The Sydney multicenter study of Parkinson's disease: The inevitability of dementia at 20 years



After 20 years follow-up of newly diagnosed patients with Parkinson's disease (PD), 100 of 136 (74%) have died. The mortality rate fell in the first 3 years of treatment, then rose compared to the general population, the standardized mortality ratio from 15 to 20 years reaching 3.1. Drug induced dyskinesia and end of dose failure were experienced by most patients, but the main current problems relate to the non-levodopa responsive features of the disease. Dementia is present in 83% of 20-year survivors. Dementia correlates with increasing age and probably reflects an interplay of multiple pathologies. Seventeen people with dementia had postmortems. Eight had diffuse Lewy bodies as the only cause of dementia, while others had mixed neuropathology. Only one person lives independently and 48% are in nursing homes. Excessive daytime sleepiness is noted in 70%, falls have occurred in 87%, freezing in 81%, fractures in 35%, symptomatic postural hypotension in 48%, urinary incontinence in 71%, moderate dysarthria in 81%, choking in 48%, and hallucinations in 74%. The challenge is to understand the cellular mechanisms underlying the diverse features of advanced PD that go far beyond a lack of dopamine. © 2008 Movement Disorder Society

Parkinson's disease (PD) is more than a nigrostriatal disorder. Although bradykinesia, tremor, and rigidity respond to dopaminergic therapy many other features do not. These include autonomic failure, hypersomnolance, imbalance, dysarthria, and dysphagia. These and most notably dementia and drug-related neuropsychiatric symptoms are what cause the most concern as the disease advances.1 Although James Parkinson stated in An Essay on the Shaking Palsy that the intellect was “uninjured,” he himself noted that, “at the very last, constant sleepiness, with slight delirium, and other marks of extreme exhaustion, announce the wished-for release.”2 This is still the picture that we see in late PD with dementia affecting most of those who do not die earlier of other causes.

How common dementia is in PD is a matter of debate. Studies from the prelevodopa era include those of Lewy, who in 1923 reported that 54 of 70 (77%) developed dementia and Monroe, who in 1951 noted that a third of people with PD aged over 61 developed a psychosis.3 Nowadays, the prevalence of dementia in community-based studies is about 30%,4 but ranges from about 10 to 80% of people with PD.5, 6 The older the patient and longer the duration of disease, the higher the prevalence of dementia.7–9 Even in young onset, PD (21–40 years) dementia affects 19% after a median of 18 years.10 Cognitive decline is noted in up to 36% of newly diagnosed cases of PD.11, 12 Dementia adds substantially to the burden of disease for the patient, caregiver, and the community.13–15

The Sydney Multicenter Study of PD is unique in having one group of clinicians follow a single cohort of newly diagnosed people with PD over 20 years. The aims of this report are to provide information on the rate of appearance of dementia and the other non-levodopa responsive features of PD, and on the mortality rate.


The design and the progression of the study have previously been reported at 5, 10, and 15 years.1, 11, 16, 17 Sydney neurologists were asked to refer their newly diagnosed PD patients for a 5-year trial comparing low-dose bromocriptine with low-dose levodopa–carbidopa. Between 1984 and 1987,149 denovo PD patients who were Hoehn–Yahr stage 1 to 3 and aged 39 to 79 years were recruited from 29 neurologists. Diagnosis of PD was based on the presence of two of the following: bradykinesia, rigidity and resting tremor in the absence of features suggestive of atypical PD. Patients were not excluded if they had intercurrent disease as long as it was not in the terminal stages. All were living in the community. After the first 6 months of drug titration, the patients were also followed by their referring neurologists.

At their base line neuropsychological assessment, 26 were found to be demented, though in all cases the presentation to their referring neurologist had been with parkinsonism not dementia or hallucinations. These patients would now be labeled as having Dementia with Lewy bodies (DLB). However, all patients considered as having PD by the London Brain Bank criteria18 were included in the study. Thirteen patients were excluded on the basis of having atypical PD (leaving 136). Six patients have been lost to follow-up (leaving 130).

