Abstract
Objectives: To investigate whether the combination of fluoro-2-deoxy-d-glucose (FDG) PET measures with the APOE genotype would improve prediction of the conversion from mild cognitive impairment (MCI) to Alzheimer disease (AD).
Method: After 1 year, 8 of 37 patients with MCI converted to AD (22%). Differences in baseline regional glucose metabolic rate (rCMRglc) across groups were assessed on a voxel-based basis using a two-factor analysis of variance with outcome (converters [n = 8] vs nonconverters [n = 29]) and APOE genotype (E4 carriers [E4+] [n = 16] vs noncarriers [E4−] [n = 21]) as grouping factors. Results were considered significant at p < 0.05, corrected for multiple comparisons.
Results: All converters showed reduced rCMRglc in the inferior parietal cortex (IPC) as compared with the nonconverters. Hypometabolism in AD-typical regions, that is, temporoparietal and posterior cingulate cortex, was found for the E4+ as compared with the E4− patients, with the E4+/converters (n = 5) having additional rCMRglc reductions within frontal areas, such as the anterior cingulate (ACC) and inferior frontal (IFC) cortex. For the whole MCI sample, IPC rCMRglc predicted conversion to AD with 84% overall diagnostic accuracy (p = 0.003). Moreover, ACC and IFC rCMRglc improved prediction for the E4+ group, yielding 100% sensitivity, 90% specificity, and 94% accuracy (p < 0.0005), thus leading to an excellent discrimination.
Conclusion: Fluoro-2-deoxy-d-glucose-PET measures may improve prediction of the conversion to Alzheimer disease, especially in combination with the APOE genotype.
Patients with mild cognitive impairment (MCI) are at higher risk for developing Alzheimer disease (AD) with an estimated conversion rate of 10 to 15% per year.1–4⇓⇓⇓ Nonetheless, MCI is a heterogeneous entity characterized by differences in cognitive profile and clinical progression, possibly due to the interplay of genetic, physiologic, and environmental factors.1,4⇓ As a result, the outcome for any patient with MCI is uncertain because many subjects may remain stable or even revert to a normal state, while others progress to dementia.1–5⇓⇓⇓⇓
There is evidence that the APOE E4 allele is the most common susceptibility gene for AD.6 It influences the course of disease by increasing the risk for developing AD and lowering the age at disease onset.6 Case-control studies have shown that the E4 allele is also overrepresented in MCI, and increased frequency of the allele turned out to be the strongest predictor of clinical progression from MCI to AD.7–9⇓⇓ Although the E4 allele alone does not implicate conversion to AD in MCI, the combination of genetic assessment with functional brain imaging is now seen as a promising preclinical AD detection strategy. Recent PET studies demonstrated an association between the E4 allele and abnormal reductions of glucose metabolism in the same brain regions as clinical AD patients in normal individuals carrying the E4 genotype.10,11⇓ It remains unknown if this is a predictor of future cognitive decline. In addition, there is evidence that the E4 allele leads to greater longitudinal metabolic decline in healthy elderly persons converting to MCI.12 Nonetheless, no study has been carried out to assess the impact of the E4 allele on brain physiology in the conversion from MCI to AD.
We investigated whether the combination of FDG-PET measures with the APOE genotype would improve prediction of the conversion from MCI to AD.
Methods.
Subjects.
Thirty-seven patients diagnosed with amnestic MCI2 were enrolled over a 2-year period, and a 1-year follow-up study was completed. The study procedures included examinations by neurologists and psychiatrists, routine laboratory tests, neuropsychological examinations, EEG, FDG-PET scans, and APOE genotyping, according to the general protocol of the Network for Efficiency and Standardization of Dementia Diagnosis research project13 (http://www.uni-koeln.de/med-fak/ neurologie/nest-dd/english/).
