Results

FOLLOW UP CLINICAL AND BRAIN PET FINDINGS

All patients were clinically monitored for a mean (SD) period of 15 (8) months. Interestingly, in two patients (patient 1 and patient 8) with initially pathological brain PET and inconspicuous MRI scan, the CNS condition worsened in the follow up period and the MRI scans became pathological after 2 and 10 months, respectively.

Follow up PET examinations were performed in 13 patients with initially pathological findings and in one patient with initially normal brain PET; the follow up periods ranged from 2 to 33 months (table 3). Patient 2 had two follow up PET examinations.

Patients with regression or improvement of all CNS symptoms (n=8) also showed improvement or normalisation of brain PET findings in all cases (fig 1). In patients with persistent neuropsychiatric symptoms (n=4) hypometabolism was unchanged (n=3) (fig 1) or partially normalised (n=1). Both patients with worsening of particular cerebral symptoms, PET examination also worsened in the affected brain regions (patients 2 and 4).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Left panel: FDG-PET and MRI studies in a 46 year old women with SLE and vertigo, gait disturbance, headache and depression (patient 5). (A) An axial T2 weighted MRI scan demonstrated infarction parieto-occipital right and brain atrophy. (B) Initial FDG-PET showed glucose hypometabolism parieto-occipital right, occipital and laterofrontal left. (C) After 30 months vertigo, gait disturbance and headache persisted, but depression improved. Follow up FDG-PET showed nearly unchanged glucose uptake defects. The right image side represents the left brain side. Right panel: FDG-PET and MRI studies in a 30 year old male patient with SLE and headache (patient 13). (A) MRI scan was normal. (B) Inital FDG-PET image showed a significant reduction in glucose metabolism in the parieto-occipital region on both sides. (C) Ten months later the headache had improved. Follow up FDG-PET showed a normalised glucose utilisation (within 1 standard deviation of normal controls) in the previously affected brain regions.

In patients with intensified or newly introduced immunosuppressive treatment (n=7), the pathological brain PET findings were improved or normalised in six and unchanged in one patient. Patients with unchanged treatment (n=3) showed worsened, improved or unchanged PET scan results. In patients in whom immunosuppressive treatment was reduced (n=3), brain PET findings worsened, did not change or improved. Patient 16 showed normalisation of brain PET after change from oral to intravenous cyclophosphamide.

The mean SLEDAI score was 15.3 (SD = 10.5) at initial PET evaluation and 5.7 (SD = 5.5) at the follow up visit. The reduction in the SLEDAI score was higher in patients with intensified (mean reduction of SLEDAI score 15.4 (9.8)) compared with unchanged or reduced immunosuppressive treatment (mean reduction of SLEDAI score 3.8 (2.9); p=0.01). PET findings were not related to the SLEDAI score. Thus, a worsened hypometabolism occurred in one patients with a SLEDAI score of 0.

Discussion

The diagnosis of CNS involvement in connective tissue diseases may be problematic and has to be differentiated from neurological impairment attributable to infections, uraemia, hypertension, drug toxicity. SLE patients with focal deficits are more likely to show MRI abnormalities than SLE patients with diffuse presentations.11 ,12 Furthermore, MRI analyses did not provide any additional clinical information on cognitively impaired SLE patients.13 In the past decade functional imaging has been proved to be more sensitive than morphological imaging. Different methods were used to measure brain function: cerebral blood flow measurement using a xenon-133 inhalation method,26 ,27technetium-99m-hexamethylpropylene amine oxime (HMPAO)14 ,15 ,28 ,29 or iodine-123-iodoamphetamine SPECT,30 glucose utilisation using fluorine-18 fluorodeoxyglucose (FDG) PET17 ,18 ,31 and magnetic resonance spectroscopy.32 The clinical value of SPECT in diagnosing cerebral lupus is being discussed controversially: some studies showed much greater sensitivity than specificity for CNS involvement; others found multifocal areas of cerebral blood flow defects also in asymptomatic cases.15 ,30 ,33

Glucose metabolism measured in PET scans provided a higher resolution and has been shown to be more sensitive than perfusion data obtained by SPECT in several diseases other than SLE.34 ,35 Earlier PET studies involving only few patients showed pathological changes in accordance with the clinical state.17 ,36 These and most following studies focused exclusively on manifest and severe CNS involvement with positive MRI scans.17 ,36 ,37 However, in recent studies hypometabolism in FDG-PET has also been shown in MRI negative neuropsychiatric SLE.18 ,33 In contrast, one study describes a significant tendency towards lower glucose utilisation only in patients with a high number of white matters lesions found by MRI.31 Recently Kaoet al compared HMPAO-SPECT and FDG-PET to detect brain involvement in SLE patients and found a higher sensitivity but lower specifity of HMPAO-SPECT.33 However, as in the latter study PET scans were interpreted visually instead of a quantitative method as we used, their PET results are not comparable to ours.

