Occipital hypoperfusion on SPECT in dementia with Lewy bodies but not AD

Methods.

SPECT methodology.

Subjects were injected with 500 MBq of ^99mTc-HMPAO (Ceretec; Nycomed Amersham, UK) while seated in quiet and relaxed surroundings. Tomographic images were acquired on an IGE CamStar XR/T gamma camera with high-resolution collimator using 360° rotation with 64 stops of 25 seconds each, 64 × 64 matrix, and 5.4 mm/pixel. Processing was carried out on a Nuclear Diagnostics Hermes computer (Sweden) with the commercially available software package BRASS.23 BRASS includes a default SPECT template registered in Talairach space24 and a set of predefined regions of interest used for quantification.25

All images were reconstructed using ramp-filtered back-projection and then three-dimensionally smoothed with a Butterworth filter order 10 (cutoff 1.2 cycles/cm). A new template image was created by registering the 20 controls to the default BRASS template using nine-parameter affine transform and summing the registered images. These registered images of control subjects were then registered to the new template and summed up, creating the final SPECT template used in this study. Both control subjects and patients were registered to this final SPECT template to make them similar in size and shape. In each of these registered images, the average counts were then calculated for each of the 19 predefined regions of interest ( figure 1). The counts for each region of interest were then referenced to the cerebellar lobe with the highest activity in each subject, providing semiquantitative rCBF ratios. Scans were also visually assessed, by consensus rating of two experienced raters (blind to diagnosis), and classified as either normal or abnormal (the latter showing temporal and/or parietal hypoperfusion). Abnormal scans were further examined for the presence of marked occipital hypoperfusion (unilateral or bilateral), which was classified as either “definitely present” or “absent.”

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. (Left) Transverse slices of the 17 regions of interest (ROI; whose counts are referenced to counts in the cerebellum). 1 = Left frontal ROI; 2 = right frontal ROI; 3 = left central ROI; 4 = right central ROI; 5 = left parietal ROI; 6 = right parietal ROI; 7 = left occipital ROI; 8 = right occipital ROI; 9 = left temporal ROI; 10 = right temporal ROI; 11 = left medial temporal ROI; 12 = right medial temporal ROI; 13 = left basal ganglia ROI; 14 = right basal ganglia ROI; 15 = left thalamus ROI; 16 = right thalamus ROI; 17 = pons ROI; 18 = left cerebellum; 19 = right cerebellum. (Right) Corresponding slices of a ^99mTc-hexamethylpropyleneamine oxime SPECT image (constructed from the summing up of the mean counts of the control group).

Results.

Demographic and clinical characteristics of individuals are summarized in table 1. Groups were comparable for age, sex, and length of history. As would be expected, CAMCOG and MMSE scores were lower in both dementia groups than in control subjects (p < 0.001). There were no significant differences between AD and DLB subjects.

The mean ± SD rCBF indexes obtained in the three groups are shown in table 2, with significance of differences between the rCBF of these groups revealed by post hoc Bonferroni tests. Both AD and DLB patients showed a significant reduction in the parietal and temporal regions compared with the control subjects, with the differences in the parietal regions of interest being more significant in the DLB group and in the temporal regions of interest more significant in the AD group. The AD group also showed a significant reduction in rCBF in the frontal and medial temporal regions. The only difference between the AD and DLB groups that reached significance was in both left and right occipital regions, with the DLB patients showing an increased rCBF deficit compared with both the AD patients (p < 0.01) and the control subjects (p < 0.001). These changes are illustrated in figure 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. (A) Demonstration of occipital region of interest on a sagittal slice (yellow arrow). (B) Sagittal SPECT image of a patient with AD. (C) Sagittal SPECT image of a patient with dementia with Lewy bodies (DLB). Note the occipital hypoperfusion (yellow arrows) in the DLB compared with the AD patient (B).

