Eating Behaviors 19 (2015) 20–23

Contents lists available at ScienceDirect

Eating Behaviors

Ability of EDI-2 and EDI-3 to correctly identify patients and subjects at

risk for eating disorders
Cristina Segura-García a,⁎, Matteo Aloi a, Marianna Rania a, Paola Ciambrone a, Antonella Palmieri a,
Valentina Pugliese a, Antonio José Ruiz Moruno b, Pasquale De Fazio a
a
Psychiatric Unit, Department of Health Sciences, University Magna Graecia, Catanzaro Italy
b
National Hospital for Paraplegics, Toledo Spain

a r t i c l e i n f o a b s t r a c t

Article history: Aims: The prevention and early recognition of eating disorders (EDs) are important topics in public health. This
Received 7 January 2015 study aims to compare the efficacy of the Eating Disorder Inventory 2 (EDI-2) with the new version, EDI-3 in
Received in revised form 18 May 2015 recognising patients and identifying subjects at risk for EDs.
Accepted 24 June 2015 Methods: The EDI-2 and EDI-3 were administered to 92 female patients with ED and 265 females from a popula-
Available online 2 July 2015
tion at risk for EDs. Experienced psychiatrists in this field held blind interviews with participants by means of the
SCID-I to determine the diagnosis.
Keywords:
EDI-2
Results: According to the cut-offs suggested by the authors, the EDI-3 correctly identified nearly all of the ED pa-
EDI-3 tients (99%), while the EDI-2 divulged less than half (48%). Both versions of the test showed comparable capabil-
Eating disorders ity to identify participants at risk for EDs but the EDI-3 seemed slightly more reliable than the EDI-2.
Clinical typification Conclusions: The EDI-2 remains a valid and very specific test. However, the new EDI-3 seems to be experimentally
Screening superior, because it typifies nearly all patients across the ED span, including those with Binge Eating Disorder and
Eating Disorder Not Otherwise Specified. In addition, it appears to be more reliable.
© 2015 Elsevier Ltd. All rights reserved.

1. Introduction
Eating disorders (EDs) are psychiatric conditions with high morbidity
and mortality rates, commonly diagnosed in teenagers and adult women.
The delay in their treatment can lead to chronic physical and psychologi-
cal consequences, so prevention and early detection are primary public
health issues.
Structured interviews [e.g. SCID-I (First, Spitzer, Gibbon, & Williams,
1997), EDE (Fairburn & Cooper, 1993)] are the best diagnostic tools, as
they allow for a very accurate assessment of the specific psychopathol-
ogy. They require much time and well-trained examiners so they are not
easily applicable to large-scale screenings.
Many validated self-administered psychometric instruments [e.g.
Eating Attitude Test (Garner & Garfinkel, 1979), Eating Disorder Exam-
ination Questionnaire (Fairburn & Beglin, 1994), Bulimic Investigatory
Test (Henderson & Freemann, 1987)] may resolve the above-
mentioned problems. Nevertheless, even these tools are not free of crit-
icisms, including denial or exaggeration of the disease by the person
who is being examined or differences in responses related to biases be-
tween subjects.
At least two main problems have not yet been solved regarding
the screening of EDs: first, there is no screening tool explicitly
aimed at the detection of subclinical ED. Second, most screening
tools are not appropriate for the recognition of at-risk behaviours
(Jacobi, Abascal, & Taylor, 2004). Even if there is no consensus
about the best protocol, the tests most frequently used are EAT,
BITE and Eating Disorder Inventory (Jauregui Lobera et al., 2009;
Manara, Manara, & Todisco, 2005).
The EDI is widely recognised as valid in the identification and char-
acterisation of patients with ED. The questionnaire is now in its third
edition (EDI-3) (Garner, 2004), due to the changes made in improving
the two previous versions.
This study has two objectives: a) to verify if the new EDI-3 is
able to reduce the rate of false negatives among ED patients; and b)
to ascertain if the EDI-3 is more efficient than the EDI-2 in correctly
identifying new cases of ED among subjects from the general
population.
2. Materials and methods
2.1. Participants

