Dementia Incidence in Individuals With Parkinson's Disease in the Framingham Heart Study

ABSTRACT

Limited information exists on incident dementia in individuals with Parkinson's disease (PD) in US community‐based samples. We examined cognitive statuses and PD diagnoses of 183 individuals in the Framingham Heart Study (FHS) to establish incident dementia, mortality rates, associations with sex, age at PD onset, and education level. Annual incidence rates of dementia (5.0%, 95% CI [3.7%–6.4%]) and mortality (10.5%, 95% CI [8.7%–12.3%]), did not differ by sex or education level. Increased age at PD diagnosis was a significant risk factor for incident dementia and mortality. Sex may not be a significant risk factor for community‐based PD dementia.

1. Introduction

Parkinson's Disease (PD), a neurodegenerative disorder characterized by progressive motor and cognitive decline [1, 2], affects more than 11 million people worldwide and is expected to reach approximately 25 million by 2050 [3, 4]. Individuals with PD face up to six times higher risk of developing dementia than age‐matched peers without PD [1, 5] A meta‐analysis of 25 longitudinal cohorts highlighted significant heterogeneity in dementia incidence in PD [5] In the United States, population‐based studies demonstrate substantial dementia burden in PD, but comparability is limited by differences in case identification and dementia adjudication [6, 7]. Existing PD‐related dementia and mortality literature is predominantly derived from specialist clinic cohorts that may over‐ or underrepresent disease severity, making estimates less comparable to community‐based samples [8, 9, 10, 11].

The Framingham Heart Study (FHS), launched in 1948 in Framingham, Massachusetts, USA, tracks adult participants longitudinally through clinical assessments, dementia surveillance, and adjudicated dementia and PD onset. Using its multigenerational, community‐based dataset, this study aimed to (1) examine dementia incidence of participants with PD, (2) analyze mortality after PD diagnosis to characterize prognosis (as a secondary outcome), and (3) identify the association between sex, age at PD onset, education, and incident dementia and mortality.

2. Methods

2.1. Study Design

The general study design of the FHS has been previously described extensively [12, 13, 14, 15]. In brief, it was initially designed to investigate the epidemiology of cardiovascular diseases but expanded to examine other health outcomes, including dementia and PD. The Original cohort (5209 mainly White participants) had a mean age at entry of 44 years. Their 5124 children (Offspring cohort) and Spouses enrolled in 1971. Gen 3, comprising 4095 children of the Offspring cohort, joined in 2002 with a mean age at entry of 41 years. Ethnically diverse participants from African, Hispanic‐Latino, Asian, Pacific Islander, and Native American origins were recruited in 1994 (506 participants, Omni Gen 1) and 2003 (410 participants, Omni Gen 2). FHS has a 99% retention rate [16]. All participants provided written informed consent before participating. This study received approval from the institutional review boards at Boston University and FHS. Recruitment details are available in Figure S1.

2.2. PD Evaluation and Adjudication

The FHS started tracking PD cases in 1988 with the Original cohort through self‐report questions at the in‐person Core exam, medical history updates, hospitalization record surveillance, and examinations conducted by an FHS neurologist [17]. PD surveillance was later introduced in the Offspring and Spouses cohort in 1991 and in 2002 for Omni Gen 1, Omni Gen 2, and Gen 3. Participants are flagged for PD consensus adjudication review if they self‐reported PD diagnosis or were flagged as having motor symptoms following clinical examination by an FHS clinician. Since 1988, PD diagnoses and the year of PD onset have been adjudicated by movement disorders specialist clinicians, according to the UK Parkinson's Disease Society Brain Bank [18]. In a prior validation study from the FHS, this approach yielded a positive predictive value (PPV) of 0.85 (95% CI, 0.78–0.90) against expert‐consensus diagnosis [19].

2.3. Dementia Evaluation and Adjudication

Surveillance techniques for incident dementia in the FHS have been previously published [15, 19, 20]. Cognitive status surveillance started with a subjective memory question in 1979, serial Mini–Mental State Examinations from 1991, and expanded to a comprehensive battery of neuropsychological tests every 4–5 years since 1999¹² (Table S1). Incident dementia diagnoses are determined through consensus adjudication by a dementia review panel (neurologist and neuropsychologist), based on Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM‐IV) criteria [21]. Functional status, including impairment in activities of daily living, was assessed using the Quick Dementia Rating System [22] and telephone interviews. This information was combined with serial neuropsychological tests, medical and hospitalization records, neuroimaging, and clinical examination findings to classify dementia subtype, assess severity, and estimate onset date. Alzheimer's Disease (AD) and vascular dementia were classified based on criteria from the NINCDS–ADRDA [23] and the NINDS–AIREN [24] criteria. Distinguishing Parkinson's Disease Dementia from Dementia with Lewy Body (DLB) requires accurate symptom timing based on the 1‐year rule of parkinsonism and dementia onset [25], which was not consistently defined in all cases. Therefore, Lewy body dementia (LBD), the umbrella term encompassing both diagnoses, was used instead. Dementia cases identified before 2001 were rereviewed to ensure compliance with updated standards.

