Patterns of Cerebral Glucose Metabolism Detected with Positron Emission Tomography Dfier in Multiple System Atrophy and Olivopontocerebellar Atrophy

Sid Gilman, MD,” Robert A. Koeppe, PhD,? Larry Junck, MD,” Karen J. Kluin, MS,*$ Mary Lohman, BA,“ and Roy T. St Laurent, PhDg

We used positron emission tomography with {‘8F7fluorodeoxyglucoseto study local cerebral metabolic rates for glucose (ICMRglc) in patients with multiple system atrophy (MSA), sporadic olivopontocerebellar atrophy (sOPCA), and dominantly inherited olivopontocerebellar atrophy (dOPCA) in comparison with normal control subjects. In MSA, absolute ICMRglc was significantly decreased in the , , putamen, , and . In sOPCA, absolute 1CMRglc was significantly decreased in the brainstem, cerebellum, putamen, thalamus, and cere- bral cortex. In dOPCA, absolute ICMRglc was significantly decreased in the brainstem and cerebellum but not in the other structures. Examination of ICMRglc normalized to the cerebral cortex in comparison with normal controls revealed in MSA significant decreases in the brainstem, cerebellum, and putamen but, in both sOPCA and dOPCA, significant decreases only in the brainstem and cerebellum. The findings indicate that these three disorders all show a marked decrease of ICMRglc in the brainstem and cerebellum but differ in the degree of hypometabolism in forebrain and cerebral cortical structures. The results are consistent with the possibility that, in many cases, sOPCA will evolve into MSA. Moreover, positron emission tomography may provide helpful diagnostic information in these neurodegenerative diseases. Gilman S, Koeppe RA, Junck L, Kluin KJ, Lohman M, St Laurent RT. Patterns of cerebral glucose metabolism detected with positron emission tomography differ in multiple system atrophy and olivopontocerebellar atrophy. Ann Neurol 1994;36:166- 175

Multiple system atrophy (MSA) is an adult-onset spo- Several neurological diseases are considered to be radic neurodegenerative disease presenting clinically subgroups of MSA, including striatonigral degenera- with symptoms and signs of extrapyramidal, cerebellar, tion (SND) 19-12], Shy-Drager syndrome (SDS) C1, and autonomic dysfunction in various combinations 3, 13-15], and olivopontocerebellar atrophy (OPCA) [1-5}. The extrapyramidal symptoms are poorly re- with autonomic insufficiency or extrapyramidal disor- sponsive or unresponsive to levodopa therapy C6, 7). ders [2, 16-23]. When detected without autonomic The neuropathological features of MSA consist of de- insufficiency or cerebellar disorder, SND is considered generative changes, including neuronal loss and , to be “possible MSA,” whereas SDS and OPCA with in the basal ganglia (putamen and globus pallidus), autonomic insufficiency or extrapyramidal disorders brainstem and cerebellum (substantia nigra, locus ceru- are considered to be “probable MSA” C4f.In SND leus, dorsal vagal nuclei, , inferior ol- the neuropathological changes affect chiefly the basal ives, , and cerebellar Purkinje cells), and ganglia and substantia nigra, in SDS the brainstem and (, intermediolateral col- spinal cord, and in OPCA the inferior olives, pons, and umns, and Onuf’s nuclei) C2, 4, 81. Other central ner- cerebellum. vous system structures that can be affected include the There is substantial overlap in symptoms, signs, and optic , cerebral cortex, caudate nuclei, subtha- neuropathological changes in SND, SDS, and OPCA laic nuclei, red nuclei, dentate nuclei, anterior horn and thus strong justification for considering them all cells, and sympathetic ganglia C2, 41. potentially as forms of MSA [4].Some patients with

From the Departments of *, ?Internal Medicine, and Received Apr 19, 1993, and in revised form Dec 27. Accepted for $Physical Medicine and Rehabilitation, University of Michigan Medi- publication Jan 6, 1994. cd Center, and $Department of Biostatistics. University of Michigan Address correspondence to D~ ~il~~~,D~~~~~~~~of ~~~~~l~~~, School of Public Health, Ann Arbor, MI. University of Michigan Medical Center, 1500 E. Medical Center Dr, Ann Arbor, MI 48109-0316.

