Dementia in Parkinson's disease

Dementia is common in Parkinson's disease (PD) and plays a significant part in the disability ^[1]. Estimation of its frequency has often been the source of controversy owing to variations in the case selection methods and diagnosis criteria used ^[1]. Using standardized criteria for dementia (DSM-III), Brown and Marsden ^[1] suggested that dementia may affect about 15% of PD cases. From this and numerous other studies, there is some general agreement that dementia occurs in 10 to 20% of cases ^[1-10]. Most studies have been hospital or clinic based ^[1,3]. 5,8,9 There are only a few studies of randomly sampled cases using a population-based methodology that can lessen selection bias ^[6,7,11]. Recently, in a population-based registry of PD cases, using strict diagnostic criteria, Mayeux et al ^[11] found that the frequency of dementia in PD cases aged 65 and older was 40%, raising the possibility that previous studies may have underestimated the frequency of dementia in PD.

As part of a large longitudinal study on brain aging (PAQUID), ^[12] we ascertained all PD cases using well-defined criteria in a large random sample of the population aged 65 years and older, living at home or in institutions, in Gironde (southwestern France) ^[13]. Dementia was diagnosed according to DSM-III-R criteria in both PD patients and individuals without PD. Our main objectives were to estimate the frequency of dementia and whether PD was an independent risk factor for dementia.

Methods. General study design. The study was conducted on a large sample of men and women over age 65 living in households and in institutions in Gironde, France (PAQUID cohort on brain aging) ^[12]. The random sampling procedure has been previously described in detail ^[12]. The household sample was obtained using a two-step random procedure. The Gironde departement was first divided into five strata based on population size, and geographical units were randomly selected in each stratum. A total of 4,050 individuals aged 65 years and older were then randomly selected from the electoral rolls. About 69% of the selected individuals agreed to participate and consisted of 1,122 men (40.2%) and 1,670 women (59.8%), with a mean age of 74.9 years. Individuals who did not agree to participate had similar age, sex, and educational attainment distributions as those who were included. The institutionalized sample was also selected using a two-step procedure. In the first step, 42 institutions were randomly selected in five strata based on the different types of institutions. A total of 380 individuals were then randomly selected from the residents of the 42 institutions. Only 6% refused to participate, and the institutionalized sample finally consisted of 91 men (25.5%) and 266 women (74.5%), with a mean age of 82.6 years.

Data were collected by means of a questionnaire administered by trained investigators. Items were related to basic demographic data, socioeconomic characteristics, personal medical history, current symptoms and diseases, and presence of hearing and visual deficiencies. Physical disability was assessed by means of different scales (Katz Scale for Activities of Daily Living, ^[14] Lawton and Brody Scale for Instrumental Activities of Daily Living, ^[15] Rosow and Breslau Scale for Physical Activities ^[16]). The Center for Epidemiologic Studies-Depression Scale (CES-D) was used to identify individuals with a high level of depressive symptomatology ^[17]. To identify individuals with an elevated level of depressive symptoms in the French population, the recommended cutoff score of 17 and above was used for men and 23 and above for women ^[17]. This was determined in a study ^[17] in which CES-D scores were compared with both clinical diagnosis, based on DSM-III criteria, and a clinician rating scale of depression ^[18]. A series of neuropsychological tests was administered to each individual, including the Mini-Mental State Examination (MMSE), ^[19] the Benton Visual Retention Test, ^[20] the Wechsler Memory Scale, ^[21] and the Isaacs Set Test of Verbal Fluency ^[22].

Diagnosis of Parkinson's disease. As previously described, ^[13] a two-phase design was used to ascertain PD cases in the household sample. During the first phase, screening for parkinsonism was made through two questions on bradykinesia and rest tremor ("Do your arms or your legs shake at rest?" and "Do you experience slowness or stiffness in your movements?"). These two questions have previously been demonstrated by others, in different countries, to be an accurate and valid screening instrument for PD in population-based studies ^[23,24]. Subjects positive for these two items were then examined by a neurologist, who confirmed or excluded parkinsonism and diagnosed PD. A diagnosis of parkinsonism was confirmed on the basis of the presence of at least two of the cardinal signs: bradykinesia, rigidity, tremor at rest, and typical postural and gait abnormalities. A clinical diagnosis of PD was reached taking into account the history and asymmetry of onset of symptoms, after exclusion of the following conditions: possible drug-induced parkinsonism (subjects taking neuroleptics at the time of neurologic examination), parkinsonism associated with a history of cerebrovascular disease, atypical disease history (lack of response to anti-parkinsonian drugs, parkinsonism associated with unusual neurologic features) ^[25]. In institutions, all individuals were examined by a neurologist using the same diagnosis procedure without the preliminary screening stage phase ^[13]. For all cases classified as PD by the examiner, records were reviewed for confirmation by an independent neurologist.

