You are on page 1of 9

Journal of Psychiatric Research 43 (2009) 10861094

Contents lists available at ScienceDirect

Journal of Psychiatric Research

journal homepage:

Association of trait-dened, eating-disorder sub-phenotypes with

(biallelic and triallelic) 5HTTLPR variations
Howard Steiger *, Jodie Richardson, Norbert Schmitz, Ridha Joober, Mimi Israel, Kenneth R. Bruce,
Lise Gauvin, Cathy Dandurand, Annelie Anestin
Eating Disorders Program, Douglas University Institute in Mental Health, 6875 LaSalle Blvd., Montreal (Verdun), Quebec, Canada H4H 1R3

a r t i c l e

i n f o

Article history:
Received 7 January 2009
Received in revised form 12 March 2009
Accepted 17 March 2009

Eating disorders
Latent classes

a b s t r a c t
Context: Efforts to classify eating-disordered individuals based on concurrent personality traits have
consistently converged on a typology encompassing over-regulated, dysregulated, and low
psychopathology subgroups. In various populations, evidence has associated personality variations of
an over-regulated/dysregulated type with differences on serotonin-system indices, and specically,
with different loadings of serotonin transporter promoter regulatory region polymorphism (5HTTLPR)
genotypes and alleles. We explored the extent to which an empirical, trait-dened typology of eatingdisordered individuals coincided systematically with variations in 5HTTLPR, assayed using biallelic and
triallelic models.
Method: We tested 185 women with a DSM-IV eating disorder (108 with Bulimia Nervosa, 17 Anorexia
Nervosa, and 60 an Eating Disorder Not Otherwise Specied) and 93 with no eating disorder on measures
reecting psychopathological traits and 5HTTLPR (biallelic and triallelic) genotypes and alleles.
Results: The highest-function, triallelic (LA/LA) genotype occurred signicantly more frequently among
eating-disordered individuals than among controls. However, a more ne-grained analysis suggested that
this association was attributable to the fact that, among eating-disordered participants, those displaying
an Inhibited/Compulsive prole (derived using latent class analysis) were more likely than those of a
Dissocial/Impulsive or a Low Psychopathology group to carry the triallelic 5HTTLPR gain-of-function
LA allele and to be LA/LA homozygotes.
Discussion: This studys empirically derived classes coincide with interpretable differences on genetic
indicesassociating an Inhibited/Compulsive group with 5HTTLPR gain-of-function genotypes (and
alleles) that have elsewhere been linked to trait compulsivity. The ndings, furthermore, suggest that
5HTTLPR, by inuencing personality-trait manifestations may, in turn, inuence eating-disorder risk
and symptom expression.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
Evidence implies that concurrent psychopathological traits
demarcate clinically relevant sub-phenotypes within the eatingdisordered population. For example, factor- or cluster-analytic
studies in the area yield consistent support for the occurrence of
compulsive (over-regulated), impulsive (dysregulated), and
psychologically intact eating-disorder (ED) subgroups (e.g., Westen and Harnden-Fischer, 2001; Steiger and Bruce, 2007; Wonderlich et al., 2005). Anorexic ED variants (especially those of the
restricter variety) tend to occur preferentially within the overregulated subgroup, whereas bulimic variants, although heterogeneously distributed, tend to occur preferentially in the dysregu* Corresponding author. Tel.: +1 514 761 6131x2895; fax: +1 514 888 4085.
E-mail address: (H. Steiger).
0022-3956/$ - see front matter 2009 Elsevier Ltd. All rights reserved.

lated subgroup. In eating- and non-eating-disordered populations

alike, traits of a compulsive or impulsive type have been associated with denable variations in serotonin (5-hydroxytryptamine: 5-HT) system function (Cloninger et al., 1993; Hollander,
1998; Steiger and Bruce, 2007; Steiger, 2004).
1.1. Psychopathological correlates of 5-HT
The 5-HT system regulates mood, social behavior, impulsivity
and eating behavior (Steiger, 2004)creating an obvious rationale
for the hypothesis that 5-HT has a role in ED pathogenesis. In support, studies in eating-disordered individuals document disorderrelevant alterations in 5-HT metabolism, receptor sensitivity and
transporter activity (Steiger, 2004; Frank and Kaye, 2005). Using
single photon emission computer tomography, studies have shown
reduced central 5-HT transporter availability in women with BN

H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

(Tauscher et al., 2001). Likewise, using platelet measures, studies

have suggested altered peripheral 5-HT reuptake in active anorexics and bulimics (Bruce et al., 2006; Steiger et al., 2005a), in binge
purge free former bulimics (Steiger et al., 2005) and even in ED patients unaffected relatives (Steiger et al., 2006). All of the preceding implicate altered 5-HT transporter kinetics in ED pathogenesis.
Consistent with this notion, some candidate-gene studies associate
low-function alleles of the 5-HT transporter promoter polymorphism (5HTTLPR)called low function because they are associated with lower levels of transcription of the transporter protein
with the EDs (Di Bella et al., 2000; Matsushita et al., 2004). Association ndings are, however, inconsistent (cf., Monteleone et al.,
2006; Steiger et al., 2005b).
One basis for inconsistent association may be that 5HTTLPR,
rather than conveying direct risk for ED development, inuences
the expression of behavioral traits that indirectly impact susceptibility to an ED. In non-eating disordered populations, data have
supported the idea that traits of compulsivity and impulsivity
may have different rates of coincidence with 5HTTLPR low- and
high-function alleles and genotypes. The low-function alleles have
been associated with impulsivity (Lesch et al., 1996), novelty seeking (Sander et al., 1998), affective instability and suicidality (Anguelova
dysregulationwhereas the high-function alleles have been
linked to obsessivecompulsive disorder (OCD: Hu et al., 2006;
Baca-Garcia et al., 2005; Bengel et al., 1999) or hyperfrontality
(Heinz et al., 2005)both, arguably, associated with cortical
over-regulation. Suggesting that the same tendency may exist
in an eating-disordered population, recent studies have shown that
bulimic individuals who carry low-function 5HTTLPR alleles are
more likely to display traits of affective instability, impulsivity
(Steiger et al., 2005b), sensation seeking (Steiger et al., 2007), or
harm avoidance (Monteleone et al., 2006).
Despite the preceding, a recent study implicating 178 bulimic
women found no association between 5HTTLPR variations and latent prole analysis-derived personality clusters characterized as
low psychopathology, affective perfectionistic and impulsive
(Wonderlich et al., 2005). We felt a replication to be warranted,
partly in view of ndings (noted earlier) showing positive association between 5HTTLPR variants and personality traits, partly because the studys sample included a disproportionate number of
low-function allele carriers, and partly because the study relied
upon a potentially imprecise biallelic conceptualization of
5HTTLPR (explanation to follow).
1.2. 5HTTLPR: biallelic or triallelic?
The 5HTTLPR polymorphism has (for many years) been conceptualized as being biallelic, with long (L) and short (S) allele
variants thought, respectively, to correspond to relatively high
or low production of 5-HT transporter protein (Lesch et al.,
1996). However, recent data support the existence of a low-frequency L-allele variant, LG (an L allele with A ? G SNP in its
sequence) whose functioning seems to be akin to that of the S
allele (Hu et al., 2006; Zalsman et al., 2006). Such data imply
that 5HTTLPR may need to be conceptualized as being triallelic,
with S and LG alleles being comparable low-function variants,
and an LA allele conferring gain-of function. Recent ndings
have, furthermore, associated the triallelic gain-of-function allele
with obsessivecompulsive disorder (OCD), showing the LA/LA
genotype and the overall frequency of the LA allele to be
substantially increased in individuals with OCD (Hu et al.,
2006). Given an existent literature based on a biallelic 5HTTLPR
assay, and new evidence that the polymorphism may be triallelic, we opted to examine effects of biallelic and triallelic
5HTTLPR formulations.


