Content uploaded by Kenneth J. Zucker
Author content
All content in this area was uploaded by Kenneth J. Zucker on Dec 20, 2017
Content may be subject to copyright.
NEW RESEARCH
Self-Harm and Suicidality in Children Referred
for Gender Dysphoria
Madison Aitken, MA, Doug P. VanderLaan, PhD, Lori Wasserman, MD,
Sonja Stojanovski, HonBSc, Kenneth J. Zucker, PhD
Objective: This study examined rates of self-harm and
suicidality (ideation and behavior) in children referred
clinically for gender dysphoria compared with their sib-
lings, and referred and nonreferred children from the
Child Behavior Checklist (CBCL) standardization sample.
Predictors or correlates of self-harm/suicidality were also
examined.
Method: The sample consisted of 572 gender-referred
children, 425 siblings, 878 referred children, and 903
nonreferred children. Parent report for 2 CBCL items was
used to assess self-harm and suicidality. CBCL total
behavior problems and a metric of peer relationship
problems were also used.
Results: The gender-referred children and the referred
children from the standardization sample had signifi-
cantly higher scores than siblings and nonreferred chil-
dren in terms of self-harm/suicidality, total behavior
problems, and poor peer relations. Based on logistic
regression analyses, gender-referred children were 5.1
times more likely than nonreferred children to talk about
suicide and 8.6 times more likely to self-harm/attempt
suicide, even after overall behavior problems and peer
relationship problems were accounted for. In the final
models, group, older age, and more total behavior prob-
lems, but not poor peer relations, were significantly
associated with an increased likelihood of self-harm/
suicidality.
Conclusion: By parent report, children with gender
dysphoria show an increased rate of self-harm/suicidality
as they get older. This risk was not simply an artifact of
the presence of behavioral and emotional problems,
although these problems were significant correlates of self-
harm/suicidality. Clinicians should routinely screen for
the presence of suicidal ideation and behavior in children
with gender dysphoria, particularly during the second
half of childhood.
Key words: gender dysphoria, suicidality, Child Behavior
Checklist
J Am Acad Child Adolesc Psychiatry 2016;55(6):513–520.
Children with gender dysphoria (GD) have a marked
incongruence between their assigned gender and
their experienced/expressed gender.
1
On standard-
ized parent- or teacher-report questionnaires of behavioral
and emotional problems, children with GD show, on
average, more problems than their siblings and nonreferred
children.
2,3
In addition, the magnitude of these problems
appears to be largely comparable in degree to that of chil-
dren referred clinically for other reasons.
2-5
For boys with
GD, there is a predominance of internalizing problems,
including separation anxiety,
6
whereas for girls with GD,
there is a more equal distribution of both internalizing and
externalizing problems.
7
Several studies of children with GD have examined cor-
relates or “predictors”of these problems.
8
For example,
1 study found that age was positively correlated with
behavioral and emotional problems on the Child Behavior
Checklist (CBCL).
9
A CBCL-derived metric of poor peer
relations (e.g., being teased) has proved to be the strongest
correlate of these problems in multiple regression
analyses.
2,3
Because peer relationship problems also increase
with age,
9
it is likely that the relationship between age and
behavioral and emotional problems is mediated by the
increase in poor peer relations. Using multiple informants,
similar findings have been reported for adolescents with GD,
including evidence that the degree of behavioral and
emotional problems is largely comparable to youth referred
for other clinical reasons.
3,10-13
Risk of suicide among adolescents with GD has recently
received a great deal of media attention, particularly
following the completed suicide of Ohio transgender teen
Leelah Alcorn in December 2014.
14-16
Indeed, there is some
evidence that adults with GD have a higher completed
suicide rate than adults without GD.
17-19
Systematic data on
completed suicides among adolescents with GD are not
known, although some clinicians have asserted that they are
“alarmingly high.”
20
Using case file information or responses to specific
questions, several studies have reported on the prevalence of
self-harm and suicidality (thoughts and behaviors) among
clinic-referred adolescents with GD, with sample sizes
ranging from 34 to 177.
21-25
For self-harm, the range was
28.8% to 41.0%; for suicidal ideation, the range was 17.5% to
42.2%; and for suicide attempts, the range was 11.9% to
15.8%. The time frame for these percentages was not
particularly clear, but 1 study reported a lifetime prevalence
This article is discussed in an editorial by Dr. Walter O. Bockting on
page 441.
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
VOLUME 55 NUMBER 6 JUNE 2016 www.jaacap.org 513
of 51% for suicidal ideation and 30.0% for suicide attempts.
26
Unfortunately, none of these studies used comparison
groups.
Steensma et al. reported the prevalence of self-harm/
suicidality in clinic-referred adolescents with GD from
Canada and the Netherlands using 2 items from the CBCL
and the Youth Self-Report (YSR) Form: CBCL Item
91 (“Talks about killing self”) and CBCL Item 18 (“Deliber-
ately harms self or attempts suicide”),
27,28
and compared the
prevalence rates with CBCL/YSR standardization data for
referred and nonreferred youth (Steensma TD, VanderLaan
DP, Cohen-Kettenis PT, et al. Suicidality in gender-dysphoric
adolescents: a cross-national, cross-clinic comparative anal-
ysis. Unpublished manuscript, 2014).
