Is ODD a developmental precursor to CD?
Criteria for ODD or CD were met at one or more study contacts by 19.1% of boys and 9.6% of girls between the ages of 9 and 16 years: CD was diagnosed on at least one occasion in 8.6% of youth (3.7% girls, 13.2% boys), and ODD in 9.7% (7.8% girls, 11.6% boys). We used a variety of approaches to assess whether ODD constituted a developmental precursor to CD, beginning with comparisons of age at first symptom onset. We found that ODD disorder was generally reported to onset earlier in development than CD. Median age at first symptom onset for cases of ODD was 2.5 years. CD first symptom onset was a full three years later, at 5.5 years. Despite these marked differences in overall ages at symptom onset, ODD symptoms preceded CD symptoms in only 56.0% (95% Confidence Interval (CI): 43.6, 67.4) of CD cases.
Next, we tested whether ODD predicted CD at both symptom and diagnostic levels. For the symptom analyses we used ODD symptoms at wave 1 to predict maximum CD symptom count at any subsequent childhood or adolescent study wave. The mean wave 1 ODD symptom count was .8 (SD = 1.3 Range: 0 - 8), and the mean maximum CD symptom count over subsequent waves was .8 (SD = 1.2, range 0 – 9). In an ordinal logistic regression model controlling for age and sex, ODD symptom count at wave 1 was a significant predictor of later maximum CD symptom count (OR for a 1 symptom increase in ODD symptoms=1.4, 95% CI: 1.3, 1.6, p<.001). This effect remained strongly significant (OR=1.2, 95% CI: 1.1, 1.4, p=.005) when wave 1 symptom counts for CD, generalized anxiety, depression and hyperactivity were included in the model. The ODD to CD prediction was similar for boys and girls (interaction p=.512).
Parallel analyses of diagnostic categories used ODD at wave 1 as the predictor, and CD at any later childhood/adolescent wave as the outcome. Controlling for age and sex, a wave 1 ODD diagnosis predicted later CD diagnosis (OR=7.9, 95% CI: 3.6, 17.6; p<.001). After controlling for baseline levels of CD the relationships were more complicated. Whereas ODD functioned as a clear risk for later CD in boys, effects in girls appeared to be attenuated by a high rate of comorbidity between the two disorders at first assessment, and strong CD persistence. After controlling for wave 1 CD diagnosis, ODD predicted later CD in boys (OR=6.5, 95% CI: 1.8, 23.2; p=.004) but not in girls (OR=.5, 95% CI: .1, 2.3; p=.407) (interaction p=.011). Initial levels of ODD and CD were more strongly related in girls (OR=64.7, 95% CI: 19.6, 213.4; p<.001) than in boys (OR=7.6, 95% CI: 3.4, 16.8; p<.001) (interaction p=.003). Continuity from wave 1 CD to later CD was also stronger in girls (OR=54.8, 95% CI: 13.7, 218.8; p<.001) than in boys (OR=8.6, 95% CI: 3.2, 23.2; p<.001) (interaction p=.032). Controlling for ADHD at wave 1 did not substantively impact upon these results.
We extended these analyses to test whether prediction from ODD to CD in boys was affected by levels of sub-threshold CD symptoms. Confining the analysis to boys without CD at Wave 1, CD symptoms were significant additional predictors of a later CD diagnosis (OR for a 1 symptom increase: 3.0, 95% CI: 1.8, 4.9, p<.001). Despite these effects, however, ODD remained independently associated with increased risk of later CD (OR=5.7, 95% CI: 2.1, 15.2; p=.001).
The analyses to this point all suggested that ODD (whether assessed at the symptom or the diagnostic level) functioned as a precursor to, and risk factor for, CD. What they could not demonstrate, however, was how commonly this progression occurred. To assess this, we classified each child into one of six mutually exclusive onset ‘pathways’ reflecting the ordering of onsets of ODD and CD across study waves. As shows, many disruptive youth only ever met criteria for one of these disorders between late childhood and mid-adolescence, and transitions between the two disorders were less common than anticipated. In particular, of the children who met criteria for CD a substantial proportion did not have ODD at any study wave. In children with ODD no diagnosis of CD during the study was also common (60.5%, 95% CI: 49.7, 70.2). shows that the confidence intervals around percentages of children falling into each of the comorbid pathways overlap. Concurrent CD and ODD onset was the most common configuration. ODD→CD was slightly less common than the pathway where CD was diagnosed before ODD. Although the confidence intervals indicate this was not significantly different, there is certainly no evidence that ODD→CD was a particularly common pathway. ODD→CD appeared especially uncommon in girls, with only 3 following this pathway.