1.3. The present study

Our rst goal was to develop an empirical classication of eating-disordered individuals based on assessment of variations along
theoretically indicated, comorbid psychopathological traits. Following from previous studies (e.g., Westen and Harnden-Fischer,
2001; Wonderlich et al., 2005), we anticipated nding over-regulated (compulsive), dysregulated (impulsive) and low psychopathology sub-phenotypes. Subsequently, we planned to
examine associations between empirically-derived sub-phenotypes and 5HTTLPR variants. The available literature led us to
anticipate that the compulsive subgroup might show stronger
loadings of high-function (LA in the triallelic model, or L in the
biallelic model) 5HTTLPR alleles than would other eating-disordered or normal-eater groups, whereas the impulsive subgroup
might show stronger loadings of the low-function (LG and/or S)

2. Methods
2.1. Participants
All participants in this institutional ethics-board approved
study differed from those assessed by Wonderlich et al. (2005).
All gave informed consent. Eating-disordered participants were recruited through a specialized Eating Disorders (ED) program in
Montreal, Quebec, Canada. Given that diagnostic heterogeneity
suited our interest in the spectrum of traits found in a broad, eating-disordered population, we included individuals with the DSMIV ED diagnoses Anorexia Nervosa (AN), Bulimia Nervosa (BN) and
Eating-Disorder Not Otherwise Specied (EDNOS). Our eating-disordered sample consisted of 185 women, 99 (53.5%) meeting DSMIV criteria for BN-Purging (BN-P) subtype, 9 (4.9%) for BN-Nonpurging (BN-NP) subtype, 9 (4.9%) for AN Restricting (AN-R) subtype, 8
(4.3%) for AN binge-eating/purging (AN-BP) subtype, 47 (25.4%) for
a bulimia-spectrum EDNOS (EDNOS-BN), and 13 (7.0%) for an anorexia-spectrum EDNOS (EDNOS-AN/R or EDNOS-AN/BP). EDNOS
disorders were dened as follows: subjects (n = 47) with BMI of
18 or more who binged or purged, but at less than the requisite
twice weekly (on average over the past 3 months), were regarded
as having a BN-spectrum EDNOS (EDNOS-BN); individuals (n = 7)
who had lower body weight, and engaged in regular bingeing
and/or purging, but either failed to meet AN criteria due to weight
above BMI of 17.5 or presence of menses were classied as having
an AN Binge/Purge spectrum EDNOS (EDNOS-AN/BP); individuals
(n = 6) who engaged in restriction and/or excessive exercise without binging or purging, but who failed to meet AN criteria due to
weight above BMI of 17.5 or presence of menses were classied
as having an AN Restricting spectrum EDNOS (EDNOS-AN/R).
When interested in comparing anorexic versus bulimic disorders,
we compared AN/R, AN/BP, EDNOS-AN/R and EDNOS-AN/BP
groups to BN and EDNOS/BN groups. When interested in comparing restricters to bingers/purgers, we compared AN/R and EDNOS-AN/R to BN, EDNOS/BN, AN/BP and EDNOS-AN/BP groups. A
major part of the sampling for this study was conducted through
a project concerned with bulimia-spectrum disorders (to which
consecutive, consenting patients with such eating syndromes were
recruited), with patients with Anorexia Nervosa added in an ad hoc
way. Therefore, proportions of cases with different diagnoses
achieve a desired degree of heterogeneity as to ED diagnoses, but
do not reect actual patterns of referral to our program.
We also recruited 93 normal-eater control women, drawn from
an age group comparable to that of our ED sample, and with
recruitment through public media and school-based announcements, so as to produce a group that included comparable propor-


H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

tions of student and non-student participants to those found in our

ED sample. To be eligible for the normal-eater group, participants
had to be free of clinical ED symptoms according to the EDE, of a
history of ED according to initial screening, and to have BMI between 18 and 34. So as not to skew the sample towards super-normalcy, we accepted 8 normal-eaters who emerged, on the
structured interviews described below (see Section 2.2) as having
had an Axis-I disorder within the past 12 months. Disorders detected included major depressive disorder (n = 4), social phobia
(n = 2), post-traumatic-stress-disorder (PTSD: n = 2), and cannabis
dependence (n = 2).
Subjects were between the ages of 17 and 50
(mean = 25.92 7.05
24.43 6.24 for controls). Body Mass Index (BMI: Kg/m2) fell between 12.5 and 34 for ED subjects (mean = 21.39 3.89) and 18
and 34 for control subjects (mean = 22.00 2.55). Predictably,
AN-spectrum (AN-R, AN-BP and EDNOS-AN) cases had a signicantly lower mean BMI (mean = 16.93 + 1.86) than did BN-spectrum (BN-P, BN-NP EDNOS-B) cases (mean = 22.26 + 3.58)
and normal-eater controls (mean = 22.00 + 2.55) [F = 37.87;
df = 2, 275; p = .000]. AN-spectrum, BN-spectrum and control cases
did not, however, differ as to mean Age. Limiting recruitment to
unmedicated individuals was impractical (and undesirable on
grounds of representativeness), and we therefore included 77 ED
women (41.6% of the eating-disordered sample) and 1 normal-eater control who were taking a psychoactive medication when
tested. The pattern of results ruled out the possibility of confounds
associated with medication effects (see Section 3).
The Quebec population (from which this sample was drawn) is
ethnically quite homogeneous. Consequently, our eating-disordered sample included mainly individuals of West-European Caucasian descent (158 individuals, or 85.4% of the sample). There
were, however, an additional 8 (4.3%) of East-European Caucasian
descent, 5 (2.7%) of South-European Caucasian descent, 4 (2.2%)
of Middle-East Caucasian descent, 4 (2.2%) of Latin-American Caucasian descent, 1 (0.5%) of mixed West-European Caucasian/Native
American (aboriginal) descent, 1 (0.5%) of mixed West-European
Caucasian/Caribbean Black descent, 3 (1.6%) of Asian descent, and
1 (0.5%) of mixed West-European Caucasian/ Asian descent. As ethnic distributions did not allow for meaningful comparisons across
ethnic groups, we instead simply conducted ancillary analyses,
repeating main investigations into associations between genotypes
and phenotypes (as described in the results), but excluding those
participants who were not of clearcut, Caucasian descent. Despite
reduced sample size (and power) in these secondary analyses,
the gist of main effects was preserved. We did not have full ethnicity data on all the normal-eater participants, and therefore did no
corresponding subject deletion in this group. However, as main
analyses concerned differences occurring within trait-dened, eating-disordered sub-samples, uncontrolled ethnic variations among
normal-eater controls were unlikely to represent a serious

