eTable 1. Diagnostic Codes for Neurological Exclusions
eTable 2. Diagnostic Codes for Exposures
eTable 3. Prevalence and Age- and Sex-Adjusted Odds Ratios of PD for Comorbidities and Risk Factors According to Time of Presentation
eTable 4. Prevalence and Age- and Sex-Adjusted Odds Ratios of PD for Prodromes and Motor Markers According to Time of Presentation
eTable 5. Prevalence of Motor Symptoms (Tremor, Rigidity, or Balance Difficulties) and Cognitive Symptoms Amongst PD Patients According to Ethnicity Together With Their Association With Future PD
eTable 6. Demographic Information of PD Cases and Unmatched Controls in UK Biobank
eTable 7. Distribution of Prediagnostic Manifestation Amongst PD and Control Groups in the ELGP and THIN Database
eFigure 1. Forest Plot Depicting Comorbidities and Risk Factors With Parkinson’s Disease Across Ethnic Groups in the East London Population
eFigure 2. Forest Plot Depicting Nonmotor Prediagnostic Manifestations of Parkinson’s Disease Across Ethnic Groups in the East London Population
eFigure 3. Forest Plot Depicting Nonmotor and Motor Prediagnostic Manifestations of Parkinson’s Disease Across Ethnic Groups in the East London Population
Customize your JAMA Network experience by selecting one or more topics from the list below.
Simonet C, Bestwick J, Jitlal M, et al. Assessment of Risk Factors and Early Presentations of Parkinson Disease in Primary Care in a Diverse UK Population. JAMA Neurol. 2022;79(4):359–369. doi:10.1001/jamaneurol.2022.0003
What are the prediagnostic manifestations of Parkinson disease (PD) presented in the primary care setting in a diverse and deprived population with universal access to health care?
This case-control study including 1055 individuals who developed PD over time noted signs and symptoms that occurred before diagnosis, with tremor and memory symptoms being the most prominent. Several recognized prediagnostic features were replicated, including temporal associations between epilepsy and hearing loss with subsequent PD.
The findings of this study suggest there is a broad range of symptoms that people present in the primary care setting up to a decade before PD diagnosis.
Early features of Parkinson disease (PD) have been described through population-based studies that overrepresent White, affluent groups and may not be generalizable.
To investigate the association between risk factors and prediagnostic presentations of PD in an ethnically diverse UK population with high socioeconomic deprivation but universal access to health care.
Design, Setting, and Participants
A nested case-control study was conducted using electronic health care records on 1 016 277 individuals from primary care practices in East London to extract clinical information recorded between 1990 and February 6, 2018. The data were analyzed between September 3, 2020, and September 3, 2021. Individuals with a diagnosis of PD were compared with controls without PD or other major neurological conditions.
Main Outcomes and Measures
A matched analysis (10 controls matched for each patient with PD according to age and sex) and an unmatched analysis (adjusted for age and sex) were undertaken using multivariable logistic regression to determine associations between risk factors and prediagnostic presentations to primary care with subsequent diagnosis of PD. Three time periods (<2, 2-<5, and 5-10 years before diagnosis) were analyzed separately and together.
Of 1 016 277 individuals included in the data set, 5699 were excluded and 1055 patients with PD and 1 009 523 controls were included in the analysis. Patients with PD were older than controls (mean [SD], 72.9 [11.3] vs 40.3 [15.2] years), and more were male (632 [59.9%] vs 516 862 [51.2%]). In the matched analysis (1055 individuals with PD and 10 550 controls), associations were found for tremor (odds ratio [OR], 145.96; 95% CI, 90.55-235.28) and memory symptoms (OR, 8.60; 95% CI, 5.91-12.49) less than 2 years before the PD diagnosis. The associations were also found up to 10 years before PD diagnosis for tremor and 5 years for memory symptoms. Among midlife risk factors, hypertension (OR, 1.36; 95% CI, 1.19-1.55) and type 2 diabetes (OR, 1.39; 95% CI, 1.19-1.62) were associated with subsequent diagnosis of PD. Associations with early nonmotor features, including hypotension (OR, 6.84; 95% CI, 3.38-13.85), constipation (OR, 3.29; 95% CI, 2.32-4.66), and depression (OR, 4.69; 95% CI, 2.88-7.63), were also noted. Associations were found for epilepsy (OR, 2.5; 95% CI, 1.63-3.83) and hearing loss (OR, 1.66; 95% CI, 1.06-2.58), which have not previously been well reported. These findings were replicated using data from the UK Biobank. No association with future PD diagnosis was found for ethnicity or deprivation index level.
