Acta Pñdiatr 92: 1453±1462.

2003

Genital anatomy in non-abused preschool girls

AK Myhre1,2, K Berntzen2 and D Bratlid1,2
Department of Laboratory Medicine, Children’s and Women’s Health1, Norwegian University of Science and Technology; Department of
Paediatrics2, University Hospital, Trondheim, Norway

Myhre AK, Berntzen K, Bratlid D. Genital anatomy in non-abused preschool girls. Acta Pædiatr
2003; 92: 1453–1462. Stockholm. ISSN 0803-5253
Aim: To describe the normal variations in genital anatomy in preschool girls selected for non-
abuse. Methods: A total of 2731 girls aged 5 or 6 y were invited to take part in the study; 195 girls
were recruited. Inclusion was based on self-selection, whereby parents who did not suspect any
occurrence of sexual abuse of their children gave informed consent to participate. Several steps
were taken to exclude abused girls and girls with previous accidental genital injuries. The genital
examination, using a colposcope and a camera, was performed in supine position using a
separation and traction technique, and in the prone knee–chest position. Results: A number of
genital anatomical features and hymenal measurements were described and found consistent with
previous studies. An important finding was outward folding of the posterior hymenal rim in many
girls, a feature that could be difficult to distinguish from attenuation of the posterior hymen. A
gaping hymenal orifice, previously suggested to be a supportive sign of sexual abuse, was fairly
frequently found and significantly associated with a large horizontal hymenal diameter.
Conclusion: To distinguish between girls with outward folding of the posterior hymen and those
with attenuated hymens, we recommend the use of the saline irrigation method. Even though
normative hymenal measurement data now exist from a reasonable number of girls, these
measurements should be used with caution in sexual abuse evaluations.
Key words: child sexual abuse, normative genital anatomy, prepubertal hymen
AK Myhre, Department of Paediatrics, St. Olav’s Hospital HF, NO-7006 Trondheim, Norway (Tel.
‡47 73 868 149, fax. ‡47 73 867 322, e-mail. arne.k.myhre@medisin.ntnu.no)

During the past twenty years, sexual abuse of children normative anatomical and microbiological data from
has been identified as a world-wide problem, with the anogenital area in preschool children (18, 19).
reported prevalences ranging from 7% to 36% for girls
and from 3% to 29% for boys (1). The forensic genital
examination is considered an important part of the
evaluation of these children (2), and during the last Patients and methods
twelve years, studies of non-abused (3–8) and probably
abused children (9–14) have provided important knowl- Study sample
edge on this topic. Following approval by the Regional Research Ethics
Despite the increasing body of knowledge about the Committee, a letter of invitation was sent to all girls
genital anatomy of prepubertal girls, controversies still born between 1 January 1992 and 23 September 1994
exist. In a newly published classification scale (15), all living in the city of Trondheim. Details about recruit-
findings related to hymenal measurements were re- ment have been presented previously (18). To avoid
moved because of uncertainty about their interpretation. including previously abused girls, self-selection based
Later, a case-control study (13) provided important data on lack of parental suspicion of sexual abuse and
about this topic. Finding a gaping hymenal orifice, not informed consent to participate was used as the
previously described in normative studies, has been inclusion criterion. Furthermore, a girl was excluded
proposed as a supportive sign of sexual abuse (16) and if one of the following criteria were present: 1) Previous
justifying validation. Still, the interpretation of a large referral to the hospital because of an allegation or
hymenal opening and a narrow posterior rim is open to suspicion of sexual abuse. 2) Anogenital findings giving
discussion (15, 17). Thus, the aim of this report is to clear evidence of blunt force or penetrating trauma,
expand the knowledge on normal variations in female defined as class 4 findings in the revised classifica-
genital anatomy in a closely age-defined cohort of girls. tion system of anogenital findings (15). 3) If a case
The report is part of a larger study, designed to provide of suspected sexual abuse was brought to our attention.

 2003 Taylor & Francis. ISSN 0803-5253 DOI 10.1080/08035250310007574

1454 AK Myhre et al. ACTA PÆDIATR 92 (2003)

Table 1. Definitions of genital terminology.

