Skip to main content

Table 2 Characteristics of studies included in this review

From: Psychometric properties of instruments assessing exercise in patients with eating disorders: a systematic review

Author (Year) (Country) Participants Factor structure* Cut-off scores Internal consistency
Studies focusing on the Exercise and Eating Disorders (EED)
 Danielsen (2012) (Norway) [3] N = 50 female inpatient ED participants (57 before exclusion). AN = 25, 50%, BN = 10, 20%, EDNOS = 15, 30%, 15 participants. N = 51 high school and university students as control group Items were divided into three subscales based on clinical exeperience: intentions to exercise (subscale 1), consequences of not exercising (subscale 2), bodily sensations (subscale 3). No factor analysis conducted Cut-off score for compulsive exercise = 50, represents ≥ average score on each item given the maximal possible total of 100 Cronbach’s α in whole sample: .92 (sum score), .66 (subscale 1), .93 (subscale 2), .87 (subscale 3). Cronbach’s α in patient group: .89 (sum score), .66 (subscale 1), .94 (subscale 2), .78 (subscale 3). Cronbach’s α in control group: .79 (sum score), .39 (subscale 1), .84 (subscale 2), .80 (subscale 3)
 Danielsen (2015) (Norway) [50] N = 235 female ED participants (244 before exclusion),
AN = 79, 32.4%, BN = 57, 23.4%, EDNOS = 84, 34.4%, BED = 24, 9.8%. N = 205 female controls
Four factor structure determined by PCA: (1) compulsive exercise (CE), (2) positive and healthy exercise (PHE), (3) awareness of bodily symptoms (BS) and (4) weight and shape related exercise (WSE) Clinical severity guide: Group 1, global score < 1.80 (no symptoms of compulsive exercise);
Group 2, global score 1.80–2.39 (low severity);
Group 3, global score 2.40–3.19 (moderate severity)
and Group 4, global score > 3.20 (high severity)
Cronbach’s α in whole sample: .90. (sum score), .93 (factor 1), .82 (factor 2), .80 (factor 3), .89 (factor 4)
 Danielsen (2018) (Norway) [51] N = 55 male ED in- and outpatients, AN = 30, 54.4%, BN = 10, 18.2%, and unspecified ED including binge eating = 15, 27.2%, N = 203 male controls included (95%) (214 before exclusion) Four factor structure found in Danielsen (2015) was confirmed by PCA. Four factor solution explained the most variance (73%). Factor 1 (Eigenvalue 6.27) explained 34.81%, factor 2 (Eigenvalue 4.02) explained 22.33%, factor 3 (Eigenvalue 1.89) explained 10.46% and factor 4 (Eigenvalue 0.97) explained 5.39% as in Danielsen (2015) Conrbach’s α in whole sample: .92 (factor 1), .89 (factor 2), .88 (factor 3), .72 (factor 4)
Studies focusing on the Compulsive Exercise Test (CET)
 Goodwin et al. (2011) (United Kingdom) [53] N = 1012 secondary school students 12–14 years, 1725 before exclusion, male = 45.3%, female = 54.7% Five factor structure confirmed by exploratory PCA. Five factors explained 64.1% of variance. Factor 1 (Eigenvalue 7.89) explained 32.88%, factor 2 (Eigenvalue 2.71) 11.28%, factor 3 (Eigenvalue 1.96) 8.17%, factor 4 (Eigenvalue 1.11) 4.62% and factor 5 (Eigenvalue 1.07) 4.46% n/a Cronbach’s α: .88 (overall scale), .87 (factor 1), .77 (factor 2), .79 (factor 3), .71 (factor 4), .77 (factor 5)
 Formby (2014) (Australia) [52] N = 104 adolescents aged 12–17, 93% female, AN = 38%, BN = 11%, OFED/ USFED = 51% Study was unable to confirm a factor structure. Four models were tested with confirmatory factor analysis, none provided adequate fit. Original five factor model suggested by Taranis et al. (2011) provided best fit out of the four n/a n/a
