Brain metabolic changes in major depressive disorder from pre- to post-treatment with paroxetine

Psychiatry Research: Neuroimaging Section 91 Ž1999. 127]139

Brain metabolic changes in major depressive disorder from pre- to post-treatment with paroxetine Arthur L. Brody a,b,U , Sanjaya Saxenaa , Daniel H.S. Silvermanc , Shervin Alborzian, Lynn A. Fairbanks a , Michael E. Phelps c , Sung-Cheng Huang c , Hsiao-Ming Wu c , Karron Maidment a , Lewis R. Baxter Jr.a,d a

UCLA Department of Psychiatry and Biobeha¨ ioral Sciences, Neuropsychiatric Institute and Hospital, Room 67-468, 760 Westwood Bl¨ d., Los Angeles, CA 90024, USA b Greater Los Angeles VA Healthcare System, Los Angeles, CA 9002, USA c UCLA Department of Medical and Molecular Pharmacology, Los Angeles, CA 90024, USA d UAB Department of Beha¨ ioral Neurobiology, Birmingham, AL 35294-0017, USA Received 24 February 1999; received in revised form 6 July 1999; accepted 27 July 1999

Abstract Functional brain imaging studies of subjects with Major Depressive Disorder ŽMDD. have suggested that decreased dorsolateral ŽDLPFC. and increased ventrolateral ŽVLPFC. prefrontal cortical activity mediate the depressed state. Pre- to post-treatment studies indicate that these abnormalities normalize with successful treatment. We performed w 18 Fxfluorodeoxyglucose positron emission tomography ŽFDG-PET. scans on 16 outpatients with MDD before and after treatment with paroxetine Žtarget doses 40 mgrday.. Regions of interest ŽROIs. for this analysis were drawn by a rater blind to subject identity on the magnetic resonance image of each subject and transferred onto their coregistered PET scans. We hypothesized that DLPFC metabolism would increase, while ventral frontal metabolism win the VLPFC, the orbitofrontal cortex ŽOFC., and the inferior frontal gyrus ŽIFG.x would decrease with successful treatment. Treatment response was defined as a decrease in the Hamilton Depression Rating Scale of ) 50% and a Clinical Global Improvement Scale rating of ‘much’ or ‘very much’ improved. By these criteria, nine of the subjects were classified as treatment responders. These responders had significantly greater decreases in normalized VLPFC and OFC metabolism than did non-responders. There were no significant effects of treatment response on change in the DLPFC or IFG in this sample. However, there was a positive correlation between change in HAM-D scores and change in normalized IFG and VLPFC metabolism. There were no significant

U

Corresponding author. Tel.: q1-310-825-5938; fax: q1-310-206-2802. E-mail address: [email protected] ŽA.L. Brody.

0925-4927r99r$ - see front matter Q 1999 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0 9 2 5 - 4 9 2 7 Ž 9 9 . 0 0 0 3 4 - 7

128

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

interactions with laterality. On pre-treatment scans, lower metabolism in the left ventral anterior cingulate gyrus was associated with better treatment response. These findings implicate ventral prefrontal]subcortical brain circuitry in the mediation of response to serotonin reuptake inhibitors in MDD. Q 1999 Elsevier Science Ireland Ltd. All rights reserved. Keywords: w 18 Fxfluorodeoxyglucose positron emission tomography; FDG-PET; Treatment response prediction; Depression; Ventrolateral prefrontal cortex; Anterior cingulate gyrus

1. Introduction Considerable heterogeneity exists in the functional brain imaging literature as to how Major Depressive Disorder ŽMDD. is mediated Žfor reviews, see Ketter et al., 1996; Kennedy et al., 1997; Drevets, 1998.. The most commonly reported findings in studies comparing subjects with MDD to normal controls are differences between the groups in prefrontal cortical activity. Specifically, in subjects with MDD, decreased activity in the dorsolateral prefrontal cortex ŽDLPFC. has been the most commonly reported finding Že.g. Baxter et al., 1989; Dolan et al., 1993; Biver et al., 1994.. However, increased activity in the ventral portion of the prefrontal cortex wincluding the ventrolateral prefrontal cortex ŽVLPFC., the inferior frontal gyrus ŽIFG., and orbitofrontal cortex ŽOFC.x has also been reported ŽUytdenhoef et al., 1983; Buchsbaum et al., 1986; Drevets et al., 1992; Biver et al., 1994.. In addition to these prefrontal differences, a recent report of a relatively large number of subjects revealed decreased activity Žand volume. in the subgenual prefrontal cortex ŽDrevets et al., 1997., an area more ventral and medial to the prefrontal regions previously described. Other areas commonly reported to have decreased activity in MDD include the anterior cingulate gyrus ŽAC. ŽBench et al., 1992; Mayberg et al., 1994., anterior temporal lobe ŽPost et al., 1987; Mayberg et al., 1994. and head of the caudate nucleus ŽCd. ŽBaxter et al., 1985; Buchsbaum et al., 1986; Drevets and Raichle, 1992; Mayberg et al., 1994.. Reports of brain metabolic changes from preto post-treatment of subjects with MDD have also been heterogeneous Žfor review, see Rubin et al., 1994; Drevets, 1998.. An increase in DLPFC

