Decline of Parkinson’s Disease Incidence in Israel: The Value of Cohort Analysis

Abstract

Background

The “birth cohort” effect (BCE) is the impact of the external environment on the health status of individuals born at a given time in relation to events that occurred during that timeframe. Cohort studies are used to identify the etiology of diseases and links between their risk factors. We undertook the first study to delineate the BCE in Israeli PD patients based upon big data.

Methods

We applied the US National Institutes of Health data analytic tool to perform a cohort analysis of PD data collected in the largest Israeli health services provider (Clalit Health Services) from 2002 to 2021. PD patients were divided into twelve 5-year age groups (40–44 to 95–99) for four five-year periods (2002–2006, 2007–2011, 2012–2016, and 2017–2021). The BCE was evaluated by counting the number of diagnosed patients and using cohort rate ratios (RRs) separately for male and female PD patients born between 1905–1909 and 1975–1979. The significance of the BCE was assessed by the Wald χ² test.

Results

There was a peak in diagnosed PD cases between 1919 and 1949, followed by a steady decline in PD incidence rates (RRs), beginning with the 1905–1909 cohorts and continuing through the 1975–1979 cohorts in both sexes. The cohort RRs decreased from 12.94 (CI: 7.76–21.55) for females and 10.83 (CI: 7.52–15.59) for males in the 1905-09 cohorts to 0.35 (CI: 0.15–0.83) and 0.45 (CI: 0.28–0.74), respectively, in the 1975-79 cohorts. A 5-year lag in BCE was found in females compared to males.

Conclusion

The BCE was characterized by a reduction in the incidence of Israeli PD patients, possibly reflecting epidemiological and sociodemographic changes that occurred in Israel.

Introduction

Parkinson’s disease (PD) is a progressive neurodegenerative disease characterized by impaired motor function, impaired balance, and injurious falls, as well as non-motor symptoms predisposing to mental and behavioral disorders, long-term disability, and premature death. In 2017 alone, the prevalence of diagnosed PD in the USA was estimated at 1 million patients with a total annual economic burden of USD 51.9 billion [1]. The number of people with PD in China was estimated at 5.08 million in 2021. Their PD rates were growing at an accelerated pace compared with 1990, with the number of new cases increasing by 455.7% and by 678.9% for all patients taken together in 2021 [2]. Due to its increasing prevalence, some authors describe PD as a major global health burden [3] and even consider PD as representing a world challenge [4].

Several factors that protect against PD have been identified, among them a strong inverse correlation between tobacco smoking and the risk of developing PD [5]. Both cohort studies and meta-analyses confirmed that smokers have a significantly lower risk of PD than nonsmokers, and that this effect is dose-dependent [6]. In addition, regular consumption of tea and coffee may protect against the development of PD, with the caffeine found in these beverages being considered the main active component [7, 8]. Finally, both the Mediterranean and the PRO-21 diets have been linked to better PD outcomes [9].

In contrast to the above-cited protective factors, numerous hereditary and environmental factors have been recognized as acting to promote the development of PD and/or parkinsonism. Genetic mutations with their epigenetic modulation of gene expression through environmental factors, as well as some infectious agents [10], exposure to chemicals, industrial and agricultural neurotoxins such as manganese, trichloroethylene, paraquat, or rotenone [11–13], air pollution containing fine and ultrafine particles [14], and even solar radiation [15], may contribute to disease pathogenesis. Other risk factors include specific medication use, such as antipsychotics, dopamine receptor blockers [16], and β-blockers [17, 18], as well as some antiepileptic drugs [19] which have been associated with the development or worsening of PD. Lifestyle influences, such as malnutrition or excess milk consumption [20], metabolic syndrome [21], and substance (methcathinone “ephedrone”) use [22, 23], were found to contribute to the risk and severity of PD. Common lifestyle-related vascular risk factors, such as hypertension and type 2 diabetes, were also shown to increase the risk of developing PD [24–29].

The birth cohort effect (BCE) is a unique historical experience of groups of people that could determine their vulnerability to PD depending on the time of birth. It is the combination and strength of exposure to factors that were significant during their childhood or even in utero, for example, large-scale outbreaks of infectious diseases, exposure to certain toxins, periods of air pollution, radiation hazard, malnutrition, etc.