 Meyer (2016) (United Kingdom) [22] N = 356 female ED patients, AN = 25.9%, BN = 31% EDNOS = 38%, BED = 5%, N = 360 female controls CFA for clinical sample marginally fitted 5-factor structure, however model was found to differ significantly from observed data (Χ2(242) = 768.50, p < .001). Removing items 8 and 12 which didn’t meet expected factor loadings didn’t improve fit. Other fit indexes marginally met criteria: RMSEA = 0.080 (90% CI = 0.073–0.086), TLI = 0.90, IFI = 0.92, CFI = 0.92 Cut-off score of 15 distinguishes between ED patients with and without features of CE Cronbach’s α: .93 (overall scale), .96 (factor 1), .77 (factor 2), .87 (factor 3), .62 (factor 4), .82 (factor 5)
 Plateau (2013) (United Kingdom) [54] N = 689 competitive athletes (male = 258, female = 431). 702 before exclusion CFA showed poor fit of five factor structure. 9 items were removed. Exploratory PCA with remaining items yielded a 3 factor solution that explained 59.90% of variance: (1) avoidance of negative affect, (2) weight control exercise and (3) mood improvement. Factor 1 explained 35.15%, factor 2 14.67% and factor 3 10.10% n/a Cronbach’s α: .62 (global score), .87 (avoidance of negative affect), .82 (weight control exercise), .71 (mood improvement)
 Plateau (2017) (United Kingdom) [55] N = 349 female athletes, N = 32 reported current or previous ED, N = 12 athletes with EDs were recruited additionally (BN = 6, AN = 3, OSFED = 3) Factor analysis not conducted Global CET-A score of 10 established as cut-off for identifying female athletes with an ED n/a
 Sauchelli (2016) (Spain) [56] N = 157 ED participants, BN = 56, 35.7%, AN = 40, 25.5%, EDNOS = 61, 38.8%, N = 128 university student controls. Female = 228, 79.9%, male = 57, 20.3% 5 factor solution was supported (RMSEA = 0.087, CFI = 0.910, TLI = 0.900, SRMR = 0.080) n/a Cronbach’s α between .79 (ER) and .96 (ARD)
 Swenne (2016) (Sweden) [57] N = 254 adolescents < 18, full data available for 210, AN = 26, 12%, BN = 9, 4%, EDNOS = 175, 84%. Male = 12, female = 198 Exploratory PCA. Kaiser criterion suggested a four factor solution, scree plot3 or 4 factor solution. Final 4 factor solution explained 69.1% of variance: (1) WCE, (2) MI and (3) LEE, (4) combination of ARD and ER n/a Cronbach’s α: .94 (ARD), .85 (WCE), .90 (MI), .81 (LEE). Not calculated for overall scale
 Taranis (2011) (United Kingdom/ Australia) [23] 3 different studies, N = 367 female participants, mostly university students, N = 101 female undergraduate students, N = 97 female undergraduate students respectively Principal components analysis yielded initial 6 factor structure, changed to 5 factors after reducing to 24 items: (1) Avoidance and rule driven behavior (ARD) explained 30.39% variance, (2) Weight control exercise (WCE) 13.72% variance, (3) Mood improvement (MI), 7.71% variance, (4) Lack of exercise enjoyment (LEE), 6.74% variance, and (5) Exercise rigidity (ER), 5.32% variance n/a Cronbach’s α: .85 (overall scale), .88 (factor 1), .86 (factor 2), .75 (factor 3), .84 (factor 4), .73 (factor 5)
 Young (2017) (Australia) [58] N = 78 ED participants with AN, female = 74, male = 4 Factor analysis not conducted n/a Cronbach’s α: .92 (CET total), .95 (ARD), .82 (WCE), .86 (MI), .83 (LEE), .85 (ER)
Studies focusing on the Exercise and Eating Disorders (EED)
