Previous psychiatric history as a risk factor for late-life dementia: a population-based case-control study

Previous psychiatric history as a risk factor for late-life dementia: a population-based case-control study

Brian Cooper

Introduction

The influence of mental illness earlier in adult life on the risk of dementia in old age or its clinical picture is still obscure. Most research on the relation of `functional’ mental disorders to abnormal cognitive decline has focused on their co-existence in old people [1]. In retrospective case-control studies, however, earlier psychiatric illness has been considered as a possible risk factor for dementia, and in particular for that of Alzheimer type. In four such studies a history of depression was found more frequently among cases than among controls [2-5], while two suggested a link with earlier psychiatric morbidity more generally [6, 7]. Some other case-control studies, however, have revealed no evidence of association [8-10].

The low statistical power of most individual studies underlines the need for analysis of pooled data. In a collaborative re-examination of 11 case-control studies, undertaken as part of the EURODEM concerted action programme, a significant association with previous depressive illness was found for Alzheimer-type dementia, the relative risk among cases being 1.82. This finding held good when only those depressive episodes were included which had occurred more than 10 years before the onset of dementia [11]. The fact that all case- control studies included in the EURODEM re-analysis were restricted to cases diagnosed as Alzheimer-type dementia narrowed its scope and could have introduced bias. Modern criteria for this diagnosis require the absence of other causes of dementia, which means that people with a history of organic brain syndrome, alcohol abuse or mental retardation may be excluded. Depressive illness has been targeted precisely because it is thought to be less subject to this kind of bias.

There are, however, some indications that any increase in risk may not be specific to depression. Henderson and co-workers [6] noted a positive association between early-onset Alzheimer-type dementia and a history of `nervous breakdown’ more than 10 years previously, Agbayewa [12], who reported a much higher frequency of previous psychiatric illness among 188 patients with Alzheimer-type dementia than in a matched comparison group of 80 non-demented medical and surgical patients, found that affective disorders (unipolar or bipolar) accounted for only 16 of the 34 cases he identified in his index group, the remainder consisting of schizophrenia, paranoid disorder, alcoholism and mixed conditions. Conversely, some forms of psychiatric illness may be associated with an increased risk for other types of late-life dementia, or for the clinical syndrome of dementia as a whole. The present study was designed to address these questions using an area-based dementia register.

Research design and method

The Camberwell dementia case register, which began operation in 1993, provides a cumulative record of all confirmed cases of dementing illness known to the mental health services covering a defined metropolitan area, with a total population of over 200 000 [13, 14]. Cases are notified by the area clinical teams for old-age psychiatry and geriatric medicine, as well as by geriatric nursing homes, old people’s homes and day centres which provide long-term care for Camberwell residents.

During the first 2 years of the register, all subjects notified to it were contacted, and wherever possible a diagnostic examination was carried out and a structured interview conducted with the next-of-kin or main caregiver, both based mainly on the standardized CAMDEX (Cambridge Examination for Mental Disorders of the Elderly) procedure [15]. The patient examination incorporated a number of rating scales: the Mini-Mental State Examination [16] and Abbreviated Mental Test [17] for cognitive impairment, the Blessed dementia scale [18] for dementia-related disability, the Hachinski scale [19] for probability of cerebrovascular disease, the Cornell scale of depression in dementia [20] and the Webster scale [21] for presence of parkinsonian features.

Non-cognitive psychopathology was grouped according to the system developed by Burns and co-workers [22], the prototype of the recently published `Mousepad’ scale [23]. Disorders of thought content, perception, mood and behaviour were distinguished, and further sub-divided as follows: persecutory and other delusions; auditory and visual hallucinations; depressive symptomatology; aggression; wandering; stereotypical behaviour; and sleep and eating disorders. Each abnormality was rated on a scale from 0 to 5, based on informant responses to a short battery of questions.

