[Skip to Navigation]
Sign In
November 3, 2008

Childhood Trajectories of Anxiousness and Disruptiveness as Predictors of Suicide Attempts

Author Affiliations

Author Affiliations: McGill Group for Suicide Studies, Douglas Hospital Research Center (Drs Brezo and Turecki), Research Unit on Children's Psychosocial Maladjustment, University of Montreal (Drs Barker, Vitaro, Tremblay, and Turecki), Department of Psychiatry, McGill University (Drs Paris), and Department of Sexology, University of Quebec (Dr Hébert), Montreal, Quebec, Canada; and King's College London, Institute of Psychiatry, MRC Social, Genetic and Developmental Psychiatry Centre, London, England (Dr Barker).

Arch Pediatr Adolesc Med. 2008;162(11):1015-1021. doi:10.1001/archpedi.162.11.1015

Objective  To investigate the association of childhood trajectories of anxiousness and disruptiveness with suicide attempts in early adulthood.

Design  Prospective cohort study.

Setting  Public francophone schools in Quebec, Canada, from the 1986 to 1988 school years.

Participants  Of 4488 French Canadian children attending kindergarten, a representative group of 1001 boys and 999 girls was chosen for follow-up. Of these, 1144 individuals participated in the study during early adulthood.

Main Outcome Measures  Suicide attempt histories by early adulthood, adjusted odds ratios (ORs) associated with membership in high- vs low-risk trajectories of anxiousness and disruptiveness, moderation (by sex), and mediation (by adolescent Axis I disorders).

Results  We observed 4 distinct developmental profiles of anxiousness and disruptiveness and a frequent co-occurrence of similar levels of these traits. In contrast to anxiousness trajectories (OR = 1.60; 95% confidence interval, 1.00-2.65), disruptiveness (OR = 1.80; 95% confidence interval, 1.03-3.13) and joint (OR = 1.88; 95% confidence interval, 1.05-3.37) trajectories made statistically significant contributions to suicide attempts. We found no support for mediation by adolescent anxiety/mood or disruptive disorders. Sex, however, moderated the effect of joint trajectories, increasing the risk of suicide attempts in women (OR = 3.60; Wald χ2 = 10.93; P < .001) but not men (OR = 0.80; Wald χ2 = 0.23; P = .64) displaying both anxious and disruptive traits as children.

Conclusions  Anxious-disruptive girls and disruptive boys appear to be more likely than their peers to attempt suicide by early adulthood. Preventive efforts will require more research into the possible mechanisms behind this early sex difference, ie, gene-environment interplays and nonpsychiatric mediators.

Evidence suggests that personality traits and their behavioral, emotional, and cognitive childhood antecedents may be involved in predisposition to psychiatric phenotypes, including suicidality.1 Understanding their contribution would be valuable in identifying vulnerable subgroups, preventing suicidal behaviors, or customizing therapy and predicting its success in suicidal individuals.2,3

Research into attempted suicide, an important predictor of completed suicide,2 has highlighted the importance of emotional dysregulation (eg, traits such as anxiousness4 and neuroticism5) and behavioral dysregulation (traits such as impulsive aggression).6-9 While useful, this empirical evidence has mostly been derived from cross-sectional observations relying on one-point snapshots rather than developmental trajectories. This has precluded elucidation of temporal relationships, predictive values, and age-dependent interactions of such traits. Although limited, relevant evidence suggests that internalizing and externalizing tendencies do interact.10,11 The question of preventive and therapeutic relevance—when might the dysregulation in both of these aspects of personality confer higher risk for suicidality—remains unanswered.

Methodological issues have also made it difficult to examine mediating mechanisms linking personality traits and their behavioral markers with suicidality. It is, for example, unclear whether their effects may be independent of or mediated by psychiatric disorders as has been suggested.12,13 Finally, moderating influences are suspected but rarely tested in a systematic manner. This knowledge is essential for identification of groups at particularly high risk and in need of timely intervention.