Patients were examined in outpatient clinics for the first 10 years, but at 15 and 20 years were often examined in their home or nursing home if they were too frail to visit hospital. Telephone interviews with the patient and carer and reports of treating physicians were obtained for four people who had moved from Sydney. Neurological assessments have previously been reported.16 Excessive daytime sleepiness was considered present if the patient had multiple sleeps during the daytime hours.

Detailed neuropsychological assessment was performed on 110 patients at study entry by the neuropsychologist (WR) and at 3, 5, 10, 15, and 20 years for people with English as their main language. The neuropsychological assessment battery was designed to sample a range of cognitive domains. Details of the tests and administration have been previously reported and neuropsychological findings will be the subject of a separate paper.11 At the 15- and 20-year follow-up the minimental state examination (MMSE), Clinical Dementia Rating (CDR), and the Boston Naming Test were added to the battery.19–22 Depression was assessed by the 30 item Geriatric Depression Scale (GDS).23 Hallucinations were detected by routine questioning at each visit of the patient, and carer when present. Hallucinations were defined as a perception in the absence of an external stimulus and included presence, passage, complex visual, and auditory hallucinations. Illusions were also included as the altered perception of an external stimulus.

Cognitive function in patients who were not seen by the neuropsychologist was assessed by the neurologist (MH). The MMSE, similarities, letter fluency (words beginning with “f”, “c”, and “s” each in 1 min; normal >9 for each),24 category fluency (animals normal >17) and clock faces were recorded at 15 and 20 years. Clock face drawing was scored from 0 to 4 with 1 point each for a closed circle, all 12 numbers, numbers in correct positions and hands correctly placed. Perseveration was tested for by asking patients to copy a three-loop spiral. A Clinical Dementia Rating was recorded after interview with family members regarding memory and other cognitive functioning and its impact on daily life. A neuropsychological diagnosis of dementia was made on the basis of impairment in memory and at least two other areas of cognitive functioning. A test score that decreased at least two standard deviations from the mean score obtained by the control group signified cognitive impairment. The control group consisted of 50 age, gender, and education matched community living people without PD, who were friends or relatives of the patients. If no neuropsychological assessment was made, a diagnosis of dementia was based on a Clinical Dementia Rating ≥1 with supporting evidence from carers of gradual cognitive decline sufficient to impair daily function. This was supported by abnormalities in MMSE, letter and category fluency, and clock face drawing.

On several occasions during the study, patients were asked to consider donating their brains for diagnostic and research purposes. Informed consent was obtained from the patient and family. The brain donation program at the Prince of Wales Medical Research Institute has both institutional and ethical approval and complies with the guidelines of the National Health and Medical Research Council of Australia. Brain only autopsy was performed and standard neuropathological methods and diagnostic criteria were used.25


The time to event data are presented as Kaplan–Meier curves, where the time to event is censored at the end of follow-up or death if the event has not occurred by that time. For the mortality analysis, the expected number of deaths is the number of deaths expected given the age and sex specific death rates seen in the general Australian population; these rates vary with calendar year, which has been taken into account for the calculations. The standardized mortality ratio (SMR) is the ratio of the observed and expected number of deaths and has been analyzed using Poisson regression. These analyses were performed in Stata (Version 9).26


Thirty patients (15 men and 15 women) have survived to be included in the 20-year follow-up that ranged from 19.8 to 22 years. Their average age is now 74 years (SD = 7.9), and was 54 years at presentation compared to 62 years for the initial total group.