166 MSA initially develop extrapyramidal symptoms and pattern similar to that seen in MSA, but the dOPCA later experience autonomic or cerebellar symptoms or group should be free of such patients. both [41. Others first manifest cerebellar symptoms The hypothesis underlying this project was that the and later develop autonomic or extrapyramidal symp- metabolic profile should be different in the three toms. Although signs of cerebellar dysfunction may groups (MSA, sOPCA, and dOPCA). Based upon the not be evident in patients who develop SND or SDS, sites of degeneration in neuropathological studies, we neuropathological studies demonstrate degenerative anticipated finding hypometabolism in the basal gan- changes in the cerebellum and brainstem in most cases glia, cerebellum, and brainstem in MSA, but only in [S}. The loss of cerebellar and brainstem is the cerebellum and brainstem in dOPCA. Many pa- greatest in patients with MSA presenting initially with tients with sOPCA progress to MSA, and thus we an- cerebellar symptoms, less in those with SDS, and least ticipated finding some degree of hypometabolism in in those with SND {S]. In some patients, especially the basal ganglia, reflecting subclinical disease, in addi- those with hereditary disease, OPCA occurs without tion to involvement of the cerebellum. A preliminary involving the extrapyramidal or autonomic systems. In communication concerning this study has been pub- others, mostly in sporadic cases, OPCA can progress lished [27]. to MSA. Currently no method is available to determine whether sporadic OPCA will progress to include other Materials and Methods manifestations of MSA. Patient Groups and Normal Subjects Previously, using [l8F1fluorodeoxyglucose (FDG) The studies were approved by the Institutional Review and positron emission tomography (PET), we demon- Board, and informed consent was obtained from each patient strated hypometabolism in the cerebellum and brain- and normal control subject. The patients were segregated stem of patients with hereditary and sporadic OPCA into three groups based upon the history and physical find- E241. The degree of hypometabolism was directly re- ings, i.e., multiple system atrophy (MSA), sporadic olivopon- lated to the severity of the and to the degree of tocerebellar atrophy (sOPCA), and dominant olivopontocer- atrophy of these structures in anatomical imaging stud- ebellar atrophy (dOPCA)(Table 1). We studied 126 subjects ies C24-261. The present study was undertaken to ex- (51 normal controls and 75 patients), age 51 -+ 16 years (mean 2 SD), including 59 men (19 normal controls and 40 amine local cerebral metabolic rates for glucose patients), age 52 -+ 15 years, and 67 women (32 normal (1CMRglc) in the central nervous system of patients controls and 35 patients), age 50 * 16 years. Significant with MSA in comparison with patients with OPCA differences in age were found among the four diagnosis who presented with no clinical evidence of involve- groups (p < 0.01) with a one-way analysis of variance (AN- ment of autonomic or extrapyramidal systems, and OVA). Three patients, two with sOPCA and one with with normal control subjects similarly distributed by dOPCA, were included in our previous studies [24-261, but age and sex. We separated the OPCA patients into two all of these patients were rescanned for the present study. groups, those with sporadic disease (sOPCA) and those One of the patients with sOPCA in the previous communica- with disease inherited in an autosomal dominant pat- tions developed extrapyramidal and autonomic symptoms tern (dOPCA). This separation was made because the and signs subsequent to the initial study and is now in the dOPCA group should constitute a group of pure group of MSA patients. In analyzing the duration of symptoms, we found that OPCA patients, that is, patients without involvement skewness in the distribution precluded reporting means and of multiple systems, since people with dOPCA gener- standard deviations. Accordingly, we report the range and ally do not exhibit extrapyramidal or autonomic symp- median duration in each group. For MSA the range of dura- toms. The sOPCA group may include patients with tion was from 2 to 27 years with a median of 5 years. For clinical symptoms only of OPCA but with pathological sOPCA, the range was from 2 to 24 years with a median of involvement in multiple systems and with a metabolic 4 years, and for dOPCA the range was from 2 to 50 years

Table I. Numbers, Ages (Mean k SD), and Sex Distributions of the Subjects Studied Normal Controls dOPCA sOPCA MSA n Age (yr) n Age (yr) n Age (yr) n Age (yr)

Men 19 44 -+ 17 11 50 -+ 14 17 56 t 12 12 63 ? 8 Women 32 44 -+ 18 8 49 -+ 13 15 51 +- 11 12 65 ? 9 Total 51 44 -+ 17 19 49 -+ 14 32 54 ? 12 24 64 * 8 dOPCA = dominantly inherited olivopontocerebellar atrophy; sOPCA = sporadic olivopontocerebellar atrophy; MSA = multiple system atrophy.

Gilman et al: Cerebral Metabolism in MSA and OPCA 167 Table 2. Clinical Features in 24 Patients with Multiple System Atrophy Patient No.

1 2 3 4 5 6 7 8 9 10 11 12

Sex F M F F M F F F M F M M Age (yr) 59 61 65 82 68 61 58 65 49 16 58 62 Duration of symptoms (yr) 4 7 4 3 11 9 2 6 2 5 4 4 Akinesia ++ ++ +++ + ++ - + +++ ++ +++ ++ +i Rigldlty ++ +++ +++ ++ ++ - ++ +++ ++ +++ ++ +i Tremor (extra- pyramidal) ------+ + + - - Hypokineric speech ++ ++ 4- - ++ - + ++ ++ ++ + ++ Response ro levodopa NT - - NT - NT NT - ++ - - - Ocular dys- metria + + + + ++ + + - - ++ ++ + Gait ataxia + + + ++ +++ +++ ++ - - +++ +++ +++ Limb ataxia + - + - ++ ++ + - - +++ ++ + Ataxic speech + + + ++ ++ ++ + + + + +++ ++ Spastic speech - + + + + + + ++ + + + + Extensor plan- tar signs - - - - ++ ++ ++ ++ - ++ ++ - Postural hypo- tension ++ ++ ++ - +++ +++ ++ ++ ++ - - - Urinary incon- tinence ++ +++ - - +++ +++ ++ ++ + + + - MRI Cerebral at- why + + + - ++ + - - - + + + Cerebellar atrophy + + - ++ - ++ ++ ++ - - ++ + ++