BCE is a component of age-period-cohort (A-P-C) analysis and represents an important tool for uncovering effects on disease incidence, independent of the other capabilities of this method (age and period), as well as other meaningful factors that may be omitted in cross-sectional studies. A-P-C analysis uses birth cohorts to understand long-term trends by separating these three different factors, providing more insight than cross-sectional studies [30].

In 1961, Poskanzer and Schwab [31] were the first to note the BCE in PD, and they proposed that PD is the result of the encephalitis lethargica pandemic that raged from 1917 to 1927 and disappeared by 1931. Those authors suggested that cohorts who had experienced mild or subclinical forms of the infection suffered delayed nervous system damage that manifested as PD 30–40 years later and predicted its disappearance as a common clinical entity by 1980 [32].

Further analysis of APCs revealed significant BCE for PD in the work of Ajdacic-Gross et al. [33] in 2012. Based upon 17,544 death certificates registered between 1921 and 2008, those authors found that individuals born between 1870 and 1920 had an increased risk of developing PD, and that mortality from PD increased both in the 1920s–1930s and more recently, beginning in the 1990s. The Rochester Epidemiology Project (1976–2005) found significant BCE on PD incidence, but only in men, identifying the 1920 birth cohort as having a higher risk. While the overall incidence of PD and parkinsonism increased between 1976 and 2005, particularly in men and older individuals, that increase was linked to a BCE for men born in the 1920s, indicating that incidence rates were predicted to decline in the future for this demographic group [34].

Although it is reasonable to expect that the number of patients with PD should increase as life expectancy of the global population and the proportion of elderly people increase, there is currently no consensus regarding the temporal trends in the incidence of PD [35, 36]. Some recent studies have found a downward trend in incidence rates of PD [37–42], while other investigations showed a stable state [43] or an increased incidence [34, 44–46].

We had earlier confirmed a progressive downward trend in the incidence of PD in Israel and discovered a previously unknown perplexing decline in the age-adjusted incidence rate of PD in the elderly after a median age of 82.5 years in males and 77.5 years in females [47]. In this study, we sought to determine whether the observed decline in PD incidence was due to shared life experiences and environmental exposures that may have influenced health status in 15 five-year birth cohorts from 1905 to 1975, and whether cohort-specific factors could be identified [48]. In this work, we examined the ratios of age-specific rates in PD patients born in a specific 5-year periods (the cohorts), in contrast to our previous study [23], in which each of the age groups was examined simultaneously for the entire 20-year period.

Methods

Study Design and Data Source

This analysis included data from the Clalit Health Services (CHS), the largest health services provider in Israel, insuring more than four million members and covering at least one-half of all PD patients registered in Israel from 2002 to 2021. Information about the CHS database and the criteria of identification of PD cases have been described in detail elsewhere [47]. Briefly, the data were collected from the chief CHS research data-sharing platform powered by the MD Clone (https://www.mdclone.com) and based upon ICD-10 diagnostic codes. The personnel of the Kaplan Medical Center’s CHS Data Research Center (see Acknowledgments) extracted the data that contained anonymized information about the age, sex, and dates of PD diagnosis from January 1, 2002, to December 31, 2021. All cases diagnosed before 2002 were excluded from analysis. The patients were identified as having PD by two essential inclusion criteria: (1) being listed by the International Classification of Diseases, 10 Revision, clinical code for PD (G20) performed by neurologists of the CHS or included in medical records of the patients by other medical specialists in both hospital and outpatient settings; (2) having a minimum of two purchases of prescribed anticholinergic drugs, MAO-B inhibitors, dopamine agonists, levodopa, or amantadine (drug-tracing approach) [49].

Data Analysis

The collected data were assessed by means of the US National Cancer Institute A-P-C analysis web tool (https://analysistools.nci.nih.gov/apc/ [50]). The A-P-C analysis is a parametric statistical method providing information about the effects of patient age, period, and cohort on trends in PD rates¹. The collected data on PD cases were entered into Microsoft Excel 2016 spreadsheets and then exported to the above-mentioned analysis tool. The input data as well as the population counts for twelve 5-year age groups (40–44 to 95–99) and four 5-year periods (2002–2006, 2007–2011, 2012–2016, and 2017–2021) were extracted separately for males and females and compared with age-matched Israeli populations for the same periods. Cohort rate ratios (CRRs) are the ratios of age-specific rates in each cohort relative to the reference cohort. The reference cohort was defined as the central value if the number of categories is odd and the lower of the two central values if it is even. The cohort analysis functions showed the pace of change in RRs in fifteen birth cohorts of PD patients from 1905–1909 to 1975–1979. The CRRs were the ratios of age-specific rates in each cohort relative to the reference cohort.