Patients were diagnosed as having amnestic MCI according to the following criteria2: 1) presence of subjective memory complaints in spite of preserved daily living activities14; 2) evidence of objective memory impairment, defined as scores of <1.5 SDs below the age-matched control mean on at least one of two tests of delayed recall, but no other significant deficits on detailed neuropsychological examination (see below); 3) National Institute of Neurological and Communication Disorders and Stroke/Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria for probable AD not met15; and 4) a Mini-Mental State Examination (MMSE)16 score of ≥24 and a Clinical Dementia Rating17 score of 0.5.
We excluded patients with major psychiatric or medical disease or using medications that could affect brain function or structure. Exclusion criteria were as follows: previous subarachnoid or intracerebral hemorrhage, intracranial tumors, hydrocephalus, psychosis (including major depression), alcoholism, epilepsy, ischemic stroke, vascular dementia and other dementing illnesses, anemia, untreated thyroid dysfunction, renal insufficiency, nonstabilized diabetes mellitus, or concomitant medications that could affect brain function. Patients with dementia were excluded according to NINCDS-ADRDA criteria.15
Outcome groups.
All MCI patients with a baseline FDG-PET received follow-up clinical evaluations after a 1-year interval (mean ± SD for follow-up was 12.1 ± 0.6 months). At the 1-year follow-up, all patients were re-evaluated for diagnosis and severity level. The diagnosis of probable AD was rendered according to NINCDS-ADRDA criteria15: 1) dementia established by clinical examination and documented by the Mini-Mental Test and confirmed by neuropsychological tests; 2) deficits in two or more areas of cognition; 3) progressive worsening of memory and other cognitive functions; 4) no disturbance of consciousness; 5) onset between ages 40 and 90, most often after age 65; and 6) absence of systemic disorders or other brain diseases that in and of themselves could account for the progressive deficits in memory and cognition. The diagnosis of AD was made blinded to APOE results.
APOE genotyping.
The participants provided informed consent, agreed that they would not be given information about their APOE genotype, and were studied under guidelines approved by human subject committees at our institutions. Venous blood samples were drawn and APOE genotypes analyzed by using standard PCR technique.18
Neuropsychological tests.
All subjects completed a detailed battery of neuropsychological tests including mnemonic and nonmnemonic tests (see table E-1 on the Neurology Web site at www.neurology.org).
The mnemonic tests administered, assessing delayed recall ability, were as follows: cued and free recall of the California Verbal Learning Test (CVLT) and recall of the Rey Complex Figure Test.19 The nonmnemonic tests administered were as follows: copy of the Rey Complex Figure Test, the 60-item Naming Test, the Mental Arithmetic, Digit Span, and Visual Span Subtests from the Wechsler Memory Scale, the Dual Task Test, the Token Test, the Stroop and Trail Making (TMT) Tests, and the Phonemic and Semantic Fluency Tests.19
Statistical analyses of cognitive and behavioral data.
The basic statistical model was a two-factor analysis of variance (ANOVA) with outcome group (converters vs nonconverters) and APOE group (E4 carriers [E4+] vs noncarriers [E4−]) as grouping factors. Follow-up statistical analysis was done with Scheffé tests. For tests of the longitudinal effects, the two-factor ANOVA was examined in a repeated-measures design. Logistic regression analyses were used to predict outcome and calculate sensitivity (ability to detect the converters to AD) and specificity (ability to recognize nonconverting MCI). Results were considered significant at p < 0.05. Analyses were done using Statistica 6.0 (StatSoft, Inc. Tulsa, OK).
PET procedures.
The detailed fluoro-2-deoxy-d-glucose (FDG) PET scanning procedure employed was previously described.13 PET scans were acquired at the Florence and Milan centers using the same GE Advance PET devices (Milwaukee, WI) and under common scanning procedures. In brief, emission scans were acquired in two-dimensional mode with an axial field of view of 152 mm, an in-plane full width at half-maximum (FWHM) of 4.6 mm, and slice thickness of 4.25 mm. Patients were injected with a dose of 110 to 370 MBq of [18F]FDG in a resting state with eyes closed and ears unplugged in a dimly lighted room with minimal background noise. A polycarbonate head holder was used to reduce head movement during the scan. The uptake interval between FDG injection and scan start was on average 42 ± 19 minutes. The average scan duration was 19 ± 3 minutes. Images were reconstructed using filtered back-projection including correction for attenuation measured by transmission scan and scatter using standard software as supplied by scanner manufacturers.