Diagnosis of dementia. At the end of the baseline interview, a diagnosis chart was completed for each subject of the cohort, allowing the DSM-III-R criteria ^[26] to be obtained by using data collected through spontaneous complaints, functional assessment scales (Katz Scale for Activities of Daily Living, ^[14] Lawton and Brody Scale for Instrumental Activities of Daily Living, ^[15] Rosow and Breslau Scale for Physical Activities ^[16]), and psychometric scales (MMSE, ^[19] Benton Visual Retention Test, ^[20] Wechsler Memory Scale, ^[21] and the Isaacs Set Test of Verbal Fluency ^[22]) ^[27,28]. All cases classified as demented were reviewed by a committee in order to reach a consensus diagnosis following the DSM-III-R criteria ^[26-28].

Data analysis. The age-specific prevalence of dementia in the elderly population of Gironde was estimated separately in individuals with PD and without PD. Overall estimates were computed taking into account the respective proportions of individuals living at home and in institutions. Using the Mantel-Haenszel procedure, we computed age-adjusted odds ratios (ORs) with a 95% confidence interval (CI) to estimate the relative risk of being demented when suffering from PD in the household sample and in the institutionalized sample separately. The prevalence of dementia in PD and non-PD cases, in the whole population, was calculated using the Cochran method for stratified samples ^[29]. Controlling for age and depressive symptomatology, a logistic regression was performed to examine whether PD was an independent risk factor for dementia. The dependent variable was the diagnosis of dementia (yes/no), and the independent variables were the diagnosis of PD (yes/no), age (in years), and depressive symptomatology (yes/no). Finally, the MMSE scores were compared in PD and non-PD subjects with an analysis of variance with covariate. The MMSE score was the dependent variable. Age was taken into account as covariate. BMDP statistical software was used for all the statistical analyses.

Results. A total of 60 PD cases were identified in the whole sample, 24 living at home and 36 in institutions ^[13]. Sixteen of these patients were previously undiagnosed. In the household sample, PD cases were significantly older than individuals without PD (78.5 years, SD = 6.7, versus 74.9 years, SD = 7.0, p < 0.05). In institutions, there were no significant differences for age between PD cases and non-PD cases (84.0 years, SD = 8.0, versus 82.6 years, SD = 7.6).

A total of 17.6% of the PD patients were demented as compared with 3.9% of individuals without PD, the overall age-standardized ratio being 4.6. In the 65- to 79-year-old group, frequency of dementia was 6.3% in PD cases and 2.8% in non-PD cases. In individuals aged 80 years and older, frequencies of dementia were 33.8% in individuals with PD and 6.7% in individuals without PD. The frequency of dementia in PD patients living at home was 16.7% compared with 33.3% for those living in institutions. The effect of age on the frequency of dementia in PD was found in both household and institutionalized samples.

Table 1shows that PD was associated with an increased risk of dementia (OR = 7.97, 95% CI = 2.4 to 26.3) in individuals living at home. When the household sample was divided into two age groups (65 to 79 years and >=80 years), we found that PD was positively associated with dementia in both groups, but ORs were not significant in the younger age group.

In the institutionalized sample, there was no association between PD and dementia (OR = 1.08, 95% CI = 0.5 to 2.4).

To complement this analysis, we examined the distribution of age-adjusted MMSE scores in PD cases and in individuals without PD. MMSE scores were divided into four classes: 0 to 9, 10 to 19, 20 to 23, and >=24. In the household sample, but not in institution residents, the proportion of PD patients increased significantly from the highest (>=24) to the lowest (<=9) MMSE values (table 2). Moreover, when all demented individuals were excluded, PD patients had significantly lower age-adjusted mean MMSE scores than individuals without PD (23.8 versus 25.8, F test, F = 6.62, p = 0.01, degrees of freedom = 1 and 2,656).