On the CBCL, the Toronto youth with GD (n ¼238) had
the highest rate of self-harm/suicidality (Item 91: 35.7%;
Item 18: 30.2%), whereas the Dutch youth with GD (n ¼250)
(Item 91: 24.8%; Item 18: 13.2%) had a rate that was more
comparable to that of the referred youth (Item 91: 18.5%;
Item 18: 13.5%) in the CBCL standardization sample. All 3 of
these groups had much higher rates than the nonreferred
youth (Item 91: 3.0%; Item 18: 0.5%). On the YSR, the pattern
was similar: the Toronto youth with GD (n ¼243) had the
highest rate of self-harm/suicidality (Item 91: 40.7%; Item
18: 28.8%), whereas the Dutch youth with GD (n ¼242)
(Item 91: 27.2%; Item 18: 20.6%) had a rate that was more
comparable to that of the referred youth (Item 91: 29.0%;
Item 18: 20.5%) in the YSR standardization sample. All 3 of
these groups had much higher rates than the nonreferred
youth (Item 91: 12.5%; Item 18: 6%).
Several reasons for this elevation among youth with GD
have been considered. One possibility is that gender
dysphoria is inherently distressing, which leads to self-harm
or suicidal thoughts and behaviors. A second possibility is
that such thoughts and behaviors are related to more global
behavioral and emotional problems
10-11,13
or to generic risk
factors for psychopathology. The most common explanation
in the extant literature is that suicidality is caused by social
ostracism or lack of social support.
19
At present, few data are available on the prevalence of
self-harm and suicidality in children with GD. Based on
case file data of 41 children with GD between the ages of 5
and 11 years, Holt et al.
22
found that 14.6% had a history of
suicidal ideation and 17.0% had a history of either self-
harm (14.6%) or suicide attempts (2.4%). These percent-
ages could have been underestimates, as the case file data
were based on varied sources of information, and it is
conceivable that suicidality was not asked about in all
cases. In addition, no comparison groups were used. The
present study examined the prevalence of suicidal thoughts
and behaviors in a consecutive series of children referred to
a specialized gender identity clinic. We compared these
rates with data from 3 control groups (siblings, children
referred for other clinical reasons, and nonreferred chil-
dren) using the same methodology used in the study of
adolescents from Toronto and the Netherlands with GD.
We then examined the effects of age, poor peer relations,
and behavioral problems in general on the prevalence of
self-harm/suicidality.
METHOD
Participants
The probands consisted of 572 gender-referred children (463 natal
boys; 109 natal girls) and 425 of their siblings (239 boys, 186 girls),
who ranged in age from 3 to 12 years, and for whom at least 1 parent
had completed the CBCL for children 4 to 18 years of age.
27
All of
the gender-referred children were evaluated in a specialized gender
identity service housed within a child and youth mental health
program at an academic health science center between 1976 and
2015 (mean year of assessment, 1997.72; SD ¼9.41). CBCLs were not
available for an additional 29 gender-referred children (e.g., because
the family dropped out of the assessment process) and 14 gender-
referred children whose parents completed the CBCL version for
2- to 3-year-olds,
29
which did not contain the relevant CBCL items
used in the current study (see below).
We also used CBCL data from the matched-pairs sample of 911
clinic-referred children (485 boys; 426 girls) and 911 nonreferred
children (485 boys; 426 girls) from Achenbach and Rescorla
30
who
ranged in age from 6 to 12 years. Referred children were recruited
from the 1999 National Survey of Children, Youth, and Adults and
from 20 outpatient and inpatient mental health services, primarily in
the United States, but also 1 site each from Australia and England.
30
The referred children were heterogeneous with regard to DSM
diagnoses. Nonreferred children were drawn from the 1999 National
Survey of Children, Youth, and Adults in the United States and had
not received mental health services in the 12 months before the
survey.
30
In the 2001 version of the CBCL, there is 1 item pertaining
to gender identity (Item 110: “Wishes to be of opposite sex”). For the
present study, we excluded the 33 referred children and 8 non-
referred children whose parents endorsed this item, leaving a total
sample of 878 referred children (472 boys; 406 girls) and 903
nonreferred children (482 boys; 421 girls).
Measures
We used CBCL ratings by the mother (if maternal ratings were not
available, ratings by father or another adult, such as a foster parent
or group home worker, were used). For this study, we examined
CBCL Item 18 (“Deliberately harms self or attempts suicide”) and
Item 91 (“Talks about killing self”) as metrics of suicidality. Like all
items on the CBCL, both items were rated on a 0- to 2-point scale
where 0 ¼“Not true,”1¼“Somewhat or sometimes true,”and
2¼“Very true or often true.”Across all groups, the correlation
between the 2 items was 0.53 (p<.001). We created an overall
suicidality index by summing the 2 items. We also dichotomized the
2 items, where 0 was dummy-coded as 0, and 1 to 2 dummy-coded
as 1. We also used a metric of poor peer relations consisting of CBCL
Item 25 (“Doesn’t get along with other kids”), Item 38 (“Gets teased
a lot”), and Item 48 (“Not liked by other kids”), which has been used
in prior studies.
2-3,31
In the present study, Cronbach’s
a
for this scale
was 0.81. Finally, we calculated the behavior problem sum scores of
all items on the CBCL (minus the 2 suicidality items). For the
gender-referred children and their siblings, the 2 CBCL items (Items
5 and 110) pertaining to gender identity from the original version
32
were artificially set to 0 and thus were not included in the total
problem score. For the gender-referred children and siblings, the
original version of the CBCL was used.
32
For the referred and
nonreferred children, however, the 2001 version of the CBCL was
used.
30
Of the 118 CBCL items, Items 2, 4, 5, 28, 78, and 99 in the
original version were replaced with 6 new items in the 2001 version.