Conclusions and Relevance
This study provides data suggesting that a range of comorbidities and symptoms are encountered in primary care settings before PD diagnosis in an ethnically diverse and deprived population. Novel temporal associations were observed for epilepsy and hearing loss with subsequent development of PD. The prominence of memory symptoms suggests an excess of cognitive dysfunction in early PD in this population or difficulty in correctly ascertaining symptoms in traditionally underrepresented groups.
Diversity is lacking in the study of complex diseases. Most research into the causes of Parkinson disease (PD) has been conducted in patients of Northern European ancestry.1,2 Little is known about how PD manifests in different ethnic groups and whether there is differential case ascertainment, response to medication, or different determinants of risk.3,4 In addition to patients from minority ethnic groups who are underrepresented in PD research,5 little is known about PD occurring in people living in areas of high socioeconomic deprivation, with increased unemployment and social isolation rates, or those residing in impoverished urban settings.6 The UK National Health Service (NHS) provides primary and secondary care to the entire UK population free at the point of access. Publicly funded health care systems, such as the NHS, are an ideal setting for the study of diseases that may vary according to the above factors, to reduce bias and improve the generalizability of results.
Early features and risk factors associated with PD diagnosis have been identified through large, population-based observational studies.7-11 In the UK, a comprehensive observational study of early features of PD was undertaken by some members of this study group in The Health Improvement Network (THIN) primary care database that involved 8166 people with PD and 46 455 healthy individuals, most of whom were White and from higher-income groups.9 In the present study, we used a similar approach but in a highly diverse population from East London with some of the highest levels of socioeconomic deprivation in the UK.
We performed a nested case-control study in a large primary care data set in East London. Primary care data were compiled from searches of Egton Medical Information Systems electronic health care records for the Secure Health Analysis and Research in East London project. The database included health records of 1 016 277 patients from general practices across 4 Clinical Commissioning Groups in East London: Hackney & City of London, Newham, Tower Hamlets, and Waltham Forest. Use of Egton Medical Information Systems began in the UK in 1990 and paper records acquired prior to this time were manually transcribed into the system. In the UK NHS, individuals are identified by a unique NHS number that enables linkage to their medical records. When individuals move between health care professionals, their records move with them. All nonemergency secondary care referrals originate from primary care, and outcomes are communicated back; thus, primary care records represent aggregated medical information about an individual throughout their lifetime. The NHS Health Research Authority waived the need for ethical approval when using anonymized data sets such as these. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
All individuals with a code diagnosis of PD were included as cases in the analysis. Patients with PD but missing a date of diagnosis were excluded, as well as those with a coded diagnosis of dementia, atypical PD, and other neurodegenerative conditions, including multiple sclerosis and motor neurone disease (eTable 1 in the Supplement). Individuals serving as controls were those without a code of PD or other chronic neurological conditions including dementia, multiple sclerosis, atypical PD, and motor neurone disease. Controls were assigned a dummy date of PD diagnosis, which was calculated as follows: the median age at PD diagnosis (69.0 years) was added to the year of birth of each control to create a dummy date of diagnosis in this cohort. The earliest date between the dummy date of PD diagnosis and February 6, 2018, which was when the database was locked, was then used as the time to categorize the selected exposures as prediagnostic risk factors.
Exposures were selected based on a comprehensive meta-analysis of prediagnostic features and risk factors for PD carried out by some members of this group in 201212 and 3 other large studies of the prediagnostic phase of PD.9,11,13 Epilepsy and hearing loss were included given preliminary evidence that these might be prediagnostic features of PD.10,11,14
Overall, 24 exposure variables were selected and subdivided into 3 categories: (1) comorbidities and risk factors, (2) prediagnostic nonmotor manifestations (metabolic, sensory, autonomic, and neuropsychiatric), and (3) prediagnostic motor manifestations. Individual patient information was extracted by the Clinical Effectiveness Group at Queen Mary University of London on February 6, 2018. All exposures were recorded up to twice in our database (earliest ever record and most recent record). If there was more than 1 observation, the earliest date was used for the analysis.
Given the cross-sectional nature of data extraction, incidence rates could not be calculated. However, we wished to examine temporal associations, and so 3 intervals were established to evaluate exposure outcome associations (<2, 2 to <5, and 5-10 years before PD diagnosis or dummy diagnosis). We selected the same periods used in the THIN primary care analysis to make findings comparable and see whether there were differences between the 2 populations with divergent socioeconomic background and ethnicity.9 Exposure variables with less than 1% prevalence among individuals with PD across all time periods were excluded from the analysis.