A physical examination which included pulse, blood pressure, vision and hearing (with and without aids), mobility, upper and lower limb tonus and tendon reflexes, plantar responses and presence of primitive reflexes. Apolipoprotein E (ApoE) genotyping was carried out using one-stage polymerase chain reaction followed by restriction enzyme digestion [24]. Consent to autopsy was sought in each instance where contact could be made with the next-of-kin.

Applying systematic diagnostic criteria to a case register collective gives rise to a number of problems of method [13]. The standardized procedure adopted for the dementia register in 1993 was based largely on the CAMDEX diagnostic guidelines [15], but supplemented by newly published criteria for dementia of Lewy-body type [25]. For all patients who had been referred to hospital specialist services, a search was made for the medical records and the consultant’s diagnosis noted, as well as the results of any special investigations including computed tomography head scan, full blood count, syphilis serology, liver function tests and concentrations of serum urea, electrolytes, vitamin B12 and folate. In each such case the hospital clinical diagnosis was compared with that made by the standardized dementia register procedure and a `consensus diagnosis’ assigned.

To test for an association between previous psychiatric history and risk for late-life dementia, an analysis was undertaken of all Camberwell residents over 60 years old, notified to the dementia register during its first 2 years (May 1993-April 1995), for whom the diagnosis had been confirmed following detailed clinical assessment. Information on earlier psychiatric illness was derived from three sources:

1. As part of the histories taken from dementia register patients and key informants;

2. From the records of the Bethlem and Maudsley Hospital Trust, the main psychiatric treatment facility for Camberwell residents;

3. From the records of the Camberwell psychiatric case register [26], which covered all mental health agencies serving the area in the years 1964-85.

The dementia register was set up quite independently of the earlier Camberwell psychiatric register and there was no transfer of cases from one to the other.

Two comparative studies were undertaken. First, clinical characteristics of subjects with a previous psychiatric history were individually matched for sex and year of birth with other dementia register patients who had no such history. Secondly, a normal control group was drawn from a list of elderly persons provided by the Family Health Services Authority for the Boroughs of Lambeth, Southwark and Lewisham, which covers all residents registered with general practitioners in the National Health Service. For each dementia register patient, a control was drawn at random from among a group of people matched by sex, completed years of life (from year of birth to 1995 or year of death) and postal district of residence. Information on psychiatric referral and treatment for both index and control groups was limited to that found by searching the hospital and Camberwell register case records. Bias due to the use of informant histories, available only for the index group, was thus avoided. Relative risks were estimated by calculating in each instance the odds ratio (OR) for individually matched pairs [27].

Research findings

The number of persons notified to the dementia register who met the case criteria was 559 (163 men and 396 women), with an age range of 60-97 years. The validity of the generic clinical diagnosis was confirmed by brain microscopy reports in 81 cases which came to autopsy by 30 April 1996 (i.e. in the register’s first 3 years), although a more detailed analysis including these and other cases showed only moderate agreement with regard to specific diagnosis of Alzheimer-type, vascular and other forms of dementia [28].

Of the 559 patients, 85 (15.2%) had previous psychiatric diagnoses, but in 15 instances the earlier illness had merged into cognitive decline, and in retrospect was judged to have marked the prodromal stage of a dementing process. Further analysis was therefore restricted to the remaining 70 cases (15 male and 55 female), which were regarded as `independent’ of dementia, in that the first known psychiatric referral had occurred at least 5 years before signs of cognitive decline were observed, and the two illnesses were clinically distinct.

Hospital and case-register files were the main source of information, yielding details of earlier psychiatric care episodes for 63 (90.0%) of the 70 patients. Informant interviews were obtained for 51 of the patients and provided independent evidence of treated mental illness in 39 instances (55.7%), including seven for whom no hospital records were available. Informants were able to name the year and the hospital for at least one psychiatric admission or referral in 31 instances (44.3%).

