Article Text

Research paper
Parkinson's disease beyond 20 years
  1. Roberto Cilia1,
  2. Emanuele Cereda2,
  3. Catherine Klersy3,
  4. Margherita Canesi1,
  5. Anna L Zecchinelli1,
  6. Claudio B Mariani1,
  7. Silvana Tesei1,
  8. Giorgio Sacilotto1,
  9. Nicoletta Meucci1,
  10. Michela Zini1,
  11. Claudio Ruffmann1,
  12. Ioannis U Isaias1,
  13. Stefano Goldwurm1,
  14. Gianni Pezzoli1
  1. 1Parkinson Institute, Istituti Clinici di Perfezionamento, Milano, Italy
  2. 2Nutrition and Dietetics Service, Fondazione IRCCS Policlinico San Matteo, Pavia, Italy
  3. 3Biometry and Clinical Epidemiology Service, Fondazione IRCCS Policlinico San Matteo, Pavia, Italy
  1. Correspondence to Dr Roberto Cilia, Parkinson Institute, Istituti Clinici di Perfezionamento, via Bignami 1, Milano 20126, Italy; roberto.cilia{at}


Background A very limited number of studies report data on the clinical features of Parkinson's disease (PD) 20 years after onset and beyond.

Objective To characterise PD 20 years after onset, investigating the impact of age at onset and disease duration on the clinical picture and the predictors of outcomes in patients reaching the 20-year time point.

Methods We conducted a retrospective, cross-sectional study and a longitudinal study. All case visits of patients with a disease duration ≥20 years (N=401) were stratified by disease duration (20–22, 23–25, ≥26 years) and by age at onset (cut-off, 50 years). Patients with a disease duration of 20–22 years (N=320) were prospectively followed up for a median of 45 months (IQR 23–89) for the new occurrence of fracture, percutaneous endoscopic gastrostomy, institutionalisation, confinement to a wheelchair or bed and death.

Results Older age at onset and longer disease duration were independently associated with a higher prevalence of major motor and non-motor milestones of disease disability (no interaction observed). In the longitudinal study, the most frequent outcomes were death (N=92), confinement to a wheelchair or bed (N=67) and fracture (N=52). Mortality was associated with the gender: male, older age, dysphagia, orthostatic hypotension, postural instability, fractures and institutionalisation. Fracture was associated with postural instability. Predictors of permanent confinement to a wheelchair or bed were older age, postural instability and institutionalisation. Comorbid dementia at the 20-year examination did not predict any of the outcomes.

Conclusions Age at onset and disease duration are independent determinants of the clinical features of PD beyond 20 years. Non-motor symptoms depend more on age at onset rather than the disease duration itself. Non-levodopa-responsive axial symptoms are the main predictors of all relevant outcomes.


Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


The mean duration of Parkinson's disease (PD) from onset to death is approximately 15 years1 and a very limited number of studies report data on its clinical features 20 years after onset and beyond.2 ,3 The emergence of non-motor disability milestones and levodopa-resistant motor symptoms (such as dementia, psychosis, postural instability and falls) dominates the late stage of PD, as they are associated with loss of independence and increased mortality.4–6 However, the clinical evolution of PD can be very heterogeneous1–3 ,7 and the relevance of the various prognostic factors is still debated.8 In particular, motor and cognitive disability milestones usually occur later in patients with younger onset,5 ,8–10 and the ageing process accelerates progression after the age of 70 years, regardless of previous disease duration or age at onset.11 In this scenario, whether or not ageing and disease duration act synergistically or merely exert additive effects is yet to be clarified.5 ,9 ,12 ,13

The primary objective of the present study was to characterise the main clinical features of the disease 20 years after onset. To address this issue, we used two different approaches. First, we investigated the relative impact of age at onset and disease duration on the main motor and non-motor symptoms in different cohorts of disease duration using a cross-section analysis. Second, we explored in a prospective study which factors evaluated at the 20-year time point predict the outcomes in individual patients as their disease progresses.


Standard protocol approvals, registrations and patient consents

The study was performed in agreement with the principles of the Declaration of Helsinki and the protocol was approved by the local Ethics Committee. We obtained written informed consent from every patient recruited.

Data source and extraction

We conducted a retrospective cross-sectional prevalence data analysis and a longitudinal study.