( - I = absent, 1 +) = mild; j + + = moderate: ( + + +) = marked: NT = nor rested. MRI = magnetic resonance maging

with a median of 10 years. Studies were performed on nor- The demonstration of changes in the putamen in magnetic mal control subjects who had no history of neurological dis- resonance imaging (MRI) scans [28, 291 provided helpful order and had no abnormalities on general physical and neu- information for the diagnosis of MSA but was not required rological examinations (see Table 1).At the time of the PET since these changes are not specific to MSA 1301. For the study, none of the patients or normal control subjects were demonstration of cerebellar dysfunction, we required at least taking medications that are known to influence cerebral me- two signs, including limb and/or gait ataxia along with ocular tabolism. Four of the 24 patients with MSA were taking dysmetria or ataxic speech. For the demonstration of auto- medications for parkinsonism (carbidopa/levodopa: 25 : 100 nomic failure, we required postural hypotension or urinary four times daily, 25 : 100 three times daily, 10: 100 four times incontinence without obstruction. The criteria for postural daily along with selegeline 5 mg twice daily, and 10: 100 hypotension were similar to those of McLeod and Tuck (3 11, three times daily). including an orthostatic drop of 30 mm or more in systolic We studied 23 patients with probable MSA and 1 patient blood pressure and 20 mm or more in diastolic blood pres- (patient 16) with possible MSA [41. The diagnosis of proba- sure with an increase in heart rate of no more than 10 beats ble MSA was based upon demonstration of clinical signs of per minute. Blood pressure and pulse were measured supine at least two of the following: (1) extrapyramidal disorder, (2) and again 2 minutes after the patient had assumed the stand- autonomic failure, and (3) cerebellar dysfunction. For the ing position. In patients who could not tolerate standing for demonstration of extrapyramidal disorder, we required at 2 minutes, these changes were accepted after the patient least two of the following signs: akinesia, rigidity, tremor, stood as long as tolerated. Blood pressure measurements and hypokinetic speech. In most cases we also assessed the were made when the patients had been withdrawn from treat- degree of responsiveness to levodopa with clinical neurologi- ment with any medication that might induce postural hypo- cal evaluations before and after treatment with carbidopa/ tension. Most of the patients with postural hypotension had levodopa at a dose of 25 : 250, gradually increased from one an associated history of sexual impotence (in the males), an- to four tablets daily. Patients were judged to be unresponsive hydrosis, or incontinence without outflow obstruction. Some or mildly, moderately, or markedly responsive. Seventeen of the patients without postural hypotension had inconti- patients were tested; 14 were unresponsive, 1 was mildly nence without outflow obstruction, and all of these patients responsive, and 2 were moderately responsive (Table 2). The who were males had sexual impotence. In Table 2, the degree remaining 7 patients could not be tested. of postural hypotension and urinary incontinence refers to

168 Annals of Neurology Vol 36 No 2 August 1994 Table 2. Continued. Patient No.

13 14 15 16 17 18 19 20 21 22 23 24 F M F M F M F M F M M M 70 73 51 72 65 50 67 64 55 71 61 61

5 4 2 10 4 20 3 16 3 5 2.5 7 +++ ++ +++ + +++ ++ +++ ++ + + - + +++ + ++ + +++ - +++ ++ + ++ + +++

+ + - + ++ ~ t t i +++ + ++ ++ +++ - + ++ + + +t

I NT - ++ NT NT - + + + + + + + - + + ++ - + + i- + - - ++ + ++ +++ +++ - + ++ - - +++ + +++ ++ + - + +++ + - +++ ++ ++ +++ + + ++ - ++ + + + + ++

4 4 - +

- ++ +++ - + + i- ++ +++ f+ +++ + - t +++ - +++ +++ - + +++ ++

++ - -

++ - +++ - ++ ++ ++ +++ +++

the severity of the symptoms from the patients’ perspective. with symptoms for less than 3 years, a search was made for Utilizing the criteria defined above, 13 of the patients had an occult malignancy, including breast and pelvic examina- extrapyramidal, cerebellar, and autonomic dysfunction, 6 had tions in women, prostate examination in men, acid phospha- extrapyramidal and autonomic, 3 had extrapyramidal and cer- tase and prostate specific antigen levels, stool guaiac tests for ebellar, 1 had cerebellar and autonomic, and 1 had extrapy- occult blood and chest x-rays. In addition, anti- ramidal alone (“possible MSA”) (see Table 2). antibodies were sought in blood samples for patients with The diagnosis of OPCA was based upon a history of pro- ataxia of less than 3 years’ duration. All patients were evalu- gressive deterioration in cerebellar function manifested by ated with MRI to determine the extent of volume loss of the ataxia of gait and speech, usually accompanied by ataxia of structures under study. The diagnosis of sOPCA was made ocular and limb movements, in the absence of a disorder of if there was no family history of a similar illness after a de- sensory function sufficient to cause ataxia, medications (e.g., tailed history had been taken with construction of a family phenytoin), toxins (e.g., ), evidence of a in tree. The diagnosis of dOPCA was based upon a clear family the cerebellum or elsewhere, or evidence of multiple sclero- history of a similar disorder in a parent. Patients with a family sis or other diseases that can cause progressive cerebellar history suggesting autosomal recessive inheritance were ex- ataxia. The diagnosis was assisted by the demonstration of cluded from this study. cerebellar and brainstem atrophy in computed tomographic The neurological examinations were conducted by a neu- (CT) or MRI scans, but this was not required since OPCA rologist (S.G.)blinded to the PET data. Speech was evaluated can occur in the absence of atrophy in anatomic imaging by a speech-language pathologist (K.J.K.) also blinded to the studies [24}. results of the PET studies. The severity of the clinical neuro- Each patient was evaluated with a complete history, physi- logical features in the patients with MSA was graded on the cal examination, neurological examination, laboratory tests to scales shown in Table 2. exclude other diseases, and MRI imaging to exclude demyeli- native disease and other structural abnormalities. Speech was Positron Emission Tomography Studies evaluated as described previously 126, 321. Laboratory tests These studies were performed with the normal control sub- included a complete blood count, serum profiles of hepatic jects and the patients lying supine and awake in a quiet room and renal function, brainstem auditory, visual, and somato- with eyes open from 5 minutes before injection until comple- sensory evoked potentials, serum levels of vitamins E and tion of the scan. To ensure that similar levels and angles of B,, and folic acid, and studies of thyroid function. [n patients scanning were obtained, all subjects were aligned parallel to