Vaginal discharge Presence of whitish, viscid fluid in the vagina

Labial adhesion Significant fusion of the adjacent edges of the labia minora
Increased friability A superficial breakdown of the skin in the posterior commissure when gentle separation or traction was applied,
causing slight bleeding
Linea vestibularis A vertical, pale/avascular line across the fossa navicularis
Fossa navicularis pallor An area with distinct pallor in the fossa navicularis mucus membrane
Hymenal configuration Defined as crescentic, annular, transitional, fimbriated, tulip, imperforate, microperforate, cribriforme and keyhole
Gaping hymenal orifice Presence of a visible hymenal orifice with the child in supine position and thighs abducted, but without labial
separation/traction
Prominent hymenal vessels Dilatation of superficial blood vessels in the hymen
External hymenal ridge A midline longitudinal ridge of tissue on the external surface of the hymen
Perihymenal band Small band of tissue lateral to the hymen that forms a connection between the perihymenal structures and the
wall of the vestibule
Hymenal bump A solid elevation of hymenal tissue, which is wider or as wide as it is long, located on the edge of the hymenal
membrane
Hymenal tag An elongated projection of tissue, rising from any location on the hymenal rim
Folded hymenal edge Part of the hymenal edge folding outwards or inwards
Hymenal notch A U- or V-shaped concavity that dips beneath the baseline but not through the entire width of the membrane.
Areas appearing as concavities because of proximity to hymenal projections are not included. A superficial
notch is defined as smaller than or equal to 50% of the width of the membrane, a deep notch more than 50%
Hymenal transection A notch that extends through the entire width of the hymenal membrane

4) History of accidental genital injury presenting with relaxation was recorded, and it was intended that at least
bleeding. one photograph would be taken at the point of maxi-
The letter of invitation was sent to 2731 girls, all aged mum relaxation in each position/method. During the
5 or 6 y, and informed consent to participate was study a Leisegang colposcope (3B3) with fixed magni-
received from the parents of 248 girls (9.1%). Of these, fication (15) and a 35-mm camera (Yashica FX-3)
10 girls did not show up for the examination and one were used.
examination was not completed, owing to technical After the examination, the children were interviewed
problems. Twenty-three girls were not examined by one of the project nurses about how they had
because they refused, and in one case the mother experienced the session. The parents remained with the
changed her mind during the interview. A total of five examiner and were then asked if they had any suspicion
girls were excluded because of an expressed concern or were concerned about their child having been
that they could have been sexually abused; three parents sexually abused. The structured parental interview
reported such a suspicion during the interview and two was completed with questions about behavioural and
girls were later referred to our sexual abuse team. Eight emotional symptoms.
girls were excluded owing to accidental genital injuries
with bleeding. None were excluded because of previous Registration and classification of anatomical variables
referrals to the child abuse team or because of
anogenital finding class 4. Finally, the first five girls To provide a consistent data collection, the same
examined were used as a pilot study, leaving a total of examiner (AKM) carried out all but one examination,
195 girls in the final study sample. and in most cases one of two project nurses assisted.
The same two physicians (the examiner being one of
them) later reviewed all photographs and classified the
Study procedure findings. Both physicians have extensive experience in
Data collection took place from March 1997 to June child abuse evaluations and have cooperated in weekly
2000. During the examination, the examiner conducted review meetings for the past 10 y. The photographs
a detailed and structured interview about demographics, were reviewed both “on screen” and with the use of a
general health and specific information about anogenital stereoscope.
and urinary health problems. Then the girls were To ensure precise and consistent classification,
examined from head-to-toe. Genital anatomy was descriptive terminology from the APSAC guideline
examined with the child first in supine position and (20) and an anatomical atlas (17) were used to define the
then in the prone knee–chest position (17). In supine variables. A summary of definitions of terminology is
position, one photo was taken with flexion and abduc- provided in Table 1. Regarding the finding of labial
tion of the hips only, then two photos with labial adhesions, only significant adhesions were recorded.
separation and labial traction methods, respectively. Posterior adhesion was defined as an extensive fusion,
Finally, two photos were taken with the girl in knee– covering all or most of the posterior hymen in all
chest position and slight separation of the labia and examination methods. Anterior adhesion was defined as
buttocks. During the examination, the degree of child a fusion covering the urethral opening. If the hymenal
ACTA PÆDIATR 92 (2003) Genital anatomy in preschool girls 1455

Fig. 2. Girl examined by supine separation technique. The horizontal

Fig. 1. Gaping hymenal orifice. The examiner can see through the line demonstrates our measure of the horizontal opening diameter,
hymenal orifice into the vagina without separation or traction of the and the arrow from the hymenal edge to the light reflex shows how
labia, but with flexion and abduction of the thighs only. Consent to we estimated the 6 o’clock hymenal rim. Owing to crescentic
publish the photograph was given by the parents. A = anterior, hymenal configuration, no vertical diameter was measured. Consent
P = posterior, L = left, R = right. to publish the photograph was given by the parents.