into four higher-order personality dimensions, based upon previously established factor analyses conducted on data from large
general, personality-disordered and twin samples (Livesley et al.,
1992; Bagge and Trull, 2003). The studies in question support the
validity of the higher-order dimensions Emotional Dysregulation
(onto which load DAPP-BQ subscales measuring Anxiousness,
Identity Problems, Social Avoidance, Affective Lability, Cognitive
Distortion, Oppositionality, Submissiveness, Insecure Attachment,
Suspiciousness and Narcissism), Dissocial Behavior (encompassing
Stimulus Seeking, Conduct Problems, Rejection, and Callousness),
Inhibition (including Intimacy Problems and Restricted Expression)
and Compulsivity (including the Compulsivity subscale alone). To
complement our assessment, we added the Barrat Impulsivity
Scale (BIS, version 11: Patton et al., 1995) and the Centre for Epidemiological Studies Depression (CES-D: Weissman et al., 1977), both
widely-known and validated for the measurement of the intended
Screening for comorbid (past 12 months) DSM-IV Axis-I disorders in control subjects was accomplished using the Structured
Clinical Interview for DSM-IV Axis-I disorders (SCID-I: First et al.,
1996), a computer-guided, interview-based version of the Diagnostic Interview Schedule, Version IV (DIS4: Bucholz et al., 1991), and/
or the Clinician-Administered Post-Traumatic Stress Disorder Scale
(CAPS: Blake et al., 1995) all industry standard measures,
exhibiting excellent reliability, and convergent and discriminant
validity. (Variations in interviews applied reected shifts in study
protocols occurring during the patient recruitment reported here).
Elsewhere, we have evaluated agreement between DIS4 and SCID-I
diagnoses, and obtained excellent Kappas (and percent agreements) for past 12-month presence of Axis-I disorders (Steiger
et al., 2006).
2.3. Genotyping
DNA samples, obtained from whole blood, were amplied by
polymerase chain reaction (PCR) in a total volume of 20 ll, which
contained 100 ng of genomic DNA, 200 lM of dNTPs, 10 pmol each
of the forward and reverse primer, 1 U of Taq DNA Polymerase
(Qiagen, Alameda, CA), 1  PCR buffer, and 1  Q solution (Qiagen).
The forward primer (50 -ATG CCA GCA CCT AAC CCC TAA TGT-30 )
and reverse primer (50 -GG ACC GCA AGG TGG GCG GGA-30 ) were
used to amplify a region encompassing 5HTTLPR; long and short
alleles were then resolved on a 2% agarose gel. The PCR protocol involved preheating the samples at 94 C for 5 min, followed by 35
cycles of denaturation at 94 C (30 s), annealing at 64 C (30 s),
and extension at 72 C (45 s), as well as a nal hold of 5 min at
72 C. The LG and LA alleles were subsequently studied by enzymatic digestion of 7 ll of the above mentioned PCR product using
5 U of MspI and incubating at 37 C for a minimum of 3 h. The LG
and LA alleles were then resolved on a 2% agarose gel.
2.4. Statistical analysis

2.2. Measures
ED diagnoses and symptoms were assessed using the widely
used Eating Disorders Examination (EDE: Fairburn and Cooper,
1993) interview. We also computed Body Mass Index (BMI: Kg/m2).
To achieve a comprehensive, dimensional assessment of personality pathology, while limiting the number of variables implicated, we applied established, higher-order factors obtained from
the Dimensional Assessment of Personality Pathology-Basic Questionnaire (DAPP-BQ: Livesley et al., 1992). The DAPP-BQ is an
empirically derived, self-report measure that systematically describes personality traits using 290 items, organized into 18 trait
subscales. In the present study, DAPP subscales were aggregated

The latent structure of psychopathology in individuals with EDs

was examined using latent class analysis (LCA) applied to scores
for the four higher-order DAPP dimensions, the BIS and the CESD. LCA models associations between observed variables and a
non-observable (latent) variable, aiming to identify the smallest
number of latent classes that adequately describes associations
among dimensions entered. Models were tted by means of an
Expectation Maximization (EM) algorithm, with the program Latent Gold 3.0 (Vermunt and Magidson, 2000). To identify the
best-tting model, we compared successive models by the Bayesian information criterion (BIC), Akaike information criterion
(AIC), and percentage classication error. BIC and AIC information

H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

statistics are global measures that weight the t and parsimony of

the model; lower observed values indicate better t. Latent class
models were applied to successively test 1- to 6-class solutions.
(The upper limit of 6 was used to ensure an adequate number of
observations in each class.) To address the problems of local minima and starting values and to ensure that a true maximum likelihood solution had been reached, we estimated the model 100
times using different random starting values to select the optimal
solution. Cases were assigned to latent classes by modal assignment (i.e., assignment to the latent class for which the a posteriori
probability was the highest). Once formed, LCA-based groups were
compared for ED diagnoses and symptoms using either ANOVAs or
Chi-Squared tests as appropriate. In addition, frequencies of genotypes and alleles were compared across eating-disordered groups
(organized to reect diagnostic and trait differences of interest)
and normal-eater groups using Chi-Squared tests.

3. Results
3.1. Latent class analysis
Criteria indicated a 3 latent-class model to provide best t to
our data (BIC values: 1 class: 4406.33; 2 classes: 4320.04; 3 classes: 4308.96; 4 classes: 4328.17; 5 classes: 4352.35; 6 classes:
4391.41). Analysis of classication errors also supported a 3-class
model. To rule out potentially confounding effects of ED diagnosis
on model estimation, we re-ran the LCA twice, once with a covariate differentiating individuals with AN-spectrum disorders (AN-R,
AN-BP, EDNOS-AN/R or EDNOS-AN/BP) from those with BN-spectrum disorders (BN-P, BN-NP, or EDNOS-B), and a second time with
a covariate differentiating individuals who binged and/or purged
(i.e., with AN-BP, EDNOS-AN/BP, BN-P, BN-NP, or EDNOS-B diagnoses) from those who did not (i.e., AN-R or EDNOS-AN/R diagnoses).
Although the covariate always emerged as a signicant predictor of
classication, both analyses yielded best-tting 3-class solutions
that differed in no substantive way from the original (no-covariate)
analysis. (For brevitys sake, we report results from the original
analysis here.)
Table 1 shows means on the six scale scores entered into the
LCA. Three classes were indicated which, in decreasing order of
group size, encompassed individuals who showed elevated (a) Dissocial Behavior, Impulsivity, Emotional Dysregulation and Depression (n = 80, or 43.2% of the sample), (b) None of the pathological
indices (n = 73, or 39.5%), or (c) Inhibition, Compulsivity, Emotional
Dysregulation and Depression (n = 32, or 17.3%). As both psychopathological groups (a and c) showed heightened affectivity, a
pattern similar to that noted by Wonderlich et al. (2005), we assumed emotional reactivity to be a non-specic characteristic,
and opted to assign labels according to the more-unique characteristics. Correspondingly, we named the three LCA derived groups
Dissocial/Impulsive, Low Psychopathology, and Inhibited/
Compulsive, respectively.
Table 1 also provides mean values for members of the control
group on the six trait/symptom scores, along with results obtained
in univariate ANOVAs that tested for group differences on each variable. Results indicate mean scores of eating-disordered individuals
to have generally been higher (in a pathological direction) than
those of normal-eater controls (signicantly so on Emotional Dysregulation, Inhibition, and Depression.) Dissocial-Impulsive group
members had higher scores than did all other groups on Emotional
Dysregulation, Dissocial Behavior, and Impulsivity, reecting these
individuals uniquely dysregulatory propensities. In contrast,
indicating a unique connection with over-regulation, Inhibited/Compulsive group members scores exceeded those of all
other groups on Inhibition and Compulsivity.


