Correlation of Regional Proton Magnetic Resonance Spectroscopic Metabolic Changes With Cognitive Deficits in Mild Alzheimer Disease FREE

ALZHEIMER DISEASE (AD) is a neurodegenerative disorder associated with a gradual deterioration of cognitive functions, personality, and behavior. Impairment of recent memory is commonly the first symptom of the disease attributable to neurochemical and pathologic changes in the medial temporal lobe (MTL). As AD progresses, language, attention, calculation, visuospatial, visuoconstructional, and executive functions become impaired because of a dysfunction of the associative cortices in both cerebral hemispheres.¹ The cause and pathogenesis of AD remain complex but it is always associated with gray matter atrophy, disruption of neuronal function, and formation of neurofibrillary tangles and neuritic plaques in the medial temporal limbic regions and isocortex.^{2 - 3} Cerebral imaging studies have revealed hippocampal atrophy consistent with neuronal loss^{4 - 5} and predictive of memory performance,⁵ and reduced metabolism in the posterior association areas of the neocortex⁶ as well as in the posterior cingulate gyrus and pericingular cortex in the early stages of AD.⁷ However, heterogeneous metabolic patterns have been reported among patients, reflecting either frontal or parietotemporal abnormalities⁸ and right- or left-sided metabolic asymmetries correlated with visuoconstructional and language discrepancies.⁹ Neuropsychological studies of patients with AD have also revealed selective and heterogeneous patterns of cognitive impairments. For example, some patients present severe language dysfunction along with milder visuoconstructional dysfunction, whereas other patients present with the opposite pattern.⁶ Such differences in the pattern of cognitive deficits suggest a variable distribution of neocortical abnormalities that can be examined with the proton (¹H) magnetic resonance spectroscopy (MRS) technique.

Most in vivo ¹H MRS investigations of patients with AD have demonstrated a reduction of the amount of N-acetylaspartate (NAA) (neuronal marker)^{10 - 22} and an increase of myo-inositol (mI) (glial marker)^{10 - 12 ,14 - 17} in different brain regions containing both white and gray matter tissues. Inconsistent results have been reported for choline-containing compounds (Cho),^{10 - 11 ,13 - 14 ,16 - 22} whereas the creatine-phosphocreatine (Cr) signal remains unchanged.^{13 - 14 ,17 ,19 ,22} Magnetic resonance spectroscopy has been used to assess the association between regional metabolic abnormalities and cognitive dysfunctions in neurologic and psychiatric disorders and in healthy individuals.^{23 - 26}

In patients with AD, conflicting results have been reported concerning the correlation between the metabolite levels and the degree of cognitive alteration.^{11 - 12 ,14 ,16 - 17 ,20 ,22} The relevance of the correlations is, however, questionable since either conclusions were based on nonspecific brief mental evaluations, such as the Mini-Mental State Examination²⁷ or global scores from the AD Assessment Scale,²⁸ thereby not reflecting specific cognitive deficits caused by regional cerebral deterioration, or the cerebral region examined is not functionally involved in early clinical symptoms of AD. To our knowledge, no investigation examined specifically the relationship between local metabolism measured by MRS and associated patterns of cognitive impairment.

In the present study, MRS data were acquired in 6 cortical regions known to be affected in patients with AD to evaluate the spatial specificity of the metabolic impairments. Correlations were established between neurometabolic variations and specific neuropsychological performances (ie, verbal memory, executive, language, and visuoconstructional abilities) associated with these regions. To determine the specificity of correlations, cognitive functions were tested in the contralateral functional region as control or in the posterior associative regions in the case of executive functions. The results indicate that pertinent lateralized correlations can be established in several regions.

Fourteen right-handed patients diagnosed as having probable AD of mild severity according to criteria for dementia of the Alzheimer type of the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition²⁹ and criteria for probable AD of the National Institute of Neurologic and Communicative Disorders and Stroke–Alzheimer's Disease and Related Disorders Association (NINCDS-ADRDA)³⁰ were compared with 14 right-handed control subjects who had no history of cognitive decline or previous neurologic or psychiatric disorder. They were submitted to physical and neurologic examinations, including laboratory tests and brain imaging (ie, magnetic resonance imaging or computed tomography). Clinical assessment of dementia included neurocognitive, behavioral, and psychiatric interviews conducted by the behavioral neurologist (R.W.B.), administration of the Mini-Mental State Examination,²⁷ and a neuropsychological evaluation. Controls were evaluated with the same neuropsychological tests to ensure healthy cognitive functions. Patients presented with no focal brain lesions detectable by computed tomography or magnetic resonance imaging, had a global staging score of 1 or less on the Clinical Dementia Rating³¹ scale and of less than 4 on the Hachinski Ischemia Scale.^{32 - 33} They were not treated with acetylcholinesterase inhibitors or other investigational drugs to enhance brain cognitive function and were not currently or previously suffering from significant systemic (including diabetes mellitus) or psychiatric conditions or traumatic brain injuries that could compromise brain functions. All controls and patients were individually paired for age and sex.