Gilman et al: Cerebral Metabolism in MSA and OPCA 169 the canthomeatal (CM) line with laser beams. Scans were completion of measurements from all studies. Absolute and performed 30 to 90 minutes after intravenous injection of normalized regional glucose cerebral metabolic rates were 10 mCi of FDG, which was synthesized by an adaptation of analyzed using multivariate repeated measures analysis of co- the method of Hamacher and colleagues 1331. variance and univariate analysis of covariance (ANCOVA). PET images were acquired with a SiemensiCTI 93 1/08-12 Absolute metabolic rates were analyzed from the following scanner, which has an intrinsic in-plane resolution of 5.5-mm seven brain regions: brainstem, cerebellar vermis, cerebellar full-width-at-half-maximum (FWHM) and a z-axis resolution hemispheres, caudate nuclei, putamen, thalamus, and cere- of 7.0-mm FWHM. Fifteen planes with 6.75-mm center-to- bral cortex. Normalized metabolic rates were analyzed for center separation were imaged simultaneously. Two sets of the first six of these regions. interleaved planes were acquired, yielding 30 planes, each In the multivariate analysis, the dependent variable was separated by 3.375 mm. Attenuation correction was calcu- regional brain metabolic activity (either absolute or normal- lated by fitting ellipses to the scalp contour. ized). Measurements on seven (for absolute data) or six (for Blood samples were collected from the radial artery for normalized data) brain regions constituted the repeated mea- estimation of the arterial radioactivity input to brain. Local surements on each patient. The primary independent variable cerebral metabolic rates for glucose (1CMRglc) were calcu- in the analysis was the categorical variable diagnosis group lated with a three-compartment model and a single scan ap- (MSA, sOPCA, dOPCA, or normal control). In addition, the proximation [34}. Volumes-of-interest (VOls) were studied factor sex and the continuous covariate age were included in in the brainstem, cerebellar vermis, cerebellar hemispheres, the model. All possible interactions between diagnosis group, thalamus, basal ganglia (caudate and putamen), and sex, and age were considered. Age was standardized by sub- cerebral cortex. Levels comprising each VOI were contigu- tracting the mean age across all subjects. ous. PET images were viewed in the transverse and sagittal The data were analyzed in several stages. In the first stage, planes. Data were collected from transverse sections with a not presented here, plots of the data were used to assess 2.6-cm2 polygon over each from 4 of gross trends and relationships between variables. Second, a the 30 levels, a 1.3-cmZ parallelogram over each thalamus repeated measures ANOVA model was fitted to the data from 2 of the 30 levels, a 0.9-cm2 square over each caudate incorporating region as the repeated (within-subject) mea- nucleus from 2 of the 30 levels, and a 1.9-cm2 parallelogram sures variable, diagnosis group, sex, and age, and all of the over each putamen from 2 of the 30 levels. Data were col- interactions among these variables to determine whether dif- lected in the midsagittal plane from the pons with a 1.3 x ferences existed between brain regions. Dependent upon the 1.5 x 1. I-cm right parallelepiped and from the superior and outcome of this stage, a univariate ANOVA or ANCOVA inferior parts of the cerebellar vermis with a 1.3 X 1.9 X model was developed to examine each region separately. The 1.1-cm right parallelepiped. Each VOI was automatically cen- factors incorporated into the univariate ANOVA model tered over a local peak in 1CMRglc. VOIs from the cerebellar were those identified in the second stage as having an overall vermis were posterior to the fourth ventricle. The brainstem effect on glucose metabolic rate across regions or having a VOI chiefly reflects the pons, but the mesencephalon could differential effect on glucose metabolic rates between re- be partially represented. gions. In the last stage, comparisons of individual diagnosis Data from the cerebral cortex were obtained in transverse group means were made, where appropriate. images by measuring ICMRglc in the cortical ribbon from 8 For all pairwise differences, a Tukey-Kramer multiple or 10 consecutive planes of 30, beginning with the plane comparison adjustment 1351 was used. Hence, the reported containing the most superior portion of the cingulate gyrus p values incorporate this adjustment. When the final model and continuing inferiorly to the plane containing the lowest included age, diagnosis group means were appropriately ad- portion of the thalamus. This was accomplished with an algo- justed for age. Data analysis was performed with the SAS rithm that detects the outer edge of the cortical rim from an statistical package (SAS Institute Inc., Cary, NC). image that has been passed through a contrast-enhancing fil- ter. The algorithm then identifies on the original image a Results cortical band that extends inward from this edge until either Analysis of Ahofute hefs ICMRgfc the metabolic rate drops below the value on the outer edge of of the rim or the band reaches a width of 15 mm. An area- Repeated measures ANCOVA revealed highly sig- weighted mean metabolic rate was computed for the mean nificant differences between brain regions (p < cortical value. Normalized values were obtained by dividing 0.0001) and strong evidence for differences in the ef- individual VOI values by the mean value from the cerebral fects of diagnosis group between regions (p < 0.0001). cortex over all slices. Data were normalized to determine Evidence was found also for differences in region by whether changes in absolute values reflect the effects of the age (p < 0.005). No evidence was adduced for differ- diseases upon specific structures or upon the brain as a whole. ences in region by sex, nor for differences between The cerebral cortex was selected for normalizing the data regions in interaction effects between age, sex, or diag- because it is a large structure not known previously to be nosis group. Examination of the overall effects of diag- involved in these disorders. nosis, age, and sex (looking across all regions simulta- neously) indicated large differences between diagnosis Data Analysis groups (p < 0.0001) but no other significant effects. A data analyst (M.L.) analyzed VOIs in the PET studies and The significant effects found between regions, and these data were shared among the investigators only after differences in the effects of diagnosis groups, indicate