3.2. Clinical features

Table 2 shows results of Chi-Squared tests (and frequency values) comparing the 3 LCA-based classes with respect to ED diagnoses (AN-spectrum disorder, Restricter subtypes, past history of AN)
and medication use. In addition, ANOVAS (and mean values) for
Age, BMI, Binge Days per Month, and Monthly Binge, Vomit and
Purge Frequencies (the latter including episodes of vomiting, or
laxative or diuretic misuse) are provided. Chi-squared tests indicated a difference across classes with respect to AN-spectrum
and Restricter diagnoses, and a trend with respect to history of
AN spectrum disorders, with the Inhibited/Compulsive class always including largest proportions of AN-spectrum (AN-R, AN-BP
and EDNOS-AN) diagnoses, Restricter (AN-R and EDNOS-AN/R)
subtypes, or past AN disorders. A nal chi-squared analysis indicated members of the Low Psychopathology class to receive significantly less psychoactive medication than did members of either
the Dissocial/Impulsive or Inhibited/Compulsive classes. The preceding tendency helps conrm that lower psychopathological
expression in the Low Psychopathology group was not due to confounding effects of medication (i.e., disproportionately higher
medication use) that cancelled symptoms in this group.
One-way ANOVAs indicated no signicant group differences as
to Age. However, in line with the nding of an increased number of
AN cases in the Inhibited/Compulsive class, there was a tendency
for this class to have a lower mean BMI relative to that in other eating-disordered groups. Potential group differences (eating-disordered groups only) as to frequencies of binge and purge
symptoms were examined using ANOVAs. Outliers on variables
recording frequencies of binge, vomit and purge episodes were
transformed to the mean plus two standard deviations and, due
to deviations from normality, square-root transformations were
performed. (Table 2 shows actual, rather than transformed, values
since both analyses revealed similar results.) Again, in line with the
nding of an increased number of AN cases in the Inhibited/Compulsive class, this class reported lowest monthly binge days and
episodes. At a statistical trend level, the Inhibited/Compulsive
group also reported fewer monthly purging episodes than did
other groups.
3.3. Genetic variables
Treating 5HTTLPR in a conventional (biallelic) fashion, frequencies (and percentages) of S/S, S/L, and L/L genotypes, respectively
occurring in 42 (22.7%), 84 (45.4%) and 59 (31.9%) of our eatingdisorder participants and 17 (18.3%), 54 (58.1%) and 22 (23.7%) of
our control participants, were in conformity with Hardy-Weinberg
equilibrium [ED: v21 1:31, n.s.; Control: v21 2:53, n.s.]. With a
triallelic model, frequencies (and percentages) of groups who were
carriers of two, one or no low-function (i.e., S or LG) alleles, 58
(31.4%), 80 (43.2%) and 47 (25.4%), respectively in ED participants,
and 27 (29.3%), 53 (57.6%) and 12 (13.0%), respectively in controls,
were also in conformity with Hardy-Weinberg equilibrium [ED:
v21 3:23, n.s.; Control: v21 3:09, n.s.].
Our next data-analytic step aimed to detect any differences in
genotype or allele frequencies that corresponded to the presence
or absence of an eating disorder. To do so, we rst compared genotype and allele frequencies occurring in the overall group of eatingdisordered individuals to those obtained in normal-eaters (see Table 3). Analyses were set up to examine rates of biallelic genotypes
and alleles (see Table 3a and b), triallelic genotypes and alleles (see
Table 3c and d), and nally triallelic genotypes organized into low
function homozygotes, heterozygotes, or high-function homozygtes, and triallelic allele frequencies when rates of low-function
and high-function alleles were compared (see Table 3e and f). A
signicant difference emerged in the analysis contrasting carriers


H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

Table 1
Mean scores (and standard deviations) for each latent-class derived cluster of eating-disordered patients and for normal-eater controls on Emotional Dysregulation, Dissocial
Behavior, Inhibition, Compulsivity, Impulsivity and Depression scales. Contribution of scale scores to cluster classications is reected by Wald test values (betas are equal among
classes). One-way ANOVAs reect differences between actual group means (including controls). Values with different letters in their superscripts differ at the p < .05 level or
better on Tukeys HSD tests. Variations in ns and dfs reect isolated missing values.
M (SD)
n = 80

Low psycho-pathology
M (SD)
n = 73

M (SD)
n = 32

Wald test p valueA

M (SD)
n = 93

F value,
p value
df = 3, 274
p < .001
40.17 p < .001
56.94 p < .001
27.78 p < .001
p < .001
df = 3, 272
72.10 p < .001

Dissocial Behavior
CES-D depression

3.57a (0.39)

2.56c (0.44)

3.34b (0.34)

161.95 p < 0.001

2.07d (0.48)

2.66a (0.52)
2.93b (0.71)
3.41b (0.81)
34.84a (9.96)

2.07b (0.38)
2.34c (0.57)
3.18b,c (0.53)
21.56b (9.95)

1.95b (0.42)
3.43a (0.59)
4.23a (0.50)
34.47a (13.45)


2.07b (0.36)
2.03d (0.58)
2.97c (0.75)
9.79c (7.63)

Barrat impulsivity

n = 80
78.47a (8.19)

n = 73
67.42b (8.35)

n = 32
58.35c (4.12)

197.80 p < .001

p < .001
p < .001
p < .001
p < .001

n = 91
61.89c (9.55)

Wald test values pertain only to eating-disordered participants, who were included in the latent class analysis.

Table 2
Frequencies of diagnoses (AN-spectrum and Restricter), diagnostic history (history of AN) and medication use, as well as means for Age, BMI and Binge Days per Month, and for
Monthly Binge, Vomit and Purge Frequencies (including episodes of vomiting, and laxative or diuretic misuse) by LCA-based groups. Chi-Square and F values reecting intergroup difference are also reported. Values with different letters in their superscripts differ at the p < .05 level or better on Turkeys HSD tests.

