Letter to the Editor

Recently, Kendler and colleagues (Reference Kendler, Myers, Torgerson, Neale and Reichborn-Kjennerud2007) replicated a high genetic correlation between major depression (MD) and generalized anxiety disorder (GAD) (1·0 in females, 0·74 in males). Interestingly, they also examined the role of neuroticism (N) and found that the genetic factors indexed by N contributed around 30% to the genetic correlation between MD and GAD, substantially less than the authors had expected. Hence, they considered their results as ‘somewhat disappointing’.

We like to argue that their results are not disappointing at all. As noted by the authors themselves, there was a large time lag of no less than 25 years between the assessment of N and the mental disorders. Furthermore, N was assessed only once. We think that the authors did not sufficiently take into account the time lag, probably because, apart from measurement error and random effects, they attribute the variance of N-measures (such as the N-scales from the EPQ and the NEO) mainly to stable personality traits. However, it can be argued that N-measures are best interpreted as mixed state-trait measures of psychological distress (Ormel et al. Reference Ormel, Rosmalen and Farmer2004): (1) Test–retest correlations of N-measures tend to decrease with increasing time intervals, even during adulthood (Watson & Clark, Reference Watson and Clark1984; Ormel & Rijsdijk, Reference Ormel and Rijsdijk2000; Caspi et al. Reference Caspi, Roberts and Shiner2005). (2) As many have noted, the item content of N-measures is very similar to that of psychological distress inventories, with one major difference: N-measures typically lack a well-defined time frame and contain vague qualifiers of frequency, intensity and duration. (3) Longitudinal studies have found that N-measures are strongly correlated with the long-term level of psychological distress (e.g. Costa & McCrae, Reference Costa and McCrae1980; Duncan-Jones et al. Reference Duncan-Jones, Fergusson, Ormel and Horwood1990; Lucas & Fujita, Reference Lucas and Fujita2000). (4) Structural equation modelling (SEM) of multi-wave N-measures covering 18 years in an adult general population sample found a 1-year autoregression of 0·967 and an estimated 25-year test–retest correlation of 0·432 (Ormel & Rijsdijk, Reference Ormel and Rijsdijk2000). Collectively, these findings suggest lasting change in N-scores in some individuals, perhaps as a result of steeling and sensitizing experiences.

The phenotypic correlation between N and MD and between N and GAD in Kendler et al.'s study is approximately 0·25 and for 75% due to the genetic correlation component (calculated from Fig. 3). This is not disappointingly weak but rather substantial given the fact that, attenuated for measurement error, the estimated 25-year longitudinal correlation of N with itself amounted to 0·43 in the Ormel & Rijsdijk (Reference Ormel and Rijsdijk2000) study.

The implication for molecular genetic studies would be that individuals with MD and/or GAD and/or high N-scores can be pooled, provided multiple assessments of N that are not too far apart in time from the assessment of mental disorder. Pooling will increase the statistical power of association and linkage studies searching for the genes involved in depression and generalized anxiety.