Diagnosing underweight in adolescent girls: Should we rely on self-reported height

and weight?
Ricarda Ohlmer , Corinna Jacobi, Eike Fittig
Department of Clinical Psychology and Psychotherapy, Technische Universitt Dresden, Dresden, Germany
a b s t r a c t a r t i c l e i n f o
Article history:
Received 13 December 2010
Received in revised form 12 July 2011
Accepted 29 September 2011
Available online 6 October 2011
Keywords:
Adolescents
Height
Weight
Self-report
Anorexia
Eating disorders
This study examines the reliability of self-reported height, weight and weight change in underweight versus
normal weight adolescent females. Self-reported height and weight were obtained from 162 schoolgirls
without an eating disorder (1216 years), and compared to objective measurements afterwards. Weight
change was assessed 4 months later. The inuence of age and current BMI on the reliability of self-reports
was analyzed by linear regression analyses. With increasing age, height and BMI were reported more accu-
rately. With increasing BMI, the underestimation of weight increased. Underweight girls overestimated
their weight signicantly compared to normal weight girls. Only 41% of the girls with a weight loss (N1 kg)
in the past 4 months reported this accurately. Therefore, especially in younger girls with low body weight,
information on height and weight as well as weight changes should be obtained objectively to identify a de-
veloping or subthreshold anorexia nervosa.
2011 Elsevier Ltd. All rights reserved.
1. Introduction
Height and weight are important indicators of underweight, which
represents an early symptom of a developing anorexia nervosa. In fe-
males aged 12 to 16 years, the only predictors for the development of
anorexia nervosa in early adulthood were low Body Mass Index (BMI)
and perfectionism (Tyrka, Waldron, Graber, & Brooks Gunn, 2002).
Compared to other risk factors, height and weight seem to be easily ob-
tainable by questioning the participants. However, several studies sug-
gest that self-reported data may be less reliable than objectively
obtained data. Although high correlations for objective measurements
and self-reports of r.87 were found in 9th to 12th grade students
(Brener, McManus, Galuska, Lowry, &Wechsler, 2003), students overes-
timated their height by 1.8 cm and underestimated their weight by
1.6 kg on average. Despite an overall underestimation of weight, about
15% of the participants overestimated their weight by more than 2.3 kg
and 6.6% by more than 4.5 kg, respectively. This, in turn, could lead to
a severe overestimation of the BMI and therefore, preclude a possible
diagnosis of anorexia nervosa and delay necessary early interventions.
Even thoughthe risk of developing an eating disorder reaches its peak
in adolescence (Hoek & van Hoeken, 2003), research on adolescents'
estimation of height and weight is scarce. In a meta-analysis on the accu-
racy of self-reported height and weight (Engstrom, Paterson, Doherty,
Trabulsi, & Speer, 2003) only four out of 34 studies included adolescent
populations. In these studies, adolescents' misestimations of height
ranged from overestimations of 1.5 cm (SD=5.9) to underestima-
tions of 0.5 cm (SD=2.7), whereas in adult women they were
more pronounced with overestimations up to 2.0 cm (SD=2.6) in
a sample of college students and underestimations up to 1.7 cm
(SD=2.4) in a sample of Scottish women. Regarding weight, adoles-
cents underestimated their weight between 0.6 kg (SD=3.8) and
2.1 kg (SD=5.9) on average, whereas adult women also showed
more pronounced underestimations up to 3.5 kg (SD=2.9) in diet-
ing college students.
Engstrom et al. (2003) also tried to examine associations between
age and BMI and the accuracy of height and weight estimations. The
majority of studies reported an increasing overestimation of height
with increasing age in adult women. Results regarding associations
between age and self-reported weight were heterogeneous. There is
no evidence regarding potential effects of age on the accuracy of
height and weight estimations in adolescents. As to BMI, almost all
of the 20 studies examining this association showed a lower reliability
of weight estimations in heavier adult women. The latter effect has
also been found in adolescents (Abraham, Luscombe, Boyd, & Olesen,
2004; Strauss, 1999). Studies addressing the reliability of self-reported
weight change could not be found.
