Epidemiologic observations on Parkinson's disease

In part to identify etiologic clues of idiopathic Parkinson's disease (PD), many studies have sought to characterize and compare patterns of occurrence, such as prevalence and incidence rates. Distinguishing between genetic, maturational, and environmental etiologies of PD by such occurrence comparisons has, however, been confounded by methodologic problems. In most studies, PD occurrence has been characterized in terms of prevalence rather than incidence, introducing the potential for biased estimates due to differential mortality and differential diagnostic precision. It is generally accepted, nevertheless, that PD occurrence rates in both developed and developing Asian nations are probably lower than in Western nations with predominantly Caucasian populations. In this study, we report PD incidence, point prevalences, and mortality rates in a prospectively followed cohort of American men of Japanese or Okinawan ancestry.

Methods.

More detailed information about cohort study methods has been provided in previous publications. ^[1,2] Data were obtained from the Honolulu Heart Study (HHS), an ongoing cohort study of 8,006 men of Japanese or Okinawan ancestry born between 1900 and 1919 who resided on O'ahu, Hawai'i in 1965. The cohort has been periodically rescreened since 1965, most recently from 1991 to 1994, and there has been continual surveillance of O'ahu hospital records and death certificates to detect health and mortality outcomes. Mortality determination is considered complete through December 31, 1992. Mortality determination is still in progress for the deaths occurring after January 1, 1993.

Incident PD cases (those detected after the baseline examination, but before November 30, 1994) were identified through four sources: (1) review of all O'ahu hospitalization records of cohort subjects for new and pre-existing diagnoses of idiopathic PD, (2) ongoing review of all Hawai'i death certificates, (3) a review of medical records at the offices of eight of 10 O'ahu neurologists (two neurologists declined to participate), and (4) a complete rescreening of the cohort (1991 to 1994). The most recent rescreening included physical examination supplemented by standardized questions about parkinsonism diagnoses, signs and symptoms, and medication usage, followed if indicated by comprehensive neurologic examination. Cases in which a known cause of parkinsonism had been established (e.g., drug-induced, post-encephalitic, lacunar infarct, post-head trauma) were eliminated from this analysis.

Cases detected before 1991 in patients not later evaluated by us were included if the diagnosis of idiopathic PD had been confirmed by a neurologist or neurosurgeon, usually including clinical response to L-dopa. Cases in patients undergoing examination by the study's two neurologists were diagnosed according to published research criteria. ^[3] For the purpose of this study, date of onset was taken to be date of diagnosis. ^[1]

Results.

Incidence.

Based on data from 92 detected cases, PD risk increased logarithmically between the ages of 50 and 75 Table 1, Figure 1. Due to the age structure of the cohort at the time of enrollment, person-years of follow-up beyond age 80 are still comparatively few, and age-specific incidence calculations are thus associated with wide standard errors Table 1. Nevertheless, PD incidence decreased after age 80 Table 1, Figure 1. Excluding five prevalent PD cases identified in 1965, the average annual incidence rate, adjusted to the age distribution of the 1970 U.S. population as a whole, was 11.1 cases per 100,000 persons, and 45.2 per 100,000 persons aged 50 or older.

View this table:

• View inline
• View popup
• Download powerpoint

Table 1 Age-specific incidence rates of idiopathic Parkinson's disease, by 5-year age interval, per 100,000 person-years of pre-illness follow-up in an all-male cohort, Honolulu Heart Study, O'ahu, Hawai'i, 1965-1994

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Age-specific incidence rates of idiopathic Parkinson's disease (--), and of actual or initially suspected parkinsonism not attributed to idiopathic Parkinson's disease (- - -), by 5-year age interval, per 100,000 person-years of pre-illness follow-up in an all-male cohort, Honolulu Heart Study, O'ahu, Hawai'i, 1965 to 1994.