opening presented with a different configuration using more than one light reflex was seen, the one consistent
different methods, the most prominent one was chosen. with other reflexes was chosen.
Hymenal rim features behind the 3 to 9 o’clock line To assess the intra-observer reliability regarding
were recorded separately in supine and prone position in classification of data, randomly selected photos were
girls where a good view of the anatomy was obtained on re-reviewed. For categorical data, photos from 20 girls
the photos. The finding of a gaping hymenal orifice (Fig. (10%) were selected. The kappa statistic ( ) was used
1) was defined according to Hobbs et al.’s study (16). to quantify the agreement, and -values ranging from
Measurements of the hymenal opening and the width 0.73–1.0 were obtained. The poorest agreement was
of the posterior rim in 6 o’clock positions were obtained observed regarding classification of “prominent hyme-
directly from the photos (Fig. 2). By comparing these nal vessels”. For continuous data, photos from 30 (15%)
relative measurements with a separate photograph of a girls were re-reviewed, and Bland Altman plots (21)
metric scale taken at the same focal length, actual were calculated to assess agreement. In accordance with
values were obtained (in mm). Measurements were a study elsewhere (13), acceptable agreement was
obtained separately in supine separation, supine traction defined as a maximum difference of 0.5 mm between
and knee–chest position, and the photo displaying the the first and second measure.
best symmetry and largest horizontal diameter was
chosen. Photos not displaying the whole hymenal
opening were not used for diameter measurements, Statistics
nor were fimbriated or tulip configurations or openings Categorical variables are presented in frequency tables
with folded hymens. Crescentic configurated hymens as numbers and percentages. To compare anatomical
were not used in measuring the vertical diameter, owing findings in supine and prone positions in the same child,
to lack of well-defined borders. The 6 o’clock hymenal a paired sample test (McNemar’s test) was used. When
rim was measured if a light reflex was observed and the comparing continuous data, parametric or non-para-
posterior fossa was well visualized (Fig. 2) (3, 13). If metric tests were used according to the observed
1456 AK Myhre et al. ACTA PÆDIATR 92 (2003)

distribution of data. Analysis was performed with SPSS Table 3. Hymenal configurations in the study group (n = 194a).
version 8.0 statistical software. P-values of less than 0.1
are reported in the tables and a p-value < 0.05 was used No. %
as the level of significance; p-values at or above 0.1 are Crescentic 152 78.4
reported with the abbreviation n.s. No correction for Annular 13 6.7
multiple comparisons was made. Transitional 12 6.2
Fimbriated 1 0.5
Tulip 1 0.5
Keyhole 1 0.5
UTDb 14 7.2
Results
a
Data provided for 194 of the 195 examinations performed. In one girl,
Of the 195 girls examined, 176 (90.3%) had a complete vestibulum could not be visualized, owing to extensive labial adhesions.
genital examination in both the supine and the knee– b
UTD = unable to detect, because the hymenal opening could not be
chest position. One child was examined in knee–chest identified with any visualization technique.
position only and 18 girls in supine position only. Of
these girls, one was examined using a saline irrigation
technique (17). All girls were at Tanner I stage of line and comparison between girls examined in both the
pubertal development. The mean age was 5.71 y (range: supine and the knee–chest position are presented in
5.15–6.75) and 183 (93.8%) were of Caucasian ethnic Table 4. Of 41 girls with hymenal bumps in supine
origin. The mean number of children in the families was position, 7 girls had more than one bump and associated
2.30 (range 1–5). The parents of approximately 80% of vaginal ridges could be seen in 17 girls. In knee–chest
the girls were either married or cohabiting. Sixty-four position, only one girl had more than one bump and 10
percent of the fathers (n = 189) and 62% of the mothers girls had visible ridges associated with the bumps.
(n = 193) had higher education defined as college or Outward folding of the hymenal edge was a frequent
university level. Compared with statistical data about finding in supine position. Of the 32 girls with this
the Norwegian population (22), the girls in the sample feature, 7 had an outward folding of the whole posterior
were from average-sized families with an average rim, giving a clinical impression of the hymen being
proportion of parents either married or cohabiting. attenuated. When these girls were examined in knee–
The parental educational level in the sample was high, chest position, the hymen unwound in two (Fig. 3a, b),
since approximately 30% of Norwegian males and and outward folding persisted in three girls (Fig. 4a, b).
nearly 40% of females in the age group 25–39 y have In one girl the hymenal folding could not be assessed in
college or university education (23). knee–chest position owing to posterior labial adhesions
In Table 2, a summary of general genital findings is and poor visualization of the hymenal edge. The last girl
provided. Signs of acute injury such as abrasions, refused to be examined in knee–chest position, but
haematomas and lacerations were not seen. One girl agreed to examination with saline irrigation. With this
with a history of recurrent genital soreness was referred technique, a small and delicate rim could be identified
to the Department of Dermatology, and diagnosed as (Fig. 5a, b).
having lichen sclerosus et atrophicus. In one girl, an Hymenal measurements are provided in Table 5. As
isolated “common skin-wart” lesion was located in the shown, the intra-observer agreement ranged from 80%
perineum and molluscum contagiosum was seen on the
labia majora in another girl.
Data about the hymenal configurations are provided Table 4. Posterior hymenal features behind the 3–9 o’clock line
in Table 3. A distinct hymenal erythema was found in 2 among the participants examined in the supine and knee–chest
girls, and 36 girls (18%) had prominent vessels in the positions.
hymen. Hymenal rim features behind the 3–9 o’clock
Supine Knee–chest
position position
Table 2. General genital findings in the study group. n = 175 n = 147
n % n % pa
n Total %
External ridges 4 2.3 3 2.0 n.s.
Gaping hymenal orifice 36 195 18.5 Perihymenal band 4 2.3 0 n.s.
Vaginal discharge 3 195 1.5 Tags 0 0 n.s.
Labial adhesion 15a 195 7.7 Bumps 41 23.4 31 21.1 n.s.
Friability of posterion fourchette 6 195 3.1 Folded outwards 32 18.3 11 7.5 p = 0.001
Linea vestibularis 3 133b 2.2 Folded inwards 3 1.7 1 0.7 n.s.
Fossa navicularis pallor 14 133b 10.3 Superficial notch 1 0.6 0 n.s.
Deep notch 0 0 n.s.
a
Twelve posterior, two anterior and one combined anterior and Transection 0 0 n.s.
posterior.
b a
Only girls with the fossa navicularis completely visualized on photos McNemar’s test performed on 147 girls examined in both supine and
were included. knee–chest positions.
ACTA PÆDIATR 92 (2003) Genital anatomy in preschool girls 1457