Basic image processing and voxel-based data analyses were performed using SPM’99 routines (Wellcome Department of Cognitive Neurology, London, UK) implemented in MATLAB (Mathworks, Sherborn, MA). PET images were spatially normalized by affine 12-parameter transformation onto a PET-FDG template that conformed to the McGill Neurologic Institute (MNI) space, which approximates the Talairach and Tournoux space.20 Normalized images were formatted to a 79 × 95 × 68 matrix with 2 × 2 × 2–mm voxel size. An isotropic Gaussian filter was used to smooth the spatially normalized PET images with an FWHM of 12 mm.21 Individual counts were normalized to mean global activity using proportional scaling to obtain relative cerebral metabolic rate for glucose (rCMRglc) values from FDG radioactivity measurements. To minimize “edge effects” without excluding hypometabolic tissue, only those voxels with values >80% of the mean for the whole brain were retained for all statistical analyses. Global calculation was obtained with respect to the mean voxel value.
Statistical analysis of PET data.
Baseline rCMRglc effects.
A two-way analysis of covariance (ANCOVA) was performed taking outcome group (converters vs nonconverters) and APOE groups (E4+ vs E4−) as grouping factors and controlling for the influence of age on rCMRglc, using SPM’99 routines. We used the general linear model (GLM), univariate procedure,21,22⇓ for the ANCOVA calculation with SPM. Each factor has an associated main effect, which is the difference between the levels of that factor, averaging over the levels of all other factors. Each pair of factors has an associated interaction. Interactions represent the degree to which the effect of one factor depends on the levels of the other factors that compose the interaction. A two-way ANOVA thus has two main effects and one interaction, which are determined by SPM for each voxel in the stereotactic space by generating factor-specific adjusted mean rCMRglc values and the associated adjusted error variance.21,22⇓ The mean rCMRglc values were then compared for each of the main effects and their interaction with the F statistics and transformed into normally distributed Z statistics. The resulting set of Z values constituted a statistical parametric map (SPM Z map). For each contrast of interest, the significances of the resulting SPMs were assessed by comparing the expected and observed distribution of the statistic under the null hypothesis of no differential group effect on rCMRglc, according to the GLM.21,22⇓ In other words, we used the GLM to create images that display main effects and interaction term, whose significance is tested for by F tests. After an overall F test had shown significance, we used post-hoc t-tests to evaluate differences among specific means. Specifically, we examined 1) the main effect of outcome, that is, rCMRglc differences between converters and nonconverters, 2) the main effect of APOE genotype, that is, rCMRglc differences between E4+ and E4− groups, and 3) outcome × APOE group interaction effects.21,22⇓ (For a detailed explanation of how the interaction term was estimated and tested, see also http://www.fil.ion.ucl.ac.uk/∼wpenny/publications/rik_anova.pdf.) We also examined whether differences between E4+ and E4− patients were detectable within the nondecliner group using an ANCOVA model controlling for age.21,22⇓
For all analyses, results were considered significant at p < 0.05, corrected for multiple comparisons according to the small-volume random field corrections.23 We used the MRIcro package (www.psychology.Nottingham.ac.uk/staff/cr1/mricro.html) to create a masking image from a set of predefined regions of interest. Based on previous FDG-PET studies10–12,24–27,29,30⇓⇓⇓⇓⇓⇓⇓⇓ we defined the precuneus (PreCu), anterior (ACC), and posterior (PCC) cingulate cortex, inferior parietal lobule (IPL), superior (STG) and middle (MiTG) temporal gyrus, and superior (SFG), middle (MiFG), and inferior frontal (IFG) gyrus, on both hemispheres, as candidate areas for possible rCMRglc alterations. Such mask was then applied to the full volume of data, and results were examined at p < 0.05 after correction for the number of comparisons in the searched volume, as defined by the masking image (156,400 mm3 corresponding to 19,550 voxels).23 Such a conservative procedure reduces the potential for Type 1 error but does not allow the assessment of significant results in the brain regions outside the masking image. We thereby assessed the results also for every voxel in the whole brain (i.e., without correction for the number of comparisons in the mask image), only for exploratory purposes. Results were considered significant at p < 0.001, uncorrected for multiple comparisons. Brain areas reaching the significance threshold were identified in terms of voxel coordinates and labeled according to the Talairach and Tournoux space,20 after coordinate conversion from MNI to the Talairach space using Brett’s set of linear transformations (http://www.mrc-cbu.cam.ac.uk/Imaging/).