For the index group of 70 patients, the mean age at onset of dementia was 75.3 years (s.d. 12.0), as against 75.4 years (s.d. 7.8) for the matched comparison group and 76.4 years (s.d. 8.1) for the dementia register collective as a whole, while the mean interval from first known psychiatric referral to the onset of dementia was estimated at 26.6 years (s.d. 10.2).

The distribution of previous psychiatric illness according to sex and diagnosis is set out in Table 1. Here, each patient has been assigned to a single category, based on the primary diagnosis recorded in the hospital case notes (based according to date of referral on either the 8th or the 9th revision of the International Classification of Diseases). For those with more than one treatment episode, the most recent diagnosis before onset of dementia was taken. Affective disorders, bipolar or monopolar, accounted for 51.4% of the total, schizophrenia and paranoid psychoses for 28.6% and a mixed group of neurosis, personality disorder and other conditions for the remaining 20.0% The frequency of, previous psychiatric illness was higher among women than among men–13.9% as against 9.1%– although this difference is not statistically significant ([chi square] = 2.35; d.f. = 1; P [is greater than] 0.10).

Table 1. Frequency of previous psychiatric illness among dementia register patients, according to sex and primary diagnostic category

Frequency (%)

Men Women All

Primary psychiatric’ diagnosis (n = 163) (n = 396) (n = 559)

Schizophrenia or

paranoid psychosis 2.4 4.0 3.6

Bipolar affective psychosis 0.0 1.5 1.1

Depression (monopolar) 3.7 6.1 5.3

Neurosis or personality

disorder 0.6 0.8 0.7

Chronic alcoholism 1.2 0.8 0.9

Other 1.2 0.8 0.9

All 9.1 13.9 12.5

No diagnosis recorded 90.9 86.1 87.5

Next, we compared the index group of 70 people with other dementia register patients, individually matched for sex and year of birth, who had no known history of earlier mental illness. Table 2 shows the distributions, according to type of dementing disorder, for the two matched groups and also for the whole sample of dementia register patients from which both were drawn. All three distributions are broadly similar. The index group contains a small excess of `secondary’ dementias and relatively fewer cases of Alzheimer-type dementia, but this disparity is accounted for by the fact that four index-group patients, whose earlier psychiatric referrals had been for chronic alcohol abuse, were each given a diagnosis of secondary (alcoholic) dementia, mainly because of their medical history.

Table 2. Distribution according to type of dementia for dementia register patients with previous psychiatric history, matched comparison group and all dementia register patients

No. (and %) of patients

With

previous Comparison All on

history group register

Type of dementia (n =: 70) (n = 70) (n = 559)

Alzheimer type 46 (65.7) 51 (72.9) 395 (70.7)

Vascular or `mixed’ 10 (14.3) 8 (11.4) 97 (17.3)

Lewy body or with 5 (7.1) 6 (8.6) 33 (5.9)

Parkinson’s disease

`Secondary’ dementia 9 (12.9) 5 (7.1) 34 (6.1)

The diagnoses on which Table 2 is based were derived from the standard dementia register examination, together with hospital case records where these were to hand. The fact that results of brain imaging and other special investigations were available for only half the patients may have resulted in some misdiagnosis, and in particular under-diagnosis of vascular and mixed forms of dementia. The proportions for whom hospital records were traced did not, however, differ significantly between the index and matched comparison groups (52.9 and 48.60% respectively), so that systematic error from this cause is unlikely.

The two matched groups were also compared on a range of clinical characteristics. Table 3, which sets out mean scores on the battery of tests and rating scales used in the assessment procedure [13], reveals no significant difference in mean scores on any of these measures. The proportions in long-term care at the time of assessment were similar for the two groups (28.6% of index and 24.3% of comparison group cases), as was their mortality experience over a 3-year period (33 and 34 deaths respectively, up to 30 April 1996). Table 4 gives reported frequencies for a number of psychiatric and behavioural disturbances in 56 matched pairs for whom full information was available, and points to generally high rates for all the abnormalities listed. The interest of the table lies, however, less in the absolute frequencies than in comparison of the matched groups. This reveals only one clear difference: a relative excess of auditory hallucinations among patients with a past psychiatric history, which is highly significant and remains significant at the 5% level (P = 0.04) when the Bonferroni correction for multiple comparisons is applied [29]. Auditory hallucinosis, which could be confirmed from the hospital records for all but one of this group of patients, was associated with a history of schizophrenic or paranoid illness in 10 cases and of severe affective disorder in four.