The two-tailed Wald χ² test was used for the hypothesis proof (the null hypothesis suggested that all CRR = 1). The p values <0.05 were considered significant.

Results

This study included 41,760 patients with PD, of whom 22,830 were men (54.7%). The average age at diagnosis was 75.5 ± 10.3 years. Three-hundred and fifteen individuals were excluded due to the small size of each cohort aged 0 to 39 years (280 cases in total) and cohorts aged >100 years (35 cases) and the inability to process them using the web tool.

The distribution of incident PD cases across 15 birth cohorts is presented in Table 1. Our dataset revealed a cohort peak in both sexes, beginning in the 1910–1914 cohorts and reaching an apex in the 1925–1939 cohorts, when the number of identified PD diagnoses was extremely high. Beginning with the cohorts born in 1940–1944, there was a decrease in the number of diagnoses, which continued until number of diagnoses in the cohort born in 1975–1979 (Table 1; Fig. 1a, b). Thus, the combined number of patients with PD born between 1925 and 1939 alone comprised 20,422 cases, which is 48.9% of the total patient population in 15 cohorts, confirming the huge magnitude of the above-mentioned peak.

Table 1.

RRs and number of diagnosed cases of PD distributed across 15 birth cohorts according to CHS data

	Males (n = 22,830)		Females (n = 18,930)
	RR	95% CI	RR	95% CI
Cohort
1905–1909	10.83, n = 86	7.52–15.60	12.94, n = 85	7.76–21.55
1910–1914	7.31, n = 488	6.23–8.59	8.76, n = 465	6.96–11.03
1915–1919	4.25, n = 1,145	3.83–4.71	5.69, n = 1,040	4.88–6.62
1920–1924	2.83, n = 2,812	2.61–3.05	3.46, n = 2,544	3.08–3.89
1925–1929	2.05, n = 3,572	1.92–2.19	2.44, n = 3,373	2.21–2.70
1930–1934	1.56, n = 3,839	1.47–1.66	1.79, n = 3,345	1.63–1.96
1935–1939	1.21, n = 3,473	1.14–1.28	1.28, n = 2,820	1.17–1.41
1940–1944	1.00, n = 2,408	1.00–1.00	1.00, n = 1,770	1.00–1.00
1945–1949	0.84, n = 2,256	0.78–0.89	0.80, n = 1,548	0.72–0.89
1950–1954	0.76, n = 1,440	0.70–0.83	0.68, n = 986	0.59–0.78
1955–1959	0.66, n = 700	0.59–0.74	0.62, n = 489	0.52–0.75
1960–1964	0.59, n = 343	0.50–0.70	0.63, n = 284	0.49–0.81
1965–1969	0.53, n = 151	0.42–0.66	0.52, n = 104	0.36–0.74
1970–1974	0.41, n = 83	0.30–0.58	0.45, n = 54	0.27–0.75
1975–1979	0.45, n = 34	0.28–0.74	0.35, n = 23	0.15–0.83

	Wald χ2 tests for estimable functions
	males			females
Null hypothesis	χ2	d.f.	p value	χ2	d.f.	p value
All CRR = 1	1,415.4	14	<0.00001	870.9	14	<0.00001

95% CI, 95% confidence interval; d.f., degree of freedom; RR, rate ratio.

Fig. 1.

a CRR with 95% CI (shaded area) in males. b CRR with 95% CI (shaded area) in females. The figures show a steep decline in the RRs for persons born before 1930. Cohort RR, cohort rate ratio; 95% CI, 95% confidence interval as shaded areas.

In accordance with the above findings, we observed a continuous decline in the CRRs of PD during the entire period from 2002 to 2021, beginning from the 1905–1909 births and continuing to the group of 1975–1979 births (Table 1). The main decline of PD RR was recorded for PD patients of both sexes in the 1905–1909 cohort, and it was more pronounced in females. This decline was especially steep until 1925–1929 after which it continued to gradually lower over time (Fig. 1a, b).

