PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptNIH Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
AJNR Am J Neuroradiol. Author manuscript; available in PMC Mar 16, 2009.
Published in final edited form as:
PMCID: PMC2656417
NIHMSID: NIHMS73915
Temporoparietal MRI Measures of Atrophy in Subjects with Mild Cognitive Impairment that Predict Subsequent Diagnosis of Alzheimer’s Disease
Rahul S. Desikan, BA,1 Howard J. Cabral, PhD,2 Bruce Fischl, PhD,3,4 Charles R. G. Guttmann, MD,5 Deborah Blacker, MD,6 Bradley T. Hyman, MD,7 Marilyn S. Albert, PhD,8 and Ronald J. Killiany, PhD1,5,6,9
1 Department of Anatomy and Neurobiology, Boston University School of Medicine, USA
2 Department of Biostatistics, Boston University School of Public Health, USA
3 Athinoula A. Martinos Center for Biomedical Imaging, Department of Radiology, Massachusetts General Hospital, USA
4 Computer Science and Artificial Intelligence Laboratory (CSAIL), Massachusetts Institute of Technology, USA
5 Department of Radiology, Brigham and Women’s Hospital, USA
6 Department of Psychiatry, Massachusetts General Hospital, USA
7 Department of Neurology, Massachusetts General Hospital, USA
8 Department of Neurology, Johns Hopkins University School of Medicine, USA
9 Department of Environmental Health, Boston University School of Public Health, USA
Correspondence should be addressed to: Ronald J. Killiany, Ph.D. Department of Anatomy and Neurobiology, Center for Biomedical Imaging, Boston University School of Medicine, Boston, Massachusetts, USA 02118, Phone: 617-638-8082, Fax: 617-638-4922, killiany/at/bu.edu
Background and Purpose
Mild cognitive impairment (MCI) represents a transitional state between normal aging and Alzheimer’s disease (AD). Our goal was to determine if specific temporoparietal regions can predict the time to progress from MCI to AD.
Methods
MRI scans from 129 individuals with MCI were analyzed to identify the volume of 14 neocortical and 2 non-neocortical brain regions, comprising the temporal and parietal lobes. In addition, three neuropsychological test scores were included to determine whether they would provide independent information. After a mean follow-up time of 5 years, 44 of these individuals had progressed to a diagnosis of AD.
Results
Cox proportional hazards models demonstrated significant effects for six MRI regions with the greatest differences being: entorhinal cortex (HR=0.54, p < 0.001), inferior parietal lobule (HR=0.64, p <0.005), and middle temporal gyrus (HR=0.64, p < 0.004), indicating decreased risk with larger volumes. A multivariable model showed that a combination of the entorhinal cortex (HR = 0.60, p < 0.001) and inferior parietal lobule (HR = 0.62, p < 0.01) was the ‘best’ predictor of time to progress to AD. A multivariable model re-iterated the importance of included both MRI and neuropsychological variables in the final model.
Conclusion
These findings reaffirm the importance of the entorhinal cortex and present evidence for the importance of the inferior parietal lobule as a predictor of time to progress from MCI to AD. The inclusion of neuropsychological performance in the final model continues to highlight the importance of using these measures in a complementary fashion.
Keywords: MRI, prodromal AD, MCI, temporal lobe, parietal lobe
Individuals classified with mild cognitive impairment (MCI) experience memory loss to a greater extent than expected for age and progress to a diagnosis of Alzheimer’s disease (AD) at a faster rate than controls.1 It has been hypothesized that MCI represents a transitional phase between normal function and AD2 for many individuals. A number of magnetic resonance imaging (MRI) studies of MCI subjects have demonstrated that selected brain regions within the medial temporal lobe, particularly the hippocampus and entorhinal cortex, are reduced in volume in MCI subjects in comparison to control subjects. Subjects with MCI that show such reductions are at increased risk for progression to AD (for a review of this topic see references 46). These findings are consistent with neuropathological reports that have demonstrated the entorhinal cortex and hippocampus have considerable neuropathology early in the course of AD.79
Fewer MRI studies have been conducted examining the role of additional temporoparietal regions (beyond the entorhinal cortex and hippocampus) in the earliest stages of AD. Of these studies, some have manually drawn regions of interest (ROIs) within temporoparietal regions, such as the fusiform and superior temporal gyrus.1011 In addition, others have utilized whole brain measures, such as voxel-based morphometry1214, fluid registration methods1516, and cortical thickness approaches1718, thus avoiding the need to measure individual anatomic areas. Taken together the results of these studies provide evidence that areas within the parietal and lateral temporal lobes may additionally be involved in the earliest stages of AD. However, it remains unclear which specific regions or which combination of these regions beyond the medial temporal region best predicts progression of disease from MCI to AD.
The present study was undertaken in order to examine which temporoparietal regions best predict progression from MCI to AD. Here, we examined 14 neocortical and 2 non-neocortical regions of interest (ROIs) on MRI scans, comprising the temporal and parietal lobes, obtained from 129 individuals with MCI, who were subsequently followed over time. After a mean follow-up interval of 5.0 years, 44 of these individuals had progressed to a diagnosis of AD. It was therefore possible, using Cox proportional hazard models, to determine which specific temporoparietal regions, alone or in combination, could be used to predict time to progress from MCI to a diagnosis of AD. Neuropsychological measures of episodic memory and executive function were included for these subjects in order to test whether the inclusion of these measures in the models provided predictive information beyond the MRI measures.