Baseline rCMRglc prediction study.
Normalized rCMRglc data (nCMRglc) were extracted from the brain regions showing significant differences across groups in the above analyses, using Marsbar toolbox (http://www.mrc-cbu.cam.ac.uk/Imaging/marsbar.html). To this aim, nCMRglc values were extracted from each set of voxels (cluster) in which significant differences were identified through the SPM analyses, and the average nCMRglc was calculated for each cluster.
Logistic regression analyses were used to assess whether baseline nCMRglc was predictive of longitudinal clinical outcome, for the whole sample group and separately for each outcome and APOE group. Results were considered significant at p < 0.05. Statistical analyses were computed with Statistica (version 6.0).
Results.
Of the 37 MCI patients enrolled, 16 were E4 allele carriers (E4+), including 14 heterozygotes (E3/E4 genotype) and 2 homozygotes (E4/E4 genotype), and 21 were noncarriers of the E4 allele (E4−), including 18 patients with the E3/E3 genotype and 3 patients with the E2/E3 genotype.
After 1-year follow-up, conversion to AD was found for 8 of 37 MCI patients (22%), a proportion of conversions similar to that reported in the literature.1–5⇓⇓⇓⇓ Of those patients who converted to AD, there were 5 of 16 E4+ (31%) and 3 of 21 E4− (14%). The frequency of converters was comparable between APOE groups (χ2 = 1.576, p = 0.21) after 1 year.
ANOVAs showed no demographic differences between outcome groups and between APOE groups for age, gender, age at onset, disease duration, education, family history of AD, and depressive symptoms at the time of either baseline and follow-up examination. Time to follow-up was equivalent. Baseline demographic and clinical characteristics of subject groups are shown in table 1.
Table 1 Baseline clinical and demographic characteristics
Neuropsychological study.
Scores on the MMSE (F[1,33] = 5.92, p < 0.05) and on the delayed Rey Complex Figure Recall Test (F[1,33] = 7.52, p < 0.01) were different between outcome groups, with those who later converted having lower scores than the nonconverters. Also, the APOE groups differed with respect to the Rey Recall Test, with the E4+ group having lower scores than the E4− group (F[1,33] = 4.04, p < 0.05) (see table E-1). However, no outcome × APOE group interaction was found for any of the other measures (see table E-1).
Only scores on the MMSE predicted outcome with 81% accuracy, 96% specificity, but only 25% sensitivity (χ2 = 6.32, p < 0.01). Performing the above analyses separately for each APOE group did not improve overall classification sensitivity, in that results were the same for both APOE groups.
After 1-year follow-up, the converters showed cross-sectional reductions with respect to the nonconverters in the MMSE (F[1,33] = 33.8, p < 0.001) and several delayed recall measures (p < 0.01) and a poorer performance on tests assessing executive functions such as the Stroop Test interference condition (F[1,33] = 11.31, p < 0.002) and the TMT (F[1,33] = 6.87, p < 0.01) (see table E-1). Reductions (p < 0.01) were found for the E4+ group on the MMSE and the delayed recall items of the Rey Complex Figure Test and CVLT as compared with the E4− group (see table E-1). There was no outcome × APOE group interaction.
The converters showed poorer follow-up performance on the MMSE (F[1,35] = 25.95, p < 0.001), long delayed free recall item of the CVLT (F[1,35] = 6.36, p < 0.02), Token Test (F[1,35] = 10.85, p < 0.002), and the Stroop Test interference condition (F[1,35] = 7.07, p < 0.02) as compared with the baseline scores.