To replicate novel associations from the main analysis, we undertook additional analysis using data from the UK Biobank. The International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) code G20 was used to define PD. As in the main analysis, patients with PD and controls were excluded if they had atypical PD, dementia, multiple sclerosis, or motor neurone disease (eTable 1 in the Supplement). We used the same time intervals as for the main analysis (<2, 2 to <5, and 5-10 years before PD diagnosis).
The variables extracted were based on identified data comprising diagnoses, laboratory test results, and demographic details coded using the Read coding system15 (eTable 2 in the Supplement). Variables were defined so that as much information as possible could be used in the modeling. For this reason, missing data were categorized as unknown in the models rather than excluded.
The age of the participants was documented as the age at data extraction (February 6, 2018). Ethnicity was defined by the self-reported UK census categories, grouped here into major ethnic groups in the East London population: Black (African, Caribbean, and other Black), South Asian (Bangladeshi, Indian, and Pakistani), White (British, Irish, and other White), other (Chinese, other, and mixed groups), and unknown.
The Index of Multiple Deprivation (IMD) is a global measure based on socioeconomic terms, including income, employment, educational level, health, crime, housing, and environment. Raw IMD scores were assigned to deciles derived from national data and converted into quintiles. Quintile 1 (IMD 1-2) represented the 20% of the population with the most socioeconomic deprivation and quintile 5 (IMD 9-10) represented the 20% of the population with the least socioeconomic deprivation. The IMD group 1-2 was used as the reference category in the analyses.
Coded diagnoses of hypertension, cholesterol level, and type 2 diabetes (T2D) were defined on 4 levels depending on whether the risk factor was never recorded, first recorded before the PD diagnosis (or dummy date of diagnosis for controls), first recorded after PD diagnosis, or unknown if the data were missing. For each risk factor, the status was determined by the earliest record unless this information was missing, in which case the status at the most recent date was used. Total cholesterol levels were taken from the clinical data and were considered valid if they ranged from 19 to 193 mg/dL (to convert to millimoles per liter, multiply by 0.0259). Values greater than 193 mg/dL were indicative of hypercholesterolemia. Hypertension and T2D were recorded according to the presence of a coded morbidity record. Patients with a diagnosis of type 1 diabetes were recorded as normal. Although being overweight is considered a vascular risk factor, it was considered separately.
Smoking status was coded as being a current smoker, ex-smoker, or never smoker, but the time before PD diagnosis was not considered given that smoking initiation in later life is rare. For that reason, data related to smoking cover all prediagnostic periods.
Body mass index (BMI) was calculated as weight in kilograms divided by height in meters squared and categorized as follows: normal (20.0-24.9), underweight (10.0-19.9), and overweight (25.0-50.0). The range included in analysis for height was 100 to 250 cm, and the range for weight was 30 to 250 kg. If height or the calculated BMI was outside these ranges, the data were deemed unlikely and were reclassified as unknown. In the same way, participants with missing BMI data were classified as unknown unless they had a diagnosis of obesity recorded, in which case they were classified as overweight. Unlike overweight, underweight is not linked to coexistent vascular disease. We therefore listed underweight as a metabolic prodrome category.
Hearing loss and epilepsy were defined by a coded diagnosis. In the case of hearing loss, a referral for assessment due to reported hearing difficulty was also included.
Clinical symptoms of PD included coded nonmotor features, such as memory problems, depression, anxiety, fatigue, erectile dysfunction, shoulder pain, neck pain, and constipation. Motor features included rigidity, tremor, and balance difficulties.
Data analysis was performed between September 3, 2020, and September 3, 2021. For the periods from less than 2 years, 2 to less than 5 years, and 5 to 10 years before the index date (date of diagnosis), the overall occurrence of prediagnostic symptoms was calculated as the absolute number and percentage. A matched case-control analysis was run by matching 10 controls for each patient with PD according to age (calendar year) and sex. This categorization was used to estimate the odds ratio (OR) for PD and 95% CI for each variable of interest in each period and in all 3 periods combined. The matched analysis was then rerun adjusting for ethnicity and IMD level. A separate multivariable logistic regression model, which included patients with PD and all controls, was performed, and estimates for each exposure of interest were adjusted for age and sex. To examine an association with PD as a trend across IMD quintiles, conditional logistic regression was undertaken, treating IMD quintiles as continuous. To examine a difference in the odds of PD by ethnicity, a comparison of the logistic regression models with and without ethnicity was undertaken using a likelihood ratio test. In a subanalysis, we examined associations between prediagnostic symptoms and subsequent diagnosis of PD, stratified by ethnic group. For this analysis, we used the unmatched analysis and the full prediagnostic period, excluding the 2 years closest to diagnosis (ie, 2-10 years) and adjusting for age and sex. A replication analysis for any novel findings was performed using data from the UK Biobank cohort (application reference 78867).16 This matched case-control analysis was done (10 controls matched to age and sex were randomly selected for each PD diagnosis bin). The analyses were performed using R, version 4.0.2 (R Foundation for Statistical Computing); Stata, version 15 (StataCorp LLC); and MATLAB (MathWorks Inc).
