Lack of Association Between Depression and Loss of Neurons in the Locus Coeruleus in Alzheimer Disease | Dementia and Cognitive Impairment | JAMA Psychiatry | JAMA Network
[Skip to Navigation]
Sign In
Table 1. 
Patient Characteristics*
Patient Characteristics*
Table 2. 
Group Characteristics
Group Characteristics
Original Article
January 1999

Lack of Association Between Depression and Loss of Neurons in the Locus Coeruleus in Alzheimer Disease

Author Affiliations

From the Netherlands Institute for Brain Research (Drs Hoogendijk, Pool, Hofman, and Swaab); Research Institute Neurosciences, Free University, Faculty of Medicine, Department of Psychiatry, Valerius Clinic (Drs Hoogendijk, Sommer, and Eikelenboom); and the Department of Neuropathology, Institute of Pathology (Dr Kamphorst), Amsterdam, the Netherlands.

Arch Gen Psychiatry. 1999;56(1):45-51. doi:10.1001/archpsyc.56.1.45

Background  Depression, one of the most frequent psychiatric disturbances in Alzheimer disease (AD), is proposed to have its neurobiological basis in neuron loss in the noradrenergic locus coeruleus, although this is not the case in idiopathic depression.

Methods  We performed image analyzer–assisted morphometry of the locus coeruleus in 6 depressed, 6 transiently depressed, and 6 nondepressed patients with AD and in 8 control subjects, emphasizing longitudinal psychiatric evaluations and matching for the clinical and neuropathological severity of dementia.

Results  The mean (±SD) number of pigmented neurons in the locus coeruleus in controls (11,607±946) was higher than in patients with AD, regardless of being depressed (5165±928;P=.001), transiently depressed (5647±1163; P=.003), or nondepressed (3717±661; P=.001). No significant difference was found in the number of pigmented neurons between patients with AD who were depressed, transiently depressed, and nondepressed. Patients who had depression at the onset of AD had a higher pigmented neuron number than other patients with AD.

Conclusions  We confirmed the loss of pigmented neurons in the locus coeruleus of patients with AD; however, no supplementary loss of pigmented neurons in the locus coeruleus was found in patients with depression and AD. This finding resembles the situation in idiopathic depression, but is in contrast with earlier studies on depression in AD.

ALZHEIMER DISEASE (AD) is generally considered a disturbance of cognition. Recently, however, depressive and other psychiatric symptoms have gained more interest, since these symptoms may seriously increase caregiver burden and are frequently the reason for hospitalization.1 This is also of practical clinical relevance since depressive symptoms in patients with AD can be successfully treated with selective serotonin reuptake inhibitors.2 These cause little postural hypotension3 and can be used relatively safely in the elderly and for AD, whereas no effective treatment is, at present, available for AD itself.

The neurobiological basis of depression in AD is unclear, while that of idiopathic depression has been related to several monoaminergic systems, including the noradrenergic, system.4,5 Neuropathological changes in the locus coeruleus (LC), which is the major source of central norepinephrine, however, have not been substantiated in patients with depression.6,7 Over the past years, 3 groups hypothesized that depression in dementia8 and especially in AD9,10 may have a neurobiological basis in the LC. Their results suggest that the LC, which is known to be severely affected in AD, is even more affected in patients with depression and dementia.

Many methodological issues in these studies, however, may have given rise to false-positive results. We therefore studied the LC in patients with AD who were depressed, transiently depressed, and nondepressed and in control subjects, emphasizing longitudinal psychiatric evaluation at the symptom level, clinical matching for severity of dementia, and neuropathological matching for the severity of AD, neurological comorbidity, and image analyzer–assisted morphometry of the LC.

Patients and methods

Patients with dementia were studied at 6-month intervals in the framework of a prospective longitudinal study of depression in AD in 8 nursing homes. Informed consent for the interviews was provided by the patients' next of kin, and the patients' consent was provided at the time of each interview. Written, informed consent for brain autopsy was obtained before patients enrolled in the study, as part of the program of the Netherlands Brain Bank. All patients with major neuropathological comorbidity were excluded. The left LC of 22 patients fulfilling National Institute of Neurological and Communicative Disorders and Stroke—Alzheimer's Disease and Related Disorders Association11 and the Consortium to Establish a Registry for Alzheimer Disease criteria12 for definite AD and 8 controls could be obtained. While blind to the depression diagnosis, 4 of these 22 patients were excluded, since Lewy bodies were abundant in substantia nigra or cortex. The remaining 18 patients with AD (Table 1) were divided into 3 groups (depressed, transiently depressed, and nondepressed).