History of AN
Psychiatric medication
Binge days/month
Binge episodes/month
Vomit episodes/month
Purge episodes/month

Dissocial/Impulsive (n = 80)

Low psychopathology (n = 73)

Inhibited/Compulsive (n = 32)

Frequency (%)

Frequency (%)

Frequency (%)

Chi-Squared df = 2

8 (10.0)
2 (2.5)
35 (44.3)
39 (48.8)
Mean (SD)
24.70 (5.36)
21.30 (3.03)
13.69a (9.74)
23.52a (23.48)
34.21 (43.50)
42.03 (46.38)

10 (13.7)
6 (8.2)
25 (35.7)
22 (30.1)
Mean (SD)
26.77 (7.48)
22.06 (4.35)
11.50a,b (9.40)
18.82a,b (23.58)
27.51 (42.87)
31.29 (43.09)

12 (37.5)
7 (21.9)
19 (59.4)
16 (50.0)
Mean (SD)
27.06 (9.21)
20.11 (4.43)
7.41b (9.57)
11.31b (17.41)
18.72 (28.46)
23.48 (28.42)

13.29 p = .001
11.52 p = .003
5.02 p = .081
6.56 p = .038
F value df = 2, 182
2.17 p = .117
2.91 p = .057
4.96 p = .008
3.39 p = .036
1.69 p = .187
2.54 p = .082

of the highest-function triallelic genotype (i.e., LA/LA) to other

genotypes, with eating-disordered participants displaying signicantly higher rates of the LA/LA genotype.
To ensure that we had not obscured potential anorexic-bulimic
differences, we repeated the set of analyses shown in Table 3, but
with eating-disordered participants organized into those displaying anorexia- (i.e., AN/R, AN/BP, EDNOS-AN/R and EDNOS-AN/BP)
or bulimia- (BN and EDNOS/BN) spectrum disorders. These analyses yielded no signicant effects for biallelic or triallelic 5HTTLPR
models, and no within-ED subtype differences. Likewise, a parallel
set of analyses that organized the eating-disordered participants
into those displaying restrictive or bulimic ED variants yielded
no signicant effects on the set of genetic indicators.
Suspecting that the difference observed between eating- and
non-eating-disordered groups may have reected different loadings of psychopathological traits within these groups, our next
data-analytic step explored the correspondence between psychopathological-trait variations, on the one hand, and genotype- or allele-frequency variations, on the other. Table 4 shows analyses
designed to detect differences among frequencies (and percentages) of 5HTTLPR genotypes and alleles (biallelic and triallelic
models) for individuals in each of the LCA, psychopathological-trait
dened groups, as well as for the overall Normal-Eater control
group. The biallelic formulation yielded no signicant results in
either (a) a 3  4 chi-squared test of association between groups
and genotypes, or (b) a 2  4 chi-squared test of associations between groups and allele frequencies (see Table 4a and b). With
the triallelic 5HTTLPR formulation, tests of association between
discreet genotypes and groups (see Table 3c) and between allele
(S, LG and LA) frequencies and groups (see Table 4d) both yielded

Statistical test

statistical trends [v2 = 21.95; df = 15; p = .109, and v2 = 12.25;

df = 6; p = .057, respectively]. More importantly, tests of association between genotypes (organized to reect homozygosity or heterozygosity for low- and high-function alleles) and groups (see
Table 4e) or between low- and high-function allele frequencies
and groups (see Table 4f) both yielded signicant results
[v2 = 15.45; df = 6; p = .017 and v2 = 9.67; df = 3; p = .022, respectively]. A main implication is that members of the Inhibited/Compulsive eating-disordered group tended to be more likely than
were members of other groups to be carriers of one or two copies
of the high-function allele (LA).
More ne-grained analyses, conducted using individual Chisquared tests to localize differences between groups indicated
the following (always in relation to the triallelic 5HTTLPR): there
were signicant pairwise group differences as to genotype frequencies obtained in the Inhibited/Compulsive group versus the
Normal Eaters [v2 = 9.57; df = 2; p = .008] and versus the Dissocial/Impulsive patients [v2 = 7.17; df = 2; p = .028]. The highestexpressing genotype (LA/LA) appeared to occur signicantly more
frequently in Inhibited/ Compulsive patients than it did in the
other two groups noted (see Table 4e for values in question). Odds
ratios indicated Inhibited/ Compulsive patients to be 4 times more
likely (95% CI: 1.5710.22) to have the highest-expressing genotype (LA/LA) than controls. Conversely, Inhibited/Compulsive patients were 3.6 times less likely (95% CI: 1.2610.33) to have the
lowest-expressing genotype (S/S, S/LG or LG/LG) than were Dissocial/Impulsive patients. Reecting the same trend, the absolute frequency of LA alleles carried by Inhibited/Compulsive patients twice
exceeded that carried by Normal Eaters [v2 = 6.95; df = 1; p = .008;
p = .009 on a Fishers exact test] and Dissocial/Impulsive patients

H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

Table 3
Frequencies (and percentages) of 5HTTLPR genotypes and alleles (biallelic and
triallelic models), comparing Eating-Disordered individuals to normal-eater control
Eating disorder
group (n = 185)

Normal-eater control
group (n = 93)

(p value)

5HTTLPR biallelic model

(a) Genotype
Frequency (%)
42 (22.7)
84 (45.4)
59 (31.9)

Frequency (%)
17 (18.3)
54 (58.1)
22 (23.7)

(p = .135)
df = 2

(b) Allele

168 (45.4)
202 (54.6)

88 (47.3)
98 (52.7)

Eating disorder
(n = 185)

Control (n = 92)

(p = .670)
df = 1

5HTTLPR triallelic model

(c) Genotype (S, LG,
and LA alleles)
42 (22.7)
69 (37.3)
15 (8.1)
1 (0.5)
11 (5.9)
47 (25.4)

17 (18.5)
43 (46.7)
10 (10.9)
0 (0.0)
10 (10.9)
12 (13.0)

(p = .099)
df = 5

(d) Allele

168 (45.4)
28 (7.6)
174 (47.0)

87 (47.3)
20 (10.9)
77 (41.8)

(p = .306)
df = 2

(e) Genotype (S/LG,

and LA alleles)

58 (31.4)
80 (43.2)
47 (25.4)

27 (29.3)
53 (57.6)
12 (13.0)

(p = .028)
df = 2

(f) Allele

196 (53.0)
174 (47.0)

107 (58.2)
77 (41.8)

(p = .249)
df = 1

[v2 = 8.07; df = 1; p = .005; p = .005 on a Fishers exact test; odds ratio of 2.17 (95% CI: 1.213.88) and 2.34 (95% CI: 1.294.24), respectively] (see Table 4f for values in question).
4. Discussion
Based on comorbid psychopathological traits, this study classied treatment-seeking eating-disordered patients, and then examined the association of resulting empirical classications with
variations in the serotonin transporter promoter polymorphism
(5HTTLPR). Our results indicated a trio of latent classes, accommodating progressively smaller proportions of our eating-disordered
sample, to which we could t the descriptors Dissocial/ Impulsive, Low Psychopathology and Inhibited/ Compulsive. The
classes derived closely resemble those obtained in comparable
classication effortsone study in an eating-disordered sample
characterizing groups obtained as emotionally dysregulated/
undercontrolled, high-functioning/perfectionistic and constricted/ overcontrolled (Westen and Harnden-Fischer, 2001)
and another, groups described as impulsive, low comorbidity,
and affective/ perfectionistic (Wonderlich et al., 2005). Evident
correspondences across these studies suggest that the tendency
for eating-disordered individuals to cluster into impulsive, relatively intact, and compulsive sub-groups is quite a replicable one.
Making such correspondences all-the-more striking, the studies
in question apply disparate psychopathological indices and statistical techniques.


