To our knowledge, this study provides the first examination of depressive symptoms and regional brain volumes within the context of longitudinal volume changes in older adults. Consistent with predictions, we observed both cross-sectional and longitudinal associations between depressive symptoms and frontal and temporal brain volumes. Higher average depressive symptoms, which may index more chronic symptoms, showed significant cross-sectional associations with smaller frontal and temporal regional volumes, including left temporal grey matter, the cingulate gyrus and orbitofrontal cortex, though effects were only observed at older ages. Furthermore, we observed longitudinal associations between depressive sypmtoms and left frontal white matter across the range of ages in our sample, with a faster rate of volume decline associated with higher depressive symptoms. In contrast to these findings, depressive symptoms were not associated with smaller hippocampus volumes or longitudinal temporal lobe volume decline.
These results are consistent with frontolimbic theories of depression43–45
and with previous investigations demonstrating associations between subthreshold depressive symptoms and smaller prefrontal volumes,8,9,27
but not hippocampus volumes.27
Our findings also extend previous results by documenting the impact of age on the association between depressive symptoms and brain volumes, and by demonstrating an association between depressive symptoms and longitudinal volume decline in left frontal white matter. The observation that age modifies the association between depressive symptoms and brain volumes is consistent with previous neuro-imaging studies that have found interactive effects of age and depression on cerebral blood flow and metabolism.46,47
These results also parallel the finding that the adverse impact of depression on some cognitive functions increases with advancing age.48,49
Perhaps older adults, who are already vulnerable to neural and cognitive changes as a function of advancing age, are particularly susceptible to the effects of depressive symptoms.
Somewhat surprisingly, the predicted acceleration of longitudinal volume decline in individuals with higher depressive symptoms was only observed for left frontal white matter. Longitudinal stability of hippocampus50
and orbitofrontal cortex51
volumes have been demonstrated in young adults with clinical depression over 1-year and 3-month follow-up, respectively. Nonetheless, we expected to find a faster rate of longitudinal decline in association with depressive symptoms owing to the older age of our participants and the longer follow-up interval. Additional longitudinal studies are needed to test the possibility that clinical depression is associated with volume decline that is greater in magnitude and occurs over a shorter time interval compared with subthreshold depressive symptoms. Our finding of longitudinal effects for left frontal white matter in the absence of significant effects for other brain regions suggests that frontal regions are particularly susceptible to subthreshold depressive symptoms. Furthermore, our finding of cross-sectional associations between sub-threshold depressive symptoms and frontal lobe structures, but not the hippocampus, suggests that frontal regions may be more vulnerable to subthreshold depressive symptoms than the hippocampus and possibly other temporal structures.
Higher depressive symptoms were associated with enlarged volumes in frontal white matter, right temporal white matter and left temporal grey matter. Although previous studies have shown an association between depressive symptoms and increased white matter lesions,29,52,53
our finding of enlarged white matter volumes is not attributable to the presence of white matter lesions in individuals with higher average depressive symptoms because white matter hyperintensity ratings available for this sample54
did not predict either depressive symptoms or regional brain volumes in supplemental analyses in a subset of the sample. Although volume enlargement has not been previously reported in subthreshold depression, previous studies have documented increased regional volumes in patients with major depression. For example, compared with controls, patients with major depression were found to have enlarged amygdala55,56
and corpus callosum57
volumes. Moreover, the number of prior episodes of depression has been associated with increased volume of the globus pallidus.58
Additional research is needed to elucidate the clinical significance of regional volume enlargement in threshold and subthreshold depressive disorders.
Given the extensive literature linking late-life depression to dementia59
and implicating hippocampal atrophy in Alzheimer disease,60
the finding of hippocampus volume reduction in some studies of late-life depression may be attributed to incipient dementia. We excluded individuals in whom dementia or mild cognitive impairment developed at any point during a 9-year follow-up period, which allowed us to more confidently attribute the observed effects to depressive symptoms rather than the prodromal phase of dementia. Furthermore, the lack of association between depressive symptoms and hippocampus volumes in our sample suggests that observed associations for other brain regions are more likely due to depressive symptoms than incipient dementia.
Associations between depressive symptoms and brain volumes in geriatric samples may also be confounded by physical symptoms and comorbid medical conditions such as cerebrovascular disease in older adults. We not only excluded individuals with cerebrovascular disease at baseline, but the longitudinal nature of the study also allowed us to exclude data for participants in whom cerebrovascular disease was diagnosed during the study interval. Our inclusion of data for such individuals before the diagnosis does not appear to have significantly impacted our results because exploratory analyses in which we eliminated prediagnosis data from the sample yielded results that did not substantially differ from the reported results. Moreover, secondary analyses with the depressed mood subscale of the CES-D yielded similar results to the CES-D total symptom score, providing confidence that our findings reflect associations with negative affect rather than physical symptoms. Sex differences in the association of depressed mood subscales scores with left frontal white matter and cingulate gyrus volumes are consistent with previous research documenting sex differences in the relations between depressive symptoms, brain volumes and cognitive impairment in the elderly.27,61,62
The measurement of brain volumes in our study was based on a semiautomated approach. Given the number of scans in this longitudinal study (783 scans in the present sample), this approach renders MRI data analysis more feasible than manual tracings. This approach is well validated37,38
and has been shown to accurately segment grey and white matter and cerebrospinal fluid compartments with error rates of less than 2%–3% when tested against our realistic digital brain phantom.37
Moreover, reliability of labelling for the ROIs in the present study exceeds 0.84. Nonetheless, the semiautomated approach is not without limitation. Labelling is limited by registration accuracy, as well as by changes in grey matter–white matter contrast with age. These limitations may result in increased noise. Our finding of significant associations of brain volumes with depressive symptoms despite this increased noise attests to the strength of our results. An additional limitation of this method is that white matter signal hypointensities on T1
-weighted MRI scans are segmented as grey matter, leading to a small overestimation of grey and underestimation of white matter. However, based on manual tracing, we have found that these signal abnormalities account for less than 1% of brain volume even in individuals with extensive white matter findings. These limitations do not outweigh the aforementioned strengths of the automated approach.
Early- and late-onset depression may have different etiological contributors and associations with brain pathology. However, information regarding the age at onset of depressive symptoms was not available for the current study and is a potential limitation. The restricted age range of our sample (i.e., age 56–85 at baseline) is an additional limitation given our interest in examining whether age modified the effects of depressive symptoms on brain volumes. Nonetheless, we found that older age was associated with an adverse impact of depressive symptoms on brain volumes, even within our limited age range. The potential for inflated type-I error is a further limitation and suggests that statistically less pronounced differences should be interpreted with caution. However, since to our knowledge this is the first study that explores the longitudinal association between depressive symptoms and volume decline in older adults, we tolerated greater type-I error rather than miss a true positive finding.41
Moreover, the large effect sizes observed for the effects would suggest that the differences are indeed meaningful rather than spurious findings.42
Our sample was not population-based, and most participants were highly educated and white, limiting the generalizability of our findings. However, the relative homogeneity of the sample may also have been a strength of this study because it diminished possible effects of demographic variables as confounds. These limitations should be viewed in the context of the strengths of the study, such as the large sample size, relatively long follow-up period with prospective assessments and well characterized nature of the sample.
Our results provide evidence that depressive symptoms, even at subthreshold levels, are associated with volume reduction in frontal and temporal brain regions, particularly with advancing age. Future longitudinal studies investigating the temporal relation between brain pathology and depressive symptoms over longer follow-up intervals and a wider age range are critical to elucidate whether structural and functional brain changes are the etiology or a consequence of depression.