Table 3. Clinical test and assessment mean scores: dementia register patients with a previous psychiatric history and matched comparison group

With previous history

Clinical test/assessment procedure n Mean score (s.d)

Mini Mental State Examination 60 10.5 (6.8)

Abbreviated Mental Test 53 3.3 (2.6)

Blessed dementia scale 54 8.8 (4.7)

Hachinski scale 46 3.4 (2.7)

Cornell scale for depression in dementia 52 5.2 (4.7)

Webster rating scale (for parkinsonism) 56 8.4 (7.2)

Comparison group

Clinical test/assessment procedure n Mean score (s.d)

Mini Mental State Examination 60 9.4 (7.0)

Abbreviated Mental Test 53 2.9 (2.7)

Blessed dementia scale 54 9.2 (4.4)

Hachinski scale 46 3.2 (2.9)

Cornell scale for depression in dementia 52 4.3 (4.3)

Webster rating scale (for parkinsonism) 56 5.8 (5.3)

Table 4. Frequency of non-cognitive psychopathology and behavioural disturbance amongst Dementia Register patients with a previous psychiatric history and in matched comparison group

Frequency of abnormality

With

previous Comparison

history group

Type of abnormality (n = 56) (n = 56) P value(a)

Auditory hallucinations 14 4 0.00

Visual hallucinations 13 15 0.78

Persecutory delusions 28 22 0.22

Other delusions 5 15 0.27

Depressive symptoms 28 34 0.24

Aggressive behaviour 37 31 0.46

Wandering behaviour 33 28 0.53

Stereotypical behaviour

disorder 42 46 0.34

Sleep disorder 42 38 O.38

Eating disorder 17 31 0.14

(a) Wilcoxon sign-rank test prior to Bonferoni correction.

In addition to these basic comparisons, the relative frequency of different ApoE alleles was examined, since risk for Alzheimer’s disease is known to vary according to ApoE type [30], and disparities between the two groups in this respect could be regarded as a possible confounding factor. Those dementia register patients with a previous psychiatric history who had been typed for ApoE were therefore matched individually by sex and year of birth (to within one year) with others who had no such history. Table 5 shows the allele frequencies for each paired group, together with corresponding distributions for the whole number of dementia register patients who had been typed (n = 263) and for an earlier-reported British population sample [31]. No significant difference was found between the pairs with and without a past psychiatric history, either for the [Epsilon] 4 allele ([X.sup.2] = 0.98; d.f. = 1; P = 0.32) or for the [Epsilon] 2 allele (Fisher exact two-tailed test, P = 0.12). The allele frequencies in both groups of matched cases were similar to those for the dementia register collective from which they were drawn, and appeared to differ to the same extent from those in the external comparison group. Hence the association between earlier psychiatric illness and dementia risk cannot be explained in terms of ApoE genotype.

Table 5. Allele frequencies of apolipoprotein E (ApoE) [Epsilon] 2, [Epsilon] 3 and [Epsilon] 4 in age- and sex-matched dementia register patients with and without a past psychiatric history, dementia register patients as a whole and an external comparison group

Group n Mean age,

years (s.d.)