In this study FDG-PET hypometabolism was found in at least one brain region of all patients with major or minor CNS symptoms. There were conflicting reports regarding the regions involved in neuropsychiatric SLE: whereas in SPECT studies a variety of abnormal brain regions have been reported (occipital, frontal, temporo-occipital, parietal cortex, basal ganglia and thalamus)14 ,15 ,28 ,33 in FDG-PET scans hypometabolism was mainly found in parietal, temporal, frontotemporal and central regions.17 ,36 ,38 ,33 In our study the most commonly affected brain region was parieto-occipital, followed by parietal, cerebellar and frontal, whereas temporal areas were rarely involved. Overall in 12 of 23 symptomatic patients and in 8 of 12 patients with normal MRI scans the parieto-occipital region was the only brain region involved. It remains unclear why these lesions were rarely described in previous studies; different scanning techniques may be the reason. One clinical reference to the parieto-occipital region in our patients may be the visual disturbances observed in six patients. Besides visual symptoms no clear correlation could be found between brain glucose utilisation and the topography of the neuropsychological complains. Taken together, our data suggest that the parieto-occipital and parietal regions are the most vulnerable zones of the cerebrum in SLE. This may be explained by the location of the parieto-occipital region at the boundary of supply of the major cerebral arteries. However, abnormalities in the parieto-occipital region are not pathognomonic for SLE with neuropsychiatric manifestation: they can also be found in the late whiplash syndrome39 or in dementia of the Alzheimer type.40

Of particular interest are longitudinal studies in patients with CNS lupus: using SPECT, Falcini et al showed improvement of perfusion alterations in six of seven patients with juvenile SLE.15 Kodama et alshowed that SPECT changes represent the activity of psychiatric manifestations,41 whereas Reiff et al found no correlation of SPECT abnormalities with patient's clinical course.29 Few case reports using FDG-PET scans showed a clear correlation of hypometabolism with clinical conditions, especially with cognitive deficits.36 ,38 ,42 Accordingly, our follow up data showed amelioration of conspicuous PET findings in 9 of 13 patients. Moreover, the dimension of hypometabolism correlated well with the clinical course of CNS disease, but not with the overall disease activity expressed by SLEDAI score or the serological activity indicator C3d. In six patients improvement of PET findings in association with a marked reduction of SLEDAI score was attributable to changes in immunosuppressive treatment. On the other hand we also observed spontaneous resolution of cognitive dysfunction paralleled by improvement of PET findings in two patients without therapeutic intervention (patients 17 and 18). This phenomenon had also been observed in a recent SPECT study15 and might be explained by studies, which have indicated, that subclinical cognitive deficits resolve spontaneously, and do not predict the subsequent emergence of either more profound cognitive deficits or other forms of clinically overt neuropsychiatric SLE.43 ,44 Persistent hypometabolism of brain areas was observed in seven patients. Cerebral infarcts were underlying these lesions in two patients (patients 4 and 5), and extended white matter lesions in another patient (patient 7). However, persistent deficits in glucose metabolism were also found in patients with normal MRI scans (patient 1, 6, 18). Active and reversible functional brain lesions cannot be differentiated from chronic lesions in a cross sectional PET study. Therefore, longitudinal studies are necessary to clarify the question whether lupus patients may suffer permanent functional brain lesions in the absence of morphological changes and whether such lesions are of clinical relevance.

Abnormalities in functional imaging have also been reported in asymptomatic SLE patients.15 ,28 ,30 In asymptomatic patients we found unconspicuous PET in three or hypometabolism in singular brain regions in two cases whereas in symptomatic patients multiple brain regions were affected. Both asymptomatic patients with pathological brain PET did not develop cerebral symptoms during follow up. In contrast two patients with major neuropsychiatric symptoms and pathological PET scans had initially normal MRI scans. They subsequently developed a severe CNS exacerbation during which MRI became positive in both (patients 1 and 8). These data suggest that PET positive patients with severe neuropsychiatric symptoms may require intensification of treatment even if MRI is negative. Progress from mild to severe CNS involvement was not observed. However, in most patients immunosuppressive treatment was intensified because of the presence of other disease manifestations. In consideration of prospective data obtained in other studies,43 we would suggest that patients with mild symptoms should be monitored carefully but do not require obligatory intensification of treatment.

The pathophysiological mechanisms that underlie the findings in PET scanning may be heterogenous. Depressed glucose utilisation is currently thought to be a result of (1) cerebral atrophy, (2) infarction, (3) reduced density of cells or (4) reduced FDG uptake of cells.