Selection of Participants
A total of 129individuals were included in this study. They were recruited through the print media (rather than from a clinical or other medical referral source). The advertisements for subjects indicated that a research study was seeking individuals both with and without memory difficulty.
Potential subjects underwent a multistage screening procedure. The details of the screening procedures have been described elsewhere.19 Briefly, to be included in the study, participants had to be aged 65 years and older, to have an informant who could provide information about their daily function, to be free of significant underlying medical, neurologic or psychiatric illness, and to be willing to participate in the study procedures. In addition, individuals with evidence of major vascular risk factors (e.g., atrial fibrillation, insulin dependent diabetes mellitus, cerebral infarcts, etc.) were excluded. The subjects in the present study were selected because they were mildly impaired but non-demented, and had a Clinical Dementia Rating (CDR) score20 of CDR=0.5.
The study procedures also included a medical evaluation (consisting of a physical examination and medical history, electrocardiogram and standard laboratory tests), a semi-structured interview, neuropsychological testing, an MRI scan, a single photon emission computed tomography scan, and blood withdrawn for genetic analysis. All subjects provided informed consent prior to the initiation of the study, in accordance with the requirements of the Human Research Committee of Massachusetts General Hospital (Boston, MA).
Assessment of Clinical Severity
The degree of clinical severity of the subjects was evaluated by an annual semi-structured interview. This interview generates both an overall CDR rating and a measure known as the CDR Sum of Boxes (CDR-SB). 21 The interview was based on the Initial Subject Protocol that was used in the development of the CDR scale20 and includes a set of questions regarding functional status asked of the subject and a collateral source (e.g., family member, friend), along with a standardized neurologic, psychiatric, and mental status evaluation of the subject. The mental status evaluation included: the Blessed Memory and Orientation Test22 which assessed episodic memory, working memory, and orientation, a set of similarities and differences, which assessed executive function, calculations that assessed arithmetic skill and general knowledge, and a standardized language evaluation, including naming, repetition, and comprehension. To be sensitive to clinical impairments at the mildest end of the spectrum, a special set of questions were added to the interview, and the reliability and validity of the revised interview was examined.23 The mean inter-rater reliability of the CDR ratings in the context of the present study was high (r = 0.99, p< 0.0001), as was the inter-rater reliability of the 6 CDR subcategories (r = 0.90) that were used to generate the overall CDR rating.23 The CDR Sum of Boxes (CDR-SB) represents the sum of the ratings in each of the 6 CDR subcategories, thus the inter-rater reliability for this measure was also high.
In the current study, each interview was administered by a masters or doctoral level clinician (e.g., psychiatrist, neuropsychologist, or physician’s assistant) and was performed without knowledge of the other study procedures, including the MRI findings. The interview took approximately 1–2 hours to complete. A consensus review of each case was conducted annually by 2 or more members of the research group (which included the interviewers mentioned above).
Group Characteristics at Baseline and at Follow-up
Baseline
A total of 129 mildly impaired individuals, with a mean CDR-SB score of 1.3 (SD=0.8), and a range of.5 to 3.5, were examined in this study. Table 1 shows the mean age, educational status, Mini-Mental State Exam (MMSE) 24 scores, gender distribution and apolipoprotein E status of the subjects. In general, the subjects were well educated and had high scores on the MMSE.
Table 1
Table 1
Descriptive statistical information for the subjects in the study (means listed with standard deviations in parentheses)
The distribution of CDR-SB scores among the mildly impaired subjects was broad (see Table 1). At the mild end of the spectrum (i.e., CDR-SB = 0.5–1.5), many subjects would not meet psychometric cut-offs commonly used to select MCI subjects in epidemiological studies and clinical trials. 2526 The subjects at the more impaired end of the spectrum (i.e., CDR-SB ≥ 2) are comparable to MCI subjects recruited from these settings, based on likelihood of progression to a diagnosis of AD.23 We use MCI here to refer to the entire group of mildly impaired subjects. A retrospective review of the cases indicated that approximately two thirds would fall into the category of amnestic MCI, while approximately one third would be considered non-amnestic MCI cases, based on the revised criteria for MCI. 27
Follow-up
Of the 129 individuals who were mildly impaired at baseline, 44 subsequently received a clinical diagnosis of AD (mean follow-up time was 5.0+/− 3.6 years), while 85 remained mildly impaired (mean follow-up time was 6.9 +/− 4.4 years). Of those who remained mildly impaired at follow-up, 54 had CDR-SB scores that declined but their impairments had not progressed to the point where they received a diagnosis of AD, 28 had CDR-SB scores that remained stable and 3 had CDR-SB scores that increased. Approximately 19% of these mildly impaired subjects (n=16) had a CDR-SB of 2 or higher, and approximately 81% (n=69) had a CDR-SB score of 0.5–1.5.
Diagnosis of Dementia on Follow-up
As part of the annual review of each case, the consensus diagnostic process determined: 1) whether the individual had sufficient impairment for a diagnosis of dementia, and if so, 2) whether the dementia was consistent with research criteria for AD28 or another known diagnostic entity, e.g., Frontotemporal dementia, Vascular dementia.2930 Diagnoses were based on findings from a combination of clinical history, medical records, laboratory evaluation, and neuroimaging studies (e.g., presence of cerebral infarcts). Only subjects with a diagnosis of probable AD on follow-up were included in the outcome group presented here.
