The effect of family structure on youth adjustment has received increasing attention as historical trends in single parenting, divorce, remarriage, and cohabitation with partners and extended family members have produced a diverse constellation of structures. African American youth are less likely than Caucasian youth to live in an “intact” family. Links between family structure and a variety of indices of youth adjustment have been established, although a relatively understudied outcome is that of substance initiation, despite its association with dependence and other negative sequelae. The dynamic effect of transitions has additionally been less studied than the static effect of structure. Differences in family structure and transitions may influence outcomes via parental socialization (monitoring and attachment) as well as strain (residential mobility and changes in income). These mechanisms may operate differently for Caucasian and African American youth, and may partially explain differences in adjustment. Relations between youth adjustment and transitions may be reciprocal in nature, a less often studied premise. This project made use of a nationally representative sample of more than 2,000 adolescents aged 12 to 13 in 1997 assessed across 3 waves. Regression analyses were employed to examine the associations among family structure and transitions, parenting, income, residential mobility, and substance initiation over time. This study found that living in non- two-parent family structures was consistently associated with higher concurrent levels of substance initiation, lower parental monitoring and relationship quality, lower income, and higher residential mobility. The effects of transitions on substance initiation and parenting were less robust than hypothesized, but reinforced the notion that consistently living outside a two-parent family, or consistently living in a single-parent family, is negatively associated with parenting, income, and residential stability over time. Evidence for mediated effects via changes in parenting, residential mobility, and income were significant but small in magnitude, and varied by race, such that they were significant for Caucasian but not African American youth . Partial evidence for reciprocal causality was found. Alcohol initiation at the first wave predicted separations, but marijuana initiation did not. These findings have important implications for parents, clinicians, and policy makers.

A presente tese aborda o tema da dependência de crack na adolescência, assunto de grande relevância clínica devido à gravidade dos pacientes e à dificuldade em realizar tratamentos eficazes. Ambos os estudos apresentados foram feitos a partir de uma amostra de 90 adolescentes, de 12 a 18 anos incompletos, internados por problemas relacionados ao uso de crack em duas enfermarias na cidade de Porto Alegre, Rio Grande do Sul, Brasil. A amostra consecutiva foi coletada de Maio de 2011 a Novembro de 2012. Também foi coletada uma amostra controle de 81 adolescentes sem uso de drogas, provenientes de um bairro de baixa-renda na região metropolitana de Porto Alegre. Como resultados do primeiro estudo, observou-se que os adolescentes internados por uso de crack apresentaram idade média de 15,6 anos (DP=1,4), e a maioria (85,5%) era do sexo masculino. Todos os pacientes haviam utilizado alguma outra substância psicoativa antes de iniciar o uso de crack: 61,4% tabaco (idade média do primeiro uso 11,6 anos), 44,3% álcool (idade média do primeiro uso 12,4 anos), e 54,5% maconha (idade média do primeiro uso 12,15 anos). A idade média de início do crack foi aos 13,4 anos. Apresentaram alta taxa de comorbidades psiquiátricas, com 81,8% dos pacientes com o diagnóstico de Transtorno de Conduta, 52,3% com Transtorno Opositor Desafiante e 44,3% apresentando Transtorno de Déficit de Atenção e Hiperatividade (todas essas com p<0,001 em relação aos controles). Evidenciamos consequências graves do uso de drogas em todas as áreas da vida (como altos escores de comprometimento nas áreas de uso de substâncias, escola, família e psiquiátrica do questionário Teen Addiction Severity Index – T-ASI). O tempo médio decorrido entre o início de uso de drogas e o início de uso de crack foi de 2,53 anos (DP=1,96). Através do modelos de regressão de Cox, encontramos os seguintes preditores de progressão precoce ao uso de crack: idade de