The Rotterdam Study
July 21, 2016
Abstract
We investigated trends in the incidence of parkinsonism and Parkinson disease (PD) by comparing data from the first 2 subcohorts of the Rotterdam Study, a prospective, population-based cohort study (first subcohort: baseline 1990 with 10 years of follow-up; second subcohort, baseline 2000 with 10 years of follow-up). From the baseline years, we observed differences in the second subcohort that were associated with a lower risk of PD for some but not all baseline risk factors. Participants in both subcohorts were followed for a maximum of 10 years and monitored for the onset of parkinsonism, the onset of dementia, or death, until January 1, 2011. We used Poisson regression models to compare the incidences of parkinsonism, both overall and by cause (PD and secondary causes), and competitive events (incident dementia and death) as well as the mortality of parkinsonism patients in the 2 subcohorts. In the 1990 subcohort, there were 182 cases of parkinsonism (84 of which were PD) during 57,052 person-years. In the 2000 subcohort, we observed 28 cases of parkinsonism (10 with PD) during 22,307 person-years. The overall age- and sex-adjusted incidence of parkinsonism was lower in the 2000 subcohort (incidence rate ratio = 0.55, 95% confidence interval: 0.36, 0.81), and PD incidence declined sharply (incidence rate ratio = 0.39, 95% confidence interval: 0.19, 0.72). Competitive event rates were lower in the 2000 subcohort, and mortality rates among persons with parkinsonism remained stable. These findings suggest that the incidence of parkinsonism in general, and of PD in particular, decreased between 1990 and 2011.
Introduction
The prevalence of parkinsonism in general and of Parkinson disease (PD) in particular is relatively low among persons younger than 50 years of age and increases sharply with age in community-dwelling elderly.[1] As the size of elderly populations increases, PD prevalence is expected to grow,[2] causing a dramatic projected increase in the burden of PD.[3] However, PD prevalence rates have remained relatively stable during recent decades,[4] suggesting that PD incidence rates might have declined, although this has never been shown empirically.
In recent decades, preventive treatment strategies for reducing the prevalence of vascular risk factors at the population level were followed by a decrease in cerebrovascular disease incidence rates.[5] Dementia incidence rates, which could be affected by similar risk factors, have also been shown to have declined in community-dwelling persons.[6] A variety of factors are associated with the risk of PD,[7] but there is little evidence regarding the causality of these associations. Factors affecting oxidative damage—such as serum cholesterol, use of lipid-lowering medication, and caffeine intake—might influence the risk of parkinsonism secondary to cerebrovascular disease as well as the most common causes of dementia, and such factors might convey similar protective influences on the risk for PD.[8–10] In contrast, other risk factors, such as serum urate and smoking, have opposite associations with the risk of PD compared with dementia and stroke.[11–14] An investigation of trends in the prevalence of risk factors and subsequent incidence of parkinsonism and PD could provide essential insight into the future burden of PD.
We investigated whether risk factors for and incidence rates of parkinsonism and PD changed over the last 2 decades by comparing a subcohort of individuals aged 55 years or older in 1990 with a subcohort followed from 2000 onward in the Rotterdam Study.
Methods
Study Population
The present study was performed on participants in the first (baseline 1990) and second (baseline 2000) subcohorts of the Rotterdam Study, a large, prospective, population-based cohort study conducted among middle-aged and elderly persons in Ommoord, a district of Rotterdam, the Netherlands.[15,16] The study was approved by the Medical Ethics Committee of Erasmus MC University Medical Center Rotterdam.
Of the 10,275 persons eligible for the first subcohort (all 55 years of age or older), 7,983 (78%) agreed to participate and signed informed consent statements. Participants were interviewed at home and subsequently invited for in-person screening for parkinsonism at our research center; 1,030 persons (13%) did not undergo in-person screening because of refusal, disease, or death.[17] For the 2000 subcohort, 4,472 persons who had attained 55 years of age or moved into the study district since the start of the study were invited; 3,011 (67%) agreed to participate and signed informed consent statements. Of these persons, 550 (18%) did not undergo in-person screening for parkinsonism.