To our knowledge, our study is the first to address these methodological limitations in the context of suicide attempts. Using a prospective, school-based cohort, we investigated childhood trajectories of anxiousness and disruptiveness as possible risk factors for lifetime suicide attempts. Drawing from previous findings,11 we hypothesized that higher levels of childhood anxiousness, disruptiveness, or both have stronger positive associations with suicide attempts than their lower levels. To prevent confounding, we controlled for factors previously linked to both personality and suicidality: adolescent and adult anxiety and mood disorder diagnoses,1,14,15 substance abuse,16 family history of suicide attempts,9 childhood sexual abuse,17 and disruptive disorders.18 Because relevant research suggests not only a substantial degree of stability between early and later markers of personality19 but also some differences,20,21 we controlled for adult anxiety and conduct problems, traits closely related to our childhood personality measures.1

Our final 2 objectives are in line with calls for a more sophisticated approach to the study of psychiatric phenotypes. First, because of its well-documented involvement with anxious and aggressive personality traits22 and suicide attempts,23 we examined whether sex moderated the strength or direction of the association between suicide attempts and developmental trajectories.23 Second, we examined psychiatric Axis I disorders in the role of mediators,14,24 variables that can elucidate underlying etiological pathways. This hypothesis was based on the evidence suggesting a “personological” component in psychiatric diathesis1: negative emotionality and inhibited temperament, for example, predict internalizing disorders,1,11,15,24 whereas early disruptiveness, undercontrolled temperament, and hyperactivity increase the risk for conduct, antisocial, and substance abuse disorders.1


In the 1986 to 1988 school years, families of 4488 children attending kindergarten in francophone schools in Quebec, Canada, were recruited using a multistage sampling procedure. Of these, a representative, randomly selected group of 1001 boys and 999 girls was followed up to adulthood. To reduce cultural heterogeneity, only children with parents born in Canada and with French as their native language were included. Eighty-nine percent were non-Hispanic white individuals.

The assessment schedule had 3 stages: wave 1 (childhood; yearly assessments at ages 6-12 years; n = 2000); wave 2 (midadolescence; mean age, 15.7 years; age range, 15-18 years; n = 1233); and wave 3 (adulthood; mean age, 21.4 years; age range, 19-24 years; n = 1144). Participants who died, refused participation, or could not be contacted accounted for an overall attrition percentage of 43%. We used 2 variables related to attrition as weights in the analyses: early socioeconomic adversity (this is a composite index consisting of parental age at first child's birth, education, economic status, and living arrangements, scored on a continuous scale from 0-1 with higher scores representing higher adversity) and sex. Mean early socioeconomic adversity was higher in nonresponders (0.32) than responders (0.25). Males represented 50% of nonresponders and 35% of responders.

The study was approved by the research ethics boards of the University of Montreal and McGill University, Montreal, Quebec. Written informed consent was obtained from all of the subjects.


Childhood Risk Factors

The Social Behavior Questionnaire25 assesses several childhood traits using teacher reports. Because teacher raters differed each year, the yearly assessments from ages 6 through 12 years were independent. We selected 2 traits: disruptiveness and anxiousness. Disruptiveness (Cronbach α = .90), encompassing hyperactive, aggressive, antisocial, and oppositional traits, was based on 13 items: is agitated, always running and jumping, or restless; destroys one's own things or those of others; fights with other children; is not liked by peers; is irritable; is disobedient; lies; mistreats or intimidates peers; does not share material used for a particular task; blames others; is inconsiderate of others; hits and kicks others; and fidgets, squirms, or cannot keep still. Anxiousness (Cronbach α = .74) was assessed with 6 items: is fearful or afraid of things or new situations; is worried or worries about many things; cries easily; has a tendency to work alone; looks sad, unhappy, or tearful; and is easily distracted.

Covariates and Mediators

Adolescent mood, anxiety, and disruptiveness diagnoses were considered as covariates and mediators and were assessed with the Diagnostic Interview Schedule for Children version 226 using Diagnostic and Statistical Manual of Mental Disorders (Third Edition Revised)27 criteria. We tested collective effects of externalizing Axis I disorders (attention-deficit/hyperactivity, oppositional-defiant, and conduct disorders) on the one hand and the internalizing ones on the other (mood disorders: major depression and dysthymia; anxiety disorders: simple and social phobias, separation anxiety, panic, avoidant, overanxious, and generalized anxiety disorder). Interviewers were psychology students who attended training for 1 to 1.5 days and a practice session.

The Diagnostic Assessment of Personality Pathology28 measures 18 narrow personality traits, each assessed with 12 to 16 questions inquiring about personal preferences and behaviors. We selected anxiousness (Cronbach α = .92) and conduct problems (Cronbach α = .94) as covariates.

The Diagnostic Interview Schedule for Adults using Diagnostic and Statistical Manual of Mental Disorders (Third Edition Revised) criteria29 allowed us to adjust our models for collective effects of internalizing Axis I disorders (anxiety disorders: generalized anxiety, panic, and phobias; mood disorders: major depression, dysthymia, and bipolar disorder) and externalizing Axis I disorders (abuse of and/or dependence on drugs, alcohol, and nicotine). Interviewers were psychology students who attended training for 1 to 1.5 days and a practice session.


