Movement Disorders

Vol. 18, No. 1, 2003, pp. 19 –31

© 2002 Movement Disorder Society

Research Review

Systematic Review of Incidence Studies of Parkinson’s Disease

Dominique Twelves, BSc,1 Kate S.M. Perkins, MCRP (UK),2 and Carl Counsell, MD2*
1
University of Edinburgh Medical School, Scotland, United Kingdom
2
Department of Neurology, Aberdeen Royal Infirmary, Scotland, United Kingdom

Abstract: Incidence studies of Parkinson’s disease (PD) are case. In 16 studies, attempts were made to confirm the diagno-
important for both health-care planning and epidemiological sis by examination of patients by a specialist as part of the
research. This report reviews the methods and results of pre- study. None of the studies used identical methods, but five were
vious incidence studies of PD and makes recommendations for sufficiently similar to merit comparison. Four of these gave a
future studies. Original articles that described the incidence of similar incidence (16 –19/100,000/year), but one from Italy had
PD were located using several strategies. The methods were a much lower incidence (8.4/100,000), the reason for which
summarised, and the results of studies with similar methodol- was unclear. Five studies found significantly greater incidence
ogies were compared on a standardised population. Twenty- in men. This review highlights the difficulties in performing
five incidence studies were included. Each used different meth- good quality incidence studies of PD. Further incidence studies
ods to identify incident patients, although most screened both using standardised methods are required. A set of minimal
primary care and hospital records. Only eight studies were scientific criteria has been devised to improve the quality and
prospective, and only two of these had any follow-up. The consistency of future studies. © 2002 Movement Disorder
diagnostic criteria for PD varied (11 studies used two or more Society
cardinal motor features, four used the UK Brain Bank criteria), Key words: Parkinson’s disease; parkinsonism; incidence;
as did the exclusion criteria and the definition of an incident epidemiology; meta-analysis

Incidence studies of disease are important for both vious reviews of such incidence studies,1– 6 but none
health-care planning and epidemiological research. Data defined their search strategy, few gave inclusion criteria
on incidence are required so that health services can be or compared study methodology so that reasonable com-
planned in advance to accommodate the predicted num- parisons could be made, and none made recommenda-
bers of patients who will require treatment. In addition, tions for future studies. Therefore, we wished to review
variations in incidence by time, place, and gender can previous incidence studies of PD before performing a
give clues to possible environmental and other causes of new incidence study to (1) assess the quality of their
the disease and follow-up of an inception cohort defined methods, (2) identify whether any had performed long-
during an incidence study is the best way to learn about term follow-up of patients to study natural history, (3)
the natural history of the disease. review results of comparable studies to determine if there
Parkinson’s disease (PD) is a common disease of was evidence of real differences in incidence between
unknown aetiology; therefore, reliable data on its inci- different areas or time periods, and (4) make recommen-
dence are particularly interesting. There have been pre- dations for further studies.

*Correspondence to: Carl Counsell, Department of Medicine and MATERIALS AND METHODS
Therapeutics, University of Aberdeen, Polwarth Building, Foresterhill,
Aberdeen AB25 2ZD, United Kingdom. The data compiled for this review were identified
Received 14 January 2002; Revised 25 June 2002; Accepted 2 July through a three-staged process: 1) collection of possibly
2002
relevant articles using several sources, 2) selection of def-
initely relevant articles, and 3) analysis of selected studies.

19
20 D. TWELVES ET AL.

Collection and Identification retrospective, the percentage of possible cases examined

Individual search strategies were designed for Medline by a specialist as part of the study, whether the cases
and Embase, and searches were run from the earliest were followed up, and the percentage confirmed by his-
possible year, 1966 and 1980 respectively, to December tology. Crude incidence rates, age and sex-specific inci-
2001.† These reference lists were downloaded into bib- dence rates, male-to-female incidence ratios (standard-
liographic software (Reference Manager v. 8; Research ised for age), and the mean age at the onset of PD were
Information Systems, Carlsbad, CA) and duplicates were also collected.
removed to give 1,648 titles and abstracts, which were
Results Comparison
read through. A preliminary list of 333 references that
possibly described incidence studies was selected. If in The results of different studies were directly compared
doubt and if abstracts were absent, the reference was only if they used sufficiently similar methods to make
included. These 333 titles and abstracts of the prelimi- such comparisons valid. It was decided that such studies
nary list were then read through by the experienced must have used similar diagnostic and exclusion criteria.
author (C.C.) and 94 references were identified that Their search strategies had to include methods to screen
looked as though they could be relevant to the incidence both the hospital and community populations for possi-
of PD (70 original studies, 24 reviews). A total of 239 ble cases of PD to maximise the number of cases found.
references were excluded, because they were either du- To improve diagnostic accuracy, attempts must have
plicates or were not relevant, for example prevalence been made to confirm or refute the diagnosis in difficult
studies. Further articles were identified from the refer- cases by examination by a study expert. Finally, the
ence lists of retrieved articles. These included articles incidence rates had to be age stratified so that they could
that were published before the earliest limit of the data- be standardised to the 1990 Scottish population for direct
bases and articles appearing in journals not indexed in comparison, thus avoiding differences in incidence due
Medline or Embase, for example, conference proceed- to differences in the age structure of each baseline pop-
ings. Finally, the Science Citation Index was used to ulation. Standardised rates and ratios were calculated
track earlier landmark articles forward in time, particu- with confidence intervals, where possible, using the Con-
larly to check whether any incidence study had carried fidence Interval Analysis program (BMJ Publications,
out long-term follow-up that was reported later. This 1989).
strategy did not identify any relevant new references.
RESULTS
Study Inclusion Criteria One hundred twenty-three articles (94 identified from
Predefined criteria were applied to select the final list electronic searches, 29 from reviewing reference lists)
of articles to be included in the review. Articles had to be were requested, and all were obtained and read in full.
full text (i.e., articles published only as abstracts were Ninety-eight of these were excluded, mainly because
excluded), and describe an original study that provided they were not original descriptions of incidence studies.
an incidence rate for the whole population (not restricted Two were excluded because, although they were original
by age) for idiopathic PD alone or for parkinsonian studies in parts of Sardinia,7,8 their results were included
syndromes including PD. in a larger study of the whole of Sardinia,9 which has
been included. One recent excellent study from Italy was
Assessment and Data Extraction excluded because it looked at the incidence only in those
All data about the methodology and results from each between 65 and 84 years old.10 Twenty-five incidence
included study were extracted by one author and then studies, therefore, were included in this review.9,11–34
double-checked by another author. Data from non-En- Their methods are described in Table 1 and their results
glish language articles were extracted after translation. in Table 2. For simplicity and to improve comparability,
Information was recorded about the study’s location, age strata were converted to a standardised set in Table
incidence period, and population, the sources used to 2; therefore, some of the figures given here are approx-
identify possible cases, methods to identify included imate as different strata were used in the original studies.
cases, the definition of incident cases, inclusion and
exclusion criteria, whether the study was prospective or Study Methodology

Methods Used to Define Base Population.

