospital or related hospitals from May through October 2003. All subjects were Japanese females. Informed consent The study was fully explained to the participants before examination, and the confidentiality of responses was assured. All participants agreed with the purpose and contents of this study. Procedures The SCOFF was developed by Morgan et al. [7] as a screening instrument for addressing the core features of anorexia nervosa and bulimia nervosa, as follows: 1. Do you make yourself (S) sick because you feel uncomfortably full? 2. Do you worry you have lost (C) control over how much you eat? 3. Have you recently lost more than (O) one stone in a 3-month period? 4. Do you believe yourself to be (F) fat when others say you are too thin? 5. Would you say that (F) food dominates your life? Question 1 indicates ‘‘self-induced vomiting’’; Question 2 indicates ‘‘fear of uncontrolled eating’’; Question 3 indicates ‘‘body weight loss’’; Question 4 indicates ‘‘body image disturbance’’; and Question 5 indicates ‘‘food domination over life’’. One point is given for each ‘‘yes’’ answer, and a subject whose total score is 2 or more is judged as likely having an
eating disorder. In this study, the SCOFF was first translated into Japanese by the authors, after which a bilingual Japanese psychiatrist assessed the accuracy of the Japanese version of the test. The SCOFF was administered orally by the psychiatrist/internist in charge during consultation. In Question 3 of this tool, ‘‘6 kg’’ was used in place of ‘‘one stone’’. The EAT-26 was also administered. Garner and Garfinkel developed the EAT [11] as an objective, self-reported, 40-item measure of the symptoms of anorexia nervosa, and the abbreviated 26-item version of the EAT (EAT-26) [12] has been proven to be valid. Subjects are required to judge whether an item should be answered as ‘‘always’’, ‘‘very often’’, ‘‘often’’, ‘‘sometimes’’, ‘‘rarely’’, or ‘‘never’’; and three points are given for an answer of ‘‘always’’, two points for ‘‘very often’’, one point for ‘‘often’’, and no points are given for other answers. The total points are the score of the EAT-26. The Japanese version of the EAT-26 [13] has been shown to have high construct validity and adequate reliability. The cut-off score for eating disorders was chosen as 15, according to a study on the validity of the EAT-26 for Japanese patients [14]. At the same time that the SCOFF and the EAT-26 were administered, the subjects were asked about the presence and frequency of the following abnormal eating patterns: engaging in dieting behavior, eating binges, vomiting after meals, chewing and spitting out, and taking laxatives or exercising to lose weight. On the basis of these data of eating patterns, diagnoses of the subjects at the time of investigation were assigned with subgroups of eating disorders according to the criteria specified in the Diagnostic and Statistical Manual of Mental Disorders, 4th ed. [15]. Analyses A Pearson’s correlation coefficient test was used to test the correlation between variables. A one-way ANOVA was used to test differences in characteristics among the subgroups of eating disorders and a Tukey’s test was selected for multiple comparisons. Differences in the detection rates between the SCOFF and EAT-26 were tested by McNemar’s test. A Kolmogorov/Smirnov test was used to test if groups showed a Gaussian distribution. The items of the EAT-26 were included in a logistic-regression model to identify items associated with each question of the SCOFF. The forward stepwise method based on likelihood-ratio statistics was used to construct the regression model. The level of significance was set at P B/0.05 in the univariate analyses and P B/0.10 in the multivariate analyses; all tests were two-tailed. All statistical analyses were performed on a personal computer with the statistical package SPSS for Windows Version 12.0J.
The SCOFF Questionnaire and Eating Attitudes Test in eating disorders Table II. SCOFF results.
Results Characteristics of the subjects
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
29
The characteristics of the subjects are shown in Table I. Values of body weight (kg) and body mass index (BMI, kg/m2) of patients with anorexia nervosa, restricting type (ANR), were significantly smaller than those of patients with bulimia nervosa, purging type (BNP), bulimia nervosa, non-purging type (BNNP), and ENDOS, and those of patients with anorexia nervosa, binge-eating/purging type (ANBP) were significantly smaller than those of patients with BNP or BNNP. Because no significant differences were found among subgroups in eating disorders on age, duration of eating disorders, or length of treatment, the subjects were considered homogeneous. Furthermore, the patients with EDNOS were classified into six subgroups: patients who had no fear of gaining weight and whose body weight was less than 85% of standard body weight (ANR without fear of getting weight), patients without body weight loss who had a fear of becoming fat and abnormal eating patterns like those of patients with ANR (ANR without weight loss), patients without body weight loss who binged and purged less than twice a week (BNP with rare binge-eating and purging), patients without body weight loss who had the habit of self-induced vomiting (habitforming vomiting), patients without body weight loss who binged in the absence of inappropriate compensatory behaviors (binge eating disorder), and others. The numbers of patients with each subtype of EDNOS were eight (29.6%), two (7.4%), five (18.5%), two (7.4%), six (22.2%), and four (14.8%), respectively.
Total n (%)
Score 0 1 2 3 4 5 Total Score ]/2
7 9 22 29 12 1 80 64
(8.8) (11.3) (27.5) (36.3) (15.0) (1.3) (100.0) (80.0)
AN and BN n (%) 1 1 15 27 9 0 53 51
(1.9) (1.9) (28.3) (50.9) (17.0) (0.0) (100.0) (96.2)
EDNOS n (%) 6 8 7 2 3 1 27 13
(22.2) (29.6) (25.9) (7.4) (11.1) (3.7) (100.0) (48.1)
AN, anorexia nervosa; BN, bulimia nervosa.
