An early study in Finland14 relied on the traditional two stage prevalence study design15 to examine descriptive epidemiology issues but did not find a significant rural/urban difference, something they do not remark upon. It does not seem this was a major concern at the time, and slightly later works, in particular that of Ali Rajput,1617 cite the Langston work as a rationale for believing rural living to be a concern. Later workers followed Rajput's lead in this.
Studies seeking to clarify etiologic or risk factors have therefore looked for the sources of apparent differences between urban and rural rates in descriptive studies. Following is a selection of the epidemiologic research on PD and agricultural factors that followed the MPTP incidents. The studies for the most part built upon each other and are therefore listed in chronological order. The community of PD epidemiology researchers, being small, communicates internally unusually well. The first studies were simply ecologic looks at rates of PD versus rural or industrial region (as the exposures). Earlier individual studies tended to use convenience or clinical samples of PD cases and broad definitions of exposure, typical of exploratory stages. Later studies used case-control design, more detailed exposure assessments, and broader case-finding techniques. Results are not always consistent, and design differences likely account for some of that.
Numerous case-control studies report elevation of risk in rural areas (for example, Rajput, 1986 and 1987; Ho, 1989; Koller, 1990; Tanner, 1986; and Svenson, 1993).1117-21 A few report elevated risk in urban areas (Tanner, 1989; Zayed, 1990; Butterfield, 1993),22-24 and there are reports of no association (Stern, 1991, and Sem-chuk, 1991).2526 Part of these differences is likely a result of geographic differences (comparing China versus Kansas, for instance), and part is due to inadequate statistical power or other methodologic problems, but part remains unexplained. The balance of opinion favors a risk associated with rural living in developed countries and city living in poorer countries (suggested by Tanner, 1990).13
The most important of the early studies following the MPTP papers was by Barbeau,12 27 who compared detailed use and sales data for pesticides in Quebec, Canada, with prevalence of PD, determined by four methods.27 These included reports to physicians, sales data for anti-PD medications, examination of death certificates, and contact of major movement disorder clinics and neurologists practicing within 100 miles of Montreal for neu-rologically confirmed cases. These latter clinicians were contacted to overcome the deficiency of the first methods, namely diagnostic accuracy. A final group of clinicians, those in general practice in one of the areas of high prevalence, were also contacted to confirm the diagnosis in their PD patients. The investigators' choice of this area, Quebec Province, cited the advantages of genetic homogeneity, equal access to medical care, and an even distribution of age groups. Prevalence estimates of Parkinson's disease as defined by these three methods in the province ranged from 0.834 to 0.967 per 1,000. The areas found to have the highest relative prevalence were those with the most land in intensive high-irrigation and chemical use agriculture. Of note, the investigators reported results by hydrographic regions rather than political divisions, stating that this gives a better depiction of environmental influences. While they seem unaware of the problem of attributing group-level exposures to population risk of disease, they do discuss the possibility of other unmeasured causal agents entering into the associations observed. They found an uneven prevalence within rural areas, contrary to previous reports at the time. After examination of possible confounders such as age and gender structure or reporting and diagnostic differences, they conclude that differences in agricultural practice and chemical use explain a large proportion of the prevalence differences observed.
The next Quebec study used individual patients rather than grouped data. Zayed et al.23 chose southern Quebec, Canada, in which to assess environmental risks in PD patients because of its stable population, mixed economy including farming, orchards, and metallurgic industries, and indications of an elevated prevalence of PD found earlier by Barbeau.27 Cases (n = 42) were found by contacting local primary care practitioners and neurologic consultants by mail and by publicity in local periodicals. Controls were drawn using municipal phone books, matched two to one on age, sex, and municipality and examined for neurologic illness. Rural residence was protective [odds ratio (OR) = 0.31]. However, there was a significant tendency for risk to increase with increasing exposure (i.e., longer residence in those areas). Some elevation was seen of risk associated with pesticide use. While Barbeau's study was ecologic (group-based), Zayed's was an individual-based study. Possible interpretations of the apparent contradiction include study design differences, including the possibility that Zayed overmatched and thereby reduced the effect of residence.s
Rajput and colleagues14 focused on early-onset (younger than age 40) PD in Saskatchewan, Canada. An important assumption that was usually explicitly stated by the investigators who studied early-onset PD was that it was etiologically the same as normal-onset disease, differing primarily in intensity of exposure or perhaps susceptibility to the exposures of interest. In any event, it was hoped that the younger subjects would be closer in time to the cause of the PD and hence better able to report such exposures. A complete assessment of drinking water, use of farm chemicals, and early onset PD was done on all early onset cases located in earlier work.17 Agricultural chemical use records were obtained from provincial records and from experts in the field at the University of Saskatchewan. Early-onset PD did not seem to rise in incidence with the use and introduction of agricultural chemicals. The weakness of this technique is the imputation of very broad group-level exposures to individuals with disease. They did nevertheless find an association of early-onset PD with both rural living and well-water use in childhood. A strength of this study is the set ting—Saskatchewan, Canada, which is a rural farming province with a very stable population, making these broad exposures easy to assess with accuracy. As might be expected, however, sample size was quite low, and Rajput did not describe the source of his control group other than to mention matching by age and sex and some exclusionary criteria. It might be reasonably expected that, in a place such as Saskatchewan, Parkinson's disease clinics might be so limited in number that were the controls to develop the disease, they would appear at Dr. Rajput's clinic and thereby have the opportunity to be included as case subjects in this study.
Sethi and colleagues28 took note of the elevation of PD risk reported by Barbeau27 in areas containing paper mills. To attempt to confirm that finding, they obtained death certificates for the U.S. state of Georgia and the total U.S.A., years 1979 through 1983. Unfortunately, they had available only primary and underlying causes of death and, as they acknowledged, the majority of people dying with PD do not have the PD noted on those first two causes. Therefore, their sample was biased, and that may have contributed to their inability to show an association of PD either with residence in counties that had paper mills or in rural areas.
A late-comer amongst the descriptive studies used administrative data in the province of Alberta, Canada11 (Svenson). The design differed from the usual cross-sectional study in that the investigators included all enrollees in the Alberta Health Care Insurance Plan, which includes the physician billing information for all "registered residents" of Alberta. The cohort's inception was April 1, 1984, and they were followed until March 31, 1989. It stayed closed during the period, during which 6% attrition was observed. All diagnoses of PD were extracted and described demographically. The investigators estimated both prevalence and "morbidity," or incidence, as a standardized morbidity ratio. Excesses of both prevalence and morbidity were observed for census divisions outside Alberta's two large cities, while urban areas were significantly reduced. They discuss the usual precautions about misclassification of PD diagnosis and an inability to detect undiagnosed PD. As an editorial comment, it seems a little odd that this work apparently was done independently of the Semchuk26 research, which was in preparation at about the same time, also in Alberta. These authors are located in Edmonton and are government employees, while Semchuk and her colleagues are academics located at that time in Calgary.
The balance of opinion favors a risk associated with rural living in developed countries and city living in poorer countries (suggested by Tanner, 1990.)13 A question not fully addressed by the early studies is precisely what is meant by "rural living." Does it refer to farming, cattle ranching, or living in the deep woods? Barbeau and Zayed both differentiated rural areas into farming and areas of rural industries, but ranches have not, to this writer's knowledge, been examined for association with PD, and the size of a community before it is termed "rural" has had only cursory examination.
Was this article helpful?