Controls and patients with AD underwent a brain ¹H MRS examination performed using a 1.5-T whole-body scanner (GE Signa; General Electric Medical Systems, Waukesha, Wis) operating at 63.85 MHz. Proton MRS data were obtained from 7- to 8-cm³ voxels localized in the left and right MTLs (ie, amygdala, the anterior half of the hippocampus, and part of the underlying subiculum, 2 × 2 × 2 cm³), left and right frontal cortices (FCs) (ie, mixed gray and white matter, 3.1 × 2.3 × 1.0 cm³), and left and right parietotemporal cortices (PTCs) (ie, mixed gray and white matter, 3.7 × 1.9 × 1.0 cm³). The voxel shapes of the FC and PTC regions were chosen to maximize their gray matter content as shown in Figure 1. The point-resolved spectroscopy pulse sequence was used with the following acquisition parameters: repetition time, 1200 milliseconds; echo time, 50 milliseconds; number of acquisitions, 64; spectral width, 2000 Hz; number of points, 1024; and total acquisition time per voxel, 1.7 minutes. The MRS data were analyzed using the Magnetic Resonance User Interface (Barcelona, Spain) software and signal intensities were calculated directly from time-domain data, using the singular value decomposition 1-dimension model function fitting. The intensity of the water (H₂O) signal was determined from the area of the H₂O peak in the H₂O-unsuppressed spectra and metabolite/unsuppressed water ratios were calculated. Owing to the significant overlap with neighboring signals, the mI signal intensity was estimated from the peak height measured manually. The mI/H₂O ratios were then calculated by dividing the mI peak height by the NAA peak height and multiplying by the NAA/H₂O ratio. Mean and SD values were calculated for all metabolite ratios.

Cognitive domains of verbal memory, language, visuoconstructional abilities, and executive functions were examined using a battery of standard neuropsychological tests as listed in Table 1.^{34 - 35} The French versions of the original tests were used and the raw test scores were used for statistical analysis. For all tests, except Trail Making Test Part A, higher scores indicate better performance.

Metabolite ratios relative to unsuppressed H₂O (NAA/H₂O, Cho/H₂O, Cr/H₂O, and mI/H₂O) were compared in all regions between patients with AD and controls by a 2-way repeated measures analysis of variance followed by a Tukey post hoc test for multiple comparisons, when allowed. Differences in metabolite ratios between the 2 groups were considered statistically significant when P<.05. The Pearson correlation coefficients followed by a Bonferroni correction for multiple correlations were used for assessment of the association between scores in the 4 cognitive domains and metabolite ratios. Correlations were considered significant at P<.05 (corrected for 2 neuropsychological tests per functional brain region) with r>0.50. Statistical analyses were performed using SPSS Version 9.0.1 (SPSS Inc, Chicago, Ill) and SIGMASTAT Version 2.03 (Rockware Inc, Golden, Colo).

The demographic and clinical data of patients with AD and controls paired for age and sex are summarized in Table 2. Significant differences in Mini-Mental State Examination scores are measured between patients with AD and controls.

Table 3 lists the mean metabolite ratios of NAA, Cho, mI, and Cr relative to unsuppressed H₂O for all 6 brain regions in patients with AD and controls. The NAA/H₂O ratio was significantly decreased for patients with AD in the left MTL only. A decrease was also measured in the left FC but the change did not reach statistical significance. In the other regions, the NAA/H₂O ratios were unchanged. The Cho/H₂O ratios of patients with AD were significantly reduced relative to controls in both the left and right MTLs (Table 3). No statistically significant Cho/H₂O changes were measured in the other regions. In the case of mI/H₂O, increases were measured in all 6 regions, but the changes were statistically significant in the right PTC only and showed a statistical tendency in the left MTL and the left PTC. Finally, both increases and decreases were measured for the Cr/H₂O ratio in patients with AD relative to the controls. Statistically significant decreases were measured in the left MTL and in the left PTC. Figure 2 presents spectra illustrating the major effects of AD on brain metabolites. No correlation could be established between metabolite ratios and age for patients with AD and controls.