Dementia register matched-pair group

With previous psychiatric illness 24 82.5 (6.8)

Without previous psychiatric illness 24 82.5 (6.8)

All dementia register cases 263 82.2 (6.6)

External comparison group(a) 77 73.2 (6.1)

ApoE allele frequency

Group [Epsilon] 2 [Epsilon] 3 [Epsilon] 4

Dementia register

matched-pair group

With previous

psychiatric illness 0.00 0.77 0.23

Without previous

psychiatric illness 0.04 0.73 0.23

All dementia register cases 0.04 0.68 0.28

External comparison group(a) O.11 0.76 0.12

(a) Liddell et al. J Med Genet (1994) [31].

In summary, those patients with a previous psychiatric history had a slightly earlier age of onset and tended to have a lower male to female ratio than the dementia register sample as a whole, but did not differ from it in clinical features (with the exception of auditory hallucinosis) or in global severity of dementia, once these differences had been corrected for by the matching procedure.

The second stage of the inquiry involved a direct comparison between the entire dementia register sample and normal population controls. Sixty-three of the 70 register patients with an `independent’ psychiatric history could be identified from the available records alone, without informant interviews, and this number (corresponding to 11.3% of the sample) was taken as the basis for numerical comparison with the normal controls, for whom no informant histories were available.

From the list of 45 000 elderly residents supplied by the Family Health Services Authority, a control was drawn at random from those matching each of the 559 dementia register patients by sex, completed years of life and district of residence. Because in the index group all known first psychiatric contacts had occurred before the age of 70 years, we took this as the cut-off age for the control group also. Records of psychiatric treatment episodes before age 70 were found for 19 members (3.4%) of the control group. For the index group, mean age at first psychiatric contact was 49.4 years (s.d. 12.7) and mean time interval to diagnosis of dementia 26.7 years (s.d. 10.3), while for the controls the mean age at first contact was 46.8 years (s.d. 13.5) and mean interval to the time of assessment 28.8 years (s.d. 10.5).

The difference between cases and controls in frequency of any previous psychiatric illness (63/559 vs 19/559, with two positively concordant matched pairs) corresponds to a risk OR of 3.59 [95% confidence interval (CI) 1.67-5.51]. If a more stringent criterion of `independence’ is applied to the previous psychiatric history by including only cases with a first referral at least 10 years before dementia onset, the number of affected index cases is reduced to 57 and, since the number of affected controls remains unchanged, the OR (with one positively concordant pair) falls to 3.11 (95% CI 1.46-4.76).

Eleven members of the control group had been clinically diagnosed as having dementia after the age of 70, but because none of these was matched with affected index cases, discarding the 11 pairs concerned from the matched case-control comparisons did not affect the above OR estimates. That no earlier psychiatric illnesses among this group were identified from the records is unsurprising, since with a frequency of 11.3% the expected number of such cases would be only 1.2.

When the association with previous psychiatric history was tested separately for Alzheimer-type and other forms of dementia, again no evidence of specificity was found: 40 of the 395 patients with Alzheimer-type dementia had psychiatric histories identifiable from the records, compared with 13 of their matched controls, while the corresponding numbers among the 164 patients with other forms of dementia and their matched controls were 23 and 6 respectively. Applying the McNemar test for differences between matched pairs to the fourfold tables yields [chi square] values of 13.25 for Alzheimer-type cases and 9.48 for all other dementia cases, both of which are significant at the 0.001 level of probability.

In Table 6, frequencies of previous psychiatric illness among dementia register patients and controls are set out as rates per 1000, according to diagnostic category and sex. Applying the McNemar test to the basic data yields [chi square] values of 8.45 for the first category (schizophrenic and paranoid psychoses), of 8.89 for that of affective disorders (bipolar and monopolar together) and of 4.27 for the remaining categories combined. The first two values are highly significant, and the third is significant at the 5% level. In this population, therefore, the association between previous psychiatric illness and dementia risk does not appear to be restricted to depressive illness. The data indicate that a history of relatively severe mental disorder, whether of schizophrenic or affective type, is associated with increase in dementia risk.

