As previously reported,18
74 depressed subjects enrolled in the original 12-week trial of sertraline had evaluable DTI data. Thirty-six of those subjects either refused to participate in this follow-up study or were lost to follow-up. Of the 38 individuals who did enroll, DTI scan data was not processable for 9 individuals, primarily due to the MRI being ended early or difficulties with image registration. This resulted in a sample of 29 individuals who were depressed at entry into the parent study who were included in this follow-up study. We found no significant baseline differences between subjects who were and were not included in this follow-up study in age, sex, education, MMSE, or medical comorbidity (by CIRS). We additionally found no significant difference in depression severity (by MADRS) at entry into the parent study or at the final MADRS score upon exiting the 12-week trial (data not shown). At the 1-year assessment of those 29 previously depressed subjects, 16 were remitted (defined as MADRS ≤ 8) and 13 were not remitted, with a MADRS > 8 and continuing to meet DSM-IV criteria for Major Depressive Disorder.
The sample additionally included 20 elderly subjects with no psychiatric history. This population was recruited from an original cohort of 27 never-depressed elders enrolled in the original study. None of this never-depressed population became depressed over the 1-year period, and no subjects were excluded because of new onset of depression. Demographic differences between groups are displayed in . There was a significant difference between cohorts in age, and a strong trend for a difference in MMSE score, likely driven by the depressed-remitted cohort, which was lower than the other cohorts. However, the largest mean difference between groups in MMSE score was 1.3, which has limited clinical significance.
Cohort Demographic Differences
Group differences in baseline DTI measures are displayed in . After controlling for age, sex, and medical comorbidity severity, only middle frontal gyri FA was significantly associated with one-year group assignment. A comparison of least square means (with 43 df, derived from the model error) showed that both depressed groups had significantly lower FA values than did the nondepressed group (nondepressed vs. remitted, p = 0.0332; nondepressed vs. nonremitted, p = 0.0143), but there was not a significant difference between depressed groups (p = 0.7078).
Baseline DTI measures by group assignment at one year
At follow-up, most of the depressed population were on antidepressants, except for 4 remitted and 2 nonremitted depressed subjects. Of the 23 subjects on antidepressants, 14 were on sertraline (6 nonremitted, 8 remitted). Nine (5 nonremitted, 4 remitted) were on other antidepressants, including venlafaxine (N = 4), bupropion (N = 1), citalopram (N = 1), escitalopram (N = 2), and duloxetine (N = 1). There was no significant difference in frequency of sertraline use at follow-up between remitted (8 of 12 (66.7%) on sertraline) and nonremitted subjects (6 of 11 (54.6%) on sertraline; 1df, χ2 = 0.35, p = 0.552).
Change in DTI Measures: Group Differences
shows mean group difference between regional ADC and FA measures, defined as Time 2 – Time 1. In models controlling for age, sex, CIRS, baseline DTI measure, and time between scans, only change in the ACC FA was significantly different between cohort groups. The depressed, nonremitted cohort exhibited less reduction in this region’s FA than did the nondepressed or depressed, remitted cohorts. Differences in the ACC ADC values approached but did not reach a level of statistical significance.
One-Year Change in DTI measures by group
To determine if differences in antidepressant use might be affecting this result, we excluded the 6 depressed subjects who were not taking antidepressants at follow-up and re-examined the models. The group differences in change in ACC FA remained significant after excluding these subjects (F 2,42 = 5.74, p = 0.0071), while the differences in ACC ADC no longer approached statistical significance (F 2,42 = 2.15, p = 0.1322).
displays change in ACC FA by each subject, for each diagnostic cohort. Although the majority of subjects for the nondepressed and depressed, remitted cohorts () exhibited a decline in this measure, in the depressed, nonremitted cohort (), 6 subjects (46%) exhibited an increase in ACC FA while 7 exhibited a decrease. When compared against those subjects who experienced a decrease in ACC FA, the group with increased ACC FA exhibited a significantly greater change in ACC FA (0.020, SD = 0.017; − 0.031, SD = 0.018; t = 5.22, 11 df, p = 0.0003) but a lower baseline ACC FA (0.317, SD = 0.034; 0.393, SD = 0.056; t = −2.91, 11 df, p = 0.0142). At alpha = 0.05, there were no significant differences between these groups in age, age of depression onset, MMSE or CIRS score, MADRS score at baseline or follow-up, sex representation, or days between scans (data not shown). All of the depressed, nonremitting subjects who exhibited a decline in ACC FA were taking antidepressants at followup, while two of the six who exhibited an increase in ACC FA were not taking antidepressants at followup (Fisher’s exact test, p = 0.1923).
Change in Anterior Cingulate Cortex Fractional Anisotropy
Change in DTI Measures: Relationship with Longitudinal Change in MADRS Score
Finally we sought to determine if change in DTI measures over 1 year was associated with change in MADRS over 1 year. In these models, change in MADRS was the dependent variable, while change in DTI measure was the independent variable. Both change variables were calculated as Time 2 – Time 1. We also controlled for age, sex, time between scans, baseline MMSE, and baseline CIRS. We created two sets of models (): one included all subjects, and had an additional diagnostic variable designating subjects as being depressed or nondepressed at study entry; the second examined only subjects who were depressed at study entry.
Models examining change in MADRS over one year
Only change in anterior cingulate cortex FA was significantly associated with change in MADRS (), exhibiting a positive correlation where a greater decline in FA was associated with greater decline in MADRS (). Removal of depressed subjects not taking antidepressants at follow-up increased the strength of the relationship between change in ACC FA and change in MADRS (all subjects: F 1,42 = 15.76, p = 0.0004; depressed only: F 1,22 = 12.93, p = 0.0029). We did not find a similar relationship between change in MADRS and change in other DTI measures (data not shown), regardless of including or excluding previously depressed subjects not taking antidepressants at follow-up.
Change in MADRS by change in Frontal Anisotropy