Statistically significant Pearson correlations between individual cognitive test scores and metabolite ratios in specific brain regions in patients with AD are listed in Table 4 and shown in Figure 3. Correlations were obtained between the NAA/H₂O ratios measured in the left MTL and the learning scores and the word recall memory scores measured by the California Verbal Learning Test (Figure 3A). There was no significant correlation between these verbal memory scores and other metabolite ratios and no correlations were found in the right MTL. In Figure 3B, the negative correlation between language performance measured by the Boston Naming Test (confrontation naming) and the mI/H₂O ratios in the left PTC is shown. No other correlation was found between Boston Naming Test or verbal fluency scores and any other metabolite ratio in the left or right PTC. Figure 3C shows the negative correlation between scores obtained for the copy of Rey Complex Figure 1 Test and mI/H₂O ratios in the right PTC. No correlation was found in the left PTC and no other metabolite ratio was correlated with visuoconstructional abilities. Executive functions were evaluated with verbal fluency test and Trail Making Test Part A. Correlations were examined between scores from these 2 tests and all metabolite ratios in the left FC, right FC, bilateral FCs, and posterior associative cortices as control. A correlation was found between Trail Making Test Part A and mI/H₂O ratios in the right FC (r = 0.55, P = .04, n = 14), the longer the time spent to complete the test, the higher the mI/H₂O ratio, but this correlation was not statistically significant when the Bonferroni correction was applied. For all tests, no significant correlation between neuropsychological performances and age or educational level could be found for patients with AD.

A strong correlation was found between NAA/H₂O ratios and both the learning and word recall parts of the California Verbal Learning Test in the left MTL (Figure 3A, Table 4) but not in the right MTL. These results are consistent with the fact that the left MTL is the site of the verbal memory function. Although nonverbal visual memory processing has not been assessed in this study, our results agree with the known laterality of the mnesic function. In patients with epilepsy, a focus in the left hippocampus was demonstrated to lead to a verbal memory deficit whereas a focus in the right hippocampus was associated with a deficit of the nonverbal visual memory.³⁶ More interestingly, verbal memory deficits were observed in patients with temporal lobe epilepsy following a right-sided excision but only in those patients who had ¹H MRS abnormalities in the contralateral (ie, left) side.³⁷ Together, these studies and our results suggest that MRS provides a sensitive measure of neuronal damage in the MTL that might be undetectable by standard imaging techniques so early in the disease course. This helps to explain anomalous memory performances that are most characteristic of the disease.

An increased mI/H₂O correlates with lower Boston Naming Test scores in the left PTC, reflecting impaired language performance (Figure 3B). This correlation is consistent with the known lateralized language function in the left PTC. Although that naming function is not exclusively localized in the left posterior associative cortex and anomia may be observed following frontal dysfunction of the left dominant hemisphere,³⁸ the absence of correlation with other brain regions lead us to conclude that the language association shown is highly specific to left PTC. Moreover, the stronger relationship between confrontation naming and mI/H₂O ratios may be partly explained by the fact that all our patients were right-handed, thus increasing the probability of left hemisphere language dominance. Visuoconstructional dysfunction measured with the Rey Complex Figure 1 Test correlated with mI/H₂O increases (Figure 3C) in the right PTC, in agreement with the known neurophysiological function of the right PTC.

In the case of executive functions, no relevant correlation was found between neuropsychological performances and metabolite ratios. Even if the Trail Making Test Part A and the verbal fluency test are considered to assess executive capabilities, the entire executive system cannot only be evaluated by these 2 tests. Indeed, in contrast to limbic cortex and posterior association areas, the concept of executive functions in the frontal lobe refers to several cognitive abilities such as planning, judgment, problem solving, abstraction, etc, that cannot be assessed by a single test. For that reason, the use of an executive composite z score constructed from specific tests such as sorting and category tests, the Porteus Maze Test, and the Tower of London Puzzle would be more accurate to test the executive functions against metabolite measures.

Recent data suggest that initial changes in the pathologic progression of AD involve an increase in the mI level and that a decrease in the NAA level and an increase in the Cho level occur later in the disease course.¹⁰ A similar hypothesis was made for Down syndrome,³⁹ a progressive neurochemical disorder in which an mI level increase was found to precede a NAA level loss, suggesting similar pathologic pathways for both diseases. Our data support these previously reported studies. The NAA and Cho level reductions along with an mI level increase, although not always statistically significant, were measured in MTL regions known to be affected early in AD (Table 3). These results seem to suggest that these regions are affected to a point where the initial mI level increase is no longer a significant marker of AD but where neuropathologic processes caused by AD are well established. Interestingly, a significant mI level increase was observed in the right PTC region and a strong tendency toward statistical significance of an mI augmentation was measured in the left PTC. These observations are consistent with Braak and Braak's model of AD pathophysiology being initiated in the MTL and then spreading to upper levels of cortical regions.^{2 ,40} In accord with this model, MRS studies have suggested that biochemical changes in AD are sequential, starting with an mI level increase and followed by decreases of both NAA and Cho levels.^{39 ,41} Further investigations will be necessary to determine if the NAA and Cho level modifications are observable in the moderate AD stage following the mI level increase in the mild stage. Data suggest that the Cr may also be reduced in the MTL and PTC regions (Table 3).