Results of SCOFF Table II shows the results of the SCOFF. Values of BMI were positively correlated with scores of the SCOFF (r /0.251, P /0.041), but age, duration of eating disorders, and length of treatment were not correlated with scores of the SCOFF. Furthermore, values of BMI and scores of the SCOFF were positively correlated in patients with EDNOS (r / 0.650, P /0.001), whereas they were not correlated in patients with anorexia nervosa or bulimia nervosa. Questions of SCOFF and subgroups of eating disorder The percentages of subjects in each subgroup who answered ‘‘yes’’ to each question of the SCOFF are shown in Table III. The rates of ‘‘yes’’ answers on Question 1 in patients with ANBP or BNP were significantly higher than those in other patients, whereas there were no significant differences among subtypes of eating disorders on questions 2, 3, 4, and 5. Furthermore, the rates of ‘‘yes’’ answers on Question 3 were lower than those on other questions in all subjects and in subjects of each subgroup.
Table I. Characteristics of the subjects (mean9/SD, range in parenthesis).
Number Age (years) Height (cm) Weight (kg) BMI (kg/m2) Duration of ED (years) Length of treatment (years)
Total 80 25.89/6.7 (15.8 /46.1) 157.99/5.2 (146.0 /170.0) 44.59/9.3 (25.4 /68.0) 17.89/3.2 (9.9 /24.7) 6.79/5.2 (0.3 /25.9) 4.59/4.5 (0.0 /22.9)
ANR 13
ANBP 10
BNP 26
BNNP 4
EDNOS 27
24.49/5.9
28.29/5.5
24.49/4.6
25.29/8.8
27.19/8.7
156.69/6.7
157.69/3.2
158.79/4.4
158.79/8.3
157.89/5.4
35.59/5.4
38.49/6.0
49.29/5.5ab
59.59/6.4ab
45.69/10.1a
14.49/1.7
15.59/2.6
19.69/1.6cd
22.09/2.2cd
18.39/3.6c
4.09/3.9
6.59/2.2
6.59/5.0
5.09/2.8
7.99/6.4
3.59/3.7
5.19/3.2
4.19/5.0
3.89/2.9
4.99/5.0
ED, eating disorders. a Significantly larger than ANR (BNP, P B/0.001; BNNP, P /0.001; EDNOS, P /0.003). b Significantly larger than ANBP (BNP, P /0.004; BNNP, P /0.005). c Significantly larger than ANR (BNP, P B/0.001; BNNP, P /0.002; EDNOS, P /0.001). d Significantly larger than ANBP (BNP, P /0.001; BNNP, P /0.014).
30
S. Noma et al.
Table III. Subjects who answered ‘‘yes’’ for each questions of SCOFF (Number, percentage in parentheses).
Question 1 (Self-induced vomiting) Question 2 (Fear of uncontrolled eating) Question 3 (Body weight loss) Question 4 (Body image disturbance) Question 5 (Food domination over life) a
Total (n /80)
ANR (n /13)
ANBP (n /10)
BNP (n /26)
BNNP (n /4)
EDNOS (n /27)
41 (51.3) 61 (76.3) 6 (7.5) 26 (32.5) 58 (72.5)
1 (7.7ab) 8 (61.5) 2 (15.4) 6 (46.2) 10 (76.9)
10 (100.0) 9 (90.0) 0 (0.0) 3 (30.0) 8 (80.0)
23 (88.5) 24 (92.3) 1 (3.8) 11 (42.3) 22 (84.6)
1 (25.0ab) 3 (75.0) 1 (25.0) 1 (25.0) 3 (75.0)
6 (22.2ab) 17 (63.0) 2 (7.4) 5 (18.5) 15 (55.6)
Smaller than ANBP (ANR, P B/0.001; ANNP, P /0.004; EDNOS, P B/0.001). Smaller than BNP (ANR, P B/0.001; BNNP, P /0.009; EDNOS, P B/0.001).
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
b
Association between each question of the SCOFF and items of the EAT-26 Table IV shows the items of the EAT-26 that were extracted as significantly associated with each question of the SCOFF by a logistic-regression model. There was no overlap among extracted items of the EAT-26, which indicates, therefore, that five questions of the SCOFF had been selected reasonably. Comparison between SCOFF and EAT-26 The scores of the SCOFF and those of the EAT-26 were positively correlated (r /0.575, P B/0.001). Table V shows the results of the SCOFF and the EAT-26 for patients with anorexia nervosa or bulimia nervosa. Although the mean score of the SCOFF for patients with ANR was significantly lower than that for patients with ANBP and that for patients with BNP, there were no significant differences among subtypes of eating disorders in the detection rate with the SCOFF. On the other hand, there were no significant differences among subtypes of eating disorders in the mean score and the detection rate with the EAT-26. For all subjects and for subjects in each subgroup, there were no significant differences between detection rates with the SCOFF and those with the EAT-26.