For this report, we excluded persons with prevalent parkinsonism or prevalent PD. For PD, we considered persons with dementia before onset of parkinsonism to be no longer at risk. Consequently, for PD analyses we excluded 277 persons with baseline dementia (259 (3.8%) in the 1990 subcohort, 18 (0.7%) in the 2000 subcohort) as well as one individual with unknown baseline dementia status. There were 128 prevalent cases of parkinsonism in the 1990 subcohort (1.8% of those at risk) and 13 in the 2000 subcohort (0.5% of those at risk). Of prevalent cases of parkinsonism, 95 had PD in the 1990 cohort (1.4% of those at risk), and there were 7 prevalent cases of PD in the 2000 subcohort (0.3%). In addition, 81 individuals retracted informed consent for further data collection during follow-up, leaving 6,752 persons at risk of parkinsonism (of whom 6,492 were also at risk of PD) in the 1990 subcohort and 2,440 (2,422 for PD) in the 2000 subcohort. Follow-up evaluation was virtually complete as of January 1, 2011.
Baseline Screening and Characteristics
At baseline, we used a 2-phase design, previously described in detail,[17] to identify subjects with parkinsonism or PD. In short, all participants were first screened at the research center for signs of parkinsonism (including hypo- or bradykinesia, resting tremor, postural instability, and cogwheel rigidity). Individuals with a positive screening result received a structured clinical examination to establish parkinsonism (including the motor examination of the Unified PD Rating Scale,[18] neurological examination, and gathering of medical history) by a research physician who was an expert in neurological disorders. We obtained additional information from medical records of specialists and general practitioners. Persons who were suspected of having PD were further evaluated by an experienced neurologist either in person, using a recorded video of their visit to the research center, or using a case report compiled as part of the study.
We also studied the baseline prevalence of risk factors for PD (smoking status, coffee intake, serum urate level, serum cholesterol level, and use of antilipid medication) and of common causes of secondary parkinsonism (primary dementia diagnosis, cerebrovascular disease, and antipsychotic medication use). (See Web Appendix 1, available at http://aje.oxfordjournals.org/, for details.)
Incident Parkinsonism
We used 4 overlapping modalities to detect potential cases of parkinsonism during follow-up: in-person screening, self-reporting of PD during in-person interviews, antiparkinsonian medication use, and clinical monitoring alerts.
Participants were screened every 4 years using the same 2-phase design.[19] In short, participants were first screened for signs of parkinsonism using standardized methods. Participants with a positive screen received a structured examination by a research physician with expertise in neurological disorders to establish parkinsonism (defined as hypo- or bradykinesia plus at least 1 other cardinal sign: resting tremor, rigidity, or postural instability in the absence of other likely causes). In addition, the cohort was continuously monitored through a surveillance system for detection of new parkinsonism cases by computer linkage with the general practitioners' automated medical record systems, which encompass diagnostic codes and narrative clinical notes from general practitioners as well as documentation from neurologists, geriatricians, and other medical specialists. Nearly all participants in the study were registered at 1 or more of the 7 community pharmacies that serve the study area, which made it possible to identify subjects who used antiparkinsonian medication at any time during follow-up. Finally, during visits to the research center, participants were asked whether they had ever been diagnosed with PD.
For all persons whose cases were detected by any of these methods, complete medical records were studied and case reports were compiled that gathered all potentially relevant information to establish the subtype of parkinsonism and the degree of certainty in the diagnosis. These case reports were evaluated by a panel led by an experienced neurologist using consistent criteria (Web Appendix 2). The diagnostic criteria for secondary parkinsonism (e.g., secondary to dementia, medication-induced parkinsonism, vascular parkinsonism, multiple system atrophy, or progressive supranuclear palsy) have been reported previously[6,17,20] and are summarized in Web Appendix 3.