metabolism has been reported with sertraline treatment ŽBuchsbaum et al., 1997. and with a variety of medications, including tricyclic antidepressants, lithium, and trazodone ŽBaxter et al., 1989.. In contrast, a decrease in ventral prefrontal and limbic activity Žincluding the IFG, the ventral portion of the middle frontal gyrus, and the ventral AC. has been reported in MDD subjects who responded to electroconvulsive therapy ŽECT. ŽNobler et al., 1994., desipramine ŽDrevets and Raichle, 1992., fluoxetine ŽMayberg et al., 1999., or paroxetine Žabstract, Kennedy et al., 1998.. In all of these studies, brain activity changes occurred primarily in treatment responders. Taken together, these findings have led to the hypothesis that MDD is mediated by decreased dorsal prefrontal metabolism and increased ventral prefrontal and paralimbic activity ŽDrevets, 1998; Mayberg et al., 1999.. In the present study, we hypothesized that subjects with MDD who responded to the selective serotonin reuptake inhibitor ŽSSRI. paroxetine would show increases in dorsal frontal metabolic activity Žin the DLPFC. and decreases in ventral frontal metabolic activity Žin the VLPFC, IFG and OFC. when compared to non-responders to paroxetine. In addition to brain metabolic changes reported previously with treatment, one group ŽMayberg et al., 1997. recently examined baseline pre-treatment brain metabolic markers of response to antidepressant medication. In that study, higher metabolism in the rostral anterior cingulate gyrus was found to be associated with treatment response. We also sought to determine if pre-treatment regional brain metabolism in any regions associated with MDD would be a baseline marker of response to paroxetine treatment.

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

2. Methods 2.1. Subjects, treatment, and rating scales Sixteen medication-free outpatients with MDD were recruited through advertisements placed in a local newspaper, flyers, and patient-initiated phone calls to a general psychiatry screening center at the UCLA Neuropsychiatric Institute. Written informed consent was obtained after a complete description of the study to the subjects, using a consent form approved by the UCLA Office for the Protection of Research Subjects. All subjects met DSM-III-RrDSM-IV criteria for MDD. Subjects were screened by the study nurse coordinator ŽK.M.. and then twice by the treating psychiatrist Žeither A.L.B. or S.S.. prior to scanning. The Schedule for Affective Disorders and Schizophrenia } Lifetime version ŽSpitzer and Endicott, 1978. was also administered to confirm diagnosis Žsee Table 1 for clinical characteristics determined from the non-structured interviews.. The primary exclusion criteria were comorbid Axis I diagnoses Žincluding substance abuse. or concurrent medical conditions that might affect brain function. All subjects were free from psychotropic medications for at least 2 weeks Žand at least 5 weeks for fluoxetine. before starting the study. Subjects were treated with paroxetine HCl with the dosage adjusted over 1]2 weeks to a target dose of 40 mgrday. Treatment was initiated on the day following the first PET scan. No other psychotropic medications were allowed during the study period. Compliance was monitored by

129

patient report during weekly medication visits with the treating psychiatrist. Symptom severity was monitored using the Hamilton Depression Rating Scale ŽHAM-D; Hamilton, 1967., the Hamilton Anxiety Rating Scale ŽHAM-A; Hamilton, 1969., and the Global Assessment of Functioning Scale ŽEndicott et al., 1976.. Clinical improvement was recorded after treatment using the Clinical Global Improvement ŽCGI. Scale. Clinical response was defined a priori as a CGI score of ‘much’ or ‘very much’ improved and a drop of 50% or greater on the first 17 items of the HAM-D. These criteria were chosen based on prior usage of these cutoffs for response in many clinical studies Že.g. Feighner et al., 1998; Wheatley et al., 1998.. Percentage changes in rating scales were calculated by subtracting the post-treatment score from the pretreatment score and dividing by the pre-treatment score. 2.2. Measurement of glucose metabolism Each subject had an FDG-PET scan before and after 8 weeks of treatment with paroxetine. The FDG-PET method used here is identical to that in previous reports from our group Žsee Baxter et al., 1992. with the exception that scans commenced between 13.00 and 14.00 h in this study. Briefly, scanning was performed on an 831 NeuroECAT III PET tomograph ŽSiemens-CTI, Knoxville, TN. with subjects in the resting state Žears and eyes open.. Each subject’s head was held in place with a special head holder ŽMaz-

Table 1 Clinical characteristics of study subjects and study treatment variables Žmean " S.D.. Variable

Responders

Non-responders

All

N Age Žyears. Gender Ž% female. No. previous depressive episodes Paroxetine dosage Žmg. Study duration Žweeks. Pre-treatment HAM-D Ž17-item. Post-treatment HAM-D Ž17-item.

9 37.7 Ž"8.4. 22.2 0.9 Ž"0.9. 35.6 Ž"7.3. 8.2 Ž"2.8. 20.7 Ž"4.1. 6.4 Ž"3.4.

7 41.1 Ž"11.1. 28.6 1.3 Ž"1.1. 32.9 Ž"9.5. 8.7 Ž"2.4. 24.4 Ž"5.7. 18.7 Ž"4.8.

