Introduction

Much research effort has been expended on trying to identify a marker of disease that would firstly help differentiate normal ageing from either MCI or mild AD and secondly help differentiate subjects with MCI due to underlying AD that will progress to dementia from those who will not. This is an important clinical problem representing one of the main reasons for referral to memory clinics. Currently no such biomarker exists although progress has been made in both biochemical (Shaw et al. 2009; Thambisetty et al. 2010) and neuroimaging markers (Fox et al. 1996; Jack et al. 1992; Jack et al. 1997; Juottonen et al. 1999; Xu et al. 2000).

One neuroimaging marker that shows promise is 1H-MRS (Kantarci 2007; Valenzuela and Sachdev 2001). It is clear that there is a pattern of metabolic differences in AD relative to normal controls detectable using 1H-MRS but to date relatively few studies have included subjects with mild AD or MCI and fewer still have included only MCI and control subjects recruited in the community. In this study we have examined hippocampal 1H-MRS in a group of patients with known memory problems (mild AD recruited through memory clinics) and in subjects not previously known to have memory problems (control and MCI subjects recruited in the community). We confirm reduced hippocampal [NAA] in mild AD relative to controls, with MCI subjects recruited in the community demonstrating an intermediate level.

Methods

Subjects

Subjects with mild dementia, diagnosed as probable AD according to NINCDS-ADRDA criteria (McKhann et al. 1984), were recruited from out-patient/memory clinics in South London. Subjects with MCI and unaffected controls were recruited from three general practice lists. All people over the age of 65 years registered with these GPs were contacted and invited to participate in the research project. Subjects were interviewed in the community and classified as normal elderly if they scored between 28 and 30 on the Mini Mental State Exam (MMSE) (Folstein et al. 1975), did not score abnormally on the Consortium to establish a registry for Alzheimer’s disease (CERAD) neuropsychological battery (Morris et al. 1989), were independent and living in the community, did not have a psychiatric or neurological condition and were not depressed using the Geriatric Depression Scale (Yesavage 1988). Subjects were defined as amnestic MCI if they met the criteria of Petersen et al. (2001).

The study was approved by the South London and Maudsley/Institute of Psychiatry Local Research Ethics Committee.

1H-MRS Protocol

Subjects were scanned using a 1.5 Tesla, GE NV/i Signa MR System (General Electric, Milwaukee, WI, USA) at the Maudsley Hospital, London. 3D T1-weighted volume images were acquired in the axial plane with 1.5 mm contiguous sections using acquisition parameters chosen using a contrast simulation tool (Simmons et al. 1996). Repetition time (TR) was 13.8 ms, inversion time (TI) 450 ms, echo time (TE) 2.8 ms, and the flip angle was 20° with one data average and a 256 × 256 × 124 voxel matrix. Acquisition time was 6 min, 27 s.

1H-MRS voxels of interest measuring 20 × 20 × 15 mm³ (6 ml) were defined in standard locations in the left and right hippocampi using previously published method (Robertson et al. 2001). We chose hippocampal regions of interest as this is one of the earliest sites of change in Alzheimer’s disease. The anterior extent of the voxel was defined as the coronal slice where the amygdala disappeared, with the posterior extent 20 mm from this (Fig. 1). The hippocampal volume of interest contained both grey and white matter and included the parahippocampal gyrus and the posterior portion of the amygdala.

Fig. 1
figure 1

Coronal T1 weighted magnetic resonance image from a healthy subject illustrating the location of the 1H-MRS voxels in the left and right hippocampi

Full size image

A point resolved spectroscopy (PRESS) pulse sequence (TE 35 ms, TR 1500 ms, 256 data averages and 2048 points) with automated shimming and water suppression and excellent reproducibility (Simmons et al. 1998) was used to obtain spectra from each voxel after CHESS water suppression with high signal to noise ratio and clearly resolved NAA, Cho, mI and Cr + PCr peaks among other metabolites. Non water-suppressed data were also collected for water referencing, but data was not collected to measure metabolite T1 and T2 relaxation times for individual subjects due to the limited tolerance of Alzheimer patients for MRI scanning. Not all subjects had spectral data from both left and right hippocampus. No significant differences were found in the metabolite content between the right and the left side of hippocampus. Therefore, we averaged the metabolite measures from the left and right hippocampus from the subjects which had data from both hemispheres.

1H-MRS Data Analysis

Differences in proportions of white and grey matter in the 1H MRS voxels may confound group differences in metabolite concentrations. Thus, to ensure that differences in tissue composition did not account for metabolic differences between subject groups, we segmented the SPGR volumes using SPM (Statistical Parametric Mapping) software (http://www.fil.ion.ucl.ac.uk/spm) to determine the percentage of grey matter, white matter and CSF within the MRS voxels. The position of the 1H MRS voxels relative to the segmented 3D dataset was determined automatically using in-house software. T1 and T2 corrections were applied for each metabolite using literature values (Christiansen et al. 1993).