In comparing age-specific incidence rates of men born before 1910 with men born after 1910, the older PD age cohort not only experienced lower PD risk, but also peak PD incidence at a younger age than men born after 1910. The men in the older age cohort had peak onset at 75 to 79 years old, versus at least 80 to 84 years old for the younger men, resulting in an apparent rightward frame-shift of the age-incidence curve for the younger men Figure 2. However, as this younger birth cohort corresponds to a current (1994) age range of only 74 to 84, there are too few data to characterize its age-specific PD risk beyond age 79. Information about ``juvenile'' (early-onset) idiopathic PD, allegedly frequent in Japan, was not provided in this study because of the age of subjects at enrollment and because prevalent PD cases had been excluded.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Age-specific incidence rates of idiopathic Parkinson's disease, by 5-year age interval, per 100,000 person-years of follow-up, determined separately for men born between 1900 and 1909 (--) and men born between 1910 and 1919 (- - -), Honolulu Heart Study, O'ahu, Hawai'i, 1965 to 1994.

The age-incidence curve for disease onset in those men suspected of having PD, but not confirmed as having PD on complete neurologic examination, is similar to the age-incidence curve of actual PD Table 2, Figure 1. This category includes men with many different conditions, including essential tremor, motor disorders other than PD, and suspected PD that does not meet diagnostic criteria.

View this table:

• View inline
• View popup
• Download powerpoint

Table 2 Age-specific incidence rates of actual or initially suspected parkinsonism not attributed to idiopathic Parkinson's disease, by 5-year age interval, per 100,000 person-years of pre-illness follow-up in an all-male cohort, Honolulu Heart Study, O'ahu, Hawai'i, 1965-1994

Prevalence.

PD age-specific prevalence rates, calculated at three points in time (December 31, 1982, 1987, and 1992, respectively), reflect incidence risks superimposed upon the effects of cohort aging and PD-related and PD-unrelated mortality Table 3, Figure 3 and might conceivably have been influenced by PD-related loss to follow-up (see Discussion). Age-specific prevalence rates appear roughly similar in 1982 and 1987, but higher for 75 to 85-year-old subjects in 1992.

View this table:

• View inline
• View popup
• Download powerpoint

Table 3 Age-specific prevalence rates of idiopathic Parkinson's disease, by 5-year age interval, per 100,000 living men, at three points in time, Honolulu Heart Study, O'ahu, Hawai'i, December 31, 1982, 1987, and 1992

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Age-specific prevalence rates of idiopathic Parkinson's disease, by 5-year age interval, per 100,000 living men, at three points in time, Honolulu Heart Study, O'ahu, Hawai'i, December 31, 1982, 1987, and 1992.

Mortality.

After age 60, PD-associated mortality rates appeared to increase logarithmically Figure 4. Comparison with the mortality experience of the rest of the cohort Figure 4 suggests that this mortality effect might not be entirely attributable to age. Increased age-related PD mortality was associated with both absolute age and duration of illness longer than 10 years Figure 5.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. Age-specific mortality rates in men with (- - -) and without (--) idiopathic Parkinson's disease, by 5-year age interval, per 100,000 person-years of pre- and post-illness follow-up, through ages 75 to 92, Honolulu Heart Study, O'ahu, Hawai'i, 1965 to 1992.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5. Age-specific mortality rates from idiopathic Parkinson's disease (PD), by 5-year age interval, per 100,000 person years of pre- and post-illness follow-up, for men who died less than 5, 5 to 10, and more than 10 years after initial diagnosis, Honolulu Heart Study, O'ahu, Hawai'i, 1965 to 1992.

Between the ages of 70 and 89, persons with PD had a two- to three-fold increase in the risk of dying (corresponding to a mortality ratio of 2.5), and a mortality risk difference of 5,414.9 excess deaths per 100,000 person-years of follow-up. Based on crude survival curves calculated from cohort mortality figures for all persons alive at age 70, PD was associated with an approximate 8-year decrease in life expectancy, from a mean of 84 years to a mean of 76 years. This difference appears to be in the same range as the pre-levodopa figures of Hoehn and Yahr, *RF 3a* although both PD patients and healthy persons are now living longer than they were 30 to 40 years ago.

Discussion.