Four DLB subjects had no history or current evidence of visual hallucinations. To determine whether occipital hypoperfusion may be related to the occurrence of hallucinations, these subjects were compared with those with hallucinations (n = 18). There was no suggestion of a relationship between occipital rCBF and hallucinations (left occipital rCBF (mean [SD]) in hallucinators 0.90 [0.08] and in nonhallucinators 0.88 [0.13]; t = −0.30, df = 23, p = 0.77). To investigate the ability of blood flow SPECT to discriminate between groups, a stepwise discriminant analysis was performed with rCBF measures as the dependent variables. Variables entered into the equation were left occipital rCBF (Wilks lambda = 9.92, df = 2, p < 0.001) and right temporal rCBF (Wilks lambda = 8.39, df = 4, p < 0.001). No other variables were entered into the analysis. Classification information for cases is shown in table 3. As can be seen, these variables correctly identified 17 of 20 (85%) control subjects and 63 of 73 (86%) demented (AD and DLB) subjects. More importantly, 15 of 23 (65%) DLB subjects were correctly identified as such, whereas only 9 AD and no control subjects were incorrectly classified as DLB (sensitivity 65%, specificity 87%). Visual assessment of scans correctly identified 16 of 20 control subjects (80%) and 69 of 73 patients with dementia (95%). Definite occipital hypoperfusion was noted in 9 of 23 DLB cases (39%) and 10 of 50 AD cases (20%).

Discussion.

We found that DLB was associated with occipital hypoperfusion on perfusion SPECT, not only in comparison with age-matched control subjects but also in comparison with AD patients of a similar age and with a similar severity of cognitive impairment. Our results are consistent with a previous report of seven DLB subjects16 and a study using [¹²³I]iodoamphetamine SPECT and statistical parametric mapping, which also found evidence of occipital hypoperfusion in DLB.19 These SPECT studies extend PET studies with fluorodeoxyglucose that have demonstrated profound occipital hypometabolism on PET,26,27⇓ albeit on very small numbers of subjects. The confirmation of this finding of occipital hypoperfusion in DLB in a larger, rigorously assessed, and more clinically representative population of DLB subjects with confirmed diagnostic accuracy supports the notion that this change on blood flow SPECT is a feature that should now be acknowledged as having a robust association with DLB.

It may be that reduced blood flow in occipital regions in DLB is linked with one of the core features of the disorder, that of visual hallucinations. Changes in occipital regions have been associated with hallucinations using functional MRI.28 However, in the current study, DLB subjects without hallucinations, although few in number, had a similar pattern of occipital hypoperfusion as those who had hallucinations. A study using PET also failed to find an association between occipital metabolism and hallucinations.27 Another explanation is that the occipital deficits relate to some other feature of the disorder, and occipital perfusion and metabolism have always been observed in patients with PD, in those both with and without dementia.29 However, Lewy bodies (the classic pathologic hallmark of both PD and DLB) rarely affect the occipital lobes. Might cholinergic loss be a possible explanation? This is significantly greater in DLB than AD in most cortical areas, including the occipital area30; with use of a SPECT ligand for the presynaptic choline transporter, reductions in occipital as well as parietal cortex in patients with PD without dementia have been shown.31 However, perfusion changes may not correlate well with cholinergic loss.31 Alternatively, it has been suggested that abnormal visual input from the retina and dopaminergic abnormalities both here and within other parts of the visual pathway may relate to this deficit.26 Occipital hypometabolism has been related to movement disorder in PD,29 suggesting an association with nigrostriatal degeneration. Although further research into the causes of occipital hypoperfusion in DLB is clearly required, we would suggest that the current evidence, based on the pattern of occipital hypoperfusion in PD and the lack of association with visual hallucinations in our study, argues that this finding is a feature of DLB itself, rather than being simply a correlate of the presence of visual hallucinations.