Thus, the total problem score used in the current study was based on
slightly different item sets. However, it should be noted that the key
items used in the present study (i.e., the 2 suicidality items and the 3
items pertaining to poor peer relations) were identical. In the 2001
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
514 www.jaacap.org VOLUME 55 NUMBER 6 JUNE 2016
AITKEN et al.
version of the CBCL, Item 5 (“Behaves like opposite sex”) from the
1991 version was deleted and replaced with a different item.
Data Analysis
All analyses were conducted using SPSS version 22 software.
Between-groups comparisons on the CBCL measures were con-
ducted with either parametric or nonparametric statistics. The pri-
mary aim of these tests was to determine whether there were
significant differences on the CBCL measures, including the focal
examination of suicidality, among the 4 groups of children. All tests
were 2-tailed. Predictors of suicidality were examined with a series
of logistic regression analyses, with the nonreferred children from
the CBCL standardization sample serving as the reference group.
The present study constituted a reanalysis of data from previous
research projects for which there was ethics approval from the
Centre for Addiction and Mental Health.
RESULTS
Because the 2001 version of the CBCL did not include data
on referred and nonreferred children under the age of 6
years, our comparative analyses described below did not
include the data on the gender-referred children and siblings
who were under the age of 6 years.
Total Behavior Problems
Table 1 shows the mean sum of CBCL behavior problems
rated as 1 or 2 (minus the 2 suicidality items) as a function of
group and sex. A 4 (Group) 2 (Sex) analysis of variance
(ANOVA) yielded a significant Group Sex interaction
(F
3,2453
¼3.91, p¼.008). To decompose the significant
interaction, we conducted a 1-way ANOVA for Group,
which was significant (F
3,2497
¼413.43, p<.001), and
Duncan post hoc tests showed that the referred group had,
on average, significantly more behavior problems in general
than the other 3 groups; the gender-referred group had
significantly more behavior problems than the siblings and
nonreferred group; and the siblings had more behavior
problems than the nonreferred group (all p<.05). For sex,
post hoc analyses showed that the referred boys had, on
average, significantly more behavior problems than the
referred girls (t
876
¼3.90, p<.001). None of the other within-
group sex differences were statistically significant.
Poor Peer Relations
Table 1 also shows the mean sum score for the Poor Peer
Relations scale as a function of group and sex. A 4 (Group)
2 (Sex) ANOVA yielded a significant Group Sex interac-
tion (F
3,2453
¼3.85, p¼.009). To decompose the significant
interaction, a 1-way ANOVA for Group was significant
(F
3,2457
¼280.34, p<.001). Duncan post hoc tests showed
that both the referred group and the gender-referred group
had, on average, poorer peer relations than both the siblings
and the nonreferred group (all p<.05). The referred group
and the gender-referred group did not differ significantly on
poor peer relations, nor did the siblings and the nonreferred
group. For sex, post hoc analyses showed that the gender-
referred boys had, on average, significantly poorer peer re-
lations than the gender-referred girls (t
365
¼2.30, p¼.022),
and the same was the case for the referred group (t
876
¼4.20,
p<.001). The sex differences for the siblings and the non-
referred group were not significant.
Suicidality
Table 1 also shows the percentage of children as a function of
group in which the 2 CBCL suicidality items were rated as 1
or 2 (within-group comparisons did not reveal any signifi-
cant sex differences in the percentage of children for whom
the 2 suicidality items were rated as 1 or 2, so all analyses
collapsed across sex). For Item 91, a 4 (Group) 2 (Suici-
dality: Yes vs. No)
c
2
test was significant (
c
2
[3] ¼210.66,
TABLE 1 Child Behavior Checklist Measures as a Function of Group and Sex (6- to 12-Year-Olds)
Measures
Gender-Referred Siblings Referred Nonreferred
Mean SD n Mean SD n Mean SD n Mean SD n
Sum Behavior Problems
a
Boys 51.28 27.49 289 30.02 24.04 168 66.36 31.00 472 24.17 17.55 482
Girls 53.08 26.22 78 25.92 20.22 145 58.05 31.86 406 22.61 16.50 421
Sum Poor Peer Relations
b
Boys 2.50 1.85 289 0.73 1.34 168 2.43 1.71 472 0.59 1.03 482
Girls 1.97 1.48 78 0.57 1.03 145 1.95 1.69 406 0.52 0.88 421
Sum Suicidality Index
c
Boys 0.30 0.67 289 0.13 0.50 168 0.55 0.95 472 0.02 0.15 482
Girls 0.31 0.69 78 0.06 0.31 145 0.46 0.92 406 0.02 0.13 421
Item 18 (% rated 1 or 2) 6.6 2.2 19.1 0.2
Item 91 (% rated 1 or 2) 19.1 5.8 23.3 1.8
Note:
a
Absolute range, 0230 (for the referred and nonreferred children) and 0228 for the gender-referred children and their siblings. The sum score excluded the 2
suicidality items for all groups and Item 110 for the referred and nonreferred children or Items 5 and 110 for the gender-referred children and their siblings (see text
for details).
b
Absolute range, 06.
c
Absolute range, 04.
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
VOLUME 55 NUMBER 6 JUNE 2016 www.jaacap.org 515
SELF-HARM, SUICIDALITY, AND GENDER DYSPHORIA
p<.001), as was the case for Item 18 (
c
2
[3] ¼223.48,
p<.001). For Item 91 (“Talks about killing self”), the
referred group had the highest percentage (22.7%), followed
by the gender-referred group (19.1%), the siblings (5.8%),
and the nonreferred group (1.7%). A similar pattern was
evident for Item 18 (“Deliberately harms self or attempts
suicide”): 18.6%, 6.5%, 2.2%, and 0.2%, respectively.