In 29-year prospective follow-up of a cohort of middle-aged and elderly American men of Japanese or Okinawan ancestry, PD was characterized as a relatively common disorder of the middle to late years associated with significant ``early'' mortality.

The age-incidence pattern of PD Figure 1 provides information on PD risk comparative with data from different population groups in other geographic regions ^[4-7] and also constitutes a model against which to fit etiologic theories.

When examined overall, or broken down by decades of life, the Hawai'i data are in general accord with accepted incidence figures from Europe ^[5,6] and the United States, ^[7] but 5- to 10-fold higher at each age stratum than age-specific incidence figures from China. ^[4] Based on incidence data, and on extrapolations of prevalence data from Japan ^[8-11] and China, ^[4,12] the notion has arisen that racially linked genetic factors may predispose to or protect against PD. ^[13] However, that American men of Asian ancestry appear to experience PD risk of the same pattern and magnitude as non-Asian American men, and different from what has been reported for Asian men living in Asia, does not provide strong support for racial/ethnic determinants of PD. Prevalence figures from Copiah County, Mississippi, and from Igbo-Ora, Nigeria, appear to be in accord: black men and women in Mississippi (most of whom are descended from members of inland and coastal West African tribes) have PD prevalence rates more like white men and women in Mississippi (comparatively high) than black men and women in Nigeria (comparatively low ^[14-16]). Zhang and Roman recorded worldwide incidence and prevalence rates of PD and parkinsonism. ^[17] Their conclusions were consistent with our data--that population PD risk may be more a function of environmental than racial/ethnic factors. However, comparative differences in occurrence rates of PD among many studies in developed countries, ^[17] generally on the order of three-fold or less, are sufficiently small that they could merely reflect methodologic differences and study biases. The five- to ten-fold difference in detected rates from China ^[4] is more problematic. Methodologic difference alone would not normally be expected to produce such a large difference, but this possibility cannot be ruled out, as the Chinese data have not yet been verified by other studies.

There are three basic epidemiologic conceptualizations of possible PD etiology. Under a ``genetic hypothesis,'' PD occurrence might differ between racial or ethnic groups, but exhibit similar epidemiologic patterns in groups of persons of the same racial/ethnic composition living in different locales. Under an ``environmental hypothesis,'' PD occurrence might be expected to differ geographically irrespective of the racial/ethnic make-up of the population under study. Under a ``maturational'' hypothesis, in which inevitable senescence of nigral cells eventually leads to PD in a purely age-dependent fashion, PD occurrence would not be expected to vary by either geographic locale or racial composition of the population. While the shape of the right tail of the age-incidence curve calculated for our data Figure 1 can only be speculated, we believe that PD incidence most likely decreases regularly after about age 75. Although somewhat controversial, this is in accord with previously published data. <sup>[18]</sup> If true, the widely held belief that PD results from normal or accelerated senescence of nigral cells <sup>[19]</sup> may not be tenable, and PD may not truly be a ``disease of aging.''

Other published studies ^[5-7] also fail to confirm strict age-dependence of PD. The incompletely characterized age-incidence curve of PD is not inconsistent with an induction or incubation period of one or more causative factors, or of a condition manifest only in genetically susceptible persons that has a normally distributed interval of time to phenotypic expression. This is consistent with the possibility that many PD cases result from widespread environmental or infectious exposures; early life exposures would seem most suspect, since widespread exposures are less likely to be encountered later in life, after individuals have retired or restricted their activities. Such a model of PD etiology would not rule out genetic factors; an environmental agent might act only, or disproportionately, in genetically susceptible persons, or the etiologic ``agent'' could be a gene itself, widely dispersed in the population, and with delayed expression. In the latter case, the age-incidence pattern might merely reflect the normal distribution of time to phenotypic expression, as documented for other genetic conditions with neurologic complications (e.g., Huntington's disease). Less plausibly, PD could be the ``early onset'' manifestation of a single underlying maturational process that leads to different or unappreciated clinical manifestations at older ages.

