Results

Among the 3777 subjects who constituted the PAQUID cohort, 102 were demented at baseline and were excluded from the study sample. The 3675 initially non-demented subjects included 2133 (58%) women and 1822 subjects (49.6%) older than 74. The main occupation during active life was distributed as follows: 376 housewives and inactives (10.2%), 561 farmers (15.3%), 315 domestic service employees (8.6%), 638 blue collar workers (17.4%), 948 other employees (25.8%), 499 craftsmen and shopkeepers (13.6%), and 338 professionals and managerials (9.2%). Repartition differed according to sex. In women, 17.5% were housewives and inactives, 14.8% were farmers, 13.7% were domestic service employees, 14.1% were blue collar workers, 23.5% were other employees, 11.3% were craftsmen and shopkeepers, and 5.1% were professionals and managerials. In men, these proportions were respectively 0.1%, 15.8%, 1.4%, 21.9%, 29.0%, 16.7%, and 14.9%.

Of the 3675 non-demented subjects, 725 (19.7%) did not participate in any follow up visit because they died (n=413; 11.2%) or were either lost to follow up or refused the follow up screenings (n=312; 8.5%). At least one complete follow up evaluation was performed on 2950 subjects (80.3%). In comparison with professionals and managerials, the reduction in size of the cohort with time was more pronounced in all the other occupational categories, but the difference was only important for housewives and inactives (25% not followed up versus 16.9%) (table 1). The mean duration of follow up of the cohort was 6.4 years.

Three hundred and ninety three subjects developed an incident dementia, of whom 251 were classified as incident cases of Alzheimer's disease, 112 as vascular dementia or Alzheimer's disease with cerebrovascular disease, and 30 as other types of dementia (16 Lewy body disease, eight dementias complicating an idiopathic Parkinson's disease, two progressive supranuclear palsies, one multiple system atrophy, three frontotemporal dementias). Because of very few frontotemporal dementias, which cannot be collapsed with another subtype of dementia, these three cases have been excluded from the analyses. The 27 other cases of other types of dementia were taken into account as a simple category in the analyses, called dementia with parkinsonism. Incidence of dementia according to occupational category and aetiology of dementia are presented in table 2. Whatever the aetiological subtype of dementia, the lowest incidence was found for professionals and managerials. Farmers had the higher incidence of dementia all causes, of vascular dementia, and of dementia with parkinsonism. As we found no case of dementia with parkinsonism in professionals and managerials, we have regrouped them with other employees, for whom incidence of dementia was the lowest after professionals and managerials.

Using a Cox model with delayed entry with adjustment for sex and interaction age by sex, the risk of dementia was significantly greater in farmers (RR=1.67; 95%CI 1.23–2.27) and in blue collar workers (RR=1.51; 95%CI 1.10–2.07) than in the reference category of professionals, managerials, and other employees, whereas it did not differ significantly for other occupational categories (table 3, model 1). Presence of occupational categories significantly improved the fit of the model, as the log likelihood ratio statistic (LRS) was 13.9 (p<0.016). These significant relations disappeared after adjustment for education (table 3, model 2), and occupational categories no longer improved the model compared with one adjusted only for sex, interaction age by sex, and education (LRS p=0.538). For models 3 and 4, the results were unchanged. According to the aetiology of dementia, occupational categories did not modify the risk of Alzheimer's disease. Although the global effect of occupational categories was not significant, there was a tendency to an increased risk of vascular dementia in farmers (RR 1.90; 95%CI 1.05–3.43), even after adjustment for education. This risk among farmers remained virtually unchanged when the occupational categories were adjusted for sex, interaction age by sex, education, and even for wine consumption and smoking habits (RR 1.88; 95%CI 1.04 to 3.41). To explain this relation between occupation and risk of vascular dementia we adjusted for two other possible confounding factors: presence of at least one vascular factor (diabetes, hypertension, or history of stroke), and level of income. The presence of at least one vascular factor was significantly associated with the risk of vascular dementia (RR=2.09; 95%C 1.38–3.18) and explained the relation between occupation and vascular dementia as the risk ratio among farmers decreased and was no longer significant (RR=1.56; 95%CI 0.82–2.97). The risk of dementia with parkinsonism was also increased among farmers (RR=4.65; 95%CI 1.58–13.65) (table 4, model 1), with a nearly significant global effect of occupation (LRS; p=0.074). This increased risk of dementia with parkinsonism was not modified when adjusted for sex, interaction age by sex, education, and for wine consumption and smoking habits (RR 4.43; 95%CI 1.39–14.14), and even after adjustment for vascular factors and level of income (RR 4.64; 95%CI 1.30–16.61, table 4, models 3 and 4).

Because of the differences according to occupation between sex, the previous results were verified for each sex separately. Risk of Alzheimer's disease seemed to be associated with occupation differently among sex: being a craftsman and shopkeeper was associated with a protective risk among women (RR 0.44; 95% CI 0.23–0.87) whereas the risk was increased among men (RR 2.05; 95%CI 1.02–4.11), even after adjustment for confounders. However, the global effect of occupational categories was not significant either for women or for men. For vascular dementia, the results were similar in both sexes; the increased risk found among farmers before introduction of adjustment factors disappeared after introduction of confounders. The increased risk of dementia with parkinsonism found among farmers in the total sample was seen particularly in women, with a significant global effect of occupational categories (p=0.004, table 5). After adjustment for education and wine consumption, RR was 7.47 (95% CI 1.80–31.07). For men there was no significant relation between farmers and dementia with parkinsonism; the apparent effect of domestic service employees was caused by only one case of dementia with parkinsonism (table 6). Moreover, the global effect of occupational categories did not reach significance among men.