16 39.3 Ž"9.5. 25.0 1.1 Ž"1.0. 34.4 Ž"8.1. 8.5 Ž"2.6. 22.1 Ž"4.9. 11.8 Ž"7.4.

130

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

ziotta et al., 1982. to minimize movement and insure accuracy of placement in the tomograph. Scanning began with a 20-min transmission scan using a 68 Ge ring source for attenuation correction. Subjects then received an injection of 5]10 mCi 18 FDG. After a 40-min uptake period, emission scanning was performed for 40 min, and each scan was reconstructed from 2 to 3 million counts. This method produced 15-plane transaxial PET images with 6-mm in-plane resolution and 6.75mm axial resolution ŽCherry et al., 1991.. 2.3. Cerebral region of interest analysis Our primary method of analysis was a magnetic resonance imaging ŽMRI.-based region of interest ŽROI. analysis. Each subject received an MRI of the brain using a double echo sequence Žproton density and T2 images; TRs 2000]2500 ms; TE s 25]30 ms and 90]110 ms, 24 cm FOV, 3-mm slices with 0-mm separation.. PET to MRI co-registration was performed using a three-dimensional MRI-PET image registration program ŽLin et al., 1994. Žsee Fig. 1.. The MRI images were first segmented into four different tissue types wi.e. gray and white matter,

cerebrospinal fluid ŽCSF., and musclex. Image values for the segmented tissue types were assigned with a relative proportion of 4:1:0:0.5 for grayrwhiterCSFrmuscle. The segmented images with the assigned image values were then smoothed Žthree-dimensionally to match the spatial resolution of the measured PET images. to generate a set of simulated FDG-PET images. The program then minimized the sum of the square of pixel value differences between image sets to align the measured FDG-PET images with the simulated PET scan. The co-registration program used the Powell algorithm for minimization with seven variable parameters Žsix for orientation adjustment and one for matching the scaling between the two image sets. ŽPress et al., 1986.. This co-registration resliced the FDG-PET images to the orientation of the MR images. After coregistration, the PET and MR images were both resliced to 23 image planes for ROI drawing. The ROIs selected for our analysis were based on the previous literature cited above. ROIs were drawn on MRI by a rater ŽS.A.. blind to subject identity, and reviewed by the primary research team. Primary ROIs were the prefrontal cortical areas most often associated with MDD, namely

Fig. 1. MRI scan and pre- and post-treatment coregistered PET images of a single subject. Regions shown here Žthe hemispheres, heads of caudate nuclei, and putamen. were drawn on MRI and transferred directly onto the coregistered PET images.

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

the OFC, IFG, DLPFC, and VLPFC. Additionally, regions less commonly associated with MDD wthe dorsal and ventral AC and the head of the caudate nucleus ŽCD.x and control regions Žthe putamen and thalamus. were drawn. Though these control regions are known to have anatomical connections to the experimental regions, they were chosen because they were easily identified on brain scans and because they have not consistently been found to be associated with MDD in prior reports. Both supratentorial hemispheres were also drawn so that ratios of ROI metabolism to ipsilateral hemisphere could be calculated to give normalized ratios ŽROIrHem.. Normalized rather than absolute metabolic values were used

131

for this study, because absolute metabolic values Žcalculated from arterialized venous blood samples. were not felt to be reliable. Regions were drawn as they appear in Fig. 2. For our primary ROIs, the OFC was drawn in two planes, while the VLPFC, IFG, and DLPFC were drawn in three planes each. The OFC included the most ventral portion of prefrontal cortex posterior and superior to the eyes, excluding the gyrus rectus. This region included primarily the medial and lateral orbital gyri, as well as the orbital part of the IFG and the most anterior and inferior part of Brodmann’s area 10. The VLPFC consisted of the ventral half of the middle frontal gyrus, while the DLPFC consisted of the dorsal

Fig. 2. Regions of interest drawn on magnetic resonance images. After transfer to coregistered PET scans, these regions were linked to give a summed value for the region, which was then normalized to the linked value for the supratentorial ipsilateral hemisphere Žregion not shown.. DLPFC, dorsolateral prefrontal cortex; VLPFC, ventrolateral prefrontal cortex; AC, anterior cingulate gyrus; IFG, inferior frontal gyrus; Cd, head of the caudate nucleus; Put, putamen; Thal, thalamus; OFC, orbitofrontal cortex

132

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

half of the middle frontal gyrus. The IFG region contained the IFG dorsal to the OFC and ventral to the DLPFC. 2.4. Statistical Parametric Mapping analysis PET data were subjected to Statistical Parametric Mapping ŽSPM. analysis by the method of Friston Ž1995., Friston et al. Ž1991, 1995a,b. and Alpert et al. Ž1993. using the software package SPM96 Žprovided by K. Friston, MRC Cyclotron Unit, London, England. for two reasons. First, the drawn ROIs encompassed relatively large regions, and SPM allowed for examination of smaller regions that might have had significant changes not found in the above analysis. And second, because our ROI analysis was based on previous literature, we used SPM to screen the rest of the brain for regions of change not seen in prior samples. SPM analysis was performed on the total group of subjects who completed the study and on the subgroup of treatment responders. PET images were coregistered and reoriented with this program into the standardized coordinate system of the Talairach and Tournoux Ž1988. coplanar stereotactic brain atlas. Both pre- and post-treatment images were globally normalized by proportional scaling. To adjust for differences in subjects’ individual neuroanatomy, a 16-mm fullwidth at half-maximum three-dimensional gaussian smoothing filter was applied so that our analysis would identify regions of metabolic change as opposed to large changes in a single voxel. Contrasts were constructed by comparing matched pairs of each subject’s PET scans. Differences between pre-treatment and post-treatment scans were then assessed with the t statistic on a voxelby-voxel basis, to identify the profile of voxels that differed significantly Ž P- 0.01. between the two conditions. A Bonferroni-type correction was applied to these results to identify regions that would meet a more rigorous threshold. For determining the location of the regions found by this method, three analyses were performed. First, the MRIs of all study subjects were transformed into Talairach space using the SPM program and significant regions were mapped onto