170 Annals of Neurology Vol 36 No 2 August 1994 that differences in metabolic rates between diagnosis Table 3. Absolute Local Cerebral Metabolic Rates for Glucase groups must be examined for each region individually. Normal The evidence for a difference in age effect by region Controls dOPCA sOPCA MSA requires that this variable be incorporated in the uni- Structure (n = 51) (n = 17) (n = 32) (n = 24) variate analysis as a covariate. Brainstem 5.98 4.74" 4.50" 4.08" Separate univariate ANCOVA analyses for each re- Cerebellar 6.25 4.77" 4.42" 4.37" gion (including the factors diagnosis and age) revealed vermis that age effects were significant in the caudate nucleus Cerebellar 6.71 5.10" 4.76" 4.24" (p = 0.0003), putamen (p = 0.021), and cerebral cor- hemispheres tex (p = 0.0008). In the thalamus there was some Thalamus 8.67 8.27ns 7.81d 7.39' indication of a trend with age (p = 0.06). No evidence Caudate nuclei 8.00 7.56"s 7.36"s 7.WnS was found in any of the remaining regions for an age Putamen 8.69 8.3OnS 7.71' 7.02" effect (p > 0.10 in each). To provide a common basis Cerebral cortex 6.88 6.64"s 6.26d 6.00' for comparison, however, subsequent analysis of abso- lute metabolic rates by diagnosis group incorporated Values are mean absolute glucose metabolic rates adjusted for age effects. Significant differences in comparison with normal controls via an adjustment for age. pairwise comparisons with a Tukey-Gamer adjustment for multiple Very strong evidence was found for overall differ- comparisons: 'p < 0.0001, 'p < 0.01, 'p < 0.05, "' = not significant. ences in diagnosis groups in the brainstem, cerebellar dOPCA = dominantly inherited olivopontocerebellar atrophy; sOPCA = sporadic olivopontocerebellar atrophy; MSA = multiple vermis, cerebellar hemispheres, and putamen (p < system atrophy. 0.0001 in each). Strong evidence for overall differ- ences in diagnosis groups was also found in the thala- mus (p = 0.0016) and cerebral cortex (p = 0.004). tween diagnosis groups (p < 0.0001) and evidence for Evidence for a difference in caudate nucleus was also age effects (p< 0.001) but no other significant effects. found (p = 0.041). To investigate these differences The significant effects found between regions and further, pairwise comparisons among mean ICMRglc differences in the effects of diagnosis groups indicated values were examined. that differences in metabolic rates between diagnosis After controlling for multiple comparisons, we groups must be examined for each region individually found that in MSA, absolute lCMRglc was markedly via a one-way ANOVA on diagnosis group. The evi- and significantly decreased relative to normals in the dence of differences in age effects by region and an brainstem, cerebellar vermis, cerebellar hemispheres, overall age effect require that age be incorporated in and putamen (p < 0.0001 in each) and also in the the univariate analysis with an ANCOVA by region. thalamus and cerebral cortex (p< 0.05 in each) (Table Separate univariate analysis for each region (includ- 3, Fig). In sOPCA, lCMRglc was decreased in the ing the factor diagnosis and the covariate age) demon- brainstem, cerebellar vermis and cerebellar hemi- strated that age effects were significant in the brainstem spheres (p< 0.0001 in each), and putamen (p< O.OI), (p = 0.012), cerebellar vermis (p = 0.005),cerebellar but somewhat less in the cerebral cortex and thalamus hemispheres (p = 0.007'), and thalamus (p = 0.017). (p< 0.05). In dOPCA, lCMRglc was greatly decreased In the caudate nucleus there was some indication of a in the brainstem, cerebellar vermis, and cerebellar trend with age (p = 0.06), but no evidence for an age hemispheres (p < 0.0001 in each), but no other re- effect was found in the putamen (p > 0.10). To pro- gions showed significant decreases. Statistically signifi- vide a common basis for comparison, subsequent anal- cant differences between MSA and dOPCA were yses of normalized metabolic rates by diagnosis group found in the putamen (p = 0.02) only. included an adjustment for age in all regions. Very strong evidence was found for overall differ- Analysis of lCMRglc Normalized ences in diagnosis groups in brainstem, cerebellar ver- to the Cerebral Cortex mis, and cerebellar hemispheres (p < 0.0001 in each). Repeated measures ANCOVA revealed highly sig- Strong evidence for overall differences in diagnosis nificant differences between brain regions (p < groups was also found in the putamen (p = 0.01). 0.0001) and strong evidence for differences in the ef- No evidence was found for differences in the caudate fects of diagnosis group between regions (p< 0.0001). nucleus or thalamus (p > 0.10). To investigate these Evidence was found also for differences in region by differences further, pairwise comparisons among mean age (p < 0.02). No evidence was adduced for differ- lCMRglc values were examined. ences in the effects of region by sex, nor for differences After controlling for multiple comparisons, we between regions in interaction effects between age, found that in MSA, normalized lCMRglc was markedly sex, or diagnosis group. Examination of the overall ef- and significantly decreased relative to normals in the fects of diagnosis, age, and sex (looking across all re- brainstem, cerebellar vermis, and cerebellar hemi- gions simultaneously) indicated strong differences be- spheres (p < 0.0001 in each) and also in the putamen