Of 1 016 277 individuals included in the data set for this case control study, 5699 were excluded and 1055 patients with PD and 1 009 523 controls were included in the analysis. Demographic information for patients and controls is summarized in Table 1. Patients with PD were more likely to be older vs controls (mean [SD] age, 72.9 [11.3] vs 40.3 [15.2] years). The PD group comprised 423 women (40.1%) and 632 men (59.9%), and the control group included 492 661 women (48.8%) and 516 862 men (51.2%). No association with PD was found when examining the data across IMD levels (eg, level 1-2 [most deprived]: 472 [44.7%] vs 453 747 [44.9%]; overall P = .71 across IMD levels), but most participants (>80%) resided in the 2 IMD quintiles with the most socioeconomic deprivation. The ethnicity of participants reflected the diversity of the local population in East London and was similar among patients with PD and controls (eg, patients: 15.7% Black, 19.7% South Asian, 50.9% White, and 8.3% other; controls: 13.3% Black, 21.5% South Asian, 43.7% White, and 11.3% other; P = .18 from the likelihood ratio test).
Table 2 summarizes associations of comorbidities or risk factors with PD over the 3 time periods. The main matched analysis (matching 10 controls for each patient according to age and sex) showed that the ORs were highest for epilepsy across each of the 3 periods analyzed (<2 years: OR, 10.00; 95% CI, 1.41-70.99; 2 to <5 years: OR, 5.00; 95% CI, 1.25-19.99; and 5-10 years: OR, 5.46; 95% CI, 2.02-14.76). When we analyzed all 3 periods together, the OR was lower (OR, 2.5; 95% CI, 1.63-3.83).
Having hypertension or T2D 5 to 10 years before diagnosis, but not closer to diagnosis, was associated with higher odds of subsequent PD (hypertension: OR, 1.29; 95% CI, 1.07-1.56; T2D: OR, 1.41; 95% CI, 1.09-1.82). In contrast with being overweight, for which no association was seen, being underweight in the period closest to PD diagnosis (<2 years) was associated with higher odds of PD (OR, 2.73; 95% CI, 1.17-6.37). High cholesterol levels were not associated with subsequent PD, even when analyzing all 3 periods together (OR, 0.96; 95% CI, 0.83-1.10). Smoking showed an inverse association (OR, 0.76; 95% CI, 0.66-0.87) with subsequent PD, as did alcohol consumption (OR, 0.79; 95% CI, 0.68-0.92).
Although ethnicity and IMD were not associated with PD, we adjusted for both covariates in the matched analysis. There was no suggestion of confounding for associations of midlife comorbidities (Table 2).
Table 3 summarizes clinical manifestations in the 3 periods from the matched case-control analysis. Cognitive symptoms were the most frequently reported nonmotor manifestations. Fifty-two (4.9%) patients who were subsequently diagnosed with PD presented with memory symptoms compared with less than 1% of controls. Within 2 years of PD diagnosis, people with memory symptoms had an approximate 9-fold increased odds of PD (OR, 8.60; 95% CI, 5.91-12.49). This association was also found up to 5 years before diagnosis but not further (2 to <5 years: OR, 3.08; 95% CI, 1.81-5.24; 5-10 years: OR, 2.06; 95% CI, 0.96-4.42). Patients with hypotension had a 7-fold increased odds of receiving a PD diagnosis within 2 years (OR, 6.84; 95% CI, 3.38-13.85). This association was also found at 2 to less than 5 years before diagnosis (OR, 4.88; 95% CI, 2.44-9.77).