Lifetime suicide attempts were assessed using both adolescent and adult reports. Adolescent suicide attempts were obtained from parental or adolescent responses to a Diagnostic Interview Schedule for Children version 2 question: “Have you already attempted suicide?” Either parental report or self-report was sufficient for a person to be classified as an attempter. Adult suicide attempts were ascertained with 1 question: “Have you already attempted suicide?”

Data analysis

Step 1: Identifying Trajectories of Anxiousness and Disruptiveness

We identified the developmental trajectories—“clusters of individuals following similar progressions of some behavior or outcome over age or time”30—with semiparametric group-based modeling, a type of growth-mixture modeling. Semiparametric group-based modeling assumes that the population is composed of a mixture of groups of youth following distinct developmental trajectories30-32 described by both the shape (low, increasing) and estimated proportions of individuals following them. Semiparametric group-based modeling can accommodate different types of data distribution by way of censored normal, Poisson, 0-inflated Poisson, and Bernoulli algorithms. Polynomial functions (ie, intercepts, slopes, and quadratic trends) model the link between age and the outcome. The Bayesian Information Criterion was used to select the most optimal from a series of models involving different numbers of trajectory groups.31 The semiparametric group-based modeling approach offers several advantages: unlike correlation-based procedures, it is less sensitive to outlier data (ie, it can accommodate nonnormal distributions); it can handle missing data through maximum likelihood estimation without losing information as would happen with listwise deletion; by allowing model parameters to differ across groups, it allows for population heterogeneity at the level of the individual at a given time and over age; and for each individual, it outputs posterior probabilities (probabilities of following trajectory subgroups) that are used as weights to account for membership uncertainty.

A generalization of semiparametric group-based modeling is the joint trajectory method.33 This routine links trajectories of 2 related but distinct outcomes into joint trajectories, allowing examination of the coevolution of 2 behaviors. We used the best-fitting trajectory models for anxiousness and disruptiveness as the starting point for the joint models whose key outputs are the conditional and joint probabilities of following given trajectories of anxiousness and disruptiveness (eg, the probability of following high trajectories of both). The analyses were performed using the SAS-TRAJ procedure (SAS Institute Inc, Cary, North Carolina).30-32,34

Step 2: Linking Trajectories to Suicide Attempts

We differentiated between high-risk (moderate or high level) and low-risk (low and very low level) trajectories of anxiousness and disruptiveness (Figure 1 and Figure 2). Their relationship with possible confounders was examined in a series of univariate (χ2 test) and multivariate (logistic regression) models. We assessed sizes and significance of odds ratios (ORs) associated with anxiousness and disruptiveness considered both independently of each other and then jointly, ie, their joint trajectories. Analyses were weighted by the inverse of each individual's probability of being in the original sample, conditional on the 2 variables related to attrition: sex and early socioeconomic adversity.

Figure 1. 
Anxiousness trajectories.

Anxiousness trajectories.

Figure 2. 
Disruptiveness trajectories.

Disruptiveness trajectories.

Sex was also tested as a moderator of the relationship between trajectories and suicide attempts.35 Briefly, using hierarchical multiple regression, we entered main effects for trajectories, sex, and their interaction. On finding a significant moderating effect, we conducted post hoc tests to quantify regression slopes and examine their statistical significance according to Wald χ2 test separately in women and men.35

Disruptive and adolescent anxiety/mood disorders were tested as mediators of the effects of anxiousness and disruptiveness trajectories on suicide attempts.24 Mediators are variables that account for a portion or all of the association between a predictor (P) and an outcome (O). Mediation is also operationalized as a mechanism through which P influences O.24 Mediation testing consists of 4 regression steps24 needed to demonstrate associations between the following: (1) P (trajectories) and O (suicide attempts); (2) P and mediators Me1 (ie, adolescent anxiety/mood disorders as mediators of the effect of anxiousness trajectories on suicide attempts) and Me2 (ie, disruptive disorders as mediators of the effect of disruptiveness trajectories on suicide attempts); (3) Me and O; and (4) P and O while controlling for Me. Where there is a decrease in the total effect when the mediator is controlled for, suggesting existence of an indirect, mediating effect, Sobel and Goodman tests are used to establish its statistical significance.36 These tests require unstandardized path coefficients and their standard errors.