this group-average MRI. Second, regions were mapped onto the MRI accompanying the SPM program. Third, voxel coordinates were located in the standard atlas ŽTalairach and Tournoux, 1988.. No differences in assignment of region location were found between these three methods. 2.5. Statistical analyses The data were screened for distributional properties, outliers, and missing values. This resulted in one subject being removed from the baseline markers of treatment response analysis for having pre-treatment mean ROI values that were ) 2 S.D.s away from the mean. For this subject, pre- to post-treatment regional metabolic changes were not significantly different from those for the rest of the group, so that this subject’s data were included in pre- to post-treatment analyses. For our MRI-based ROI analysis, a MANOVA was performed ŽSPSS version 8.0. for the change scores Žfrom pre- to post-treatment. of the normalized ROI values, with the prefrontal cortical experimental regions ŽDLPFC, VLPFC, IFG and OFC. and hemisphere Žleft vs. right. as withinsubject factors, and treatment response status Žresponder vs. non-responder. as a between-subject factor. A significant interaction of treatment response by ROI was followed by four univariate ANOVAs for change in each of the hypothesized regions, with treatment response as a betweensubject factor and hemisphere as a within-subject factor. Alpha levels for the hypothesized contrasts were set at P- 0.05. Kendall’s tau Žtwo-tailed. was then used to assess the relationship between the degree of change in HAM-D and change in ROI metabolism for each of the four regions. In addition to the four hypothesized regions, exploratory ANOVAs were run for the three ROIs less commonly associated with MDD Ždorsal and ventral AC and head of the Cd. and the two control regions Žputamen and thalamus.. We also performed exploratory Kendall’s correlations both pre- and post-treatment between ROIs found to be significant in the above analysis and anatomically related regions Že.g. frontal cortical to caudate and thalamus. to elucidate functional

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

changes in brain circuitry from pre- to post-treatment. For our analysis of pre-treatment metabolic markers of treatment response, we used a Stepwise Variable Selection Ža multiple regression technique . with pre-treatment ROIrHem metabolic values as the independent variables and change in HAM-D score Ž17-item. as the dependent variable.

3. Results 3.1. Clinical response By our criteria, nine of the 16 subjects responded to treatment. Of these, one of the treatment non-responders fell asleep during the second PET scan, so that this subject’s data were excluded from the pre- to post-treatment comparisons but were used for the baseline markers of treatment response analysis. 3.2. Regional metabolic changes with paroxetine treatment The overall MANOVA of the effects of treatment response on change in metabolism across the four regions ŽDLPFC, VLPFC, IFG, OFC. and two hemispheres revealed a significant interaction between region and treatment response Ž F s 5.45, d.f.s 3,39, Ps 0.015.. This result indicated that the metabolic changes associated with response to paroxetine differed across the four regions. The univariate ANOVAs indicated that the region most responsible for this interaction was the VLPFC. There was a significant effect of treatment response on change in this region Ž F s 6.49, d.f.s 1,13, Ps 0.024., with the responders decreasing as hypothesized, and the non-responders increasing in normalized metabolic rate ŽTable 2.. There was a tendency for the effect of response on change in VLPFC to be stronger in the right than in the left hemisphere, but the hemisphere by response interaction did not reach statistical significance Ž F s 4.77, d.f.s 1,13, Ps 0.097.. There was also a significant effect of treat-

133

ment response on change in metabolism in the OFC Ž F s 4.77, d.f.s 1,13, Ps 0.048. in the hypothesized direction. The effects of treatment response on changes in metabolism in the DLPFC and the IFG were not statistically significant. Correlations between the degree of change in HAM-D and change in ROI values for individual regions supported the above findings for the VLPFC. Decreases in metabolism correlated with greater improvement of depressive symptoms for both the right Ž t s 0.62, Ps 0.002. and left VLPFC Ž t s 0.60, P s 0.003.. Although the effect of treatment response on change in the IFG was not significant in the above ANOVA, there was a significant correlation between change in the HAM-D and change in normalized metabolism in the right IFG Ž t s 0.43, Ps 0.036. and a trend in the left IFG Ž t s 0.34, Ps 0.093.. There was also a trend toward a significant correlation between change in HAM-D and change in normalized metabolism in the OFC on the left Ž t s 0.36, Ps 0.075., but not on the right side Ž t s 0.21, n.s... Pre- to post-treatment metabolic changes in the five exploratory ROIs had no significant or trend level associations with treatment response, except for a significant interaction between hemisphere and treatment response in the thalamus Ž F s 7.18, d.f.s 1,13, Ps 0.019.. In the exploratory examination of correlations between regional activity in the ROIs found to be significant in the above analysis and associated structures Ži.e. between the VLPFC, IFG, and OFC and the Cd and thalamus., we found a significant change in one of these correlations from pre- to post-treatment. On the right side, there was a strong pre-treatment correlation between metabolism in the Cd and thalamus Ž r s 0.67, Ps 0.003. and this correlation was not present post-treatment Ž r s 0.11, n.s... The difference between these two correlations was statistically significant Žtwo-tailed t s 2.31, d.f.s 12, P0.05.. SPM analysis from the entire group of subjects Žboth responders and non-responders. revealed no significant metabolic changes from pre- to post-treatment. However, the SPM analysis did reveal an area of metabolic increase in the treat-