Constipation was found to be more frequent among people with future PD compared with controls (4% vs 1%). In the matched analysis, patients presenting with constipation were greater than 3 times more likely to be diagnosed with PD compared with the controls. This association was found across all 3 analysis periods (<2 years: OR, 3.29; 95% CI, 2.32-4.66; 2 to <5 years: OR, 2.68; 95% CI, 1.97-3.66; 5-10 years: OR, 2.96; 95% CI, 2.16-4.06). In contrast with other autonomic symptoms, an association between erectile dysfunction and future PD in men was observed only 5 to 10 years before diagnosis (OR, 1.51; 95% CI, 1.11-2.05).
The OR for depression and subsequent PD was highest up to 2 years before diagnosis of PD (OR, 4.69; 95% CI, 2.88-7.63), and the OR decreased over time (5-10 years: OR, 1.97; 95% CI, 1.31-2.97). Anxiety, fatigue, and insomnia were associated with PD, but the ORs were lower than depression and mainly presented closer to PD diagnosis. Overall, patients in East London presenting to primary care physicians with psychiatric symptoms, including depression, anxiety, fatigue, and insomnia, within 2 years before diagnosis had about 2- to 4-fold increased odds of receiving a PD diagnosis (depression: OR, 4.69; 95% CI, 2.88-7.63; anxiety: OR, 3.08; 95% CI, 2.06-4.60; fatigue: OR, 1.91; 95% CI, 1.25-2.93; and insomnia: OR, 2.18 ; 95% CI, 1.36-3.51) compared with individuals without PD.
Associations between pain and subsequent PD varied according to the location of pain. Shoulder pain was associated with a doubling of the odds of PD diagnosis up to 5 years before diagnosis (<2 years: OR, 2.23; 95% CI, 1.50-3.30; 2 to <5 years: OR, 1.88; 95% CI, 1.33-2.66), whereas neck pain was not associated with PD across the 3 periods.
Hearing loss was more common in individuals who developed PD (<2 years: 2.2% vs 1.3%; 2 to <5 years: 2.7% vs 1.6%, 5-10 years: 2.3% vs 1.5%). Although ORs were small, an association with PD was observed up to 5 years before PD diagnosis but not after that (<2 years: OR, 1.66; 95% CI, 1.06-2.58; 2 to <5 years: OR, 1.73; 95% CI, 1.16-2.57; and 5-10 years: OR, 1.48; 95% CI, 0.96-2.29). Anosmia was present in less than 1% of patients with PD across all time periods and was not included in the analysis.
Tremor was associated with subsequent PD across the 3 time periods. Most presentations of tremor were within 2 years before PD diagnosis (25% of patients and <1% of controls: OR, 145.96; 95% CI, 90.55-235.28). The prevalence of tremor was also higher in the PD group compared with the control group in the 2 later periods (2 to <5 years: OR, 14.48; 95% CI, 9.02-23.25; 5-10 years: OR, 11.66; 95% CI, 6.59-20.64). Individuals who received a diagnosis of PD had a higher prevalence of balance difficulties across the 3 periods (4%) compared with controls (2%). Although rigidity was more frequently reported in the PD group than controls, it was not as common as other motor features, with a prevalence of 1.2% during the first period (<2 years) and less than 1% within the other 2 periods. There was no suggestion of ethnicity and IMD confounding the association of prediagnostic features with the risk of PD, for example, tremor (unadjusted OR, 145.96; 95% CI, 90.55-235.28; adjusted OR, 151.24; 95% CI, 93.74-244.02) (Table 3).
eTable 3 and eTable 4 in the Supplement summarize the same midlife comorbidities and prediagnostic manifestations using an unmatched analysis adjusted for age and sex. In general, the results from the unmatched analysis were similar to those from the matched analysis. However, in contrast with the matched analysis, when increasing the sample size by including all controls, the association between memory symptoms and risk of PD was also found up to 10 years before diagnosis (OR, 2.78; 95% CI, 1.38-5.61). The ORs for hearing loss and future PD diagnosis were higher and consistent across all periods (<2 years: OR, 1.84; 95% CI, 1.21-2.79; 2 to <5 years: OR, 2.24; 95% CI, 1.54-3.26; and 5-10 years: OR, 1.75; 95% CI, 1.17-2.64). In addition, in the unmatched analysis, neck pain was associated with subsequent PD, although the ORs were lower than the association with shoulder pain (ORs ranging between 1.50 and 1.71) (eTable 4 in the Supplement).