Gilman et al: Cerebral Metabolism in MSA and OPCA 171 Positron emission tomography scans showing local cerebral meta- bolic rate for glucose ClCMRglc) as detected with ['8F}f2tloro- deoxyglucose in a mule normal control subject age 52 (left upper (p < 0.01) (Table 4). In both sOPCA and dOPCA, and lower images), a male patient age 61 with multiple system lCMRglc was decreased in the brainstem, cerebellar atrophy (MSA, upper and lower images second from the vermis, and cerebellar hemispheres (p < 0.0001 in left), a male patient age 46 with sporadic olivopontocerebellar at- each). No statistically significant differences between rophy (sOPCA, upper and lower images third from the Ieft), MSA and dOPCA were found. and a male patient age 37 with dominant4 inherited olivopon- tocerebellar atrophy (dOPCA, right upper and lower images). Neuropathological Findings All scans in the upper row shouJ horizontal sections at the level Four of the 14 patients with MSA in the present study of the cerebral cortex, basal ganglia, and thalamus, and all have died and come to neuropathological examination. scans in the lower row show horizontal sections at the level of The diagnosis of MSA was confirmed in 3 of these the cerebellum and the base ofthe temporal and frontal lobes. patients by the marked degenerative changes, including The color bar indicates the rate of cerebral glucose utilization of for all scans illustrated expressed as milligrams per 100 grams loss neurons and gliosis, in the inferior olives, pons, per minute extending from 0.0 to 10.5. Note the decreased cerebellum, substantia nigra, and posterior lateral as- lCMRglc in the brainstem and cerebellum of all 3 patients in pects of the putamen. There was no involvement of comparison with the normal control subject. There is decreased the anterior thalamus, caudate nucleus, hypothalamus, lCMRglc also in the cerebral cortex and basal ganglia of the pa- or the frontal, parietal, temporal, and occipital regions tients with MSA and sOPCA but not dOPCA. of the cerebral cortex. The fourth patient survived car- diac arrest and lingered in a vegetative state for 4 weeks before succumbing. The preexistent neuropath- ological changes in the forebrain could not be estab-

172 Annals of Neurology Vol 36 No 2 August 1994 Table 4. Local Cerebral Metabolic Rates for Glucose Normalized rebral cortex, all three groups of patients showed a to the Cerebral Cortex marked and significant decrease in the brainstem and Normal cerebellum. In MSA patients a mild but significant de- Conrrols dOPCA sOPCA MSA crease was found also in the putamen. In sOPCA and Structure (n = 51) (n = 19) (n = 32) (n = 24) dOPCA patients, no other regions showed significant changes. These findings reflect the severe involvement Brainstem 0.86 0.71" 0.73a 0.68" of the brainstem and cerebellum in all three disorders Cerebellar 0.91 0.72" 0.71" 0.73" vermis and the greater involvement of the putamen in MSA Cerebellar 0.98 0.76" 0.77" 0.71" than in the other two entities. None of the other re- hemispheres gions that were significantly affected in the analysis of Thalamus 1.26 1.24"s 1.26ns 1.24"= absolute data in sOPCA and MSA groups showed sig- Caudate nuclei 1.16 1.13"s 1.1SnS 1.1SnS nificant changes in the normalized data because these Putamen 1.26 1.25"s 1.24"s 1.17' regions were all mildly affected to about the same degree. Values are mean cerebral glucose metabolic rates normalized to cere- The finding of markedly reduced 1CMRglc in the bral cortex and adjusted for age effects. Significant differences in comparison to normal controls via pairwise comparisons with Tu- putamen as well as the brainstem and cerebellum in key-Gamer adjustment for multiple comparisons: 'p < 0.0001, MSA patients is in keeping with the neuropathological 'p < 0.01, "' = not significant. changes in this disease [4, 8, 9, 12, 14, 153 and, to dOPCA = dominantly inherited olivopontocerebellar atrophy; sOPCA = sporadic olivopontocerebellar atrophy; MSA = multiple some extent, with previous PET studies 136-381. De- system atrophy. generative changes in MSA are prominent in the puta- men, substantia nigra, and locus ceruleus and also com- monly affect the inferior olives, pons, and cerebellum lished because of extensive hypoxic-ischemic tissue 12, 4, 81. Thus, hypometabolism in the putamen, as damage, but degenerative and atrophic changes were well as the brainstem and cerebellum, reflects the loss found in the inferior olives, pons, and cerebellum. of neurons and synaptic connections in these sites. Many other sites can be affected neuropathologically Discussion in MSA, including the cerebral cortex, caudate nucleus, We found in all three groups of patients, MSA, sOPCA and globus pallidus (especially the external portion), and dOPCA, a marked and significant decrease of ab- but the thalamus is only rarely involved. Thalamic solute lCMRglc in the brainstem and cerebellar vermis involvement when present is probably due to deaffer- and hemispheres. MSA patients showed additionally a entation secondary to degeneration of striatothalamic marked and significant decrease in the putamen, and a projections, which is evident in the mild but significant decrease in the thalamus and cere- and fasciculus lenticularis. The finding of normal glu- bral cortex, but no significant change in the caudate cose metabolic rates in the caudate nucleus of MSA nucleus. Patients with sOPCA showed, in addition, a patients in the present study indicates that this site was mild but significant decrease in the putamen, thalamus, not appreciably affected in our cases, but the hypome- and cerebral cortex and no change in the caudate nu- tabolism found in the cerebral cortex suggests that this cleus. Patients with dOPCA showed no significant dif- site is either directly affected or shows the effects of ferences in other structures, including the thalamus, deafferentation. The finding of decreased glucose met- putamen, caudate nucleus, and cerebral cortex. Thus, abolic rates in the cerebral cortex suggests that PET distinctive regional patterns of decreased ICMRglc are studies with FDG can demonstrate disturbances not found in all three disorders, with more similarity be- detected with conventional neuropathological tech- tween sOPCA and MSA than between dOPCA and niques. either MSA or sOPCA. The finding of significantly reduced ICMRglc in the Comparison of 1CMRglc in the putamen among the brainstem and cerebellum in both sOPCA and dOPCA three groups of diseases disclosed (1) a large and sig- is in keeping with the primary neuropathological pro- nificant reduction in MSA, (2) a significant reduction cess in OPCA, which involves neuronal loss and gliosis in sOPCA but not as large as that seen in MSA, and (3) within the inferior olives, pons, and cerebellum {2, no significant change in dOPCA. Since patients with 19-23]. The decreased metabolic rates in these struc- sOPCA can progress to develop MSA, it is possible tures reflect the reduced numbers of neurons and syn- that the moderate reduction of lCMRglc in the puta- aptic connections, though partial volume effects from men in these patients may represent the development tissue atrophy may be important. Although many other of neuropathological changes typical of MSA, although central nervous system structures can show neuropath- clinical signs diagnostic of MSA were not detectable at ological findings in OPCA [2, 19-23}, the finding of the time of study. reduced ICMRglc so widely throughout the brain in When the ICMRglc data were normalized to the ce- patients with sOPCA in this study, including the basal