By 20 years, 100 of 136 (74%) had died and three have died since. In the first 3 years, five died compared with an expected 9.2 deaths (SMR = 0.5, 95% CI 0.2–1.1). Between 3 and 20 years, 95 died compared with an expected 38.7 deaths (SMR = 2.5, 95% CI = 2.0–3.0). The SMR was similar for the periods 3 to 5, 5 to 10, 10 to 15, and 15 to 20 years (chi-square test statistic for between period differences = 2.6 for 3 degrees of freedom, P = 0.5) but showed a trend to increase with duration of disease compared to the age matched Australian population after falling in the early years of treatment: SMR = 0.5 at 0 to 3 years, 1.8 at 3 to 5 years, 2.3 at 5 to 10 years, 2.7 at 10 to 15 years and 3.1 at 15 to 20 years. The SMR did not show variation with age and sex (chi-square test statistic for differences with sex or age = 1.0 for 2 degrees of freedom, P = 0.6).

The mean age at death was 76 years (SD = 7.5), 75 years for men and 78 years for women after a mean of 121 and 132 months from entry to the study, respectively. The difference between genders is not significant when compared with the longer survival of women in the general population. The median time from historical onset of disease to death was 12.4 years. Pneumonia was the most common cause of death, 26 of 103 (25%), generally occurring in Hoehn and Yahr stage 5. PD was considered to have been a significant contributor to the death of 55 of 103 (54%). A survival curve is shown in Figure 1.

Figure 1.

Kaplan–Meier plot of time to death and of nursing home placement.

L-Dopa-Induced Motor Complications

The mean levodopa intake is 729 mg/day (SD = 453) with a daily average of 4.7 intakes at 20 years. Eleven still take bromocriptine or an alternative agonist (mean dose equivalent to 18 mg/day (SD = 8) bromocriptine). No patients took anticholinergic medications at 20 years but 3 used selegiline, 7 entacapone, and 5 amantadine. All of those on at least 300 mg of levodopa per day had experienced end of dose failure and dyskinesia. At 20 years, dyskinesia was graded as mild to moderate (UPDRS 1–2) for both duration and severity in all, but 3 patients where it was graded as 3. Three others had pallidotomies for dyskinesia. Sudden “offs” were reported by 5 of 30 and an equal number lacked a good “on” due to inadequate levodopa dosage because of dementia. All others were on for more than half the day.

Hoehn and Yahr Stage and Level of Independence

At 20 years, the mean Hoehn and Yahr stage “on” was 4.2 and “off” was 4.6. Only 1 patient was stage 2 when “on,” and he was the only patient living alone. Now aged 65, he was the last patient to cease work just prior to 15 years of disease. Two patients were stage 3 when “on,” 15 stage 4, and 12 stage 5. Fourteen lived with family or had fulltime live in carers. Fifteen people (48%) lived in nursing homes. A survival curve for nursing home placement is shown in Figure 1. Patients lived there for a mean of 34 months (SD = 27) before death. The mean activities of daily living score (UPDRS items 5–17) “on” was 19 (SD = 8) and “off” was 27 (SD = 8). Excessive daytime sleepiness was recorded in 21 (70%).

Falls, Fractures, and Freezing

Falls were experienced by 27 (87%) of the 20 year survivors and 11 (35%) had sustained fractures, often multiple. This was in spite of active interventions with physiotherapy, walking aids, and home modification. Survival curves for these features are shown in Figure 2. Freezing occurred in 25 people (81%).

Figure 2.

Kaplan–Meier plot of time to falls, dysphagia, symptomatic postural hypotension, and first fracture.

Bladder, Bowels, Blood Pressure, and Bulbar Problems

At 20 years autonomic problems were common with 15 (48%) noting symptomatic postural hypotension, although only 6 required fludrocortisone. Urinary incontinence occurred in 22 (71%), fecal incontinence in 5 (17%), and constipation requiring daily laxatives was present in 12 (40%). Dysarthria (>UPDRS1) was present in 25 (81%), 13 gade 3 and 2 gade 4. Choking was present in 15 (48%). Survival curves for dysphagia and postural hypotension are shown in Figure 2.