2.4. Statistical Analysis

Data were extracted on participants adjudicated as having PD from January 1, 1988 until December 31, 2022. Individuals with unknown dementia diagnosis dates were excluded. Time‐to‐event analyses began at the adjudicated year of PD motor symptom onset and continued until the earliest of dementia adjudication, death, or loss to follow‐up. Participants were right‐censored at whichever event occurred first, ensuring that no individuals were assumed immortal beyond their observed follow‐up and avoiding time‐bias issues.

An exact Poisson test‐derived 95% confidence interval was used to calculate dementia incidence rates. Kaplan–Meier (K‐M) curves estimated survival probabilities and dementia incidence post‐PD diagnosis over a 25‐year follow‐up period, by comparing males and females distributions using the log‐rank (Mantel‐Cox) test. Cox proportional hazards models were adjusted for sex and age at PD onset to account for confounding from baseline age‐related mortality risk. Proportional hazards assumption was verified via log–log survival plots. All analyses were performed using R software (version 4.4.2) [26]. Survival analyses, including Cox regression models for hazard ratios and predictors of mortality and dementia incidence, utilized the R “survival” package (version 3.8.3) [27]. K‐M survival curves were generated by R “survminer” package (version 0.5.0) [28].

3. Results

3.1. Demographics and Clinical Characteristics

Among 15,447 individuals, 183 participants (110 men and 73 women) were diagnosed with PD, yielding an overall PD prevalence of 1.2%. The majority (60.1%) were from the Original cohort, and no cases were identified in the Omni 2 cohort. The mean age at PD onset was 74.83 years (SD = 9.84). Most participants (27.3%) had high school education. The most common dementia was LBD (34.4%). Further characteristics are summarized in Table 1.

TABLE 1.

Demographics data of participants with Parkinson's Disease in the Framingham Heart Study.

	All participants with PD since 1988 until 2022 (n = 183)
Cohort, n (%)
Original	110 (60.1%)
Offspring	65 (35.5%)
Spouse	2 (1.1%)
Gen 3	5 (2.7%)
Omni 1	1 (0.5%)
Age at PD diagnosis, mean (SD)	74.83 (9.84)
Sex, n (%)
Male	110 (60.1%)
Female	73 (39.9%)
Highest level of education attained, n (%) a
High school—did not graduate	47 (25.7%)
High school graduate	50 (27.3%)
Some college/university	42 (23.0%)
College/university graduate	42 (23.0%)
Dementia category, n (%)
No dementia	98 (53.6%)
LBD	63 (34.4%)
AD (without stroke)	10 (5.5%)
AD (with stroke present)	1 (0.5%)
Vascular dementia (without AD)	3 (1.6%)
Dementia (uncertain phenotype) b	8 (4.4%)

Abbreviations: AD, Alzheimer's Disease; APOE, apolipoprotein E; LBD, lewy body dementia; PD, Parkinson's Disease; SD, standard deviation.

^a Two participants are unknown in the highest level of education attained.

^b Uncertain phenotype indicates dementia where the available clinical information was insufficient to assign a confident subtype (e.g., overlapping features or limited documentation).

3.2. Incidence of Dementia and Mortality

Over 1131 person‐years following PD diagnosis, 56 dementia cases were identified, yielding an overall annual dementia incidence rate of 5.0% (95% CI [3.7, 6.4]). When stratified by sex, males accounted for 33 cases over 651 person‐years (5.1% per year; 95% CI [3.5, 7.1]), and females had 23 cases over 480 person‐years (4.8% per year; 95% CI [3.0–7.2]). Dementia incidence did not differ significantly by sex (males compared to females showed Hazard ratio [HR] = 1.06, 95% CI 0.61–1.86, p = 0.837) or by level of education (log rank p = 0.21) (Figure 1A,C). Similarly, overall survival probability was comparable for sex (HR = 0.74, 95% CI, 0.53–1.04, p = 0.788) and education (log rank p = 0.54) (Figure 1B,D). In our multivariable Cox models (Table 2), each 1‐year increase in age at PD diagnosis was associated with a 3% higher hazard of dementia (HR 1.03; 95% CI, 1.00–1.06; p = 0.033) and a 7% higher hazard of mortality (HR 1.07; 95% CI, 1.05–1.09; p = 9.77e‐13). The number at risk declined progressively over time, with few surviving beyond 20 years post‐diagnosis. Incidence rates for mortality showed 153 deaths over 1455 person‐years for a 10.5% annual incidence rate (95% CI [8.9, 12.3]). Males had higher mortality rates than females at 92 deaths over 830 person‐years (11.1% annual incidence, 95% CI [8.9–13.6]). The median (mean) follow‐up duration from baseline to death or last contact was 6.71 (8.15) years, while the median (mean) survival time and median (mean) time to dementia were 6.21 (7.37) and 5.34 (7.07) years, respectively.

