Thirty-six MCI patients including 18 MCI_D and 18 MCI_ND cases matched for age and gender were recruited from a cohort regularly followed at the Dementia & Age-Associated Cognitive Decline Clinic at Seoul National University Hospital. MCI was diagnosed according to current consensus criteria for amnestic MCI.¹⁰ Depression was diagnosed when a subject met the major or minor depressive disorder criteria of the fourth edition of the Diagnostic Statistical Manual of Mental Disorders 4th edition. All MCI individuals had an overall clinical dementia rating (CDR)¹¹ of 0.5. Sixteen cognitively normal (CN) elderly individuals were also recruited. All CN subjects received a CDR score of 0 and did not meet depression criteria.

The following exclusion criteria were applied to all subjects: any present serious medical or neurological disorder that could affect mental function; any major psychiatric disorders except depression, evidence of focal brain lesions on magnetic resonance imaging (MRI) including lacunes and white matter hyperintensity lesions of grade 2 or more on the Fazeka scale,¹² the presence of severe behavioral or communication problems that would make a clinical or cerebral imaging of MRI and PET examination difficult; ambidextrousness or left-handedness; and the absence of a reliable informant. The Institutional Review Board of Seoul National University Hospital, approved the study protocol and informed consent was obtained from all study subjects and their relatives.

All subjects received a standardized clinical assessment and MRI according to the protocol of the Korean version of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) Assessment Packet.¹³

The CERAD neuropsychological battery includes eight neuropsychological tests: Verbal Fluency; 15-item Boston Naming Test; Mini-Mental State Examination (MMSE); World List Memory; Word List Recall; Word List Recognition; Constructional Praxis (CP); Constructional Recall, and was administered by experienced clinical psychologists. Depression severity was assessed using the 17-item Hamilton Rating Scale for Depression (HRSD)¹⁴ included in Clinical Assessment of Depression in Dementia.² A panel consisting of four neuropsychiatrists with expertise in dementia research made clinical decisions, including the assignment of clinical diagnosis and CDR rating. All clinical assessments were carried out within 4 weeks of PET examination.

PET studies were performed using an ECAT EXACT 47 scanner (Siemens-CTI, Knoxville, TN) with an intrinsic resolution of a 5.2 mm full width at half maximum and 47 contiguous transverse plane images with a 3.4 mm thickness for a longitudinal field of view of 16.2 cm. Before administering FDG, transmission scanning was performed, using 3 germanium-68 rod sources to correct attenuation. Static emission scans began 30 minutes after the intravenous injection of 370 MBq (10 mCi) FDG and were continued for 30 minutes. All FDG-PET scans were performed in a dimly lit room with minimal auditory stimulation during PET scanning, with the subject in the supine position with their eyes closed to minimize confounding effects due to activity. Transaxial images were reconstructed using a filtered back-projection algorithm employing a Shepp-Logan filter with a cutoff frequency of 0.3 cycles/pixel and 128×128×47 matrices of size of 2.1×2.1×3.4 mm.

Imaging data were analyzed using Statistical Parametric Mapping (SPM) 2 (Institute of Neurology, University College of London, UK) implemented in the Matlab (Mathworks Inc, USA). Before statistical analysis, all images were spatially normalized to the Montreal Neurological Institute (MNI, McGill University, Montreal, CA, USA) space to correct intersubject anatomical variabilities. An affine transformation was performed to determine the 12 optimal parameters essential for registering the brain on the MNI template. Subtle differences between transformed images and the template were removed by a nonlinear registration method using the weighted sum of predefined smooth basis functions used in a discrete cosine transformation. Using an in-house Matlab-based program, the glucose metabolism value of each voxel was normalized for pontine value, which was extracted for each scan, because glucose metabolism in the pons tends to be relatively preserved in AD.¹⁵ In this program, the volume of interest for the pons was predefined in the MNI space and applied to the individual spatially normalized PET images to measure the mean pontine activity. Normalized images were smoothed by convolution using an isotropic Gaussian kernel with 16 mm full width at half maximum to accommodate intersubject differences in gyral and functional anatomies and to increase dataset signal-to noise ratios.

Differences in glucose metabolic values between MCI_D, MCI_ND and CN were estimated on a voxel-by-voxel basis using the t-test. The resulting set of t values constituted the SPM(t) map. The SPM(t) was then transformed into a normal distribution to give a SPM(Z).¹⁶ We applied both p value <0.001 (uncorrected for multiple comparisons) as a significance height threshold and 50 voxels as a cluster size threshold to decrease the probability of detecting false positives.^17,18 The MNI coordinates of the local maximum of each voxel cluster were automatically calculated in SPM, and could be transformed to Talairach and Taurnoux¹⁹ by "min2tal" program (http://imaging.mrc-cbu.cam.ac.uk/downloads/MNI2tal/).

To further characterize regional glucose metabolic alterations associated with depression, we extracted normalized metabolism values at the local maximum of voxel cluster showing a significant difference between MCI_D and MCI_ND in voxel-by-voxel analysis using the voxel of interest module of SPM. The extracted metabolism values of the region were compared between MCI_D, MCI_ND and CN using analysis of variance (ANOVA) and post hoc contrasts with Tukey's method using Statistical Package for the Social Sciences (SPSS, SPSS Inc., Chicago, IL, USA). A partial correlation between depression severity, as measured by HRSD, and the extracted metabolism values was also assessed, controlling for age as a covariate for all the MCI subjects, using SPSS. In these analyses of extracted values, a p value of <0.05 was applied as the threshold of statistical significance.

The demographic and clinical data from MCI_D, MCI_ND and CN subjects were compared by ANOVA and post hoc contrasts with Tukey's method for continuous variables and χ² tests for categorical variables using SPSS.

Demographic and clinical characteristics for subjects are presented in Table 1. MCI_D, MCI_ND and CN did not differ in regard to age, gender or education. For all eight neuropsychological tests, including MMSE, there was no significant difference between the MCI_D and MCI_ND groups. In contrast, both MCI_D and MCI_ND had significantly lower scores than CN for most of the neuropsychological tests except CP. Only MCI_D showed lower CP score than CN, but MCI_ND did not. MCI_D subjects also had significantly higher HRSD scores than did either MCI_ND or CN individuals.

Voxel-wise comparison revealed that MCI_D had significantly lower glucose metabolism in the right superior frontal region [Brodmann's area (BA) 6] than MCI_ND (Table 2, Figure 1). In contrast, there was no brain region where MCI_D showed increased glucose metabolism compared with MCI_ND. Voxel-based analysis also showed that MCI_D had significantly lower glucose metabolism in the bilateral precuneus (BA 31 and BA 7), right superior parietal gyrus (BA 6), left postcentral gyrus (BA 47), left fusiform gyrus (BA 39), and left parahippocampal gyrus (BA 6) than CN. MCI_ND also showed marginally significant hypometabolism in the right precuneus (BA 31), compared with CN (Table 2).

To characterize further glucose metabolic alteration associated with depression on the right superior frontal region, we extracted normalized metabolism values at the local maximum of voxel cluster showing a significant difference between MCI_D and MCI_ND groups using the voxel of interest module of SPM. There was a significant between-group difference in the extracted regional metabolism value between MCI_D, MCI_ND, and CN groups (F=7.87, df=2; p<0.005). As shown in Figure 2, both MCI_ND and CN groups had significantly higher glucose metabolism than MCI_D, while there was no significant difference between MCI_ND and CN groups. There was also a significant correlation between HRSD scores and the extracted metabolism values in overall MCI subjects (partial correlation coefficient=-0.418; p=0.012)(Figure 3).