The detection rates with the SCOFF and the EAT-26 for EDNOS are shown in Table VI. The detection rate for ANR without fear of gaining weight was significantly lower than that for BNP with rare binge-eating and purging, habit-forming vomiting, and binge eating disorder. The detection rate for patients with low body weight and EDNOS, namely, ANR without fear of gaining weight and ANR with normal weight, was only 10% (1 /10). There were no differences among subtypes of EDNOS in the detection rates with the EAT-26. Discussion This study has found that results of the SCOFF precisely reflect the characteristics of patients with eating disorders, because scores of the SCOFF were strongly correlated with those of the EAT-26, the gold standard instrument for eating disorders, and there were no significant differences in the detection rates between the SCOFF and the EAT-26 for all subjects and for subjects with each subtype of eating disorder. The detection rate with the SCOFF for anorexia nervosa and bulimia nervosa was high, namely 96.2%. Furthermore, each question of the SCOFF is considered to have been selected
Table IV. Association between each question of the SCOFF and items of the EAT-26 by a logistic-regression model.
SCOFF
Associated Item number of EAT-26
Item content
P
Question 1
9 7 26 20
Vomit after I have eaten Avoid foods with high carbohydrate content Have the impulse to vomit after meals Feel that others pressure me to eat
Question 2
1 4
Am terrified about being overweight Have gone on eating binges where I feel that I may not be able to stop
0.001 0.025
Question 3
17
Eat diet foods
0.081
B/0.001 0.001 0.002 0.006
Question 4
11
Am preoccupied with a desire to be thinner
B/0.001
Question 5
18
Feel that food controls my life
B/0.001
The SCOFF Questionnaire and Eating Attitudes Test in eating disorders
31
Table V. Comparison of detection rates of eating disorders between SCOFF and EAT-26.
Mean score of SCOFF EAT-26 Detective rate of SCOFF (%) (score ]/2) EAT-26 (%) (score ]/15)
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
a
AN and BN (n /53)
ANR (n /13)
ANBP (n /10)
2.779/0.824 28.79/14.2 96.2 (N /51) 88.7 (N /47)
2.089/0.862 25.19/11.0 84.6 (N /11) 92.3 (N /12)
BNP (n /26)
3.009/0.667a 34.39/18.1 100.0 (N /10) 90.0 (N /9)
BNNP (n /4)
3.129/0.653a 29.69/13.9 100.0 (N /26) 88.5 (N /23)
2.259/0.500 21.09/13.7 100.0 (N /4) 75.0 (N /3)
Larger than ANR (ANBP, P /0.016; BNP, P B/0.001).
appropriately because there was no overlap among items of the EAT-26 associated with questions of the SCOFF. On the other hand, the two subjects in eating disorders other than EDNOS who could not be detected with the SCOFF were patients with anorexia nervosa, restricting type. Furthermore, among subjects with EDNOS, all eight patients with restricting-type anorexia nervosa without fear of gaining weight were excluded with the SCOFF, although the EAT-26 could detect only one of these patients. The result that values of BMI were significantly correlated with scores of the SCOFF could be considered to reflect a low detection rate with this screening tool for patients with low body weight and EDNOS. It was understandable that the rate of subjects who answered ‘‘yes’’ to Question 3, which asks about recent extreme body weight loss, was low because the subjects of this study were patients who had already received treatment for some time. Of course, the SCOFF was developed as a screening tool to detect patients with eating disorders at an early stage in the general population, but there are patients with anorexia nervosa whose body weight has been decreasing gradually for a long period. If Question 3 of the SCOFF is modified to ask not about body weight loss but about low body weight, for example, ‘‘Is your body mass index less than 17.5 kg/m2?’’, the detection rate for EDNOS with this tool would be higher, although the questions would be less simple and patients who have only recently started to lose weight might be excluded.
This study has found that the SCOFF questionnaire is exceedingly valid for screening eating disorders, especially anorexia nervosa and bulimia nervosa in the narrow sense, but the question about body weight loss should be modified to detect patients with EDNOS more accurately. The SCOFF would then be expected to in greater detail investigate the characteristics and psychopathology of EDNOS. Key points . The SCOFF and the EAT-26 were administered to 80 patients with eating disorders in Japan . The scores of the SCOFF and those of the EAT26 were positively correlated . The detection rates of the SCOFF for patients with anorexia nervosa or bulimia nervosa and that for patients with EDNOS were 96.2 and 48.1%, respectively . Each question of the SCOFF has been selected appropriately because there was no overlap among items of the EAT-26 associated with questions of the SCOFF . The detection rate with the SCOFF of patients with low body weight and EDNOS was low (10%) Statement of interest The authors have no conflict of interest with any commercial or other associations in connection with the submitted article.
Table VI. Comparison of detection rates of EDNOS specified between SCOFF and EAT-26 (number, percentage in parenthesis).
