Fig 1.
CONSORT Diagram for the Randomized Controlled Trial (RCT).
We enrolled 57 patients and acquired MRS data in 41 patients at two time points: at a baseline before the start of the trial and immediately following the end of the trial. All patients using psychotropic medications (N = 4) underwent a 4-week wash out period before starting the trial. In addition, we acquired MRS data at one time point in 29 healthy controls. The patients were randomized to receive either the active medication duloxetine or placebo for 10 weeks. Once the trial was complete and MRS data were acquired in every participant, we processed MRS data blind to the order of data acquisition, participant characteristics, and patient assignment to a treatment arm. Subsequently, all statistical analyses were conducted on the processed MRS data.
Table 1.
Participant demographics and symptom severity.
Dysthymic patients did not differ significantly from healthy controls in age (p = 0.41) or sex (χ2 = 0.17, df = 1, p = 0.68). Duloxetine-treated patients did not differ from placebo-treated patients in age (p = 0.44), symptom severity (p = 0.26), or lifetime history of Major Depressive Disorder (χ2 = 1.17, df = 1, p = 0.27), but differed significantly on sex (χ2 = 7.15, df = 1, p = 0.007). The excellent remission rate in duloxetine-treated patients is likely because of their treatment with a high dose (~95 mg/day) of duloxetine. Responder is defined as a patient whose symptom severity at week 10 decreased by 50% or more relative to its baseline severity. The table lists the average and standard deviation for age and symptom severity.
Fig 2.
Placement of the Six Slabs in Multiplanar Chemical Shift Imaging (MPCSI).
We acquired MR spectroscopy data in six axial oblique slabs through the brain, with each slab comprising a 24 x 24 matrix of 1 cm3 voxels with 2 mm gap between adjoining slabs. The second slab from the bottom was placed such that it contained the anterior commissure (AC) and the posterior commissure (PC). We placed one slab below and 4 slabs above the AC-PC slab, thereby generating metabolite maps across the entire brain in each participant.
Fig 3.
Baseline abnormalities in NAA concentrations and associations with symptom severity.
At baseline, we compared NAA concentrations in 41 patients with dysthymic disorder (PDD) relative to 29 healthy controls (Dx Effects, left column) and correlated levels with symptom severity in the 41 patients (HamD, right column). The major white matter (WM) fiber pathways labeled in the Johns Hopkins atlas[65] (WM Tracts) were used to localize WM findings; the gray matter (GM) gyri labeled in the CCB atlas[66] (GM Regions) were used to localize GM findings. The WM atlas and the GM atlas were each normalized within the coordinate space of the template brain, thereby allowing us to localize precisely the findings to specific brain regions and WM fiber pathways. The findings are shown on 5 representative slices through the brain. Dx Effects: Patients, relative to controls, had higher metabolite concentrations across large portions of the brain, especially in the caudate nucleus (CN), anterior corona radiata (aCR), corpus callosum (CC), thalamus (Thal), right lenticular nucleus (putamen and globus pallidus), and posterior corona radiata (pCR). HamD Correlations: Symptom severity, measured using the Hamilton Depression Rating Scale (HDRS), correlated positively with NAA concentrations (i.e., patients with higher metabolite concentrations had more severe symptoms), especially in the sCR and CN. Transverse brain slices are shown in the radiological orientation. We controlled for false positives in all analyses using a false discovery rate (FDR) procedure and covaried for age and sex. We subsequently applied a cluster threshold that suppressed all findings of spatial extent smaller than 100 contiguous voxels. P-values are color coded, with positive associations displayed in warm colors (orange and red) and inverse associations in cool colors (cyan and blue). Abbreviations: R, Right Hemisphere; L, Left Hemisphere; NAA, N-acetyl Aspartate; CC, corpus callosum; sCR, superior corona radiata; CN, caudate nucleus; aCR, anterior corona radiata; Ins, insular; Th, thalamus; LN, lenticular nucleus (putamen and globus pallidus); pCR, posterior corona radiata.
Fig 4.
Changes in NAA concentrations over the 10-Week trial.
Left Panel: We applied repeated measures analyses to assess whether NAA concentrations in duloxetine-treated patients changed differently than those in placebo-treated patients. The analyses showed significant treatment-by-time effects on metabolite levels: that is, metabolite levels in duloxetine-treated patients changed differently than those in placebo-treated patients across large portions of the brain. We therefore assessed separately in duloxetine-treated (DLX) and in placebo-treated (PBO) patients how NAA concentrations changed during the clinical trial. These analyses showed that NAA concentrations in duloxetine-treated patients declined significantly across large portions of the brain, whereas in placebo-treated patients concentrations increased across large portions of the brain. Right Panel: At the end of the trial, we assessed separately how NAA concentrations in duloxetine-treated or in placebo-treated patients differed from those in healthy controls. Concentration levels were normalized across large portions of the brain in duloxetine-treated (DLX) but increased and deviated further away from healthy values on placebo-treated patients (PBO). P-values in all analyses that survived the FDR procedure for multiple comparisons and a cluster-level threshold of 100 contiguous voxels were color encoded using warm colors (orange and red) for increased and cool colors (cyan and blue) for decreases in NAA concentrations. Abbreviations: R, Right Hemisphere; L, Left Hemisphere; NAA, N-acetyl Aspartate; CC, corpus callosum; sCR, superior corona radiata; CN, caudate nucleus; aCR, anterior corona radiata; Ins, insular, CC, corpus callosum, Th, thalamus; LN, lenticular nucleus (putamen and globus pallidus); pCR, posterior corona radiata; IFG, inferior frontal gyrus.
Fig 5.
Changes in NAA concentrations and symptom severity.
Left Column: Using repeated measures analyses, we assessed within duloxetine-treated patients how NAA concentrations changed with changes in symptom severity. These analyses showed that in general the change in NAA concentrations were positively associated with the change in symptom severity: i.e., NAA concentrations decreased towards healthy values as symptom severity decreased in duloxetine-treated patients. Right Column: We subsequently applied longitudinal mediation analyses to assess whether changes in symptom severity mediated the treatment effects on changes in NAA concentration. These analyses showed that symptom severity significantly mediated the change in NAA concentrations as a consequence of treatment in the caudate nucleus (CN) and putamen (PUT). Mediation analyses provided no statistical evidence for the alternate hypothesis, that the change in NAA concentrations mediated the treatment effects on change in symptom severity. All maps are FDR-corrected at p<0.05, and analyses included age and sex as covariates. We subsequently applied a cluster threshold that suppressed all findings of spatial extent smaller than 100 contiguous voxels. Positive associations were coded in warm colors (orange and red); inverse associations were coded in cool colors (cyan and blue). Abbreviations: R, Right Hemisphere; L, Left Hemisphere; NAA, N-acetyl Aspartate; CC, corpus callosum; sCR, superior corona radiata; CN, caudate nucleus; aCR, anterior corona radiata; Ins, insular, Th, thalamus; LN, lenticular nucleus (putamen and globus pallidus); pCR, posterior corona radiata.