(1) The spatial resolution of current PET scanners does not allow a distinction between cerebrospinal fluid containing spaces and contiguous brain tissue.45 Our data suggest that mild atrophy on MRI scan may have no influence on PET findings. However, if cerebral atrophy is moderate or severe, appropriate corrections for atrophy should be used to measure actual brain tissue metabolism more accurately. (2) Multiple infarctions may lead to a patchy pattern in PET, single lesions result in focal deficits of glucose metabolism. Further changes in cerebral metabolism may also occur at intracranial sites distant from ischaemic lesions, a phenomenon known as diaschisis.46 Diaschisis is attributed to a decrease in local neuronal activity caused by the loss of afferent projections from the damaged region. A loss of afferent projections may also be the cause of some changes in glucose metabolism seen in patients with white matter lesions. (3) A reduced density of cells as a common cause of hypometabolism in SLE patients seems to be unlikely, because hypometabolism was reversible in most cases. (4) Finally, a reduced FDG uptake of cells may represent either a reversible vascular disorder or a primary abnormality in brain structure.47 Widespread and diverse vascular lesions are a pathological hallmark of SLE. Thrombotic lesions would require antiphospholipid antibodies leading to small infarcts. Accordingly, antiphospholipid antibodies have been associated with neurological disorders of vascular origin,48 ,49 as seen in some of our patients. In contrast, inflammatory lesions would require either circulating immune complexes or significant complement activation of neutrophils.50 In studies on the pathology of cerebral lupus, vasculopathy, infarctions and haemorrhages are often observed, whereas vasculitis is rare.51 Otherwise, in lupus patients with CNS involvement a complement mediated vascular injury was described independent from immune complex deposition.52 Activation of complement may lead to an occlusive vasculopathy via activation, aggregation and adherance of neutrophils and platelets to vascular endothelium.53However, complement mediated vascular injury does not seem to be the main cause of brain hypometabolism as we were not able to demonstrate a correlation to C3d levels or to a consumption of total haemolytic complement.

Antibodies against neuronal structures have been described by several authors in lupus patients, especially in those with organic brain syndromes, psychosis or seizures.3 ,8 ,9 We screened patients sera for the presence of antibodies against beef brain tissue (anti-CNS) by a well established method.9 Interestingly, we found a correlation between anti-CNS antibodies and the presence of severe and to a lesser extent with mild neuropsychiatric symptoms. Antibodies against ribosomal P protein, an intracytoplasmic constituent of neuronal and other human cells, were reported to be highly specific for psychosis,6 ,54 or depression in lupus patients.6 ,55 The specificity of anti-ribosomal P antibodies to lupus psychosis or depression was not confirmed by other studies.56 ,57 We found only a weak association between anti-ribosomal P antibodies and mood disorders, but a statistically significant correlation with the presence of overall psychiatric symptoms (including reactive depression and neurosis). The discrepancy between our findings and those by others6 ,55 might be influenced by different assays used to detect anti-ribosomal P antibodies or by the small number of patients included in our study. The role of autoantibodies against neuronal cell constituents in the pathogenesis of cerebral SLE is unclear. It is assumed that autoantibodies exert their effects by interfering with the cell's ability as a responder in the neuronal network.58Microvascular injury induced by immune complexes, complement or antiphospholipid antibodies may change the blood-brain barrier and provide access of systemically produced antibodies that would otherwise not reach the CNS. However, we were not able to demonstrate any significant association between the PET or MRI findings and presence of distinct autoantibodies. Moreover, we observed brain glucose hypometabolism also in three patients (17, 18 and 24) without antibodies against cardiolipin, CNS tissue, ribosomal P protein or signs of complement consumption.

Finally, cytokines may play a part in the development of functional brain lesions seen in PET. Intrathecal synthesis of interferon α59 and interleukin 660 has been reported in neuropsychiatric lupus. Interestingly, in a recent study changes in brain glucose metabolism have been demonstrated in patients with chronic viral hepatitis during treatment with interferon α.61 This issue requires further study.

We conclude that the combination of PET and MRI constitutes the most useful diagnostic procedure to differentiate functional from morphological changes in CNS involvement of SLE. While MRI scans are highly sensitive in detecting infarction and especially white matter lesions, we suggest that brain PET is primarily indicated in patients with cerebral symptoms and unconspicuous MRI. In these patients FDG-PET may establish the diagnosis of CNS involvement and monitor the brain function. Despite lack of specificity, PET may also be helpful in diagnosis of unclear cases, when only few ACR criteria and minor neuropsychiatric symptoms are present.