Gilman et al: Cerebral Metabolism in MSA and OPCA 173 ganglia, thalamus, and cerebral cortex, is unexpected involvement in extrapyramidal and cerebellar disorders. Clin but may suggest that (1) this disorder affects many Auton Res 1991;1:147-155 more structures in the central nervous system than 6. Izume K, Inoue N, Shirabe T, et al. Failed levodopa therapy in striato-nigral degeneration. Lancet 197 1;1:1955 conventional neuropathological studies have disclosed, 7. Rajput AH, Kazi KH, Rotdilsky B. Striatonigral degeneration (2) the sites showing primary degenerative changes in- response to levodopa therapy. J Neurol Sci 1972;16:331-341 fluence other structures through transneuronal effects, 8. Kume A, Takahashi A, Hashizume Y, Asai J. A histometrical or (3) these patients are beginning to develop MSA. and comparative study on Purkinje cell loss and olivary nucleus Several investigators have examined ICMRglc in cell loss in multiple system atrophy. J Neurol Sci 1991;lOl: small numbers of MSA patients. De Volder and col- 178-186 9. Takei Y, Mirra SS. Striatonigral degeneration: a form of multi- leagues 1361 studied 7 patients with probable SND ple system atrophy with clinical Parkinsonism. In: Zimmerman compared with 16 neurologically normal control sub- HM, ed. Progress in neuroparhology. New York: Grune and jects by using FDG and PET. All 7 patients had extra- Stratton, 1973:217-251 pyramidal symptoms poorly responsive or unrespon- 10. Borir A, Rubinstein LJ, Urich H. The striatonigral degenera- tions: putaminal pigments and nosology. Brain 1975;98: 101- sive to levodopa, and 2 of the 7 had signs of cerebellar 112 disorder accompanied by autonomic disturbances. Glu- 11. Gibb WRG. Accuracy in the clinical diagnosis of parkinsonian cose utilization was significantly decreased in the pa- syndromes. Postgrad Med J 1988;64:345-35 1 tients with SND in the putamen, caudate nucleus, and 12. Fearnley JM, Lees AJ. Striatonigral degeneration. A clinicopath- frontal regions of the cerebral cortex but not in the ological study. Brain 1990;113:1823-1842 13. Bannister R, Oppenheimer DR. Degenerative diseases of the thalamus or cerebellum. In the 2 patients with cerebel- nervous system associated with autonomic failure. Brain 1972; lar symptoms, however, hypometabolism was found 95:457-474 within the cerebellum. Moreover, a third patient 14. Sung JH, Masrri AR, Segal E. of Shy-Drager syn- scanned twice showed clinical signs of cerebellar disor- drome. J Neuropathol Exp Neurol 1979;38:353-368 der at the time of the second scan, and only this scan 15. Spokes EGS, Bannister R, Oppenheimer DR. Multiple system atrophy with autonomic failure. J Neurol Sci 1979;43:59-82 demonstrated cerebellar hypometabolism. Our pa- 16. Petito CK, Hart MN, Porro RS, Earle KM. Ultrastructural stud- tients differed from those of De Volder and colleagues ies of olivopontocerebellar atrophy. J Neuropathol Exp Neurol [36] in that we had more who had involvement of the 1973;32:503-522 cerebellum. Fulham and associates 137) studied 8 MSA 17. Landis DMD, Rosenberg RN, Landis SC, et al. Olivopontocere- patients in comparison with 18 normal control subjects bellar degeneration. Clinical and ultrastructural abnormalities. Arch Neurol 1974;31:295-307 by using FDG and PET and found hypometabolism in 18. Chokrovercy S. Autonomic dysfunction in olivopontocerebellar the cerebellar vermis, cerebellar hemispheres, stria- atrophy. In: Duvoisin RC, Plaitakis A, eds. The olivopontocere- turn, and frontal cortex. Four of their 8 patients had bellar arrophies. New York: Raven, 1984: 105-141 prominent signs of cerebellar disorder. Their findings 19. Koeppen AH, Baron KD. The neuropathology of olivoponto- are similar to those in our MSA patients. Eidelberg cerebellar atrophy. In: Duvoisin RC, Plaitakis A, eds. The olivo- ponrocerebellar atrophies. New York: Raven, 1984; 13-38 and co-workers 1381 reported finding hypometabolism 20. Gilman S, Bloedel JR, Lechtenberg R. Disorders of the cerebel- in the caudate nucleus and putamen of 10 patients lum. Philadelphia: Davis, 198 1 with striatonigral degeneration in comparison with 10 21. Harding AE. The hereditary araxias and related disorders. Lon- normal control subjects and 20 patients with Parkin- don: Churchill Livingsrone, 1384 son’s disease by using FDG and PET. They found no 22. Eadie MJ. Olivo-ponto-cerebellar atrophy (Dejerine-Thomas type). In: Vinken PJ, Bruyn GW, eds. Handbook of clinical metabolic changes in other cortical or subcortical re- neurology. Amsterdam: North-Holland, 1975:4 15-43 1 gions. 23. Eadie MJ. Olivo-ponto-cerebellar atrophy (Menzel type). In: Vinken PJ, Bruyn GW, eds. Handbook of clinical neurology. This study was supported in part by grants NS 15655 and AGO867 1 Amsterdam: North-Holland, 1975:433-449 from the National Institutes of Health. 24. Gilman S. Markel DS, Koeppe RA, et al. Cerebellar and brain- stem hypometabolism in olivopontocerebellar atrophy detected with positron emission tomography. Ann Neurol 1988;24:223- References 230 1. Polinsky RJ. Multiple system atrophy. Clinical aspects, pdtho- 25. Rosenthal G, Gilman S, Koeppe RA, et al. Motor dysfunction physiology and treatment. Neurol Clin 1984;2:487-498 in olivopontocerebellar atrophy is related to cerebral metabolic 2. Oppenheimer SE. Diseases of the basal ganglia, cerebellum and rate studied with positron emission tomography. Ann Neurol motor neurons. In: Humes AJ. Corsellis JAN, Duchen LW, 1988;24:414-419 eds. Greenfield’s neuropathology. 4th ed. New York: Wiley, 26. Kluin KJ, Gilman S, Markel DS, et al. Speech disorders in 1984:699-747 olivopontocerebellar atrophy correlate with positron emission 3. Cohen J, Low P, Fealey R, et al. Somatic and autonomic function romography findings. Ann Neurol 1988;23:547-554 in progressive autonomic failure and mulriple system atrophy. 27. Gilman S, Koeppe RA, Junck L, et al. Extensive cerebral hypo- Ann Neurol 1987;22:692-699 merabolism in multiple system atrophy detected with [18F3flu- 4. Quinn N. Multiple system atrophy-the nature of the beast. orodeoxyglucose and positron emission tomography. Neurology J Neurol Neurosug Psychiatry 1989;(spec suppl):78-89 1992;42(suppl 31:398-399 5. Sandroni P, Ahlskog JE, Fealey RD, Low PA. Autonomic 28. Savoiardo M, Strada L, Girotti F, et al. MR imaging in progres-

