Among youth who did show both disorders, the retrospective age of onset data showed just more than half reported that an ODD symptom emerged prior to their first CD symptom. This result must, of course, be interpreted in the context of known inaccuracies in dating symptom onsets retrospectively (Angold, Erkanli, Costello, & Rutter, 1996). Our pathway analyses showed that the assumed transition from full disorder ODD to CD diagnosis was not common. It was no more common than concurrent onset or the pathway where CD onset prior to ODD. We considered a number of GSMS design features that might have contributed to this pattern of results. First, we checked whether selective attrition might have underestimated ODD development to CD but found no evidence to support this effect. Participation rates were high across all study waves, and did not differ in the ODD never CD and ODD→CD pathways. Second, the CAPA uses a three-month reporting period. As a result, annual assessments may miss some periods of disorder and therefore underestimate prevalence in the population. This may be particularly relevant for CD