Identifying trajectories

Trajectories were available for 1869 individuals who had at least 3 childhood data points. The Bayesian Information Criterion suggested the 4-group trajectory solution as the most optimal for both anxiousness and disruptiveness (Figure 1 and Figure 2). Of the 4 trajectories of disruptiveness, the most frequent one was the very low trajectory, representing 40% of the sample, in contrast to the high trajectory followed by only 10% (Figure 2). Moderate (20%) and low (31%) disruptiveness had a more stable character than their anxiousness equivalents, representing 27% and 30% of the sample, respectively (Figure 2). The most commonly followed trajectory of anxiousness was the low trajectory, comprising individuals whose anxiousness declined from moderate to low levels (30%) (Figure 1).

The joint trajectories procedure allowed us to examine the co-occurrence of the 2 traits by estimating proportions of individuals following different combinations of the trajectories of anxiousness and disruptiveness (Figure 3). About one-third of the sample had low or very low levels of both traits, whereas only 5% displayed the highest levels of both. In general, participants were most likely to follow equivalent (prevalence range, 6%-14%) or similar (very low and low or high and moderate; prevalence range, 4%-15%) levels of both traits. Dissimilar anxiousness and disruptiveness levels, eg, very low or low with high or moderate trajectories, were least frequent, occurring in less than 4% of participants.

Figure 3. 
Dual trajectories.

Dual trajectories.

Descriptive statistics

High-risk (moderate- or high-level) trajectories of anxiousness, disruptiveness, or both were associated with higher mean socioeconomic adversity (P < .001) in childhood than their low-risk counterparts: 0.30 vs 0.24, 0.32 vs 0.24, and 0.33 vs 0.25, respectively. All of the 3 high-risk groups also had higher prevalence of suicide attempts relative to low-risk groups (Table 1). Furthermore, males, family history of suicide attempts, and substance dependence or abuse disorders were overrepresented in high-risk joint and disruptiveness trajectories (Table 1). Individuals who were highly disruptive as children had a lower prevalence of adolescent mood and anxiety disorders and a higher prevalence of disruptive disorders (Table 1). Childhood sexual abuse was more common in the high-risk joint trajectories than in the low-risk ones (Table 1).

Table 1. 
Univariate Differences Between High- and Low-Risk Trajectoriesa
Univariate Differences Between High- and Low-Risk Trajectoriesa

Multivariate models

Trajectories of Anxiousness

Relative to their low-risk counterparts, high-risk anxiousness trajectories were associated with a 60% increase in the likelihood of attempted suicide, exhibiting a statistically significant trend in the presence of covariates. As for the latter, adolescent anxiety and mood disorder diagnoses (OR = 1.41), family history of suicide attempts (OR = 2.50), and experiences of childhood sexual abuse (OR = 1.37) made statistically significant contributions (Table 2).

Table 2. 
Adjusted Models for High-Risk Anxiousness and Disruptiveness as Risk Factors for Lifetime Suicide Attempts
Adjusted Models for High-Risk Anxiousness and Disruptiveness as Risk Factors for Lifetime Suicide Attempts

Trajectories of Disruptiveness

High-risk disruptiveness was significantly associated with suicide attempts (OR = 1.80). In addition to sex (OR = 2.28) and substance abuse or dependence (OR = 1.50), disruptive disorders were also statistically relevant (OR = 2.39). Similarly, familial suicide attempts (OR = 2.83) and childhood sexual abuse (OR = 1.36) had positive associations with personal history of suicide attempts. In contrast, the personality trait dimension of conduct problems, assessed in adulthood by means of the Diagnostic Assessment of Personality Pathology, was not significant (OR = 1.03) (Table 2).

Joint Trajectories

Moderate to high levels of disruptiveness and anxiousness doubled the risk of suicide attempts (OR = 1.88). Adolescent internalizing (OR = 1.40) and disruptive (OR = 2.21) disorders, substance abuse or dependence (OR = 1.41), and conduct problems (OR = 1.04) were also statistically significant. Family history of suicide attempts tripled the likelihood of personal history of suicide attempts (OR = 2.96), whereas childhood sexual abuse had a more modest effect size (OR = 1.35) (Table 2).

Mediating and moderating effects

Childhood anxiousness and disruptiveness in this study did not influence the risk for suicide attempts by increasing the likelihood of adolescent anxiety/mood or disruptive disorders, respectively. The 4 criteria required to reject the null hypothesis of no mediation were not met. Only disruptive and disruptive/anxious children had a higher prevalence of disruptive disorders than their low-trajectory counterparts (disruptive children: 10.0% vs 5.0%, respectively; P < .001; and disruptive/anxious children, 9% vs 6%, respectively; P = .11), but the mediating/indirect effects were not statistically significant.