†
Individual search strategies designed for Medline and Embase are In six studies, the methods used to define the base
available from the authors. population were not clearly stated.16,17,19,20,24,26 Most of

Movement Disorders, Vol. 18, No. 1, 2003

 TABLE 1. Methods of incidence studies of Parkinson’s disease
% Possible
Study cases
(population Incidence Sources to identify Methods to identify examined by Definition of incident Inclusion/diagnostic
size) period possible cases included cases specialist* cases criteria Exclusion criteria Prospective
Rochester, 1935–66 Medical records Case notes review 0 Symptom onset and PD (not defined) Isolated tremor, familial No
USA11 -Mayo clinic diagnosis Vascular parkinsonism tremor, drug-induced
(33,400) record linkage Postencephalitic parkinsonism
Death certificates
Iceland12 1954–63 Medical records Diagnosis confirmed Many, actual Not defined PD—an unclear Essential tremor, senile No
(171,500) -Neurology by a doctor figures not combination of rest tremor, drug-induced
-Some other doctors stated tremor, rigidity, parkinsonism,
Social security ⫹ masked face, akinesia, postencephalitic,
disability records flexion posture, and traumatic
Inquiries to all the other signs of
doctors ⫹ parkinsonism
pharmacists about Vascular Parkinsonism—
PD drugs an unclear
Tour of country and combination of tremor,
word of mouth rigidity, and other
neurological signs (eg,
pyramidal or
cerebellar), dementia,
hypertension and
retinal changes
Carlisle, UK13 1955–61 Medical records Case notes review Some, actual Symptom onset Not specified Not specified No
(69,400) -Hospital figures not
-GP stated
Death certificates
Victoria, 1958–64 Medical records Not stated Not stated Diagnosis Parkinsonism (not Drug-induced No
Australia14 -Hospital defined) parkinsonism
(83,001) -GP -PD
Mental institutions -Vascular
Geriatric institutions -Postencephalitic
GP referrals
Sardinia9 1961–71 Medical records Examination by 61 Not defined ⱖ2 of 4 cardinal signs Essential tremor, drug- No
(1,448,319) -Hospital neurologist of Note: People who PD induced
REVIEW: INCIDENCE OF PARKINSON’S DISEASE

-GP those not had died or moved Postencephalitic parkinsonism,

-Neurologist previously seen away by end of cerebrovascular
Health insurance by neurologist ⫹ study period were disease
records consenting. excluded
Pension records Case notes review
of others
Hawaii15 1965–94 Medical records Case notes review Most, actual Diagnosis ⱖ2 cardinal signs Secondary parkinsonism Partial
(8,006) -Hospital Examination by figures not Improvement after (including drug-
-Neurologist neurologist stated levodopa induced,
Death certificates postencephalitic,
Re-screening of traumatic)
cohort with
questionnaire ⫹
examination

Movement Disorders, Vol. 18, No. 1, 2003

21
 22
TABLE 1. (Continued)
% Possible
Study cases
(population Incidence Sources to identify Methods to identify examined by Definition of incident Inclusion/diagnostic
size) period possible cases included cases specialist* cases criteria Exclusion criteria Prospective
Rochester, 1967–79 Medical records Case notes review 0 Not defined 1 or more of: Transient drug-induced No
USA16 -Record linkage 1) diagnosis of parkinsonism
(53,885) system Parkinsonism by
neurologist
2) all 3 cardinal signs
3) facial akinesia and

Movement Disorders, Vol. 18, No. 1, 2003

asymmetrical
bradykinesia and
rigidity
4) histological findings
characteristic of IPD
Ferrara, 1967–87 Medical records Case notes review 32 Symptom onset ⱖ2 of 4 cardinal signs Essential tremor, No
Italy17 -Hospital Examination by Progressive deterioration drug-induced
(187,381) -Rehabilitation neurologist parkinsonism,
centres vascular
-Neurologists parkinsonism, MSA
Health insurance
records
Nursing homes
Drug prescriptions
GP telephone survey
Southwest 1968–70 Medical records Case notes review 89 Not defned ⱖ2 of 4 cardinal signs Essential/senile tremor, No
Finland18 -All physicians Examination by PD drug-induced
(402,988) -All Hospitals neurologist Postencephalitic parkinonism, vascular
D. TWELVES ET AL.

