In this commentary, we use comorbidity to refer to the co-occurrence of two diagnoses at the same time for a single patient, independently of etiological and/or pathway considerations (see Banaschewski et al. Reference Banaschewski, Neale, Rothenberger and Roessner2007). In children's mental disorders, comorbidity plays an even greater role than in adults: about 80% of children develop at least one comorbid condition compared with 45% in adults (Cramer et al., target article; Freeeman et al. 2007; Gillberg et al. Reference Gillberg, Gillberg, Rasmussen, Kadesjö, Söderström, Raståm, Johnson, Rothenberger and Niklassen2004). However, parents and physicians prefer to restrict the necessary multimodal treatment regime to only one disorder (i.e., a clinical entity). Therefore, in children it is highly important to determine whether or not the coexistence of two mental disorders represents a separate clinical entity (i.e., true comorbidity) with the possibility of a specific treatment (Banaschewski et al. Reference Banaschewski, Neale, Rothenberger and Roessner2007; Cramer et al., target article). To approach this issue, the concept of “overlapping” or “bridging” psychopathology between two problem areas comes into play. For example, the coexistence of tic disorders (TDs) and attention-deficit hyperactivity disorder (ADHD) presents an important practical problem for diagnosis and treatment in child psychiatry (Rothenberger et al. Reference Rothenberger, Roessner, Banaschewski and Leckman2007). First, some direct overlap exists between symptoms of hyperactivity, impulsivity, and inattention, and, second, there is some indirect overlap because TD and ADHD are both associated with the same disorders (e.g., Asperger Syndrome, anxiety, depression, obsessive-compulsive disorder [OCD]).

Thus, it is of practical importance to disentangle which symptom dimensions are actually overlapping and which ones are distinct between TD and ADHD. And does an overlap of symptoms indicate that TD+ADHD represents a separate clinical entity?

In order to answer these questions, we think that a 2×2 factorial design, analyzing “pure” groups and their combination (i.e., without further comorbid conditions like anxiety or depression) is more appropriate and practically relevant than a network approach, because the factorial design creates a clearer and clinically adaptable picture for guiding treatment. So far, several studies using this approach have shown that TD+ADHD is not a clinical entity and that its comorbidity needs to be explained within the framework of an additive model (Rothenberger et al. Reference Rothenberger, Roessner, Banaschewski and Leckman2007). Using the Child Behavior Checklist as a psychopathological screener, both TD and ADHD showed similar scores for the anxiety/depression, schizoid/obsessive, and social withdrawal scales, whereas only ADHD reached statistical main effects for aggression, delinquent behavior, and attention and social problems (Roessner et al. Reference Roessner, Becker, Banaschewski and Rothenberger2007b). These results also clarify that internalizing problems should not be neglected in TD+ADHD, nor in primarily externalizing disorders like ADHD (see also, Sobanski et al., in press). Further, the results underline that boundaries between both mental disorders are partly fuzzy while also displaying distinct features, a conclusion which, methodologically, might also be reached by using the network approach. Applying both approaches to the same data set could be helpful in order to detect their scientific strengths and weaknesses. At least, both statistical approaches stress that bridging symptoms, the correlation of latent variables and interacting networks, must be carefully controlled for in clinical trials since they are possible confounders of mental disorders.

Further, relations between symptoms may vary according to a patient's age, his/her stage of development, gender, and changes in other symptoms. To provide a long-term dynamic view of comorbid psychopathology, it would be of great interest to evaluate the advantages of a network approach in developmental psychopathology research when investigating the stability and change of certain disorders, symptom clusters, and/or behavioral dimensions along the life-span. For example, there is a high stability of ADHD symptoms from ages 9–14 (Larsson et al. Reference Larsson, Larsson and Lichtenstein2004). Subsequent follow-up at ages 16–17 has indicated that hyperactivity-impulsivity decreases, while inattention remains the same (Larsson et al. Reference Larsson, Lichtenstein and Larsson2006). Changes in ADHD subtypes and comorbidity have also been reported (Lahey et al. Reference Lahey, Perlham, Loney, Lee and Willcutt2005; Steinhausen et al. Reference Steinhausen, Rothenberger and Döpfner2010). So far, logistic regression analyses have mainly been used and have already provided some practically useful answers in regard to the development of comorbidity (Roessner et al. Reference Roessner, Becker, Banaschewski, Freeman and Rothenberger2007a; Wanderer et al., submitted).

In their introduction (target article, sect. 1), Cramer et al. ask whether there is a general order in which people develop a particular disorder first and then another next. We would add here that there might be a third disorder, as can be seen in daily clinical practice in child psychiatry when faced with psychosocial impairment caused by comorbidity.

For example, TD has a childhood onset and follows a remitting course into adulthood. Comorbid ADHD commonly antecedes TD and remains after tic remission, whereas comorbid OCD starts later than tics and often remains longer than ADHD. Unfortunately, developmental psychopathology with more than one comorbidity has received minor attention to date, and it would be a challenge for a network approach to disentangle this complex issue. Using logistic regression, Roessner et al. (Reference Roessner, Becker, Banaschewski, Freeman and Rothenberger2007a) analyzed, in children and adolescents with TD, the impact of comorbid ADHD diagnosis on the frequency of additional (third) comorbidities during development. The main finding was that ADHD generally increased comorbidities in all age groups (especially the externalizing problems), with the exception of adolescent OCD and anxiety disorders. In addition, there was a higher annual rate of change in emotional problems compared to externalizing disorders. This indicates the importance of an increase in internalizing problems with age for comorbidity research in youngsters and underlines that multiple comorbidity is a relevant issue that needs to be dealt with.

It is hard to imagine how a network approach could solve this problem in a scientifically better and practically more useful way, since too many interactions over time may reduce clarity. On the other hand, these manifold interactions between symptoms, changing over time, reflect reality. Hence, a network approach can be successful in this matter only if a statistical way is found to reduce complexity without loosing validity.

In sum, the network approach should be evaluated in developmental psychopathology mainly for theoretical reasons while its practical value for a better understanding and treatment of young patients has yet to be proven. Such a research perspective should include not only the level of clinical symptoms, but also the area of neuropsychological and neurobiological parameters. This may allow us to define new endophenotypes, which might in turn help to further elucidate the comorbidity pathway from genes to behavior.