In our opinion, a likely explanation for the substantial age-associated risk of developing a suspected condition not confirmed as PD but exhibiting a similar age-incidence pattern Figure 1, is that some of these persons have actual, but prediagnostic, PD. Alternatively, the same etiologic agents might cause a spectrum of disease outcomes including PD and conditions confused with PD (e.g., essential tremor, vascular parkinsonism). The overall and age-specific incidence rates of suspected but unconfirmed PD should not be considered precise, because we did not attempt to identify cases using methods free of selection biases, and because they represent a mix of diseases and probably normal patterns of aging.

Although the apparent birth cohort effect on both PD incidence and age of onset Figure 2 could be interpreted in light of the events of 1916 to 1926 (encephalitis lethargica and pandemic influenza ^[20,21]), almost all HHS cohort members were living in Hawai'i at that time, and would presumably have had an equivalent chance for exposure to any pandemic infectious agent. We believe that the differences resulted from study artefact. Improved medical detection associated with formal re-examination of the entire cohort, which occurred in 1968 to 1970, 1971 to 1974, and 1991 to 1994, could explain not only the incidence divergence at age 70 to 74, but also the ``right-shifting'' of the left part of the curve for the older men. Intensive re-examination was conducted twice while the younger birth cohort was in the higher-risk age range for PD (60 to 80 years old). The older birth cohort, however, had not yet entered the maximal risk years during the second cohort re-examination, and was past it by the fourth (1991 to 1994) re-examination. That is, the cohort re-examinations may have achieved better PD case detection for younger subjects because they were conducted during the subjects' ages of greatest risk. Similarly, the left tail of the incidence curve for younger men may be explained by the occurrence of the third cohort re-examination while these men were 52 to 64 years old. The older birth cohort was not being examined while in this age range, but may have experienced an increased frequency of catch-up diagnosis when examined at older ages in the third cohort re-examination. We also examined PD risk by 1- and 5-year birth cohorts, and using different cut-off dates, but saw no evidence for any birth cohort effect (data not shown).

Under an assumption of no birth cohort effect on PD incidence, comparison of the age-incidence curves Figure 2 suggests that, because of less intensive cohort surveillance at the ages of peak risk, many true cases of PD could have been missed in older cohort members. Such persons might have died, declined follow-up evaluation, or developed obscuring medical conditions before the time of the subsequent re-screening. This possibility can be neither proved nor disproved without further follow-up of the cohort. If age-specific incidence rates between the birth cohorts are identical, the conclusion that apparent differences in ages of onset and in age-specific incidence rates resulted from untimely and incomplete diagnosis would be inescapable.

Correction of the total incidence curves to achieve the same degree of case detection as in the younger cohort would also result in an increased expected lifetime detection rate of PD in this cohort. Using the 8,006 man denominator, 92 detected PD cases would correspond to a lifetime PD occurrence rate of one in 87. Considering the additional 69 initially suspected case subjects would correspond to a one in 50 lifetime occurrence rate of suspected parkinsonian conditions. (As noted above, this latter figure is likely to be a substantial underestimate, as surveillance for parkinsonism not attributable to idiopathic PD was not as intensive as for idiopathic PD itself.) The corrected figures, using person-time denominators and projecting age-specific rates to living cohort members, would result in an estimated lifetime PD occurrence rate of one in 38, and a lifetime estimated occurrence rate of total suspected parkinsonian conditions of at least one in 31. This compares with standard U.S. estimates of lifetime occurrences of about one in 63 for PD, and one in 40 for parkinsonism of all causes. ^[7,22] Thus the HHS cohort of Japanese-American men appears to be experiencing a PD/parkinsonism occurrence rate that is at least as much as, and possibly 1 1/2 to 2 times, what has been estimated for the U.S. population as a whole. While some of this difference can be explained by Hawai'i's increased life expectancy (about 3 years longer than for all American men combined ^[23]), undoubtedly optimal case detection also plays a role.