Fig. 3. (a) When examined in supine traction position, an outward folding of the whole hymenal rim can be seen in this girl. The hymenal edge
is indicated by the arrow. (b). When examined in knee–chest position, the left and posterior part of the hymen did unwind (arrow), while the
right part still rolled outward. Consent to publish the photograph was given by the parents. A = anterior, P = posterior, L = left, R = right.

Fig. 4. (a) In supine traction position, an outward folding of the whole hymenal rim gives an impression of a thickened, irregular and narrow
hymen with exposed vaginal contents. Hymenal edge indicated by an arrow. (b). In knee–chest position, the hymen did not unwind but still
rolled outward and was adherent to the vestibular floor. The edge of the hymen (arrow) has a scalloped appearance and can be distinguished
from the vestibular tissue. Consent to publish the photograph was given by the parents. A = anterior, P = posterior, L = left, R = right.
1458 AK Myhre et al. ACTA PÆDIATR 92 (2003)

Fig. 5. (a) In supine traction position, an outward folding of most of the hymenal rim exposed the posterior vaginal wall and two vaginal
ridges. (b) After installation of saline, a delicate hymenal rim (arrow) can be seen partly covering the posterior vaginal wall and the ridges.
Consent to publish the photograph was given by the parents. A = anterior, P = posterior, L = left, R = right.

to 100%. Differences between the examinations the examination, t-tests were performed to compare
methods were analysed with paired samples t-tests. measurements between girls observed as relaxed or
The horizontal diameter was significantly larger with non-relaxed. The mean horizontal diameter was found
the traction method compared with separation significantly larger for relaxed girls compared with non-
(p = 0.000), in traction compared with knee–chest relaxed girls examined within traction (5.22 and
(p = 0.000) and in knee–chest compared with the 4.27 mm, respectively, p = 0.002) and knee–chest
separation method (p = 0.000). For the vertical di- (4.09 and 2.52 mm, respectively, p = 0.000) methods.
ameter, no significant differences were found. The 6 No other differences were found.
o’clock rim was significantly wider with the separation The data were further examined for possible associa-
method compared with the traction method (p = 0.000), tions between hymenal measurements and the finding of
in separation compared with knee–chest (p = 0.000) and a gaping hymenal orifice. It was found that the mean
in knee–chest compared with traction method horizontal diameters obtained with all three methods
(p = 0.032). To explore the relationship between hyme- were significantly larger in girls with hymenal gaping
nal measurements and degree of child relaxation during compared with those without (separation method: 4.18

Table 5. Hymenal measurements (mm) in the study group.