The relationship between high-risk joint trajectories and suicide attempts was moderated by sex (test of interaction: Wald χ2 = 5.90; P = .01). While the average OR in the sample was 1.88 (Table 2), when stratified by sex, the odds of suicide attempts in children following high-risk joint trajectories were higher among girls (OR = 3.60; Wald χ2 = 10.93; P < .001) than among boys (OR = 0.80; Wald χ2 = 0.23; P = .64).


Using a developmental, person-centered approach, we examined temporal trends in childhood anxiousness and disruptiveness in relation to lifetime suicide attempts assessed in early adulthood. To our knowledge, this is the first study to examine the codevelopment of anxiousness and disruptiveness and risk for suicide attempts. This is also one of the first efforts to formally evaluate relevant mediating and moderating effects.

We observed 4 distinct developmental profiles of anxiousness and disruptiveness and a frequent co-occurrence of similar levels of the 2 traits. Four trajectories of disruptiveness were also reported in a different French-Canadian cohort focusing on disadvantaged boys and individual components of disruptiveness.31 Our data agree with prior studies estimating that highly anxious temperaments occur in 15% to 20% of any population,37 with highly disruptive behaviors being less frequent.38 Intriguingly, we have observed that while high anxiousness can and most often did coexist with very low disruptiveness, the converse was rarely the case. This relational asymmetry requires research into the biological and environmental contributions to each trait as well as epidemiological validation in other samples.

Anxious, disruptive, and, in particular, anxious/disruptive children had a higher prevalence of and higher risk for suicide attempts. The relationship between high-risk joint and high-risk disruptiveness trajectories with suicide attempts persisted in the presence of demographic, psychiatric, experiential, and adult personality covariates. Our conclusions are similar to previous research whose support for anxious traits appears to be less consistent than that for externalizing traits related to disruptiveness.39

As for our covariates, we provide further support for the relevance of Axis I disorders to suicide attempts. This was evident for externalizing diagnoses in adolescence and adulthood (disruptive and substance abuse disorders) and for internalizing diagnoses in adolescence (mood and anxiety disorders) but not in adulthood. The association of internalizing disorders (mainly major and bipolar depression) and suicidality has been consistently corroborated.40 As for externalizing diagnoses, impulse control dysregulation, a feature common to substance abuse and disruptive disorders, may explain their involvement in both suicidal behaviors and psychiatric comorbidity.41,42

Second, of our 2 adult personality constructs, the trait of conduct problems exhibited a stronger relationship with suicide attempts than did adult anxiety. Considered together with the significant involvement of the disruptiveness and joint but not anxiousness trajectories, this suggests that early externalizing behaviors may be more useful markers of the risk for suicidality than both the internalizing ones and those assessed in adulthood.

Third, sex exhibited a main effect in relation to disruptiveness but a moderating one in relation to joint trajectories. (The absence of sex differences in preadolescent anxiousness is not too surprising given that such differences in internalizing behaviors may not appear until adolescence.43) Sex-based differences in joint and disruptiveness trajectories are consistent with the reported sex differences in externalizing behaviors and suicide attempts.22,23,43 Together, these findings suggest that sex merits serious consideration in both clinical and research contexts. Clinically, girls at risk for suicide attempts appear to display both anxious and disruptive traits and boys appear to display mostly the latter. Before this can be used in tailoring preventive and management programs, mechanisms underlying sex moderation, such as gene-environment interactions, require further research as they may account for sex-specific profiles of personality markers of suicide attempts.

Among candidate endophenotypes that can explain associations between childhood traits and suicidal behaviors, psychopathology is particularly promising because on the one hand it may represent an extreme expression of temperament44 and on the other it is strongly associated with suicidality. Nevertheless, our hypotheses predicting that adolescent mood/anxiety and disruptive disorders mediate the effects of childhood anxiousness and disruptiveness, respectively, on suicide attempts were not supported. One explanation may be methodological. Underestimation of the effect of the mediator and overestimation of the effect of the predictor are directly related to the measurement error in the mediator.24 More powerful samples, either larger in size or lower in attrition, will be necessary before we can exclude this possibility.