Residential homes parkinsonism, other

Questionnaires extrapyramidal
-Health visitors diseases
-District nurses
Health insurance
records
Inquiries to all GPs
& physicians
Aarhus, 1968–72 Medical records Examination by 86 Symptom onset ⱖ2 of 3 cardinal signs Essential tremor, drug- Yes (with
Denmark19 -Hospital neurologist PD induced parkinsonism follow-
(242,151) -GP Postencephalitic up
-Rehabilitation planned)
centres
Nursing homes
Yonago City, 1975–79 Medical records Case notes review 40 Not defined ⱖ2 of 4 cardinal signs Essential tremor, drug- No
Japan20 -University hospital Examination by induced
(121,812) Inquires to neurologist if parkinsonism,
-Other hospitals diagnosis in doubt cerebrovascular
-GP disease
Health insurance
records
 TABLE 1. (Continued)
% Possible
Study cases
(population Incidence Sources to identify Methods to identify examined by Definition of incident Inclusion/diagnostic
size) period possible cases included cases specialist* cases criteria Exclusion criteria Prospective
Olmsted 1976–90 Medical records Case notes review 0 Symptom onset Parkinsonism: ⱖ2 of 4 Drug-induced No
County, -Mayo clinic record cardinal signs parkinsonism,
USA21 linkage PD: All of vascular
(95,000) -Response to levodopa parkinsonism, PSP,
-No prominent/early MSA, Lewy-body
signs of other dementia
Parkinsonian syndrome
-No secondary cause
Benghazi, 1982–84 Medical records Examination by 100 Not defined PD: ⱖ2 of 4 cardinal Not defined Unclear
Libya22 -Clinics neurologist signs
(518,745) -Hospitals
-Rehabilitation
centres
-Neurology centre
Disability records
Medical care abroad
records
Netherlands23 1983–85 GP reports sent to Examination by 100 Diagnosis PD (not defined) Secondary parkinsonism Yes
(163,577) disease register of neurologist
the sentinel
stations
Poznan, 1985–88 Medical records Case notes review 0 Symptom onset PD defined as a Drug-induced No
Poland24 -Neurology progressive parkinonism,
(1,308,000) -Neurosurgery neurological problem postencephalitic,
-Chronic disease including rigidity, vascular parkinonism
institute akinesia, and rest
-GP tremor
China25 1986 Door-to-door Examination by 100 Diagnosis PD: Schoenberg’s Toxin/drug-induced Yes
(3,869,162) questionnaire neurologist criteria, ie, symptoms/ parkinonism,
distributed by signs of insidiously postencephalitic,
medical workers progressive rest vascular parkinsonism
REVIEW: INCIDENCE OF PARKINSON’S DISEASE

tremor, rigidity and

emotional hypokinesia,
cases without definite
causes and attacks
occurring after middle
age
Southeast 1986–88 Medical records Case notes review Most, actual Symptom onset PD: All of the following, Essential tremor, drug- No
Sweden26 -Neurologist Examination by figures not 1) at least 1 of: tremor, induced parkinonism,
(147,777) Inquiries to all neurologist if not stated rigidity, hypokinesia parkinsons-plus
neurologists, previously seen 2) progression syndromes
geriatricians ⫹ by expert 3) no neuroleptics
GPs 4) positive response to
Drug prescriptions levodopa
Nursing homes

Movement Disorders, Vol. 18, No. 1, 2003

23
 24

TABLE 1. (Continued)
% Possible
Study cases
(population Incidence Sources to identify Methods to identify examined by Definition of incident Inclusion/diagnostic
size) period possible cases included cases specialist* cases criteria Exclusion criteria Prospective
Northampton, 1986–90 Medical records Examination by 89 Symptom onset Based on UK PDS Brain Essential tremor, drug- No
UK27 -Hospital neurologist of all Bank criteria for PD induced
(298,985) Inquiries to GPs ⫹ patients consented parkinsonism,

Movement Disorders, Vol. 18, No. 1, 2003

consultants Case notes review secondary
of those refusing parkinsonism
Turku, 1987–91 Medical records Case notes review 39 Not defined PD: UK PDS Brain UK PDS Brain Bank No
Finland28 -Neurologist of all possible Bank (step 1) (step 2), essential
(196,864) -Hospital cases tremor, drug-induced
Finnish PD Examination by parkinsonism, other
association neurologist of parkinsonism
Residential homes doubtful cases
Inquiries to health
centres
Social insurance
records
Manhattan, 1989–91 Medical records Examination by 100 Symptom onset PD: UK PDS Brain UK PDS Brain Bank Yes (with
USA29 -HMO neurologist to Bank (step 1 and 3) (step 2), essential 85%
(213,000) -Hospital confirm diagnosis tremor, drug-induced follow-
-Private GP parkinsonism, up at 1
-Neurologist postencephalitic, year)
Social services secondary
D. TWELVES ET AL.

Disability/pension parkinsonism, PSP,

records MSA, dementia prior
Nursing homes to motor
Health insurance manifestations
records
Yonago City, 1989–92 Medical records Not specified 0 Not defined PD: ⱖ2 of 4 cardinal Toxin/drug-induced No
Japan30 -University hospital signs ⫹ improvement parkinsonism,
(131,704) -Questionnaire after levodopa vascular parkinsonism
-Other hospitals
-Medical
practitioners
Disability records
Death certificates
London, UK31 1993 Recording of GPs Review of Not stated Not defined Not specified Not specified Yes
(26,636) consultations for information sent
neurological by GP, neurology
problems referral
GP notes screening encouraged for
in a proportion difficult cases
 TABLE 1. (Continued)
% Possible
Study cases
(population Incidence Sources to identify Methods to identify examined by Definition of incident Inclusion/diagnostic
size) period possible cases included cases specialist* cases criteria Exclusion criteria Prospective
Ilan County, 1993–97 Door-to-door Case notes review 0 Diagnosis PD: ⱖ2 of 4 cardinal Secondary parkinsonism Yes
Taiwan32 questionnaire ⫹/⫺ signs (vascular, drug-
(75,579) examination to induced, MSA, PSP,
identify people other, eg, tumour or
without trauma)
Parkinonism at
baseline.
New cases then
identifed from
Bureau of National
Health Insurance
records
Navarra, 1994–95 Medical records Case notes review Not stated Not defined PD: UK PDS Brain Drug-induced Unclear
Spain33 -Neurologists Bank, modified by parkinsonism, other
(523,563) -GPs Hughes secondary
Questionnaire to parkinonism
neurologists
Telephone inquiries
to residential care
doctors
London, UK34 1995–96 Medical records GP case notes 100 Diagnosis ⱖ2 of 4 cardinal signs Drug-induced Yes
(100,230) -All GP notes review parkinsonism,
screened Examination by essential tremor,
-Hospital neurologist vascular parkinsonism
Drug prescriptions
GP referrals (any
REVIEW: INCIDENCE OF PARKINSON’S DISEASE

neurological cases)
to linkage clinic
Inquiries to GPs
GP databases

*Patients examined specifically for inclusion in the study.