In a subanalysis stratifying exposure variables by ethnic group, fairly consistent associations were observed for midlife comorbidities (eFigure 1 in the Supplement), nonmotor manifestations, and early motor markers (eTable 5, eFigure 2 and eFigure 3 in the Supplement). Some differences were observed, with the 95% CI crossing the null across all ethnic groups for motor and memory symptoms. We found that epilepsy was associated with subsequent PD in patients who were Black or White but not in those of South Asian ethnicity. In contrast to tremor and balance impairment, rigidity was not reported in Black patients before diagnosis (sparse data) and had lower ORs in South Asian patients (OR, 4.62; 95% CI, 0.63-33.88).
Demographic details of the UK Biobank data set used are provided in eTable 6 in the Supplement. For this replication analysis, we used a matched design (1:10) that included 110 cases and 1100 controls in the first period (<2 years), 274 cases and 2740 controls in the second period (2 to <5 years), and 1074 cases and 10 740 controls in the third period (5-10 years). Epilepsy and hearing loss again were associated with subsequent PD diagnosis over the 3 time periods (ORs for epilepsy approximately 3; ORs for hearing loss approximately 1.2), except for hearing loss in the period closer to PD diagnosis (OR, 1.03; 95% CI, 0.69-1.52). A similar range of ORs was observed to that from the main analyses (Table 4).
We used a large primary care data set to explore risk factors and early clinical manifestations of PD in a highly diverse population with general socioeconomic deprivation. Based on the 2011 UK Census, London had the greatest ethnic diversity of anywhere in the UK, with the highest proportion of Black, South Asian, and mixed/other ethnic groups, which comprise approximately 45% of residents in East London compared with 14% in the rest of the UK. East London has one of the highest unemployment rates in the UK (6.7%), and 80% of patients included in this analysis resided in the lowest 2 quintiles of national wealth. There was no association between ethnic group or Index of Multiple Deprivation and odds of PD in our data, suggesting that ethnicity and deprivation may not play a major role in PD risk, in contrast to what has recently been reported for dementia.17
We observed a constellation of symptoms noted by general practitioners up to a decade before diagnosis of PD. Similar findings have been reported in more homogeneous populations.8,9,18 In the main (matched) analysis, associations noted before PD diagnosis were motor (up to 10 years) and cognitive (up to 5 years) symptoms. When we repeated the analysis using an unmatched but age- and sex-adjusted approach (with a larger sample size), there was an association with both factors up to 10 years before diagnosis.
To our knowledge, this is the first study focusing on the prediagnostic phase of PD in such a diverse population with universal access to health care. Plouvier and colleagues7 carried out a nested case-control study using primary care data in a small sample size and a period up to 2 years before diagnosis. They reported that patients with PD presented more often with functional symptoms, autonomic symptoms, and sleep problems than controls. A study conducted in the Netherlands used primary care data to compare a group of 60 patients with PD and 58 controls.8 The investigators identified a prediagnostic period of 4 to 6 years, comprising a wide range of nonmotor manifestations. Neither of these studies reported on motor manifestations before diagnosis. A UK study using THIN primary care data shared a similar approach to our study.9 Both compared the medical records of a large sample of people with PD with healthy controls using the same time periods (<2, 2-5, and 5-10 years) and had similar age and sex distribution. The THIN data differ substantially from East London data in terms of wealth and ethnicity. Autonomic symptoms and mood disturbances were on average between 2 and 4 times more commonly reported in the THIN database than in the East London sample (eTable 7 in the Supplement). Tremor was the primary marker of subsequent PD in both studies. Prescribing data, which were not available in our study, were used in the THIN analysis to help define exposures. Inclusion of this information may explain the higher prevalence of certain symptoms in the THIN analysis. However, there may be other reasons for underascertainment of symptoms in the East London sample. Minoritized ethnic and lower-income groups may be more resilient or less likely to consult primary care physicians about symptoms such as constipation, fatigue, insomnia, and erectile dysfunction. A recent cross-sectional study of patients with PD from different ethnic groups found that those who were Asian scored higher in sleep/fatigue and mood/apathy questions on direct questioning.19 Together, these findings suggest nonmotor symptoms might be experienced but underreported among certain ethnic groups.
Epilepsy occurring before PD was a notable finding, which we went on to replicate using a similar matched design in the UK Biobank. Some evidence for this association has been identified in previous studies, which prompted its inclusion in the present study.10,11 The prevalence of epilepsy in patients with PD has been reported as being higher than the estimated prevalence in the general population.20-22 In the present study, drug-induced parkinsonism could not be ruled out owing to the lack of information about medication. Certain antiepileptic drugs have been associated with tremor and PD,23 and coexisting vascular disease might also plausibly link epilepsy with PD, especially in the older population. Further research is needed to investigate a potential link.