FIGURE 1.

Kaplan–Meier curves of participants with PD showing: (A) dementia incidence by sex; (B) overall survival probability by sex; (C) dementia incidence by education; (D) overall survival probability by education.

TABLE 2.

Cox models for dementia and mortality in individuals with Parkinson's Disease.

	Dementia		Mortality
	Hazard ratio (95% CI)	p	Hazard ratio (95% CI)	p
Sex	1.06 (0.605, 1.86)	0.837	0.738 (0.526, 1.04)	0.788
Age	1.03 (1.00, 1.06)	0.0337	1.07 (1.05, 1.09)	9.77E‐13
Education	1.23 (0.944, 1.60)	0.125	0.913 (0.786, 1.06)	0.238

4. Discussion

The annual dementia incidence in PD within the FHS was 5.0%, aligning with a recent meta‐analysis across 25 longitudinal cohort studies, which reported a pooled annual incidence risk rate of 4.45% [5]. Several characteristics in our study are noteworthy, including the presence of PD diagnoses adjudicated before 2007, older age at PD onset, and an extended follow‐up period. Incident dementia rates in PD did not vary significantly by education level, consistent with another study [29], or by sex, consistent with two earlier studies [12, 30]. However, another study of younger participants with a mean age of mid‐50s at PD onset associated male sex with PD‐related cognitive decline [31]. Our sample, with an older age of PD onset, may reflect additional co‐pathologies such as AD, where female sex is a greater risk factor for incident dementia [32]. Our annual mortality rate of 10.5% is somewhat comparable to prior evidence of increased all‐cause mortality in PD [33, 34, 35], with differences in absolute rates likely reflecting our cohort's older age and community‐based sample.

Given the smaller sample size than comparable longitudinal studies, educational attainment and sex were not statistically significantly associated with dementia incidence in this PD cohort, likely reflecting limited power. Dementia subtype was classified following adjudication by a review panel without neuropathological confirmation, a common limitation in many large prospective cohort studies without autopsy data [30]. Cognitive surveillance at FHS intensified over time. Consequently, earlier follow‐up relied more on the MMSE, which is less sensitive to PD‐related cognitive deficits, potentially underestimating dementia prior to broader neuropsychological testing. As PD motor symptom onset is inherently difficult to date precisely in observational cohorts, onset timing was adjudicated to the year rather than an exact date. Survival heterogeneity attributable to motor phenotype (e.g., tremor‐dominant vs. postural instability gait difficulty‐dominant) or alternative causes of parkinsonism‐related dementia (e.g., corticobasal syndrome) was not possible, as these were not systematically captured in the structured examination forms used in this study. Nonetheless, our findings align with others in showing that later age at PD onset predominantly influences dementia conversion and mortality [36, 37, 38].

A key strength of the FHS is its longitudinal prospective design, which has collected comprehensive evaluations throughout participants' lifespans. Although this study included limited racial and ethnic diversity, the data provide important information regarding dementia incidence and mortality rates and clinical characteristics of dementia in individuals with PD in a North American community‐based cohort. Despite the limitations, the combination of community‐based sampling and adjudicated PD and dementia outcomes provides an important complement to prior clinic‐derived literature and supports a more comparable estimation of dementia incidence following incident PD in a US setting. Future studies to assess incident dementia rates may require larger samples to effectively examine the impact of other exposures that may influence dementia and mortality in PD.

Author Contributions

Joshi Dookhy: research project – conception, organization, execution; statistical analysis – review and critique; manuscript – writing of the first draft, review and critique. Michael Cummings: research project – execution; statistical analysis – design, execution, review and critique; manuscript – review and critique. Eden Price: research project – execution; manuscript – review and critique. Yulin Liu: statistical analysis – execution. Santiago Gutierrez‐Gomez: manuscript – review and critique. Sarah A. O‘Shea: manuscript – review and critique. Samuel A. Frank: manuscript – review and critique. Rhoda Au: manuscript – review and critique. Ashita S. Gurnani: statistical analysis – review and critique; manuscript – review and critique. Ting Fang Alvin Ang: research project – execution; statistical analysis – design, execution, review and critique; manuscript – review and critique. Jesse Mez: statistical analysis – review and critique; manuscript – review and critique. Ludy C. Shih: research project – conception, organization, execution; statistical analysis – review and critique; manuscript – review and critique.

Funding

This work was supported by the National Heart, Lung, and Blood Institute, 75N92019D00031, HHSN269201500001I, and N01‐HC‐25195; The National Institute on Aging, R01‐AG016496, R01‐AG062109, and U19AG068753.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1. STROBE diagram showing process of participant recruitment.

Table S1. Comprehensive Battery of neuropsychological tests.

Funding Statement

This work was funded by National Heart, Lung, and Blood Institute grants 75N92019D00031, HHSN269201500001I, and N01‐HC‐25195; National Institute on Aging grants R01‐AG016496, R01‐AG062109, and U19AG068753.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.