The comments of Hynes & McCune raise pertinent questions. As they point out, it is possible that sexual abuse in childhood might have influenced the onset of juvenile depression, and also the likelihood of adult depression recurrence in our sample. In this study, we have collected data on sexual abuse, using both a review of medical charts at the time of Maudsley attendance and from adult interviews based on the Childhood Experience of Care and Abuse (CECA) measure. The effect of sexual abuse in childhood on patterns of adult depression recurrence will be investigated in the next analyses of this data-set, with particular attention given to differential risk processes according to childhood comorbidity.

Regarding comorbid ADHD as a risk factor for adolescent depression, particularly in the depression group with comorbid conduct disorder, we found a significantly increased rate of ADHD in the comorbid group, as we reported (Reference Fombonne, Wostear and CooperFombonne et al, 2001a , Table 2). Yet, it is plausible that the rate of ADHD in this sample was underestimated as many cases were ascertained before ADHD or hyperkinetic disorders were fully recognised as valid diagnostic entities. Nevertheless, our findings suggest that it is possible that ADHD might have been implicated in the development of conduct symptoms in the comorbid group although, because of the small sample size and likely underestimation of ADHD in that group, we cannot test for the specific contribution of (untreated) ADHD in the onset and recurrence of depression.

We had provided explicit data on the use of tricyclic antidepressant drugs during childhood years and found that the rate of prescriptions of these drugs was significantly higher in the non-comorbid group than do the comorbid group (48.4% v. 30.2%, P=0.032; see Reference Fombonne, Wostear and CooperFombonne et al, 2001a ). Most of these prescriptions were for amitriptyline and relied on dosages much lower than those considered appropriate by today's standards. Although the rate of antidepressant use was lower in the comorbid group, antidepressants were nevertheless often prescribed in that group too. Obviously, we could not assess whether or not use of tricyclic medications in that sample influenced long-term out-comes, since our study relied on an observational design. The interesting aspect of these data was to point to the frequent use by practising child psychiatrists of antidepressant drugs (irrespective of their known efficacy) in this sample of youths with depression assessed in the 1970s at a time when child and adolescent depression was largely ignored in professional training and in the literature. Furthermore, the data indirectly validated our diagnostic procedures.

Most of the comments by Hynes & McCune raise questions about the mechanisms underlying recurrence of depression in adulthood following a first episode in childhood or adolescence. The findings of our study (Fombonneet al, Reference Fombonne, Wostear and Cooper2001a , b) indicated that relapse rates were similar, irrespective of the presence of comorbid conduct disorder in childhood. This result is important in its own right as it refutes previous hypotheses that depression, when occurring in the context of conduct disturbances, reflected mostly local psychosocial circumstances and was not associated with long-term heightened risk of affective disorders in adulthood.

This study was designed to assess mechanisms underlying recurrence of depression in adult life and further reports will address the role of early childhood experiences (such as sexual abuse), life events, family history and individual psychological characteristics on the patterns of adult depressive recurrence. It could well be that, in spite of having similar rates of relapse in adulthood, the mechanisms of depressive recurrence differ for the two groups included in this study, according to childhood comorbidity.