Although actual and self-reported data were highly correlated in
previous studies (Brener et al., 2003; Engstrom et al., 2003; Strauss,
1999), misestimations of height and weight might be substantial
enough to be of clinical relevance. We therefore analyzed the inuence
Eating Behaviors 13 (2012) 14
Corresponding author at: Technische Universitt Dresden, Professur fr Essstrungen
und assoziierte Strungen, Institut fr Klinische Psychologie und Psychotherapie, Chem-
nitzer Str. 46, 01187 Dresden, Germany. Tel.: +49 351 463 38570; fax: +49 351 463
37208.
E-mail addresses: ohlmer@psychologie.tu-dresden.de (R. Ohlmer),
cjacobi@psychologie.tu-dresden.de (C. Jacobi), ttig@psychologie.tu-dresden.de
(E. Fittig).
1471-0153/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.eatbeh.2011.09.005
Contents lists available at SciVerse ScienceDirect
Eating Behaviors
of height and weight misestimations on the resulting BMI percentiles of
adolescent girls to appraise the usability of self-reported data for the di-
agnosis of underweight.
1.1. Critical limits of height and weight misestimation
On the basis of the few previous ndings in adolescents, misestima-
tions in self-reported height and weight are to be expected. However,
nothing is known about howmisestimations can affect the classication
of weight status, especially diagnosing underweight in adolescents. To
demonstrate this potential inuence and to dene critical limits for
misclassications based on self-reported height and weight, we cal-
culated age adjusted BMI percentiles (Kromeyer-Hauschild et al.,
2001) based on ctive data for a 14-year old girl and a BMI percentile
of 10, which equals underweight (b15th percentile) (World Health
Organization, 1995).
Table 1 demonstrates the impact of misestimations of 14 cm in
height and 14 kg in weight on the resulting BMI percentiles assuming
different heights. Especially in short girls, an overestimation of weight
of only 2 kg can become precarious. The resulting BMI percentiles will
misclassify the participant as of normal weight instead of as under-
weight, whereas the impact of the misestimation of height on BMI
classication is rather minor.
Height and weight are considered to be correctly reported, if their
deviation from the actual value does not lead to a considerable devia-
tion of the resulting BMI percentile (N15th percentile in cases of actual
underweight). In the present study, deviations up to 2 cm, 1 kg
are considered to be correct.
Ingeneral, while previous studies reportedhighcorrelations between
actual and self-reported data, considerable weight overestimations have
also been observed. Especially in the phase of adolescence with its
challenge of mastering bodily changes and thus elevated risk of eat-
ing disorder onset, precise knowledge of body weight can enable
early interventions. However, research on the reliability of self-
reported height and weight of this age-group is scarce. In addition,
because misestimations of height and weight may have different
clinical implications in underweight girls compared to normal
weight girls (e.g., result in the false negative classication of under-
weight as core symptom of anorexia nervosa, as demonstrated
above), the knowledge on misestimations in underweight girls is of
crucial importance. Potential moderators of the accuracy of self-
reported height and weight could be age and actual BMI (Abraham
et al., 2004; Engstrom et al., 2003).
Accordingly, the purpose of this study is to examine the reliability of
self-reported height, weight, and weight change in adolescent girls of
different age and weight groups (underweight versus normal weight).
2. Methods
2.1. Study design and procedure
Data for this study were collected longitudinally at baseline (t
1
) and
4 months later (t
2
). The study was approved by the human subjects
committee of the Technische Universitt Dresden. Headmasters and
teachers of the respective schools were informed about the study.
Teachers forwarded consent forms for both students and parents to
the potential participants, who brought them to their assessments. Eat-
ing disorder risk status and self-reported height and weight were
assessed by self-report questionnaires; height and weight were mea-
sured objectively within 30 min thereafter. Assessments took place in
the schools' assembly halls, measurements were taken separately and
condentially for each girl.