We next calculated an overall suicidality index by sum-
ming the ratings across the 2 items (Table 1). A 1-way
ANOVA for Group was significant (F
3,2457
¼94.06,
p<.001). Duncan post hoc tests showed that the referred
group had, on average, a significantly higher score than the
other 3 groups, and the gender-referred group had a
significantly higher score than the siblings and the non-
referred group (all p<.05), who did not differ significantly
from each other. For the gender-referred probands, the
correlation between year of assessment and the suicidality
index was r¼0.02, which was not significant.
Logistic Regression
Age Effects on Suicidality. Table 2 shows the percentage of
children in which the 2 suicidality items were endorsed as a
function of age. From Table 2, it can be seen that the highest
percentages were among the 10- and 11-year-olds for Item 91
and among the 11-year-olds for Item 18 (with the exception
of the nonreferred children, where there were floor effects).
For Item 91, the gender-referred group had the largest range
(9.333.3%) across the age range of 6 to 12 years, followed
by the referred group (12.629.4%). The same was true for
Item 18: for the gender-referred group, the range was from
1.7% to 17.2%, and for the referred group, the range was
from 14.8% to 21.6% (for completeness, Table 2 includes the
data on gender-referred children and their siblings who
were under the age of 6 years).
To evaluate age effects, we conducted 2 logistic re-
gressions in which group and age were entered as predictor
variables. Item 91 or Item 18 was the criterion variable. For
these analyses, the nonreferred group served as the reference
group. The results of the logistic regression are shown in
Table 3. Each coefficient, B, represents the change in the log
odds of scoring positive for suicidality for a 1-unit increase
in the corresponding predictor, controlling for all other
predictors in the model. The next column presents the
standard error (SE) for each B. The Wald statistic was the
quantity used to determine the significance level of each
predictor variable. The quantity, e
B
, is the multiplicative
change in the odds of scoring positive for suicidality for a
1-unit increase in the corresponding predictor, and thus
100 (e
B
- 1) represents the percentage change in the odds
for a 1-unit increase in that predictor.
33
For Item 91 (ideation), both group and age were signifi-
cant (p<.001). The gender-referred group was 15.21 times
more likely to talk about killing oneself than the nonreferred
group (p<.001); the siblings were 3.57 times more likely
(p¼.001), and the referred group was 17.65 times more
likely (p<.001). With each 1-unit increase in age, children
across all groups were 1.16 times more likely to talk about
killing oneself. For Item 18 (behavior), group was a
TABLE 2 Parental Endorsement of Child Behavior Checklist Suicidality Items as a Function of Group and Age
Age Group (y)
Item 91: Talks about killing self
Group
34 5 6 7 8 9 10 11 12
%n%n%n%n%n%n%n%n%n
Gender dysphoria 0.9 110 7.4 95 9.3 86 10.2 59 18.3 71 26.0 50 33.3 39 27.6 29 27.3 33
Siblings 1.4 69 2.3 43 2.6 39 4.2 48 4.8 62 8.9 45 2.3 43 11.9 42 5.9 34
Referred 15.5 116 12.6 111 18.2 121 27.0 122 25.3 98 29.4 136 27.2 173
Nonreferred 2.5 119 1.7 115 1.6 129 1.6 123 3.8 104 0.7 137 0.6 176
Item 18: Deliberately harms self or attempts suicide
Group
34 5 6 7 8 9 10 11 12
%n%n%n%n%n%n%n%n%n
Gender dysphoria 0.0 110 5.3 95 2.3 86 1.7 59 9.9 71 4.0 50 12.8 39 17.2 29 6.1 33
Siblings 1.4 69 2.3 43 2.6 39 2.1 48 1.6 62 4.4 45 0.0 43 2.4 42 2.9 34
Referred 19.8 116 13.5 111 18.2 121 20.5 122 16.2 98 21.3 136 19.1 173
Nonreferred 0.0 119 0.0 115 0.8 129 0.0 123 0.0 104 0.7 137 0.0 176
Note: Referred and nonreferred children are from Achenbach and Rescorla.
30
Raw data provided by T. M. Achenbach in an SPSS data file.
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
516 www.jaacap.org VOLUME 55 NUMBER 6 JUNE 2016
AITKEN et al.