The E4+ showed lower follow-up scores for the MMSE (F[1,35] = 7.27, p < 0.01) than baseline scores. No outcome × APOE group interactions were observed in any of the measures.
PET study.
Baseline rCMRglc effects.
On two-way ANOVA, outcome group effects were found in the right IPL (Brodmann area [BA] 40), with the converters having lower rCMRglc than the nonconverters (see table E-2; figure 1A).
Figure 1. PET voxels that demonstrate significant relative cerebral metabolic rate for glucose (rCMRglc) reductions for the converters as compared with the nonconverters (A), significant rCMRglc reductions for the E4 carriers as compared with the noncarriers (B), and significant interactions between outcome and APOE group effects (C).
Areas of significant hypometabolism are displayed onto the anterior, posterior, medial, lateral, inferior, and superior views of a volume-rendered spatially normalized MR image. Results are displayed at p < 0.05, corrected for multiple comparisons. Local maxima of the above brain regions can be found in table E-2 on the Neurology Web site (www.neurology.org).
Baseline APOE group effects were found within the PreCu (BA 7) and PCC (BA 31), MiTG (BA 21), and IPL (BA 39/40), bilaterally, with the E4+ patients having lower rCMRglc than the E4− patients (see table E-2; see also figure 1B). Conversely, no rCMRglc reduction was found for the E4− group as compared with the E4+ group, even by assessing results without correction for multiple comparisons.
An interaction between outcome and APOE groups was found in the left ACC (BA 32) and IFG (BA 47) (see table E-2; see also figure 1C). Post-hoc analysis revealed that the interaction was driven by the E4+/converters, who had lower rCMRglc than the other subgroups (figure 2).
Figure 2. Results from post-hoc analysis of the outcome × APOE interaction effects. The figure was obtained by plotting the percentage relative cerebral metabolic rate for glucose (rCMRglc) change at the peak height of significance from the brain regions listed in table E-2 (on the Neurology Web site at www.neurology.org), in which an interaction between outcome and APOE groups was found on two-way analysis of variance (p < 0.05, corrected for multiple comparisons). As displayed, such interaction was driven by the E4 carriers who converted to Alzheimer disease, who show lower rCMRglc as compared with the other subgroups. E4+ = E4 carriers (black); E4− = noncarriers (white).
Given the small number of conversions, ACC and IFG rCMRglc data were re-examined with the nonparametric Kruskal–Wallis ANOVA by ranks with Dunn post-hoc tests,31 using Statistica 6.0. The Kruskal–Wallis ANOVA assesses the hypothesis that the different samples in the comparison are drawn from the same distribution or from distributions with the same median.31 The Dunn post-hoc tests compare the difference in the sum of ranks between the paired groups with the expected average difference (based on the number of groups and their size).31 Results were assessed at p < 0.05. Metabolic differences between groups were found for both the ACC (Kruskal–Wallis H[3, n = 37] = 9.02, p < 0.03) and IFG (Kruskal–Wallis H[3, n = 37] = 9.39, p < 0.02). Post-hoc tests showed that the E4+/converters had lower median ACC and IFG rCMRglc than the E4+/nonconverters (ACC: p < 0.005, IFG: p < 0.02), E4−/converters (ACC: p < 0.05, IFG: p < 0.03), and E4−/nonconverters (ACC: p < 0.05, IFG: p < 0.04). No difference was found across the other groups (figure 3).
Figure 3. Results from Kruskal–Wallis analysis of variance by rank test. Significant relative cerebral metabolic rate for glucose (rCMRglc) differences between groups were found for both (top) anterior cingulate cortex (p < 0.03) and (bottom) inferior frontal gyrus (p < 0.02). The results show that the E4+/converters have the lowest median metabolic rate as compared with the other subgroups. Whiskers = min–max; boxes = 25 to 75% percentiles; black squares = median values. E4+ = E4 carriers; E4− = noncarriers; Conv = converters; Non-Conv = nonconverters.