The main goal of the present study was to explore the association of 5HTTLPR variations with presence of an eating disorder, and
with trait proles that characterize eating-disorder sufferers.
Although we found the highest-function 5HTTLPR genotype
(LA/LA) to be more common in eating-disorder sufferers than in
controls, ner-grained analyses examining associations of genetic
variations with variations in psychopathological traits suggested
that the association observed was not so much characteristic of
our eating-disordered sample (in a wholesale fashion), or of particular diagnostic subgroups (e.g., those with Anorexia-spectrum syndromes), as it was of those members of the sample who were
markedly Inhibited/Compulsive. In other words, our results associate genetic variations more specically with trait variations (such
as the Inhibited/Compulsive versus Dissocial/Impulsive distinction
we derived) than with presence or absence of an ED syndrome per
se. In this respect, our results corroborate other ndings (obtained
in eating- and non-eating-disordered populations) suggesting that
heightened inhibition/compulsivity coincides with the gain-offunction (LA) allele (triallelic model) of 5HTTLPR (Hu et al., 2006),
and heightened impulsivity and affective instability with the
low-function 5HTTLPR alleles (Anguelova et al., 2003; Lesch
et al., 1996; Sander et al., 1998; Steiger et al., 2005b, 2007). Our
study may, furthermore, take a step beyond the simple corroboration of the association noted, in suggesting a correspondence between the polymorphism of interest and phenotypic variations
(or latent classes) that are validated empiricallyour hope being
that this effort may improve the stability (and hence replicability)
of the ndings. Members of an empirically derived Inhibited/
Compulsive subgroup (when compared to those of Dissocial/
Impulsive or Low Psychopathology ED subgroups, or members of
a normal-eater control group) were signicantly more likely to carry at least one copy of the triallelic 5HTTLPR gain-of-function LA allele, and to be high-function homozygotes (i.e., LA/LA genotype
carriers) which for 5HTTLPR represents the highest-expressing
In a similar vein, Inhibited/Compulsive group members were
more likely than other LCA-based groups to exhibit Anorexia
Nervosa (versus a bulimic ED variant). Important clues as to
the interpretation of this nding would seem to lie in the observation that the LA allele is, elsewhere, associated with heightened
risk of obsessivecompulsive disorder (Hu et al., 2006), and that
obsessivecompulsive characteristics predominate in anorexic
(and particularly restrictive-anorexic) ED variants (Westen and
Harnden-Fischer, 2001; Steiger and Bruce, 2007). Together, these
observations suggest that the 5HTTLPR LA allele, while not a specic factor in eating-disorder risk, may exercise a pathoplastic
effectheightening the likelihood of expression of obsessive
compulsiveness and pronounced dietary restraint in eatingdisordered individuals. In other words, we may be observing a
shaping impact of a genetic factor (in this case, of the 5HTTLPR
high-function alleles) upon expression of traits (in the Inhibited/Compulsive versus Dissocial/Impulsive spectrum) which, in
eating-disorder sufferers shapes eating-symptom expression
(i.e., anorexic versus bulimic forms). In addition, our results support the notion that Anorexia Nervosa may not only resemble
obsessivecompulsive disorder in a phenomenological sense
(bodily obsessions motivating dieting compulsions), but that
these disorders may also have at least one common, moleculargenetic determinant (i.e., the 5HTTLPR LA allele).
On the opposite side of the same coin, we note that members
of our Dissocial/Impulsive subgroup were signicantly more likely
than members of the Inhibited/Compulsive group to carry lowfunction genotypes and alleles. The preceding corroborates parallel
ndings obtained in eating-disordered (Steiger et al., 2005b,
2007) and non-eating-disordered populations (Lesch et al., 1996;


H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

Table 4
Frequencies (and percentages) of 5HTTLPR genotypes and alleles (biallelic and triallelic) across LCA-dened eating-disordered groups and normal-eater controls.
eating disorder (n = 80)
Frequency (%)

Low comorbidity
eating disorder (n = 73)
Frequency (%)

eating disorder (n = 32)
Frequency (%)

Normal-eater control
group (n = 93)
Frequency (%)

5HTTLPR biallelic model

(a) Genotype

21 (26.3)
37 (46.3)
22 (27.5)

18 (24.7)
31 (42.5)
24 (32.9)

3 (9.4)
16 (50.0)
13 (40.6)

17 (18.3)
54 (58.1)
22 (23.7)

8.96 p = .176 df = 6

(b) Allele

79 (49.4)
81 (50.6)

67 (45.9)
79 (54.1)

22 (34.4)
42 (65.6)

88 (47.3)
98 (52.7)

4.34 p = .227 df=3

Chi-Square (p value)

Dissocial/Impulsive (n = 80)

Low comorbidity (n = 73)

Inhibited/Compulsive (n = 32)

Controls (n = 92)

5HTTLPR triallelic model

(c) Genotype (S, LG, and LA alleles)

21 (26.3)
27 (33.8)
10 (12.5)
1 (1.3)
5 (6.3)
16 (20.0)

18 (24.7)
28 (38.4)
3 (4.1)
0 (0.0)
5 (6.8)
19 (26.0)

3 (9.4)
14 (43.8)
2 (6.3)
0 (0.0)
1 (3.1)
12 (37.5)

17 (18.5)
43 (46.7)
10 (10.9)
0 (0.0)
10 (10.9)
12 (13.0)

21.95 p = .109 df = 15

(d) Allele

79 (49.4)
17 (10.6)
64 (40.0)

67 (45.9)
8 (5.5)
71 (48.6)

22 (34.4)
3 (4.7)
39 (60.9)

87 (47.3)
20 (10.9)
77 (41.8)

12.25 p = .057 df = 6

(e) Genotype (S/LG, and LA alleles)


32 (40.0)
32 (40.0)
16 (20.0)

21 (28.8)
33 (45.2)
19 (26.0)

5 (15.6)
15 (46.9)
12 (37.5)

27 (29.3)
53 (57.6)
12 (13.0)

15.45 p = .017 df = 6

(f) Allele

96 (60.0)
64 (40.0)

75 (51.4)
71 (48.6)

25 (39.1)
39 (60.9)

107 (58.2)
77 (41.8)