Clinical evaluation

Patients' next of kin and nursing home physicians were interviewed about medical history and the age at onset of AD. Possible and probable AD were diagnosed according to the National Institute of Neurological and Communicative Disorders and Stroke—Alzheimer's Disease and Related Disorders Association11 andDSM-III-R criteria.13 The DSM-III-R criteria (inclusive approach) were also used to diagnose a major depressive episode (except for the absence of an organic factor, ie, AD) or dysthymia (except for the 2-year duration criterion). The severity of the depressive symptoms was determined at 6-month intervals by the Hamilton Depression Rating Scale14 and the Cornell Scale for the Assessment of Depression in Dementia.15 To estimate the influence of symptom overlap, we interviewed our patients by using the standard (inclusive) method,16 and by using the diagnostic–etiologic (exclusive) method17 At 6-month intervals and immediately post mortem the Global Deterioration Scale18 and Functional Assessment Staging19 were used to determine the severity of clinical dementia. The mean (± SD) interval between the last clinical measurement and death was 4.3 ± 0.9 months and did not differ between the 3 AD groups (P =.7).

Six of the remaining 18 patients with AD suffered from a major depressive episode according to DSM-II-R at the time of death, with a minimum duration of 5 months and high scores on the Cornell Scale for the Assessment of Depression in Dementia (>13). This depressed AD group was matched for the duration of clinical dementia (P =.5) with 6 nondepressed AD patients having low scores on the Cornell Scale for the Assessment of Depression in Dementia (<9 (Table 1 and Table 2). Patients with AD and depression were also matched with 6 patients with AD and transient depression, who, in contrast to the depressed AD group, did not have a major depressive episode in the last 4 months of life. The patients with transient depression and AD had a transient major depressive episode or dysthymia during the course of AD (patients 7, 9, 10, and 11) or a history of depression requiring admission prior to AD with depressive symptoms during the course of AD (patients 6 and 12)(Table 1). Moreover, patients with AD and depression, transient depression, or no depression did not differ in severity of clinical dementia, age, age at onset, psychiatric comorbidity, medication used, and severity of AD (Table 1 and Table 2). Patients with AD and transient depression were all female, but this did not influence our data since sex differences in LC neuron counts have not been found in humans.20 Finally, patients with AD were matched with 8 controls without any psychiatric or neurological diseases, except for patient 9, who had psychosis, and patient 21, who had used haloperidol for a transient delirium. Excluding these controls would only have intensified our results. All controls were free of neuropathologic changes, except for small or multiple infarctions. The mean (± SD) age of controls was lower (P =.004) than that of the entire AD group (84.6 ± 1.6 years), but this was not considered a relevant difference, since LC neuron number decreases by only 7% between the ages of 75 and 85 years.21-23 Postmortem delay was longer in controls than in AD, but no difference was found between the 3 AD groups (Table 2).

Neuropathologic findings

After a standard fixation time of 4 weeks, a semiquantitative estimation of the severity of AD changes according to the classification of Braak et al24 was performed by one of us (W.K.). A more differentiated semiquantitative sumscore of neurofibrillary tangles, neuritic plaques, and neuropil disruption was established in a Bodian staining of frontal, temporal, parietal, and occipital cortex and hippocampus. In each area all AD changes were separately scored as 0, absent; 1, absent to moderate; 2, moderate (ie, 2-3 neurofibrillary tangles, 2-3 neuritic plaques or 30%-60% of the normal network replaced by neuropil threads per 0.4 mm2 area); and 3, more than moderate.