TABLE 3 Logistic Regression Predictors for the 2 Child Behavior Checklist (CBCL) Suicidality Items (Group and Age; Group, Age,
and Poor Peer Relations; Group, Age, Poor Peer Relations, and General Behavior Problems)
Item 91: “Talks about killing self”
Step 1 BSE Wald df p e
B
95% CI
Group 140.17 3 <.001
Gender-referred 2.72 0.29 85.85 1 <.001 15.21 8.55e27.05
Siblings 1.27 0.35 12.74 1 <.001 3.57 1.77e7.18
Referred 2.87 0.27 110.52 1 <.001 17.65 10.33e30.15
Age 0.15 0.03 22.35 1 <.001 1.16 1.09e1.24
Item 18: “Deliberately harms self or attempts suicide”
Step 1 BSE Wald df p e
B
95% CI
Group 93.35 3 <.001
Gender-referred 3.47 0.73 22.13 1 <.001 32.34 7.59e137.65
Siblings 2.32 0.80 8.36 1 .004 10.24 2.11e49.56
Referred 4.63 0.71 42.20 1 <.001 102.85 25.41e416.24
Age 0.056 0.038 2.12 1 .14 1.05 0.98e1.14
Item 91: “Talks about killing self”
Step 1 BSE Wald df p e
B
95% CI
Group 69.98 3 <.001
Gender-referred 1.96 0.30 40.74 1 <.001 7.13 3.90e13.05
Siblings 1.19 0.36 11.11 1 .001 3.31 1.63e6.70
Referred 2.23 0.28 672.38 1 <.001 9.34 5.36e16.27
Age 0.11 0.03 10.61 1 .001 1.11 1.04e1.19
Poor Peer Relations 0.36 0.04 82.85 1 <.001 1.43 1.32e1.55
Item 18: “Deliberately harms self or attempts suicide”
Step 1 BSE Wald df p e
B
95% CI
Group 73.53 3 <.001
Gender-referred 2.62 0.74 12.28 1 <.001 13.79 3.18e59.87
Siblings 2.24 0.80 7.73 1 .005 9.41 1.93e45.71
Referred 3.95 0.71 30.20 1 <.001 51.97 12.70e212.65
Age 0.005 0.04 0.01 1 NS 1.01 0.93e1.08
Poor Peer Relations 0.38 0.04 65.80 1 <.001 1.47 1.34e1.61
Item 91: “Talks about killing self”
Step 1 BSE Wald df p e
B
95% CI
Group 28.32 3 <.001
Gender-referred 1.63 0.31 27.08 1 <.001 5.10 2.76e9.44
Siblings 0.97 0.36 6.99 1 .008 2.65 1.28e5.47
Referred 1.38 0.29 21.74 1 <.001 4.00 2.23e7.16
Age 0.10 0.03 7.86 1 .005 1.10 1.03e1.18
Poor Peer Relations e0.00 0.05 0.01 1 NS 0.99 0.90e1.10
General behavior problems 0.03 0.00 129.58 1 <.001 1.03 1.03e1.04
Item 18: “Deliberately harms self or attempts suicide”
Step 1 BSE Wald df p e
B
95% CI
Group 31.37 3 <.001
Gender-referred 2.15 0.75 8.12 1 .004 8.58 1.95e37.63
Siblings 1.90 0.81 5.45 1 .02 6.73 1.35e33.36
Referred 2.99 0.72 16.96 1 <.001 19.97 4.80e83.03
Age e0.02 0.04 0.22 1 NS 0.98 0.90e1.06
Poor Peer Relations e0.03 0.06 0.22 1 NS 0.97 0.85e1.09
General behavior problems 0.04 0.00 112.12 1 <.001 1.04 1.03e1.04
Note: The reference group was the nonreferred children in the CBCL standardization sample. NS ¼not significant.
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
VOLUME 55 NUMBER 6 JUNE 2016 www.jaacap.org 517
SELF-HARM, SUICIDALITY, AND GENDER DYSPHORIA
significant predictor (p<.001), but age was not. The gender-
referred group was 32.34 times more likely to deliberately
self-harm or attempt suicide than the nonreferred group
(p<.001), the siblings were 10.24 times more likely
(p¼.004), and the referred group was 102.85 times more
likely (p<.001).
Effects of Poor Peer Relations on Suicidality. The correlation
between the measure of poor peer relations and suicidality
was 0.36 (p<.001). To evaluate the effects of poor peer re-
lations, it was added to the logistic regression models. As
shown in Table 3, for Item 91 (ideation), group, age, and
poor peer relations were all significant. The gender-referred
group was 7.13 times more likely to talk about killing oneself
than the nonreferred group (p<.001); the siblings were 3.31
times more likely (p¼.001), and the referred group was 9.34
times more likely (p<.001). For Item 18 (behavior), group
and poor peer relations were significant predictors (both
p<.001), but age was not. The gender-referred group was
13.79 times more likely to deliberately self-harm or attempt
suicide than the nonreferred group (p<.001), the siblings
were 9.41 times more likely (p¼.005), and the referred
group was 51.97 times more likely (p<.001). These findings
show that controlling for poor peer relations reduced the
odds of suicidality (both in ideation and in behavior) in all
3 groups compared to the nonreferred group, but the odds
ratios were still statistically significant.
Effects of Behavior Problems in General on Suicidality. The
correlation between the number of behavior problems and
the suicidality index was 0.51 (p<.001). To evaluate the
effects of general behavior problems, it was added to the
logistic regression models. As shown in Table 3, for Item 91
(ideation), group, age, and general behavior problems were
all significant, but poor peer relations was not. The gender-
referred group was 5.10 times more likely to talk about
killing oneself than the nonreferred group (p<.001), the
siblings were 2.65 times more likely (p¼.008), and the
referred group was 4.00 times more likely (p<.001). For
Item 18 (behavior), group and general behavior problems
were significant predictors (both p<.001), but age and poor
peer relations were not. The gender-referred group was 8.58
times more likely to deliberately harm self or attempt suicide
than the nonreferred group (p¼.004), the siblings were
6.73 times more likely (p¼.02), and the referred group was
19.97 times more likely (p<.001). These findings show that
controlling for general behavior problems reduced the odds
of suicidality (in both ideation and in behavior) in all
3 groups compared to the nonreferred group, but the odds
ratios were still statistically significant. With general
behavior problems added to the regression equation, poor
peer relations was no longer significant.
DISCUSSION
Based on parent report, this study examined the prevalence
of both suicidal ideation and self-harm/suicidal behavior in
a sample of children referred for gender identity concerns
and compared these rates to those of their siblings, referred,
and nonreferred children. The referred children in the CBCL
standardization sample had, on average, a significantly
higher suicidality sum score than the gender-referred chil-
dren, but both groups had a significantly higher score than
the siblings of the gender-referred children and the non-
referred children in the CBCL standardization sample,
which did not differ significantly from each other. As noted
above, Holt et al.