9.67 p = .022 df = 3

Anguelova et al., 2003) that link low-function 5HTTLPR alleles with

traits of impulsivity or dysregulation.
How should such ndings be interpreted? Evidence of lack of
association of the triallelic 5HTTLPR with PET-determined transporter binding in normal humans has led to the proposal that associations of 5HTTLPR with clinical phenotypes may depend, not
upon direct (momentary) effects on serotonin binding, but effects
of 5HTTLPR upon brain development (Parsey et al., 2006). Following from this idea, we speculate that it is not opportune to interpret 5HTTLPR variants as having any simple or direct connotation
for high versus low levels of intrasynaptic 5-HT. Rather, presence of high- or low-function 5HTTLPR alleles may have implications for the adaptability, resilience, or other global
propensities of an individuals 5-HT systemand may therefore
have relevance for the likelihood of specic psychopathological
expressions when the adaptive capacities of this system are overtaxed. Available ndings implying that 5HTTLPR variants may
inuence psychopathological responses to environmental stressors
in ED patients (see Steiger et al., 2007, 2008) are consistent with
this type of explanation.
A nal comment is necessary, concerning the meaning of the
Inhibited/Compulsive construct we have applied. This higher-order DAPP factor encompasses lower-order traits of Restricted
Expression (emotional constriction or guardedness), Compulsivity
(traits such as orderliness, conscientiousness, and hypervigilance),
and Intimacy Problems (insecure or avoidant attachments). The
inhibited/ compulsive construct thus appears to more closely
represent obsessivecompulsive personality disorder (OCPD) than
it does OCD. That previous work has linked the high-function
5HTTLPR allele to OCD, whereas our study suggests a link to a construct more closely approximating OCPD, suggests that there may
be a need, in future work, to dene the implications of 5HTTLPR
high-function alleles and genotypes for the expression of OCD,

OCPD, and other phenotypes in a putative obsessivecompulsive

spectrum (e.g., Hollander, 2007). Awaiting this clarication, we
note that OCPD traits and syndromic OCD are both common in eating-disordered patients (Westen and Harnden-Fischer, 2001; Steiger and Bruce, 2007).
4.1. Limitations
Our sample is relatively small for a multivariate exploration,
and this may limit stability of ndings and power to detect certain
effects of potential interest. Furthermore, in this study, we have
not addressed geneenvironment interaction effects that may be
relevant to the understanding of associations between genetic factors, on the one hand, and psychopathology-dened classications,
on the other. Suggesting that such effects may be important, our
group has recently documented several instances of gene-environment interaction effects linking 5HTTLPR and childhood abuse to
such traits as novelty seeking (Steiger et al., 2007) or Dissocial
Behavior (Steiger et al., 2008). In addition, we reiterate a point
made earlier that, due to recruitment factors, our sample cannot
be assumed to be representative of the population seeking treatment in specialized eating-disorders programs. Nonetheless, our
samples heterogeneity allowed for exploration of possible relationships between genetic variables, on the one hand, and personality and eating-symptom variables, on the other.
4.2. Conclusions
Our ndings provide evidence of association amongst traitbased, eating-disorder sub-phenotypes (inhibited/compulsive, dissocial/impulsive, or low psychopathology) and variations in
5HTTLPR genotypes- and less so, with categorical ED phenotypes
(e.g., Anorexia- versus Bulimia-spectrum syndromes). We believe

H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

that this pattern of ndings may guide the understanding of the

relationships between genotypic variations and complex phenotypes, like eating disorders and their subtypes. A lesson to be derived from the present ndings may be that the correct loci at
which to seek correspondences between genotypes and phenotypes may be those linking genetic effects to higher-order, traitbased phenotypes (e.g., Inhibited/Compulsive) rather than to syndromic variations (e.g., Anorexia Nervosa). Findings such as the
present ones suggest, for example, that 5HTTLPR (and presumably
other polymorphisms in the 5-HT system) may inuence psychopathological and behavioral traits, and in so doing, may indirectly
channel vulnerable individuals thinness and body-image preoccupations into either restrictive or bulimic expressions. This interpretation is consistent with general ndings on serotonin indices in
eating-disordered individuals suggesting that serotonin-system
variations may be more powerful correlates of generalized psychopathological-trait variations seen in ED sufferers than they are of
eating-symptom variations and severities per se (see Steiger,
2004; Steiger and Bruce, 2007).
Howard Steiger was Principal Investigator (PI) on the grants
that supported this research and on the research itself, and oversaw all aspects of this work. Jodie Richardson assisted in the
studys conceptualization and design, and conducted main data
analyses. With Steiger, she helped co-write the rst draft of this
manuscript. Norbert Schmitz contributed to statistical aspects of
the report, carried out data analyses, and offered particular guidance on latent class analyses. Ridha Joober was responsible for genetic assays and contributed to the conceptualization and design of
the study. Mimi Israel contributed to the conceptualization and design of the study, and supervised clinical assessments and medical
acts. Kenneth Bruce contributed to conceptualization and design of
the study, to data analyses, and to training and supervision of research assistants. Lise Gauvin contributed to the conceptualization
and design of the study, and to the design and execution of data
analyses. Cathy Dandurand and Annelie Anestin contributed to
participant recruitment, testing, data entry and management, literature searches, and design and execution of certain data analyses.
All authors have contributed to different sections of draft manuscripts, and all have approved the nal manuscript.
Role of funding sources
Funding agencies had no further role in study design; in the collection, analysis and interpretation of data; in the writing of the report, or in the decision to submit the paper for publication.
Preliminary results from this study were presented at the annual
meeting of the Eating Disorders Research Society, Montreal, Quebec, Canada, Sept 26, 2008.
Conict of interest statement
Dr. Joober has received honoraria for consultations from Janssen
Ortho and Pzer Canada.
This research was supported by grants awarded to Dr. Steiger
and colleagues (No. 6313A from the Fonds de la Recherche en Sant
du Qubec, and nos. MOP-79490 and MOP-57929 from the Canadian Institutes for Health Research), and by a doctoral bursary
awarded to Jodie Richardson from the Canadian Institutes for
Health Research. Funding agencies had no further role in study de-