The left brainstem, containing the entire LC, was dissected and serial 6-µm paraffin sections were cut transversely. Every 100th section was stained with 0.1% hematoxylin and 0.2% eosin. Cell counts of the pigmented norepinephrine–producing25 neurons and volume measurements of the LC were performed as described in detail elsewhere.26 Briefly, the analyses were performed on an image analysis system (Imago Bild Analyse System; Kontron, Munich, Germany) connected to a television camera on a Zeiss microscope and equipped with a scanning stage; both were controlled by a joystick and the image analyzer. Reliability and calibration studies of the method are described elsewhere.26 Counting neurons at high magnification (×500) was done in an area-weighed, nonselective sample taken by the image analysis system. The scanning stage moved automatically to the positions included in the sample. The total number of pigmented neurons was estimated using an unfolding procedure on the profiles of the pigmented neuron nucleus.26 This procedure determines neuron density, precluding the double counting of neurons, multiplied by the LC volume. Since the rostral and middle part of the LC have projections to mood-related limbic structures, and the caudal part to the spinal cord, separate measurements were performed for the rostral, middle, and caudal part of the LC as defined by Chan-Palay and Asan.25 During morphometry one of us (W.J.G.H.) was blind to clinical findings.

Statistical analysis

For some variables the population variance in the AD and control groups was different. Differences among groups were therefore statistically evaluated by the nonparametric Kruskal-Wallis multiple comparisons test (2 tailed). Relationships between variables were assessed by Spearman rank correlation coefficients. P values less than .05 were considered to be statistically significant. All values are expressed as the mean ± SD.


Patients with AD and depression were notably more depressed on different depression scales during the last measurement before death than patients with AD and either transient or no depression. Patients with AD and transient depression were more depressed than patients with AD and no depression. Patients with depression and AD who had also had a depression at the onset of AD (patients 2, 4, and 5) had the highest scores of all patients on the Hamilton Depression Rating Scale (26.3 ± 0.3), the Cornell Scale for the Assessment of Depression in Dementia (23.0 ± 2.5), and the exclusive Cornell Scale for the Assessment of Depression in Dementia (21.7 ± 2.3) (Table 1 and Table 2).


The intraobserver reliability (100% × SD per mean value) for the measurement of the number of pigmented neurons in the LC (8.2%), determined by measuring the same section 5 times in the course of the study, was good.

The volume of the LC in the control group was larger than in the AD groups, but no difference was found among AD groups (Table 2). The mean total number of pigmented neurons of the LC in patients in the control group was higher than in patients with AD, regardless of their being depressed, transiently depressed, or nondepressed (Table 1 and Table 2). The control group had higher neuron numbers than the AD group in the rostral, middle, and caudal part of the LC, but no differences were found among the 3 AD groups in these areas (Table 2). The mean neuron number in the entire LC in the depressed and transiently depressed AD group was relatively high, but not statistically different from the nondepressed AD group. The 3 patients with AD and depression who had a depression at the onset of AD had a mean pigmented neuron number (6453 ± 1024) that was relatively high, but not statistically different from the other patients with AD (Table 1). Taking all patients with AD together (n = 18), no correlation was found between severity of depression (inclusive Cornell) and neuron number (r = 0.14, P =.57). In addition, no correlation was found between neuron number and the duration of depression (r = −0.38; P =.46).


The number of pigmented neurons in the LC in controls in our study, as well as the severity of neuronal loss in AD, is in agreement with the literature.20,22 Our finding that the total number of pigmented neurons in patients with AD and depression is not different or lower than that of patients with AD and no depression is in contrast with earlier studies.8-10 These studies suggested that the neuronal loss in the noradrenergic LC in AD was even more severe if the patient was also depressed, providing support for the noradrenergic hypothesis of depression and offering a theoretical rationale for noradrenergic antidepressant treatment of patients with AD and depression. In idiopathic depression, no neuron loss in the LC was found6,7; however, in AD 50% to 80% of the neurons are lost.20,26 It is therefore difficult to detect a mild supplementary neuron loss due to depression, on top of the severe neuron loss that is already present in the LC of patients with AD. For the same reason, it is crucial that the patients with AD in the depressed group have the same degree of dementia as patients with AD in the nondepressed group, since a slightly more severe degree of AD in the depressed AD group would go together with a supplementary neuron loss in this group. Therefore, these previous studies8-10 were susceptible to generating false-positive results due to a mismatch between the depressed and nondepressed dementia group regarding the clinically and neuropathologically established severity of AD.