134

Experimental ROI

VLPFC IFG OFC DLPFC a

Right Left Right Left Right Left Right Left

Responders

Non-responders

Mean Ž"S.D..

Mean Ž"S.D..

ANOVA

Pre-treatment

Post-treatment

Change

Pre-treatment

Post-treatment

Change

1.190 Ž"0.112. 1.154 Ž"0.099. 1.201 Ž"0.050. 1.196 Ž"0.077. 1.040 Ž"0.083. 1.066 Ž"0.070. 1.240 Ž"0.056. 1.226 Ž"0.081.

1.151 Ž"0.083. 1.127 Ž"0.089. 1.184 Ž"0.062. 1.173 Ž"0.088. 1.026 Ž"0.078. 1.059 Ž"0.066. 1.233 Ž"0.071. 1.232 Ž"0.094.

]0.039 ]0.027 ]0.017 ]0.023 ]0.014 ]0.007 ]0.007 q0.006

1.223 Ž"0.069. 1.210 Ž"0.067. 1.268 Ž"0.051. 1.246 Ž"0.024. 1.064 Ž"0.029. 1.077 Ž"0.024. 1.288 Ž"0.090. 1.288 Ž"0.057.

1.261 Ž"0.089. 1.230 Ž"0.090. 1.270 Ž"0.060. 1.251 Ž"0.075. 1.094 Ž"0.030. 1.106 Ž"0.041. 1.285 Ž"0.073. 1.284 Ž"0.037.

q0.038 q0.020 q0.002 q0.005 q0.030 q0.029 y0.003 y0.004

Resp vs.non-resp P s 0.024 n.s. Resp vs. non-resp P s 0.048 n.s.

Abbre¨ iations. ROI, region of interest; S.D., standard deviation; VLPFC, ventrolateral prefrontal cortex; IFG, inferior frontal gyrus; OFC, orbitofrontal cortex; DLPFC, dorsolateral prefrontal cortex; resp, responder; non-resp, non-responder; n.s., non-significant

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

Table 2 Summary of changes in normalized regional brain metabolism in the four frontal regions of interest a

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

ment responder group that passed Bonferronitype correction Žsee Fig. 3.. This area contained voxels within the deepest gray matter gyri of the left premotorrsupplementary motor area and extended inferomedially to the adjacent white matter. The location of the voxel of peak significance in this region was xs y24 mm, y s y14 mm, zs 40 Ž Z s 5.3, P- 0.001 after Bonferroni-type correction.. Regional metabolic decreases shown in Fig. 3 which had a trend toward significance Ž P- 0.01, uncorrected. were in the primary occipital cortex Žbilaterally ., the posterior cerebellum Žbilaterally ., and the left posterior superior frontal cortex Žroughly corresponding to Brodmann’s area 9.. Other regions Žnot shown in the figure. which had trend-level increases Ž P- 0.01,

135

uncorrected. in metabolism were the mesial temporal lobe Žbilaterally ., the lateral aspect of the midbrain Žbilaterally ., and the right pyramidal tracts. 3.3. Cerebral metabolic markers of response to paroxetine As for pre-treatment markers of treatment response, the Stepwise Variable Selection selected the left ventral AC as the region that was most associated with treatment response. Lower normalized metabolism in this region was a marker for better response to paroxetine Ž F s 5.38, d.f.s 1,15, Ps 0.039. ŽFig. 4..

Fig. 3. Decreases seen in treatment responders Ž P- 0.01, uncorrected., including the primary occipital cortex Žbilaterally ., the posterior cerebellum Žbilaterally ., and the left posterior superior frontal cortex Žroughly corresponding to Brodmann’s area 9.. The significant increase Ž P- 0.01 after Bonferroni-type correction. in the deepest gray matter gyri of the left premotorrsupplementary motor area Žinset. is also shown.

136

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

Fig. 4. Correlation between pre-treatment metabolism in the left ventral anterior cingulate gyrus ŽLVAC. and change in the HAM-D. Lower pre-treatment metabolism in this region was associated with greater decrease in the HAM-D.