2.2. Participants
The sample consisted of 162 girls, recruited fromten different high
schools in Dresden, Germany. Participants were included if they did
not fulll criteria for eating disorder high risk (dened by scoresN20
in the German version of the Eating Attitudes Test-26 (EAT-26)
(Meermann & Vandereycken, 1987) and a negative response to the
Eating Disorder Examination Questionnaire (EDE-Q) diagnostic
items (Garner, Olmsted, &Polivy, 1983)). At baseline, mean age of par-
ticipants was 13.8 years (SD=1.1), mean BMI was 18.7 (SD=2.3). At
the time of the second assessment 4 months later, 51 girls could no
longer be reached. As a result, data from 111 girls (mean age=13.8,
SD=1.2; mean BMI =18.7, SD=2.4) were available at t
2
. Girls who
could not be reached for follow-up did not differ signicantly in
weight status (underweight versus normal weight), EAT scores or
other relevant variables such as age, BMI percentiles, or EDE-Q diag-
nostic items.
2.3. Measures
As part of a larger set of questionnaires, participants' age, EAT-26
and EDE-Q scores were obtained by self-report. Self-reported height
and weight were used to calculate self-reported BMI. At t
2
, girls were
asked about any weight change in the past 4 months. Self-reported
data were veried after both points of assessment by objectively mea-
suring their height using an ultrasonic measuring device (0.5 cm pre-
cise, rounded to whole cm) and weight using a digital scale (0.05 kg
precision, rounded to 0.1 kg). For reasons of comparison, participants
were asked to remove shoes, headpieces and heavy jackets. To control
for the remaining clothes, 1 kg of the objectively obtained weight was
deducted from each participant's weight.
2.4. Data analysis
The reliability of the participants' self-reported height and weight
was examined by an intra class correlation of actual and self-reported
data. The inuence of age and BMI on the reliability of self-reported
height and weight was assessed via linear regression analysis using
age and BMI as predictors and the difference between actual and
self-reported data (height, weight, and BMI) as outcome. In addition
to the linear regression analysis, we compared categories of normal
weight an underweight girls by t-tests. Because only four of the in-
cluded participants had a weight of N85th percentile, overweight
was not included as a separate weight group in these analyses.
3. Results
3.1. Reliability of self-reported height and weight
Average measured height was 1.66 m compared to an average self-
reported height of 1.65 m, while average measured weight was
51.7 kg compared to an average self-reported weight of 51.2 kg. On
the basis of these data, meanactual BMI was 18.7 kg/m (corresponding
the 40th percentile), and mean self-reported BMI was 18.8 kg/m
(corresponding the 44th percentile).
Table 1
Effects of height and weight misestimations in short, middle-sized and tall girls with
objective underweight.
Actual
height
Actual
weight
Height or weight
misestimation
Impact on BMI
Actual BMI
percentile
Self-reported BMI
percentiles
1.50 m 37.5 1/2/3/4 cm 10 12/14/17/20
1.65 m 45.5 1/2/3/4 cm 10 12/14/17/19
1.80 m 54.0 1/2/3/4 cm 10 11/13/16/18
1.50 m 37.5 +1/+2/+3/+4 kg 10 14/19/26/32
1.65 m 45.5 +1/+2/+3/+4 kg 10 14/18/23/28
1.80 m 54.0 +1/+2/+3/+4 kg 10 13/16/20/24
2 R. Ohlmer et al. / Eating Behaviors 13 (2012) 14
On average, height was underestimated by about 1.0 cm
(SD=2.6 cm) and weight was underestimated by about 0.5 kg
(SD=2.2 kg). However, considering the distribution of misestimations,
a considerable percentage of girls (11.1%; N=18) showed deviances of
4 cm in their height estimation and 16 girls (9.9%) showed deviances
of 4 kg in their weight estimation.
The results of the intra class correlation show the expected highly
signicant concordance between actual and self-reported height
(r
ICC
=.93; pb.001). Actual and self-reported weight are also signi-
cantly correlated (r
ICC
=.96; pb.001).
3.2. Reliability of self-reported weight change
The intra class correlation between actual and self-reported weight
change over the past 4 months was signicant (p=.05) but small
(r
ICC
=.16). Changes inactual and self-reported weight are summarized
in Table 2. Compared to the baseline assessment, the majority of the
girls (59%; N=65) showed no considerable weight change (1 kg).