Spectra were processed using LCModel (Provencher 1993) and metabolite concentrations were automatically corrected for CSF contamination of the voxel by dividing by the tissue fraction of the MRS voxel determined using SPM. These corrected concentrations were then calibrated to absolute molar units with respect to a phantom of known concentration, which was scanned in the same scanning session as the subject, using a PRESS acquisition with the same TE and TR.

Statistical Analysis

Comparisons between age, education level and MRS metabolites between the groups were made using univariate general linear models. Differences between gender were tested for using a Chi Squared test. Group differences in metabolite concentrations were tested with one-way analysis of variance (ANOVA), with group as the between subject factor. There were no significant interactions between the side from which 1H-MRS metabolites were measured (left or right hippocampus) or gender and group. Therefore, mean hippocampal metabolite values were considered in the analysis. Differences in metabolite concentrations were considered significant if P < 0.05. Least significant difference (LSD) post hoc tests were employed. The association between neuropsychological tests and the MRS variables in the patients with AD were analyzed using partial correlations, covarying for age. We used logistic regression to assess whether MRS metabolite concentrations could reliably classify individuals according to AD status, compared to the MMSE. Statistical analyses were performed using SPSS version 11.0.1 (SPSS Inc, Chicago, Ill).

Results

We studied 98 subjects; 39 with AD, 21 with MCI and 38 healthy elderly controls. The three groups were not significantly different with respect to age, years of education or male to female ratio. The subject group with AD had a MMSE score indicating mild dementia (mean 23; range 16–30; Table 1). Composition of the MRS voxel did not differ between the groups with respect to grey or white matter, but the proportion of CSF was greater in the patients with AD than in the normal elderly controls (F = 12.01, P = 0.0004).

Table 1 Demographic and neuropsychological characteristics of healthy controls subjects, patients with Alzheimer’s disease and subjects with mild cognitive impairment
Full size table

People with AD had a substantial reduction in [NAA] compared to normal controls (F = 13.33, P = 0.0005) and a significant reduction compared to those with MCI (P < 0.05). There was a non-significant trend towards a difference in NAA between MCI and normal controls (P = 0.06; Fig. 2). [Cr + PCr] was reduced in both AD (F = 4.33, P < 0.005) and MCI (P < 0.05) relative to controls but there was no difference in [Cr + PCr] between those with AD and MCI. We found no significant between-group differences in [mI] or [Cho] (Table 2) Fig. 3.

Fig. 2
figure 2

Hippocampal [NAA] within the hippocampi of healthy controls subjects, patients with Alzheimer’s disease and subjects with mild cognitive impairment. One-way ANOVA indicated that people with Alzheimer’s disease had significantly lower [NAA] compared to the healthy control subjects (P = 0.0004) and subjects with mild cognitive impairment (P = 0.024). † differs from controls ‡ differs from MCI

Full size image
Table 2 Mean hippocampal metabolite concentrations of healthy controls subjects, patients with Alzheimer’s disease and subjects with mild cognitive impairment
Full size table
Fig. 3
figure 3

Hippocampal [Cr + PCr] within the hippocampi of healthy controls subjects, patients with Alzheimer’s disease and subjects with mild cognitive impairment. The healthy control subjects had higher [Cr + PCr] than people with Alzheimer’s disease (P = 0.003) and subjects with mild cognitive impairment (P = 0.023). † differs from controls ‡ differs from MCI

Full size image

Logistic regression analysis showed that mean hippocampal [NAA] and [Cr + PCr] correctly categorised 75% of normal controls with a negative predictive probability of 73% and correctly categorised 72% of the patients with AD with a positive predictive probability of 74%.

Within the combined AD and MCI groups we carried out a planned exploration of the relationship between overall cognitive ability (as measured by the MMSE) and hippocampal [NAA]. The correlation between mean [NAA] and overall cognitive ability showed a trend which did not reach significance (r = 0.255, P = 0.056). However, there were significant correlations between other neuropsychological parameters and [NAA], with hippocampal [NAA] correlating with both delayed recall of a learned word list (r = 0.330, P = 0.018) and delayed praxis (r = 0.341, P = 0.015). These correlations remained significant after correcting for age (r = 0.345, P = 0.016 and r = 0.334, P = 0.020, respectively).

Discussion

In this study we have found that changes in [NAA] and [Cr + PCr], measured by 1H-MRS, differ between mild AD, MCI and unaffected elderly controls and correlates with neuropsychological measures of cognition in the impaired groups. We chose to study MRS changes in the hippocampus as this area is affected early in the Alzheimer’s disease process and studied only subjects early in the disease process as these present the most clinically relevant group with respect to a diagnostic biomarker.

Previously a decrease in [NAA] in AD has been well established in different brain regions (reviewed in Valenzuela and Sachdev 2001) as well as in whole brain (Falini et al. 2005). Many of these studies have used MRS in regions of the brain other than the hippocampus. Obtaining a reliable MRS spectrum in hippocampus in AD can be challenging as this is the site of earliest pathology and hence shows increased atrophy. However, some previous studies have focused on hippocampus and these also have shown a decrease in NAA (Ackl et al. 2005; Dixon et al. 2002; Schuff et al. 1997). Our results confirm and extend this work, demonstrating that hippocampal [NAA] is reduced even in very mild AD compared to both healthy controls and to those with MCI. We have shown that mean hippocampal NAA values in MCI fall between those of mild AD and normal elderly controls.

