Suicide is the third leading cause of death in adolescents and young adults in the United States and the second leading cause in European countries 1 . Suicidal behaviors are also the most common reason for adolescent psychiatric hospitalizations in many countries 2 . Reducing suicide and suicide attempts is therefore a key public health target. In the United States, the death rate by suicide is 6.9/100 000 in adolescents aged 15 to 19 3 . In France, recent epidemiological data showed that the suicide rate in adolescents aged 15 to 19 is 4.1/100 000 inhabitants 4 . Considerable variability exists among the European countries that published their statistics regarding death rates by suicide in 2008 5 . Prevalence of suicidal ideations ranges from 15 to 25% in the general population, whereas the lifetime estimates of suicide attempts among adolescents range from 1.3 to 3.8% in males and from 1.5 to 10.1% in females, with higher rates in females than in males in the older age range 6 .

Current models of suicide phenomena in adolescents emphasize: (i) the importance of distinguishing suicidal ideation, non-suicidal self-harm, suicide attempt and completed suicide 7 8 (ii) the key role of depression in the transition from suicidal ideations to suicide attempts, in which depression is a strong proximal factor 9 ; (iii) the fact that the numerous risk factors identified do not capture the whole risk leading to the idea that protective factors should be taken into account for suicide risk prediction 10 . Risk factors for completed and attempted suicide have been widely studied. First, psychiatric disorders are present in about 90% of suicidal adolescents 6 . Depressive disorders are consistently the most prevalent psychiatric disorder among adolescents who commit suicide with a prevalence ranging from 49% to 64% and among adolescents who attempt suicide 6 11 12 . Secondly, adolescents who attempted suicide in the past are up to 60 times more likely to commit suicide than those who have not 6 . Also, self-harm is an important predictor of future completed suicide 13 . Thirdly, substance abuse plays a significant role in adolescent suicide and in suicide attempts, especially in older adolescent males when it is comorbid with mood disorders or disruptive disorders 14 15 . Fourthly, social factors such as socio-economic status, school exclusion and social isolation have been also implicated 16 17 . Finally, several studies pointed a significant association with family factors, including family psychopathology, abuse, loss of a parent (death, divorce), intrafamilial relationships, familial cohesion, support and suicidality 16 18 19 20 .

Indeed, the family factors, and especially the perceived quality of family relationships, have been pinpointed as an important risk or protective factor in clinical and community samples of adolescents 1 2 6 21 22 23 24 25 26 . However, only few population-based studies have examined family factors 19 . They showed several predictive or associated factors, like: poor family environment (low satisfaction with support, communication, leisure time, low parental monitoring) 27 , low family support 28 , low family cohesion 29 , poor family functioning, poor parent–child attachment and problems of parental adjustment 1 19 . On the contrary, higher family cohesion has been reported as a protective factor against future suicide attempt 26 as well as having positive relationships with a parent 30 31 . Improved family connectedness was related to less severe depressive symptoms and suicidal ideation 32 . Nevertheless, equivocal findings exist with regard to the relationship between adolescents’ suicidal behaviours and family variables. This is mainly due to methodological limitations, such as considering only parental marital status (e.g. 22 ) or parents together (e.g. 33 ), and ignoring other common risk factors from multivariate analysis (e.g. 16 19 34 ). Moreover, data suggest a different effect of family factors on suicidal behaviours according to gender (e.g. 34 ), clinical severity (e.g. 34 ), parental marital status (e.g. 22 ), dissatisfaction with relationship with parents (e.g. 33 ), and different relationship with mother vs. father (e.g. 34 ).

Notwithstanding these interesting results, the complex association between family factors, depression and suicidal behaviors among adolescents remains to be explored in samples large enough to allow multivariate analysis, so as to understand specific contributions (e.g. mother vs. father; conflict vs. no conflict; separation vs. no separation) taking into account other risk factors and severity of depression and suicidal behaviors. The aim of the present study was to assess the link between family factors and suicidal behaviors, adjusting for several potential confounding factors, in a large community-based sample of adolescents aged 17 years. Given that the prevalence of suicide differs substantially between boys and girls, we hypothesized that the impact of familial risk factors would differ according to gender. Similarly, given the role of current depression, we hypothesized that family risk would be related to depression severity, defined as depression associated with suicidal ideation in the last year and/or life-time suicide attempt.