início de uso de qualquer substância e idade no momento da internação. Dando continuidade ao tema desta tese, o segundo artigo abordou os níveis de quatro potenciais biomarcadores para o uso de crack na adolescência. São eles: o Fator de crescimento derivado do cérebro (BDNF), as interleucinas 6 e 10 (IL-6 e IL-10), e as substâncias reativas ao ácido tiobarbitúrico (TBARS). Foram medidos, na mesma amostra de pacientes do primeiro estudo, em dois momentos: no dia seguinte à internação e no momento da alta. Também foram medidos em controles. As análises foram controladas para possíveis confundidores. Encontrou-se níveis significativamente mais baixos de BDNF entre os pacientes (16,61; DP=2,06), em relação aos controles (25,12; DP=175) (p<.001). Os níveis de TBARS estavam significativamente elevados entre os pacientes (25,07; DP=21,83), em relação aos controles (16,46; DP=23,25) (p=0,005). Os níveis de IL-6 estavam significativamente elevados entre os pacientes (577,34; DP=156,11) em relação aos controles (151,91; DP=64,7) (p=0,027). Os níveis de IL-10 estavam significativamente elevados entre os pacientes (383,86; DP= 97,24), em relação aos controles(109,89; DP=40,89) (p=0,025). Concluiu-se que adolescentes internados por uso de crack apresentam problemas em diversas áreas da vida, altas taxas de comorbidade e alterações em biomarcadores específicos relacionados com inflamação e estresse oxidativo. Sugere-se que esta população necessita cuidados especiais para evitar comprometimentos maiores. / This thesis is about crack-cocaine dependence in adolescence, a subject that is of high clinical interest due to the severity of the clinical presentation of those patient and the challeng of providing effective treatment. Both studies presented were conducted with a ample of 90 adolescents, from 12 to 18 years age, of both sexes, who had been admitted for problems related to crack-cocaine use in two impatient treatment units for adolescents. In the city of Porto Alegre, south of Brazil. This consecutive sample was collected from May 2011 to November 2012. A control sample of drug-naïve adolescents has been recruited in a low income neighborhood. As results of this first study, it was observed that adolescents admitted for crack-cocaine use had a mean 15.6 years of age, and most of them (85.5%) were male. All patients had used at least one other psychoactive substance before initiating crack-cocaine use: 61.4% had used tobacco (mean age of first use=11.6 years), 44.3 had used alcohol (mean age of first use=12.4 years), and 54.5 5 had used cannabis (mean age of first use = 12.15 years). Mean age of crack-cocaine initiation was 13.4 years. Patients had high rates of comorbid conditions, with 81.8% of patients having a lifetime diagnosis of Conduct Disorder, 52.3% with Oppositional Defiant Disorder, and 44.3% with Attention Deficit and Hyperactivity Disorder. All the comorbid conditions found were more prevalent in the patients group than in controls. The T-ASI interview showed severe consequences of crack-cocaine abuse in all life areas. Mean time from first substance use to first use of crack cocaine was 2.5 years. By using Cox regression moles, we found that predictors to early crack-cocaine initiation were: age of first substance use and current age. The second study is about four potential biomarkers of crack-cocaine use in adolescence. They are the Brain Derived Neurotrophic Factor (BDNF), the interleukins 6 and 10 (IL-6 and IL-10), and Thiobarbituric acid reactive substance (TBARS). They were measured in peripherical blood, in the same sample of patients from the first study, at two moments: the day after admittance and in the day of discharge from the unit. They were also measured in controls. They were compared using Generalized Estimating Equations, and analyses where controlled by possible confounders. BDNF levels were lower in patients than in controls. TBARS levels were higher in patients than in controls. IL=6 was higher in patients than in controls. IL-10 was higher in patients than in controls. In conclusion, adolescents that are admitted to inpatient units for crack-cocaine-related problems show high degree of impairment in multiple life areas, high rates of comorbid conditions and alterations in biomarkers related to inflammation and oxidative stress. It is suggested that this population need special care to prevent further impairments.