⁎ Corresponding author at: Department of Health Sciences, University “Magna Græcia” Two studies were carried out in order to verify the objectives.
of Catanzaro, Campus Universitario “Salvatore Venuta”, Viale Europa, 88100-Catanzaro,
Italy. Tel.: +39 096 1712408, +39 3316718232 (mobile); fax: +39 096 1712393. • Study 1 consisted in comparing the efficiency of EDI-2 and EDI-3 to
E-mail address: segura@unicz.it (C. Segura-García). correctly identify EDs among patients. A sample of 92 ED patients

http://dx.doi.org/10.1016/j.eatbeh.2015.06.010
1471-0153/© 2015 Elsevier Ltd. All rights reserved.
 C. Segura-García et al. / Eating Behaviors 19 (2015) 20–23
according to DSM-IV TR diagnostic criteria (APA, 2000) (25 AN, 21 BN,
35 BED, 11 ED-NOS) was enrolled.
• Study 2 screened EDs within a population at risk. As female gender,
adolescence/early adulthood and living in a western society are
thought to be general risk factors for the development of these disor-
ders (Fairburn & Harrison, 2003), a second sample of 265 Italian fe-
male students were evaluated.
All partakers, or their parents on behalf of those younger than 18,
signed a written informed consent according to the Ethical Committee's
guidelines.
2.2. Measures
2.2.1. EDI-2 (Garner, 1991)
The EDI-2 is a self-report questionnaire that assesses the psychopa-
thology of EDs by means of 91 items on a Likert-type six-point scale,
using a 3-point system where ‘sometimes’, ‘rarely’ and ‘never’ are
assigned zeros while ‘often’, ‘usually’, and ‘always’ are assigned a score
of 1, 2 and 3, respectively. It consists of 11 scales of which only Drive
for Thinness (DT) was considered in this study for diagnostic purposes
according to authors (Garner, Garfinkel, Rockert, & Olmsted, 1987).
2.2.2. EDI-3 (Garner, 2004)
The questionnaire consists of the same 91 items of EDI-2, although
item 71 is not scored. Three scales are specific for ED (i.e. DT, Bulimia
and Body Dissatisfaction), while the other nine are relevant but not ex-
plicit. It also provides six composite scores of which the Eating Disorder
Risk Composite (EDRC) that is specific for ED risk was considered in this
study, while the other five scales represent general psychological con-
structs. Finally, three response style indicators [i.e. Inconsistency (IN),
Infrequency (IF) and Negative Impression (NI)] were added to the
EDI-3 in order to alert the examiner about possible irregular, deviant
or infrequent responses and they were also contemplated in this study.
2.3. Procedures
As the Italian version of EDI-3 (Giannini, Pannocchia, Dalle Grave,
Muratori, & Vignone, 2008) uses the same items of EDI-2 Italian version
(Garner, 1995), participants were asked to complete the EDI-2. Each an-
swer sheet was then tabulated according to both scoring procedures.
Regarding EDI-2, although other cut-off points have been proposed
(Vetrone, Cuzzolaro, Antonozzi, & Garfinkel, 2006), a DT score N14,
the one suggested by the authors (Garner et al., 1987) and the most
widely used in the literature (Leon, Fulkerson, Perry, & Early-Zald,
1995), was applied as the threshold to identify cases in Study 1, as
well as participants at risk for ED in Study 2. For EDI-3 we referred to
the most restrictive thresholds suggested by Garner (2004): a combined
clinical score of EDRC (percentage) N24 was considered in Study 1 and
an EDRC N75 and N90 for adults (older than 18) and adolescents, re-
spectively, was considered in Study 2.
Experienced psychiatrists subsequently held blind interviews with
SCID-I in order to determine the diagnosis of each participant.
2.4. Statistical analyses
The comparison of frequencies was performed with a Chi-squared test
or Fisher's exact test as appropriate. The distribution of quantitative vari-
ables across different groups was compared by a t-test or ANOVA, as ap-
propriate. Where significant differences were found, Bonferroni post-
hoc pairwise comparisons were used. Level of significance was set at
p b .05.
• Study 1: First, the frequencies of patients positive for EDI2-DT and
EDI3-EDRC were calculated and compared according to diagnosis.
The mean values of the three response style indicators of EDI-3 were