MRI Image Acquisition
The MRI scans utilized in this study were acquired on a 1.5 Tesla scanner (General Electric, Milwaukee, USA). T1-weighted 3D spoiled gradient echo (SPGR) scans were acquired using the following sequence: one coronal acquisition, TR=35 msec, TE=5 msec, FOV=220 mm, flip angle=45 °, slice thickness = 1.5 mm, matrix size= 256 * 256 NEX=1.
Regions of Interest
The MRI scans obtained at baseline were processed using the FreeSurfer software package (http://surfer.nmr.mgh.harvard.edu). 3132 First, each scan was normalized for spatial intensity changes, using the N3 algorithm, followed by an intensity normalization procedure.32 Next, the skull was removed using a skull-stripping algorithm.33 The images were then segmented to identify the dorsal, ventral and lateral extent of the gray/white matter boundary, in order to provide a surface representation of the cerebral white matter.3132 The quality of the skull stripping and the accuracy of the gray/white matter tissue boundary for each subject was reviewed by an anatomically-knowledgeable operator (RSD) and edited, as needed, to produce an anatomically accurate surface representation of the cortical white matter (i.e. to assure the exclusion of bone, and other non-neocortical matter from white matter, and to fill in artifactual ‘holes’ in the white matter surface that were inconsistent with known neuroanatomy). Once the white matter representation was complete, an automatic topology correction algorithm was applied that corrects for small topological defects (i.e. voxel misclassifications that result in erroneous ‘bridges’ or ‘connections’ between areas in the white matter). 34 The topologically corrected white matter surface was then used in a deformation algorithm that identified the neocortical (i.e., gray matter) surface of the brain.35 Lastly, the white and gray matter surfaces were visually inspected and further edited, as needed, for anatomical accuracy, by a trained operator (RSD) (i.e. to assure the exclusion of skull from gray matter, and the proper outward deformation of the white matter).
The neocortex of the brain on the MRI scans was then automatically subdivided into 32 gyral-based ROIs (in each hemisphere). To accomplish this, a registration procedure was used that aligns the cortical folding patterns36 and probabilistically assigns every point on the cortical surface to one of the 32 ROIs.37 The ROIs generated were examined for anatomical accuracy and edited by an anatomically-knowledgeable operator (RSD), as needed, in order to ensure that they adhered to previously published boundary definitions.37 For the purposes of this study, we focused on the 14 ROIs that corresponded to neocortical regions from the temporal and parietal cortices, since pathological evidence of AD is primarily evident in the temporal and parietal regions early in the course of disease. The regions selected included: (1) the banks of superior temporal sulcus, (2) entorhinal cortex, (3) fusiform gyrus, (4) inferior parietal lobule, (5) inferior temporal gyrus, (6) isthmus of cingulate cortex (i.e., the caudal portion of the posterior cingulate), (7) posterior cingulate cortex (i.e., the rostral portion of the posterior cingulate), (8) middle temporal gyrus, (9) parahippocampal gyrus, (10) precuneus cortex, (11) superior parietal lobule, (12) superior temporal gyrus, (13) supramarginal gyrus, and (14) temporal pole.
The non-neocortical regions of the brain were subdivided into 19 ROIs (in each hemisphere). As with the neocortical regions, an algorithm automatically assigned each voxel in this portion of the scan to one of 19 neuroanatomical ROIs.38 These ROIs were edited by an anatomically-knowledgeable operator (RJK), as needed, in order to ensure that they adhered to previously published boundary definitions.3940 Since pathological evidence of AD is primarily evident in the temporal and parietal regions early in the course of disease, for the purposes of the current study, we selected 2 of the 19 non-neocortical ROIs corresponding to regions in the temporal lobe: (1) the amygdala, and (2) hippocampus.
In total, 16 neocortical and non-neocortical temporoparietal ROIs were used in this study. Figure 1 depicts the location of one of these ROIs, namely the inferior parietal lobule. For all of the analyses performed here, the volume of the right and the left hemispheres for each individual ROI, were added together. It should be noted that editing of the ROIs was necessary because SPGR scans have lower contrast to noise than do the specific sequences on which the image analysis algorithms utilized here were optimized. Z-scores were computed for each ROI based on the distributions of volumes found in the sample.
Figure 1
Figure 1
Illustration of the location of the inferior parietal and medial temporal regions of the brain on a three-dimensional image on one hemisphere of the brain. Please see Figure 5 from Fischl et al., 2002 and Figure 1 from Desikan et al., 2006 for a detailed (more ...)
Neuropsychological Measures
As part of participation in this study, all subjects were also administered a neuropsychological battery that was independent of the assessment of clinical severity. The composition of the entire battery has been previously described.19 Three test scores from this battery were selected for analysis in the present study because they had previously been shown to be sensitive predictors of time to progression from MCI to AD.41 These three tests included two tests of episodic memory: (1) the total number of words learned across the five learning trials of the California Verbal Learning Test (CVLT) 42, and (2) the total number of words learned across the four learning trials of the Selective Reminding Test (SRT) 43; and an executive function test: (3) the time to complete Part B of the Trail Making Test (Trails B). 44 All neuropsychological measures were standardized to have zero mean and unit variance, averaged over the combined study sample, to facilitate interpretable coefficients in the Cox proportional hazards models and facilitate comparisons of effect sizes across tests. Prior to standardization, the raw time to complete Trails B was log-transformed so that the distribution was more normal.