n Mean (SD) Min 10a 50a 90a Max Intra-observer agreementb

Horizontal diameter
Separation method 104 3.27 (1.30) 0.7 1.55 3.25 5.30 6.2 95%
Traction method 142 5.03 (1.61) 0.8 3.10 5.00 6.97 9.5 95%
Knee–chest method 95 3.85 (1.54) 1.0 1.60 3.80 5.80 7.7 100%
Vertical diameterc
Separation method 16 5.58 (2.32) 1.5 1.78 5.60 8.63 8.7 100%
Traction method 23 5.71 (2.28) 0.8 2.12 6.00 8.52 10.7 100%
Knee–chest method 14 6.59 (2.63) 0.9 2.05 6.55 10.85 11.1 100%
Hymenal rim 6 o’clock
Separation method 87 3.75 (1.17) 1.7 2.20 3.70 5.44 6.9 92%
Traction method 91 3.37 (1.24) 1.1 1.80 3.20 4.96 7.9 80%
Knee–chest method 59 3.41 (0.94) 1.8 2.10 3.40 4.80 5.40 100%
a
Percentiles.
b
Based on re-review of 30 girls. Agreement defined as a maximum difference of 0.5 mm between measure one and two.
c
Only annular and transitional configuration.
ACTA PÆDIATR 92 (2003) Genital anatomy in preschool girls 1459

since it can be mistaken as scar tissue, was seen in

10.3% of the girls.
Of the posterior hymenal features reported in Table 4,
the outward folding of the hymen needs to be discussed.
Previous normative studies have not emphasized on this
finding, but it is classified as “normal anatomy” in
McCann’s Atlas (17). In our study, outward folding of
the hymen was frequently found and significantly more
common in the supine position (Table 4). In some of the
girls, an extensive outward folding of the posterior
hymen in supine position exposed the vaginal contents
and gave a false impression of the posterior hymen
being attenuated or even absent. Even though most
hymens did unwind in the knee–chest position, not all
did (Fig. 4a, b). In our opinion, the saline irrigation
technique (17) should be the method of choice to
differentiate between a true lack of posterior hymen
(which is considered to be clear evidence of penetrating
trauma) (15) and the normal finding of a folded hymen
(17).
Unfortunately, the saline irrigation technique was not
included in our study protocol. However, it was used in
one girl who refused to be examined in knee–chest
position, and a narrow hymenal rim was identified (Fig.
5b). The girl whose hymen did not unwind in knee–
chest position (Fig. 4b) refused further examination
Fig. 6. Fossa navicularis pallor (arrow) seen in a girl examined in
with saline irrigation. In this case, we consulted an
supine traction position. Consent to publish the photograph was external expert before classifying this finding as out-
given by the parents. A = anterior, P = posterior, L = left, R = right. ward folding of the hymen (J. McCann, pers. comm.).
After becoming familiar with the benefits of the saline
irrigation technique, this technique is now frequently
used in our centre.
vs 2.95 mm, p = 0.000, traction method: 5.87 vs Consistent with other studies (3, 4, 12), notches in the
4.79 mm, p = 0.001, knee–chest method: 4.99 vs posterior hymen were rarely seen. Using our definition,
3.45 mm, p = 0.000). No differences were found for only one girl was classified as having a superficial notch
vertical diameters and 6 o’clock rim. Finally, a possible in supine position. Unfortunately, this girl refused to be
association between hymenal measurements and height/ examined in the knee–chest position, so persistence of
weight of the girls was assessed using the Pearson the finding could not be assessed. When classifying
correlation analysis. A positive correlation between the hymenal notches, it is important, and often difficult, to
horizontal diameter using the supine traction method discriminate between real concavities in the hymen and
and height (r = 0.205, p = 0.015) as well as weight concavities caused by proximity to hymenal projec-
(r = 0.183, p = 0.029) was found, but no other associa- tions. In our opinion, careful examination of the photos
tions were revealed. through a stereoscope is recommended.
The use of hymenal measurements as a diagnostic
tool to differentiate between abused and non-abused
Discussion girls is a much-debated topic (13, 15, 26, 27). The small
number of participants in normative studies, the large
Morphological findings degree of variation within the same child depending on
Most of the general findings described in Table 2 have patient relaxation and examination method used and
been reported in previous studies of non-abused girls. poor accuracy of the measurements are important
Our rates of vaginal discharge (3, 5, 12), labial adhe- arguments against using this method. Measurement of
sions (4, 12), posterior fourchette friability (3) and linea the width of the hymenal rim is especially considered
vestibularis (4, 5, 12) are all consistent with previous inaccurate (13, 15), as the point where the hymen meets
studies. The finding of “fossa navicularis pallor” (Fig. the posterior vaginal wall can normally not be visua-
6) might be the same finding previously described in lized. Instead, an estimate based on a light reflex is used
newborns and infants as a partial linea vestibularis (24, (Fig. 2) (3, 13).
25), and it is defined in an anatomical atlas (17) as non- Recently, a case-control study compared hymenal
specific. In our study, this finding, which is important, measurements in 189 prepubertal children with a history
1460 AK Myhre et al. ACTA PÆDIATR 92 (2003)

Table 6. Comparison of hymenal measurements in studies of non-abused girls.