SCOFF (score ]/2) EAT-26 (score]/15)
Total (n /27)
ANR without fear of getting weight (n /8)
ANR with normal weight (n /2)
BNP with rare bingeeating and purging (n /5)
Habit-forming vomiting (n /2)
Binge eating disorder (n /6)
Others (n /4)
13 (48.1) 10 (37.0)
0 (0.0) 1 (12.5)
1 (50.0) 2 (100.0)
5 (100.0) 4 (80.0)
2 (100.0) 1 (50.0)
4 (66.7) 2 (33.3)
1 (25.0) 0 (0.0)
32
S. Noma et al.
References [1] Nakai Y. Epidemiology of eating disorders. Psychosom Med 2000;4:1 /9. [2] Whitehouse AM, Cooper PJ, Vize CV, Hill C, Vogel L. Prevalence of eating disorders in three Cambridge general practices: Hidden and conspicuous morbidity. Br J Gen Pract 1992;42(355):57 /60. [3] Rathner G, Messner K. Detection of eating disorders in a small rural town: An epidemiological study. Psychol Med 1993;23(1):175 /84. [4] Steinhausen HC, Winkler C, Meier M. Eating disorders in adolescence in a Swiss epidemiological study. Int J Eat Disord 1997;22(2):147 /51. [5] Santonastaso P, Zanetti T, Sala A, Favaretto G, Vidotto G, Favaro A. Prevalence of eating disorders in Italy: A survey on a sample of 16-year-old female students. Psychother Psychosom 1996;65(3):158 /62. [6] Nakai Y. The epidemiology of eating disorders: Data from Japan. In: Maj M, Halmi K, Lo´pez-Ibor J, Sartorius N, editors. Eating disorders. WPA series, evidence and experience in psychiatry, vol. 6. Chichester: John Wiley & Sons Ltd; 2003. p 126 /8. [7] Morgan JF, Reid F, Lacey JH. The SCOFF questionnaire: Assessment of a new screening tool for eating disorders. Br Med J 1999;319(7223):1467 /8. /
/
/
/
/
/
/
/
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
/
/
/
/
[8] Luck AJ, Morgan JF, Reid F, O’Brien A, Brunton J, Price C, Perry L, Lacey JH. The SCOFF questionnaire and clinical interview for eating disorders in general practice: comparative study. Br Med J 2002;325(7367):755 /6. [9] Perry L, Morgan J, Reid F, Brunton J, O’’Brien A, Luck A, Lacey H. Screening for symptoms of eating disorders: Reliability of the SCOFF screening tool with written compared to oral delivery. Int J Eat Disord 2002;32(4): 466 /72. [10] Fairburn CG, Harrison PJ. Eating disorders. Lancet 2003; 361(9355):407 /16. [11] Garner DM, Garfinkel PE. The Eating Attitudes Test: An index of the symptoms of anorexia nervosa. Psychol Med 1979;9(2):273 /9. [12] Garner DM, Olmsted MP, Bohr Y, Garfinkel PE. The eating attitudes test: Psychometric features and clinical correlates. Psychol Med 1982;12(4):871 /8. [13] Nagata T, Kiriike N, Yoshino S, Nishiwaki S, Takeuchi N, Tanaka M, Kawakita Y. Reliability and validity of the Eating Attitudes Test in patients with eating disorders. Rinshoseishinigaku 1989;18(8):1279 /86. [14] Nakai Y. Validity of Eating Attitudes Test (EAT). Seishinigaku 2003;45(2):161 /5. [15] American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 4th ed. American Psychiatric Association, Washington; 1994. /
/
/
/
/
/
/
/
/
/
/
/
/
/
ORIGINAL ARTICLE
Comparison between the SCOFF Questionnaire and the Eating Attitudes Test in patients with eating disorders
SHUN’ICHI NOMA1, YOSHIKATSU NAKAI2, SEIJI HAMAGAKI3, MINAKO UEHARA1, AKIKO HAYASHI1 & TAKUJI HAYASHI1 Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
1
Department of Psychiatry, Graduate School of Medicine, Kyoto University, Kyoto, Japan, 2School of Health Sciences, Faculty of Medicine, Kyoto University, Kyoto, Japan, and 3Takagi Psychiatric Clinic, Kyoto, Japan
Abstract Objective: The SCOFF was developed as a simple, five-question screening tool for eating disorders to be used in primary care. The aim of this study was to examine the appropriateness of each question in comparison with the Eating Attitudes Test-26. Methods: The SCOFF and the EAT-26 were administered to 80 patients with eating disorders who had received treatment from May through October 2003 in Japan. Results: The scores of the SCOFF and those of the EAT-26 were positively correlated (P B/0.001), and each question of the SCOFF was strongly associated with different items of the EAT26. The detection rates with the SCOFF of the patients with anorexia nervosa or bulimia nervosa and that of the patients with eating disorders not otherwise specified (EDNOS) were 96.2 and 48.1%, respectively. The scores of the SCOFF and the values of body mass index were significantly correlated (P /0.041), and the detection rate with the SCOFF of patients with low body weight and EDNOS was low (10%). Conclusion: Each question of the SCOFF has been selected appropriately for reflecting characteristics of eating disorders, although one question about body weight loss should be modified to detect patients with anorexia nervosa more accurately.
Key Words: Eating disorders, screening tool, SCOFF, Eating Attitudes Test-26, primary care
Introduction Eating disorders are prevalent among young females in Western countries and in several Asian countries. According to a nationwide epidemiological survey in Japan, the estimated prevalences of anorexia nervosa and bulimia nervosa among females aged 10 /29 years in 1998 were 0.06 and 0.03%, respectively [1]. These rates were lower than those in Western countries: among young white females the prevalence of anorexia nervosa is between 0.2 and 0.7% [2/4], and that of bulimia nervosa is between 0.5 and 1.5% [2,4,5]. The situation of eating disorders in Japan, however, is expected to become similar to that in Western countries in the near future because the prevalence of eating disorders in Japan has increased markedly in recent years: the estimated prevalences of anorexia nervosa and bulimia nervosa in females aged 10 /29 years in 1992 were only 0.02 and 0.007% [6]. Doctors in Eastern and Western countries have opportunities to encounter patients with eating disorders in daily primary care.