4. Discussion This study is, to our knowledge, the first reported pre- to post-treatment study of MDD using an MRI-based ROI analysis of FDG-PET scans. The use of anatomical information from MRIs to localize cortical brain regions decreases the intersubject variability in ROIs associated with regions drawn directly on functional brain images, as well as the potential confounding factor of anatomical standardization across subjects with SPM. Both variations in region placement on functional images and anatomical standardization are thought to be confounding factors in cognitive ŽNadeau and Crosson, 1995. and language processing ŽSteinmetz and Seitz, 1991; Rajkowska and Goldman-Rakic, 1995. studies without MRIbased localization of ROIs. In this study, ventral prefrontal metabolism Žin the OFC and VLPFC. was found to decrease in MDD subjects who responded to paroxetine. Decreases in ventral prefrontal metabolism Žin the VLPFC and right IFG. correlated with improvement in the severity of MDD. No other experimental or control regions changed significantly other than the right thalamus. These findings indicate that changes in ventral prefrontal]subcortical brain circuit function are

associated with response to serotonin reuptake inhibitors, though the lack of a control group limits the degree to which this connection can be firmly established. Multiple, parallel frontal]subcortical circuits are thought to exist, with strong input from the frontal cortex into the caudate and reciprocal connections between the frontal cortex and thalamus ŽAlexander et al., 1986; Parent and Hazrati, 1995a,b.. The significant decrease found here in the correlation between activity in the right Cd and thalamus, therefore, may indicate that right ventral prefrontal input into the Cd or thalamo-ventral cortical tone is altered by treatment with an SSRI. SSRIs preferentially bind to that part of the striatum Žventromedial Cd. that receives input from ventral prefrontal and paralimbic cortex, including the OFC ŽInsel, 1992.. Hence, enhanced serotonergic neurotransmission would be expected to preferentially modulate ventral prefrontal]subcortical circuit function. The SPM finding of increased supplementary motor area metabolism with successful treatment might represent an increase in movement or intention to move that took place in treatment responders only. This finding is consistent with the improvement in psychomotor retardation seen in most of the successfully treated subjects. Finally, though our finding that lower ventral AC metabolism was associated with better treatment response is different from that of Mayberg et al. Ž1997., methodological differences may explain the divergent findings. In particular, though the region found significant by both their group and ours would be considered to be in the affectiverrostral portion of the AC ŽDevinsky et al., 1995., our ventral AC ROI was mostly ventral to the perigenual anterior cingulate region described by Mayberg et al. Žreported as corresponding to Brodmann’s area 24 arb.. Therefore, different regions of the AC may have different relationships to clinical response. Additionally, in the other report, subjects all required hospitalization, had higher mean HAM-D score pre-treatments than in our study, and were not treated with paroxetine. The limitations of this study were small sample size, the lack of reliable absolute metabolic values, the absence of a control group, and the

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

general limitation of ROI placement. A larger sample size may help clarify findings that did not reach significance in the present study, such as the lack of a significant correlation between decrease in OFC metabolism and decrease in HAM-D. Reliable absolute metabolic data would have enabled the determination of global brain changes and improved the interpretation of regional findings. Several of the significant regional findings in this study were the result of decreases in normalized metabolism in responders and increases in non-responders. If global decreases in metabolism were found, this would help explain why treatment non-responders had increases in some of the ROIs. More accurate absolute metabolic data in future studies would help clarify this issue. A control group of subjects would have helped insure that changes seen here were specific to subjects with MDD and were not nonspecific changes due to paroxetine treatment or repeated PET scans. However, a previous report of repeated FDG-PET scans in healthy subjects during a similar time frame to that used here Ž2]3 months. indicated that regional metabolic changes over that period are small Žand smaller than global changes. ŽMaquet et al., 1990.. Finally, ROI placement is a limitation of this study, given that the region boundaries we defined may have differed from those in previous studies. This may explain why no significant changes were found in the DLPFC, a region often found to be associated with MDD in prior work.

Acknowledgements Supported by the National Alliance for Research in Schizophrenia and Depression ŽNARSAD. ŽA.L.B. and D.H.S.S..; the Charles A. Dana Foundation Consortium on Neuroimaging Leadership ŽS.S..; and R01 MH-53565 ŽL.R.B... The authors thank Edythe D. London, John Matochik, and Peter C. Whybrow, for their suggestions on the manuscript. References Alexander, G.E., DeLong, M.R., Strick, P.L., 1986. Parallel