A presente tese aborda o tema da dependência de crack na adolescência, assunto de grande relevância clínica devido à gravidade dos pacientes e à dificuldade em realizar tratamentos eficazes. Ambos os estudos apresentados foram feitos a partir de uma amostra de 90 adolescentes, de 12 a 18 anos incompletos, internados por problemas relacionados ao uso de crack em duas enfermarias na cidade de Porto Alegre, Rio Grande do Sul, Brasil. A amostra consecutiva foi coletada de Maio de 2011 a Novembro de 2012. Também foi coletada uma amostra controle de 81 adolescentes sem uso de drogas, provenientes de um bairro de baixa-renda na região metropolitana de Porto Alegre. Como resultados do primeiro estudo, observou-se que os adolescentes internados por uso de crack apresentaram idade média de 15,6 anos (DP=1,4), e a maioria (85,5%) era do sexo masculino. Todos os pacientes haviam utilizado alguma outra substância psicoativa antes de iniciar o uso de crack: 61,4% tabaco (idade média do primeiro uso 11,6 anos), 44,3% álcool (idade média do primeiro uso 12,4 anos), e 54,5% maconha (idade média do primeiro uso 12,15 anos). A idade média de início do crack foi aos 13,4 anos. Apresentaram alta taxa de comorbidades psiquiátricas, com 81,8% dos pacientes com o diagnóstico de Transtorno de Conduta, 52,3% com Transtorno Opositor Desafiante e 44,3% apresentando Transtorno de Déficit de Atenção e Hiperatividade (todas essas com p<0,001 em relação aos controles). Evidenciamos consequências graves do uso de drogas em todas as áreas da vida (como altos escores de comprometimento nas áreas de uso de substâncias, escola, família e psiquiátrica do questionário Teen Addiction Severity Index – T-ASI). O tempo médio decorrido entre o início de uso de drogas e o início de uso de crack foi de 2,53 anos (DP=1,96). Através do modelos de regressão de Cox, encontramos os seguintes preditores de progressão precoce ao uso de crack: idade de início de uso de qualquer substância e idade no momento da internação. Dando continuidade ao tema desta tese, o segundo artigo abordou os níveis de quatro potenciais biomarcadores para o uso de crack na adolescência. São eles: o Fator de crescimento derivado do cérebro (BDNF), as interleucinas 6 e 10 (IL-6 e IL-10), e as substâncias reativas ao ácido tiobarbitúrico (TBARS). Foram medidos, na mesma amostra de pacientes do primeiro estudo, em dois momentos: no dia seguinte à internação e no momento da alta. Também foram medidos em controles. As análises foram controladas para possíveis confundidores. Encontrou-se níveis significativamente mais baixos de BDNF entre os pacientes (16,61; DP=2,06), em relação aos controles (25,12; DP=175) (p<.001). Os níveis de TBARS estavam significativamente elevados entre os pacientes (25,07; DP=21,83), em relação aos controles (16,46; DP=23,25) (p=0,005). Os níveis de IL-6 estavam significativamente elevados entre os pacientes (577,34; DP=156,11) em relação aos controles (151,91; DP=64,7) (p=0,027). Os níveis de IL-10 estavam significativamente elevados entre os pacientes (383,86; DP= 97,24), em relação aos controles(109,89; DP=40,89) (p=0,025). Concluiu-se que adolescentes internados por uso de crack apresentam problemas em diversas áreas da vida, altas taxas de comorbidade e alterações em biomarcadores específicos relacionados com inflamação e estresse oxidativo. Sugere-se que esta população necessita cuidados especiais para evitar comprometimentos maiores. / This thesis is about crack-cocaine dependence in adolescence, a subject that is of high clinical interest due to the severity of the clinical presentation of those patient and the challeng of providing effective treatment. Both studies presented were conducted with a ample of 90 adolescents, from 12 to 18 years age, of both sexes, who had been admitted for problems related to crack-cocaine use in two impatient treatment units for adolescents. In the city of Porto Alegre, south of Brazil. This consecutive sample was collected from May 2011 to November 2012. A control sample of drug-naïve adolescents has been recruited in a low income neighborhood. As results of this first study, it was observed that adolescents admitted for crack-cocaine use had a mean 15.6 years of age, and