Subjects provided written informed consent as approved by the University of Pittsburgh Biomedical Institutional Review Board. The first 10 healthy mothers enrolled comprised an independent sample from which the ROI was established (23
). Group comparison was conducted on the next 12 healthy mothers enrolled versus all 12 depressed mothers enrolled, without prospective subject matching. The structured clinical interview for DSM-IV(24
) was used to assess psychiatric status. Healthy subjects had no present or past history of an Axis I disorder, no family history of a mood or psychotic disorder, and a 25-item Hamilton Rating Scale for Depression score (HAM25
) ≤ 7. Depressed subjects had no psychotic or bipolar illness, met DSM IV criteria for major MDD, and had a HAM25
≥ 15 in the past month. Both prevalent(beginning antenatally) and incident(new onset postpartum) cases of postpartum depression were included to maximize generalizability, since the disorder commonly begins antenatally(25
). Subjects were excluded if they had medical or neurological illnesses likely to affect cerebral physiology or anatomy, gross abnormalities of brain structure evident by magnetic resonance images, suicidal intent, substance abuse within one year, lifetime history of substance dependence(other than nicotine), eating disorders, use of hormonal contraception, or exposure to medications likely to alter cerebral physiology within 3 weeks. Subjects, of whom 50% were primiparous, delivered a healthy, term infant in the preceding 10 weeks, were medication-free, and breastfeeding or bottlefeeding.
On the scan day, clinical severity was established with the Hamilton depression scale, the Fawcett-Clark Pleasure Scale, the Edinburgh Postnatal Depression Scale (a well-validated, 10-item self-report measure of perinatal depression, anxiety, and function(26
)), and the parent-to-infant attachment questionnaire(a reliable and valid self-report of attachment quality, hostility, and pleasure in interaction during the first postpartum year(27
)). Statistical tests on group differences in demographic, reproductive, psychiatric, and behavioral data were performed with Pearson chi-square for categorical and Mann-Whitney U exact tests for continuous variables ().
We used a fast event-related version of a well known monetary reward number guessing task that activates ventral striatum (22
) and that was previously used in a block design in MDD(29
). Subjects guess whether the value of a hidden card is less than or greater than 5(range 1–9). Subjects receive monetary gains(rewards) for correct guesses and incur monetary losses(punishment) for incorrect ones. Gains and losses may be high (+$0.80, −$0.50) or low magnitude(+$0.30, −$0.20). Unbeknownst to the subjects trial outcomes are predetermined and there is no way to optimize winning. Feedback is given as upward point green (reward) or downward pointing red (punishment) arrows that are presented for 750 ms. The size of the arrow indicates magnitude(high or low). The task consists of 160 trials with 40 trials per outcome.
Scanning was performed on a Siemens 3 Tesla Trio (Erlangen, Germany). High-resolution, T1-weighted anatomical images were acquired using an MPRAGE sequence (TR=1630ms; TE=2.48 ms; FOV=20.4 cm; α=8°; image matrix=2562; voxel size = 0.8×0.8×0.8 mm; 224 slices). High resolution functional scans(blood oxygenation level-dependent: BOLD) were taken in the same plane as the anatomical images(28, 3 mm slices, TR 1.5s, 3.5 mm in plane resolution). In each run, 160 successive brain volumes were acquired (5 runs, total=800).
The NeuroImage Software package(NIS) and AFNI were used to preprocess and analyze the data. Data was transformed using standard anatomical landmarks(anterior and posterior commisures) to conform to the atlas of Talairach and Tournoux. Functional data were concatenated across runs and analyzed using a general linear model (3dDeconvolve). Covariates for the model included time onsets for each trial type(positive and negative feedback, high and low) with separate parameters estimated for each time point from 0 to 16.5s after trial onset(TRs 1–11), motion estimates, a model of linear drift and baseline activity. Beta values for each of the time point covariates were calculated by least squares regression to the BOLD signal. These values constitute a time series from time 0–16.5s for each voxel in the brain and were used for all subsequent within and between subject statistical analyses.
The analysis of the functional data from the independent sample of ten healthy mothers was used to localize a functional region of interest(ROI) in the striatum. For each subject, we generated a voxel-wise contrast between the estimated BOLD response to reward trials(high and low magnitude) and the estimated BOLD response to punishment trials(high and low magnitude). A voxelwise t-test versus the null hypothesis then determined clusters of voxels where this difference was significant. The AFNI AlphaSim program(Montecarlo method) was used on the contrast T-maps to set the contiguity thresholds such that the map wise probability of a false detection remained lower than 0.01(30
). Repeated-measures ANOVA of time series data extracted from this ROI was next used to compare depressed(n=12) with healthy mothers(n=12), separately for each trial type(low reward, high reward, low punishment, or high punishment). A sensitivity analysis was conducted in which subjects were matched on behavioral performance(depressed n=10, healthy n=12).