The RRs decreased from 12.94 (confidence interval [CI]: 7.76–21.55) for females and 10.83 (CI: 7.52–15.59) for males recorded for in the 1907 cohort to 0.35 (CI: 0.15–0.83) and 0.45 (CI: 0.28–0.74), respectively, recorded for the 1977 cohort, indicating that the later the subjects were born, the lower were their RRs for PD (Fig. 1a, b). Interestingly, the birth cohort changes in males outpaced those changes in females by 5 years. The Wald χ² test results were highly significant for both RRs and CDs in males and females, thereby rejecting the null hypotheses (Table 1).

Discussion

The findings of this analysis confirm those of our previous report that the incidence of PD in Israel has fallen gradually during the past century, reaching unexpectantly very low levels. One possible explanation is that this is an “artifact,” and that the reduction in PD incidence will disappear over the next decades; i.e., some of those currently unaffected people with birthdates from 1940 onward will also develop PD. However, we realize that this possibility implies that the preclinical stage of PD may be considerably longer than is currently accepted and may have become longer over the past decades for some undefined reasons.

Our cohort analysis demonstrated a pronounced BCE in Israeli PD patients starting from those born in 1905–1909 and which continued in both sexes throughout the observed period and was especially robust until 1925–1929 (Fig. 1a, b). These data suggest a continuous decline in CRRs (Table 1). The amplitude of that peak contrasted sharply with the measurements of subsequent age groups, which could indicate that the causative factor(s) were temporary and significantly weakened soon after 1949.

In our previous work on age-period analysis, we found that the incidence of PD increased with age only until about age 80 years, after which it gradually declined [47]. However, the more detailed analysis performed in the current study demonstrated that this age effect also depended upon the year of birth. The age-adjusted incidence rate was highest for PD patients belonging to the oldest-old group. The same effect was also seen in later ages, although to a lesser degree, until it disappeared altogether in the younger groups [47].

One possible reason that should be discussed is an environmental factor. The oldest population in our cohort were living in the period of the encephalitis lethargica and may have been exposed to the virus, which could have reduced the number of dopaminergic cells in their substantia nigra, leading to vulnerability to other factors. If this were the case, we would expect a sharp decline in the incidence in those born after the late twentieths of the last century, which was not nevertheless seen. However, exposure to other unknown substances that may have an anti-dopaminergic effect cannot be excluded.

It is possible that the reductions in PD incidence that we identified are a result of the disappearance of a large number of patients who had sustained some infectious disease of the central nervous system and/or its consequences. The rapid and dramatic decline in а number of PD patients and RR observed after 1949 indicates that the risk factors responsible for the peak in incidence were significantly reduced. Rod et al. [51] observed a similar peak in PD diagnoses among Danish PD patients born in the prewar years followed by a decline in the 1940–1950s. In the 1960s, Poskanzer and Schwab [31] postulated that this phenomenon could be postencephalitic parkinsonism following the encephalitis lethargica pandemic of 1917–1927. These authors believed that a decline in the PD rate was beginning at the time of their report and that it should disappear in the 1980s [32] since many of those who survived the initial illness were left with long-term neurological deficits that did not become apparent until years or even decades later [52]. Although this hypothesis was not verified by a study by Estupionen et al. [53], it is possible that other types of infections played a role, such as childhood infections or seasonal waves of influenza. Stolzenberg et al. [54] found that the common norovirus infection in the gastrointestinal tract of children leads to increased levels of α-synuclein in enteric neurons, which may help mobilize the immune response to fight pathogens. Those authors hypothesized that infections with certain enteric pathogens may represent a risk factor for PD, due to the putative triggering of increased neuronal and extracellular α-synuclein. In a case-control study conducted in Denmark, influenza was significantly associated with PD more than 10 years after the acute stage of the disease, which supports the hypothesis of a role of the influenza virus in the development of neurodegenerative diseases [55]. Since the hypothesis of a direct viral etiology of PD has not been convincingly confirmed, the most promising assumption is that of a postinfectious autoimmune genesis of PD based upon the concept of “molecular mimicry” which is supported by updated data [56, 57].

Our data showed a delay of approximately 5 years in epidemiological changes in females. This could be due to a lower incidence of the disease compared to males, as well as other factors that slow the progression of PD in females, such as hormonal influences or lower exposure to environmental toxins.