Hallucinations and Other Psychiatric Problems

In this group of 20-year survivors, visual hallucinations occurred in 23 (74%) necessitating reduced dopaminergic medication in all, introduction of atypical antipsychotics in 10 (7 quetiapine, mean dose 43 mg/day; 2 olanzapine 10/mg/day; 1 risperidone 1 mg/day) and cholinomimetics in 3 (2 donepezil 10/mg/day; 1 galantamine 16 mg/day). Of the original cohort, 78 were known to have suffered visual hallucinations prior to death, see Figure 3. One patient was physically violent in a psychotic episode. One patient had committed suicide. Two patients developed gambling problems after 10 years of treatment. One had hypersexuality. Other hedonistic behaviors were not noted. Depression was not formally assessed in all patients at 20 years due to cognitive decline, however, 15 (50%) were using antidepressants all at the recommended starting dose or less (8 on serotonin reuptake inhibiters and 7 on tricyclics). Seven of the 10 patients able to be assessed with the GDS showed evidence of depression (mean17, range 11–25).

Figure 3.

Kaplan–Meier plot of time to hallucinations and dementia.


At 20 years, 25 of 30 (83%) surviving patients are demented and 2 have developed dementia after this. The increasing prevalence of dementia is shown in Figure 3. Table 1 shows the pattern of deficits in tests performed by the neurologist compared to the CDR. Results of the neuropsychological assessments will be the subject of a separate paper. However, the neuropsychological assessment confirmed dementia in 5 of 30 prior to 20 years and 11 others at 20 years. Eleven were not seen neuropsychologically in the last 5 or more years necessitating the use of the CDR and MMSE. Three of the 5 listed as CDR 0–0.5 were tested neuropsychologically, and did not meet neuropsychological criteria for dementia. Another 70 patients were demented before they died prior to 20 years, giving a total of 97 (75%) demented predeath. Of the patients who died never having demented, 14 did so early in the course of their PD from unrelated causes. Sixteen died more than 1 year after their last assessment so that their cognitive state just prior to death was unknown.

Table 1. Comparison of CDR, MMSE, and brief cognitive testing for patients examined at 20 years
CDRNo.Age at 20 yearsMMSEMean letter fluency N > 9Animal category fluency, N > 17Mean clockface normal = 4Const. apraxia (%)Perseveration (%)

The mean age at time of diagnosis of dementia was 71.5 years for those 26 demented at presentation (DLB) and 71.6 years for those diagnosed later (PDD). There were 6 people aged 50 to 59 years at the time dementia was diagnosed, 27 aged 60 to 69 years, 56 aged 70 to 79 years, and 8 aged 80 to 89 years, see Figure 4. Excluding those demented at baseline, the mean follow-up until dementia intervened was 10.9 years (SD = 5.5). Once dementia was diagnosed (excluding those diagnosed at baseline), the median survival was 54 months.

Figure 4.

Decade of presentation to the study and of dementia. Age over 80 was an exclusion criteria of the study.


Brain only postmortems were performed for 17 patients who died with dementia (mean age at death 79) and for 4 who died without dementia early in their course from unrelated health problems (mean age at death 70). All had brainstem Lewy body pathology consistent with the diagnosis of PD. Additional large vessel disease and hippocampal sclerosis were found in each of one case without dementia. Limbic and/or neocortical Lewy bodies were prominent features in 8/17 with dementia. In addition to limbic/neocortical Lewy bodies, 3/17 fulfilled NIA-Regan criteria for Alzheimer's disease (AD) and 2 of these had vascular disease including amyloid angiopathy. This latter was also present in 3/17 others as the most prominent additional neuropathology to PD. One had frontotemporal lobar degeneration with Pick bodies as well as PD.25


This is the longest prospective study of PD and is based on 136 newly diagnosed community living patients referred by 29 neurologists in Sydney from diverse social and educational backgrounds. The referring neurologists also continued to manage the patients after the initial drug study. The average age at presentation (62 years, range 37–79) was similar to other studies of this period.18, 27 We feel that these patients are representative of the general PD population.