most of them (85.5%) were male. All patients had used at least one other psychoactive substance before initiating crack-cocaine use: 61.4% had used tobacco (mean age of first use=11.6 years), 44.3 had used alcohol (mean age of first use=12.4 years), and 54.5 5 had used cannabis (mean age of first use = 12.15 years). Mean age of crack-cocaine initiation was 13.4 years. Patients had high rates of comorbid conditions, with 81.8% of patients having a lifetime diagnosis of Conduct Disorder, 52.3% with Oppositional Defiant Disorder, and 44.3% with Attention Deficit and Hyperactivity Disorder. All the comorbid conditions found were more prevalent in the patients group than in controls. The T-ASI interview showed severe consequences of crack-cocaine abuse in all life areas. Mean time from first substance use to first use of crack cocaine was 2.5 years. By using Cox regression moles, we found that predictors to early crack-cocaine initiation were: age of first substance use and current age. The second study is about four potential biomarkers of crack-cocaine use in adolescence. They are the Brain Derived Neurotrophic Factor (BDNF), the interleukins 6 and 10 (IL-6 and IL-10), and Thiobarbituric acid reactive substance (TBARS). They were measured in peripherical blood, in the same sample of patients from the first study, at two moments: the day after admittance and in the day of discharge from the unit. They were also measured in controls. They were compared using Generalized Estimating Equations, and analyses where controlled by possible confounders. BDNF levels were lower in patients than in controls. TBARS levels were higher in patients than in controls. IL=6 was higher in patients than in controls. IL-10 was higher in patients than in controls. In conclusion, adolescents that are admitted to inpatient units for crack-cocaine-related problems show high degree of impairment in multiple life areas, high rates of comorbid conditions and alterations in biomarkers related to inflammation and oxidative stress. It is suggested that this population need special care to prevent further impairments.

A presente tese aborda o tema da dependência de crack na adolescência, assunto de grande relevância clínica devido à gravidade dos pacientes e à dificuldade em realizar tratamentos eficazes. Ambos os estudos apresentados foram feitos a partir de uma amostra de 90 adolescentes, de 12 a 18 anos incompletos, internados por problemas relacionados ao uso de crack em duas enfermarias na cidade de Porto Alegre, Rio Grande do Sul, Brasil. A amostra consecutiva foi coletada de Maio de 2011 a Novembro de 2012. Também foi coletada uma amostra controle de 81 adolescentes sem uso de drogas, provenientes de um bairro de baixa-renda na região metropolitana de Porto Alegre. Como resultados do primeiro estudo, observou-se que os adolescentes internados por uso de crack apresentaram idade média de 15,6 anos (DP=1,4), e a maioria (85,5%) era do sexo masculino. Todos os pacientes haviam utilizado alguma outra substância psicoativa antes de iniciar o uso de crack: 61,4% tabaco (idade média do primeiro uso 11,6 anos), 44,3% álcool (idade média do primeiro uso 12,4 anos), e 54,5% maconha (idade média do primeiro uso 12,15 anos). A idade média de início do crack foi aos 13,4 anos. Apresentaram alta taxa de comorbidades psiquiátricas, com 81,8% dos pacientes com o diagnóstico de Transtorno de Conduta, 52,3% com Transtorno Opositor Desafiante e 44,3% apresentando Transtorno de Déficit de Atenção e Hiperatividade (todas essas com p<0,001 em relação aos controles). Evidenciamos consequências graves do uso de drogas em todas as áreas da vida (como altos escores de comprometimento nas áreas de uso de substâncias, escola, família e psiquiátrica do questionário Teen Addiction Severity Index – T-ASI). O tempo médio decorrido entre o início de uso de drogas e o início de uso de crack foi de 2,53 anos (DP=1,96). Através do modelos de regressão de Cox, encontramos os seguintes preditores de progressão precoce ao uso de crack: idade de início de uso de qualquer substância e idade no momento da internação. Dando continuidade ao tema desta tese, o segundo artigo abordou os níveis de quatro potenciais biomarcadores para