Study McCann 90 Berenson 92 Berenson 02a Present study

Population 5–8 y 4–7 y 3–8 y 5–7 y
Method Supine traction Supine traction Supine tractionb Supine traction
Horizontal diameter (mm)
n 43 25 181 142
Mean 5.6 3.6 5.57 5.03
SD ±1.8 ±1.2 ±1.63 ±1.6
Range 1.0–9.0 2.0–4.8 1.0–10.5 0.8–9.5
6 o’clock rim (mm)
n 75c 24 185 91
Mean 2.3 2.7 2.68 3.37
SD ±0.7 ±0.88 ±1.2
Range 0–6.0 1.0–3.8 1.25–6.5 1.1–7.9
Method Knee–chest Knee–chest Knee–chest
Horizontal diameter (mm)
n 41 177 95
Mean 5.6 4.55 3.85
SD ±1.5 ±1.87 ±1.5
Range 2.5–8.5 1.0–10.5 1.0–7.7
6 o’clock rim (mm)
n 75c 189 59
Mean 2.8 2.65 3.41
SD ±0.99 ±0.9
Range 1.0–8.0 1.0–7.5 1.8–5.4
a
Data from the control group of non-abused girls.
b
Examined in stirrups.
c
Age 0–10 y.

of genital penetration with 197 non-abused girls (13). We do agree with other investigators that hymenal
Although some significant differences were found, the measurements should never be used as the only tool in
investigators concluded that hymenal measurements the diagnosis of previous penetration (13).
demonstrated a high degree of overlapping between The finding of a gaping hymenal orifice (Fig. 1) was
abused and non-abused girls. If cut-off values for introduced as a supportive sign of sexual abuse (16),
horizontal diameters with acceptable specificity were and it was considered related to enlarged horizontal
chosen, the sensitivity became extremely low (1–2%) diameters and narrow hymenal rims (attenuated hy-
and the measurement became useless as a tool to detect mens). To the best of our knowledge, this finding has
abuse among the abused girls. not been reported in previous normative studies. In our
In Table 6, we have compared hymenal measure- study this feature was common (18.5%) and signifi-
ments from our study with the results from studies cantly associated with a large horizontal diameter, but
elsewhere. Interestingly, the data seem consistent, not with narrow hymenal rims. Our data suggest that
supporting their validity. The 6 o’clock rim shows a this finding is a common occurrence in girls with large
variation in mean width from 2.3 to 3.37 mm with horizontal hymenal diameters, and it should not be
supine traction and from 2.65 to 3.41 mm using the classified as suggestive of sexual abuse.
knee–chest method. With the exception of some girls in
McCann’s study with 6 o’clock rims <1.0 mm in supine
Methodological issues
position (3), no non-abused girls with 6 o’clock rims
<1.0 mm are described, either in supine or in knee– Selection bias. In our study, the parental educational
chest position in this age group. After consulting the level was well above the average for Norwegian adults
author, we were told that these girls had outward folded (23). Over-representation of well-educated parents or
hymens in supine position, which stretched out and middle-class families has also been reported in norma-
became more than 1.0 mm in knee–chest position (J. tive studies elsewhere (3, 5). From our experience, the
McCann, pers. comm.). In Berenson et al.’s study (13), main reason for this type of selection is that well-
hymenal rims at 6 o’clock <1.0 mm were only seen educated parents in general may have a positive attitude
among the abused girls. In the literature today (Table 6) towards research. Thus, most parents reported that the
we now have hymenal measurement data on a reason- main reason for their participation in the study was a
able number of girls within a defined age range. general concern about the problem of child sexual abuse
However, in our opinion the interpretation of these data in society, and they considered our study as an oppor-
should be used with caution in sexual abuse evaluations. tunity to contribute to its prevention. However, an
ACTA PÆDIATR 92 (2003) Genital anatomy in preschool girls 1461