Alternatively, childhood antecedents may indeed primarily act directly and any mediating effect through adolescent psychopathology may be weak, a situation that is plausible in young suicide attempters identified in the general population. Furthermore, early anxiousness and disruptiveness are undifferentiated, broad, and probably quite heterogeneous,45 and their mediators may be further affected by different moderators (eg, turning-point events or sex). Also, given that the effect of temperament on psychopathology may be mediated by both internal and external factors,44 multiple mediators—such as nonshared environmental factors and relational variables—acting collectively may also be involved. Because our study population was normative and young, some of these effects may be cumulative, contingent on continued developmental challenges and maturational patterns. In sum, not only can a mediational effect be restricted to specific subgroups but it may also require longer time to manifest. Future studies should therefore investigate moderated mediation and test multiple mediators simultaneously in suitably sized samples.

Our findings need to be considered in light of several methodological limitations. Given our culturally homogeneous community sample, present conclusions may have limited generalizability to other populations.

Attrition may have affected our internal validity, although we conducted weighted analyses to minimize its effect. Our estimates of ORs are probably larger than their risk ratio counterparts given that they are not equivalent for an outcome whose frequency is more than 10%.

Lastly, mediational tests require temporal ordering of the predictor, mediator, and outcome such that the mediator must follow the predictor and precede the outcome. Because we focused on lifetime suicide attempts and did not have information on their precise time of onset, it is possible that some suicide attempts had occurred before age 16 years, the average age at which our mediators were assessed. However, because less than 1% of suicide attempts occur before age 15 years, we do not believe that this had a major effect on our findings.46

Although our trajectories covary with and precede suicide attempts and their association is theoretically plausible and suggested in the literature, our design was not experimental. We may have failed to rule out other variables responsible for their relationship.

These limitations were balanced by a number of methodological strengths. We used a comprehensive, hypothesis-driven approach, relying on multipoint assessments by independent raters and conducting analyses weighted for attrition. Lastly, in establishing the optimal number of trajectories, we used statistical criteria rather than arbitrary cutoffs.

In conclusion, we have demonstrated that the effects of childhood personality markers (primarily externalizing) on suicide attempts are largely independent of related covariates. Moreover, these effects were direct rather than mediated by adolescent psychopathology. Pending further research, preventive programs may benefit from considering sex differences in personality markers as early as childhood.

Correspondence: Gustavo Turecki, MD, PhD, McGill Group for Suicide Studies, 6875 LaSalle Blvd, Montreal, QC H4H 1R3, Canada (gustavo.turecki@mcgill.ca).

Accepted for Publication: April 22, 2008.

Author Contributions: Drs Tremblay and Turecki share senior authorship of this work. Dr Brezo had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Brezo, Paris, Vitaro, and Turecki. Acquisition of data: Hébert, Vitaro, Tremblay, and Turecki. Analysis and interpretation of data: Brezo, Barker, Paris, Vitaro, and Turecki. Drafting of the manuscript: Brezo and Paris. Critical revision of the manuscript for important intellectual content: Barker, Paris, Hébert, Vitaro, Tremblay, and Turecki. Statistical analysis: Brezo, Barker, and Vitaro. Obtained funding: Paris, Hébert, Vitaro, Tremblay, and Turecki. Administrative, technical, and material support: Paris. Study supervision: Turecki.

Financial Disclosure: None reported.

Funding/Support: This work was supported by grant MOP 151060 from the Canadian Institutes of Health Research and by the National Consortium on Violence Research, Carnegie Mellon University, the Programme National de recherche et de développement en matière de santé, the Conseil Québécois de la recherche sociale, and the Fonds Québécois de la recherche sur la société et la culture.