A total of 32.7% of PD cases expressed a significant level of depressive symptomatology. In the household sample, frequency of depressive symptomatology was significantly higher in PD cases (33.3%) than in individuals without PD (13.3%), and also higher in younger subjects (65 to 79 years: 40%) than in the older ones (>=80 years: 22%). In institutionalized subjects, there was no association between PD and depressive symptomatology.

The co-occurrence of dementia and high depressive symptomatology was observed in 4.8% of PD patients living at home versus 0.4% of individuals without PD. In subjects living in institutions, this co-morbidity was independent of PD.

Finally, controlling for age and depressive symptomatology, we examined whether PD was an independent risk factor for dementia in the household sample and the institution sample separately. Logistic regressions showed that PD was associated with an increased risk of dementia in the elderly population living at home (OR = 8.2, 95% CI = 2.2 to 30.4). In the institutionalized subjects, no significant relationship between PD and dementia was found (OR = 1.26, 95% CI = 0.5 to 3.2).

Discussion. This is the first study of dementia in subjects with and without a diagnosis of PD conducted through a "door-to-door" survey in household and institutionalized elderly populations. We diagnosed both dementia and PD according to conventional, well-defined criteria, including neurologic examination, extensive neuropsychological testing and comprehensive functional assessment, ^{[1,9,11,26-28]} and we found that the frequency of dementia in PD cases was 17.6%. This is in accordance with the commonly accepted average rate of about 15% derived from numerous previous studies ^[1-10]. In contrast, the frequency of dementia in PD in this study was lower than the 40% rate found by Mayeux et al ^[11] in a population-based register of PD cases in Manhattan, New York. Although both studies were population-based, they used different case-finding strategies. Also, since standardized prevalence rates were not compared, the estimates might not be comparable due to differences in the age structure or gender distribution of the population compared. In addition, social and educational differences between the PAQUID and the Manhattan samples might contribute to differences in dementia frequency. The high frequency of dementia in the Manhattan PD sample could also be compared with the extremely high overall prevalence of dementia reported in other US populations ^[30].

In agreement with other studies, ^[11] we found that the frequency of dementia in PD increased with age (prevalence ratio between age groups >=80 years and 65 to 79 years = 4.4); this increase is similar to that of the prevalence of dementia in individuals without PD (prevalence ratio between age groups >=80 years and 65 to 79 years = 4.9).

Depression is often considered an important factor associated with the cognitive impairment in PD ^[31]. The clinical diagnosis of depression was not formally made in our study, but significant depressive symptoms were noted through the CES-D scale ^[17]. Such symptoms were present in 32.7% of PD cases, which is rather close to the 40% generally accepted ^[32]. The rate of co-occurrence with dementia is also close to that previously proposed by Sano et al ^[33] (about 5%).

Household and institution samples showed different patterns of relationship between PD and dementia or depression. One likely explanation of the observed differences is a Berkson's bias: in France, cognitive impairment is the first cause of institutionalization, and approximately 40% of institution residents are demented ^[12]. Bias resulting from this pattern decreases the power of testing hypotheses regarding dementia. This is likely the reason why neither dementia nor depression were significantly associated with PD in institutionalized residents.

Our study design permitted us to examine whether PD was an independent risk factor for dementia and to estimate the risk value. In the household sample, we found that PD was associated with an eight-fold increase in the risk of dementia.

We did not determine the precise nature of the dementia. Dementia in PD may be a direct or indirect consequence of the specific pathologic features of the disease, but there are alternative hypotheses. Mayeux et al ^[34] found that first-degree relatives of Alzheimer's disease patients were not at greater risk for dementia when compared with relatives of patients with PD. Similarly, both Hofman et al ^[35] and Marder et al ^[36] reported that dementia and PD aggregated in families of patients with PD or Alzheimer's disease. These findings may suggest that two independent pathologic consequences of accelerated brain aging could aggregate due to genetic or environmental factors, or to a combination of the two.