137

organization of functionally segregated circuits linking basal ganglia and cortex. Annual Review of Neuroscience 9, 357]381. Alpert, N.M., Berdichevsky, D., Weise, S., Tang, J., Rauch, S.L., 1993. Stereotactic transformation of PET scans by nonlinear least squares. In: Uemura, K., Jones, J., Lassen, N.A., Kanno, I. ŽEds.., Quantification of Brain Function: Tracer Kinetics and Image Analysis in Brain PET. Elsevier Science Publishers, Amsterdam, The Netherlands, pp. 459]463. Baxter, L.R., Phelps, M.E., Mazziotta, J.C., Schwartz, J.M., Gerner, R.H., Selin, C.E., Sumida, R.M., 1985. Cerebral metabolic rates for glucose in mood disorders. Archives of General Psychiatry 42, 441]447. Baxter, L.R., Schwartz, J.M., Phelps, M.E., Mazziotta, J.C., Guze, B.H., Selin, C.E., Gerner, R.H., Sumida, R.M., 1989. Reduction of prefrontal cortex glucose metabolism common to three types of depression. Archives of General Psychiatry 46, 243]250. Baxter, L.R., Schwartz, J.M., Bergman, K.S., Szuba, M.P., Guze, B.H., Mazziotta, J.C., Alazraki, A., Selin, C.E., Ferng, H.K., Munford, P., Phelps, M.E., 1992. Caudate glucose metabolic rate changes with both drug and behavior therapy for obsessive]compulsive disorder. Archives of General Psychiatry 49, 681]689. Bench, C.J., Friston, K.J., Brown, R.G., Scott, L.C., Frackowiak, R.S., Dolan, R.J., 1992. The anatomy of melancholia } focal abnormalities of cerebral blood flow in major depression. Psychological Medicine 22, 607]615. Biver, F., Goldman, S., Delvenne, V., Luxen, A., De Maertelaer, V., Hubain, P., Mendlewicz, J., Lotstra, F., 1994. Frontal and parietal metabolic disturbances in unipolar depression. Biological Psychiatry 36, 381]388. Buchsbaum, M.S., Wu, J., DeLisi, L.E., Holcomb, H., Kessler, R., Johnson, J., King, A.C., Hazlett, E., Langston, K., Post, R.M., 1986. Frontal cortex and basal ganglia metabolic rates assessed by positron emission tomography with w 18 Fx2-deoxyglucose in affective illness. Journal of Affective Disorders 10, 137]152. Buchsbaum, M.S., Wu, J., Siegel, B.V., Hackett, E., Trenary, M., Abel, L., Reynolds, C., 1997. Effect of sertraline on regional metabolic rate in patients with affective disorder. Biological Psychiatry 41, 15]22. Cherry, S.R., Dahlbom, M., Hoffman, E.J., 1991. 3D PET using a conventional multislice tomograph without septa. Journal of Computer Assisted Tomography 15, 655]668. Devinsky, O., Morrell, M.J., Vogt, B.A., 1995. Contributions of the anterior cingulate cortex to behaviour. Brain 118, 279]306. Dolan, R.J., Bench, C.J., Liddle, P.F., Friston, K.J., Frith, C.D., Grasby, P.M., Frackowiak, R.S., 1993. Dorsolateral prefrontal cortex dysfunction in the major psychoses; symptom or disease specificity? Journal of Neurology, Neurosurgery and Psychiatry 56, 1290]1294. Drevets, W.C., 1998. Functional neuroimaging studies of depression: the anatomy of melancholia. Annual Review of Medicine 49, 341]361.

138

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139

Drevets, W.C., Raichle, M.E., 1992. Neuroanatomical circuits in depression: implications for treatment mechanisms. Psychopharmacology Bulletin 28, 261]274. Drevets, W.C., Videen, T.O., Price, J.L., Preskorn, S.H., Carmichael, S.T., Raichle, M.E., 1992. A functional anatomical study of unipolar depression. Journal of Neuroscience 12, 3628]3641. Drevets, W.C., Price, J.L., Simpson, J.R., Todd, R.D., Reich, T., Vannier, M., Raichle, M.E., 1997. Subgenual prefrontal cortex abnormalities in mood disorders. Nature 386, 824]827. Endicott, J., Spitzer, R.L., Fleiss, J.L., Cohen, J., 1976. The Global Assessment Scale. A procedure for measuring overall severity of psychiatric disturbance. Archives of General Psychiatry 33, 766]771. Feighner, J., Targum, S.D., Bennett, M.E., Roberts, D.L., Kensler, T.T., D’Amico, M.F., Hardy, S.A., 1998. A double-blind, placebo-controlled trial of nefazodone in the treatment of patients hospitalized for major depression. Journal of Clinical Psychiatry 59, 246]253. Friston, K.J., 1995. Statistical parametric mapping: ontology and current issues. Journal of Cerebral Blood Flow and Metabolism 15, 361]370. Friston, K., Frith, C., Liddle, P., Frackowiak, R., 1991. Comparing functional ŽPET. images: the assessment of significant change. Journal of Cerebral Blood Flow and Metabolism 11, 690]699. Friston, K.J., Ashburner, J., Frith, C.D., Poline, J., Heather, J.D., Frackowiak, R.S.J., 1995a. Spatial registration and normalisation of images. Human Brain Mapping 2, 165]189. Friston, K.J., Holmes, A.P., Worsley, K.J., Poline, J.P., Frith, C.D., Frackowiak, R.S.J., 1995b. Statistical parametric maps in functional imaging: a general linear approach. Human Brain Mapping 2, 189]210. Hamilton, M., 1967. Development of a rating scale for primary depressive illness. British Journal of Social Psychology 6, 278]296. Hamilton, M., 1969. Diagnosis and rating of anxiety. British Journal of Psychiatry 3, 76]79. Insel, T.R., 1992. Toward a neuroanatomy of obsessive]compulsive disorder. Archives of General Psychiatry 49, 739]744. Kennedy, S.H., Javanmard, M., Vaccarino, F.J., 1997. A review of functional neuroimaging in mood disorders: positron emission tomography and depression. Canadian Journal of Psychiatry 42, 467]475. Kennedy, S.H., Evans, K.R., Arifuzzaman, A.I., Vaccarino, F.J., Houle, S., 1998. Enhanced hypofrontality in depressed patients following chronic but not acute SSRI therapy: a w 18 FxFDG-PET study. Abstract 484.15 Annual Meeting of the Society for Neuroscience. Ketter, T.A., George, M.S., Kimbrell, T.A., Benson, B.E., Post, R.M., 1996. Functional brain imaging, limbic function and affective disorders. Neuroscientist 2, 55]65. Lin, K.P., Huang, S.C., Baxter, L., Phelps, M.E., 1994. A general technique for inter-study registration of multi-func-