Morgan et al. [7] introduced the SCOFF questionnaire as a simple screening tool for eating disorders. The SCOFF consists of only five questions, and its high sensitivity and specificity have been confirmed [8]. Because the validity of the SCOFF administered as a written questionnaire as well as an oral interview has also been established [9], this scale can be used in various clinical situations. However, Morgan et al. did not explain in their original paper why they chose these questions as items of the SCOFF, and subsequent studies have not verified the appropriateness of each question. Furthermore, detecting patients who do not meet the full criteria for anorexia nervosa or bulimia nervosa may be more difficult and less pertinent for this screening tool [8]. Although this supposition is understandable, eating disorder not otherwise specified (EDNOS) is now the most common diagnostic category of eating disorders [10]. The characteristics of patients with EDNOS who cannot be detected with the SCOFF may reveal weak points of this tool, and screening for eating disorders
Correspondence: Shun’ichi Noma, M.D., Department of Psychiatry, Graduate School of Medicine, Kyoto University, 54 Shogoin-Kawaharacho, Sakyo, Kyoto, 606-8507 Japan. Tel: /81 75 751 3373. Fax: /81 75 751 3246. E-mail: [email protected]
(Received 21 September 2004; accepted 18 July 2005) ISSN 1365-1501 print/ISSN 1471-1788 online # 2006 Taylor & Francis DOI: 10.1080/13651500500305275
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
28
S. Noma et al.
would be more effectively if these weak points were eliminated. The purpose of this study was to evaluate each question of the SCOFF by administering it to patients with eating disorders, including EDNOS. For that purpose, the Eating Attitudes Test-26 (EAT-26) was administered at the same time. The EAT-40, and then the EAT-26, were originally developed by Garner et al. [11,12] as a rating scale that accurately assesses the eating patterns of patients with anorexia nervosa, but it can be used to screen for eating disorders because there is little overlap in the frequency distributions of patients with eating disorders and healthy subjects [11,12]. In this paper, the results of the SCOFF were compared with those of the EAT-26 in patients with eating disorders, and the effectiveness of the SCOFF was verified.
Method Subjects The subjects were 80 patients with eating disorders who attended Kyoto University Hospital or related hospitals from May through October 2003. All subjects were Japanese females. Informed consent The study was fully explained to the participants before examination, and the confidentiality of responses was assured. All participants agreed with the purpose and contents of this study. Procedures The SCOFF was developed by Morgan et al. [7] as a screening instrument for addressing the core features of anorexia nervosa and bulimia nervosa, as follows: 1. Do you make yourself (S) sick because you feel uncomfortably full? 2. Do you worry you have lost (C) control over how much you eat? 3. Have you recently lost more than (O) one stone in a 3-month period? 4. Do you believe yourself to be (F) fat when others say you are too thin? 5. Would you say that (F) food dominates your life? Question 1 indicates ‘‘self-induced vomiting’’; Question 2 indicates ‘‘fear of uncontrolled eating’’; Question 3 indicates ‘‘body weight loss’’; Question 4 indicates ‘‘body image disturbance’’; and Question 5 indicates ‘‘food domination over life’’. One point is given for each ‘‘yes’’ answer, and a subject whose total score is 2 or more is judged as likely having an
eating disorder. In this study, the SCOFF was first translated into Japanese by the authors, after which a bilingual Japanese psychiatrist assessed the accuracy of the Japanese version of the test. The SCOFF was administered orally by the psychiatrist/internist in charge during consultation. In Question 3 of this tool, ‘‘6 kg’’ was used in place of ‘‘one stone’’. The EAT-26 was also administered. Garner and Garfinkel developed the EAT [11] as an objective, self-reported, 40-item measure of the symptoms of anorexia nervosa, and the abbreviated 26-item version of the EAT (EAT-26) [12] has been proven to be valid. Subjects are required to judge whether an item should be answered as ‘‘always’’, ‘‘very often’’, ‘‘often’’, ‘‘sometimes’’, ‘‘rarely’’, or ‘‘never’’; and three points are given for an answer of ‘‘always’’, two points for ‘‘very often’’, one point for ‘‘often’’, and no points are given for other answers. The total points are the score of the EAT-26. The Japanese version of the EAT-26 [13] has been shown to have high construct validity and adequate reliability. The cut-off score for eating disorders was chosen as 15, according to a study on the validity of the EAT-26 for Japanese patients [14]. At the same time that the SCOFF and the EAT-26 were administered, the subjects were asked about the presence and frequency of the following abnormal eating patterns: engaging in dieting behavior, eating binges, vomiting after meals, chewing and spitting out, and taking laxatives or exercising to lose weight. On the basis of these data of eating patterns, diagnoses of the subjects at the time of investigation were assigned with subgroups of eating disorders according to the criteria specified in the Diagnostic and Statistical Manual of Mental Disorders, 4th ed. [15]. Analyses A Pearson’s correlation coefficient test was used to test the correlation between variables. A one-way ANOVA was used to test differences in characteristics among the subgroups of eating disorders and a Tukey’s test was selected for multiple comparisons. Differences in the detection rates between the SCOFF and EAT-26 were tested by McNemar’s test. A Kolmogorov/Smirnov test was used to test if groups showed a Gaussian distribution. The items of the EAT-26 were included in a logistic-regression model to identify items associated with each question of the SCOFF. The forward stepwise method based on likelihood-ratio statistics was used to construct the regression model. The level of significance was set at P B/0.05 in the univariate analyses and P B/0.10 in the multivariate analyses; all tests were two-tailed. All statistical analyses were performed on a personal computer with the statistical package SPSS for Windows Version 12.0J.