PD, Parkinson’s disease; MSA, multiple systems atrophy; PSP, progressive supranuclear palsy; GP, general practitioner; UK PDS, UK Parkinson’s Disease Society; HMO, Health maintenance
organisation.

Movement Disorders, Vol. 18, No. 1, 2003

25
 26

TABLE 2. Results of incidence studies of Parkinson’s disease

Crude incidence Age
Type and Age strata (yr) incidence/100,000/year rate per 100,000 Mean time standardised
number of per year Mean age at from onset to M:F ratio
Study (population size) incident cases 30–39 40–49 50–59 60–69 70–79 80⫹ (95% CI)* onset (yr) diagnosis (yr) (95% CI)

Rochester, USA11 (33,400) By sx onset P Onset 0.1 6 17 73 76 34 12 65

125
By diagnosis P Diagnosis 1 5 18 65 174 167 18
191
Iceland12 (171,500) PD & VP 272 1 6 33 97 136 69 16 61
Carlisle, UK13 (69,400) PD/Pa 60 0 7 20 43 65 37 12

Movement Disorders, Vol. 18, No. 1, 2003

Victoria, Australia14 P 35 0.3 6 16 25 59 59 7 58
(83,001)
Sardinia9 (1,448,319) PD & PE 796 Not stratified 5 56
Hawaii15 (8,006) PD 92 0 0 23 37 93 107 11 (US 1970 pop)
Rochester, USA16 (53,885) P 138 0 5 32 113 254 155 20 72 1.6 (1.3–1.9)
Ferrara, Italy17 (187,381) PD 394 1 6 18 36 22 22 10 (9–11) 63 (SD 9) (M 62, 1.0 (0.9–1.2)
F 63)
Southwest Finland18 PD & PE 179 0.1 3 20 62 93 49 15 62 (SD 0.5) (M 0.9 (0.7–1.1)
(402,988) 59, F 63)
Aarhus, Denmark19 PD & PE, Not stratified 9 62 (M 60, F 63)
(242,151) number not
stated
Yonago city, Japan20 PD 62 Not stratified 10 64 (M 65, F 64)
(121,812)
Olmstead County, USA21 PD 154 1 1 17 52 93 79 11 3.6 2.0 (1.6–2.6)
(95,000)
Benghazi, Libya22 PD 58 Not stratified 4.5 64 (M 65, F 63)
(518,745)
Netherlands23 (163,577) PD 51 Not stratified M 11, F 12
Poznan, Poland24 PD 163 1 1 26 63 75 34 13
D. TWELVES ET AL.

(1,308,000)
China25 (3,869,162) PD 58 0 1 3 3 19 16 1.5 0.9 (0.6–1.4)
Southeast Sweden26 PD 49 2 3 9 22 59 79 11 66 (M 65, F 66) 2.8
(147,777)
Northampton, UK27 PD 175 Not stratified 12
(298,985)
Turku, Finland28 (196,864) PD, number not M 0 10 22 62 140 90 17 65 1.9 (1.4–2.6)
stated F 0 5 10 40 78 20
Manhattan, USA29 PD 83 0 11 11 54 133 213 13 (10–16) 69 (M 67, F 70) 1.4 (SD 0.9) 1.6 (1.3–2.1)
(213,000)
Yonago City, Japan30 PD 79 Incomplete 15 70 (M 70, F 71)
(131,704) stratified
data
London, UK31 (26,636) PD 7 Not stratified 26
Ilan County, Taiwan32 PD 15 0 0 18 47 100 0 10 1.1 (0.5–2.7)
(75,579)
Navarra, Spain33 (523,563) PD 86 0 0.3 7 30 42 34 8 69 1.9 (1.4–2.8)
London, UK34 (100,230) PD, number not 0 10 0 43 161 116 19 (UK 1991
stated pop)

*If only a standardised incidence was given, the population used to standardised is given in parentheses.
Sx, symptom; PD, Parkinson’s disease; P, parkinsonism; VP, vascular parkinsonism; PE, postencephalitic parkinsonism; M, male; F, female; CI, confidence interval; SD, standard deviation.
a
“PD” and “P” used interchangeably.
 REVIEW: INCIDENCE OF PARKINSON’S DISEASE 27