The regional metabolite pattern observed in patients with AD, compatible with Braak and Braak's model of pathologic evolution, ^{2 ,40} is also consistent with the progression of cognitive impairment and reflects a gradual process of brain degeneration. Deficits of episodic memory, that is, the ability to acquire and retain new information associated with the medial structures of the temporal lobe (which includes the hippocampus and the adjacent neocortex),⁴² usually mark the onset of clinical symptoms of AD. The correlation observed between decreased NAA/H₂O ratios measured in the left MTL and verbal memory deficits (learning and word recall memory scores, Figure 3A) in patients with AD supports the role of the MTL in recent memory function. Moreover, since NAA is a neuronal marker, our results suggest a correlation between memory dysfunction and the degree of neuronal loss or damage in the left MTL of patients with AD. This is consistent with previously reported correlations between memory impairment and hippocampal pathology determined by neurofibrillary tangle counts⁴³ as well as hippocampal atrophy estimated by volumetric magnetic resonance imaging in patients with AD.⁵ Higher cortical functions such as language, visuoconstructional and executive functions that decline later in the disease course, are associated with destruction of neocortical association areas. Although neuronal integrity of the associative cortices seems preserved, as suggested by the absence of significant changes in NAA/H₂O levels in both the PTC and the right FC, the mI/H₂O increases in the left and right PTCs correlated inversely with language and visuoconstructional performances, respectively, in accord with an early AD phase. In Braak and Braak's staging model of AD,^{2 ,40} the neocortical stages (V and VI) that correspond to the conventional neuropathologic criteria for clinical AD diagnosis, are marked by severe destruction of neocortical association areas while limbic stages (III and IV) show a much less extensive destruction of the neocortex with no macroscopically detectable atrophy. Since cognitive functions are impaired, as demonstrated by correlations with increased mI/H₂O ratios (Figure 3B, C), and since the disease is still in its early stages, as evidenced by the relatively high Mini-Mental State Examination scores (Table 2) and Clinical Dementia Rating global staging scores, it seems as if the disease of our patients corresponds to a clinical incipient AD presumably with mild changes in the associative cortices. Indeed, it is well known that patients in the early stages of AD present heterogeneous patterns of cognitive impairment with considerable between-patients score discrepancies for language, visuoconstructional, and executive dysfunction assessment.⁶ This cognitive dysfunction variability among patients tends to fade with the evolution of the disease. Our metabolic-functional correlations reinforce the hypothesis that the mI level increase characterizes the initial stages of the disease and precedes the NAA level loss.¹⁰

The results of this study demonstrate that clear correlations can be observed when MRS data are compared with scores from specific cognitive tests associated with AD-relevant regions. The use of tests measuring specific cognitive functions instead of a screening test of the general mental status, which has been used in most reported studies, provides a more detailed characterization of functional brain abnormalities in AD. Because of the substantial overlap for some metabolite ratios between the 2 groups, our findings have limited value for clinical diagnosis in individual patients. However, our data on specific brain regions support established models of disease progression within the brain and are consistent with functional laterality of the brain. More support for understanding the course of neurochemical changes in AD could be provided by a longitudinal follow-up study of the same cohort of patients.

Accepted for publication February 11, 2002.

Author contributions: Study concept and design (Ms Chantal, Mr Labelle, and Drs Bouchard, Braun, and Boulanger); acquisition of data (Ms Chantal, Mr Labelle, and Dr Boulanger); analysis and interpretation of data (Ms Chantal, Mr Labelle, and Dr Boulanger); drafting of the manuscript (Ms Chantal and Dr Boulanger); critical revision of the manuscript for important intellectual content (Ms Chantal, Mr Labelle, and Drs Bouchard, Braun, and Boulanger); statistical expertise (Ms Chantal); obtained funding (Ms Chantal and Dr Bouchard); administrative, technical, and material support (Mr Labelle and Drs Braun and Boulanger); study supervision (Drs Bouchard, Braun, and Boulanger).

This work was supported by grants from the Fondation and Département de la Recherche, Hôpital de l'Enfant-Jésus du CHA, Quebéc City, and by a scholarship from the Alzheimer Society of Canada, Toronto, Ontario (Ms Chantal).

We thank Andrée Morin, RN, for her assistance with patient recruitment and Abdesslem Khiat, PhD, for helpful discussions.

Corresponding author and reprints: Sophie Chantal, MPs, Centre de recherche, bureau H-441, Hôpital de l'Enfant-Jésus du CHA, 1401 18^e Rue, Quebec City, Quebec G1J 1Z4, Canada (e-mail: sophie.chantal@cha.quebec.qc.ca).