Krueger  RFCaspi  AMoffitt  TE Epidemiological personology: the unifying role of personality in population-based research on problem behaviors.  J Pers 2000;68 (6) 967- 998PubMedGoogle ScholarCrossref
Brezo  JParis  JTurecki  G Personality traits as correlates of suicidal ideation, suicide attempts, and suicide completions: a systematic review.  Acta Psychiatr Scand 2006;113 (3) 180- 206PubMedGoogle ScholarCrossref
Coté  STremblay  RENagin  DSZoccolillo  MVitaro  F Childhood behavioral profiles leading to adolescent conduct disorder: risk trajectories for boys and girls.  J Am Acad Child Adolesc Psychiatry 2002;41 (9) 1086- 1094PubMedGoogle ScholarCrossref
van Heeringen  CAudenaert  KVan Laere  K  et al.  Prefrontal 5-HT2a receptor binding index, hopelessness and personality characteristics in attempted suicide.  J Affect Disord 2003;74 (2) 149- 158PubMedGoogle ScholarCrossref
Beautrais  ALJoyce  PRMulder  RT Personality traits and cognitive styles as risk factors for serious suicide attempts among young people.  Suicide Life Threat Behav 1999;29 (1) 37- 47PubMedGoogle Scholar
Koller  GPreuss  UWBottlender  MWenzel  KSoyka  M Impulsivity and aggression as predictors of suicide attempts in alcoholics.  Eur Arch Psychiatry Clin Neurosci 2002;252 (4) 155- 160PubMedGoogle ScholarCrossref
Kausch  O Suicide attempts among veterans seeking treatment for pathological gambling.  J Clin Psychiatry 2003;64 (9) 1031- 1038PubMedGoogle ScholarCrossref
Turecki  G Dissecting the suicide phenotype: the role of impulsive-aggressive behaviours.  J Psychiatry Neurosci 2005;30 (6) 398- 408PubMedGoogle Scholar
Brent  DAMann  JJ Family genetic studies, suicide, and suicidal behavior.  Am J Med Genet C Semin Med Genet 2005;133C (1) 13- 24PubMedGoogle ScholarCrossref
Gil  S The role of personality traits in understanding of suicide attempt behavior among psychiatric patients.  Arch Suicide Res 2003;7 (2) 159- 166Google ScholarCrossref
Caspi  AMoffitt  TENewman  DLSilva  PA Behavioral observations at age 3 years predict adult psychiatric disorders: longitudinal evidence from a birth cohort.  Arch Gen Psychiatry 1996;53 (11) 1033- 1039PubMedGoogle ScholarCrossref
Lonigan  CJPhillips  BMHooe  ES Relations of positive and negative affectivity to anxiety and depression in children: evidence from a latent variable longitudinal study.  J Consult Clin Psychol 2003;71 (3) 465- 481PubMedGoogle ScholarCrossref
Klein  MHWonderlich  SShea  MT Models of relationships between personality and depression: toward a framework for theory and research. Klein  MHKupfer  DJShea  MT Personality and Depression A Current View. New York, NY Guilford Press1993;1- 54Google Scholar
Boden  JMFergusson  DMHorwood  LJ Anxiety disorders and suicidal behaviours in adolescence and young adulthood: findings from a longitudinal study.  Psychol Med 2007;37 (3) 431- 440PubMedGoogle ScholarCrossref
Watson  DClark  LAHarkness  AR Structures of personality and their relevance to psychopathology.  J Abnorm Psychol 1994;103 (1) 18- 31PubMedGoogle ScholarCrossref
Kelly  TMCornelius  JRLynch  KG Psychiatric and substance use disorders as risk factors for attempted suicide among adolescents: a case control study.  Suicide Life Threat Behav 2002;32 (3) 301- 312PubMedGoogle ScholarCrossref
Santa Mina  EEGallop  RM Childhood sexual and physical abuse and adult self-harm and suicidal behaviour: a literature review.  Can J Psychiatry 1998;43 (8) 793- 800PubMedGoogle Scholar
Renaud  JBrent  DABirmaher  BChiappetta  LBridge  J Suicide in adolescents with disruptive disorders.  J Am Acad Child Adolesc Psychiatry 1999;38 (7) 846- 851PubMedGoogle ScholarCrossref
Caspi  AHarrington  HMilne  BAmell  JWTheodore  RFMoffitt  TE Children's behavioral styles at age 3 are linked to their adult personality traits at age 26.  J Pers 2003;71 (4) 495- 513PubMedGoogle ScholarCrossref
Jones  CJMeredith  W Patterns of personality change across the life span.  Psychol Aging 1996;11 (1) 57- 65PubMedGoogle ScholarCrossref
Lewis  M Altering Fate: Why the Past Does Not Predict the Future.  New York, NY Guilford Press1997;
Costa  PT  JrTerracciano  A McCrae  RR Gender differences in personality traits across cultures: robust and surprising findings.  J Pers Soc Psychol 2001;81 (2) 322- 331PubMedGoogle ScholarCrossref