calculated and compared in relation to the diagnosis.
21
• Study 2: Apparent and Real Prevalence, Sensitivity (Se), Specificity
(Sp), Positive Predictive Value (PPV), Negative Predictive Value
(NPV), Positive Likelihood Ratio (PLR), Negative Likelihood Ratio
(NLR), Correct Classification, Area under de ROC curve (AUC) and
Odds Ratio (OR) were calculated for the results of EDI-2 and EDI-3
using the clinical evaluations by psychiatrists and the results of the
SCID-I as gold standards. Cohen's kappa was calculated to measure
the agreement between the gold standard and EDI-2 DT and EDI-3
EDRC.
3. Results
3.1. Study 1: identifying cases in a clinical sample
Very significant differences regarding age and current, minimum,
maximum and ideal Body Mass Index (BMI) are evident, consistent
with the participants' diagnoses: patients with AN are the youngest
with lowest BMI; patients with BED are the oldest with the highest
BMI; patients with a diagnosis of either BN or ED-NOS present interme-
diate values.
Fig. 1. shows the percentages of patients positive to each test accord-
ing to their diagnosis. EDI-2 only identified 44/92 (48%) patients as pos-
itive for ED, while the EDI-3 correctly identified up to 91/92 (99%) of
them (X2 = 51.02; p b .001).
No significant differences according to diagnosis emerge from the
comparison of IN, IF, NI of the EDI-3.
3.2. Study 2: screening for at-risk individuals
The average age of the sample is 19.7 ± 3.4 years. The current, ideal,
lowest and highest BMI are 22.4 ± 3.4, 20.4 ± 1.8, 19.3 ± 2.5 and
22.9 ± 2.9 kg/m2 respectively.
Table 1 reassumes the characteristics of both tools. The AUCs were
similar for EDI-2 (97.0) and EDI-3 (94.2). The cut-offs used for each
test reach paragonable values in most descriptors (i.e. Sp, PPV, NPV, Cor-
rect Classification, and AUC). However, the EDI-2 shows a higher sensi-
tivity than does the EDI-3 and, conversely, the EDI-3 has a higher PLR,
NLR and OR than does the EDI-2.
Cohen's kappas for the EDI-2 DT and EDI-3 EDRC are .739 (p b .001)
and .695 (p b .001) respectively, indicating good agreement with the
gold standard.
Significant differences emerge from the comparison of the three re-
sponse style indicators between the participants who score above and
below the EDI-3 EDRC threshold: higher scores are evident in
Fig. 1. Percentage of patients correctly identified by EDI-2 and EDI-3.
22 C. Segura-García et al. / Eating Behaviors 19 (2015) 20–23
Table 1
Study 2: properties of the two diagnostic tools.
EDI-2 DT N 14 EDI-3 EDRC
Apparent prevalence (%) 17.76 14.95
Real prevalence (%) 22.43 22.43
Sensitivity (%) 70.83 62.5
Specificity (%) 97v59 98.8
Positive predictive value (%) 89.47 93.75
Negative predictive value (%) 92.04 90.11
Positive likelihood ratio 29.39 52.08
Negative likelihood ratio 0.274 0.379
Correct classification (%) 91.59 90.65
AUC 97.0 94.2
Odds ratio 98.36 136.67
Infrequency (1.9 ± 1.8 vs. 0.7 ± 1.1; t = 5.689; p b .001) and Negative
Impression (23.9 ± 10.0 vs. 7.1 ± 8.3; t = 11.419; p b .001) for partic-
ipants positive to screening, whereas the average score for Inconsisten-
cy is similar (10.7 ± 4.6 vs. 11.1 ± 6.0; t = 0.401; p = .688) for females
either positive or negative to screening with the EDI-3.
4. Discussion
The theoretical purpose of the EDI-3 was to offer a more extensive
description of the psychological constructs of ED and to improve the di-
agnostic capabilities of the previous version.
In Study 1 our results demonstrated that EDI-3 is able to classify ED
patients more precisely than will the EDI-2, as the EDI-3 correctly
characterised all patients with BN, BED and ED-NOS, and nearly all pa-
tients with AN diagnosis, while the EDI-2 showed a high percentage of
false negatives. Besides the fact that not all patients with ED have a
high DT (Ramacciotti et al., 2002; Strober, Freeman, & Morrell, 1999),
especially those with atypical EDs, the BED and ED-NOS seem to have
benefited most while using the last version of the EDI.
Several explanations can be given for this result. This capability is
probably related to the wider scoring system of the new version from
