LAUSR

In his Review, Parker questions the usefulness of a staging approach for bipolar disorders (BD) given the state of knowledge and warns of the risk of overdiagnosis and overtreatment of early stage presentations. Parker raises the point that childhood anxiety and sleep disorders are common, making it difficult to ascertain when these presentations represent antecedents to BD and that by identifying individuals at increased risk, unnecessary worry and possible stigma may ensue. Although concerns about going beyond the evidence are valid and core to the values of medicine to do no harm, a developmental approach to understanding the evolution of BD has been extremely informative, advancing both clinical practice and research. The Canadian Flourish longitudinal offspring study started in 19971 in direct response to questions from BD parents about the risk of illness in their children. At the time, there was insufficient data to inform accurate individual risk prediction together with an appreciation that, given the substantial genetic and phenotypic heterogeneity, the risk would vary significantly between families and among individuals within families. An unexpected finding from the Flourish offspring study was the elevated rate of childhood anxiety and sleep disorders in high risk children compared with children of well parents. The highrisk children came from mostly intact, middleclass families, with only one BD parent (i.e., the other parent had no lifetime history of mental illness). With the longer observation of more children over the peak risk period, we found evidence of an increased risk of major mood disorder of about 2.5fold in highrisk offspring with, compared to those without, childhood anxiety and sleep disorders.2 This finding has since been independently replicated. In contradiction to concerns raised by Parker, parents found it helpful and reassuring to understand that the risk of their child(ren) developing BD was much lower than anticipated and that families could be signposted to lowintensity support when first indicated— which likely explains the low (under 10%) attrition over decades of observational research. Several groups around the world have invested in longitudinal studies of children at familial risk of mood disorders, and as a result, the approaches and precision of individualized risk prediction have advanced; taking into account heterogeneity, and being more honest about statistical uncertainty of predictions, largely related to sample sizes.3,4 One concern raised by Parker is the possibility that variables used to assess the risk of BD may have non causal or indirect relationships with the onset of illness. While understanding causal mechanisms is important and may inform future directions in prevention and treatment, it is not necessary that variables used to assess risk be strictly causal in nature. In practice, the causal variable may be difficult to measure for various reasons; however, a proxy variable, confounded by the truly causal variable, will still contribute useful information in assessing future risk. Further, by taking a developmental approach to mapping psychopathology in children at confirmed familial risk, we have been able to develop a refined conceptual framework to advance progress in both clinical practice and research (Figure 1).5 The staging approach has provided evidence that while not all highrisk children who develop BD will manifest each and every