Moreover, we removed all the patients who declined from the analysis and compared the remaining E4+ patients (n = 11) with the E4− patients (n = 18). As figure 4 displays, among nonconverters, the E4+ patients show lower rCMRglc within the MiTG (BA 21), IPL (BA 40), and PreCu (BA 7), bilaterally, as compared with the E4− patients (see table E-3).
Figure 4. PET voxels that demonstrate significant relative cerebral metabolic rate for glucose (rCMRglc) reductions for the E4 carriers (n = 11) as compared with the E4 noncarriers (n = 18) after removal of those patients who later converted to Alzheimer disease. (Left) Results (black areas) are displayed as statistical parametric mapping projections in the three orthogonal views: right sagittal (top), posterior coronal (middle), and superior axial (bottom). Results are displayed at p < 0.05, corrected for multiple comparisons. (Right) Percentage rCMRglc values (normalized to a mean value of 50 μmol/100 g/min) extracted at the peak height of significance from the brain regions listed in table E-3 (on the Neurology Web site at www. neurology.org) are plotted for the E4 carriers (E4+, gray) and noncarriers (E4−, white). PreCu = precuneus; IPL = inferior parietal lobule; STG = superior temporal gyrus.
For all analyses, the results remained substantially unchanged by resetting the probability threshold at p < 0.001, uncorrected for multiple comparisons, and by correcting for the baseline MMSE scores.
Baseline prediction effects.
For the whole MCI group, baseline rCMRglc in the right IPL predicted conversion to AD with 84% accuracy, 38% sensitivity, and 97% specificity (χ2 = 6.04, p < 0.01) (see table E-4). Adding the IPL rCMRglc to the MMSE scores, the only neuropsychological measure that predicted conversion increased the overall diagnostic sensitivity for the logistic regression model from 25 to 50%, with 87% accuracy and 97% specificity (χ2 = 11.71, p < 0.003).
As table E-4 on the Neurology Web site shows, such low sensitivity was due mainly to the poor discrimination between the E4+ who converted from those who did not. We then performed the logistic regression analysis separately for the APOE groups and restricted the results to the E4+ group because of the small sample size of the E4− group. Metabolism in the IFG and ACC, in which we found an outcome × APOE interaction, discriminated the converters from the nonconverters for the whole MCI group with 87% accuracy (75% sensitivity and 89% specificity) (χ2 = 4.26, p < 0.05) and was the best predictor of outcome for the E4+ group with 100% sensitivity, 90% specificity, and 93% overall accuracy (χ2 = 18.46, p < 0.0005), thus leading to excellent discrimination (see table E-4). Combining rCMRglc data with MMSE scores did not further improve classification.
For the sake of completeness, we also performed the logistic analyses for the E4− group and found that IPL rCMRglc predicted outcome with an overall accuracy of 86% (χ2 = 6.37, p < 0.01), with the combination of IPL rCMRglc and baseline MMSE scores resulting in a diagnostic accuracy of 95% (χ2 = 14.5, p < 0.001) (see table E-4).
Discussion.
These results demonstrate that FDG-PET measurements predict conversion to dementia among MCI patients and are especially sensitive in combination with the APOE genotype. We showed that the reduced baseline parietal rCMRglc is a sensitive predictor of conversion to AD. Moreover, there is an interaction with the APOE genotype, such that ACC and IFG rCMRglc predicted conversion for the E4 allele carriers with excellent accuracy (93%) and discrimination capacity (100% sensitivity). These results suggest that the identification of different regional patterns of hypometabolism may help identifying prognostic genetic subgroups and stress the importance of biologic markers in the early assessment of AD.
The initial pattern of cerebral metabolism allowed differentiation of distinct prognostic groups in a population of patients with MCI with comparable neuropsychological status. MCI patients who converted to AD within 1 year showed significantly lower rCMRglc in the right IPL as compared with those who did not convert. Such pattern of reduced rCMRglc in the converters is consistent with previous PET reports of parietal hypometabolism in patients with very early AD24,26⇓ and is consistent with the few prior PET studies that showed an association between temporoparietal hypometabolism and the conversion from MCI to AD.27,29,30⇓⇓ However, none of the above studies examined the impact of the APOE genotype on such metabolic abnormalities.