a 4 to a 5-point format (Lehman et al., 2013; van Strien & Ouwens,
2003). It can also clarify why patients with subthreshold symptoms
that could not satisfy all criteria for a full syndrome according to DSM-
IV TR as well as patients with ED-NOS are better identified as patholog-
ical by this new version. Perhaps the wider range of symptoms that the
new composite-score EDRC screens better explains this result.
Denial and social desirability can affect self-report measures. Present
data showed that the three response style indicators were low and sim-
ilar among the different diagnostic groups; as a result, these scores help
to confirm that tests were valid.
In Study 2 both tools were found to be much more specific than sen-
sitive in the screening, to have comparable PPV and NPV, and to correct-
ly classify comparable percentages of participants. It is noteworthy that
the prevalence of an event within a population at risk can influence the
predictive value of a diagnostic procedure; i.e. if the prevalence is low, as
in the case of EDs, a negative result allows for more safely excluding the
diagnosis because the NPV is higher. Contrarily, a positive result will not
allow for the diagnosis to be confirmed. As a result, other indices that do
not depend on the prevalence are needed. The PLR and NLR measure
how likely a particular outcome is (positive or negative), according to
the presence or absence of disease, and indicate how a result modifies
the pre-test probability into the post-test probability. In our case, the
EDI-3 EDRC showed higher PLR and NLR, as well as a higher general
OR, than did the EDI-2 DT.
The AUCs of the EDI-2 and EDI-3 were very high and similar, demon-
strating a very good overall accuracy. Good agreement of both tests with
the gold standard was further confirmed by the kappa values.
The comparison of the response style indicators showed significant
differences between students that were found to be positive or negative
for EDs according to the screening, with higher mean scores among
those who were positive, comparable to those of the clinical sample
from the Study 1. This result further confirms that EDI-3 is a psycho-
diagnostic assessment tool that can detect pathological subjects in the
screening (Clausen, Rosenvinge, Friborg, & Rokkedal, 2011).
To our knowledge, this is the first study that compares the diagnostic
capabilities of the EDI-2 and EDI-3, both in a clinical sample and in the
screening of ED within a population at risk.
The EDI-2 remains a valid and very specific test, however, the EDI-3
seems to be experimentally superior because: a) it is able to identify al-
most all patients within the ED categories, including patients with BED
and ED-NOS, diagnoses that frequently are not given adequate recogni-
tion on previous versions of the EDI and b) it has similar properties re-
garding Se, Sp, PPV and NPV in the screening of populations at risk for
EDs, but the OR, and PLR and NLR show that the EDI-3 EDRC is more re-
liable than is the EDI-2 DT.
In light of the above, we can conclude that the EDI-3 is more reliable
than the EDI-2 in the characterisation of patients, as it correctly typifies
nearly all patients across the ED spectrum. As for the screening of at-risk
individuals, both the EDI-3 EDRC and the EDI-2 show comparable effica-
cy, although the new version of EDI is more reliable.
Role of funding sources
The authors received no financial support for the research and/or authorship of this
article.
Contributors
CSG and PDF designed the study and wrote the protocol. MA, MR, PC, AP, and VP col-
lected the data. MA conducted literature searches and provided summaries of previous re-
search studies. CSG, MA, AJRM conducted the statistical analysis. CSG, MA, MR wrote the
first draft of the manuscript and all authors contributed to and have approved the final
manuscript.
Conflict of interest
The authors declared no potential conflicts of interest with respect to the research, au-
thorship, and/or publication of this article.
Acknowledgements
Authors are grateful to participants for the time they have given to this study.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.eatbeh.2015.06.010.