Table 6. Cases of previous psychiatric illness identified from hospital records, according to diagnostic category: dementia register patients and matched control group

No. of patients (and rate per 1000)

Register Matched

patients controls

Diagnostic category (n = 559) (n = 559)

Schizophrenia or paranoid

psychosis 18 (32.2) 4 (7.2)

Bipolar affective psychosis 6 (10.7) 0

Depression (unipolar) 27 (48.3) 12 (21.5)

Neurosis or personality

disorder 4 (7.1) 1 (1.8)

Chronic alcoholism 5 (8.9) 2 (3.6)

Other 3 (5.4) 0

All psychiatric diagnoses 63 (112.3) 19 (34.0)

Table 7, which also presents population-based rates, provides some supporting evidence, in terms of raised probabilities among the dementia register patients of having been admitted to psychiatric in-patient care and undergone electro-convulsive therapy, prefrontal leucotomy, or treatment with lithium or neuroleptic drugs. While the data summarized in Table 7 are reliable within the limits of the present study, they cannot be assumed to correspond to total lifetime experience of psychiatric treatment.

Table 7. Types of care and treatment recorded for previous psychiatric illness: dementia register patients and members of matched control group

No. of patients (and rate per 1000)

Register Matched

patients controls

Type of care/treatment (n = 559) (n = 559)

Psychiatric inpatient care 47 (84.0) 11 (19.7)

Outpatient or day-hospital

care only 16 (128.6) 8 (14.3)

Electro-convulsive therapy 17 (30.4) 2 (3.6)

Prefrontal leucotomy 3 (5.4) 0

Lithium treatment 4 (7.2) 1 (1.8)

Neuroleptic drugs 22 (39.3) 6 (10.7)

Anti-depressants 31 (55.5) 12 (21.5)

Discussion

The advantage of this study was that it could make use of both an existing dementia register for a defined population and the earlier Camberwell psychiatric case register, which covered the same population and for over 20 years recorded all episodes of care by mental health agencies serving the area. In addition, we were able to draw individually matched normal controls at random from the background elderly population.

The problems of method inherent in retrospective case-control designs and record-linkage studies are now well recognized. A possible source of inaccuracy arises here because dementia register patients and normal controls could not be matched for duration of residence in the area. Date of registration with the current medical practitioner is recorded by the Family Health Services Authority, but this is a highly unreliable guide to the period of time at risk, because many old people who are admitted to long-term care (a high proportion of the dementia register patients) must then change their doctors. Bias of this nature can be excluded only by gathering information at first hand, on both index cases and controls. Further investigation therefore calls for case-control and cohort studies, in which direct contact can be made with persons at risk and key informants.

More generally, studies based solely on hospital and other service records are bound to be subject to selective bias [32]. Those dementing old persons who have a history of psychiatric illness may be especially prone to behaviour disorders, and hence more likely to become known to specialist agencies. Specialist referral could be facilitated in cases where there have been psychiatric contacts earlier in life. Such people might also be over-represented among residents of long-stay homes. Each of these tendencies could lead to a bias in favour of a positive association. Although the comparisons made here with a matching group of other dementia register patients have provided little evidence for such trends, the research needs to be extended to include representative groups of dementia cases unknown to the specialist services.

Retrospective case-control studies which rely on biographical data from informant accounts may seriously under-estimate associations with this particular risk factor. In this study, little more than half the previous psychiatric illnesses which were identified could have been detected by informant interviews alone. There is a continuing need for research based, at least in part, on service-agency records.

The nature of any causal connections which might explain the reported association is so far obscure. An apparent lack of specificity with respect both to type of dementing disorder and to psychiatric illness category could result if a number of different risk exposures were involved, whether interrelated or not. Equally, however, it could reflect the limitations of a clinical diagnostic classification which makes too little allowance for the extensive overlapping between Alzheimer-type and other categories of dementia, on the one hand, or between different forms of functional mental illness, on the other. Some clarification of this question can be anticipated as it becomes possible to examine the previous psychiatric history of dementia patients in relation to neuropathological findings post mortem. Further research in this field should not be concentrated too narrowly on depression and Alzheimer’s disease, but should test also for associations between other mental illnesses and late-life dementia, and explore their significance. A logical next step would be to plot the occurrence of different dementing disorders in cohorts of people with well-documented histories of major psychiatric illness and treatment, and to compare these with matched control groups.