15.3% (N=17) of the girls lost onaverage 2 kg of weight; the remaining
26% (N=29) gained on average 2.1 kg of weight. Overall, weight
change was overestimated by about 0.2 kg (SD=2.0 kg). Of girls with
actual weight loss in the past 4 months, less than half reported to
have lost weight while an equal number of girls reported no weight
change andevery sixth girl reported weight gain. Ingirls withno weight
change, every fourth girl reported weight gain.
3.3. Relation between age and accuracy of self-reported height and weight
The linear regression analysis showed a signicant decrease in mean
height misestimationof 4 mmwith each additional year of age (=.23,
t=2.93, df=160, pb.005). Mean weight misestimation also decreased
non-signicantly with every year by 0.1 kg (=.07, t=.84, df=160,
p=.4, n. s.) and the resulting mean BMI misestimation decreased by
0.11 kg/m per year of added age (=.16, t=2.0, df=160, pb.05).
3.4. Relation between BMI and accuracy of self-reported height and weight
The linear regression analysis showed a signicant underestimation
of weight of about 0.3 kg (=.33, t=4.5, df =160, pb.001) with every
additional BMI point. Similarly, the underestimation of the BMI also in-
creased by about 0.14 kg/m (=.31, t =4.2, df =160, pb.001) with
each BMI point. Height misestimation was not related to BMI.
For the comparison of categories of normal weight and under-
weight girls, normal weight was dened as weight between the
15th and 85th percentile (N=121, mean BMI percentile=48,
SD=20.4); underweight was dened as weight below the 15th per-
centile (N=36, mean BMI percentile=7, SD=4.4) (World Health
Organization, 1995). Normal weight and underweight girls did not
differ signicantly in their eating disorder status as dened by their
EAT-26 scores (Meermann & Vandereycken, 1987). For the results
of the t-test see Fig. 1.
Both normal weight and underweight girls showed an underestima-
tion in height (t=0.83, p=0.4, d=.16). Normal weight and under-
weight girls differed signicantly in weight and resulting BMI
estimation. Effect sizes were in the medium range. Normal weight girls
underestimated their weight on average by about 0.7 kg, whereas
underweight girls overestimated their weight by about 0.3 kg (t=2.5,
pb.05, d=.49) on average. The resulting self-reported BMI was exact
in normal weight girls, whereas underweight girls overestimated their
BMI by about 0.44 BMI points (t=2.4, pb.05, d=.45) on average.
Table 3 shows the distribution of misestimations in both groups in
detail.
Fifteen of the 43 girls with a BMI overestimationN0.5 kg/m had
an actual BMI below the 15th percentile. Of those, 8 girls would
have been false negatively classied as normal weight on the basis
of their self-report. Underweight girls overestimated their BMI more
frequently compared to normal weight girls; however, this result
just missed the signicance level (
2
=5.67; p=.059).
4. Discussion
The objective of this explorative study was to examine the reliabil-
ity of self-reported height, weight and their effect on BMI and/or BMI
percentiles in adolescent girls. Overall, we found that the correlation
between actual and self-reported data at baseline was comparably
high compared to results of previous studies (Brener et al., 2003).
However, despite the high concordance of actual and self-reported
data, height and weight were correctly estimated by only seventy
ve and fty percent of the girls, respectively. Compared to the previ-
ously reported results of an overestimation of height and an underes-
timation of weight (Brener et al., 2003; Engstromet al., 2003) only the
underestimation of weight could be conrmed by the present study,
whereas height was underestimated on average. This may be caused
by the fact that girls of the included age group grow faster than girls
of older age groups without being aware of this. In accordance with
this assumption, Stolzenberg, Kahl, and Bergmann (2007) found a de-
cline in speed of growth in a large group of girls between 12 and
16 years who were part of their children and adolescents health sur-
vey. Whereas 12 year old girls grew on average 4.1 cm per year, girls
aged 13 grew only 2.3 cm on average, and 14-year-olds only 1.3 cm.
Beyond age 15, the annual growth of girls was less than 1.0 cm.
During adolescence, while girls are still in the period of growth, the
absence of weight gain could also be an important diagnostic indicator
of anorexia nervosa. Consequently, we also examined the accuracy of
Table 2
Actual and self-reported weight change after 4 months.