A surveillance bias effect similar to that suspected for different age cohorts (discussed above) can be imagined for the PD prevalence estimates taken at 5-year intervals Figure 3. Although the more complete curve of 1992 prevalence must obviously reflect the effect of five additional years of high-incidence risk, with the exception of the 1917-centered both cohort (i.e., men between 62.5 and 67.5 in 1982, who are at highest age-associated risk between 1982 and 1992), the major prevalence differences for each of the five traceable birth cohorts (i.e., persons who were 62.5, 67.5, etc., in 1982) occurs in 1992, at which time prevalence in each birth cohort appears to increase Figure 3. As the third and fourth rescreenings of the entire cohort were conducted from 1971 to 1974 and from 1991 to 1994, respectively, the age-specific prevalence rates for 1992 include many newly incident cases detected as a result of more intensive case-finding. The bulk of the excess prevalent cases in 1992 may thus reflect not only expected effects of cohort aging and enhanced surveillance, but also differentially improved detection in cases of recent onset in men over age 80. This presumed effect of detection bias is a powerful reminder of the difficulties associated with studying conditions of insidious onset, for which medical care may not be sought, and which may not be diagnosed either correctly or early.

There are several other limitations of these data. Lack of a diagnostic test for PD, coupled with insidious and prolonged onset, probably make complete case detection unachievable in this or any study. Although our methods of follow-up and case detection are among the most intensive ever reported--with only five losses to follow-up of 8,006 men over 29 years--actual cases were undoubtedly missed because of nondiagnostic signs early in the course of illness, deaths due to PD and other conditions in undiagnosed case subjects, and, conceivably, differences in completeness of follow-up (and thus of the chance of detecting medical conditions like PD) for persons with PD-associated debility or dementia.

Finally, mortality associated with PD was not only substantial, but appeared to increase logarithmically with age beyond 70. Although many population data on cause-specific mortality are available, there are relatively few good comparative data on age-specific/cause-specific mortality from PD in either the levodopa or pre-levodopa era. The mean age at death for cohort members with PD who reached age 70 with or without PD was 76, approximately 4.5 years older than the mean age at death in a 1968 to 1982 series reported by Hoehn, ^[24] and 9 years older than in a 1967 pre-levodopa study. ^[25] Our mortality figures are roughly 2- to 4-fold lower at older ages than Danish population-based data recently published by Kurtzke and Murphy. ^[26] Although we cannot entirely explain these survival differences, three factors deserve consideration: (1) as life expectancy in Hawai'i is markedly longer than elsewhere in the United States, ^[23] unappreciated factors in our cohort may be influencing background mortality, PD-associated mortality, or both; (2) the Danish figures included cases of parkinsonism, ^[19] which may conceivably have increased the mortality risk (e.g., by inclusion of cases with severe vascular disease); and (3) the 1968 to 1982 Hoehn figures ^[24] may have included mortality effects from cases with onset in the pre-levodopa era.

Although we have not examined either the nature of the interaction of age and disease duration on mortality, or the actual causes and correlates of death in PD patients, between the ages of 70 and 90 cohort subjects with PD clearly face an approximate two- to three-fold increased risk of dying, a figure similar to estimates from the pre-levodopa era. ^[25] This increase in mortality corresponds to an 8-year decrease in life expectancy for cohort members. Most Americans enjoy good to excellent health in their 70s and 80s, and may find these retirement years among their most enjoyable and fulfilling; it is thus discouraging to realize that apart from the debilitating effects of the disease, PD apparently also greatly shortens life. Given the ages and length of follow-up of these cohort subjects, the effect of PD-associated mortality will probably not be fully appreciated until 5 to 10 additional years of surveillance have been accrued.

Acknowledgments

The authors acknowledge the efforts of many individuals participating in the Honolulu Heart Study and the Honolulu-Asia Aging Study, Kuakini Medical Center, Honolulu, Hawai'i; these studies are funded, respectively, by the National Heart, Lung and Blood Institute, and the National Institute on Aging, National Institutes of Health. Jan Hansen searched medical records of 8,006 study subjects. We also thank Lenora W.M. Loo and Kathy Tomasu for their help in record review. Helen Petrovitch, MD, has provided considerable research support.