22
reported that 14.6% of their sample of
children with GD aged 5 to 11 years had a history of suicidal
ideation. We obtained a similar rate of 15.8% for the 5- to 11-
year-old children with GD in our sample (Table 2). For self-
harm/suicide attempts, Holt et al. reported a higher rate
than what was obtained in our sample (17.0% versus 6.2%).
If, compared to the sample in the Holt et al. study, our
sample had a much higher proportion of younger children
within this age range (which it most likely did), it would
account for our lower rate of self-harm/suicide attempts.
In the gender-referred group, suicidal ideation (Item 91)
showed a relatively linear relationship with age: as shown in
Table 2, there was relatively little such ideation in the very
young children, but by ages 10 to 12 years, it was present in
about 30% of the sample. This age-related pattern was also
present among the referred children in the CBCL standardi-
zation sample. Self-harm/suicide attempts (Item 18) were less
common than talk about suicide in the gender-referred group:
at no age was the prevalence of the former higher than the
latter. However, the age-related pattern among the referred
children was more complex: for example, at ages 6 to 8 years,
the percentage of children who engaged in self-harm/suicide
attempts was quite similar to talk about suicide; it was only
starting at age 9 years that the latter exceeded the former.
In the first regression model, in which only group and age
were entered as predictors, both variables were significant for
Item 91, and group was significant for Item 18. Compared to
the reference group of nonreferred children, the gender-
referred children, their siblings, and the referred children all
showed significantly increased odds of suicidal ideation and
behavior and the odds ratios were higher for self-harm/
attempts suicide than for talk about killing oneself. When we
added poor peer relations to the model, the odds ratios were
reduced in magnitude, but still significant, suggesting that
poor peer relations are 1 potential correlate of suicidal ideation
and behavior. However, when we added general behavior
problems to the model, poor peer relations was no longer sig-
nificant. Thus, inthe present study, the results suggest that the
most important correlate of suicidal ideation and behavior
among gender-referred children, their siblings, and referred
children was the presence of behavior problems in general.
Our metric of suicidality was relatively crude, as it was
based on only 2 items from the CBCL. In addition, it should
be noted that Item 18, as written, indexes either actual sui-
cide attempts or self-harming behavior (or both), and the
latter may not be a marker of suicidality per se. Nonetheless,
the correlation between Items 91 and 18 was reasonable,
with an rof 0.53. It would, of course, be desirable in future
studies to use a psychometrically superior measurement tool
that better differentiates self-harm from bona fide suicide
attempts.
34
It would also be important to obtain self-report
of suicidal thoughts, at least for older children, using
dimensional measures such as the Suicidal Ideation Ques-
tionnaireJunior.
35
Qualitative data about the reasons for
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
518 www.jaacap.org VOLUME 55 NUMBER 6 JUNE 2016
AITKEN et al.
self-harm and suicidality among gender-referred children
should also be examined to determine to what extent the
proximal precipitants are related to gender identity issues
per se and not to other factors.
As noted above, the referred children had a significantly
higher mean sum score on the suicidality index and also had
more behavioral and emotional problems in general than the
gender-referred group. In prior studies, we have shown that
when gender-referred children are demographically
matched to clinical controls, they appear to have similar
levels of CBCL behavior problems.
5
One reason that the
referred children may have been more extreme than the
gender-referred children in the present study, with regard to
both behavior problems in general and suicidality in
particular, pertains to sampling. As noted earlier, the
referred children were obtained from both outpatient and
inpatient settings, and it is quite likely that the inclusion of
children from inpatient settings resulted in more severe
CBCL ratings. In contrast, all of the gender-referred children
were seen as outpatients, although a small percentage of
these children were in day-treatment or in the care of a child
protection agency at the time of assessment. Compared to
CBCL standardization data on referred children from prior
samples,
32,36
it appears that the CBCL 2001 sample was
somewhat more extreme in degree of behavior problems in
general and suicidality in particular.
30
Unfortunately, the
raw data did not include a variable denoting outpatient
versus inpatient status, so it is not possible to formally test
this conjecture. In any case, it is important to note that the
gender-referred children were significantly higher in suici-
dality than their siblings and nonreferred children.
In the present study, once we controlled for general
behavior problems, poor peer relations was no longer a
significant predictor of suicidal ideation and behavior.
Thus, we cannot argue that social ostracism of gender-
referred children was a unique correlate of suicidality.
This, perhaps, is not surprising, as our metric of poor peer
relations was based on only 3 items from the CBCL,
whereas the measure of general behavior problems was
based on a much larger number of items and thus may have
been a more robust measure. Our measure of poor peer
relations also did not take into account the potential dif-
ferential influence of same-sex versus opposite-sex peers,
which may have diluted its power. There is some evidence
that gender-referred children, particularly natal boys, are
more likely to have trouble with same-sex peers than
opposite-sex peers.
37
Thus, future studies should take this
into account when attempting to gauge the potential effects
of social ostracism within the peer group on suicidal idea-
tion and behavior.
In summary, the results of the present study suggest that
self-harm and suicidality are part of the clinical presentation
of a sizeable minority of gender-referred children, particu-
larly in the second half of childhood. As noted in Table 2,
33.3% of 10-year-old gender-referred children were rated by
the parent as expressing suicidal ideation, and 17.2% of the
11-year-olds were rated as engaging in self-harm or suicide
attempts. These data were obtained over a period of 40
years, which was made possible by the use of a standardized
measure. One could, however, ask if secular changes (e.g., a
greater societal acceptance of gender-variant children) might
cloud the use of a data set from such a long period of time. In
this regard, it is of note that our suicidality metric showed a
correlation of almost 0 with year of assessment, suggesting
that whatever secular changes have taken place, it does not
appear to have an impact on parent report of self-harm/
suicidality in this particular clinical population. To some
extent, this finding is consistent with broader analyses of
CBCL behavior problems, for which there have been only
very small changes in sampled cohorts.
