Authors' reply

We thank Dr Mosley for his interest in our paper. As noted in the discussion to the paper, we agree with him that the relationship between physical disease and mood disorder is complex. However, we do not agree regarding the other points raised.

Our two-group statistic presented in the paper is quite appropriate. We compared progression of white matter changes between those with and without depression at the 3-year point, and demonstrated with a high significance that those with depression had a greater progression of white matter changes prior to depression. If, as suggested by Mosley we dichotomise the white matter change progression, then high white matter change progression is still significantly associated (P<0.01) with both depression at 3 years, and any depression over the 3 years.

Patient drop-out is a problem in any longitudinal study, and our drop-out rate is fairly typical for the study population. However, participants who drop out are typically less well than those who do not, and it is likely that those developing depression are more likely to drop out. The effect of drop-out is thus more likely to have weakened the association between white matter changes and depression than otherwise.

We agree that executive dysfunction is associated with depression. However, it is also associated with white matter changes, ¹ and the purpose of the regression analysis was to investigate whether depression could be accounted for by factors other than white matter changes, rather than attempting to identify the best risk factors, and hence we used the Mini-Mental State Examination as a well-recognised measure of general cognitive ability.

When combined with the previous findings in our cohort ² demonstrating that baseline white matter changes predict incident depression, we feel confident that, in this cohort at least, our findings robustly demonstrate that vascular disease as measured by white matter changes is a risk factor for depression.