o uso de crack na adolescência. São eles: o Fator de crescimento derivado do cérebro (BDNF), as interleucinas 6 e 10 (IL-6 e IL-10), e as substâncias reativas ao ácido tiobarbitúrico (TBARS). Foram medidos, na mesma amostra de pacientes do primeiro estudo, em dois momentos: no dia seguinte à internação e no momento da alta. Também foram medidos em controles. As análises foram controladas para possíveis confundidores. Encontrou-se níveis significativamente mais baixos de BDNF entre os pacientes (16,61; DP=2,06), em relação aos controles (25,12; DP=175) (p<.001). Os níveis de TBARS estavam significativamente elevados entre os pacientes (25,07; DP=21,83), em relação aos controles (16,46; DP=23,25) (p=0,005). Os níveis de IL-6 estavam significativamente elevados entre os pacientes (577,34; DP=156,11) em relação aos controles (151,91; DP=64,7) (p=0,027). Os níveis de IL-10 estavam significativamente elevados entre os pacientes (383,86; DP= 97,24), em relação aos controles(109,89; DP=40,89) (p=0,025). Concluiu-se que adolescentes internados por uso de crack apresentam problemas em diversas áreas da vida, altas taxas de comorbidade e alterações em biomarcadores específicos relacionados com inflamação e estresse oxidativo. Sugere-se que esta população necessita cuidados especiais para evitar comprometimentos maiores. / This thesis is about crack-cocaine dependence in adolescence, a subject that is of high clinical interest due to the severity of the clinical presentation of those patient and the challeng of providing effective treatment. Both studies presented were conducted with a ample of 90 adolescents, from 12 to 18 years age, of both sexes, who had been admitted for problems related to crack-cocaine use in two impatient treatment units for adolescents. In the city of Porto Alegre, south of Brazil. This consecutive sample was collected from May 2011 to November 2012. A control sample of drug-naïve adolescents has been recruited in a low income neighborhood. As results of this first study, it was observed that adolescents admitted for crack-cocaine use had a mean 15.6 years of age, and most of them (85.5%) were male. All patients had used at least one other psychoactive substance before initiating crack-cocaine use: 61.4% had used tobacco (mean age of first use=11.6 years), 44.3 had used alcohol (mean age of first use=12.4 years), and 54.5 5 had used cannabis (mean age of first use = 12.15 years). Mean age of crack-cocaine initiation was 13.4 years. Patients had high rates of comorbid conditions, with 81.8% of patients having a lifetime diagnosis of Conduct Disorder, 52.3% with Oppositional Defiant Disorder, and 44.3% with Attention Deficit and Hyperactivity Disorder. All the comorbid conditions found were more prevalent in the patients group than in controls. The T-ASI interview showed severe consequences of crack-cocaine abuse in all life areas. Mean time from first substance use to first use of crack cocaine was 2.5 years. By using Cox regression moles, we found that predictors to early crack-cocaine initiation were: age of first substance use and current age. The second study is about four potential biomarkers of crack-cocaine use in adolescence. They are the Brain Derived Neurotrophic Factor (BDNF), the interleukins 6 and 10 (IL-6 and IL-10), and Thiobarbituric acid reactive substance (TBARS). They were measured in peripherical blood, in the same sample of patients from the first study, at two moments: the day after admittance and in the day of discharge from the unit. They were also measured in controls. They were compared using Generalized Estimating Equations, and analyses where controlled by possible confounders. BDNF levels were lower in patients than in controls. TBARS levels were higher in patients than in controls. IL=6 was higher in patients than in controls. IL-10 was higher in patients than in controls. In conclusion, adolescents that are admitted to inpatient units for crack-cocaine-related problems show high degree of impairment in multiple life areas, high rates of comorbid conditions and alterations in biomarkers related to inflammation and oxidative stress. It is suggested that this population need special care to prevent further impairments.