the other studies used recent census figures. There were diagnosis. In most other studies, it was clear that some
three exceptions, one of which used a predefined cohort patients had been examined as part of the study, usually
of men from the Honolulu Heart Study,15 and the other those patients who had not previously been seen by a
two of which used populations based on specific general specialist or in whom the case notes did not provide
practices.23,34 sufficient information to make a diagnosis. The percent-
age of patients seen by the study specialist varied from
Sources Used to Identify Possible Cases. 32% to 100%, although it was frequently not stated.
Patients with PD may be managed by hospital specialists
but many, particularly the elderly, may be managed in the Definition of Incident Case.
community by their primary care physician or remain un- To define an incident case, the date of onset of the
diagnosed. Therefore, to identify as many incident cases as disease must be known. Parkinson’s disease has an in-
possible, a variety of different search strategies should be sidious onset, often with rather nonspecific symptoms
used that cover both patients managed in the community initially; therefore, the definition of disease onset can be
and those referred to specialists. Most studies did this. difficult. The optimal definition is probably the date of
However, their exact searches varied and frequently relied specific symptom onset, provided the patient is seen
on retrospective reviews of primary care or hospital case relatively soon after onset so they can remember when
notes, which may not contain the relevant information to the symptoms started. The alternative method is to
make the appropriate diagnosis. There were differences in choose the date of diagnosis, but this method is depen-
how active the search methods were. For example, some dent on access to medical care, which will vary between
studies simply sent requests to doctors, asking them to and possibly within studies. The choice of definition is
report any of their patients with PD.17,18,26,27 In other important because different definitions of disease onset
studies, a more active approach was taken, with regular do appear to alter the final incidence figures (using the
screening of GP records.31,34 Some studies relied heavily date of diagnosis inflates the incidence figure).11 In 10
on hospital data,11,16,21 and these searches were likely to studies, the definition of incident cases was unclear. Of
miss early cases and those not referred for specialist the studies that did give a definition, eight used date of
opinion. Other studies did not search hospital or specialist symptom onset13,17,19,21,24,26,27,29 and six used date of
records.23,25,31,32 One study conducted a door-to-door sur- diagnosis.14,15,23,25,32,34 One study gave figures for both.11
vey of all people living in a defined area, asking about
symptoms of Parkinson’s disease.25 This method should Diagnostic Criteria for Inclusion and Exclusion.
have identified more early cases, and those who had not There are a variety of causes of parkinsonism other
presented to a doctor. It was uncertain, however, how than idiopathic Lewy body Parkinson’s disease. Differ-
this survey was accomplished, as it appears from the entiating between these different syndromes can be dif-
report of this study that nearly four million people were ficult without histologic confirmation, but there are clin-
interviewed, which seems unlikely. ical criteria that can help. Clearly, the incidence found in
each study could vary, depending on how broad or nar-
Methods for Identifying Included Cases. row the inclusion criteria were. Unfortunately, the stud-
It is often difficult to make a clinical diagnosis of PD, ies used various definitions of parkinsonism or Parkin-
and for the sake of accuracy in an epidemiological study, son’s disease. In six studies, the diagnostic criteria were
the diagnosis should ideally be confirmed in all possible not clear.11–14,23,31 Of the others, the commonest defini-
cases by an examination by a study neurologist or move- tion of PD was the presence of two or more of the
ment disorder expert. In practice of course, this confir- cardinal motor signs (rest tremor, bradykinesia, rigidity,
mation is usually not possible as some of the incident impaired postural and righting reflexes) with no atypical
cases may have died, moved away, or refused examina- features,9,15,17,18,20 –22,32,34 whereas four used the UK
tion. However, we tried to extract from each study how Brain Bank criteria (see Appendix).27–29,33 Exclusion cri-
many possible cases had been assessed in this way. In six teria differed between the studies. For example, some
studies,11,16,21,24,30,32 none of the patients were examined included other parkinsonian syndromes such as posten-
as part of the study. In one of these, it was not clear how cephalitic parkinsonism, whereas others excluded them.
included cases were identified,30 and the others relied on Most studies tried to exclude other common causes of
case notes review. Some patients in these studies had parkinsonism, for example drug-induced or vascular par-
been seen by a specialist as part of their routine clinical kinsonism, but there was substantial variation, and in
care, but this may not be as accurate as being seen several studies, the exclusion criteria were not explicit.
specifically as part of the incidence study to confirm the To overcome some of these difficulties, the results in

Movement Disorders, Vol. 18, No. 1, 2003

28 D. TWELVES ET AL.

Table 2 show whether the figures refer to PD or to a Study Results

wider group of parkinsonian conditions. Where studies In most studies (12), the peak incidence was between
give a figure for PD and for other groups, the results have 70 and 79 years of age; although in three studies, the
been given for PD alone. incidence continued to increase even in those 80 years of
The time span covered by our report presented a age or older (Table 2). The mean age of symptom onset
further problem in that diagnostic criteria have changed was 60 to 65 years in eight studies and over 65 years in
over the years. For example, although some of the later five studies. The onset in men was often slightly earlier
studies clearly tried to exclude causes of parkinsonism than in women (Table 2). Five of nine studies in which
such as progressive supranuclear palsy and multiple sys- age-standardised sex ratios were available found a sig-
tem atrophy, these diagnoses were unknown or under- nificantly greater incidence in men (ratio, 1.5 to 2.0).
recognised when some of the earlier studies were done; To determine whether there was any evidence of real
therefore, such patients would probably have been in- differences in incidence over time and between different
cluded as PD. Similarly, some of the later studies used places, the results of studies that used similar methods
diagnostic criteria, such as the UK Brain Bank criteria, were compared. Three similar studies were identified that
which had not been developed at the time of many of the diagnosed parkinsonism based on at least two cardinal
earlier studies. motor signs,17,18,34 and two similar studies were identi-
fied that used the UK PDS Brain Bank clinical diagnostic
Prospective Follow-up and Histology. criteria.28,29 None of the studies used identical methods
Prospective studies where patients are identified for- (the precise exclusion criteria and search strategies dif-
ward in time are preferable to retrospective studies where fered slightly, some defined incidence by symptom on-
much of the information has to be extracted from clinical set, others by diagnosis, and some were prospective,
notes that were not designed for epidemiological re- others retrospective), but it was believed they were suf-
search. However, only eight studies were prospec- ficiently similar to merit comparison (Table 3). Age-
tive,15,19,23,25,29,31,32,34 and of these, only two followed up specific incidence rates are shown in the original age
the patients for any length of time.19,29 Follow-up is strata, which varied between studies. All these studies
important because symptoms or signs may develop that gave similar standardised incidences of approximately
change the underlying diagnosis (and also because it 16 to 19 per 100,000 per year except one in Italy, which
allows the natural history of the disease to be studied in gave a much lower incidence (approximately half).17 The
a representative group of patients). Only one study only obvious significant difference between the Italian
planned long-term follow-up, but the report of this study study and the others was the duration of case ascertain-
did not describe the natural history.19 ment (20 years compared to 2 to 4 years for the other
The clinical diagnosis of Parkinson’s disease is often studies). Although the Italian study was retrospective, so
difficult, and the criterion standard remains histological were some of the other studies that gave a much higher
confirmation on postmortem material. Brain bank data incidence.18,28 Similarly, although it used date of symp-
have shown that even experts get the diagnosis wrong up tom onset to define an incident case, so did one of the
to 25% of the time.35 This finding emphasises the need other studies.29
for long-term follow-up and the need for postmortems in DISCUSSION
a high percentage of cases. Unfortunately, only two We identified as many eligible incidence studies as
studies reported any data on histology.11,16 In one early possible using several search strategies. The sensitivity
study (1935–1966) only 27 of 51 cases were confirmed of the electronic search was checked by comparing rel-
pathologically to be idiopathic PD.11 evant references found in the bibliographies of obtained
articles against those in the database. All relevant articles
Other Methodological Issues. except one12 had been retrieved, provided they were
One study had a further significant methodological published within the time period of the databases. We
flaw.9 In this combined incidence and prevalence study, identified many more prevalence studies of PD, but these
it seems that any patient who had died or moved away are less useful than incidence studies for epidemiological
from the area by prevalence day was excluded from the research because patients are included at different stages
incidence study. However, if these patients had an onset of the disease process.
of disease during the incidence period, they should have The best incidence studies gave a rate of approxi-
been included. This strategy may partly explain the low mately 17 per 100,000 per year but this rate may be an
incidence rate found in Sardinia in this study. underestimate because a significant proportion of pa-