End of Follow-up
Person-time at risk for incident parkinsonism ended at the first of the following: a diagnosis of parkinsonism, death, 10 years after baseline, or January 1, 2011. Person-time at risk for incident PD ended at the first of any of the above or at the onset of dementia. Because of the substantial overlap between the 4 detection methods, the consistently high response rates for follow-up in-person examinations, and the fixed maximum time period of 10 years, we considered persons who were not screened in person during one of the follow-up rounds to be still at risk of parkinsonism and PD. For age at onset of parkinsonism, we took the age at the midpoint between the date on which parkinsonism was first observed (either during in-person screening or in medical records) and the date of the preceding in-person examination.
Statistical Analysis
We compared the prevalence of risk factors for PD and common causes of secondary parkinsonism in the 1990 and 2000 subcohorts using linear regression for continuous variables and logistic regression for dichotomous variables, with adjustment for age and sex. Fisher's exact test was used for dichotomous variables if a percentage was zero.
We calculated person-years at risk per 10-year age group by adding each participant's contribution of follow-up time within that age group; individuals could contribute person-time to more than 1 age group. We estimated incidence rates and rate ratios between subcohorts for any parkinsonism, PD, secondary parkinsonism, and a composite of competing events (the first of death or incident dementia) per 10-year age group. We used Poisson regression models with person-time at risk as the offset variable and additional adjustment for sex and starting age within each age stratum. We repeated analyses of overall parkinsonism incidence rates after excluding possible cases of parkinsonism (Web Appendix 3 http://aje.oxfordjournals.org/content/183/11/1018/suppl/DC1). We also repeated overall PD incidence rate analyses after narrowing diagnostic criteria for PD (i.e., excluding patients whose PD diagnosis was confirmed by a neurologist but for whom we had no information about response to antiparkinsonian medication consistent with PD). In sensitivity analyses, we compared age-stratified PD incidence rate ratio estimates obtained by Poisson regression with the same estimates obtained by the model of Fine and Gray,[21] which takes into account competitive events (first of death or incident dementia). Separately, we estimated overall incidence rates and incidence rate ratios for parkinsonism by causes other than PD, adjusting for age and sex. Also, we calculated mortality rates and age- and sex-adjusted mortality rate ratios for parkinsonism patients in both subcohorts, in 10-year age strata. For this analysis, the time scale ran from the onset of parkinsonism to death, 10 years after baseline, or January 1, 2011.
Results
Baseline characteristics of participants at risk of PD are presented in Table 1 . Compared with the 1990 subcohort, the age distribution in the 2000 subcohort was skewed toward younger participants (mean age of 64.5 years in the 2000 subcohort vs. 69.3 years in the 1990 subcohort; P < 0.001). After stratification by age group, residual age differences between subcohorts persisted in the age strata 55–64 years and 75–84 years. The 2000 subcohort generally comprised more ever smokers but fewer current smokers at the time of study entry. Caffeinated coffee intake was significantly higher in the 55–64 years and 75–84 years age strata and nonsignificantly higher in the 65–74 years age stratum, but it was lower among the oldest participants. Serum urate levels were lower in 3 of 4 age strata, but higher in the 65–74 years age stratum. The use of lipid-lowering medication was more prevalent in the 2000 subcohort, and serum cholesterol levels were lower. There were fewer antipsychotic medication users in the 2000 subcohort. Stroke prevalence was similar in all strata except in the youngest participants, in whom it was higher, whereas dementia prevalence was lower in all age strata, although this difference was not significant in the 65–74 years age stratum. Exclusion of individuals with prevalent dementia did not meaningfully alter differences in baseline characteristics between subcohorts.