In East London, having hypertension, T2D, and being underweight were associated with increased odds of developing PD. Of the vascular risk factors associated with an increased risk of PD, T2D is the one with the most evidence.24 Antidiabetic treatments, such as glucagonlike peptide 1 analogues, are being investigated as potential treatments for PD.25,26 Hypertension has been associated with a small increase in the risk of PD according to a recent large meta-analysis,27 but a protective association was observed in a previous meta-analysis involving only case-control studies.28 In the present study, being underweight was associated with increased odds of future PD. There is conflicting evidence for BMI as a factor in PD risk. A meta-analysis of prospective studies indicated no association,29 but case-control studies consistently show an inverse association between BMI and PD.28 A large prospective study performed in Korea found an inverse association between BMI and risk of PD, such that higher BMI was protective and lower BMI was a risk factor.30 One issue that arises is that weight loss is commonly encountered in the clinical course of PD, and reverse causation may be a factor in at least some of the association. In an attempt to mitigate this possibility, members of our group reported a potential causal effect of lower BMI and increased risk of PD using mendelian randomization.31
Of the prediagnostic clinical manifestations, tremor showed the highest association with subsequent PD, which was also found up to 10 years before diagnosis. Various studies have reported that tremor may be an early feature of PD.9,32-34 In people reporting tremor years before diagnosis, one must consider the possibility that diagnosis is simply delayed in primary care. That is, individuals seeing general practitioners may not be referred early enough to a movement disorders specialist. Rigidity is a sign rather than a symptom, which might explain why it is rarely reported by patients. Indirect symptoms of rigidity (shoulder pain but not neck pain) were more common in people with subsequent PD than controls, with an association observed up to 10 years before diagnosis.
Compared with THIN primary care data, the association between cognitive symptoms and subsequent PD was higher in the East London population (incidence risk ratio, 2.13; 95% CI, 1.44-3.11) and may be a population-specific observation.9 Patients with PD who are Black have been reported as being more likely to have cognitive impairment.1,19 In a study comparing patients with PD who were Black or Asian with those who were White, it was found that Black individuals with PD had greater cognitive impairment assessed using the Mini-Mental State Examination than White patients.19 However, the method of assessment used may have produced artificially lower scores in certain groups owing to bias induced by language, cultural, and social determinants. It is notable that both ethnicity and socioeconomic deprivation appear to be factors associated with dementia risk (albeit apparently not PD risk), which may explain the higher prevalence of cognitive symptoms in our population compared with the THIN database.17 Depression and anxiety were associated with subsequent PD, but the ORs were not as high as those in other previous studies.7-9
Hearing loss was a novel association with PD identified by this study and occurring up to 5 years before diagnosis. This finding was replicated in the analysis in the UK Biobank. There is an emerging literature on auditory processing difficulties in PD35,36 and impaired recognition of musical and nonverbal vocal emotions.37,38 Although the role of early hearing loss requires further research, it is possible that this factor represents another deficit in sensory processing that occurs as part of PD pathogenesis, similar to visual impairment.39 Subjective anosmia was not prevalent enough to be included in this analysis and is likely to be underrecognized and underreported in primary care. Objective smell testing is the best method of identifying hyposmia, and short smell tests could be conceivably applied in a primary care setting to support earlier diagnosis of PD.
We believe our findings raise potentially important practical considerations for primary care physicians and the opportunity to address patient concerns at an earlier stage of the disease. It is not a case of screening for asymptomatic disease but correctly identifying the underlying cause in patients who are presenting with symptoms and may seek timely onward referral. Patients might otherwise wait for up to 10 years for an explanation for their symptoms. Early treatment of symptoms (motor and nonmotor) may improve quality of life even if patient symptoms do not yet fulfill the clinical diagnostic criteria of PD.40 To this end, more focus on timely diagnosis rather than simply earlier diagnosis may be useful.41 We have outlined some practical advice and observations that might serve a purpose in primary care (Table 5).
There are limitations to this study. The main limitation is that these data were derived from routinely collected primary care data with underascertainment of factors of interest and high missingness. Although the year of recording for each variable was available, data were extracted in a cross-sectional manner, meaning that incidence rates could not be calculated. Another caveat to our study is that, although the NHS provides free care at the point of access to all patients, there may still be underascertainment of PD. For example, there is preliminary evidence for atypical presentations in ethnic minority groups42 and a higher likelihood of being mislabeled with vascular mimics of neurodegenerative disease.1 However, we did not observe significant differences in PD prevalence by ethnic group. Another limitation is the lack of information regarding prescriptions of medication; therefore, it was not possible to create a more robust definition of PD or include additional patients not recorded as having PD but who were prescribed anti-parkinsonian medication or exclude patients with drug-induced parkinsonism. Ascertainment of exposure variables and risk factors may also be incomplete, with mild or transient symptoms not being reported or recorded. Furthermore, some symptoms lack context. For example, memory problems recorded in primary care often lack supportive neurological examination or standardized neuropsychological testing to support a formal diagnosis of cognitive impairment.