Statistical Analysis of Data
The time-to-progression data were analyzed using Cox proportional hazards models as implemented in the PHREG procedure in SAS version 8 (SAS Institute, Cary, NC).These models tested whether specific predictors (i.e., z-scores based on MRI measures from temporoparietal regions) are associated with time to a diagnosis of AD). The hazard ratio indicates the differential risk per one unit difference in the predictor. For instance, if the hazard ratio is 1.06 for the volume of the entorhinal cortex, each 1 SD decrease increases risk by 6%, or if the hazard ratio is 0.57 for the entorhinal, each 1 SD decrease in the volume of this region increases risk by 43%.
The primary focus of the analyses was time from study entry to the endpoint of interest, i.e., the diagnosis of AD. A set of univariate and multivariable Cox models were performed. The initial set of analyses included two bivariate (single predictor) models and one multivariable (multiple predictor)model. The two bivariate models were as follows: (1) The first bivariate model for each of the MRI measures was ‘crude’ in that it only included an adjustment for Intracranial Cavity (ICC) size. (2) The second bivariate model for each of the MRI measures adjusted for both ICC and age. The multivariable model, which was designed to be the “best” multivariable Cox model (given the set of variables), was then developed as follows: (1) it began with the inclusion of ICC and age, which were “forced” into the model (i.e., these two variables were entered into the model and were retained even if they were not significant.). The 16 z-scores based MRI measures were then added, but were only retained if they were significant at the 0.05 level. Age was controlled for linearly. Subsequent analyses repeated these models using binary variables based on the z-scores for each ROI, with those with z-scores smaller than 1 standard deviation below the mean placed in one group and those with larger volumes placed in a comparison group. In these analyses, hazard ratios greater than 1 indicated increased risk of progression to AD with ROI volumes below the 1 SD cutoff The models were also repeated with the hippocampus forced into the models.
A separate set of analyses involved recalculation of the models with the inclusion of the 3 neuropsychological measures. These variables were added in order to determine whether the inclusion of neuropsychological data provided redundant or additive information to the MRI data concerning prediction of progression. In addition to these proportional hazards analyses, Spearman rank correlation coefficients were utilized to examine the relationship between the sixteen temporoparietal MRI volumes and the three neuropsychological measures.
The proportional hazards assumption was evaluated descriptively by checking whether the negative log of survival probabilities associated with higher level(s) of each covariate were constant multiples of those of the lower level(s), across the entire range of event time. This approach was extended to the multivariate Cox model by examining such patterns across the levels of each linear predictor. Model fit was also examined descriptively by checking the distribution of the martingale residuals, as well as deviance residuals. 45
MRI Volumetric Temporoparietal Measures and Time to Diagnosis of AD
Bivariate Cox models were first constructed to assess likelihood of time to progression to a diagnosis of AD for each of the 16 temporoparietal MRI measures, using only an adjustment for ICC. Of the 16 variables, 7 were statistically significant at the p < 0.05 level or greater (Table 2). A second set of bivariate Cox models were then completed that included an adjustment for both ICC and age (Table 2). Of the 16 variables, 6 of the same 7 variables were statistically significant at the p <0.05 level or greater. The only variable that was significant in the first univariate model but was not significant in the second univariate model was the hippocampus. There were large effects for the entorhinal cortex [HR=0.54 {0.37–0.78}, p < 0.001], the inferior parietal lobule [HR=0.64 {0.46–0.88}, p <0.005], and the middle temporal gyrus [HR=0.64 {0.47–0.86}, p < 0.004]. As anticipated, the level of statistical significance was decreased by the inclusion of the age adjustment.
Table 2
Table 2
Bivariate analyses of z-scores for MRI measures and risk of progression to a diagnosis of AD
Identifying the Best Temporoparietal Predictors of Time to a Diagnosis of AD Using a Multivariable Model
A multivariable model of predictors of time to progression to diagnosis of AD among subjects with MCI at baseline was then constructed. This model examined which z-scores, in combination, were the ‘best’ predictors of time to diagnosis. In this model, only the entorhinal cortex (HR = 0.57 {0.39–0.83}, p < 0.004) and the inferior parietal lobule (HR = 0.68 (0.47–.98), p < 0.05) were statistically significant. Figure 2 shows survival curves for time to progress from MCI to a diagnosis of AD, as a function of the volume of the inferior parietal lobule.
Figure 2
Figure 2
Survival curves for prediction of time to progression from MCI to a diagnosis of AD [based on the adjusted univariate model (shown at the mean and one standard deviation above and below the mean)], as a function of variation in the MRI volume of the inferior (more ...)
Assessment of Uniformity of Risk in Time to Diagnosis of AD
Several analyses were performed in order to further examine the uniformity of risk for time to a diagnosis of AD based on the two temporoparietal MRI variables that were significant in the multivariable model (i.e., the entorhinal cortex and the inferior parietal lobule). First, the multivariable model was recalculated, with the volume of the hippocampus forced into the model. The presence of the hippocampus did not reduce the hazard ratio of the entorhinal cortex [HR = 0.57 {0.27–0.94}, p < 0.005] and slightly reduced the hazard ratio of the inferior parietal lobule [HR = 0.68 (0.48–99), p < 0.04].