Discussion

OCCUPATION AND ALZHEIMER'S DISEASE

This 10 year follow up of 2950 initially non-demented French elderly people with prospective screening of incident dementia shows that occupational category has no major impact on the risk of Alzheimer's disease. These findings contrast with the strong relation found at the baseline screening phase of the cohort, between cognitive performance and principal lifetime occupation.14 15 This confirms previous analyses conducted on population based cohort studies with active search for incident cases of dementia in Canberra12 and east Boston,11 but is in conflict with the New York City study6 which is very close to the present study from a methodological point of view. Sternet al 6 found a significant effect of low occupational attainment on the risk of Alzheimer's disease independently of education. A lack of power cannot explain our negative results as the number of people followed up and the number of incident cases of Alzheimer's disease seen in the present study (2950 and 251 respectively) are far greater than in the New York City study (593 and 97 respectively). It could be argued that our study lacks validity in the aetiological diagnosis of dementia, as the diagnosis was essentially based on the clinical evaluation made at home by a senior neurologist, and brain imaging was available in only 29% of the 393 incident cases (v 78% in the New York city study). However, the good accuracy of the NINCDS-ADRDA clinical criteria for the diagnosis of Alzheimer's disease is well accepted by the scientific community.29 An indirect confirmation is given by the total absence of increased risk of Alzheimer's disease associated with the presence of vascular factors (RR 0.98). Moreover, the neurologist was not aware of the research hypothesis tested in this paper, and the clinical classification was thus conducted independently of occupational category. Differences in the occupational categories considered in the analysis could be another explanation. However, if the same classification as the New York City study (high occupation=professional and managerials; low occupation=other occupations) is taken, the relation remained non-significant (RR 1.18; 95%CI 0.66–2.11). The differences could be explained by the fact that adjustment for the diploma validating primary school education which we used in our study might be more precise than the adjustment for the number of years of schooling used by Stern et al.6 We have recently reported that the evaluation of level of education by this diploma was the best classification for predicting Alzheimer's disease in France.3 It was essentially based on linguistic and conceptual ability assessed by reading, writing, and spelling tests and mathematical problem solving. The Nun study30 has also shown the major impact of these abilities on the risk of Alzheimer's disease. Indeed, even when we used the definition of low education used by Sternet al 6 (8 years of schooling) we did not find any significant relation between occupation and Alzheimer's disease (RR 1.11; 95%CI 0.59–2.07). Thus it seems that the risk of Alzheimer's disease late in life has more to do with cognitive abilities in childhood and adolescence than with lifestyle and environment in adult life.

OCCUPATION AND NON-ALZHEIMER'S DISEASE DEMENTIA

The risk of vascular dementia tended to be increased among farmers, but this increased risk was explained by a higher proportion of vascular risk factors among farmers than among professionals, managerials, and other employees. On the contrary, we found an increased risk of dementia with parkinsonism among farmers, even after adjustment for confounders, but significant only for women.

Several methodological issues may have influenced these results. The reduction in the cohort size with time could interfere with our findings. However, our proportion of subjects who did not participate in any follow up is close to that in other cohort studies.11 21 Farmers refused to continue to participate or died before follow up slightly more than professionals, managerials, and other employees. However, the proportion of subjects successfully contacted remained high in each occupational category (>75%) and the differences in follow up of the subjects might essentially lead to an underestimation of the risk ratios.

As in all cohort studies, selective survival before entry into the cohort has to be considered. However, according to occupation, in France, as in other developed countries, survival is longer among professionals and managerials than among farmers; thus, if a selective survival bias exists, it would tend to underestimate the risk of dementia among farmers compared with professionals, managerials, and other employees.

On the other hand, farmers may simply perform worse on the psychometric tests used to screen for dementia whereas those with other occupations might perform better. However, in such a cohort study, the diagnosis at follow up necessarily implies decline from baseline performance and the final diagnosis was made by a senior neurologist. These procedures should minimise the detection bias. Moreover, reduction in the cohort size over time and detection biases, if present, were as great in Alzheimer's disease as in other types of dementia whereas significant relations were found only in non-Alzheimer's disease dementia.

Cognitive stimulation and premorbid intelligence have been proposed to explain the relation between education and incident dementia or Alzheimer's disease. A similar part played by occupation is unlikely as an association was found only for dementia with parkinsonism. Other explanations can be discussed and should be explored, such as differences in health behaviour according to occupation and differences in physical activity or nutrition. The toxic effect of chronic exposure to chemicals related to occupation and particularly pesticides in farmers already implicated in Parkinson's disease,31could also be an explanation for this specific association.

The relation between farmers and dementia with parkinsonism in women has been found in a small sample, as we found only 16 cases of dementia with parkinsonism in women. Among them, eight cases were found in farmers, leading to a large confidence interval. Lack of power could explain why we did not find any association among men. Thus, these results need to be confirmed in larger sample.

In conclusion, risk of Alzheimer's disease does not seem to be influenced by occupation, and the relation between occupation and dementia seems to be limited to dementia with parkinsonism. The increased risk for this disease among farmers could be due partially to differences in health behaviour and perhaps because of exposure to environmental factors. The reason why these results were found particularly among women should be explored further. A larger sample of subjects with dementia with parkinsonism with more precise measurements of the type and duration of exposures to environmental factors—for instance, by means of a job exposure matrix—are needed to confirm these findings.