When patients reach the later stages of their disease, physical frailty and cognitive decline mean they can no longer attend hospital clinics and private neurologists. Their extreme disability is therefore a hidden problem. In this group, only 14/30 (47%) still see their neurologist after 20 years. The later disease stages can often last 5 years and so, at their most vulnerable period, patients with PD are often lacking specialist care. Careful adjustment of medication can markedly improve quality of life even for the nursing home patient, for example, reducing cognitive side effects or dyskinesia. An outreach service from health professionals with knowledge of PD would be of value.


It is notable that, although not statistically significant, mortality briefly declines in the early years of treatment then continues to rise with the increasing duration of disease. This presumably reflects disease progression and the appearance of disease features that do not respond to our current medications such as falls, fractures, dysphagia (leading to pneumonia), autonomic failure, and dementia. All of these increase in prevalence after 10 years of disease and are not discussed here, as they formed the basis of the 15-year report.1


Dementia occurred in almost every patient, even in those with a younger onset of disease, who did not succumb to intercurrent disease. The current article includes the whole available group using the neurologist's cognitive assessment to supplement the neuropsychological findings, as not all patients were assessed neuropsychologically. Direct comparison of the neurologist's and neuropsychologist's diagnosis of dementia revealed that neurologists are more likely to underestimate than overestimate the prevalence of dementia in their PD patients as neuropsychological criteria were at times reached before the MMSE fell below 25 or the neurologist was aware of significant decline in social functioning. Neuropsychological assessment early is invaluable in detecting cognitive decline and guiding interventions but is not readily available in the community. Our findings suggest that unless dementia is actively sought and excluded, it should not be assumed to be absent. It is notable that while it is accepted that the MMSE is a poor tool for examining cognitive decline in PD with dementia (PDD), the CDR scale also tended to under-diagnose dementia (usually rating people 0.5, cognitive decline but not dementia) compared to the formal neuropsychological assessment. The latter is taken for the diagnosis of dementia where performed.

Clinical criteria for the diagnosis of PDD have recently been proposed that include the core features of idiopathic PD and a slowly progressive dementia syndrome that includes impairment in more than 1 cognitive domain that is a decline from premorbid levels. Cognitive deficits must be severe enough to impair daily functioning independent of deficits due to motor or autonomic failure.28 Memory loss is not essential for diagnosis unlike DSM IV criteria or as defined in our neuropsychological diagnosis. Although memory loss is noted to occur commonly as the dementia progresses, it is not always an early feature.29 As in our cases, it is clearly of value to have continuity of follow-up to detect the evolution of these changes and to be able to weight the various contributions of cognitive, autonomic, and motor deficits that impair the social, personal, and occupational functioning of people with PD. The history of a carer is also invaluable and delirium from drugs or systemic illness, and major depression need exclusion. Although neuroimaging may show changes of vascular disease in this older population, the typical insidious onset of PDD suggests that this is usually not the major cause. Extreme bradykinesia with poor motor skills and dysarthria can hamper cognitive assessment late in PD. It is, therefore, essential to perform serial brief regular assessments throughout the disease course to detect cognitive decline, and we encourage neurologists to add some of the simple tests used here to their assessments.