Second, using binary variables based on the z-scores for each ROI, based on the set of 129 MCI subjects, we computed an additional set of Cox regressions. Among these binary variables, we found that 17 subjects had z-scores for volume of the entorhinal cortex that were 1 standard deviation below the mean and 24 subjects fell into this category based on the z-score for the volume of the inferior parietal lobule. The results of these additional Cox model showed a substantial increase in the hazard ratio for both of the variables in the model: the entorhinal cortex [HR = 4.75 {2.27–9.94}, p < 0.0001] and the inferior parietal lobule [HR = 2.33 (1.16–4.69, p < 0.01]. This model was then recalculated with the binary variable based on the z-score for the hippocampus forced into the analysis. The results showed minimal change in the hazard ratio for both of the variables in the model: the entorhinal cortex [HR = 4.27 {1.91–9.54}, p < 0.0004) and the inferior parietal lobule [HR = 2.65 (1.29–5.47, p < 0.008].
Contribution of Neuropsychological Variables to the Bivariate and Multivariable Models and Correlations with MRI volumes
Further analyses were performed to assess whether the addition of specific neuropsychological variables, previously shown to be significant predictors of time to progress from MCI to AD41 would provide additional predictive information, above and beyond the MRI measures, when added to the bivariate and multivariate models. First, the bivariate models were recalculated, with the inclusion of four additional variables: years of education and the three neuropsychological test scores (i.e., the CVLT, the SRT, and Trails B). The same 6 z-scores for MRI variables identified above were still statistically significant at the p <0.05 level or greater, after the addition of these 4 variables. The hazard ratios for each of the 6 MRI variables from the bivariate analyses were similar in magnitude to those observed in the models in which neuropsychological variables had not been included. Significant effects were observed for the entorhinal cortex [HR = 0.56 {0.40–0.81}, p < 0.01], amygdala [HR = 0.60 {0.41–0.88}, p < 0.01], inferior parietal lobule [HR = 0.61 {0.44–0.87}, p < 0.01], supramaginal gyrus [HR = 0.61 {0.42–0.87}, p < 0.01], middle temporal gyrus [HR = 0.63 {0.45–0.89}, p < 0.01], and the fusiform gyrus [HR = 0.68 {0.49–0.94}, p < 0.05].
Second, the same 4 variables (i.e., years of education and the 3 neuropsychological test scores) were added to the multivariable model in order to determine whether any of them would be selected instead of the MRI variables as the best predictors of progression from MCI to AD. In this multivariable model, only the entorhinal cortex [HR = 0.63 {0.44–0.91}, p < 0.01], and Trails B [HR = 2.75 {1.17–6.65}, p < 0.05] were statistically significant. The inferior parietal lobule [HR = 0.70 {0.48–1.02}, p = 0.06] demonstrated a trend towards statistical significance.
For the correlations between the MRI volumes and CVLT, the parahippocampal gyrus (r = 0.25, p = 0.005) and temporal pole (r = 0.20, p = 0.05) demonstrated a significant relationship, with the hippocampus (r = 0.15, p = 0.09) demonstrating a trend towards statistical significance. With SRT, the parahippocampal gyrus (r = 0.21, p = 0.01), temporal pole (r = 0.23, p = 0.01), and hippocampus (r = 0.31, p = 0.001) demonstrated a significant relationship. With Trails B, none of the ROIs demonstrated a significant relationship.
These findings reaffirm the importance of an MRI measure of the entorhinal cortex as a predictor of progression from MCI to a diagnosis of AD. In every analysis that was performed, the volume of the entorhinal cortex was a better predictor of progression from MCI to AD than any of the other 15 temporoparietal MRI measures. These data are in agreement with a number of previous reports that have concluded that the volume of entorhinal cortex is better at predicting likelihood of progression from MCI to AD than the hippocampus. 4648
These results also emphasize the value of a volumetric measure of the inferior parietal lobule. This measure, when used in combination with the entorhinal cortex, was the best predictor of time to progress from MCI to AD. Moreover, it remained statistically significant even when the volume of the hippocampus was forced into the model. Prior studies using fluid-registration, cortical thickness and voxel-based morphometry have implicated areas within the lateral parietal cortex to be involved in the earliest stages of AD and as a predictor of progression1218 but this is the first volumetric study, to our knowledge, to demonstrate the relative importance specifically of the inferior parietal lobule in predicting progression in comparison to the many other brain regions within the temporal and parietal lobes.
This finding is consistent with pathological studies of AD showing that specific laminae in the inferior parietal lobule are preferentially affected in the early stages of the disease. 4950 Moreover, projections from the inferior parietal lobule target several subfields within the medial temporal lobe 5153, suggesting that atrophy in the inferior parietal lobule likely reflects the spread of AD pathology from the temporal lobe to an interconnected region in the parietal lobe.
The analyses presented here also demonstrate that MRI volumetric measures may be useful in identifying the subset of MCI subjects who are at a particularly high risk of progression to AD. Those MCI subjects whose entorhinal and inferior parietal lobule volumes were 1 SD below the mean for the group as a whole at baseline had markedly increased risk of progression to AD than those whose volumetric measures did not fall 1 or more SD below the mean. It is increasingly recognized that MCI subjects from a community volunteer-based cohort generally include a broad range of severity. The present findings suggest that it should be possible to use MRI measures, independent of clinical and neuropsychological measures, to identify the subset of MCI subjects at greatest risk for progression.
These findings also suggest that MRI volumetric data provide information concerning time to progress from MCI to AD that is independent of neuropsychological measures that have previously been shown to be significant predictors of progression. A number of temporoparietal regions, including the entorhinal cortex and inferior parietal lobule, continued to significantly predict time to progression, even after the addition of the neuropsychological variables to the bivariate models. Moreover, the entorhinal cortex was retained as one of the best predictors of conversion in the multivariable model that also included a neuropsychological variable, suggesting that MRI and neuropsychological data may provide complimentary information in relation to prediction of progression from MCI to AD. This finding differs somewhat from a recent report suggesting that once neuropsychological measures are considered, the added value of MRI measures is small.47 The difference between the findings reported here and the previous study may be related to the fact that the earlier study examined subjects that were more mildly impaired than those examined in the present study and additionally did not include a test of executive function, such as that included here.