Limitations and Strengths of the Study

Although our study used a robust cohort analysis and a comprehensive CHS database to investigate PD trends in Israel, CHS data are limited in establishing definitive causal relationships between observed trends and possible influencing factors. These data are compiled from a variety of sources (e.g., ambulatory services and specialized treatment centers) and may vary in quality and completeness. Our reliance upon reported diagnoses may have introduced biases influenced by differences in healthcare and coding practices over time. Data sources that differentiate between early-onset and late-onset PD should also be considered in order to understand the significance of timing in the setting of PD as well as improved imaging methods for diagnosing the presence of Lewy bodies in the nervous system [58] in patients with certain symptoms of hereditary and sporadic parkinsonism, given that it is well recognized that PD is actually a syndrome with many forms and causes [59].

Our data did not include genetic information, which could have identified subgroups. We also did not consider other possible causes, such as a decrease in the number of individuals working in agriculture and an increase in the efforts against air pollution. Lifestyle changes, such as drinking more purified tap water instead of well water, greater physical activity, and weight training, may also have produced some anti-Parkinsonism effects. Tobacco smoking is recognized as a means to counteract the onset of PD and its progression, and quitting smoking, which has occurred worldwide in recent decades, may affect PD epidemiology [60]. However, unlike most other countries, smoking habits in Israel have not changed in recent years.

One important limitation of this study that should be noted is that not all birth cohorts were examined equally across all age groups. Later birth cohorts may not have reached the age at which PD is most frequently diagnosed during the study period. This may artificially lower the incidence rate in later cohorts. However, since a decline in incidence is also observed in cohorts with overlapping ages (1925–1954), the observed incidence rate may be real, but may be overestimated by the earliest birth cohorts and underestimated by the latest birth cohorts.

Also we were unable to assess the combined influence of the entire spectrum of positive and negative influences on the development of PD in a given birth cohort of individuals, since the listed factors have their own power, time of activation, and duration of action. In addition, Israel’s population has changed substantially since the State’s founding due to massive immigration from many countries of Europe, Africa, and Asia, which brought their own genetics and customs. The subsequent significant demographic changes within the Israeli society in the 21st century complicated the assessment of the influence of the above factors on the development of PD. Additional studies are needed to explore the complex interaction of genetic, epigenetic, and environmental factors, as well as to establish definitive causal relationships between the observed trend and potential influencers. All that said, our study is the first to show the presence of BCE in PD patients in Israel, explaining the observed RR changes by probable events, and pointing to possible etiological causes of at least some types of parkinsonism.

Conclusion

Our analysis revealed a sharp decline in the incidence of PD in Israel across all birth cohorts, despite the peak in diagnosed cases occurring in cohorts before 1939. We believe that this is, at least in part, a result of a decline in the incidence of PD among older Israelis due to possible cessation of exposure to infectious agents, improved environmental conditions, and demographic changes in Israeli society. Further studies are needed to explore the processes underlying these observations and to better understand the factors influencing the incidence of PD in further cohorts in Israel and elsewhere.

Acknowledgments

We would like to thank personnel of the Kaplan Medical Center’s CHS Data Research Center and personally Tamar Zohar, for providing the primary data on Parkinson’s disease patients which made the A-P-C analysis possible. Also we would like to thank Dr. Philip Rosenberg from the Biostatistics Branch, Division of Cancer Epidemiology and Genetics, Department of Health and Human Services, National Cancer Institute, NIH, Bethesda, MD, USA, one of the authors of the A-P-C Web tool, for his important advice in interpreting the results of this study. In addition, the authors are grateful to Esther Eshkol, MA, medical and scientific copyeditor, for editorial assistance.

Statement of Ethics

This study was approved by the Kaplan Medical Center Institutional Review Board (Protocol No. KMC-0071-22). It was based on anonymous databases and involved no direct interaction with patients. Informed consent was not required for this study, which used anonymized data.

Conflict of Interest Statement

Prof. Amos D. Korczyn is a Neuroepidemiology Review Board Member. Other authors have no conflicts of interest to declare.

Funding Sources

The authors declare no funding was obtained or used for this study.

Author Contributions

Y.B.: study concept and design; analysis and interpretation of data; and drafting/revision of the manuscript for content. R.G. and A.E.: revision of the manuscript for content. A.D.K.: study concept and design and analysis; and revision of the manuscript for content.

Funding Statement

The authors declare no funding was obtained or used for this study.

Data Availability Statement

The primary data for the findings of this study are unavailable due to the CHS restrictions and requirements for anonymity.