Matching for covariables of severity of dementia is, however, complicated by many methodological issues. First, depression and dementia have some symptoms in common, including loss of interest, decreased energy, difficulty in thinking or concentrating, and psychomotor agitation or retardation.27 To control for symptom overlap we used the Cornell scale, which was specifically developed for the assessment of depression in all stages of dementia. Furthermore, in our 6-month follow-ups we interviewed our patients using not only the standard (inclusive) method, but also the diagnostic–etiologic (exclusive) approach. The patients with high Cornell scores (>9) according to the exclusive approach had pigmented neuron numbers that were even among the highest of the AD group. The exclusive approach showed that an attempt to purify depression symptoms, thus diminishing symptom overlap, intensified our results. In most earlier studies, it was difficult to avoid symptom overlap, since retrospective interviews and medical chart reviews were used10 and the severity of depression was not determined.8,10

The second methodological issue is that it may be difficult to clinically distinguish levels of severity of dementia in the final stages of AD (floor effect).19 Therefore, we obtained Functional Assessment Staging scores every 6 months, and 1 additional Global Deterioration Scale score concerning the last 2 weeks of life was performed immediately post mortem. Since some floor effect of the clinical assessment may still be present, we also matched the depressed and nondepressed AD groups for the severity of neuropathological AD changes, according to Braak et al24 and to a more differentiated semiquantitative analysis of AD changes (Table 1 and Table 2). In 1 of the previous studies,10 the patient groups were not exactly matched for AD severity, which may also have influenced their finding in the raphe nucleus.

Third, groups should be matched for neurological comorbidity.28 To control for neurological comorbidity, we excluded all patients with dementia that showed neuropathological signs of major neurological comorbidity. This was done before the 3 AD groups were composed, thus precluding a selection bias concerning depression. Multiple cerebral infarctions were found in 1 patient with AD and depression and 1 patient with AD and no depression, while 2 more patients with AD and depression only showed 1 small infarction. One patient with AD suffered from alcoholism (Table 1), which may theoretically affect the LC,29 but post hoc exclusion of these subjects would only have intensified our results. One of the earlier studies8 included more patients with neuropathological signs of Parkinson disease in the depressed (n = 4) than in the nondepressed dementia group (n = 3), while a more severe LC neuron loss has been reported in patients with both parkinsonian and dementia signs.30 According to this study28 the increased number of Lewy bodies in the LC, as part of a composite score, shows that the LC is more affected in patients with dementia and depression than in patients with dementia and no depression. It may thus be that the LC in this study8 is more affected by Parkinson disease than by depression. Groups should also be matched for the presence of vascular pathologic condition, since this may attribute to the severity of clinical dementia, although it has never been shown to affect the LC.31 In the nondepressed dementia group of the latter study8 1 patient was included with both AD and multiple infarcts and 3 patients with vascular dementia. The depressed group in this study8 may have been as clinically demented as the nondepressed group, even though having more neuropathologic AD changes. The more severe degree of neuropathologic AD changes in the depressed group may explain the supplementary number of tangles and neuron loss in the LC, rather than depression.

The fourth methodological issue is that the distribution of neurons in the LC is heavily dependent on the level in rostrocaudal direction where counting performed.20 This may result in a large variability in neuron counts if the LC is delineated by hand26 or if the number of sections in the sample is small or restricted to a certain portion of the LC as in most earlier studies.8,9 To avoid biased outlining that might influence the results, we performed image analyzer–assisted morphometry26 and counted the pigmented neurons in a large sample at all 3 levels of the LC (Table 1and Table 2). Moreover, since in our study mean values in patients with AD and depression were not lower than in patients with AD and no depression, it is hazardous to perform 1-tailed statistics, as was done in 1 previous study.9

It is unclear whether the subtle complaints of sadness of a patient with mild dementia signify the same depressive syndrome as the emotional incontinence of a patient with severe dementia. Therefore, we also studied patients with AD and transient depression in our study, who did not have a major depressive episode during the last stage of AD and who showed a mean neuron number that was not different or lower than that of patients with AD and no depression. Therefore, it is unlikely that our discrepant results are caused by a different (more severe) AD group, resulting in a floor effect in our study and precluding the detection of any additional neuron loss associated with depression. Moreover, the loss of LC neurons in AD in this study (59%) was not as extreme as in some other studies.26 Finally, the duration and severity of dementia, the severity of neuropathological changes, and the percentage of LC neuron loss in AD resembled that of the other studies.8-10 Although taking our patients with AD and depression or transient depression together (n = 12) only intensified our results, more studies in larger populations of patients with AD will be needed to confirm the absence of differences in locus coeruleus neuron counts.