We used the Parkinson Institute research database, which contains detailed demographic, clinical and lifestyle information on all patients assessed. Idiopathic PD was diagnosed according to the UK Brain Bank criteria.14 We excluded all patients with major vascular abnormalities at the visual analysis of brain MRI (97.5% of cases) or CT (namely stroke, diffuse confluent lesions in hemispheric white matter, and lacunar infarcts in the brainstem and/or basal ganglia—as the primary aim of the present study was to investigate the role of ageing, we did not adopt a threshold for vascular load that was too low because it would have been a source of selection bias).

We extracted data on all visits (January 2000—January 2013) of patients with PD with a disease duration ≥20 years. For every patient, we considered the most recent record within the period for each of the following cohorts of disease duration: 20–22, 23–25 and ≥26 years (with a patient potentially being included in different cohorts of disease duration).

The following information was extracted: demographics, family history of PD (first-degree relatives), genetic status (see online supplementary material), lifestyle and exposure (cigarette smoking) and clinical assessment data. Clinical assessment included the complete Unified Parkinson Disease Rating Scale (UPDRS) from parts I to IV, the Hoehn and Yahr (HY) staging system and the complete pharmacological and surgical therapy (ie, deep brain stimulation) of PD and related non-motor symptoms. UPDRS-Part II scores (activities of daily living) and UPDRS-Part III scores (motor examination) were rated in the morning in the patient's worst-OFF and best-ON conditions, while the HY stage and the motor phenotype were assessed in the medication-OFF state.15 Patients with incomplete clinical data (ie, where no UPDRS score in the ON state was available) were excluded from the analysis. UPDRS-Part III scores in the ON and OFF states could have been collected during two consecutive visits carried out within 6 months.

Information on pharmacological therapy included the calculation of total levodopa daily dose, taking into account equivalent doses of other antiparkinsonian medications.16

Assessment of clinical features and study end points

Clinical features of the disease were defined according to clinical assessment in agreement with internationally established diagnostic criteria,17 ,18 UPDRS scores and current pharmacological therapy (table 1). As not all patients underwent extensive neuropsychological assessment, an internal validation of the diagnosis of dementia according to the Diagnostic and Statistical Manual, Fourth Edition, Text Revision (DSM-IV-TR) was performed by using data on Mini-Mental State Examination (score <24) from 222 patients consistent with an almost perfect agreement (Cohen's κ=0.803 (95% CI 0.723 to 0.882)19). In the cross-sectional data analysis, these characteristics were grouped as follows: non-motor symptoms, non-levodopa-responsive motor symptoms and motor complications.

Table 1

Definition of clinical features

Vital status and the cause (according to the rules of the International Classification of Diseases, 10th Revision) and date of death were ascertained by means of record linkages with death indices or by active follow-up (inquiries by telephone to participants, proxy respondents and municipal registries). Occurrence, date and site of first fracture were ascertained by active follow-up (inquiries by telephone to participants or proxy respondents) and confirmed by direct reference to specific discharge charts. The same method was used for the ascertainment of percutaneous endoscopic gastrostomy (PEG). Finally, confinement to a wheelchair or bed (HY stage, 5) and any-type institutionalisation (nursing home placement or admission to any other long-term care facility) were ascertained by active follow-up (inquiries by telephone to participants or proxy respondents).

Statistical analysis

Descriptive statistics were provided for continuous and categorical variables. Prevalence of clinical end points across the cohorts was computed together with their binomial 95% CIs.

Cross-sectional analysis

Case patients were initially grouped into three cohorts according to disease duration at assessment (20–22, 23–25 and ≥26 years). The conservative approach of a panel data analysis (with a patient potentially being included in different cohorts of disease duration) was chosen to achieve the highest sample size and statistical power. General linear models, either with identity or logit link, according to the dependent variable, were fitted to assess the association of age at onset dichotomised at median age of the cohort assessed at a disease duration of 20–22 years (<50 years (younger-onset PD) and ≥50 years (older-onset PD)) and disease duration (3 cohorts) with patient characteristics. Interaction of age at onset and disease duration was tested. Huber-White robust SEs were computed to account for within-patient correlation.

Longitudinal data analysis

Patients with a disease duration at assessment of 20–22 years were included in this analysis. The risks (reported as HR and 95% CI) of death, fracture, permanent confinement to a wheelchair or bed, PEG and institutionalisation were analysed with the use of Cox regression models, including time-dependent variables. For predictors presenting during follow-up, time-dependent Cox models were fitted. Non-collinear variables considered to be clinically relevant or having a p value ≤0.1 at univariate analysis were retained in the final model. Exclusion of collinear variables was based on consensus among the authors and consistency with literature data. Prior to inclusion in the model, the interaction between any predictor and age at PD onset (<50 and ≥50 years) was also investigated. The Harrell's c-statistic and the Royston's explained variation were computed for model validation. A performance index (PI) was computed as the linear combination of the coefficients estimated in the Cox model to stratify patients in the low, intermediate and high-risk groups based on the tertiles of its distribution. Kaplan-Meier cumulative survival by PI was computed accordingly and was used to assess model discrimination. Finally, to minimise the confounding effect of genetic status on disease severity and progression, sensitivity analyses were performed by refitting all models after excluding patients with a known PD-related pathogenic mutation.