 Danielsen (2012) (Norway) [3] Mean EED sum score patients = 58.5 (±16.5), control group 33.4. (± 11.2). Mean difference = 25.1, p < .001. Subscale 1 patients = 21.6 (±6.4), control group 15.8 (±4.4), mean difference = 5.9. Subscale 2 patients = 24.0 (±9.0), control group 12.1 (±6.9), mean difference = 11.9. Subscale 3 patients = 12.9 (±4.7), controls 5.6 (±3.9), mean difference 7.4. All p < .001 EED total (Spearman’s ρ = .84) and all subscales (intentions to exercise ρ = .65, consequences of not exercising ρ = .76, bodily sensations ρ = .71) were significantly correlated with Body Attitudes Test (BAT total) in the whole sample. All p < .001. Correlations between EED subscales and BAT subscales ρ = .55 to .76. Correlations of total scores also significant for patients and controls separately. All p < .001 n/a
 Danielsen (2015) (Norway) [50] EED global score patients = 2.49 (±.96), controls = 1.40 (±.65), mean difference = 1.09. Factor 1 mean score patients = 2.64(±1.40), controls 1.30 (±.95), mean difference = 1.35. Factor 2 patients = 2.41 (± 1.36), controls = 1.59 (±.1.11, mean difference = .81. Factor 3 patients = 1.86 (±1.18), controls 1.01 (±.90), mean difference = .86. Factor 4 patients = 3.00 (±1.60), controls (1.40(±.65), mean difference = 1.09. All p < .001 n/a EED global score correlated significantly with EDE-Q global score for whole sample (r = .79) and for patients (r = .66) and controls (r = .73) separately. All p < .01. EED subscales CE (r = .70), PHE (r = .36), BS (r = .39) and WS (r = .65) correlated significantly with EDE-Q global score for the whole sample as well as patients and controls separately. All p < .01
 Danielsen (2018) (Norway) [51] EED global score patients = 2.00 (±.76), controls = 1.16 (±51), mean difference = .84. Factor 1 mean score patients = 1.89 (±1.45), controls = .65 (±.91), mean difference = 1.23. Factor 2 patients = 1.77 (±.94), controls = 1.13 (±1.08), mean difference = .64. Factor 3 patients = 2.27 (±1.40), controls = 1.56 (±1.30), mean difference = .71. Factor 4 patients = 2.16 (±1.61), controls 1.28 (±.94), mean difference = .84. All p < .001 n/a EED global score correlated significantly with EDE-Q global score for whole sample (r = .66) and for patients (r = .65) and controls (r = .35) separately. EED CE and WSE subscales correlated significantly with EDE-Q score for the whoe sample (r = .65 and r = .61) as well as patients (r = .67 and r = .65) and controls (r = .39 and r = .54). No correlation between EDE-Q scores and PHE subscale. Only EDE-Q scores for whole male sample correlated with BS subscale (r = .24). p < .01 for all correlations
Studies focusing on the Compulsive Exercise Test (CET)
 Goodwin et al. (2011) (United Kingdom) [53] n/a CET total (r = .54,), and all subscales (ARD (r = .65), WCE (r = .27), MI (r = .54), LEE (r = −.33) and ER (r = .56)) were significantly correlated with CES total. All p < .01 EDI subscales drive for thinness (r = .54), bulimia (r = .21) and body dissatisfaction (r = .24) were significantly correlated with CET total. All p < .01
 Formby (2014) (Australia) [52] n/a n/a Global EDE (r = .68, p < .001) and EDI subscales body dissatisfaction (r = .62, p < .001), drive for thinness (r = .70, p < .001), bulimia (r = .32, p = .01) and perfectionism (r = .42, p = .001) were significantly correlated with CET total
 Meyer (2016) (United Kingdom) [22] CET global score patients = 14.6 (±4.71), controls = 11.4 (±3.37). Factor 1 mean score patients = 2.75 (±1.71), controls = 1.74 (±1.28). Factor 2 patients = 3.47 (±1.34), controls = 2.59 (±1.17). Factor 3 patients = 3.37 (±1.28), controls = 3.26 (±1.12). Factor 4 patients = 2.21 (±1.19), controls 1.48 (±1.09). Factor 5 patients = 2.90 (±1.55), controls = 2.37 (±1.21). All differences significant at p < .001 level except for factor 3 n/a n/a
 Plateau (2013) (United Kingdom) [54] n/a n/a Strong correlations between all EDE-Q subscales and weight control exercise (r(685) ≥ .53), avoidance of negative affect (r(685) ≥ .31) and global score (r(685) ≥ .47). Smaller positive correlation with mood improvement (r(685) ≥ .16. All p < .01