sign; in the collection, analysis and interpretation of data; in the

writing of the report, or in the decision to submit the paper for
publication. Preliminary results from this study were presented
at the annual meeting of the Eating Disorders Research Society,
Montreal, Quebec, Canada, Sept. 26, 2008. Authors are grateful to
Patricia Groleau, Catherine Villenneuve-Tang and Catherine Senecal for their assistance in various aspects of this research.
Anguelova M, Benkelfat C, Turecki G. A systematic review of association studies
investigating genes coding for serotonin receptors and the serotonin
transporter II: suicidal behavior. Molecular Psychiatry 2003;8:64653.
Baca-Garcia E, Rodriguez Aalgado B, Dolengvich Segal H, Vaquero Lorenzo C, Navio
Acosta M, Arrojo Romero M, et al. A pilot genetic study of the continuum
between compulsivity and impulsivity in females: the serotonin transporter
promoter polymorphism. Progress in Neuropsychopharmacology and Biological
psychiatry 2005;29:7137.
Bagge C, Trull T. DAPP-BQ: factor structure and relations to personality disorder
symptoms in a non-clinical sample. Journal of Personality Disorders
Bengel D, Greenberg BD, Cora-Locatelli G, Altemus M, Heils A, Li Q, et al. Association
of the serotonin transporter promoter regulatory regions polymorphism and
obsessivecompulsive disorder. Molecular Psychiatry 1999;4:4636.
Blake DD, Weathers FW, Nagy LM, et al. The development of a clinicianadministered PTSD scale. Journal of Traumatic Stress 1995;8:7590.
Bruce KR, Steiger H, Ng Ying Kin NMK, Israel M. Reduced platelet [3H] paroxetine
binding in anorexia nervosa: relationship to eating symptoms and personality
pathology. Psychiatry Research 2006;142:22532.
Bucholz KK, Robins LN, Shayka JJ, Przybeck TR, Heltzer JE, Goldring E, et al.
Performance of two forms of a computer psychiatric screening interview:
version 1 of the DISSI. Journal of Psychiatry Research 1991;25:11729.
Cloninger CR, Svrakic DM, Przybeck TR. A psychobiological model of temperament
and character. Archives of General Psychiatry 1993;50:97590.
Di Bella DD, Catalano M, Cavallini MC, Riboldi C, Bellodi L. Serotonin transporter
linked polymorphic region in anorexia nervosa and bulimia nervosa. Molecular
Psychiatry 2000;5:23341.
Fairburn C, Cooper P. The eating disorders examination. In: Fairburn C, Wilson G,
editors. Binge eating: nature. Assessment and treatment. New York: Guilford;
1993. p. 31760.
First MB, Spitzer RL, Gibbon M, Williams J. Structured Clinical Interview for DSM-IV
Axis I Disorders Patient Edition (SCID-I/P, Version 2.0). New York: Biometrics
Research Department, New York State Psychiatric Institute; 1996.
Frank GK, Kaye WH. Positron emission tomography studies in eating disorders:
multireceptor brain imaging, correlates with behavior and implications for
pharmacotherapy. Nuclear Medicine and Biology 2005;32:75561.
Heinz A, Braus DF, Smolka MN, Wrase J, Puls I, Hermann D, et al. Amygdalaprefrontal coupling depends on a genetic variation of the serotonin
treansporter. Nature Neuroscience 2005;8:201.
Hollander E. Treatment of obsessivecompulsive spectrum disorders with SSRIs.
British Journal of Psychiatry 1998;35(Suppl.):712.
Hollander E. Anxiety and OC spectrum disorders over life cycle. International
Journal of Psychiatry in Clinical Practice 2007;11(Suppl.):510.
Hu XZ, Lipsky RH, Zhu G. Serotonin transporter promoter gain of function genotypes
are linked to obsessivecompulsive disorder. American Journal of Human
Genetics 2006;78:81526.
Lesch KP, Bengel D, Heils A, Sabol SZ, Greenberg BD, Petri S. Association of anxietyrelated traits with a polymorphism in the serotonin transporter gene regulatory
region. Science 1996;274:152730.
Livesley WJ, Jackson DN, Schroeder ML. Factorial structure of traits delineating
personality disorders in clinical and general population samples. Journal of
Abnormal Psychology 1992;101:43240.
Matsushita S, Suzuki K, Murayama M, Nishiguchi N, Hishimoto A, Takeda A, et al.
Serotonin transporter regulatory region polymorphism is associated with
anorexia nervosa. American Journal of Medical Genetics (Neuropsychiatric
Genetics) 2004;128:1147.
Monteleone P, Santonastaso P, Mauri M, Nellodi L, Erzegovesi S, Fuscino A, et al.
Investigation of the serotonin transporter regulatory region polymorphism in
bulimia nervosa: relationships to harm avoidance, nutritional parameters and
psychiatric comorbidity. Psychosomatic Medicine 2006;68:99103.
Parsey RV, Hastings RS, Oquendo MA, Hu X, Goldman D, Huang Y-y, et al. Effect of a
triallelic functional polymorphism of the serotonin-transporter linked promoter
region on expression of serotonin transporter in the human brain. American
Journal of Psychiatry 2006;163:4851.
Patton JH, Stanford MS, Barrat E. Factor structure of the Barrat impulsiveness scale.
Journal of Clinical Psychology 1995;51:609774.
Sander T, Harms H, Dufeu P, Kuhn S, Hoehe M, Lesch KP, et al. Serotonin transporter
gene variants in alcohol-dependent subjects with dissocial personality disorder.
Biological Psychiatry 1998;43:90812.
Steiger H. Eating disorders and the serotonin connection: state, trait and
developmental effects. Journal of Psychiatry and Neuroscience 2004;29:209.
Steiger H, Bruce KR. Phenotypes, endophenotypes, and genotypes in bulimiaspectrum eating disorders. Canadian Journal of Psychiatry 2007;52:2207.


H. Steiger et al. / Journal of Psychiatric Research 43 (2009) 10861094

Steiger H, Gauvin L, Joober R, Israel M, Ng Ying Kin NMK, Bruce KR, et al.
Intrafamilial correspondences on platelet [3H-] paroxetine-binding indices in
Neuropsychopharmacology 2006;31:178592.
Steiger H, Joober R, Isral M, Young SN, Ng Ying Kin NMK, Gauvin L, et al. The
5HTTLPR polymorphism, psychopathological symptoms, and platelet [3H-]
paroxetine binding in bulimic syndromes. International Journal of Eating
Disorders 2005b;37:5760.
Steiger H, Richardson J, Israel M, Ng Ying Kin NMK, Bruce K, Mansour S, et al.
Reduced density of platelet-binding sites for 3H-paroxetine in remitted bulimic
women. Neuropsychopharmacology 2005a;30:102832.
Steiger H, Richardson J, Joober R, Gauvin L, Israel M, Bruce KR, et al. The 5HTTLPR
polymorphism, prior maltreatment, and dramaticerratic personality
manifestations in women with bulimic syndromes. Journal of Psychiatry and
Neuroscience 2007;32:35462.
Steiger H, Richardson J, Joober R, Israel M, Bruce KR, Ng Ying Kin NMK, et al.
Dissocial behavior, the 5HTTLPR polymorphism and maltreatment in women
with bulimic syndromes. American Journal of Medical Genetics Part B

Tauscher J, Pirker W, Willeit M, de Zwaan M, Bailer U, Neumeister A, et al. Beta-CIT

and single photon emission computer tomography reveal reduced brain
serotonin transporter availability in bulimia nervosa. Biological Psychiatry
Vermunt JK, Magidson J. Latent GOLD 2.0 users guide. Massachusetts; 2000.
Weissman MM, Sholomskas D, Pottenger M, Prusoff BA, Locke BZ. Assessing
depressive symptoms in ve psychiatric populations: a validation study.
American Journal of Epidemiology 1977;106:20314.
Westen D, Harnden-Fischer J. Personality proles in eating disorders: rethinking the
distinction between axis I and axis II. American Journal of Psychiatry
Wonderlich SA, Crosby RD, Joiner T, Peterson CB, Bardone-Cone A, Klein M, et al.
Personality subtyping and bulimia nervosa: psychopathological and genetic
correlates. Psychological Medicine 2005;35:64957.
Zalsman G, Huang Y-y, Oquendo MA, Burke AK, Hu X-z, Brent DA, et al. Association
of a triallelic serotonin transporter gene promoter region (5-HTTLPR)
polymorphism with stressful life events and severity of depression. American
Journal of Psychiatry 2006;163:158893.