174 Annals of Neurology Vol 36 No 2 August 1094 sive supranuclear palsy and Shy-Drager syndrome. J Comput 34. Hutchins GD, Holden JE, Koeppe RA, et al. Alternative ap- Assist Tomogr 1989;13:550-560 proach of single-scan estimation of cerebral glucose metabolic 29. Savoiardo M, Strada L, Girotti F, et al. Olivopontocerebellar rate using glucose analogs, with particular application to isch- atrophy: MR diagnosis and relationship to multisystem atrophy. emia. J Cereb Blood Flow Metab 1984;4:35-40 Radiology 1990;174:693-696 35. Hochberg Y,Tamhane AC. Multiple comparison procedures. 30. Brooks DJ. Reply to the diagnosis of multiple system atrophy New York Wiley, 198791-92 by T Caraceni. Ann Neurol 1991;29:690 (Letter) 36. De Volder AG, Francart J, Laterre C, et al. Decreased glucose 31. McLeod JG, Tuck RR. Disorders of the autonomic nervous utilization in the and frontal lobe in probable striatoni- system: Part 2. Investigation and treatment. Ann Neurol 1987; gral degeneration. Ann Neurol 1989;26:239-247 211519-529 37. Fulham MJ,Dubinsky RM, Polinsky RJ, et al. Computed to- 32. Kluin KJ, Foster NL, Berent S, Gilman S. Perceptual analysis mography, magnetic resonance imaging and positron emission of speech in progressive supranuclear palsy. Neurology 1993; tomography with [’8F]fluorodeoxyglucosein multiple system at- 43:563-566 rophy and pure autonomic failure. Clin Auton Res 1991;l: 33. Hamacher K, Coenen HH, Stocklin G. Efficiency stereospe- 27-36 cific synthesis of NCA 2-{’8F]-fluoro-2-deoxy-~-glucoseusing 38. Eidelberg D, Takikawa S, Moeller JR, et al. Striatal hypometab- aminopolyether supported direct nucleophilic substitution. J olism distinguishes striatonigral degeneration from Parkinson’s Nucl Med 1986;27:235-238 disease. Ann Neurol 1993;33:518-527

Gilman et al: Cerebral Metabolism in MSA and OPCA 175