Movement Disorders, Vol. 18, No. 1, 2003

 REVIEW: INCIDENCE OF PARKINSON’S DISEASE 29

TABLE 3. Standardised incidence of Parkinson’s disease in studies using similar methodologies

Standardised incidence
No. incident to Scottish population,
Study cases Incidence/100,000/year in original age strata (actual no. of cases if stated) 1990 (95% CI)

ⱖ2 Cardinal signs
Ferrara, Italy17 394 35–39 40–44 45–49 50–54 55–59 60–64 65–69 70–74 75⫹ 8 (7–9)
1967–87 1 4 7 14 21 44 29 32 13
(3) (12) (20) (40) (64) (96) (68) (61) (30)
Southwest 179 0–39 40–49 50–59 60–69 70–79 80⫹ 17 (14–19)
Finland18 0.1 3 20 62 93 49
1968–70 (1) (4) (27) (78) (60) (9)
London, UK34 N.S. 45–49 50–59 60–64 65–69 70–74 75–79 80–84 85⫹ 18a
1995–96 20 0 50 37 222 100 0 116
UK PDS Brain Bank
criteria
Turku, Finland28 N.S. 30–39 40–49 50–59 60–69 70–79 80⫹ 16a
1987–91 0 10 22 62 140 90
M 0 5 10 40 78 20
F
Manhattan, USA29 83 45–64 65–74 75–84 85⫹ 16.5 (13–20)
1989–91 11 54 133 213
(12) (20) (33) (20)
a
Unable to calculate confidence intervals due to insufficient data.
N.S., number not stated.

tients remain undiagnosed without population screen- that few good quality studies exist. Further incidence
ing.10,36 The lack of comparable incidence studies means studies, therefore, are required, and on the basis of this
that few inferences can be made about time trends or review, we have produced a provisional set of minimal
geographical differences in the incidence of PD from this scientific criteria, which we think would improve the
review. However, there are some data to suggest that, at quality and consistency of such studies.
least in parts of the USA and Japan, the incidence has not 1. The base population should be neither too small (too
changed significantly over the past few decades.20,30,37 few cases), nor too large (difficulty achieving adequate
Although we have highlighted a low incidence in part of detection). We would suggest 250,000 to 500,000 people
Italy, we believe this should be interpreted cautiously. as a suitable size. For example, to detect a difference in
The Italian study identified patients over a very long incidence rates of 5 per 100,000 between two different
period of time and, thus, relied heavily on retrospective populations (␣ ⫽ 0.05, 80% power), a million patient-
identification of patients from case records, which may years of surveillance are required assuming an average
have reduced ascertainment. In addition, the Europarkin- incidence of 17 per 100,000 per year. This could be
son’s Group36 showed there were over twice as many achieved by screening populations of 250,000 for 4 years
patients in Italy who had undiagnosed PD compared with or 500,000 for 2 years.
Northern European countries (30% vs. approximately 2. The studies should be prospective to maximise case
10%) which would also reduce incidence. ascertainment and data accuracy.
The peak incidence is generally between 70 and 79 3. Multiple sources should be used to identify possible
years of age, but this may be because of the difficulty in cases. Primary care physicians should be encouraged,
identifying very elderly patients. A recent small Italian and reminded regularly, to refer any patients they suspect
study that carefully screened a random sample of 6,000 of having parkinsonism to the study. Regular searches
people 65 to 84 years of age found a much higher should also be made of routine medical data for both
incidence in these age groups than previous studies (220 inpatients and outpatients with possible parkinsonism
to 670 per 100,000 per year) and showed the incidence (including data from hospitals, rehabilitation centres,
doubled between 75 and 79 years and 80 and 84 years.10 public/private neurologists, geriatricians, primary care
We found conflicting evidence of an increased incidence clinicians, death certificates). Records of disability, pen-
in men. Several studies found a 1.5- to 2-fold increase, sions, health insurance data, and drug prescriptions could
but the largest study found no difference between men be searched if available. To minimise the number of very
and women.17 elderly patients who might be missed, contact should be
This review has highlighted the difficulties in perform- established with local residential or nursing homes. Pa-
ing good quality incidence studies in PD and has shown tient support groups could also be contacted.