The SCOFF Questionnaire and Eating Attitudes Test in eating disorders Table II. SCOFF results.
Results Characteristics of the subjects
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
29
The characteristics of the subjects are shown in Table I. Values of body weight (kg) and body mass index (BMI, kg/m2) of patients with anorexia nervosa, restricting type (ANR), were significantly smaller than those of patients with bulimia nervosa, purging type (BNP), bulimia nervosa, non-purging type (BNNP), and ENDOS, and those of patients with anorexia nervosa, binge-eating/purging type (ANBP) were significantly smaller than those of patients with BNP or BNNP. Because no significant differences were found among subgroups in eating disorders on age, duration of eating disorders, or length of treatment, the subjects were considered homogeneous. Furthermore, the patients with EDNOS were classified into six subgroups: patients who had no fear of gaining weight and whose body weight was less than 85% of standard body weight (ANR without fear of getting weight), patients without body weight loss who had a fear of becoming fat and abnormal eating patterns like those of patients with ANR (ANR without weight loss), patients without body weight loss who binged and purged less than twice a week (BNP with rare binge-eating and purging), patients without body weight loss who had the habit of self-induced vomiting (habitforming vomiting), patients without body weight loss who binged in the absence of inappropriate compensatory behaviors (binge eating disorder), and others. The numbers of patients with each subtype of EDNOS were eight (29.6%), two (7.4%), five (18.5%), two (7.4%), six (22.2%), and four (14.8%), respectively.
Total n (%)
Score 0 1 2 3 4 5 Total Score ]/2
7 9 22 29 12 1 80 64
(8.8) (11.3) (27.5) (36.3) (15.0) (1.3) (100.0) (80.0)
AN and BN n (%) 1 1 15 27 9 0 53 51
(1.9) (1.9) (28.3) (50.9) (17.0) (0.0) (100.0) (96.2)
EDNOS n (%) 6 8 7 2 3 1 27 13
(22.2) (29.6) (25.9) (7.4) (11.1) (3.7) (100.0) (48.1)
AN, anorexia nervosa; BN, bulimia nervosa.
Results of SCOFF Table II shows the results of the SCOFF. Values of BMI were positively correlated with scores of the SCOFF (r /0.251, P /0.041), but age, duration of eating disorders, and length of treatment were not correlated with scores of the SCOFF. Furthermore, values of BMI and scores of the SCOFF were positively correlated in patients with EDNOS (r / 0.650, P /0.001), whereas they were not correlated in patients with anorexia nervosa or bulimia nervosa. Questions of SCOFF and subgroups of eating disorder The percentages of subjects in each subgroup who answered ‘‘yes’’ to each question of the SCOFF are shown in Table III. The rates of ‘‘yes’’ answers on Question 1 in patients with ANBP or BNP were significantly higher than those in other patients, whereas there were no significant differences among subtypes of eating disorders on questions 2, 3, 4, and 5. Furthermore, the rates of ‘‘yes’’ answers on Question 3 were lower than those on other questions in all subjects and in subjects of each subgroup.
Table I. Characteristics of the subjects (mean9/SD, range in parenthesis).
Number Age (years) Height (cm) Weight (kg) BMI (kg/m2) Duration of ED (years) Length of treatment (years)
Total 80 25.89/6.7 (15.8 /46.1) 157.99/5.2 (146.0 /170.0) 44.59/9.3 (25.4 /68.0) 17.89/3.2 (9.9 /24.7) 6.79/5.2 (0.3 /25.9) 4.59/4.5 (0.0 /22.9)
ANR 13
ANBP 10
BNP 26
BNNP 4
EDNOS 27
24.49/5.9
28.29/5.5
24.49/4.6
25.29/8.8
27.19/8.7
156.69/6.7
157.69/3.2
158.79/4.4
158.79/8.3
157.89/5.4
35.59/5.4
38.49/6.0
49.29/5.5ab
59.59/6.4ab
45.69/10.1a
14.49/1.7
15.59/2.6
19.69/1.6cd
22.09/2.2cd
18.39/3.6c
4.09/3.9
6.59/2.2
6.59/5.0
5.09/2.8
7.99/6.4
3.59/3.7
5.19/3.2
4.19/5.0
3.89/2.9
4.99/5.0
ED, eating disorders. a Significantly larger than ANR (BNP, P B/0.001; BNNP, P /0.001; EDNOS, P /0.003). b Significantly larger than ANBP (BNP, P /0.004; BNNP, P /0.005). c Significantly larger than ANR (BNP, P B/0.001; BNNP, P /0.002; EDNOS, P /0.001). d Significantly larger than ANBP (BNP, P /0.001; BNNP, P /0.014).
30
S. Noma et al.
Table III. Subjects who answered ‘‘yes’’ for each questions of SCOFF (Number, percentage in parentheses).