In the 1990 subcohort, after 57,052 person-years of follow-up, 182 persons were diagnosed with incident parkinsonism. In the 2000 subcohort, 28 incident cases of parkinsonism were observed during 22,307 person-years of follow-up. Compared with the 1990 subcohort, parkinsonism incidence rates were lower in the 2000 subcohort (incidence rate ratio (IRR) = 0.55, 95% confidence interval (CI): 0.36, 0.81). The age difference at baseline between participants in both subcohorts was paralleled by a difference in mean age at onset of parkinsonism (mean age 78.0 years vs. 74.9 years; P = 0.68). Stratification by sex showed lower overall parkinsonism incidence estimates in both women and, although nonsignificant, men (IRR in women = 0.40, 95% CI: 0.20, 0.71; IRR in men = 0.74, 95% CI: 0.42, 1.22). Formal testing showed no significant statistical interaction between subcohort and sex for incidence (P = 0.17). Age-specific parkinsonism incidence was much lower in the highest age category and, albeit nonsignificantly, in the 65–74 years and 75–84 years age strata ( Table 2 ). The proportion of possible cases of parkinsonism among all parkinsonism cases was lower in the 2000 subcohort (7%) than in the 1990 subcohort (30%). After exclusion of possible cases of parkinsonism, the overall parkinsonism incidence rate ratio between subcohorts was 0.71 (95% CI: 0.46, 1.08).
As shown in Table 3 , the incidences of PD and parkinsonism due to vascular causes were remarkably lower, while the incidence of parkinsonism due to other rare causes was higher. There were no significant differences in the incidence of parkinsonism secondary to dementia or multiple system atrophy, whereas the incidences of medication-induced parkinsonism and of persons without a clear cause for parkinsonism were lower. Incident PD rate ratio estimates remained similar when competing risks were taken into account (Web Table 1).
Participants in the 1990 subcohort amassed 55,920 person-years at risk for PD, during which 84 persons developed incident PD. In the 2000 subcohort, there were 10 PD cases in 22,224 person-years at risk. The overall incidence of PD was remarkably lower in the 2000 subcohort (IRR = 0.39, 95% CI: 0.19, 0.72) ( Table 3 ). As shown in Table 4 , the incidence of PD was lower among all age groups. Stratified analyses showed a more pronounced but not significantly different lower overall PD incidence in women than in men (IRR in women = 0.18, 95% CI: 0.03, 0.58; IRR in men = 0.57, 95% CI: 0.25, 1.16; P for interaction by sex = 0.17). Forty-one of 84 PD cases (49%) in the 1990 subcohort fulfilled narrower diagnostic criteria for PD, and 6 of 10 PD cases (60%) in the 2000 subcohort fulfilled those criteria. We had no DaTscan (GE Healthcare, Cardiff, United Kingdom) data for any of these patients. Using narrower diagnostic criteria for PD did not alter the observed overall lower PD incidence (overall age- and sex-adjusted IRR = 0.41, 95% CI: 0.16, 0.91).
In the 1990 subcohort, 281 participants were diagnosed with incident dementia (4%), and 1,521 (23%) died while at risk for PD. Of persons at risk for PD in the 2000 subcohort, 73 (3%) were diagnosed with dementia before parkinsonism onset and 324 (13%) died. Overall, the composite competitive risk was lower in the 2000 subcohort (IRR = 0.79, 95% CI: 0.70, 0.88). As shown in Table 5 , the age- and sex-adjusted composite competitive event rate estimates for each 10-year stratum were 20%–24% lower in the 2000 subcohort than in the 1990 subcohort.
In the 1990 subcohort, 100 persons with incident parkinsonism died after 885 person-years with parkinsonism (mortality rate = 113.0, 95% CI: 92.0, 137.5). In the 2000 subcohort, 13 patients with incident parkinsonism died after 114 person-years with parkinsonism (mortality rate = 113.6, 95% CI: 60.5, 194.2). The age- and sex-adjusted mortality rate ratio was 1.37 (95% CI: 0.71, 2.45).
Discussion
Comparing 2 subcohorts of the population-based Rotterdam Study, we found that the incidence of parkinsonism in the general population declined over the course of 20 years. PD remains the most frequently recognized cause of parkinsonism, and its incidence has decreased sharply.