Movement Disorders, Vol. 18, No. 1, 2003

30 D. TWELVES ET AL.

These methods will miss patients who do not present induced parkinsonism, vascular parkinsonism, and par-
to the medical profession. The only way to overcome this kinsonism with dementia), this variation would be un-
problem is to screen the population for parkinsonian likely to have a major effect on the incidence. It is
symptoms with a validated questionnaire at the begin- important to avoid dependence on high technology as-
ning and end of the incidence period and confirm any sessments such as PET scanning, which are expensive
suspected patients identified with a medical examination. and not widely available and, hence, not practical to
This approach has been taken in several prevalence stud- perform in large epidemiological studies.
ies36 and in one incidence study restricted to a small 7. Ideally, studies should have some follow-up to give
elderly population10 but would require huge resources if information on response to dopaminergic therapy, disease
applied to large populations. A compromise would be to progression (both part of step 3 of the UK Brain Bank
screen a random sample of the population at the end of diagnostic criteria), and the development of other features
the incidence period to give some idea of how many suggestive of a parkinsonian plus syndrome. However, pro-
patients had been missed by the other search strategies. longed follow-up will not be practical in most studies, and
4. As many of the possible cases as possible should be given that multiple system atrophy and progressive su-
seen by an expert in movement disorders to confirm the pranuclear palsy make up less than 10% of parkinsonian
diagnosis. However, if some patients cannot be seen, as syndromes,39 including some patients with these syndromes
much information about their signs and symptoms should not significantly alter the incidence rate.
should be gathered from their medical records as possi- 8. Incidence rates should be reported by standard age
ble, and if these data are strongly suggestive of PD, the strata to enable comparisons between studies (0 –39,
patient should be included. 40 – 49, 50 –59, 60 – 69, 70 –79, 80 – 89, 90⫹), and confi-
5. Ideally, an incident case should be defined by spe- dence intervals should be given. We would welcome
cific symptom (tremor, bradykinesia, rigidity) onset further discussion on these provisional criteria so that
rather than date of diagnosis, because the latter is less more definitive standards can be established for future
generaliseable across studies because it depends on ac- incidence studies.
cess to health care. However, this approach may not be
Acknowledgments: We thank our translators, Professor Jan
feasible in prospective studies, because there is often a Deregowski, Department of Psychology, University of Aber-
significant delay between symptom onset and diagnosis deen, and Miriam Brazelli, Health Services Research Unit,
(mean, 1 to 4 years; see Table 2). Hence, in the first year University of Aberdeen. K.P. was supported by the Neurology
of a prospective study, few patients would be eligible Research Endowment Fund, Grampian University Hospitals
based on symptom onset and case ascertainment would Trust.
need to be extended for at least 4 years beyond the
incident period to ensure that patients whose symptoms APPENDIX
came on during the incident period were identified. UK Parkinson’s Disease Society Brain Bank
Symptom onset could be used in studies where popula- Clinical Diagnostic Criteria
tions were screened at the beginning and end of the
incidence period, but this strategy will not be feasible in Step 1: Diagnosis of Parkinsonian Syndrome.
large studies. The date of diagnosis, therefore, may be Bradykinesia and at least one of the following: muscular
the most practical definition in most prospective studies. rigidity; 4 – 6 Hz rest tremor; postural instability not caused by
6. Clear and consistent inclusion and exclusion criteria primary visual, vestibular, cerebellar, or proprioceptive dys-
should be applied. To minimise the number of cases function.
missed, we suggest broad criteria for selecting the pri-
mary cohort of patients with possible PD,38 i.e., at least Step 2: Exclusion Criteria for Parkinson’s Disease.
two cardinal motor signs or an isolated rest tremor (some History of repeated strokes with stepwise progression of
patients with PD may only have a rest tremor). Further parkinsonian features, history of repeated head injury; history
of definite encephalitis; oculogyric crises; neuroleptic treatment
clinical criteria such as the UK Brain Bank criteria at onset of symptoms; more than one affected relative; sus-
should then be used to identify those with probable or tained remission; strictly unilateral features after 3 years; su-
possible idiopathic PD and those with probable or pos- pranuclear gaze palsy; cerebellar signs; early severe autonomic
sible other forms of parkinsonism. The results should involvement; early severe dementia with disturbances of mem-
clearly state how many patients were finally included in ory, language, and praxis; Babinski sign; presence of cerebral
tumour or communicating hydrocephalus on computed tomog-
each diagnostic category. The exact definition of PD may raphy scan; negative response to large doses of levodopa (if
vary slightly but, provided the most common alternative malabsorption excluded); 1-methyl-4-phenyl-1,2,3,6-tetrahy-
diagnoses were excluded (e.g., essential tremor, drug- dropyridine hydrochloride (MPTP) exposure.