References
American Psychiatric Association (2000). Diagnostic and statistical manual of mental disor-
ders (4th ed.). Washington, DC: American Psychiatric Association (text revision).
Clausen, L., Rosenvinge, J. H., Friborg, O., & Rokkedal, K. (2011). Validating the Eating Dis-
order Inventory-3 (EDI-3): A comparison between 561 female eating disorder pa-
tients and 878 females from the general population. Journal of Psychopathological
Behavior Assessment, 33, 101–110.
Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders: Interview or self-
report questionnaire? International Journal of Eating Disorders, 16, 363–370.
Fairburn, C. G., & Cooper, Z. (1993). The eating disorder examination. In C. G. Fairburn, &
G. T. Wilson (Eds.), Binge eating: Nature, assessment, and treatment (pp. 317–360)
(12th ed ). New York: Guilford Press.
Fairburn, C. G., & Harrison, P. J. (2003). Eating disorders. Lancet, 361, 407–416.
First, M. B., Spitzer, R. L., Gibbon, M., & Williams, J. B. W. (1997). Structured clinical inter-
view for DSM-IV axis I disorders. Clinician version (SCID-CV). Washington, D.C.:
American Psychiatric Press, Inc.
Garner, D. M. (1991). EDI 2. Eating Disorder Inventory-2. Professional manual. Odessa, FL:
Odessa: Psychological Assessment Resources Inc.
Garner, D. M. (2004). The Eating Disorder Inventory-3: Professional manual. Odessa, FL:
Psychological Assessment Resources Inc.
Garner, D. M., & Garfinkel, P. E. (1979). The eating attitudes test: An index of the
symptoms of anorexia nervosa. Psychological Medicine, 9, 273–279.
Garner, D. M., Garfinkel, P. E., Rockert, W., & Olmsted, M. P. (1987). A prospective study of
eating disturbances in the ballet. Psychotherapy and Psychosomatics, 48, 170–175.
Garner, D. M. (1995). EDI-2. Eating Disorder Inventory-2. O.S. Firenze: Organizzazioni
Speciali.
Giannini, M., Pannocchia, P., Dalle Grave, R., Muratori, F., & Vignone, V. (2008). Eating Dis-
order Inventory-3. Firenze: Giunti O.S.
Henderson, M., & Freemann, C. P. (1987). A self-rating scale for bulimia. The ‘BITE’. British
Journal of Psychiatry, 150, 18–24.
 C. Segura-García et al. / Eating Behaviors 19 (2015) 20–23
Jacobi, C., Abascal, L., & Taylor, C. B. (2004). Screening for eating disorders and high-risk
behavior: Caution. International Journal of Eating Disorders, 34, 280–295.
Jauregui Lobera, I., Bermudo Parra, S., Martinez Lopez, M. D., Cuadri Galan, D., Garrido
Casals, O., & Alvarez Bautista, E. (2009). Screening for eating disorders in a psychiatric
outpatient population. The European Journal of General Practice, 15, 107–111.
Lehman, V., Ouwens, M. A., Braeken, J., Danner, U. N., van Elburg, A. A., Bekker, M. H. J.,
et al. (2013). Psychometric properties of the Dutch version of the Eating Disorder
Inventory-3. SAGE Open, 3.
Leon, G. R., Fulkerson, J. A., Perry, C. L., & Early-Zald, M. B. (1995). Prospective analysis of
personality and behavioral vulnerabilities and gender influences in the later develop-
ment of disordered eating. Journal of Abnormal Psychology, 104, 140–149.
Manara, F., Manara, A., & Todisco, P. (2005). Correlation between psychometric and bio-
logical parameters in anorexic and bulimic patients during and after an intensive
day hospital treatment. Eating and Weight Disorder, 10, 236–244.
23
Ramacciotti, C., Dell'Osso, L., Paoli, R. A., Ciapparelli, A., Coli, E., Kaplan, A. S., et al. (2002).
Characteristics of eating disorder patients without a drive for thinness. International
Journal of Eating Disorders, 32, 206–212.
Strober, M., Freeman, R., & Morrell, W. (1999). Atypical anorexia nervosa: Separation
from typical cases in course and outcome in a long-term prospective study.
International Journal of Eating Disorders, 25, 135–142.
van Strien, T., & Ouwens, M. (2003). Validation of the Dutch EDI-2 in one clinical and two
nonclinical populations. European Journal of Psychological Assessment, 19, 66–84.
Vetrone, G., Cuzzolaro, M., Antonozzi, I., & Garfinkel, P. E. (2006). Screening for eating dis-
orders: False negatives and eating disorders not otherwise specified. The European
Journal of Psychiatry, 20, 13–20.