To date, 3 articles have been published on trends in PD incidence in white populations after 1990, reporting stable incidence rates[22,23] and a small decrease in PD incidence.[24] However, those studies relied largely on general practitioner records and diagnosis codes, methods that have a high risk of misclassification of cases and that cannot distinguish between secondary parkinsonism and PD.[25] Strengths of our study include its population-based design, extensive case-detection methods (which yielded detailed clinical information on causes of parkinsonism), and standardized assessment for both parkinsonism and PD throughout the study period.
Limitations of our study include the low number of PD cases in the 2000 subcohort and the lack of histological confirmation of parkinsonism subtype diagnoses. In addition, our case-finding methods were directed mainly at detecting PD, which might have caused an underestimate of parkinsonism due to causes other than PD in both subcohorts. Also, we lacked data on some known risk factors for PD (e.g., exposure to pesticides), and our study population comprised almost exclusively white people, which might limit the generalizability of our findings to other populations.
We considered 5 possible explanations for the remarkable decline in parkinsonism and PD incidence. First, a lower prevalence of risk factors in the 2000 subcohort could have caused a drop in incidence rates of parkinsonism and PD in the following decade. In participants in the younger 3 age strata, we observed higher caffeinated coffee intake, which has been associated with a lower risk of PD.[10]Similarly, there was a rise in the use of lipid-lowering medication and a concomitant drop in cholesterol levels, which might reduce the risk of PD.[8,9] However, we observed lower serum urate levels, which has been associated with a higher risk of PD.[12] In addition, prospective cohort studies have consistently shown an inverse association between smoking and the risk of incident PD, and the association is stronger for current smoking than for former smoking.[14] In the present study, we observed a higher prevalence of ever smoking in the 2000 subcohort, but the proportion of current smokers was lower, suggesting that changes in baseline risk profiles were not uniform. More importantly, evidence supporting the causality of the association of the above risk factors with incident PD is limited, especially for smoking. In addition, even assuming causal associations with PD for risk factors such as serum urate and caffeine, there is a lack of data on the population attributable risk they represent for PD. To further understand the mechanisms responsible for the decline in parkinsonism and PD incidence, we need high-quality data on trajectories toward the onset of clinical PD.
Second, the decline in parkinsonism and PD incidences could merely reflect a change in application of diagnostic criteria. Although case-detection methods for parkinsonism and PD within the Rotterdam Study remained similar throughout the study period, PD diagnosis in the research setting relies in part on exclusion of secondary causes of parkinsonism.[26] Therefore, we examined the prevalence of secondary causes of parkinsonism and the incidence of parkinsonism by cause. Lower baseline antipsychotic medication use was paralleled by a lower incidence of medication-induced parkinsonism, while the lack of a major, unidirectional shift in prevalence rates of stroke within our population and the previously reported decrease in stroke incidence in the Rotterdam Study[5] were followed by a decrease in vascular parkinsonism. Furthermore, the increase in rare causes of parkinsonism and the decrease in unspecified parkinsonism might reflect advanced availability of diagnostic modalities and better understanding of secondary parkinsonism by neurologists. Therefore, we repeated our analyses after excluding patients whose PD diagnosis was confirmed by a neurologist but for whom we had no information about response to antiparkinsonian medication consistent with PD. We observed a similar decline in PD incidence, suggesting that these changes did not account for the decline in PD incidence. We also compared overall parkinsonism incidence between subcohorts after exclusion of possible cases of parkinsonism, which rendered the difference between subcohorts no longer significant. However, we note that the incidence rate ratio between subcohorts (IRR = 0.71) was not vastly different from the estimate obtained when all cases were included (IRR = 0.55).