important methodological question is whether this obtain health information about the child. As a way of
selection bias affects the possibility to generalize the exploring our data for the possibility of undetected
results. sexual abuse, we used the interview score as an
To address this question, we developed a score indicator for abuse. With a Mann–Whitney U-test, the
combining both parents’ education. Mann–Whitney U- score was compared between children with and without
tests were then used to compare parental educational rare categorical findings (findings seen in less than 5%),
scores for children with and those without each and no significant differences were found (data not
categorical finding reported. For continuous findings shown). Finally, a non-parametric correlation analysis
reported, association with educational score was (Spearman’s rho) was used to examine any associations
assessed using a non-parametric correlation analysis between continuous data and the interview score.
(Spearman’s rho). No significant association between Interestingly, significant correlations were revealed
any specific finding and parental education was revealed between the interview score and two hymenal measure-
(data not shown). We therefore believe that the ments obtained in knee–chest position: vertical dia-
selection bias did not have any influence on the validity meter (Spearman’s rho = 0.632, p = 0.015) and the 6
of our results. o’clock rim (Spearman’s rho = 0.304, p = 0.019). Thus,
the data indicate that a high interview score (supposed
to predict abuse) correlates with a small vertical
Undetected previous sexual abuse diameter and a wide hymenal rim. This is actually the
Another important question is whether some of the opposite of what would be expected, since sexual abuse
children in the sample could have been exposed to is considered associated with large hymenal diameters
undetected previous sexual abuse. The fact that we have and narrow rims (17). We find no plausible explanation
a high educational level among the parents does not for these findings, and the phenomena could be due to
preclude this possibility since epidemiological research chance, because of multiple comparisons. Thus, the
has not found that sexual abuse is less common when analysis based on the interview score supports our
parents have a higher level of education (28). In our assumption that undetected sexual abuse did not have an
study, it could be argued that parents abusing their influence on our results.
children would not participate out of fear of being
discovered. In the letter of information we provided a
telephone number for parents who suspected previous
abuse of their children, and some parents did actually
make contact. To prevent including children whose Conclusion
parents entered the study in order to have an “examina- In our study of assumed non-abused girls, a number of
tion for sexual abuse”, all parents were asked (after the genital anatomical features were described. An impor-
physical examination) if they had such suspicions or tant observation was outward folding of the posterior
concerns. The parents of three children admitted to hymenal rim in many girls, a finding that in some cases
some concern, and these children were excluded from was difficult to distinguish from attenuation or even
the study and referred to our sexual-abuse team. In absence of the posterior hymen. The hymenal folding
agreement with the Research Ethics Committee, no in- did stretch out in the knee–chest position in most, but
depth interview of the children in private about the topic not all, girls. In these girls we recommend the use of the
of sexual abuse was conducted. The main reason for this saline irrigation method. Another important observation
was a fear of embarrassing both the children and their was the fairly frequent finding of a gaping hymenal
parents. orifice and its significant association with a large
In the normative study conducted by McCann et al. horizontal diameter. Furthermore, the study provides
(3), a strategy for exclusion based on a combination of additional normative data on hymenal measurements.
results from a structured interview (SIPBES) and Even though such data now exist from a reasonable
occurrence of questionable physical findings was used. number of girls, we do agree with other investigators
The SIPBES score consists of the following data: that such measurements should be used with caution in
physical and psychosomatic problems, anal complaints, sexual abuse evaluations.
emotional functioning, sleep problems, social problems, In the literature of today, there is general agreement
sexual behaviour problems and genital problems. A that the history from the child provides the single most
high score was considered predictive of previous sexual important information when diagnosing child sexual
abuse. The questionnaire has later been validated and abuse (12–15). In cases with a history of genital
used to compare groups of abused and non-abused girls penetration, the genital examination may provide some
(29). supportive evidence, but usually the examination of
The structured interview used in our study was abused girls will reveal only normal or unspecific
designed along the lines of a model of the SIPBES, findings (10, 12–14). It is therefore important to empha-
but this modification has not been validated as yet. In size that a normal genital examination never rules out
our study, the main purpose of the interview was to the possibility of sexual abuse.
1462 AK Myhre et al. ACTA PÆDIATR 92 (2003)

