Fewer than twenty years ago, developmental psychopathologists and psychiatrists believed that there was little overlap between externalizing childhood psychiatric disorders, such as oppositional defiant disorder (ODD), and internalizing disorders, such as depression. There is now ample evidence from my own research (Boylan et al., 2007, 2010) and research of others (Copeland et al., 2009; Nock et al,.2007; Burke et al., 2010; Stringaris and Goodman, 2009), that childhood ODD shows significant concurrent and prospective associations or comorbidity with later depression. This discovery is changing the way we treat youths with ODD, and think about the causes of the disorder.

Despite these advances, important questions remain unanswered: Do all youths with ODD develop depression or will this only occur in sub-groups? What are the distinguishing characteristics of children belonging in one or other subgroups? The answers to these questions have profound clinical implications. If all youths with ODD develop depression, then perhaps we should begin prophylactic depression treatment, even before symptoms occur. On the other hand, if we can identify sub-groups of youths with ODD that develop depression, then prevention and treatment protocols can be applied in a more sophisticated and effective manner.

Based on my earlier work and the works of others which have demonstrated ODD symptoms cluster in various ways in the population , I propose that there are sub-groups of youths with ODD that will develop depression over time and the ability to identify and validate these groups is of substantial importance.

Study Aims:

(1) To determine whether oppositional symptom sub-groupings identified in previous factor analytic studies can be linked to groups of children who are oppositional.

(2) To determine whether these groups show differential risk for developing depression over time. If there is not strong evidence for person-based groupings, we will test how oppositional behaviours are associated with risk of depression across childhood.

Research objectives:

(1) To assess whether oppositional symptom subgroups identified by others (ie. a predominantly negative emotion and a predominantly antisocial symptom type) can be reliably identified in individual children using cross sectional and longitudinal data from an epidemiologic cohort sample.

(2) To determine if one, or more, of these groups is differentially associated with depression in late childhood and adolescence.

(3) To assess whether membership in these groups can be predicted on the basis of neurobiological and psychosocial risk factors that are associated with the ability to regulate negative emotions, in each of the following domains:

The child themselves (physiology of stress response, early life temperament, cognitive abilities, aggression, anxiety comorbidity)

The family context (family functioning, early attachment, parent discipline, family socioeconomic and family structure, parent mental health)

The peer and social context: peer relationships, school connectedness, social skills

Proposed methods:

We propose to use previously-collected questionnaire data for children from the ALSPAC study. The main outcome variables are the ODD and major depression and dysthymia symptoms from the DAWBA (at ages 7, 10 & 16) as reported by parents. Additional covariables, predominantly from questionnaire data, will be requested to distinguish the ODD groups and or predict relationships between ODD symptoms and depression based on study hypotheses. Longitudinal data on the DAWBA responses for the entire cohort will be requested to understand the impact of sample loss over time as complete data on DAWBA assessment periods is not necessary to conduct the analyses.

Analytic strategy:

Objective 1: A two step approach will test for the presence of groups of children with i) different types of ODD symptoms and ii) the stability (or reliability) of these prototypes/groups over time. The clustering of ODD symptoms will be identified from the DAWBA ODD symptoms data using factor analysis, followed by a confirmatory method (longitudinal confirmatory factor analysis). In step 2, we will test whether these (?#) identified ODD groups are developmentally distinct in that they can be distinguished based on the course of their symptom trajectory over time. To do this, we will use the group-based trajectory approach (with the Proc Traj procedure in SAS) which estimates the number and shape (course) of longitudinal trajectory groups in the sample, the proportion of children in each group and a probability for each child to belong to each group. Trajectories will be estimated for each ODD symptom cluster that is identified. Following this step, an extension of Proc Traj allows trajectory groups of one symptom cluster to be linked probabilistically to groups of another symptom cluster as joint trajectories to describe how the course of "comorbid" symptoms potentially overlap within individual children. As many children are likely to have both clusters of ODD symptoms over time, but most will have one or the other cluster (ie not have symptom overlap), assessing joint trajectories will establish how common these "comorbid" and non-comorbid children are. If comorbidity between ODD clusters is common, this suggests that there are not distinct ODD subtypes or subgroups.

In summary, the outputs of these analyses will identify: 1) are there likely ODD clusters? 2) how common are they? 3) how distinct are they from each other in terms of developmental course or how commonly do they overlap?

Objective 2: The identified joint oppositional symptom trajectory groups will be used to predict youth self-report of depression as an outcome at age 16 (measured as the youth DAWBA) using an extension in Proc Traj for this purpose. The output here will consist of a regression coefficient describing the strength of association between the various identified joint trajectory with the outcome of depression.

Objective 3: The same joint trajectory groups will be compared for their differential association with covariables (based on hypotheses which we do not outline here given space issues) as a means of exploring how they may differ etiologically from each other and how they may be differentially associated with risk of depression. Logistic regression will be conducted to examine if the prevalence rate of each covariate differs across trajectory groups.