For both the total MCI group and the nonconverters separately, in comparison with the E4 noncarriers, MCI patients carrying the E4 allele showed the largest rCMRglc reduction in the PCC and PreCu, according to the well-established observation of reduced metabolism within these areas both in MCI27,28,30⇓⇓ and in very early AD.24–26⇓⇓ Moreover, the E4 carriers presented with metabolic reductions within the temporoparietal areas, which are considered a reliable marker of a very early AD stage,13 the extent of hypometabolism being correlated with severity of cognitive impairment.13,32⇓
Overall, the current results indicate that the APOE E4 allele, the main genetic risk factor for AD,6,18⇓ contributes to heterogeneity in MCI by regionally affecting brain metabolic activity, so that MCI patients who are carriers of the E4 allele present more extended rCMRglc reductions than the noncarriers, with a metabolic pattern suggestive of AD, which may reflect the known E4-related increased vulnerability to dementia. These data provide evidence for APOE E4-related alterations of brain metabolism in clinical MCI patients that could close the gap between previous observations in presymptomatic elderly individuals at risk for familial AD10,11⇓ and clinical AD patients.33,34⇓ These cross-sectional findings are in agreement with prior FDG-PET studies that have shown how at least one copy of the E4 allele is associated with lowered rCMRglc in healthy elders at familial risk for AD, who demonstrated a faster cognitive and metabolic decline with respect to the noncarriers.10,11⇓ The metabolic reduction involved the posterior cingulate, precuneus, parietotemporal, and frontal areas and was evident in the absence of dementia and cortical atrophy.10 Our findings are thus consistent with prior reports of reductions in brain metabolism related to the E4 genotype.
Clues about the mechanisms and pathways by which the E4 allele may influence brain activity in MCI have been provided in a recent FDG-PET study in which the authors monitored the transition from healthy aging to MCI and examined the contribution of the APOE genotype to the onset of cerebral hypometabolism.12 At follow-up and among those subjects who declined to MCI, the E4 carriers showed marked temporal metabolic reductions.12 Our data are consistent with their findings of an E4-related temporal hypometabolism emerging in MCI and extend this association to the frontal and anterior cingulate cortex.
Overall, these findings extend data from the above investigations on normal aging and AD to MCI, showing that MCI patients with an E4 genotype present with intermediate metabolic features between MCI and AD. In addition, we found a significant interaction between outcome and APOE genotype within two frontal lobe regions, the IFG and ACC. On post-hoc analysis, such interaction was driven by the converted E4 carriers, who showed the lowest rCMRglc as compared with the other subgroups. Although our data were obtained with a small sample of patients and therefore deserve to be replicated, these results are consistent with previous PET studies that demonstrated an association between conversion to AD and rCMRglc decreases in the frontal areas27 and ACC,25 thus suggesting that the symptomatic shift between MCI and AD may be related to the appearance of such metabolic dysfunction. The frontal areas have been involved with AD severity and progression,35 although several PET and MRI studies showed that loss of function within these areas is preferentially associated with the aging process per se.13,36⇓ The current findings support the hypothesis that such additional E4-related reductions in frontal rCMRglc, which couple prior observations in healthy elders10–12⇓⇓ and AD patients,37 may reflect an accelerated aging process. Considerable literature has indeed associated the E4 allele with ineffectual aging-related cellular mechanisms and poorer neuroreparative ability; enhanced deposition of abnormal protein aggregates throughout the brain, and disrupted neurotransmission, hence an earlier onset sign of AD pathophysiology.38,39⇓
A main finding from the cross-sectional PET analysis was the diagnostic gain when rCMRglc measurement was added to the MMSE measure, replicating previous findings.29,40,41⇓⇓ Including APOE genotypes further improves diagnostic accuracy and sensitivity from good to excellent discrimination between outcome groups. Specifically, we found that rCMRglc within the ACC and IFG predicted conversion to AD for the E4 carriers with excellent discrimination capacity and without overlap, improving diagnostic sensitivity over the MMSE scores from 25 to 100%.