This study assessed the associations between depression, family factors and suicidality in a large representative community-based sample of adolescents aged 17 (n = 39,542), adjusting for confounding variables. Given data in the literature regarding depression as a proximal risk factor in suicidality 9 and the relevance of classifying suicidality (ideations and suicide attempt) 7 , we divided the sample into 3 grades of suicide risk severity combining depression and suicidality (grade 1 = depressed without suicidal ideation and without suicide attempts, grade 2 = depressed with suicidal ideations and grade 3 = depressed with suicide attempts). The results confirmed previous risk factors for depression/suicidality in adolescents. Previously, school exclusion and academic difficulties have been implicated in suicidality in young people 17 41 . In France, given the high frequency of repeated years in school, this educational data also needs to be taken into account in assessment of suicidality among adolescents. All substance use including tobacco and marijuana use was associated with increased suicide risk in depressed adolescents. It has been shown that, unless comorbid, substance abuse disorders were not proximally associated with suicidality 9 . Adjusting on confounding variables (educational data, socio economic status, substance use), the results here showed that negative relationships with either or both parents, and parents’ living together with a negative relationship were significantly associated with depression and/or suicide risk in both genders (all risk severity grades) and that odds ratios increased according to risk severity grade. This means that what affects depression and suicidality is not parental separation per se, but rather parental harmony on the one hand, and perceived quality of the adolescents’ relationships with mother and father, on the other. Although we hypothesized different familial risk factors between girls and boys because of differential epidemiology, we found similar family risk factors in the two genders.

We found depression rates similar to those reported in the literature (e.g. the Center for Disease Control, for the year 2005–2006, found a depression prevalence between 4% and 6.4% in adolescents aged 12 to 17, without testing for gender differences ) 42 . We also had a higher prevalence in girls than boys, as found in many epidemiological studies 43 44 45 . Therefore, the higher prevalence of depression in girls than boys may not be a consequence of differential perceptions of family relationships. It should rather be interpreted as a consequence of other factors that were not assessed in the current study: e.g. genetic vulnerability, hormonal changes, gender specific social constraints, differential comorbid psychopathology 45 46 47 48 49 .

The importance of family factors is strengthened by the fact that we found increases in odds ratios for most factors according to severity grade. Recent data suggested that defining and classifying suicidality could provide a better understanding of risk factors (proximal and distal) and interactions among them 7 . The recommended classification distinguishes depression, suicidal ideation and suicidal behavior in a hierarchical model 7 50 . Previous studies have underlined the role of family factors in suicidality in young people. First, adolescents who commit suicide are more likely to come from a family with a history of suicide and/or family psychopathology 17 19 20 51 . Second, childhood abuse, a history of separation and loss (by death or divorce) and exposure to physical and/or sexual violence are also associated with suicidality 16 52 53 54 55 . Third, adolescents with suicidal behaviors are more likely to be living in non-intact families 17 22 33 56 57 58 59 and their environment is characterized by problematic communication, poor attachment and high levels of conflict 14 16 17 28 29 34 51 57 60 61 62 . In depressed adolescents, poor family function is predictor of suicide attempts 1 , and suicidal ideations and family conflict were independently associated with a suicidal event over a one-year follow-up 26 . Another recent study showed that the most common proximal risk factor for completed suicide for subjects younger than 30 years was conflict with family members, partners or friends 63 . Here, we focused on perceived intrafamilial relationships and found that negative relationship with either or both parents, and parents living together but with discord were significantly associated with suicide risk and/or depression in the two genders.

The current results have important clinical implications. Practitioners working with young people presenting depression and suicidal behaviors (ideation and/or attempts) should take the family factors into account, in particular aspects such as the adolescent’s relationships with either or both parents and relationships between parents whether or not they are living together. Assessing suicide risk in adolescents should include the assessment of family relationships and this could enable appropriate care to be provided for the adolescent and his family. A recent study assessed treatment of adolescent suicide attempters 64 . Depressed adolescents with prior suicide attempts were treated with a combination of medication and psychotherapy. After treatment, rates of improvement and remission of depression appeared comparable to those in non-suicidal depressed adolescents. The treatment included antidepressant medication and CBT (specifically developed to address suicide risk) including both individual and parent-adolescent sessions. Parent-adolescent sessions had probably contributed to this improvement. Of course, other psychotherapies have empirical evidences for its effectiveness such as family therapy.