38,39
Our data suggest that it would be prudent for the prac-
ticing clinician to routinely ask about the presence of suici-
dality during the assessment of children with gender
dysphoria, along with an evaluation of behavioral and
emotional problems more generally. The presence of these
problems then requires a case formulation in which both
distal (e.g., biological vulnerability to psychiatric disorder)
and proximal (e.g., social problems) factors are taken into
account in developing a plan for clinical management. &
Accepted April 1, 2016.
Ms. Aitken is with the Ontario Institute for Studies in Education of the University
of Toronto, Ontario, Canada. Dr. VanderLaan is with University of Toronto
Mississauga, Ontario, and the Child, Youth, and Family Program, Centre for
Addiction and Mental Health, Toronto. Dr. Wasserman is with the Repro-
ductive Life Stages Program, Women’s College Hospital, Toronto. Ms. Sto-
janovski was with the Gender Identity Service, Child, Youth, and Family
Program, Centre for Addiction and Mental Health, Toronto and is now with the
Research Institute, Hospital for Sick Children, Toronto. Dr. Zucker is with Uni-
versity of Toronto.
Dr. VanderLaan was supported by a Canadian Institutes of Health Research
postdoctoral fellowship.
Disclosure: Drs. VanderLaan, Wasserman, Zucker, and Mss. Aitken and
Stojanovski report no biomedical financial interests or potential conflicts of
interest.
Correspondence to Kenneth J. Zucker, PhD, Department of Psychiatry, Uni-
versity of Toronto, 250 College Street, Toronto, ON M6J 1H4, Canada;
e-mail: ken.zucker@utoronto.ca
0890-8567/$36.00/ª2016 American Academy of Child and Adolescent
Psychiatry
http://dx.doi.org/10.1016/j.jaac.2016.04.001
REFERENCES
1. American Psychiatric Association. Diagnostic and Statistical Manual of
Mental Disorders (Fifth ed.). Arlington, VA: American Psychiatric Pub-
lishing; 2013.
2. Cohen-Kettenis PT, Owen A, Kaijser VG, Bradley SJ, Zucker KJ.
Demographic characteristics, social competence, and behavior problems
in children with gender identity disorder: a cross-national, cross-clinic
comparative analysis. J Abnorm Child Psychol. 2003;31:41-53.
3. Steensma TD, Zucker KJ, Kreukels BPC, et al. Behavioral and emotional
problems on the Teacher’s Report Form: a cross-national, cross-clinic
comparative analysis of gender dysphoric children and adolescents.
J Abnorm Child Psychol. 2014;42:635-647.
4. Bates JE, Bentler PM, Thompson SK. Gender-deviant boys compared with
normal and clinical control boys. J Abnorm Child Psychol. 1979;7:243-259.
5. Zucker KJ, Bradley SJ. Gender Identity Disorder and Psychosexual
Problems in Children and Adolescents. New York: Guilford Press; 1995.
6. Zucker KJ, Bradley SJ, Lowry Sullivan CB. Traits of separation anxiety in
boys with gender identity disorder. J Am Acad Child Adolesc Psychiatry.
1996;35:791-798.
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
VOLUME 55 NUMBER 6 JUNE 2016 www.jaacap.org 519
SELF-HARM, SUICIDALITY, AND GENDER DYSPHORIA
7. Zucker KJ. Gender identity disorder in girls. In: Bell DJ, Foster SL,
Mash EJ, eds. Handbook of Behavioral and Emotional Problems in Girls.
New York: Kluwer Academic/Plenum Publishers; 2005:285-319.
8. Zucker KJ, Wood H, VanderLaan DP. Models of psychopathology in
children and adolescents with gender dysphoria. In: Kreukels BPC,
Steensma TD, de Vries ALC, eds. Gender Dysphoria and Disorders of Sex
Development: Progress in Care and Knowledge. New York: Springer;
2014:171-192.
9. Zucker KJ. Psychosocial and erotic development in cross-gender identi-
fied children. Can J Psychiatry. 1990;35:487-495.
10. de Vries AL, Steensma TD, Cohen-Kettenis PT, VanderLaan DP,
Zucker KJ. Poor peer relations predict parent- and self-reported behav-
ioral and emotional problems of adolescents with gender dysphoria: a
cross-national, cross-clinic comparative analysis [published online ahead
of print 15 September 2015]. Eur Child Adolescent Psychiatry. 2015.
http://dx.doi.org/10.1007/s00787-015-0764-7.
11. Shiffman M, VanderLaan DP, Wood H, et al. Behavioral and emotional
problems as a function of peer relationships in adolescents with gender
dysphoria: a comparison with clinical and nonclinical controls. Psychol
Sex Orientation Gend Diversity. 2016;3:27-36.
12. Skagerberg E, Davidson S, Carmichael P. Internalizing and externalizing
behaviors in a group of young people with gender dysphoria. Int J
Transgenderism. 2013;14:105-112.
13. Zucker KJ, Bradley SJ, Owen-Anderson A, et al. Demographics, behavior
problems, and psychosexual characteristics of adolescents with gender
identity disorderor transvestic fetishism. J Sex Marital Ther. 2012;38:51-189.
14. Boylan JF. How to save your life: a response to Leelah Alcorn’s suicide
note. New York Times. January 6, 2015. Available at: http://www.
nytimes.com/2015/01/07/opinion/a-response-to-leelah-alcorns-suicide-
note.html?_r¼0. Accessed November 7, 2015.