Correlations between tests of episodic memory function (CVLT and SRT) and volumes of the parahippocampal gyrus, temporal pole, and hippocampus are consistent with the fact that these temporal lobe regions are critical for normal memory function (for a discussion of this topic see reference 41). Of interest, Trails B, a test of executive function, did not demonstrate any significant correlations with any of the temporoparietal regions but was one of the best predictors in the multivariable model when combined with the MRI volumes. This suggests that regions beyond the temporal and parietal lobes are potentially responsible for executive function and may additionally be significant predictors of progression.
A concern in this study pertains to the difference in APOE-ε4 between the two groups. Since more MCI-Converters were APOE-ε4 positive than the MCI-Nonconverters and the ε4 allele of this gene is overrepresented in AD patients compared with the general population54, one possibility is that the presence of APOE-ε4 alone can best account for the time to progress from MCI to AD. Prior work from our research group has demonstrated the influence of the ε4 allele on the time to progress from MCI to AD is largely accounted for by neuropsychological measures and assessments of clinical severity41 thus disputing the notion that the presence of this allele can solely account for the time to progress from MCI to AD.
The present study has several strengths. The subjects were followed prospectively and then categorized after their symptoms had evolved by clinicians with no access to the MRI data. The image analysis methods presented here permit a comparison of the relative strengths of prediction for each anatomic region within the temporal and parietal lobe and can be combined with survival analyses to determine which individual or combination of ROIs best predict time to progress from MCI to AD.
One limitation of this study is that only brain regions within the temporal and parietal cortices were examined. It is therefore possible that regions elsewhere in the brain may also be significantly related to time to progression from MCI to AD. In addition, a longer follow-up interval may have resulted in a larger number of subjects progressing to AD; as a result other ROIs, in addition to the ones presented here, may have been identified as significant predictors of time to progress from MCI to AD.
CONCLUSION
Taken together these findings suggest the importance of examining brain regions not emphasized in previous MRI studies, such as the inferior parietal lobule, a region selected as one of the best predictors of time to progress from MCI to AD. These MRI measures may also be useful in identifying individuals at particularly high risk for progression, and could readily be employed for selecting subjects for clinical trials in MCI, or for guiding for treatment decisions, when improved medications become available.
Acknowledgments
The authors would like to thank Dr. Mary Hyde for invaluable assistance with data analysis and Dr. Svetlana Egorova, Amanda Dow, and Marisa Tricarico for assistance with data management.
This work was supported grants from the National Institute on Aging (P01-AG04953), the National Center for Research Resources (P41-RR14075, R01-RR16594, U24-RR021382), the National Institute for Biomedical Imaging and Bioengineering (R01-EB001550), the National Institute for Neurological Diseases and Stroke (R01 NS052585), the BIRN Mophometric Project, and the Mental Illness and Neuroscience Discovery (MIND) Institute.
1. Petersen R, Smith G, Waring S, Ivnik R, Tangalos E, Kokmen E. Mild cognitive impairment: clinical characterization and outcome. Arch Neurol. 1999;56:303–308. [PubMed]
2. Petersen RC, Doody R, Kurz A, Mohs RC, Morris JC, Rabins PV, Ritchie K, Rossor M, Thal L, Winblad B. Arch Neurol. 2001;58:1985–1992. [PubMed]
3. Atiya M, Hyman B, Albert M, Killiany R. Structural magnetic resonance imaging in established and prodromal Alzheimer’s disease: A review. Alzheimer Dis Assoc Disord. 2003;17:177–195. [PubMed]
4. Chetelat G, Baron JC. Early diagnosis of Alzheimer’s disease: contribution of structural neuroimaging. Neuroimage. 2003;18:525–541. [PubMed]
5. Kantarci K, Jack C. Neuroimaging in Alzheimer’s disease: an evidenced-based review. Neuroimaging Clin N Am. 2003;13:197–209. [PubMed]
6. Anderson VC, Litvack ZN, Kaye JA. Magnetic resonance approaches to brain aging and Alzheimer disease-associated neuropathology. Top Magn Reson Imaging. 2005;16:439–452. [PubMed]
7. Hyman BT, Van Hoesen GW, Damasio AR, Barnes CL. Alzheimer’s disease: cell-specific pathology isolates the hippocampal formation. Science. 1984;225:1168–1170. [PubMed]