Question 1 (Self-induced vomiting) Question 2 (Fear of uncontrolled eating) Question 3 (Body weight loss) Question 4 (Body image disturbance) Question 5 (Food domination over life) a
Total (n /80)
ANR (n /13)
ANBP (n /10)
BNP (n /26)
BNNP (n /4)
EDNOS (n /27)
41 (51.3) 61 (76.3) 6 (7.5) 26 (32.5) 58 (72.5)
1 (7.7ab) 8 (61.5) 2 (15.4) 6 (46.2) 10 (76.9)
10 (100.0) 9 (90.0) 0 (0.0) 3 (30.0) 8 (80.0)
23 (88.5) 24 (92.3) 1 (3.8) 11 (42.3) 22 (84.6)
1 (25.0ab) 3 (75.0) 1 (25.0) 1 (25.0) 3 (75.0)
6 (22.2ab) 17 (63.0) 2 (7.4) 5 (18.5) 15 (55.6)
Smaller than ANBP (ANR, P B/0.001; ANNP, P /0.004; EDNOS, P B/0.001). Smaller than BNP (ANR, P B/0.001; BNNP, P /0.009; EDNOS, P B/0.001).
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
b
Association between each question of the SCOFF and items of the EAT-26 Table IV shows the items of the EAT-26 that were extracted as significantly associated with each question of the SCOFF by a logistic-regression model. There was no overlap among extracted items of the EAT-26, which indicates, therefore, that five questions of the SCOFF had been selected reasonably. Comparison between SCOFF and EAT-26 The scores of the SCOFF and those of the EAT-26 were positively correlated (r /0.575, P B/0.001). Table V shows the results of the SCOFF and the EAT-26 for patients with anorexia nervosa or bulimia nervosa. Although the mean score of the SCOFF for patients with ANR was significantly lower than that for patients with ANBP and that for patients with BNP, there were no significant differences among subtypes of eating disorders in the detection rate with the SCOFF. On the other hand, there were no significant differences among subtypes of eating disorders in the mean score and the detection rate with the EAT-26. For all subjects and for subjects in each subgroup, there were no significant differences between detection rates with the SCOFF and those with the EAT-26.
The detection rates with the SCOFF and the EAT-26 for EDNOS are shown in Table VI. The detection rate for ANR without fear of gaining weight was significantly lower than that for BNP with rare binge-eating and purging, habit-forming vomiting, and binge eating disorder. The detection rate for patients with low body weight and EDNOS, namely, ANR without fear of gaining weight and ANR with normal weight, was only 10% (1 /10). There were no differences among subtypes of EDNOS in the detection rates with the EAT-26. Discussion This study has found that results of the SCOFF precisely reflect the characteristics of patients with eating disorders, because scores of the SCOFF were strongly correlated with those of the EAT-26, the gold standard instrument for eating disorders, and there were no significant differences in the detection rates between the SCOFF and the EAT-26 for all subjects and for subjects with each subtype of eating disorder. The detection rate with the SCOFF for anorexia nervosa and bulimia nervosa was high, namely 96.2%. Furthermore, each question of the SCOFF is considered to have been selected
Table IV. Association between each question of the SCOFF and items of the EAT-26 by a logistic-regression model.
SCOFF
Associated Item number of EAT-26
Item content
P
Question 1
9 7 26 20
Vomit after I have eaten Avoid foods with high carbohydrate content Have the impulse to vomit after meals Feel that others pressure me to eat
Question 2
1 4
Am terrified about being overweight Have gone on eating binges where I feel that I may not be able to stop
0.001 0.025
Question 3
17
Eat diet foods
0.081
B/0.001 0.001 0.002 0.006
Question 4
11
Am preoccupied with a desire to be thinner
B/0.001
Question 5
18
Feel that food controls my life
B/0.001
The SCOFF Questionnaire and Eating Attitudes Test in eating disorders
31
Table V. Comparison of detection rates of eating disorders between SCOFF and EAT-26.
Mean score of SCOFF EAT-26 Detective rate of SCOFF (%) (score ]/2) EAT-26 (%) (score ]/15)
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
a
AN and BN (n /53)
ANR (n /13)
ANBP (n /10)
2.779/0.824 28.79/14.2 96.2 (N /51) 88.7 (N /47)
2.089/0.862 25.19/11.0 84.6 (N /11) 92.3 (N /12)
BNP (n /26)
3.009/0.667a 34.39/18.1 100.0 (N /10) 90.0 (N /9)
BNNP (n /4)
3.129/0.653a 29.69/13.9 100.0 (N /26) 88.5 (N /23)
2.259/0.500 21.09/13.7 100.0 (N /4) 75.0 (N /3)
Larger than ANR (ANBP, P /0.016; BNP, P B/0.001).
appropriately because there was no overlap among items of the EAT-26 associated with questions of the SCOFF. On the other hand, the two subjects in eating disorders other than EDNOS who could not be detected with the SCOFF were patients with anorexia nervosa, restricting type. Furthermore, among subjects with EDNOS, all eight patients with restricting-type anorexia nervosa without fear of gaining weight were excluded with the SCOFF, although the EAT-26 could detect only one of these patients. The result that values of BMI were significantly correlated with scores of the SCOFF could be considered to reflect a low detection rate with this screening tool for patients with low body weight and EDNOS. It was understandable that the rate of subjects who answered ‘‘yes’’ to Question 3, which asks about recent extreme body weight loss, was low because the subjects of this study were patients who had already received treatment for some time. Of course, the SCOFF was developed as a screening tool to detect patients with eating disorders at an early stage in the general population, but there are patients with anorexia nervosa whose body weight has been decreasing gradually for a long period. If Question 3 of the SCOFF is modified to ask not about body weight loss but about low body weight, for example, ‘‘Is your body mass index less than 17.5 kg/m2?’’, the detection rate for EDNOS with this tool would be higher, although the questions would be less simple and patients who have only recently started to lose weight might be excluded.