Les données d’enquêtes jouent un rôle primordial dans la production scientifique en sciences sociales. Cependant, la présence de biais dans les données au moment de la collecte, y compris les erreurs de réponse et de non-réponse, pourrait affecter la fiabilité des résultats obtenus. Bien que le problème des erreurs de réponse et de non-réponse soit largement discuté, la littérature existante s’intéresse peu aux mécanismes par lesquels ces erreurs influencent les mesures estimées. Par conséquent, l’objectif principal de cette thèse est d’apporter une contribution méthodologique à la compréhension des estimations d’enquêtes en présence de données erronées et manquantes, en déterminant leur part explicative dans les modèles estimés. À l’aide de deux mesures répétées de l’âge de la première consommation de substances psychoactives recueillies par l’ELNEJ (1994-2009) auprès de jeunes Canadiens à l’âge de 12-13 ans, puis à l’âge de 14-15 ans, cette thèse par articles visait à étudier : (1) les types d’incohérences ou de biais imputés dans la deuxième mesure comparativement à la première (mesure de référence) et ainsi déduire les caractéristiques des répondants dont les déclarations sont incohérentes (Article 1); (2) l’impact de ces biais dans la détermination des prédicteurs de la consommation précoce, c’est-à-dire à l’âge de 13 ans ou moins (Article 2); et (3) leur impact sur la prédiction de la consommation à l’âge de 16-17 ans en fonction de l’âge d’initiation (Article 3). L’impact du biais est déterminé en corrigeant (1) le biais de sélection dû à la censure des répondants ayant des déclarations incohérentes de l’échantillon de l’étude, lorsque l’âge de l’initiation est la variable dépendante dans le modèle estimé (Article 2) ou (2) le biais d’endogénéité dû à la présence de valeurs erronées et manquantes dans l’âge d’initiation, lorsque l’âge d’initiation est une variable explicative dans le modèle estimé (Article 3). Le premier article révèle qu’au deuxième passage de l’enquête, les jeunes de 14-15 ans ne fournissent pas nécessairement des âges d’initiation cohérents avec ceux fournis lorsqu’ils avaient 12-13 ans. La proportion d’incohérence enregistrée n’est pas négligeable; elle est de 43 % pour l’alcool, 33 % pour le tabac et 32 % pour la drogue. Ces jeunes sont susceptibles de déclarer des âges d’initiation plus tardifs ou d’omettre l’expérimentation antérieure (biais télescopique vers l’avant: BTA), des âges d’initiation plus hâtifs (biais télescopique vers l’arrière: BTR), et peuvent également ne pas déclarer leur initiation à au moins un des deux cycles d’enquête (Biais non-déclaré : BND). Les résultats de la régression multinomiale montrent que le risque de détecter ces biais n’est pas le fruit du hasard; il varie en fonction des caractéristiques socio-démographiques et personnelles des répondants, notamment le genre, la structure de la famille et la région de résidence. En raison du biais dans la deuxième déclaration de l'âge d'initiation, le deuxième article démontre que l'identification des groupes à risque de consommation précoce est affectée par le potentiel de biais au sein de ces groupes. En utilisant l'approche de Heckman, il a été conclu que les âges d'initiation déclarés plus hâtifs (plus tardifs) génèrent une surestimation (sous-estimation) des risques de consommation précoce (à 13 ans et moins) dans les groupes les plus susceptibles de fournir des âges biaisés vers l'arrière BTR (biaisés vers l'avant BTA). Cependant, ces risques sont sous-estimés dans les groupes qui n’ont pas déclaré leur âge d’initiation lors du premier passage de l’enquête (BND). Ceci indique que pour ces groupes, l’âge d’initiation qui n’a pas été déclaré lors du premier passage est probablement un âge précoce (autour de 12-13 ans). Le troisième article conclut que les biais attribués à l'âge de l'initiation affectent l’estimation de la relation entre l’âge d’initiation et la fréquence de consommation à l'âge de 16-17 ans. Le fait que cette relation soit surestimée ou sous-estimée dépend spécifiquement du type de biais et de sa corrélation avec la consommation ultérieure. Enfin, cette thèse tente de fournir des preuves empiriques mettant en évidence le biais de