8. Gomez-Isla T, Price JL, McKeel DW, Jr, Morris JC, Growdon JH, Hyman BT. Profound loss of layer II entorhinal cortex neurons occurs in very mild Alzheimer’s disease. J Neurosci. 1996;16:4491–4500. [PubMed]
9. Kordower J, Chu Y, Stebbins G, DeKosky S, Cochran E, Bennett D, Mufson E. Loss and atrophy of layer II entorhinal cortex neurons in elderly people with mild cognitive impairment. Ann Neurol. 2001;49:202–213. [PubMed]
10. Convit A, de Asis J, de Leon MJ, Tarshish CY, De Santi S, Rusinek H. Atrophy of the medial occipitotemporal, inferior, and middle temporal gyri in non-demented elderly predict decline to Alzheimer’s disease. Neurobiol Aging. 2000;21:19–26. [PubMed]
11. Killiany R, Gomez-Isla T, Moss M, Kikinis R, Sandor T, Jolesz F, Tanzi R, Jones K, Hyman B, Albert M. Use of structural magnetic resonance imaging to predict who will get Alzheimer’s disease. Ann Neurol. 2000;47:430–439. [PubMed]
12. Bell-McGinty S, Lopez OL, Meltzer CC, Scanlon JM, Whyte EM, Dekosky ST, Becker JT. Differential cortical atrophy in subgroups of mild cognitive impairment. Arch Neurol. 2005;62:1393–1397. [PubMed]
13. Bozzali M, Filippi M, Magnani G, Cercignami M, Franceschi M, Schiatti E, Castiglioni S, Mossini R, Falautano M, Scotti G, Comi G, Falini A. The contribution of voxel-based morphometry in staging patients with mild cognitive impairment. Neurology. 2006;67:453–460. [PubMed]
14. Karas G, Sluimer J, Goekoop R, van der Flier W, Rombouts S, Vrenken H, Scheltens P, Fox N, Barkhof F. Amnestic Mild Cognitive Impairment: Structural MR Imaging Findings Predictive of Conversion to Alzheimer Disease. AJNR Am J Neuroradiol. 2008 Feb 22 [PubMed]
15. Scahill RI, Schott JM, Stevens JM, Rossor MN, Fox NC. Mapping the evolution of regional atrophy in Alzheimer’s disease: unbiased analysis of fluid-registered serial MRI. Proc Natl Acad Sci U S A. 2002;99:4703–4707. [PubMed]
16. Buckner RL, Snyder AZ, Shannon BJ, LaRossa G, Sachs R, Fotenos AF, Sheline YI, Klunk WE, Mathis CA, Morris JC, Mintun MA. Molecular, structural, and functional characterization of Alzheimer’s disease: evidence for a relationship between default activity, amyloid, and memory. J Neurosci. 2005;25:7709–7717. [PubMed]
17. Singh V, Chertkow H, Lerch JP, Evans AC, Dorr AE, Kabani NJ. Spatial patterns of cortical thinning in mild cognitive impairment and Alzheimer’s disease. Brain. 2006;129:2885–2893. [PubMed]
18. Dickerson BC, Bakkour A, Salat DH, Feczko E, Pacheco J, Greve DN, Grodstein F, Wright CI, Blacker D, Rosas HD, Sperling RA, Atri A, Growdon JH, Hyman BT, Morris JC, Fischl B, Buckner RL. The Cortical Signature of Alzheimer’s Disease: Regionally Specific Cortical Thinning Relates to Symptom Severity in Very Mild to Mild AD Dementia and is Detectable in Asymptomatic Amyloid-Positive Individuals. Cereb Cortex. 2008 Jul 16 [PMC free article] [PubMed]
19. Albert M, Moss M, Tanzi R, Jones K. Preclinical prediction of AD using neuropsychological tests. J Internatl Neuropsychol Soc. 2001;7:631–639. [PubMed]
20. Hughes CP, Berg L, Danziger WL, Coben LA, Martin RL. A new clinical scale for the staging of dementia. Brit J Psychiatry. 1982;140:566–572. [PubMed]
21. Morris JC. The Clinical Dementia Rating (CDR): current version and scoring rules. Neurology. 1993;43:2412–2414. [PubMed]
22. Blessed G, Tomlinson B, Roth M. The association between quantitative measures of dementia and of senile change in cerebral grey matter of elderly subjects. Br J Psychiatry. 1968;114:797–811. [PubMed]
23. Daly E, Zaitchik D, Copeland M, Schmahmann J, Gunther J, Albert M. Predicting ‘conversion’ to AD using standardized clinical information. Arch Neurol. 2000;57:675–680. [PubMed]
24. Folstein M, Folstein S, McHugh P. “Mini-Mental State”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res. 1975;12:189–198. [PubMed]
25. Davis H, Rockwood K. Conceptualization of mild cognitive impairment: a review. Internatl J Geriatr Psychiatr. 2004;19:313–19. [PubMed]
26. Petersen R, Thomas R, Grundman M, Bennett D, Doody R, et al. Vitamin E and donepezil for the treatment of mild cognitive impairment. New England Journal of Medicine. 2005;352:2379–88. [PubMed]
27. Petersen R. Mild cognitive impairment. J Intern Med. 2004;256:183–194. [PubMed]
28. McKhann G, Drachman D, Folstein MF, Katzman R, Price D, Stadlan E. Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work group under the auspices of Department of Health and Human Services Task Force. Neurology. 1984;34:939–944. [PubMed]
29. McKhann G, Albert M, Grossman M, Miller B, Dickson D, Trojanowski J. Clinical and pathological diagnosis of frontotemporal dementia: report of the Work Group on Frontotemporal Dementia and Pick’s Disease. Arch Neurol. 2001;58:1803–1809. [PubMed]