This study has found that the SCOFF questionnaire is exceedingly valid for screening eating disorders, especially anorexia nervosa and bulimia nervosa in the narrow sense, but the question about body weight loss should be modified to detect patients with EDNOS more accurately. The SCOFF would then be expected to in greater detail investigate the characteristics and psychopathology of EDNOS. Key points . The SCOFF and the EAT-26 were administered to 80 patients with eating disorders in Japan . The scores of the SCOFF and those of the EAT26 were positively correlated . The detection rates of the SCOFF for patients with anorexia nervosa or bulimia nervosa and that for patients with EDNOS were 96.2 and 48.1%, respectively . Each question of the SCOFF has been selected appropriately because there was no overlap among items of the EAT-26 associated with questions of the SCOFF . The detection rate with the SCOFF of patients with low body weight and EDNOS was low (10%) Statement of interest The authors have no conflict of interest with any commercial or other associations in connection with the submitted article.
Table VI. Comparison of detection rates of EDNOS specified between SCOFF and EAT-26 (number, percentage in parenthesis).
SCOFF (score ]/2) EAT-26 (score]/15)
Total (n /27)
ANR without fear of getting weight (n /8)
ANR with normal weight (n /2)
BNP with rare bingeeating and purging (n /5)
Habit-forming vomiting (n /2)
Binge eating disorder (n /6)
Others (n /4)
13 (48.1) 10 (37.0)
0 (0.0) 1 (12.5)
1 (50.0) 2 (100.0)
5 (100.0) 4 (80.0)
2 (100.0) 1 (50.0)
4 (66.7) 2 (33.3)
1 (25.0) 0 (0.0)
32
S. Noma et al.
References [1] Nakai Y. Epidemiology of eating disorders. Psychosom Med 2000;4:1 /9. [2] Whitehouse AM, Cooper PJ, Vize CV, Hill C, Vogel L. Prevalence of eating disorders in three Cambridge general practices: Hidden and conspicuous morbidity. Br J Gen Pract 1992;42(355):57 /60. [3] Rathner G, Messner K. Detection of eating disorders in a small rural town: An epidemiological study. Psychol Med 1993;23(1):175 /84. [4] Steinhausen HC, Winkler C, Meier M. Eating disorders in adolescence in a Swiss epidemiological study. Int J Eat Disord 1997;22(2):147 /51. [5] Santonastaso P, Zanetti T, Sala A, Favaretto G, Vidotto G, Favaro A. Prevalence of eating disorders in Italy: A survey on a sample of 16-year-old female students. Psychother Psychosom 1996;65(3):158 /62. [6] Nakai Y. The epidemiology of eating disorders: Data from Japan. In: Maj M, Halmi K, Lo´pez-Ibor J, Sartorius N, editors. Eating disorders. WPA series, evidence and experience in psychiatry, vol. 6. Chichester: John Wiley & Sons Ltd; 2003. p 126 /8. [7] Morgan JF, Reid F, Lacey JH. The SCOFF questionnaire: Assessment of a new screening tool for eating disorders. Br Med J 1999;319(7223):1467 /8. /
/
/
/
/
/
/
/
Int J Psych Clin Pract Downloaded from informahealthcare.com by CDL-UC Santa Cruz on 11/24/14 For personal use only.
/
/
/
/
[8] Luck AJ, Morgan JF, Reid F, O’Brien A, Brunton J, Price C, Perry L, Lacey JH. The SCOFF questionnaire and clinical interview for eating disorders in general practice: comparative study. Br Med J 2002;325(7367):755 /6. [9] Perry L, Morgan J, Reid F, Brunton J, O’’Brien A, Luck A, Lacey H. Screening for symptoms of eating disorders: Reliability of the SCOFF screening tool with written compared to oral delivery. Int J Eat Disord 2002;32(4): 466 /72. [10] Fairburn CG, Harrison PJ. Eating disorders. Lancet 2003; 361(9355):407 /16. [11] Garner DM, Garfinkel PE. The Eating Attitudes Test: An index of the symptoms of anorexia nervosa. Psychol Med 1979;9(2):273 /9. [12] Garner DM, Olmsted MP, Bohr Y, Garfinkel PE. The eating attitudes test: Psychometric features and clinical correlates. Psychol Med 1982;12(4):871 /8. [13] Nagata T, Kiriike N, Yoshino S, Nishiwaki S, Takeuchi N, Tanaka M, Kawakita Y. Reliability and validity of the Eating Attitudes Test in patients with eating disorders. Rinshoseishinigaku 1989;18(8):1279 /86. [14] Nakai Y. Validity of Eating Attitudes Test (EAT). Seishinigaku 2003;45(2):161 /5. [15] American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, 4th ed. American Psychiatric Association, Washington; 1994. /
/
/
/
/
/
/
/
/
/
/
/
/
/