In a recent editorial, Esiri32 emphasized the difficulty and importance of postmortem brain research of the aminergic brainstem nuclei in patients with AD, who had behavioral symptoms that had been assessed over a longer period. The most frequent finding in the English-language literature on postmortem brain research in patients with AD and behavioral symptoms has been that the degenerated noradrenergic system in AD is even more affected if the patient is also depressed. In our study we could not confirm this. Even if LC neuron number was additionally decreased in patients with AD and depression, the pathogenetic relation to depression was unclear, since neuron loss in the LC may well be related to some overlapping symptoms of depression and dementia (eg, insomnia), while not being related to depressed mood. Our findings are in line with the normal LC neuron number found in idiopathic depression. In the same patient sample, we also found decreased postmortem norepinephrine concentrations in the cortex and increased activity of the remaining LC neurons, but no difference between patients with AD and depression and patients with AD and no depression (W.J.G.H., Matthijs G.P. Feenstra, Margriet H.A. Botterblom, et al, unpublished data, January 1, 1992-December 1, 1996). These findings confirm our morphometric results in the LC. Our data do not support the hypothesis that the comorbidity of depression in AD has a neuropathological substratum in the LC.

Accepted for publication September 29, 1998.

This study was supported by grant 900557007/716019, from the Netherlands Scientific Council, The Hague.

We thank Olga Pach and Matthijs G. P. Feenstra, MD, for their advice. Brain tissue was obtained from the Netherlands Brain Bank (Rivka Ravid, MD).

Corresponding author: Witte J. G. Hoogendijk, MD, PhD, Valerius Clinic, Valeriusplein 9, 1075 BG, Amsterdam, the Netherlands (e-mail:

Steele  CRovner  BChase  GAFolstein  M Psychiatric symptoms and nursing home placement of patients with Alzheimer's disease.  Am J Psychiatry. 1990;1471049- 1051Google Scholar
Nyth  ALGottfries  CGLyby  KSmedegaard-Anderson  LGylding-Sabroe  JKristensen  MRefsum  HEOfsti  EEriksson  SSyverson  S A controlled multicenter clinical study of citalopram and placebo in elderly depressed patients with and without concomitant dementia.  Acta Psychiatr Scand. 1992;86138- 145Google ScholarCrossref
Naranjo  CAHerrmann  NMittmann  NBremner  KE Recent advances in geriatric psychopharmacology.  Drugs Aging. 1995;7184- 202Google ScholarCrossref
Schildkraut  JJSchanberg  SMBreese  GRKopin  IJ Norepinephrine metabolism and drugs used in the affective disorders: a possible mechanism of action.  Am J Psychiatry. 1967;124600- 608Google Scholar
Van Praag  HM Depression.  Lancet. 1982;21259- 1264Google ScholarCrossref
Hankoff  LDPeress  NS Neuropathology of the brain stem in psychiatric disorders.  Biol Psychiatry. 1981;16945- 952Google Scholar
Klimek  VStockmeier  COverholser  JMeltzer  HYKalka  SDilley  GOrdway  GA Reduced levels of norepinephrine transporters in the locus coeruleus in major depression.  J Neurosci. 1997;178451- 8458Google Scholar
Zubenko  GSMoossy  J Major depression in primary dementia: clinical and neuropathological correlates.  Arch Neurol. 1988;451182- 1186Google ScholarCrossref
Förstl  HBums  ALuthert  PCairns  NLantos  PLevy  R Clinical and neuropathological correlates of depression in Alzheimer's disease.  Psychol Med. 1992;22877- 884Google ScholarCrossref
Zweig  RMRoss  CAHedreen  JCSteele  CCardillo  JEWhitehouse  PJFolstein  MFPrice  DL The neuropathology of aminergic nuclei in Alzheimer's disease.  Ann Neurol. 1988;24233- 242Google ScholarCrossref
McKhann  GDrachman  DFoistein  MKatzman  RPrice  DStadlan  EM Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease.  Neurology. 1984;34939- 944Google ScholarCrossref
Mirra  SSHeyman  AMcKeel  DSumi  SMCrain  BJBrownlee  LM Vogel FS, Hughes JP, Van Belle G, Berg L.  The Consortium to Establish a Registry for Alzheimer's Disease (CERAD), II: standardization of the neuropathologic assessment of Alzheimer's disease. Neurology 1991;41479- 486Google Scholar
American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition.  Washington, DC American Psychiatric Association1987;
Hamilton  M Development of a rating scale for primary depressive illness.  Br J Soc Clin Psychol. 1967;6278- 296Google ScholarCrossref
Alexopoulos  GSAbrams  RCYoung  RCShamoian  CA Cornell Scale for Depression in Dementia.  Biol Psychiatry. 1988;23271- 284Google ScholarCrossref
Claman  DLRadebaugh  TS Neuropsychological assessment in clinical trials of Alzheimer's disease.  Alzheimer Dis Assoc Disord. 1991;5(suppl 1)S49- S56Google ScholarCrossref
Alexopoulos  GS The treatment of depressed demented patients.  J Clin Psychiatry. 1996;57(suppl 14)14- 20Google Scholar
Reisberg  BFerris  SHDe Leon  MJCrook  T Global Deterioration Scale (GDS).  Psychopharmacol Bull. 1988;24661- 663Google Scholar
Reisberg  B Functional assessment staging (FAST).  Psychopharmacol Bull. 1988;24653- 659Google Scholar
German  DCManaye  KFWhite  CLWoodward  DJMcIntire  DDSmith  WKKalaria  RNMann  DM Disease-specific patterns of locus coeruleus cell loss.  Ann Neurol. 1992;32667- 676Google ScholarCrossref
Lohr  JBJeste  DV Locus ceruleus morphometry in aging and schizophrenia.  Acta Psychiatr Scand. 1988;77689- 697Google ScholarCrossref
Vijayashankar  NBrody  H A quantitative study of the pigmented neurons in the nuclei locus coeruleus and subcoeruleus in man as related to aging.  J Neuropathol Exp Neurol. 1979;38490- 497Google ScholarCrossref
Mouton  PRPakkenberg  BGundersen  HJPrice  DL Absolute number and size of pigmented locus coeruleus neurons in young and aged individuals.  J Chem Neuroanat. 1994;7185- 190Google ScholarCrossref
Braak  HBraak  EYilmazer  DVos  RAJansen  ENBohl  J Pattern of brain destruction in Parkinson's and Alzheimer's disease.  J Neural Transm. 1996;103455- 490Google ScholarCrossref
Chan-Palay  VAsan  E Quantitation of catecholamine neurons in the locus coeruleus in human brains of normal young and older adults and in depression.  J Comp Neurol. 1989;287357- 372Google ScholarCrossref
Hoogendijk  WJGPool  ChWTroost  DVan Zwieten  ESwaab  DF Image analyser-assisted morphometry of the locus coeruleus in Alzheimer's disease, Parkinson's disease and amyotrophic lateral sclerosis.  Brain. 1995;118131- 143Google ScholarCrossref
Burke  WJRubin  EHMorris  JCBerg  L Symptoms of "depression" in dementia of the Alzheimer type.  Alzheimer Dis Assoc Disord. 1988;2356- 362Google ScholarCrossref
Jellinger  K Neuropathological substrates of Alzheimer's disease and Parkinson's disease.  J Neural Transm. 1987;(suppl 24)109- 129Google Scholar
Arango  VUnderwood  MDMann  JJ Fewer pigmented locus coeruleus neurons in suicide victims: preliminary results.  Biol Psychiatry. 1996;39112- 120Google ScholarCrossref
Zweig  RMCardillo  JECohen  MGiere  SHedreen  JC The locus ceruleus and dementia in Parkinson's disease.  Neurology. 1993;43986- 991Google ScholarCrossref
Gottfries  CGAdolfsson  RAquilonius  SMCarlsson  AEckernas  SANordberg  AOreland  LSvennerholm  LWiberg  AWinblad  B Biochemical changes in dementia disorders of Alzheimer type (AD/SDAT).  Neurobiol Aging. 1983;4261- 271Google ScholarCrossref
Esiri  MM The basis for behavioural disturbances in dementia.  J Neurol Neurosurg Psychiatry. 1996;61127- 130Google ScholarCrossref