30. Roman GC, Tatemichi TK, Erkinjuntti T, et al. Vascular dementia: diagnostic criteria for research studies. Report of the NINDS-AIREN International Workshop. Neurology. 1993;43:250–260. [PubMed]
31. Dale AM, Fischl B, Sereno MI. Cortical surface-based analysis. I. Segmentation and surface reconstruction. Neuroimage. 1999;9:179–194. [PubMed]
32. Fischl B, Sereno MI, Dale AM. Cortical surface-based analysis. II: Inflation, flattening, and a surface-based coordinate system. Neuroimage. 1999a;9:195–207. [PubMed]
33. Segonne F, Dale AM, Busa E, Glessner M, Salat D, Hahn HK, Fischl B. A hybrid approach to the skull stripping problem in MRI. Neuroimage. 2004;22:1060–1075. [PubMed]
34. Segonne F, Grimson E, Fischl B. Genetic algorithm for the topology correction of cortical surfaces. Proceed Internatl Conf Inf Process Med Imaging. 2005;3964:393–405. [PubMed]
35. Fischl B, Dale AM. Measuring the thickness of the human cerebral cortex from magnetic resonance images. Proc Natl Acad Sci USA. 2000;97:11050–11055. [PubMed]
36. Fischl B, Sereno MI, Tootell RB, Dale AM. High-resolution intersubject averaging and a coordinate system for the cortical surface. Hum Brain Map. 1999b;8:272–284. [PubMed]
37. Desikan RS, Ségonne F, Fischl B, Quinn BT, Dickerson BC, Blacker D, Buckner RL, Dale AM, Maguire RP, Hyman BT, Albert MS, Killiany RJ. An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. Neuroimage. 2006;31:968–980. [PubMed]
38. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Haselgrove C, van der Kouwe A, Killiany R, Kennedy D, Klaveness S, Montillo A, Makris N, Rosen B, Dale AM. Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain. Neuron. 2002;33:341–355. [PubMed]
39. Killiany R, Moss M, Albert M, Sandor T, Tieman J, Jolesz F. Temporal lobe regions on magnetic resonance imaging identify patients with early Alzheimer’s disease. Arch Neurol. 1993;50:949–954. [PubMed]
40. Killiany RJ, Hyman BT, Gomez-Isla T, Moss MB, Kikinis R, Jolesz F, Tanzi R, Jones K, Albert MS. MRI measures of entorhinal cortex vs hippocampus in preclinical AD. Neurology. 2002;58:1188–1196. [PubMed]
41. Blacker D, Lee H, Muzikansky A, Martin EC, Tanzi R, McArdle JJ, Moss M, Albert M. Neuropsychological measures in normal individuals that predict subsequent cognitive decline. Arch Neurol. 2007;64:862–871. [PubMed]
42. Delis D, Kramer J, Kaplan E, Ober B. The California Verbal Learning Test. New York: Psychological Corp; 1987.
43. Grober E, Buschke H. Genuine memory deficits in dementia. Develop Neuropsychol. 1987;3:13–36.
44. Reitan RM. Validity of the Trail Making Test as an indicator of organic brain damage. Percept Motor Skills. 1958;8:271–276.
45. Yin DY, Wei LJ, Ying Z. Checking the Cox model with cumulative sums of martingale-based residuals. Biometrika. 1993;80:557–572.
46. DeToledo-Morrell L, Stoub TR, Bulgakova M, Wilson RS, Bennett DA, Leurgans S, Wuu J, Turner DA. MRI-derived entorhinal volume is a good predictor of conversion from MCI to AD. Neurobiol Aging. 2004;9:1197–203. [PubMed]
47. Devanand DP, Pradhaban G, Liu X, Khandji A, De Santi S, Segal S, Rusinek H, Pelton GH, Honig LS, Mayeux R, Stern Y, Tabert MH, de Leon MJ. Hippocampal and entorhinal atrophy in mild cognitive impairment: prediction of Alzheimer disease. Neurol. 2007;68:828–836. [PubMed]
48. Tapiola T, Pennanen C, Tapiola M, Tervo S, Kivipelto M, Hänninen T, Pihlajamäki M, Laakso MP, Hallikainen M, Hämäläinen A, Vanhanen M, Helkala EL, Vanninen R, Nissinen A, Rossi R, Frisoni GB, Soininen H. MRI of hippocampus and entorhinal cortex in mild cognitive impairment: a follow-up study. Neurobiology of Aging. 2008;29:31–38. [PubMed]
49. Pearson RC, Esiri MM, Hiorns RW, Wilcock GK, Powell TP. Anatomical correlates of the distribution of the pathological changes in the neocortex in Alzheimer disease. Proc Natl Acad Sci. 1985;82:4531–4534. [PubMed]
50. Hof PR, Bouras C, Morrison JH. Cortical neuropathology in aging and dementing disorders: Neuronal typology, connectivity, and selective vulnerability. In: Peters A, Morrison JH, editors. Cerebral Cortex. Vol. 14. New York: Kluwer Academic Press; 1999. pp. 175–311.
51. Van Hoesen GW, Pandya DN. Some connections of the entorhinal (area 28) and perirhinal (area 35) cortices of the rhesus monkey. I. Temporal lobe afferents. Brain Res. 1975;95:1–24. [PubMed]
52. Seltzer B, Pandya DN. Further observations on parieto-temporal connections in the rhesus monkey. Exper Brain Res. 1984;55:301–312. [PubMed]
53. Ding SL, Van Hoesen G, Rockland KS. Inferior parietal lobule projections to the presubiculum and neighboring ventromedial temporal cortical areas. J Compar Neurol. 2000;425:510–530. [PubMed]
54. Saunders AM, Strittmatter WJ, Schmechel D, et al. Association of apolipoprotein E allele epsilon 4 with late-onset familial and sporadic Alzheimer’s disease. Neurology. 1993;43:1467–1472. [PubMed]