In this study, associations between comorbidities and prediagnostic symptoms of PD were noted in a diverse, urban-dwelling population with universal access to health care. Tremor and memory symptoms occurring up to 10 and 5 years, respectively, before diagnosis had the highest level of associations with subsequent PD. Further research is needed to explore the associations between epilepsy and PD as well as hearing loss and PD.
Accepted for Publication: December 3, 2021.
Published Online: March 7, 2022. doi:10.1001/jamaneurol.2022.0003
Corresponding Author: Alastair J. Noyce, PhD, Preventive Neurology Unit, Wolfson Institute of Population Health, Barts and The London School of Medicine and Dentistry, Queen Mary University of London, Charterhouse Square, London EC1M 6BQ, United Kingdom (email@example.com).
Author Contributions: Drs Simonet and Noyce had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Simonet, Jitlal, Ben-Joseph, Marshall, Jacobs, Belete, Giovannoni, Cuzick, Schrag, Noyce.
Acquisition, analysis, or interpretation of data: Simonet, Bestwick, Jitlal, Waters, Ben-Joseph, Marshall, Dobson, Marrium, Robson, Jacobs, Lees, Noyce.
Drafting of the manuscript: Simonet, Jitlal, Ben-Joseph, Marrium, Jacobs, Giovannoni, Noyce.
Critical revision of the manuscript for important intellectual content: Simonet, Bestwick, Waters, Marshall, Dobson, Robson, Jacobs, Belete, Lees, Cuzick, Schrag.
Statistical analysis: Simonet, Bestwick, Jitlal, Waters, Ben-Joseph, Marrium, Jacobs, Belete, Cuzick, Noyce.
Obtained funding: Giovannoni, Cuzick, Noyce.
Administrative, technical, or material support: Robson.
Supervision: Marshall, Lees, Giovannoni, Schrag, Noyce.
Conflict of Interest Disclosures: Dr Marshall reported receiving grants from Barts Charity during the conduct of the study and educational content from Biogen and GE Healthcare outside the submitted work. Dr Dobson reported receiving grants from the MS Society of Great Britain and Northern Ireland, NMSS, BMA Foundation, Horne Family Charitable Trust, and MRC; fees for serving on the advisory boards paid into a university account from Biogen, Roche, Merck, Teva, Novartis, and Janssen; speaking fees from Biogen, Merck, Novartis, and Janssen; and grants from Biogen and Merck outside the submitted work. In addition, Dr Dobson works within the Preventive Neurology Unit of the Wolfson Institute of Population Health, which is partly funded by Barts Charity. Dr Lees reported receiving personal fees from Britannia Pharmaceuticals and Bial Portela for advisory board participation and speaking outside the submitted work. Dr Noyce reported receiving grants from Parkinson’s UK, Aligning Science Across Parkinson's/Michael J. Fox Foundation (GP2), Cure Parkinson’s, Alchemab, Michael J. Fox Foundation/National Institute for Health Research, and Barts Charity; fees for consultancy from uMed, AbbVie, Charco Neurotech, and AstraZeneca; fees for advisory board participation from Britannia, Alchemab, Biogen, Roche, and UCB; and speaker’s fees from Bial outside the submitted work. No other disclosures were reported.
Funding/Support: The Preventive Neurology Unit is funded by Barts Charity. Dr Robson was funded by Barts Charity and by Health Data Research UK, an initiative funded by UK Research and Innovation, Department of Health and Social Care (England) and the devolved administrations, and leading medical research charities. Dr Schrag was supported by the National Institute for Health Research UCLH Biomedical Research Centre.
Additional Contributions: We are grateful to the general practitioners and patients in East London for the use of data from their electronic health records and the Clinical Effectiveness Group, Queen Mary University of London, who provided a deidentified curated extract of the relevant coded data. The code sets used were supported by the Secure Health Analysis and Research in East London study funded by Barts Charity and are available from the authors.
Additional Information: The code used for the statistical analysis is available in the following link: https://github.com/Wolfson-PNU-QMUL/PD_Riskfactors_ELGP_UKB.