15. Abeni C. San Diego mourns fourth trans teen lost to suicide this year.
Advocate. October 7, 2015. Available at: http://www.advocate.com/
transgender/2015/10/07/san-diego-mourns-fourth-trans-teen-lost-suicide
-year. Accessed November 7, 2015.
16. Pesce NL. After Caitlyn Jenner addresses transgender teen suicide, Kyler
Prescott’s mother shares her son’s story. New York Daily News. July 27,
2015. Available at: http://www.nydailynews.com/entertainment/
caitlyn-jenner-addresses-trans-suicide-prescott-family-article-1.2305827.
Accessed December 7, 2015.
17. Asscheman H, Giltay EJ, Megens JA, de Ronde WP, van
Trotsenburg MA, Gooren LJ. A long-term follow-up study of mortality in
transsexuals receiving treatment with cross-sex hormones. Eur J Endo-
crinol. 2011;164:635-642.
18. Dhejne C, Lichtenstein P, Boman M, Johansson AL, Långstr€
om N,
Land
en M. Long-term follow-up of transsexual persons undergoing sex
reassignment surgery: cohort study in Sweden. PLoS One. 2011;6:e16885.
19. Zucker KJ, Lawrence AA, Kreukels BPC. Gender dysphoria in adults.
Annu Rev Clin Psychol. 2016;12:217-247.
20. Karasic D, Ehrensaft D. We must put an end to gender conversion
therapy for kids. WIRED. July 6, 2015. Available at: http://www.
wired.com/2015/07/must-put-end-gender-conversion-therapy-kids/.
Accessed November 7, 2015.
21. Becker M, Gjergji-Lama V, Romer G, M€
oller B. Merkmale von Kindern
und Jugendlichen mit Geschlechtsdysphorie in der Hamburger
Spezialsprechstunde [Characteristics of children and adolescents with
gender dysphoria referred to the Hamburg Gender Identity Clinic].
Praxis Kinderpsychol Kinderpsychiatr. 2014;63:486-509.
22. Holt V, Skagerberg E, Dunsford M. Young people with features of gender
dysphoria: demographics and associated difficulties. Clin Child Psychol
Psychiatry. 2016;164:108-118.
23. Khatchadourian K, Amed S, Metzger L. Clinical management of youth
with gender dysphoria in Vancouver. J Pediatr. 2014;164:906-911.
24. Skagerberg E, Parkinson R, Carmichael P. Self-harming thoughts and
behaviors in a group of children and adolescents with gender dysphoria.
Int J Transgenderism. 2013;14:86-92.
25. Kaltiala-Heino R, Sumia M, Ty€
ol€
aj€
arvi M, Lindberg N. Two years of
gender identity service for minors: overrepresentation of natal girls with
severe problems in adolescent development. Child Adolesc Psychiatry
Ment Health. 2015;9:9.
26. Olson J, Schrager SM, Belzer M, Simons LK, Clark LF. Baseline physio-
logic and psychosocial characteristics of transgender youth seeking care
for gender dysphoria. J Adolesc Health. 2015;57:374-380.
27. Achenbach TM. Manual for the Child Behavior Checklist/4-18 and 1991
Profile. Burlington, VT: University of Vermont, Department of Psychia-
try; 1991.
28. Achenbach TM. Manual for the Youth Self-Report. Burlington, VT: Uni-
versity of Vermont, Department of Psychiatry; 1991.
29. Achenbach TM, Edelbrock C, Howell CT. Empirically based assessment
of the behavioral/emotional problems of 2-3-year-old children. J Abnorm
Child Psychol. 1987;15:629-650.
30. Achenbach TM, Rescorla LA. Manual for the ASEBA School-Age Forms
and Profiles. Burlington, VT: University of Vermont, Research Center for
Children, Youth, and Families; 2001.
31. Zucker KJ, Bradley SJ, Sanikhani M. Sex differences in referral rates of
children with gender identity disorder: some hypotheses. J Abnorm
Child Psychol. 1997;25:217-227.
32. Achenbach TM, Edelbrock CS. Manual for the Child Behavior Checklist
and Revised Child Behavior Profile. Burlington, VT: University of Ver-
mont, Department of Psychiatry; 1983.
33. DeMaris A. Logit Modeling: Practical Applications. Newbury Park, CA:
Sage; 1992.
34. Hill RM, Castellanos D, Pettit JW. Suicide-related behaviors and anxiety
in children and adolescents: a review. Clin Psychol Rev. 2001;31:
1133-1144.
35. Reynolds W. Suicidal Ideation QuestionnaireJunior. Odessa, FL: Psy-
chological Assessment Resources; 1987.
36. Achenbach TM, Edelbrock CS. Behavioral problems and competencies
reported by parents of normal and disturbed children aged four through
sixteen. Monogr Soc Res Child Dev. 1981;46:1-82.
37. Wallien MSC, Veentra R, Kreukels BPC, Cohen-Kettenis PT. Peer group
status of gender dysphoric children: a sociometric study. Arch Sex Behav.
2010;39:553-560.
38. Achenbach TM, Howell CT. Are American children’s problems getting
worse? A 13-year comparison. J Am Acad Child Adolesc Psychiatry.
1993;32:1145-1154.
39. Achenbach TM, Dumenci L, Rescorla LA. Are American children’s
problems still getting worse? A 23-year-comparison. J Abnorm Child
Psychol. 2003;31:1-11.
JOURNAL OF THE AMERICAN ACADEMY OF CHILD &ADOLESCENT PSYCHIATRY
520 www.jaacap.org VOLUME 55 NUMBER 6 JUNE 2016
AITKEN et al.