It is important to acknowledge possible limitations of the current study. First, a correction for brain atrophy was not performed, and results of reduced metabolism could thus represent either pronounced metabolic dysfunction or the combined effects of hypometabolism and atrophy. However, most studies in which voxel-based atrophy correction of resting glucose metabolism was performed reported a relative independence of metabolic measurements from brain atrophy.42 Nonetheless, in previous studies on normal elderly,12 correction for atrophy generally led to increased statistical values, improving the detection of significant rCMRglc alterations. In fact, it has been recently shown that voxel-based analysis possibly fails to detect hypometabolism within small brain structures such as entorhinal cortex and hippocampus, especially during predementia stages when alterations in brain metabolism are still of moderate extent.12,28⇓ On the basis of the increasingly important role assigned to the limbic structures in MCI and early AD28,43,44⇓⇓ and of previous findings of greater hippocampal atrophy in E4 carrier AD patients,45 future MRI-guided region-of-interest studies with FDG-PET are essential to assess whether MCI carriers of the E4 allele are also more hypometabolic within the hippocampus and surrounding structures.
In addition, studies with larger numbers of MCI converters are needed to directly compare frontal metabolism in subjects with and without the E4 allele, consistent with the hypothesis that a separate metabolic threshold in the frontal lobes can portend the conversion of E4 carriers to AD from MCI. The same would be useful to assess the accuracy of memory measures to predict conversion. Actually, in the current study, only the MMSE scores predicted conversion, possibly reflecting low statistical power. We thereby examined the other baseline neuropsychological measures using the Cohen d test,46 which, unlike significance tests, is independent of sample size. Although d is only a descriptive measure, the results showed that the differences between converters and nonconverters on the Rey Figure Delayed Recall Test, Stroop Test, TMT, and semantic fluency measures are not negligible (table 2).
Table 2 Baseline cognitive performance scores for MCI patients
The current findings show that FDG-PET measures improve prediction of the conversion to AD and that these measures are especially sensitive among APOE E4 carriers. Longer follow-up examination of our patients is necessary to determine whether such E4-related rCMRglc reductions are predictive of conversion to AD or other clinical outcomes. These rCMRglc reductions may precede dementia onset. Thus, PET examination in E4 carriers may have the potential for preclinical detection of following clinical changes and might allow intervention before irreversible brain damage occurs.
Acknowledgments
Supported by European Union grants QLK-6-CT-1999-02178 and QLK-6-CT-1999-02112. Work at Florence was also supported by grants from MURST, Telethon Italia Fondazione ONLUS (grant E.0980), and Regione Toscana, Progetto Ministero della Sanità, “Diagnosi Tempestiva e Differenziale della Malattia di Alzheimer.” Work at CRC Liege was also supported by grants from FNRS and FMRE, Belgium.
The authors thank those people who contributed to the subjects’ care and to the collection of PET images and clinical reports. They also thank Drs. Mony de Leon and Susan De Santi for the critical revision of the paper and Dr. Juan Li for statistical advice.
Footnotes
Additional material related to this article can be found on the Neurology Web site. Go to www.neurology.org and scroll down the Table of Contents for the December 28 issue to find the title link for this article.
- Received April 23, 2004.
- Accepted in final form September 2, 2004.
References
Letters: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Alert Me
Recommended articles
Neuroanatomic basis of amnestic MCI differs in patients with and without Parkinson disease
Longitudinal modeling of frontal cognition in APOE ε4 homozygotes, heterozygotes, and noncarriers
Association of Dipeptidyl Peptidase-4 Inhibitor Use and Amyloid Burden in Patients With Diabetes and AD-Related Cognitive Impairment
Topographic Distribution of Amyloid-β, Tau, and Atrophy in Patients With Behavioral/Dysexecutive Alzheimer Disease