Beautrais  AL Gender issues in youth suicidal behaviour.  Emerg Med (Fremantle) 2002;14 (1) 35- 42PubMedGoogle ScholarCrossref
Baron  RMKenny  DA The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations.  J Pers Soc Psychol 1986;51 (6) 1173- 1182PubMedGoogle ScholarCrossref
Màsse  LCTremblay  RE Behavior of boys in kindergarten and the onset of substance use during adolescence.  Arch Gen Psychiatry 1997;54 (1) 62- 68PubMedGoogle ScholarCrossref
Breton  JJBergeron  LValla  JPBerthiaume  CSt-Georges  M Diagnostic Interview Schedule for Children (DISC-2.25) in Quebec: reliability findings in light of the MECA study.  J Am Acad Child Adolesc Psychiatry 1998;37 (11) 1167- 1174PubMedGoogle ScholarCrossref
American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders. 3rd ed, revised Washington, DC American Psychiatric Association1987;
Livesley  WJJang  KLVernon  PA Phenotypic and genetic structure of traits delineating personality disorder.  Arch Gen Psychiatry 1998;55 (10) 941- 948PubMedGoogle ScholarCrossref
Robins  LNCottler  LBucholtz  KCompton  W Diagnostic Interview Schedule for DSM-IV.  St Louis, MO Washington University1995;
Jones  BLNagin  DS Advances in group-based trajectory modeling and an SAS procedure for estimating them.  Sociol Methods Res 2007;35 (4) 542- 571Google ScholarCrossref
Nagin  DTremblay  RE Trajectories of boys' physical aggression, opposition, and hyperactivity on the path to physically violent and nonviolent juvenile delinquency.  Child Dev 1999;70 (5) 1181- 1196PubMedGoogle ScholarCrossref
Jones  BLNagin  DSRoeder  K A SAS procedure based on mixture models for estimating developmental trajectories.  Sociol Methods Res 2001;29 (3) 374- 393Google ScholarCrossref
Nagin  DSTremblay  RE Analyzing developmental trajectories of distinct but related behaviors: a group-based method.  Psychol Methods 2001;6 (1) 18- 34PubMedGoogle ScholarCrossref
Jones  BL SAS PROC TRAJ. http://www.andrew.cmu.edu/user/bjones/index.htm. Accessed November 15, 2007
Holmbeck  GN Post-hoc probing of significant moderational and mediational effects in studies of pediatric populations.  J Pediatr Psychol 2002;27 (1) 87- 96PubMedGoogle ScholarCrossref
Kenny  DA Mediation. http://davidakenny.net/cm/mediate.htm. Accessed November 15, 2007
Prior  MSmart  DSanson  AOberklaid  F Sex differences in psychological adjustment from infancy to 8 years.  J Am Acad Child Adolesc Psychiatry 1993;32 (2) 291- 304PubMedGoogle ScholarCrossref
Mun  EYFitzgerald  HEvon Eye  APuttler  LIZucker  RA Temperamental characteristics as predictors of externalizing and internalizing child behavior problems in the contexts of high and low parental psychopathology.  Infant Ment Health J 2001;22 (3) 393- 415Google ScholarCrossref
Brezo  JParis  JTremblay  R  et al.  Personality traits as correlates of suicide attempts and suicidal ideation in young adults.  Psychol Med 2006;36 (2) 191- 202PubMedGoogle ScholarCrossref
Cavazzoni  PGrof  PDuffy  A  et al.  Heterogeneity of the risk of suicidal behavior in bipolar-spectrum disorders.  Bipolar Disord 2007;9 (4) 377- 385PubMedGoogle ScholarCrossref
Turecki  G Suicidal behavior: is there a genetic predisposition?  Bipolar Disord 2001;3 (6) 335- 349PubMedGoogle ScholarCrossref
Verona  ESachs-Ericsson  NJoiner  TE  Jr Suicide attempts associated with externalizing psychopathology in an epidemiological sample.  Am J Psychiatry 2004;161 (3) 444- 451PubMedGoogle ScholarCrossref
Bongers  ILKoot  HMvan der Ende  JVerhulst  FC The normative development of child and adolescent problem behavior.  J Abnorm Psychol 2003;112 (2) 179- 192PubMedGoogle ScholarCrossref
Rettew  DC McKee  L Temperament and its role in developmental psychopathology.  Harv Rev Psychiatry 2005;13 (1) 14- 27PubMedGoogle ScholarCrossref
Crijnen  AAMAchenbach  TMVerhulst  FC Problems reported by parents of children in multiple cultures: the Child Behavior Checklist syndrome constructs.  Am J Psychiatry 1999;156 (4) 569- 574PubMedGoogle Scholar
Goldman  SBeardslee  WR Suicide in children and adolescents. Jacobs  DG The Harvard Medical School Guide to Suicide Assessment and Intervention. San Francisco, CA Jossey-Bass Publishers1999;Google Scholar