The prevalence of dementia in PD is poorly appreciated in the wider neurological community. It behoves neurologists to examine patients with PD most carefully for dementia as it has a number of consequences including: its association with drug and procedure induced hallucinations and psychoses; excessive daytime sleepiness and carer distress; loss of insight and poor judgment; the inability of patients to make sound financial decisions; impairment of driving skills; unreliability in following advice concerning medication; visuospatial problems that contribute to an increased risk of falls and fractures; unsuitability for deep brain stimulation, and it should suggest that elective surgery for intercurrent problems is best avoided. In association with increasing age, dementia and hallucinations predict nursing home placement.1 Although dementia was predictive of death (P = 0.001) univariately, it is so strongly associated with increasing age as to render it insignificant when age is corrected for multivariately.17 When correcting for severity of motor symptoms, Levy et al. found that incident dementia was associated with a doubling of mortality risk in a group of older PD patients.30

Dementia is unusual in younger patients at presentation. At baseline, 8% of patients aged 40 to 69 showed evidence of dementia by neuropsychological criteria. However, dementia was found in 39% of those aged 70 to 79 years at presentation.11 Today, these people would be called DLB, although their presenting symptoms were motor, not dementia or hallucinations. These patients died early and so do not feature in the percentages of demented patients at 15 to 20 years. Others have also noted cognitive impairment in a proportion of patients with early PD.12 The cognitive profile of PDD and DLB is very similar.31 In a group with the early motor signs of PD, the London Brian Bank was unable to distinguish differences in the physical and cognitive signs and drug responsiveness of those with early dementia of DLB and later dementia of PDD.32

We had no patients with dementia in their 40s but 6 became demented in their 50s, including 1 at 53 after 8 years of disease but still alive at 20 years, and 1 at baseline at age 58 who at postmortem had AD and PD with limbic Lewy bodies. Dementia that is in excess of the general population has been noted in 19% of young onset patients (aged less than 40) after a median of 18 years on a telephone interview: 13% of those still aged under 60 and 43% of those aged over 60.10 In contrast, 83% of our patients are demented after 20 years with an average age now of 74. The strong relationship to age, also noted by others,8, 33, 34 rather than disease duration suggests that additional pathology to Lewy body parkinsonism may have a role in the appearance of dementia. A combination of pathologies is likely to be additive and thus cross the threshold for expression of dementia at an earlier stage than a single disease.

The neuropathology of dementia in PD is heterogeneous.35–38 Staining for cortical Lewy bodies with α-synuclein suggests that limbic and cortical Lewy bodies are the most important association of dementia in PD with or without additional Alzheimer's pathology.37, 39 However, the picture is complicated by the fact that neocortical Lewy body involvement does not always correlate with dementia in PD,32 although high numbers of temporal and parahippocampal Lewy bodies may be better correlates of dementia in PD40; and the suggestion that Lewy body inclusions may represent a protective mechanism.41 The fundamental cause of clinical dysfunction in PDD may lie within as yet undiscovered cellular mechanisms.32

In PDD, treatment is limited. Dopaminergic medication is frequently restricted by the appearance of visual hallucinations. Cholinergic deficits from the nucleus basalis of Meynert are believed to contribute to the fluctuations in levels of alertness in DLB42 and cortical cholinergic loss is found to be more severe in PD than in AD.43 This may explain the mild benefit noted in PDD from rivastigmine.44, 45


In this 20-year study of 136 patients with PD, 30 survive. All have shown an improvement in motor function in response to dopaminergic therapy but clinical features, particularly dementia have now emerged for which this form of treatment is unhelpful. Until we have a better understanding of and can treat dysfunction in the nondopaminergic neuronal systems, we have less than we would wish to offer our patients with advanced PD. What we can offer is an early recognition of dementia and the other non-dopamine responsive problems to anticipate the measures that need to be taken to minimize their consequences for the patient and the carer. Ultimately, a cure for PD will depend on understanding the basic biochemistry of cellular dysfunction that affects so many aspects of the central and autonomic nervous systems.


The National Health and Medical Research Council of Australia funds Professor Halliday. We thank the neurologists who referred patients but most importantly, we wish to thank the patients and their families who were so tolerant in allowing us into their lives over so many years. The initial drug study was supported by Sandoz, Australia (now Novartis) and subsequently by the Australian Brain Foundation. Our nurse, Beverley Zielinski, was invaluable in liaising with patients and families.