tion registration of multi-function and multimodality images. IEEE Transactions in Nuclear Science 41, 2850]2855. Maquet, P., Dive, D., Salmon, E., von Frenckel, R., Franck, G., 1990. Reproducibility of cerebral glucose utilization measured by PET and the w 18 Fx-2-fluoro-2-deoxy-D-glucose method in resting, healthy human subjects. European Journal of Nuclear Medicine 16, 267]273. Mayberg, H.S., Lewis, P.J., Regenold, W., Wagner Jr., H.N., 1994. Paralimbic hypoperfusion in unipolar depression. Journal of Nuclear Medicine 35, 929]934. Mayberg, H.S., Brannan, S.K., Mahurin, R.K., Jerabek, P.A., Brickman, J.S., Tekell, J.L., Silva, J.A., McGinnis, S., Glass, T.G., Martin, C.C., Fox, P.T., 1997. Cingulate function in depression: a potential predictor of treatment response. Neuroreport 8, 1057]1061. Mayberg, H.S., Liotti, M., Brannan, S.K., McGinnis, S., Mahurin, R.K., Jerabek, P.A., Silva, J.A., Tekell, J.L., Martin, C.C., Lancaster, J.L., Fox, P.T., 1999. Reciprocal limbic-cortical function and negative mood: converging PET findings in depression and normal sadness. American Journal of Psychiatry 156, 675]682. Mazziotta, J.C., Phelps, M.E., Meadors, A.K., Ricci, A., Winter, J., Bentson, J.R., 1982. Anatomical localization schemes for use in positron computed tomography using a specially designed headholder. Journal of Computer Assisted Tomography 6, 848]853. Nadeau, S.E., Crosson, B., 1995. A guide to functional imaging of cognitive processes. Neuropsychiatry, Neuropsychology and Behavioral Neurology 8, 143]162. Nobler, M.S., Sackheim, H.A., Prohovnik, I., Moeller, J.R., Mukherjee, S., Schnur, D.B., Prudic, J., Devanand, D.P., 1994. Regional cerebral blood flow in mood disorders. III: Treatment and clinical response. Archives of General Psychiatry 51, 884]897. Parent, A., Hazrati, L.-N., 1995a. Functional anatomy of the basal ganglia. I. The cortico-basal ganglia-thalamo-cortical loop. Brain Research: Brain Research Reviews 20, 91]127. Parent, A, Hazrati, L.-N., 1995b. Functional anatomy of the basal ganglia. II. The place of the subthalamic nucleus and external pallidum in basal ganglia circuitry. Brain Research: Brain Research Reviews 20, 128]154. Post, R.M., DeLisi, L.E., Holcomb, H.H., Uhde, T.W., Cohen, R., Buchsbaum, M.S., 1987. Glucose utilization in the temporal cortex of affectively ill patients: positron emission tomography. Biological Psychiatry 22, 545]553. Press, W.H., Flannery, B.P., Teukolsky, S.A., Vetterling, W.T., 1986. Numerical Recipes. Cambridge University Press, New York. Rajkowska, G., Goldman-Rakic, P.S., 1995. Cytoarchitectonic definition of prefrontal areas in the normal human cortex: II. Variability in locations of areas 9 and 46 and relationship to the Talairach coordinate system. Cerebral Cortex 5, 323]337. Rubin, E., Sackeim, H.A., Nobler, M.S., Moeller, J.R., 1994. Brain imaging studies of antidepressant treatment. Psychiatric Annals 24, 653]658.

A.L. Brody et al. r Psychiatry Research: Neuroimaging Section 91 (1999) 127]139 Spitzer, R.L., Endicott, J., 1978. Schedule for Affective Disorders and Schizophrenia. New York State Psychiatric Institute, New York, NY. Steinmetz, H., Seitz, R.J., 1991. Functional anatomy of language processing: neuroimaging and the problem of individual variability. Neuropsychologia 29, 1149]1161. Talairach, J., Tournoux, P., 1988. Co-planar Stereotaxic Atlas of the Human Brain. Thieme, New York, NY. Uytdenhoef, P., Portelange, P., Jacquy, J., Charles, G.,

139

Linkowski, P., Mendlewicz, J., 1983. Regional cerebral blood flow and lateralized hemispheric dysfunction in depression. British Journal of Psychiatry 143, 128]132. Wheatley, D.P., van Moffaert, M., Timmerman, L., Kremer, C.M., 1998. and the Mirtazapine-Fluoxetine Study Group. Mirtazapine: efficacy and tolerability in comparison with fluoxetine in patients with moderate to severe major depressive disorder. Journal of Clinical Psychiatry 59, 306]312.