Movement Disorders, Vol. 18, No. 1, 2003

 REVIEW: INCIDENCE OF PARKINSON’S DISEASE 31

Step 3: Supportive Prospective Positive Criteria for 21. Bower JH, Maraganore DM, McDonnell SK, Rocca WA. Inci-
dence and distribution of parkinsonism in Olmsted County, Min-
Parkinson’s Disease (Three or More Required for nesota, 1976 –1990. Neurology 1999;52:1214 –1220.
Diagnosis of Definite Parkinson’s Disease). 22. Ashok PP, Radhakrishnan K, Sridharan R, Mousa ME. Epidemi-
Unilateral onset, rest tremor present, progressive disorder, ology of Parkinson’s disease in Benghazi, North-East Libya. Clin
Neurol Neurosurg 1986;88:109 –113.
persistent asymmetry affecting side of onset most, excellent
23. Hofman A, Collette HJ, Bartelds AI. Incidence and risk factors of
response (70 –100%) to levodopa, severe levodopa-induced Parkinson’s disease in The Netherlands. Neuroepidemiology 1989;
chorea, levodopa response for 5 years or more, clinical course 8:296 –299.
of 10 years or more. 24. Wender M, Pruchnik D, Kowal P, Florczak J, Zalejski M. [Epide-
miology of Parkinson’s disease in the Poznan province] (Polish).
REFERENCES Przegl Epidemiol 1989;43:150 –155.
25. Wang YS, Shi YM, Wu ZY, He YX, Zhang BZ. Parkinson’s
1. Marttilla RJ, Rinne UK. Epidemiology of Parkinson’s disease: an disease in China. Coordinational Group of Neuroepidemiology,
overview. J Neural Transm 1981;51:135–148. PLA. Chin Med J 1991;104:960 –964.
2. Pedro-Cuesta J. Parkinson’s disease occurrence in Europe. Acta 26. Fall P-A, Axelson O, Fredriksson M, Hansson G, Lindvall B,
Neurol Scand 1991;84:357–365. Olsson J-E, et al. Age-standardized incidence and prevalence of
3. Pedro-Cuesta J, Stawiarz L. Parkinson’s disease incidence: mag- Parkinson’s disease in a Swedish community. J Clin Epidemiol
nitude, comparability, time trends. Acta Neurol Scand 1991;84: 1996;49:637– 641.
382–388. 27. Sutcliffe RL, Meara JR. Parkinson’s disease epidemiology in the
4. Zhang Z-X, Roman GC. Worldwide occurrence of Parkinson’s Northampton District, England, 1992. Acta Neurol Scand 1995;
disease: an updated review. Neuroepidemiology 1993;12:195–208. 92:443– 450.
5. Tanner CM, Goldman SM. Epidemiology of Parkinson’s disease. 28. Kuopio AM, Marttila RJ, Helenius H, Rinne UK. Changing epi-
Neurol Clin 1996;14:317–335. demiology of Parkinson’s disease in southwestern Finland. Neu-
6. Rajput AH, Birdi S. Epidemiology of Parkinson’s disease. Parkin- rology 1999;52:302–308.
sonism Relat Disord 1997;3:175–186.
29. Mayeux R, Marder K , Cote LJ, Denaro J, Hemenegildo N, Mejia
7. Rosati G, Granieri E, Pinna L, Aiello I, De Bastiani P, Tola R, et
H, et al. The frequency of idiopathic Parkinson’s disease by age,
al. Parkinson’s disease. Prevalence and incidence in the Province
ethnic group, and sex in Northern Manhattan, 1988 –1993. Am J
of Sassari, North Sardinia. Acta Neurol 1978;33:201–207.
Epidemiol 1995;142:820 – 827.
8. Rosati G, Granieri E, Pinna L, Devoto MC. The frequency of
30. Kusumi M, Nakashima K, Harada H, Nakayama H, Takahashi K.
Parkinson’s disease in the Province of Nuoro (Sardinia). Acta
Epidemiology of Parkinson’s disease in Yonago City, Japan: com-
Neurol 1979;1:303–308.
parison with a study carried out 12 years ago. Neuroepidemiology
9. Rosati G, Granieri E, Pinna L, Aiello I, Tola R, De Bastiani P, et
1996;15:201–207.
al. The risk of Parkinson’s disease in Mediterranean people. Neu-
rology 1980;30:250 –255. 31. Cockerell OC, Goodridge DM, Brodie D, Sander JW, Shorvon SD.
10. Baldereschi M, Di Carlo A, Rocca WA, Vanni P, Maggi S, Neurological disease in a defined population: the results of a pilot
Perissinotto E, Grigoletto F, Amaducci L, Inzitari D. Parkinson’s study in two general practices. Neuroepidemiology 1996;15:73– 82.
disease and Parkinsonism in a longitudinal study: two-fold higher 32. Chen RC, Chang SF, Su CL, Chen THH, Yen MF, Wu HM, Chen
incidence in men. ILSA Working Group. Italian Longitudinal ZY, Liou HH. Prevalence, incidence, and mortality of PD: a
Study on Aging. Neurology 2000;55:1358 –1363. door-to-door survey in Ilan County, Taiwan. Neurology 2001;57:
11. Nobrega F, Glattre E, Kurland LT, Okazaki H. Comments on the 1679 –1686.
epidemiology of Parkinsonism including prevalence and incidence 33. Vines JJ, Larumbe R, Gaminde I, Artazcoz MT. Incidencia de la
statistics for Rochester, Minnesota, 1935–1966. In: Brunette JR, enfermedad de Parkinson idiopatica y secundaria en Navarra.
Barbeau JR, editors. Progress in Neurogenetics. Amsterdam: Ex- Registro poblacional de casos. [Incidence of idiopathic and sec-
cerpta Medica; 1969. p 474 – 485. ondary Parkinson disease in Navarre. Population-based case reg-
12. Gudmundsson KR. A clinical survey of parkinsonism in Iceland. istry.] (Spanish). Neurologia 1999;14:16 –22.
Acta Neurol Scand 1967;43(Suppl. 33):9 – 61. 34. MacDonald BK, Cockerell OC, Sander JW, Shorvon SD. The
13. Brewis M, Proskanzer DC, Rolland C, Miller H. Neurological dis- incidence and lifetime prevalence of neurological disorders in a
eases in an English city. Acta Neurol Scand 1966;42(Suppl. 24):1– 89. prospective community-based study in the UK. Brain 2000;123:
14. Jenkins AC. Epidemiology of parkinsonism in Victoria. Med J 665– 676.
Aust 1966;2:496 –502. 35. Hughes AJ, Daniel SE, Kilford L, Lees AJ. Accuracy of clinical
15. Morens DM, Davis JW, Grandinetti A, Ross GW, Popper JS, diagnosis of idiopathic Parkinson’s disease: clinico-pathological
White LR. Epidemiologic observations on Parkinson’s disease: study of 100 cases. J Neurol Neurosurg Psychiatry 1992;55:181–
incidence and mortality in a prospective study of middle-aged men. 184.
Neurology 1996;46:1044 –1050. 36. de Rijk MC, Tzourio C, Breteler MMB, Dartigues JF, Amaducci L,
16. Rajput AH, Offord KP, Beard CM, Kurland LT. Epidemiology of Lopez-Pousa S, Manubens-Bertran JM, Alpérovitch A, Rocca WA,
parkinsonism: incidence, classification, and mortality. Ann Neurol for the EUROPARKINSON Study Group. Prevalence of parkin-
1984;16:278 –282. sonism and Parkinson’s disease in Europe: the EUROPARKIN-
17. Granieri E, Carreras M, Casetta I, Govoni V, Tola MR, Paolino E, SON collaborative study. J Neurol Neurosurg Psychiatry 1997;62:
et al. Parkinson’s disease in Ferrara, Italy, 1967 through 1987. 10 –15.
Arch Neurol 1991;48:854 – 857. 37. Rocca WA, Bower JH, McDonnell SK, Peterson BJ, Maraganore
18. Marttila RJ, Rinne UK. Epidemiology of Parkinson’s disease in DM. Time trends in the incidence of Parkinsonism in Olmsted
Finland. Acta Neurol Scand 1976;53:81–102. County, Minnesota. Neurology 2001;57:462– 467.
19. Dupont E. Epidemiology of parkinsonism. The Parkinson investi- 38. de Rijk MC, Rocca WA, Anderson DW, Melcon MO, Breteler
gation, Aarhus, Denmark (preliminary results). In: Worm-Petersen MMB, Maraganore DM. A population perspective on diagnostic
J, Bottcher J, editors. Symposium on parkinsonism. Coppenhagen: criteria for Parkinson’s disease. Neurology 1997;48:1277–1281.
Merck Sharp & Dohme; 1977. p 65–75. 39. Schrag A, Ben-Shlomo Y, Quinn NP. Prevalence of progressive
20. Harada H, Nishikawa S, Takahashi K. Epidemiology of Parkin- supranuclear palsy and multiple system atrophy: a cross-sectional
son’s disease in a Japanese city. Arch Neurol 1983;40:151–154. study. Lancet 1999;354:1771–1775.

Movement Disorders, Vol. 18, No. 1, 2003