All p values presented are two-tailed, and p values <0.05 were considered to indicate statistical significance. Analyses were performed with the software STATA V.13 (StataCorp, College Station, Texas, USA).


Cross-sectional analysis

In total, data from 401 patients were included in the panel data analysis. General features (clinical and demographic) are presented in table 2. Two and three evaluations were available for 146 and 55 of these patients, respectively, providing a total of N=596 case visits (figure 1). Pathogenic gene mutations were present in 28 patients, of whom 24 had younger-onset PD (online supplementary material). The exclusion of patients with pathogenic mutations did not affect the results (data not shown).

Table 2

General clinical and demographic characteristics of the study population by disease duration and age at onset

Figure 1

Study flow chart of the study population. aEleven patients of 438 (2.5%) underwent a brain CT scan instead of MRI for safety reasons (cardiac pacemakers, N=6; heart valve prostheses, N=3; other devices or implants, N=2). Of these, three patients were excluded from the study population for major vascular abnormalities (PD, Parkinson's disease).

Age at onset was independently associated with most motor and non-motor symptoms, patients with younger-onset PD generally presenting with less severe disease and a better response to levodopa (ie, UPDRS motor scores lower in the ON state only, but similar in the OFF state; figure 2 and table e-1). Disease duration was associated with worse non-levodopa-responsive motor symptoms and non-motor symptoms, along with a further worsening of motor fluctuations and dyskinesias (figure 2 and table e-1). All patients experienced motor fluctuations during the course of the disease and overall dopaminergic therapy remained stable across the time points considered. Most treatment features were unrelated to disease duration, with the exception of the use of antipsychotics and drugs for the management of orthostatic hypotension (table e-1). No significant interaction between disease duration and age at onset was observed for all the end points considered.

Figure 2

Clinical and demographic characteristics of the study population (cross-sectional panel study) by disease duration and age at onset (*p value for disease duration; **p value for age at onset). Patients with age at Parkinson's disease onset <50 years are in grey, age at onset >50 years in black colour.

There was no interaction between age at onset and disease duration for all motor and non-motor symptoms also between 10 and 20 years of disease. The only exception was disability related to levodopa-related motor complications, as patients with younger-onset PD had more motor fluctuations and dyskinesias as the disease progressed (for interaction between age at onset and disease duration, p=0.009 and 0.044, respectively; table e-2).

Longitudinal study

Among patients with a disease duration at assessment of 20–22 years (N=320), 17 were lost to follow-up. Accordingly, prospective data analysis relied on a final sample of 303 patients (figure 1).

The results of hazard regression analysis of predictors of the most frequent outcomes occurring during a median follow-up of 45 months (25th–75th centile, 23–89) are reported in table 3. None of the predictors tested in bivariate analysis showed an interaction with age at onset prior to inclusion in hazard models. All-cause mortality was positively associated with both baseline features (gender, age, dysphagia, orthostatic hypotension and postural instability) and time-dependent events (fracture and institutionalisation). Fracture was significantly associated only with postural instability. Predictors of permanent confinement to a wheelchair or bed were institutionalisation (time-dependent), age and postural instability. Risk discrimination based on regression coefficients derived by Cox's models predicting these outcomes was found to be good (figure e-1). PEG (N=11) and institutionalisation (N=21) were less frequent outcomes, and were mainly predicted (gender and age-adjusted) by dysphagia (HR=14.44 (95% CI 3.55 to 58.74); p<0.001) and dementia (HR=2.81 (95% CI 1.07 to 7.34); p=0.036), respectively. Psychosis was also a predictor of institutionalisation, but was not included in the same model with dementia due to high collinearity: at multivariate analysis (gender and age-adjusted), HR=3.42 (95% CI 1.34 to 8.74; p=0.011). Sensitivity analyses excluding patients with a known pathogenic gene mutation yielded similar results (data not shown).

