Demographic and Clinical Characteristics of the Sample
MR scans were aborted in one control and one depressed participant when they reported anxiety. Data from two controls had significant motion artifacts. Therefore, results are presented for 32 control, 22 high-risk and 29 depressed participants. Demographic and clinical features of the sample are outlined in . The groups did not differ significantly with respect to age, gender, ethnicity/race or pubertal status. The depressed and high-risk groups had lower SES, and a higher magnitude of early-life adversity and chronic stress during adolescence, than controls. Depressed adolescents had higher BDI and HDRS scores than the control and high-risk groups. Depressed adolescents also had a lower CGAS score than controls, and the high-risk group was intermediate. Of 29 adolescents with depression, 15 (51.7%) had a history of parental depression, and only two (6.9%) had a prior history of antidepressant treatment. By definition, all high-risk subjects had a parent with a history of depression.
Baseline demographic and clinical characteristics by diagnosis
Relationship between Demographic/Clinical Variables and Hippocampal Volume
The right and left hippocampal volumes correlated highly with each other (r = .89, df = 83, p = .0001). Age, SES, BDI, HDRS and CGAS did not correlate significantly with the hippocampal volume. Gender, ethnicity/race and pubertal status also did not have a significant effect on hippocampal volume.
Group Comparisons on Hippocampal Volume
After controlling for age, gender, ethnicity/race, pubertal status and SES, diagnostic status had a significant effect on hippocampal volume (F4,148 = 5.32, p = .0001). None of the covariates had a significant effect on hippocampal volume even in the multi-variate analysis. In post-hoc analyses, the depressed and high-risk groups had smaller hippocampal volume than controls bilaterally (Left: F2,75 = 8.54, p = .0001; Right: F2,75 = 10.26, p = .0001; see ). Compared with controls, the left hippocampal volume was smaller by 4.2% and the right by 4.9% in the high-risk group, and 3.1% and 3.0%, respectively in the depressed group.
Figure 2 Left and right hippocampal volumes (mean ± SEM) in control, high-risk and depressed groups (Significant differences in hippocampal volume after controlling for age, gender, ethnicity/race, pubertal status and socioeconomic status; Left: controls (more ...)
Effect of Early-life Adversity on Hippocampal Volume
After controlling for chronic stress during adolescence, higher scores on early-life adversity were associated with smaller hippocampal volumes in the controls (Left: r = −.29, df = 29, NS; Right: r = −.44, p = .02), and in high-risk participants (Left: r = −.52, df = 18, p = .02; Right: r = −.54, p = .02). In the depressed cohort, the association between early life-adversity and hippocampal volume was moderated by parental depression. Among adolescents with parental history of depression, early-life adversity correlated negatively with hippocampal volume (Left: r = −.68, df = 12, p = .007; Right: r = −.43, NS). In contrast, early-life adversity correlated only modestly in adolescents without family history (Left: r = .25, df = 11, NS; Right: r = .05, NS).
Follow-up information was available for 79/83 (95.2%) participants. Subjects who did not participate in follow-up assessments did not differ significantly from those with follow-up information on any baseline characteristics. Recruitment did not occur simultaneously and, therefore, not all subjects were studied longitudinally for the same period of time. Of the 79 adolescents who had follow-up information, 7.6% were followed for 1 year, 20.2% for 2 years, 22.8% for 3 years, 21.6% for 4 years and 27.8% for 5 years. The three groups were comparable on the mean follow-up interval (3.1 ± 1.3 years in controls, 3.0 ± 1.2 in the high-risk group, and 3.2 ± 1.3 in the depressed group; F2,76 = 0.39, NS).
Depressive Episode during Follow-up
Of 31 controls with follow-up information, two (6.5%) developed a depressive episode compared to 6/21 (28.6%) high-risk subjects. In the depressed cohort, 11/27 (40.7%) developed a recurrent episode. Participants with and without recurrent episodes did not differ significantly on treatment history. The groups differed significantly on the probability of depression during follow-up (Mantel-Cox χ2 = 9.45, df = 2, p = .009), with controls differing having significantly lower risk than high-risk (χ2 = 5.67, p = .02) and depressed youth (χ2 = 10.18, p = .001). In the controls, the probability of depression was 0% in the first two years, 6.7% in the third year and 14.4% in the fourth year. In the high-risk group, the probability of depression was 4.8% in the first year, 24.8% in the second year, and 32.3% in the third year. Among depressed adolescents, the probability of a recurrent episode was 7.4% in the first year, 21.9% in the second year, 44.7% in the third year, and 57.9% in the fourth year.
Associations among Early-life Adversity, Hippocampal Volume and Depression at Follow-up
The relationships among early-life adversity, hippocampal volume and depressive disorder are depicted in (also see ). After controlling for age, gender, ethnicity/race, pubertal status and SES, early-life adversity was associated with depressive disorder (path C) and hippocampal volume (path A). When hippocampal volume was included in the model, the effect of early-life adversity on depression was reduced (path C′). After controlling for early-life adversity, hippocampal volume predicted depression (path B) during follow-up. However, the mediational effect was not statistically significant (Left: AB = 0.08, CI = 0.00–0.16, z = 1.85, Sobel Test, p = .06, Cohen’s d = 1.11; Right: AB = 0.07, CI = −0.01–0.15, z = 1.80, Sobel Test, p = .07, Cohen’s d = 1.44). Although the Sobel Test was not statistically significant, the coefficient of path C′ for the left hippocampus (beta = 0.18) was less than the coefficient of path C (AB+C′ = 0.26). This also was true for the right hippocampus (path C′ = 0.20; AB+C′ = 0.28). Hippocampal size did not moderate the relationship between early-life adversity and depression (interaction effect: Left: β = −0.09, SE = 0.30, OR = 0.92, CI = 0.51–1.66, F = 0.01, NS; Right: β = −0.28, SE = 0.29, OR = 0.76, CI = 0.43–1.33, F = 0.94, NS).
Mediational analyses predicting major depressive episode at follow-up1
For the purpose of graphical representation, the sample was stratified into four groups based on a median split of early-life adversity and right hippocampal size: (1
) low adversity-normal hippocampus (n = 24); (2
) low adversity-small hippocampus (n = 16); (3
) high adversity-normal hippocampus (n = 17); and (4
) high adversity-small hippocampus (n = 21). The groups then were compared on the probability of a depressive episode during follow-up (see ). Among adolescents with low-adversity and normal hippocampal volume, 20.3% were likely to develop a depressive episode by 3.2 years (mean follow-up period), whereas 56.6% of those with high adversity and small hippocampal volume were likely to develop a depressive episode during the corresponding period (Mantel-Cox χ2
= 8.79, df = 3, p = .03). The probability of depression during that period was 30.4% in subjects with low adversity and small hippocampal volume, and 21.4% in those with high adversity and normal hippocampal volume.
Figure 3 Probability of a major depressive disorder (MDD) episode during follow-up, stratified by early-life adversity and hippocampal volume (right) measured during the initial evaluation (high adversity-small hippocampal volume vs. low adversity-normal hippocampal (more ...)
Although the interaction of early-life adversity and hippocampal volume was not statistically significant, these data suggest that the combination of early-life adversity and smaller hippocampal size significantly increases the vulnerability to depressive episodes. Examination of the data separately in control and high-risk groups (new episode of depression) and depressed youth (recurrent episode) showed the same pattern.
There were no significant group differences between the two sites on any of the major variables of interest. Additionally, the analyses were repeated separately at both sites, and the same pattern of results emerged.