Key points

* A past psychiatric history, particularly of major mental disorder, appears to be a risk factor for the development of late-life dementia.

* This association is not specific for a previous history of depression, but occurs with schizophrenia and other major psychiatric disorders.

* The association is not restricted to dementia of the Alzheimer type.

* A history of psychiatric illness increases the probability of auditory hallucination in the course of dementia, but not of other forms of non-cognitive psychopathology.

Acknowledgements

The Wellcome Trust supported the Camberwell dementia case register during its first 2 years by a project grant (no. 036000) made to its director, R. Levy, and also awarded a grant (no. 039652) to B.C. for epidemiological research based on the dementia register. We are grateful to J. P. Left, Director of the Medical Research Council Social Psychiatry Unit, for providing access to the archives of the Camberwell psychiatric case register and to Geoff Der, formerly of that unit, for valuable assistance in searching the records. J. F. Powell of the Department of Neurosciences, Institute of Psychiatry gave technical advice on DNA typing, and money for laboratory consumables was donated by the Psychiatry Research Trust. Permission to make use of the FHSA data base was kindly given by S. Langford, Director of Commissioning, Lambeth, Southwark and Lewisham Health Commission, and we are particularly indebted to Sue Renshaw (Information Services), who processed the data, and David Shaw (Patient Services) for helpful advice.

References

[1.] Wragg RE, Jeste DV. Overview of depression and psychosis in Alzheimer’s disease. Am J Psychiatry 1989; 146: 577-87.

[2.] Barclay LL, Kheyfets S, Zemcov A, Blass JP, McDowell FH. Risk factors in Alzheimer’s Disease. In: Fisher A, Hanin I, Lachman C eds. Alzheimer’s and Parkinson’s Diseases. New York: Plenum, 1985; 141-6.

[3.] Broe GA, Henderson AS, Creasey H et al. A case-control study of Alzheimer’s disease in Australia. Neurology 1990; 40: 1698-707.

[4.] French LR, Schuman LM, Mortimer JA et al. A case-control study of dementia of the Alzheimer type. Am J Epidemiol 1985; 121: 414-21.

[5.] Kokmen E, Beard CM, Chandra V, Offord KP, Schoenberg BS, Ballard DJ. Clinical risk factors for Alzheimer’s disease: a population-based case-control study. Neurology 1991; 41: 1393-7.

[6.] Henderson AS, Jorm AF, Korten AE et al. Environmental risk factors for Alzheimer’s disease: their relationship to age of onset and to familial or sporadic types. Psychol Med 1992; 22: 429-36.

[7.] Heyman A, Wilkinson WE, Stafford JA et al. Alzheimer’s disease: a study of epidemiological aspects. Ann Neurol 1984; 15: 335-41.

[8.] Shalat SL, Seltzer B, Pidcock C, Baker EL. Risk factors for Alzheimer’s disease: a case-control study. Neurology 1987; 37: 1630-33.

[9.] Canadian Study of Health and Aging Risk factors for Alzheimer’s disease in Canada. Neurology 1994; 43: 1467-72.

[10.] Fuhrer R, Dufouil C, Antonucci TC, Datigues JF. Is depression predictive of the incidence of dementia? Results from a longitudinal community survey in France. Submitted for publication (1996).

[11.] Jorm AF, van Duijn CM, Chandra V et al. Psychiatric history and related exposures as risk factors for Alzheimer’s disease: a collaborative re-analysis of case-control studies. Int J Epidemiol 1991; 20 (suppl. 1): 43-47.