Table 3

Predictors (multivariable Cox's model) of outcomes in patients with disease duration of 20–22 years (prospective study)

Of the 303 patients entering the longitudinal study, 121 had a second evaluation at our institute (follow-up (mean±SD), 31.5±15.9 months) between 23 and 25 years of disease duration; of these, 41 had a third evaluation. Clinical features of patients who had one or more follow-up visits (N=121) were similar to those who had been examined only once (N=199; data not shown).


In the present study, we investigated a large cohort of patients reaching a disease duration of 20 years using two different methodological approaches, such as a retrospective cross-sectional analysis and a longitudinal study. In the first part of the study, we highlighted the pivotal role of age at PD onset in the equation representing the rate of all major motor and non-motor symptoms, related to the differential age-related impairment of dopaminergic and non-dopaminergic neuronal systems. In the second part, we investigated the prognostic impact of demographic and clinical variables recorded at the 20-year time point on a number of PD-related long-term outcomes. Overall, the clinical picture was dominated by levodopa-resistant motor features, particularly axial symptoms, which were the most important predictors of all relevant outcomes. It is important to emphasise, however, that levodopa did not reduce its effectiveness on the dopaminergic motor symptoms of PD in the long term.

The aim of the present investigation was not to provide a description of the disease course up to 20 years of duration, but to describe the disease and the implications for outcome of a number of disease milestones when a patient finally reaches this time point. This population is unique, as it has survived despite a number of selection pressure factors.20 A key element for such a long disease course is age at onset, which was confirmed to be a source of variability in the major milestones of disability.3 ,7 ,21 In comparison with previous studies on long-standing PD, we stratified analysis on all major PD-related motor and non-motor symptoms for age at onset. This approach enabled the identification of a subgroup of patients with PD that was fairly representative of the general PD population, in view of their comparable age at onset.1 Interaction analysis consistently showed that age at onset and disease duration are independent contributors to all motor and non-motor milestones of disability. It is likely that this effect applies at least beyond 10 years of PD duration. On the one hand, longitudinal studies with a follow-up ≤10 years demonstrated a significant interaction between age at onset and disease duration on both cognitive impairment and levodopa-resistant motor symptoms.9 ,13 ,22 ,23 The hypothesis that the main impact of age at onset on the progression of the degenerative process occurs in the early to mid stages of the disease is further supported by a recent clinicopathological study.5 On the other hand, however, our data in patients with a history of 10–20 years of disease confirmed only a higher likelihood of developing motor complications (motor fluctuations and levodopa-induced dyskinesias) in patients with younger-onset PD.10 ,24 ,25 Taken as a whole, these data suggest a differential age-related rate of progression only in the short term, while these factors are likely to have an additive effect on the natural history of PD in the long term.12 ,26 This differential progression of clinical features is supported by a slower progression of neuronal loss in patients with earlier PD onset, who have more efficient compensatory mechanisms.27 Besides delaying the onset of motor symptoms, it might be speculated that better dopaminergic and non-dopaminergic compensatory mechanisms might additionally modulate symptoms progression over time,28 and possibly explain the reduced severity of motor symptoms (including levodopa-resistant ones) beyond 20 years in patients with younger age at onset. Likewise, as compensatory mechanisms occur also in Alzheimer's disease and dementia with Lewy bodies,28 ,29 it might be hypothesised that this age-related effect could similarly apply to PD non-motor symptoms.

Compared with older-onset PD, patients with earlier onset responded better to levodopa and had a lower incidence of major non-motor symptoms (ie, dementia and psychosis) and axial symptoms (ie, gait disturbances and imbalance), with subsequent loss of independence according to the HY stage.7 ,10 ,24 ,25 Nonetheless, mean age at onset of the older-onset PD subgroup at the 20-year assessment was comparable to the general PD population1 and, most notably, a large number of patients with older-onset PD were still independent 20 years after onset. Accordingly, dementia and levodopa-resistant motor features should not be considered inevitable after 20 years of disease, even in patients with older age at onset. Our findings support the conceptualisation of late-stage PD as a definition to be applied to patients who are highly dependent on caregivers for their activities of daily living,6 irrespective of their disease duration. Besides age at onset, a number of other prognostic factors are likely to influence the progression of disability in the long term and still need to be identified. Among these factors, genetic status13 ,30 and tremor-dominant motor phenotype10 did not play a major role in our cohort. The latter observation is consistent with a clinicopathological study where the tremor-dominant presenting phenotype did not predict a longer survival.10