Ordering of ODD and CD diagnoses across the 9-16 age range.
As these findings were unexpected, we checked whether they might be an artifact, arising from differential non-response rate of children with ODD at Wave 1 at later study waves. We found no evidence for a pattern of this kind. Later participation rates were high across all the pathway groupings, and did not differ between the ODD→CD (89.7%, 95% CI: 85.4, 94.1) and ODD never CD paths (88.8%, 95% CI: 84.9, 92.7).
Next, to assess whether these transition patterns varied by age, we compared pathway classifications in the three age cohorts who entered the study at ages 9, 11 and 13 respectively. As shows, relative to other pathways to CD, the likelihood of simultaneous CD and ODD onset was stable across cohort and a CD diagnosis with no history of ODD became more usual with increasing age. The ODD→CD pathway, by contrast, became significantly less common across age, relative to the other pathways to CD (OR for a 1 year increase in age =.6, 95% CI: .3, .9, p=.019).
The CD→ODD pathway was also unexpected, so we explored it further in longitudinal analyses. In a model predicting future ODD, wave 1 ODD (OR=7.7, 95% CI: 3.1, 19.4, p<.001) and CD (OR=2.3, 95% CI: 1.1, 4.8; p=.034) were both significant predictors. When wave 1 ODD symptom count (OR=1.8, 95% CI: 1.5, 2.0, p<.001) was included as a predictor instead of ODD diagnosis, CD diagnosis no longer predicted future ODD (OR=1.1, 95% CI: .5, 2.3; p=.819).
Characteristics of youth with pure and combined CD and ODD diagnoses
The analyses thus far suggested that, although some disruptive youth meet diagnostic criteria for both ODD and CD across the late childhood/mid-adolescent years, many present with only one of these disorders. We explored whether these groups could be distinguished on the basis of symptom patterns, selected risk factors and rates of comorbid disorders in childhood and adolescence. Because numbers in the three sub-groups meeting criteria for both ODD and CD over the study period were small, we combined these cases (ODD→CD, ODD/CD and CD→ODD) into a single grouping for these comparisons.
shows mean levels of ODD symptoms were similar in youth with ODD only and combined ODD+CD, and mean levels of CD symptoms (total and split into aggressive and non-aggressive symptoms) differed little between cases with CD only and the combined pattern. As also makes clear, however, youth who only ever met full diagnostic criteria for one disorder (ODD or CD) during the study period nonetheless typically showed elevated levels of sub-threshold symptomatology for the other disorder. We used the retrospectively reported age at first CD symptom to classify CD cases into childhood and adolescent onset sub-types. Onset of first CD symptom was prior to age 10 for 86.3% of CD diagnoses (childhood onset CD) with the remaining 13.7% classified as adolescent onset CD according to DSM-IV. The mix of child and adolescent onset CD in the CD only (16.4%) and CD+ODD (10.5%) groups did not differ significantly (p=.572).
Symptomatology, risk factors and comorbidity of pure CD, pure ODD and children who displayed both CD and ODD (CD+ODD)
also shows levels of selected correlates for youth in the different pathways. These were analyzed as dependent variables predicted by pathway. Boys were significantly over-represented in the CD and CD+ODD groups, while ODD only cases showed no male preponderance. In terms of other social and familial risks the diagnosed groups showed significantly elevated rates by comparison with youth with no disruptive disorders. Contrasts among disorder groupings revealed few significant differences, however. The only exception to this pattern was parental scape-goating of the child, which was specifically associated with ODD. There were some tendencies that inadequate supervision was more strongly associated with ODD than CD but the statistical comparisons among the disorder groups were not fully supportive of this relationship.
The final section of shows levels of comorbid disorders in childhood and adolescence. Youth in all the disruptive disorder groupings were at increased risk of emotional disorders (anxiety and depression) and substance disorder relative to the no ODD/CD group, but did not differ significantly from each other. Risk for ADHD was also elevated in all disorder groups. There was some evidence that risk was particularly high in the CD+ODD group relative to the CD group, but the CD+ODD comparison with ODD group was non-significant.