 Plateau (2017) (United Kingdom) [55] Global CET-A score of 10 successfully discriminated female athletes with an eating disorder from those without. This cutoff score represented suitable levels of sensitivity (0.92) and specificity (0.73) n/a n/a
 Sauchelli (2016) (Spain) [56] Control group and all ED subgroups (AN, BN and EDNOS) differed significantly in subscales ARD, WCE and in CET total score. All p < .001. No differences between groups on other CET subscales n/a Partial correlations between CET scores and EDI. Correlations with moderate to good effect sizes are reported here. ARD, WCE and CET total correlated with drive for thinness (r = .408, r = .578 and r = .487), body dissatisfaction (r = .277, r = .399 and r = .301) and EDI total (r = .325, r = .404 and r = .316). ER correlated with drive for thinness (r = .300)
 Swenne (2016) (Sweden) [57] n/a n/a ARD (R2 = .27), WCE (R2 = .53), MI (R2 = .08) and LEE (R2 = .04) all correlated with EDE-Q global score. All p < .001 except LEE p < .01
 Taranis (2011) (United Kingdom/ Australia) [23] n/a Significant correlations (Spearman’s ρ) between CET total (ρ = .62) and subscales (ARD ρ = .70, WCE ρ = .41, MI ρ = .44, LEE ρ = −.42, ER ρ = .51) and Commitment to Exercise Scale (CES) total. All p < .001. Significant correlations between CET total (r = .58) and subscales (ARD r = .74, WCE r = .27, MI r = .36, LEE r = −.27 ER r = .52) and Obligatory Exercise Questionnaire (OEQ) total. All p < .01 and p < .001 Significant correlations between CET total, EDI total (ρ = .47) and EDI drive for thinness (ρ = .53) and body dissatisfaction (ρ = .40) subscales. Only the CET subscale WCE correlated with EDI total (ρ = .77). All p < .001. CET total correlated significantly with, EDE-Q total (r = .55) and EDE-Q subscales restraint (r = .49), eating concern (r = .50), shape concern (r = .53), weight concern (r = .48) All p < .001. CET subscales ARD (r = .33), WCE (r = .65) correlated significantly with EDE-Q total. All p < .001
 Young (2017) (Australia) [58] n/a CET total correlated significantly with CES mean (ρ = .78), EBQ total (ρ = .52) and REI total (ρ = .33). All p < .01 CET total (ρ = .64) and CET subscales ARD (ρ = .72), WCE (ρ = .39), MI (ρ = .27) and ER (ρ = .54) correlated significantly with EDEQ total. All p < .01. No significant correlation with LEE
  1. Note, AN Anorexia nervosa, ARD Avoidance and rule driven behavior, BAT Body Attitudes Test, BED Binge eating disorder, BN Bulimia nervosa, BS Awareness of bodily symptoms, CE Compulsive exercise, CES Commitment to Exercise Scale, CET Compulsive Exercise Test, CFA Confirmatory factor analysis, CFI Comparative Fit Index, EDNOS Eating disorder not otherwise specified, ED Eating disorder, EDE-Q Eating Disorder Examination Questionnaire, EDI Eating Disorder Inventory, EED Exercise and Eating Disorders, ER Exercise rigidity, IFI Incremental Fit Index, LEE Lack of exercise enjoyment, MI Mood improvement, OEQ Obligatory Exercise Questionnaire, OSFED Other unspecified feeding and eating disorder, PCA Principal components analysis, PHE Positive and healthy exercise, REI Reasons for Exercise Inventory, RMSEA Root Mean Square Error of Approximation, SRMR Standardized Root Mean Square Residual, TLI Tucker-Lewis Index, WCE Weight control exercise, WSE weight and shape exercise, WS Weight and shape related exercise
  2. *Factor 1, Factor 2 etc. refers to factors identified by confirmatory or exploratory factor analysis, or principal components analysis