Longitudinal analysis confirmed the pivotal role of dysphagia, postural instability and fractures in prognosis (morbidity and mortality) in advanced-stage PD.5 ,31–35 In contrast to previous studies, dementia did not predict mortality in patients who had already survived 20 years. Despite being associated with mortality in univariate analysis, dementia predicted only institutionalisation,11 but not mortality20 or fractures36 in multivariate analysis. This discrepancy might be due to the mean shorter disease duration at death in most of these studies.5 ,10 ,11 It might be argued that the predictive value of cognitive impairment may be higher in the earlier stages of the disease,37 ,38 while other factors are more likely to influence prognosis once a patient reaches the 20-year PD duration time point. In particular, our data suggest that another non-motor symptom such as orthostatic hypotension is likely to be underestimated in asymptomatic patients and emphasise the importance of its assessment in a routine clinical setting, even in patients with younger-onset PD.31

We acknowledge that our study has some limitations. First, inclusion bias could not be fully excluded due to the cross-sectional study design and the hospital-based setting, which were likely to capture patients in better prognostic groups than longitudinal community-based studies. On the other hand, however, our specialised tertiary movement disorder centre is a referral clinic for advanced-stage therapies (ie, deep brain stimulation and infusion therapies) and nutritional issues (ie, dysphagia), thus increasing the relative frequency of patients presenting with long-standing disease. Longitudinal studies aiming to go beyond 20 years after onset are clearly underpowered if large populations are not included, and following patients starting from the very first years of the disease may also lead to inclusion bias (eg, cases with early dementia fulfilling diagnostic criteria for Lewy body dementia2 ,21 ,38). Second, the retrospective nature of the study did not allow accounting for factors potentially affecting the results. Particularly, given that significant changes in the therapeutic approach occurred in the last decades, it could not be excluded that the pharmacological history of the patients could have affected the course of the disease to some extent. Third, this study lacks a systematic neuropsychological assessment using a comprehensive cognitive testing battery. However, dementia was diagnosed by established criteria and cross-validated in a subgroup with appropriate testing.

The main strengths of the study are the large cohort from a single tertiary referral institute and the homogeneous and standardised approach to the patient in terms of assessment and treatment. Our findings in younger-onset PD are free from the possible confounding effect of genetic status.3 ,29 The role of concomitant major cerebrovascular disease in the aetiology of cognitive decline and levodopa-resistant motor features26 ,39 ,40 and the related loss of independence have never been examined beyond 20 years from onset.2 ,3 Finally, only the systematic assessment in the medication ‘OFF’ and ‘ON’ states provides a reliable clinical picture of disease severity and related disability, and enables minimisation of the confounding effect of medication in the relative balance between levodopa-responsive versus non-responsive symptoms.13 ,37


Age at onset and disease duration are independent determinants of the clinical features of PD in the long term (20 years of disease duration and beyond). Non-levodopa-responsive axial symptoms are the main contributors to the clinical picture and the main predictors of all relevant outcomes, while non-motor symptoms largely depend on age at onset rather than disease duration per se. This different profile in younger patients might be due either to a slower progression of neuronal degeneration or to better dopaminergic and non-dopaminergic compensatory mechanisms,27–29 or both. Further research is needed to elucidate the prognostic factors for long-term survival in this segment of the PD patient population.


The authors wish to thank Jennifer S Hartwig, MD, for assistance in editing the manuscript, Mrs Franca Felici for active contribution in database management, and Angelo Antonini, MD, for data acquisition and helpful comments.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • RC and EC contributed equally to this study.

  • Contributors EC and RC contributed to the study concept and design, and drafted the manuscript. MC, RC, SG, IUI, CBM, NM, GP, CR, GS, ST, ALZ and MZ contributed to the acquisition of the data. EC, RC, CK and GP contributed to the analysis and interpretation of the data, and a critical revision of the manuscript for important intellectual content. CK conducted statistical analysis. EC, RC and GP provided administrative, technical or material support. GP supervised the study. RC is the study guarantor.

  • Funding This work was supported by the “Fondazione Grigioni per il Morbo di Parkinson” and the Fondazione IRCCS Policlinico San Matteo.

  • Competing interests RC received honoraria for symposia from Boehringer-Ingelheim, Lundbeck and UCB Pharma. Emanuele Cereda received consultancy honoraria and investigator grants from the “Fondazione Grigioni per il Morbo di Parkinson”, the Fondazione IRCCS Policlinico San Matteo and Nutricia Italia. GP obtained the funding.

  • Ethics approval Ethics Committee ICP.

  • Provenance and peer review Not commissioned; externally peer reviewed.