Unlike previous SPECT studies, the sample size in the current investigation allowed some estimation of the diagnostic potential of perfusion SPECT imaging. This is an extremely important clinical question, given that DLB and AD are common causes of dementia and perfusion SPECT is widely available in most clinical settings. In keeping with numerous previous reports, we replicated the robust bilateral temporal and parietal deficits that have been described for AD subjects when compared with control subjects. Using a discriminant function analysis, we tried to separate all three groups (control, AD, and DLB). Variables that significantly discriminated between groups were right temporal and left occipital CBF. With use of these measures, discrimination between control subjects and those with dementia (whether AD or DLB) had 86% sensitivity for dementia with 80% specificity, results very much in keeping with previous studies. However, arguably, the distinction between control subjects and those with established dementia has already been made on neuropsychological grounds and is not of particular interest. Of more importance is the ability of perfusion SPECT to discriminate between those with AD and DLB, particularly when subjects are matched for age and degree of cognitive impairment, as in this study. Perhaps surprisingly, given that others have tended to dismiss the value of perfusion SPECT scanning in differentiating between AD and DLB,32 14 of the 22 DLB subjects (sensitivity 64%) were correctly identified using SPECT, whereas 39 of the 51 AD subjects (specificity 76%) were correctly classified as not having DLB. On blinded visual rating of scans, occipital hypoperfusion could be observed twice as often in DLB subjects (39%) compared with those with AD (20%). Although these figures are encouraging, they also suggest a degree of overlap, which means that perfusion SPECT will not be diagnostically specific in an individual case. Of interest, hypometabolism in the visual association cortex on PET showed a sensitivity of 86% and specificity of 91% in separating DLB from AD.18 Our results, particularly when combined with others, would suggest that the presence of occipital hypoperfusion on SPECT in a patient with progressive dementia should, at the very least, raise a high level of awareness that DLB may be the underlying cause and prompt a very careful search for other core features of the disorder.

The strengths of this study are the relatively large sample size, the inclusion of an age-matched control group, the inclusion of subjects from a representative (geographically defined) catchment area (rather than from a specialist research clinic), and the standardized acquisition and assessment (using an automated SPECT template system linked to Talairach space) method of analysis. Potential weaknesses include reliance on clinical rather than neuropathologic diagnosis in most cases, the use of a single-head rotating gamma camera, and the inclusion of a relatively large number of NINCDS/ADRDA possible rather than probable AD subjects. Whereas reliance on clinical diagnosis can always be criticized, we did have neuropathologic confirmation in six cases. Importantly, we have recently published a prospective validation study of the criteria adopted here, with clinical diagnosis performed in precisely the same way, by consensus between the same three investigators.3 Not only did we demonstrate a high sensitivity and specificity for the NINCDS/ADRDA criteria for probable and possible AD, as have others, but we also demonstrated a sensitivity of 0.83 and specificity of 0.91 for the diagnosis of DLB. As subjects in this study were recruited and diagnosed in the same way, we feel confident that the diagnostic accuracy in the current study mirrors these findings. A substantial number of our subjects did have possible as opposed to probable AD. The main reason for this was our inclusion of AD subjects from a representative sample of patients presenting to old-age psychiatry services. To simply choose those with probable AD, which many other studies have done, would have led to a selection bias. However, in case the inclusion of possible AD cases confounded our results, we repeated analysis including only probable AD cases, and results were unchanged.

Frontal hypoperfusion has been reported to be associated with DLB,17 though it was not found in this study. Indeed, AD rather than DLB was associated with frontal hypoperfusion. Visuospatial and attentional impairments are certainly characteristic of DLB: Both can be associated with frontal pathology, and an early CT study did suggest frontal atrophy in DLB.33 However, DLB is not associated with other features suggestive of frontal lobe dysfunction (e.g., perseveration, disinhibition, language impairment), whereas more sophisticated and much larger volumetric MR studies have shown that frontal lobe volume is not reduced in DLB. This finding, together with other negative SPECT studies,19,27,32⇓⇓ supports our own conclusions that frontal lobe dysfunction and reduced perfusion on rCBF SPECT are not characteristic of DLB. Indeed, we have argued that the attentional impairments in DLB may be related to cholinergic loss and so have a neurochemical, rather than a structural, basis.34

Other imaging changes have been suggested as helping to separate DLB from AD. These include relative preservation of the temporal lobe and hippocampus on structural MRI35 and decreased density of the dopamine transporter on PET or SPECT.16,36⇓ However, it remains to be seen whether structural MRI, perfusion SPECT, decreases in dopamine transporter, or other techniques will ultimately have potential to assist with the differential diagnosis of DLB not only from AD but also from vascular dementia. Only future longitudinal studies combining all modalities will be able to realistically address this issue.