Self-reported weight change Total
Weight lossN1 kg No alteration1 kg Weight gainN1 kg
Actual weight change Weight lossN1 kg 7 (41.2%) 7 (41.2%) 3 (17.6%) 17 (100%)
No alteration1 kg 3 (4.6%) 45 (69.2%) 17 (26.2%) 65 (100%)
Weight gainN1 kg 2 (6.9%) 19 (65.5%) 8 (27.6%) 29 (100%)
Total 12 (10.8%) 71 (64.0%) 28 (25.2%) 111 (100%)
Fig. 1. Mean differences of actual and self-reported height, weight and BMI in normal
weight and underweight girls.
3 R. Ohlmer et al. / Eating Behaviors 13 (2012) 14
self-reported weight change. Weight change was accurately reported
by half of the girls. Of the other half, especially those girls reporting
an actual weight loss as weight gain, warrant specic attention, be-
cause their misestimation could be an indicator of a disturbed self-
perception or body image distortion.
Previous research has found inconsistent results with regards to
the relation between age and the accuracy of self-reported height
and weight (Engstrom et al., 2003). The present results indicate that
height and weight are being reported more precisely in older girls.
Considering the decline in speed of growth, a rare self-measurement
of height still leads to comparatively accurate estimations in older
girls. In addition, because older girls are on average taller and heavier,
the same misestimation will impact their self-reported BMI to a lesser
degree compared to younger girls. Consequently, especially in youn-
ger and more lightweight girls practitioners should rely on objectively
measured data of height and weight rather than on self-reported data.
In addition to age, the present study also addressed the inuence of
actual BMI on the reliability of self-reported height and weight. Results
of the linear regression analyses are in line with previous research
(Abraham et al., 2004; Engstrom et al., 2003; McCabe, McFarlane,
Polivy, & Olmsted, 2001): the higher the actual BMI, the larger weight
and BMI underestimations. However, when weight status was
addressed by comparing underweight girls (b15th BMI percentile)
with those in the normal weight range, underweight girls in contrast
underestimated their height and overestimated their weight, resulting
in a BMI overestimation. This is in accordance with the study by McCabe
and colleagues (McCabe et al., 2001), where only individuals with an-
orexia nervosa overestimated their weight compared to normal weight
individuals with bulimia nervosa and college students. In the present
sample, for a quarter of the underweight girls this would have resulted
in a false negative classication of their weight as in the normal range. Al-
thoughnone of the participants inthe present study scoredabove the cut-
off of 20 for anorexia nervosa in the EAT-26 (Meermann &Vandereycken,
1987) a possible reason for their weight overestimation could be the de-
nial of their own critically low body weight.
The present study demonstrates that despite high correlations be-
tweenself-reportedandobjectively measuredheight andweight consid-
erable misestimations are frequently observed. For a correct diagnosis of
anorexia nervosa, objective measurements of height and weight are,
therefore, of crucial importance and would facilitate early interventions
before the development of the full-syndrome disorder. The results are
of practical relevance for future assessments in the context of screening
for at-risk participants and preventive interventions.
While previous studies have been criticized because of time lags
between the assessment of objective and self-reported data, our
study assessed both variables subsequently within 30 min, thus min-
imizing risks of objective changes in weight due to other inuences.
Furthermore, none of the previous studies addressed the reliabili-
ty of self-reported weight changes in adolescent girls longitudinally.
However, some limitations also need to be mentioned. Because
only high-school students were included, the generalizability of the
results to other age groups is unclear. The reliability of self-reported
weight change may also depend on the time interval chosen. In the
present study, a longer follow-up period for the second assessment
was not possible for practical reasons. Also, we cannot rule out that
smaller weight changes, chosen here to yield highest sensitivity for
the classication of weight status, might reect natural weight uctu-
ations. Finally, to further explore the present ndings in underweight
girls, a larger sample of participants with a BMI b15th percentile
would be necessary.
Role of funding sources
Funding for this study was provided by a grant from the Swiss Anorexia Nervosa
Foundation. The Swiss Anorexia Nervosa Foundation had no role in the study design, col-
lection, analysis or interpretation of the data, writing the manuscript, or the decision to
submit the paper for publication.