Early adult outcomes
Psychiatric outcomes at ages 19 and 21 years are shown in . ASPD symptoms showed stronger associations with a history of childhood and adolescent CD than with prior ODD. Both the CD and CD+ODD groups had significantly higher levels of ASPD symptoms than youth with no history of CD/ODD, and did not differ from each other. Youth on the ODD only pathway, by contrast, were not at significantly increased risk. Early adult substance disorder appeared more common in the CD and CD+ODD groups relative to the ODD only and no CD/ODD groups, although the only significant difference was between the CD+ODD and the no CD/ODD groups.
Adult outcomes of pure CD, pure ODD and children who displayed both CD and ODD (CD+ODD)
In contrast, increased risk for anxiety disorders in early adulthood was confined to youth with a history of ODD. Both the ODD and ODD+CD groups were at significantly higher risk of anxiety than the no CD/ODD and CD only groups. Youth on the CD only pathway showed significantly lower rates of early adult anxiety disorders than those with no history of disruptive behavior problems. A somewhat similar pattern emerged for depression, although contrasts between the disordered groups were non-significant.
Symptom dimensions in ODD
In the final stage of the analyses we explored whether there were separable symptom dimensions within the ODD symptoms. In an exploratory factor analysis of ODD symptoms in the full sample at Wave 1, a two factor solution provided an acceptable fit to the data (RMSEA=.021). Computation of a three factor model failed, showing a negative estimated residual variance for the spiteful symptom which may indicate that too many factors had been specified. Factor loadings from the two factor model are shown in , following Promax rotation. The two factors correlated at .55, and high loading items were similar to the irritable and headstrong dimensions proposed by Stringaris and Goodman (2009b)
; the symptom of spitefulness loaded clearly on the headstrong factor in this solution.
Promax rotated factor loadings from an exploratory factor analysis of the ODD symptoms at wave 1.
We next examined whether pure cases of irritable and headstrong ODD could be identified within the DSM-IV diagnostic framework. The factor analysis identified 3 symptoms loading onto the irritability factor, and 5 symptoms loading onto the headstrong factor. Because DSM-IV criteria for ODD require 4 symptoms to be present, at least one headstrong symptom must be endorsed in all cases to reach the diagnostic threshold. Across the full 9-16 age-range, almost all cases that met ODD criteria also showed at least some level of irritability: only 3.3% had no irritability symptoms, 18.8% had one, 52.9% had two and 25.0% had three. These figures suggest few cases of ‘pure’ headstrong ODD exist within the DSM framework; most youth reaching the diagnostic threshold showed a mixed symptom pattern.
Given these findings, we took a dimensional approach to examine the predictive utility of the irritability/headstrong distinction. We summed items loading onto each factor at Wave 1 and examined the extent to which these two dimensions were differentially related to risk for behavioral and emotional disorders at subsequent childhood and adolescent study waves. Both symptom dimensions were standardized and age, sex and the Wave 1 level of the outcome variable were included as predictors.
Beginning with behavioral disorders, shows the headstrong dimension significantly predicted later CD. Prediction from irritability fell only a little below significance, and there was no evidence that effects of the two dimensions differed. In further analyses we examined prediction to aggressive and non-aggressive CD symptom counts. An ordinal logistic regression model showed that the irritability (OR=1.3, 95% CI 1.1, 1.6; p=.01) and headstrong (OR=1.4, 95% CI: 1.2, 1.7; p<.001) dimensions predicted maximum aggressive CD symptom count over future observations. Non-aggressive CD symptoms were also predicted by headstrongness (OR=1.4, 95% CI: 1.2, 1.6; p<.001) while irritability (OR=1.2, 95% CI: 1.0, 1.5; p=.058) fell just below significance. Given the overlap in the confidence intervals, however, there was no convincing evidence of differential prediction of aggressive and non-aggressive CD symptoms.
Prediction from a 1 SD increase in irritable and headstrong ODD symptom dimensions at wave 1 to later psychiatric outcomes up to age 16.
Links between the dimensions and later full ODD diagnosis were also very similar for the two dimensions. The headstrong dimension did, however, predict later substance disorder significantly more strongly than irritability (which had an OR below 1 in relation to substance disorder risk). We ran a further model including Wave 1 CD diagnosis as an additional predictor of substance disorder: headstrong symptoms remained highly significant predictors even with prior CD controlled (OR=1.4, 95% CI: 1.1, 1.7; p=.004) and CD diagnosis was non-significant (OR=1.3, 95% CI: .6, 2.7; p=.484). An opposite pattern emerged in relation to anxiety: irritability predicted risk for later anxiety diagnoses while headstrongness did not. Depression was significantly predicted by headstrongness; prediction from irritability fell short of significance, but comparisons between the two dimensions were non-significant.