The current study has several limitations. First, we could only focus on and assess a limited number of risk factors. Regarding adolescent psychopathology, 70 to 91% of young people who attempt or commit suicide present a psychiatric disorder 60 65 . Depression is the most common diagnosis in adolescents who commit suicide and it is highly prevalent in those with suicidal ideations and suicide attempts 15 65 . However, other conditions can interfere, but were not assessed in the current study (e.g. generalized anxiety disorder; disruptive behaviors; borderline personality disorder) 9 15 49 66 67 . Similarly, many non-clinical risk factors were not assessed (e.g. life stressors, problems with authorities, relationship problems with peers, sexual and physical abuse, low socio-economic status) 7 . Second, as our study was cross sectional, meaning that the assessment of suicidal behaviors and changes in family structure was retrospective and that the mechanisms underpinning the associations could not be investigated. Only longitudinal studies are able to explore the different effects of the potential moderators of associations. Third, we had no data available on ethnicity because in France it is not allowed by ethics committees. It can however be noted that the present data only concerned French people from metropolitan France (i.e. excluding overseas territories). The sample nevertheless included 5% of the French metropolitan population aged 17 and was representative of it. Fourth, we had a differential temporal focus for our clinical variables. Current depression was measured for the previous 2 weeks, suicide ideations concerned the past 12 months and suicide attempts concerned lifetime. However, (1) given that subjects were 17 years old, suicide attempts mostly concerned the previous 5 years; (2) prior suicide attempt is an important risk factor for suicidality in young people. In addition, we did not differentiate single suicide attempt and lifetime history of several attempts because of the small numbers of subjects in each subgroup. Thus, grade 3 risk severity included adolescents with a history of one or several suicide attempts. Finally, our aim to investigate current depression as a proximal risk factor led us to exclude many adolescents who had experienced suicidal ideations in the past 12 months and/or lifetime suicide attempt(s) but were not depressed at the time of assessment (see Figure 1). The same analyses (multivariate analysis) as those conducted in this study were performed on the excluded sample and showed similar results for family risk factor (see Additional file 1: Figure S1). Therefore the exclusion of these subjects did not radically modify our results. Finally, self-report of family functioning was also a limitation because depression may lead to a negative perception bias regarding relationships with parents.

The study also has some strength. First, the study included a large representative population-based sample of French adolescents aged 17 which allowed an exhaustive investigation of suicide risk severity in depressed adolescents. In addition, the setting in which the study was implemented (JAPD) was a good guarantee of methodological thoroughness for sampling and conditions of administration. Compared to studies conducted in adult populations, we were able to restrict recall bias because subjects were all 17 years old. Second, depression assessment was performed on a scale specific to adolescents 38 . In previous studies, depression has often been lifetime depression so that it was difficult to know if depression reported by a subject was present before, during or after suicide attempts. Third, results regarding family factors were adjusted on several confounding variables (educational level, repeat school years, socio-economic status, and substance use). Fourth, because of the good statistical power, we were able to (1) run multivariate analyses on each gender; (2) distinguish family separation, family discord and perceived parental relationship.

This study has provided keys regarding the epidemiology of suicidal behaviours in a large community-based sample of French adolescents aged 17. Substance use, repeat school years and family factors were associated with an increased suicide risk in depressed adolescents, with no difference according to gender. Specifically, perceived negative relationships with either or both parents, and a negative relationship between parents, whether living together or not, were associated with suicidality. So it appears essential to take into account real and perceived intrafamilial relationships among depressed adolescents to assess and prevent suicidal behaviours.

Dr Consoli reported receiving travel support from Bristol-Myers Squibb and Dr. Cohen reported past consultation for or the receipt of honoraria from Schering-Plough, Bristol-Myers Squibb, Otsuka, Janssen, Shire, and Sanofi-Aventis. Dr Revah-Levy, Moro, Peyre, Speranza, Falissard, Mme Hassler, Touchette have no relationships that might have interest in the submitted work. No authors have any non-financial interests that may be relevant to the submitted work.

Study concept and design: ARL, ET, AC, HP, BF. Acquisition of data: CH, ARL, HP, ET. Statistical analysis: BF, DC, HP, CH. Interpretation of data: all authors. Drafting the manuscript: AC, ARL, DC. Critical revision of the manuscript revision for important intellectual content: MRM, ARL, MS, DC, BF. Final draft: all authors. All authors read and approved the final manuscript.