Third, improved awareness of early parkinsonian signs among clinicians could have led to a shift in the timing of a parkinsonism diagnosis, appearing as a spurious decline in parkinsonism incidence rates. Participants were all 55 years of age or older at the time of study entry and were at risk of parkinsonism and PD for a fixed maximum time period (10 years). Therefore, a left-shift in the timing of parkinsonism diagnosis would have resulted in young patients being considered to have incident cases in the 1990 subcohort and prevalent cases in the 2000 subcohort. However, the prevalences of parkinsonism and PD were much lower in the 2000 subcohort, and although the difference in parkinsonism and PD incidence between subcohorts was particularly pronounced in the highest age stratum, incidence rate ratios in other strata suggested lower parkinsonism and PD incidence rates among all age groups. Similarly, our observation that persons diagnosed with incident parkinsonism and PD in the 2000 subcohort were generally younger at the time of diagnosis was accounted for by the difference in age at baseline. We also investigated the mortality rates of parkinsonism patients, because a left-shift in the timing of parkinsonism diagnosis would artificially increase the survival time of patients within a fixed maximum time period (in the absence of major changes in parkinsonism treatment practices). We observed similar mortality rates for patients in both subcohorts, although the low number of patients who died limited our power to detect modest overall changes and prevented us from performing subgroup analyses (e.g., PD patients only). Taken together, we found little evidence to support a left-shift in the timing of parkinsonism diagnosis as an explanation for the remarkably lower in incidence rates.
Fourth, selection bias may have occurred. Although standardized and uniform ascertainment methods were used throughout the study period, and follow-up was virtually complete in both subcohorts, the baseline response proportion was higher in the 1990 subcohort than in the 2000 subcohort (78% vs. 67%). In 2000, most inhabitants of the study area who were aged 65 years or older were already enrolled in the 1990 subcohort of the Rotterdam Study (and a further minority had already refused to participate). Therefore, the majority of persons invited in 2000 were 55–64 years of age, and a substantial proportion of these persons were still working. We believe that this might have contributed to the difference in response rates between the subcohorts, and the dissimilar response rates, as well as the age distribution, may have contributed to the major, unadjusted difference in prevalence of parkinsonism (1.8% vs. 0.5%) and PD (1.4% vs. 0.3%) between subcohorts. Persons with greater disability (for example, individuals with parkinsonism) might be less likely to participate in an observational study such as the Rotterdam Study. In the absence of a left-shift in the timing of diagnosis, one would expect that the difference in response proportion would likely not directly affect age-specific incidence rates of parkinsonism and PD, unless the individuals who declined to participate, in both subcohorts, comprised a large proportion of persons at high risk of clinical parkinsonism and PD. The latter appears unlikely, because some common risk factors that are inversely associated with parkinsonism, particularly with PD, are associated with higher risk of much more common outcomes in the elderly (e.g., smoking and serum urate with cardiovascular disease). However, in view of the lack of data on population attributable risk of these factors for PD, even persons without any of these known risk factors could theoretically be at high risk for the disease, and their prodromal features (e.g., depressive symptoms) could have decreased the likelihood of their participation.
Fifth, in persons with dementia it remains difficult to clinically distinguish PD from parkinsonism that occurs secondary to the disease they already have (often Alzheimer's disease).[27] We considered persons with a dementia diagnosis to be no longer at risk for PD. As a consequence, incident dementia, as well as death, can be considered a competitive event for incident PD. Because underlying risk factors for dementia and mortality may also predispose to PD (e.g., genetic risk factors or oxidative agents), we examined age-specific changes in the incidence rates of competitive events. We observed lower rates among all age groups, suggesting that the composite competing risk for PD was higher in the 1990 subcohort. Had the competitive risk remained stable, an even larger decrease in PD incidence would have occurred. In addition, overall parkinsonism and PD incidence rates were lower among women in the 2000 subcohort than in the 1990 subcohort; incidence rates among men were also lower, but the difference was not significant. We did not observe statistically significant interaction with sex, and we could not investigate age-specific differences for men and women separately. Therefore, we consider our observation on sex-specific differences preliminary, and this remains an area for future study.
In conclusion, the present study suggests that parkinsonism rates have declined over 2 decades, largely as a consequence of a sharp drop in the incidence of PD. Further insight into the factors that caused this shift may open the door to protective strategies at the population level.
http://www.medscape.com/viewarticle/864662
No comments:
Post a Comment