This finding of an intermediate value of [NAA] in MCI also mirrors previous studies. Thus, Chantal et al. (2004) reported a difference in the NAA/Cr + PCr ratio between subjects with MCI and controls in the left medial temporal lobe and Ackl et al. (2005) found a similar difference in hippocampus. However, neither found a difference between subjects with MCI and patients with AD. A decrease in NAA/Cr + PCr in MCI relative to controls was also found by (Kantarci et al. 2006) but (Catani et al. 2001) and (Kantarci et al. 2000) found no difference in NAA/Cr + PCr between subjects with MCI and controls. Indeed NAA/Cr + PCr was found to be the most sensitive measure distinguishing between MCI and AD (Kantarci et al. 2002).

It is clear from these studies that measures of NAA and NAA/Cr + PCr are decreased in established AD but the findings from very mild AD and from MCI are less certain. One possible reason for this discrepancy is the regions of brain examined in different studies—it is possible that changes in MCI are confined to those regions of the brain affected early in the AD process. Our finding of an intermediate NAA value in MCI in hippocampus is in line with this hypothesis.

Another source of variability in studies of MCI is the means of subject recruitment. In a previous systematic review (Bruscoli and Lovestone 2004) we noted the substantial difference in conversion rates from MCI to AD—ranging from less than 5% to more than 30% per year. The only variable distinguishing high from low conversion rates was the source of the recruitment to the study, with memory clinic attendees showing a conversion rate twice that of community dwelling volunteers. This finding suggests that MCI subjects entering studies via memory clinics are substantially different to subjects with MCI in the community even though clinical and neuropsychological assessments fail to distinguish between these groups. Most, if not all, previous 1H-MRS studies of MCI have recruited subjects through memory clinics or similar specialist services raising questions as to the generalisability of these findings as well as perhaps explaining some of the apparent discrepancies in the literature. All of the subjects with MCI in our study were recruited through direct contact and not through self or professional referral to a memory clinic. The only difference between controls and MCI subjects in the present study was the MCI inclusion criteria and therefore we are confident that this group represents, more fully than in previous studies, those people with a mild cognitive impairment as it occurs frequently in the community. As such the group of people with MCI represented in this study closely resemble an important target group in the community—those concerned about their memory but not yet acting upon these concerns. We find that in this representative group MRS metabolites are indeed altered in the hippocampus with a reduction in [Cr + PCr] and a trend towards a reduction in [NAA]. Thus, MCI seems to be an intermediate state and in line with this we found that within the impaired group hippocampal [NAA] was related to overall cognitive ability, as measured by the MMSE, and more specifically to memory, as measured by the delayed recall of a learnt word list and the delayed praxis subtests of the CERAD neuropsychological battery.

[NAA] has also been used to discriminate between patients with AD and controls. We found that we could best distinguish between our patients and controls using a model including hippocampal [NAA] and [Cr + PCr] (74% classified correctly). These findings are in line with others (Dixon et al. 2002; Schuff et al. 1997) although these previous studies included more severely affected patients. Despite this we found similar classification accuracy even in very mild AD.

We also found a reduction in [Cr + PCr] in the subjects with AD and MCI compared to normal controls. The Cr + PCr peak measures creatine and phosphocreatine, which are present in both neurons and glial cells and are thought to reflect energy use and storage by neurons, and are relatively stable over time and largely unaffected by disease states. However, our finding of a significantly lower [Cr + PCr] in people with AD and those with MCI may suggest an impairment in energy metabolism which in AD is in accordance with some (Adalsteinsson et al. 2000; Ernst et al. 2004) but not all (Huang et al. 2001), previous findings.

We found no difference in hippocampal [mI] between AD, controls and MCI. Previously, increases in mI in AD have been reported in parietal lobe (Rose et al. 1999) and temporal lobe (Chantal et al. 2002; Parnetti et al. 1997). Only one study has measured mI levels within the hippocampus, however, and this reported no change in mI/NAA or mI/Cho + Cr + PCr (Dixon et al. 2002).

Kantarci et al. (2000) found that mI/Cr + PCr was different between MCI and control groups in posterior cingulate gyrus and paratrigonal white matter and Chantal et al. (2004), between MCI and AD in parietotemporal cortex. No one has previously examined hippocampal mI in people with MCI.

Conclusions

We have demonstrated that there is a significant reduction in hippocampal [NAA] and [Cr + PCr] in early AD, and that people with MCI are intermediate between early AD and control groups. The test characteristics of the MRS data do not meet the criteria for a definitive stand-alone biomarker (The Ronald and Nancy Reagan Research Institute of the Alzheimer’s Association and, National Institute on Aging Working Group 1998) but nonetheless this data is promising considering that we have studied a group of patients in the severity range where diagnosis can be difficult and in people not yet referred by themselves or by their physicians to specialist services. These findings suggest that in this group of patients MRS, alone or in combination with other imaging and non-imaging biomarkers may prove to be useful in the diagnostic process and may also be valuable in assessing treatments for dementia.