Contributors
Authors Ohlmer and Fittig designed the study under supervisionof author Jacobi. Ohl-
mer collected data, conducted the statistical analysis and wrote the rst draft of the man-
uscript. All authors contributed to and have approved the nal manuscript.
Conict of interest
The authors declare that they have no conicts of interest.
Acknowledgments
The authors wish to thank Dipl.-Psych. Heidi Pftze for her support in data
ascertainment.
References
Abraham, S., Luscombe, G., Boyd, C., & Olesen, I. (2004). Predictors of the accuracy of
self-reported height and weight in adolescent female school students. International
Journal of Eating Disorders, 36(1), 7682.
Brener, N. D., McManus, T., Galuska, D. A., Lowry, R., & Wechsler, H. (2003). Reliability
and validity of self-reported height and weight among high school students. Journal
of Adolescent Health, 32(4), 281287.
Engstrom, J. L., Paterson, S. A., Doherty, A., Trabulsi, M., & Speer, K. L. (2003). Accuracy
of self-reported height and weight in women: An integrative review of the litera-
ture. Journal of Midwifery & Womens Health, 48(5), 338345.
Garner, D. M., Olmsted, M. P., & Polivy, J. (1983). Development and validation of a
multidimensional eating disorder inventory for anorexia nervosa and bulimia.
International Journal of Eating Disorders, 2, 1534.
Hoek, H. W., & van Hoeken, D. (2003). Review of the prevalence and incidence of eating
disorders. International Journal of Eating Disorders, 34(4), 383396.
Kromeyer-Hauschild, K., Wabitsch, M., Kunze, D., Geller, D., Geiss, H. C., Hesse, V., et al.
(2001). Perzentile fr den Body-Mass-Index fr das Kindes- und Jugendalter unter
Heranziehung verschiedener deutscher Stichproben. Monatsschrift Kinderheilkunde,
149(8), 807818.
McCabe, R. E., McFarlane, T., Polivy, J., & Olmsted, M. P. (2001). Eating disorders, dieting,
and the accuracy of self-reported weight. International Journal of Eating Disorders, 29
(1), 5964.
Meermann, R., & Vandereycken, W. (1987). Therapie der Magersucht und der Bulimia
Nervosa. Berlin: De Gruyter.
Stolzenberg, H., Kahl, H., &Bergmann, K. E. (2007). Body measurements of children and ad-
olescents in Germany. Results of the GermanHealth Interviewand Examination Survey
for Children and Adolescents (KiGGS). Bundesgesundheitsblatt, Gesundheitsforschung,
Gesundheitsschutz, 50(56), 659669.
Strauss, R. S. (1999). Comparison of measured and self-reported weight and height in a
cross-sectional sample of young adolescents. International Journal of Obesity, 23(8),
904908.
Tyrka, A. R., Waldron, I., Graber, J. A., & Brooks Gunn, J. (2002). Prospective predictors of
the onset of anorexic and bulimic syndromes. International Journal of Eating Disorders,
32(3), 282290.
World Health Organization (1995). Physical status: The use and interpretation of an-
thropometry. WHO Technical Report Series, 0(854), I-452.
Table 3
Absolute misestimations of height, weight and BMI depending on weight-status.
Normal weight Underweight
Height Underestimation3 cm 19 (15.7%) 12 (33.3%)
Correct estimation2 cm 101 (83.5%) 22 (61.1%)
Overestimation3 cm 1 (0.8%) 2 (5.6%)
Weight Underestimation1.1 kg 46 (38.0%) 9 (25.0%)
Correct estimation1 kg 63 (52.1%) 16 (44.4%)
Overestimation1.1 kg 12 (9.9%) 11 (30.6%)
BMI UnderestimationN0.5 kg/m 28 (23.1%) 4 (11.1%)
Correct estimation0.5 kg/m 65 (53.7%) 17 (47.2%)
OverestimationN0.5 kg/m 28 (23.1%) 15 (41.7%)
N
total
121 (100%) 36 (100%)
4 R. Ohlmer et al. / Eating Behaviors 13 (2012) 14