Acknowledgements.—First and foremost, we thank the children and 14. Heger A, Ticson L, Velasquez O, Bernier R. Children referred for
their parents who participated in the study. We also thank Harald possible sexual abuse: medical findings in 2384 children. Child
Kielland for his assistance in developing the letter of invitation, the Abuse Negl 2002; 26: 645–59
project nurses, Gerd Eva Fenheim and Kari Gulla, for their help with 15. Adams JA. Evolution of a classification scale: medical evaluation
data collection and Eirik Skogvoll (MD), at the Unit for Applied of suspected child sexual abuse. Child Maltreat 2001; 6: 31–6
Clinical Research, for his assistance in data analysis. The study was 16. Hobbs CJ, Wynne JM, Thomas AJ. Colposcopic genital findings
supported by the Libero Diapter Fund (Mölnlycke SCA). in prepubertal girls assessed for sexual abuse. Arch Dis Child
1995; 73: 465–9
17. McCann JJ, Kerns D. The anatomy of child and adolescent sexual
abuse. A CD-ROM Atlas/Reference. St. Louis: InterCorp; 1999
References 18. Myhre AK, Berntzen K, Bratlid D. Perianal anatomy in non-
1. Finkelhor D. The international epidemiology of child sexual abused preschool children. Acta Pædiatr 2001; 90: 1321–8
abuse. Child Abuse Negl 1994; 18: 409–17 19. Myhre AK, Bevanger LS, Berntzen K, Bratlid D. Anogenital
2. Palusci VJ, Cox EO, Cyrus TA, Heartwell SW, Vandervort FE, bacteriology in non-abused preschool children: a descriptive
Pott ES. Medical assessment and legal outcome in child sexual study of the aerobic genital flora and the isolation of anogenital
abuse. Arch Pediatr Adolesc Med 1999; 153: 388–92 Gardnerella vaginalis. Acta Pædiatr 2002; 91: 885–91
3. McCann J, Wells R, Simon M, Voris J. Genital findings in 20. Terminology Subcommittee of the APSAC Task Force on Medi-
prepubertal girls selected for nonabuse: a descriptive study. cal Evaluation of Suspected Child Abuse. Chicago: APSAC;
Pediatrics 1990; 86: 428–39 1995
4. Berenson AB, Heger AH, Hayes JM, Bailey RK, Emans SJ. 21. Bland JM, Altman DG. Statistical methods for assessing agree-
Appearance of the hymen in prepubertal girls. Pediatrics 1992; ment between two methods of clinical measurement. Lancet
89: 387–94 1986; 1(8476)307–10
5. Gardner JJ. Descriptive study of genital variation in healthy, 22. Statistics Norway. http://www.ssb.no/emner/02/01/20/barn/tab-
nonabused premenarchal girls. J Pediatr 1992; 120: 251–7 2002-11-06-01.html (17.01.03). (Norwegian only)
6. Berenson AB. A longitudinal study of hymenal morphology in 23. Statistics Norway. http://www.ssb.no/utniv/tab-2002-10-31-
the first 3 years of life. Pediatrics 1995; 95: 490–6 03.html (17.01.03). (Norwegian only)
7. Berenson AB, Grady JJ. A longitudinal study of hymenal 24. Kellogg ND, Parra JM. Linea vestibularis: a previously
development from 3 to 9 years of age. J Pediatr 2002; 140: 600–7 undescribed normal genital structure in female neonates.
8. Heger AH, Ticson L, Guerra L, Lister J, Zaragoza T, McConnell Pediatrics 1991; 87: 926–9
G, et al. Appearance of the genitalia in girls selected for 25. Kellogg ND, Parra JM. Linea vestibularis: follow up of a normal
nonabuse: review of hymenal morphology and nonspecific genital structure. Pediatrics 1993; 92: 453–6
findings. J Pediatr Adolesc Gynecol 2002; 15: 27–35 26. Pugno PA. Genital findings in prepubertal girls evaluated for
9. McCann J, Voris J, Simon M. Genital injuries resulting from sexual abuse. A different perspective on hymenal measurements.
sexual abuse: a longitudinal study. Pediatrics 1992; 89: 307–17 Arch Fam Med 1999; 8: 403–6
10. Adams JA, Harper K, Knudson S, Revilla J. Examination 27. Ingram DM, Everett VD, Ingram DL. The relationship between
findings in legally confirmed child sexual abuse: it’s normal to the transverse hymenal orifice diameter by the separation
be normal. Pediatrics 1994; 94: 310–7 technique and other possible markers of sexual abuse. Child
11. Bowen K, Aldous MB. Medical evaluation of sexual abuse in Abuse Negl 2001; 25: 1109–20
children without disclosed or witnessed abuse. Arch Pediatr 28. Finkelhor D. Epidemiological factors in the clinical identification
Adolesc Med 1999; 153: 1160–4 of child sexual abuse. Child Abuse Negl 1993; 17: 67–70
12. Berenson AB, Chacko MR, Wiemann CM, Mishaw CO, 29. Wells RD, McCann J, Adams J, Voris J, Ensign J. Emotional,
Friedrich WN, Grady JJ. A case-control study of anatomic behavioral, and physical symptoms reported by parents of
changes resulting from sexual abuse. Am J Obstet Gynecol 2000; sexually abused, nonabused, and allegedly abused prepubescent
182: 820–31 females. Child Abuse Negl 1995; 19: 155–63
13. Berenson AB, Chacko MR, Wiemann CM, Mishaw CO,
Friedrich WN, Grady JJ. Use of hymenal measurements in the Received Mar. 10, 2003; revisions received Sept. 2, 2003; accepted
diagnosis of previous penetration. Pediatrics 2002; 109: 228–35 Oct. 6, 2003