Eight questionnaires tested for binge-eating disorder and one for atypical anorexia
- Systematic review from the University of Sydney, with co-authors at Western Sydney University and the University Medical Centre Rostock, updating the same group's 2022 review of the same question. Five databases were searched from 2020 to November 2025 and the protocol was registered as CRD420251186115. A study was eligible if it reported the diagnostic accuracy of a self-report questionnaire against a clinical interview, for eating disorders or for disordered eating behaviours, in adolescents or adults living with higher weight. Synthesis was narrative, with the new studies set alongside those found the first time. Thirty-two studies were included, five of them new, covering 13 questionnaires in adults and 5 in adolescents. Published online on 20 May 2026 in the International Journal of Eating Disorders.
- Counted by what each questionnaire was asked to detect, adults first and adolescents second: any eating disorder 5 and 0; binge-eating disorder 8 and 2; disordered eating behaviours such as binge eating and purging 4 and 1; loss-of-control eating 1 and 2. Bulimia nervosa, atypical anorexia nervosa, purging disorder and night eating syndrome have one adult questionnaire each and none in adolescents.
- The three questionnaires used most often, with the ranges as printed. Eating Disorder Examination Questionnaire, 7 studies, sensitivity 0.16 to 0.88, specificity 0.54 to 1.0. Binge Eating Scale, 6 studies, 0.37 to 0.98 and 0.48 to 0.96. Questionnaire on Eating and Weight Patterns, 6 studies, 0.07 to 1.0 and 0.0 to 1.0. The SCOFF, five yes-or-no questions that take about half a minute, ran 0.67 to 1.0 on sensitivity and 0.59 to 0.83 on specificity across three studies of 63 to 178 participants; those SCOFF figures come from the group's 2022 review, not from the 2026 abstract.
- The authors state that progress in evaluating the diagnostic accuracy of eating disorder questionnaires in people with higher weight has been limited, that evidence in adolescents is lacking, as are questionnaires sensitive enough to identify eating disorders other than binge-eating disorder, and that assessing accuracy in this population remains a research priority.
An adult whose BMI sits in the obesity range comes in for a weight appointment, and somewhere in the intake pack there is a short eating questionnaire. It gets scored. Above the cut-point the patient is referred on; below it the subject closes and the visit moves to diet, medication or surgery. The question the Sydney group set is the one that comes before all of that. For which diagnoses, in which age group, has that form ever been checked against a clinical interview in patients who look like this one?
Eight diagnoses, and the studies sit on one
Five databases from 2020 to November 2025, a registered protocol, and the same team's 2022 review underneath it. Thirty-two studies qualified, five of them new, covering 13 questionnaires in adults and 5 in adolescents. Sorted by target condition, the counts are not spread evenly across the diagnoses a clinician might be looking for.
Binge-eating disorder has 8 adult questionnaires behind it and 2 in adolescents. Any eating disorder taken as a whole has 5 in adults and none in adolescents. Disordered eating behaviours such as binge eating and purging have 4 and 1, loss-of-control eating 1 and 2. Then the list runs out. Bulimia nervosa, atypical anorexia nervosa, purging disorder and night eating syndrome rest on one adult study apiece, and on nothing at all in adolescents.
The instruments used most often carry wide ranges. The EDE-Q appears in 7 studies at sensitivity 0.16 to 0.88 and specificity 0.54 to 1.0. The Binge Eating Scale appears in 6, at 0.37 to 0.98 and 0.48 to 0.96. The Questionnaire on Eating and Weight Patterns appears in 6, at 0.07 to 1.0 and 0.0 to 1.0. The SCOFF, three studies of 63 to 178 participants, runs 0.67 to 1.0 and 0.59 to 0.83. These are ranges across separate samples, cut-points and interviews, not pooled estimates; the review synthesises narratively and does not combine them.
Reading a negative score in the room
Sensitivity is the share of the people a clinical interview diagnoses whom the questionnaire flags. At 0.16, the bottom of the EDE-Q range, most of them are not flagged. At 0.88, most are. Both figures describe the same instrument in the same broad population, in different samples, and the clinician holding the completed form has no way to tell which sample the patient in front of them resembles.
That is the state of the diagnosis with the most work behind it. For the patient who restricts while weighing more than average, for the one who purges without bingeing, for the one who is eating at three in the morning, there is a single questionnaire each to consult. In adolescents living with higher weight the tested repertoire is binge eating and loss of control, and nothing else: no questionnaire in that age group has been measured against an interview for any eating disorder taken as a whole, or for bulimia nervosa, atypical anorexia nervosa, purging disorder or night eating syndrome.
So the reference standard in all 32 studies is worth naming, because it is also the fallback. Every accuracy figure in this review exists only because somebody sat down and conducted a clinical interview. For binge-eating disorder in an adult, a questionnaire score has been tested as a stand-in for that interview and its performance is on record, wide range and all. For the other diagnoses, and for adolescents, the interview is not the comparator. It is the whole of the tested evidence.
Eight questionnaires have been checked against a clinical interview for binge-eating disorder in adults living with higher weight, while atypical anorexia nervosa, purging disorder and night eating syndrome have one questionnaire each, and in adolescents the tested ground stops at binge eating and loss of control.
A count of studies is not a measure of performance. A questionnaire with one study behind it in this population, or with none, has not been shown to fail there; it has not been tested there, and the two are different findings. This review counts and describes, and its ranges are not pooled estimates: the numbers come from separate samples with different BMI thresholds, different cut-points and different diagnostic interviews, so the low end of a range belongs to one study rather than to the instrument. The abstract of the 2026 update carries the counts and the EDE-Q, Binge Eating Scale and Questionnaire on Eating and Weight Patterns ranges quoted here; the SCOFF figures and the participant numbers behind them are from the open full text of the same group's 2022 review, whose studies are carried into this one, and the 2026 full text sits behind the publisher's paywall. The search closed in November 2025, so anything published since is outside it. Accuracy is reported for adolescents and adults living with higher weight and says nothing about how these questionnaires behave in patients at lower weight.