réponse et de non-réponse comme une source d'information supplémentaire qui caractérise l'échantillon de l'étude et qui sa propre part explicative dans les modèles estimés. La validité des données d'enquête est donc d'une grande utilité pour la validité des résultats des études. / Survey data play an essential role in scientific production in the social sciences. However, the presence of bias in the data at the time of collection, including response and non-response errors, could affect the reliability of the results obtained. Although the problem of response and nonresponse errors is widely discussed, little attention has been paid in the existing literature to the mechanisms by which these errors influence the estimated measures. Therefore, the main objective of this thesis is to make a methodological contribution to the understanding of survey estimates in the presence of erroneous and missing data, by determining their explanatory part in the estimated models. Using two repeated measures of age at first substance use collected by the NLSCY (1994-2009) from Canadian youth, when they were 12-13 years old and again at 14-15 years old, this article-based dissertation aimed to investigate: (1) the types of inconsistencies or biases imputed in the second measure relative to the first (baseline) measure and thus infer the characteristics of respondents whose reports are inconsistent (Paper 1); (2) the impact of these biases in determining predictors of early use, i.e., at age 13 and younger (Paper 2); and (3) their impact on predicting use at age 16-17 as a function of age of initiation (Paper 3). The impact of bias is determined by correcting for (1) selection bias due to censoring of respondents with inconsistent reports from the sample, when age of initiation is the dependent variable in the estimated model (Paper 2) and (2) endogeneity bias due to the presence of erroneous and missing values in age of initiation, when age of initiation has an explanatory variable in the estimated model (Paper 3). The first article reveals that at the second survey round, 14–15 years old do not necessarily provide ages of initiation consistent with those provided when they were 12-13 years old. The proportion of inconsistency recorded is not negligible; it is 43% for alcohol, 33% for tobacco, and 32% for drugs. These youth are likely to report later ages of initiation or omit previous experimentation (telescopic forward bias: BTA), earlier ages of initiation (telescopic backward bias: BTR) and may also fail to report initiation on at least one of the two survey rounds (unreported bias: BND). Multinomial regression results show that the risk of detecting these biases is not random; it varies with respondents' socio-demographic and personal characteristics, including gender, family structure, and region of residence. Because of the bias in the second report on age of initiation, the second article demonstrates that the identification of groups at risk for early use is affected by the potential for bias within these groups. Using Heckman's approach, it was concluded that earlier (later) reported initiation ages generate an overestimation (underestimation) of the risks of early use (at age 13 and younger) in the groups most likely to provide backward-biased (forward-biased) ages. However, this risk is underestimated in groups that did not report their age of initiation at the first survey round (BND). This indicates that for these groups, the age of initiation that was not reported in the first round is likely to be an early age (around 12-13 years). The third paper concludes that biases attributed to age of initiation affect the estimate of the relationship between age of initiation and frequency of use at age 16-17. Whether this relationship is overestimated or underestimated depends specifically on the type of bias and its correlation with later use. Finally, this thesis attempts to provide empirical evidence highlighting response and nonresponse bias as an additional source of information that characterizes the study sample and has its own explanatory part in the estimated models. The validity of survey data is thus of great benefit to the validity of studies results.