[12.] Agbayewa MO. Earlier psychiatric morbidity in patients with Alzheimer’s disease. J Am Geriatr Soc 1986; 34:561 -64.

[13.] Holmes C. The Camberwell Dementia Case Register. Int J Geriatr Psychiatry 1996; 11: 369-75.

[14.] Holmes C, Cooper B, Levy R. Dementia known to mental health services: first findings of a case register for a defined elderly population. Int J Geriatr Psychiatry 1995; 10: 875-81.

[15.] Roth M, Huppert FA, Tym E, Mountjoy CQ. CAMDEX: the Cambridge Examination for Mental Disorders of the Elderly. Cambridge: Cambridge University Press, 1988.

[16.] Folstein ME Folstein SE, McHugh PR. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J Psychiatric Res 1975; 12: 189-98.

[17.] Hodkinson HM. Evaluation of a mental test score assessment of mental impairment in the elderly. Age Ageing 1972; 1: 233-8.

[18.] Blessed G, Tomlinson BE, Roth M. The association between quantitative measures of dementia and senile change in the cerebral grey matter of elderly subjects. Br J Psychiatry 1968; 114: 797-811.

[19.] Hachinski VC, Iliff LD, Zilka E et al. Cerebral blood flow in dementia. Arch Neurol 1975; 32: 632-37.

[20.] Alexopoulos GS, Abrams RC, Young RC, Shamoian CA. Cornell Scale for Depression in Dementia. Biol Psychiatry 1988; 23: 271-84.

[21.] Webster DD. Critical analysis of the disability in Parkinson’s disease. Modern Treatment 1968; 5: 257-82.

[22.] Burns A, Jacoby R, Levy R. Psychiatric phenomena in Alzheimer’s disease, parts. I-IV. Br J Psychiatry 1990; 157: 72-94.

[23.] Allen NHP, Gordon S, Hope T, Burns A. Manchester and Oxford Universities Scale for the Psychopathological Assessment of Dementia (MOUSEPAD). Br J Psychiatry 1996; 169: 293-307.

[24.] Wenham PR, Price WH, Blundell G. Apolipoprotein E genotyping by one-stage PCR. Lancet, 1991; 337:1158-9.

[25.] McKeith IG, Perry RH, Fairburn AF et al. Operational criteria for senile dementia of Lewy body type (SDLT). Psychol Med 1993; 22: 911-22.

[26.] Wing JK, Hailey AM. Evaluating a community psychiatric service: the Camberwell Register 1964-1971. Oxford: Oxford University Press, 1972.

[27.] Fleiss JL. Statistical Methods for Rates and Proportions, 2nd edition. New York: Wiley, 1981.

[28.] Cooper B, Holmes C. The Camberwell Dementia Case Register: a valuable research tool. IPA Bulletin 1997; 14: 23-4.

[29.] Savitz DA, Olshan AO. Multiple comparisons and related issues in the interpretation of epidemiologic data. Am J Epidemiol 1995; 142: 204-8.

[30.] Saunders AM, Strittmatter WJ, Schmechel D et al. Association of apolipoprotein E allele 4 with late-onset familial and sporadic Alzheimer’s disease. Neurology 1993; 43: 2073-80.

[31.] Liddell M, Williams J, Baker A, Kaiser F, Owen M. Confirmation of association between the e4 allele of apolipoprotein E and Alzheimer’s disease. J Med Genet 1994; 31: 197-200.

[32.] Berkson J. Limitations of the application of fourfold table analysis to hospital data. Biometrics 1946; 2: 47-53.

Received 28 February 1997

BRIAN COOPER, CLIVE HOLMES

Section of Old Age Psychiatry, Institute off Psychiatry, De Crespigny Park, London SE5 8AF, UK

Address correspondence to: B. Cooper: Fax: (+44) 171 701 0167

COPYRIGHT 1998 Oxford University Press

COPYRIGHT 2000 Gale Group