UC San Diego
UC San Diego Previously Published Works
Title
MDS research criteria for prodromal Parkinson's disease.
Permalink
https://escholarship.org/uc/item/2v885924
Journal
Movement disorders : official journal of the Movement Disorder Society, 30(12)
ISSN
0885-3185
Authors
Berg, Daniela
Postuma, Ronald B
Adler, Charles H
et al.
Publication Date
2015-10-01
DOI
10.1002/mds.26431
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California
MDS Research Criteria for Prodromal Parkinson’s Disease
Daniela Berg, MD,
1
* Ronald B. Postuma, MD, MSc,
2
* Charles H. Adler, MD, PhD,
3
Bastiaan R. Bloem, MD, PhD,
4
Piu Chan, MD, PhD,
5
Bruno Dubois, MD, PhD,
6
Thomas Gasser, MD,
1
Christopher G. Goetz, MD,
7
Glenda Halliday, PhD,
8
Lawrence Joseph, PhD,
9
Anthony E. Lang, OC, MD, FRCPC,
10
Inga Liepelt-Scarfone, PhD,
1
Irene Litvan, MD,
11
Kenneth Marek, MD,
12
Jos
e Obeso, MD, PhD,
13
Wolfgang Oertel, MD,
14
C. Warren Olanow, MD, FRCPC,
15
Werner Poewe, MD,
16
Matthew Stern, MD,
17
and G
€
unther Deuschl, MD
18
1
Department of Neurodegeneration, Hertie-Institute for Clinical Brain Research and German Center for Neurodegenerative Diseases, Tuebingen,
Germany
2
Department of Neurology, Montreal General Hospital, Montreal, Quebec, Canada
3
The Parkinson’s Disease and Movement Disorders Center, Department of Neurology, Mayo Clinic, Scottsdale, Arizona, USA
4
Department of Neurology, Radboud University Medical Center, Donders Institute for Brain, Cognition and Behavior, Nijmegen, The Netherlands
5
Xuanwu Hospital of Capitol of Medical University, Beijing, China
6
Hopital De La Salpetriere, Paris, France
7
Rush University Medical Center, Chicago, Illinois, USA
8
Neuroscience Research Australia & University of NSW, Randwick, Australia
9
Department of Epidemiology and Biostatistics, McGill University, Montreal, Quebec, Canada
10
Division of Neurology, Toronto Western Hospital, Toronto, Ontario, Canada
11
Department of Neurosciences, University of California San Diego, La Jolla, California, USA
12
Institute for Neurodegenerative Disorders, New Haven, Connecticut, USA
13
University of Navarra-FIMA, Pamplona, Spain
14
Department of Neurology, Philipps University of Marburg, Marburg, Germany
15
Department of Neurology, The Mount Sinai Hospital, New York, New York, USA
16
Department of Neurology, Innsbruck Medical University, Innsbruck, Austria
17
Penn Neurological Institute, Philadelphia, Pennsylvania, USA
18
Department of Neurology, Christian-Albrechts University, Kiel, Germany
ABSTRACT: This article describes research criteria
and probability methodology for the diagnosis of pro-
dromal PD. Prodromal disease refers to the stage
wherein early symptoms or signs of PD neurodegenera-
tion are present, but classic clinical diagnosis based on
fully evolved motor parkinsonism is not yet possible.
Given the lack of clear neuroprotective/disease-modify-
ing therapy for prodromal PD, these criteria were devel-
oped for research purposes only. The criteria are based
upon the likelihood of prodromal disease being present
with probable prodromal PD defined as 80% certainty.
Certainty estimates rely upon calculation of an individu-
al’s risk of having prodromal PD, using a Bayesian na
€
ıve
classifier. In this methodology, a previous probability of
prodromal disease is delineated based upon age. Then,
the probability of prodromal PD is calculated by adding
diagnostic information, expressed as likelihood ratios.
This diagnostic information combines estimates of back-
ground risk (from environmental risk factors and genetic
findings) and results of diagnostic marker testing. In order
to be included, diagnostic markers had to have prospec-
tive evidence documenting ability to predict clinical PD.
They include motor and nonmotor clinical symptoms,
clinical signs, and ancillary diagnostic tests. These crite-
ria represent a first step in the formal delineation of early
stages of PD and will require constant updating as more
information becomes available.
V
C
2015 International Par-
kinson and Movement Disorder Society
Key Words: Parkinson’ s disease; diagnosis; prodromal
------------------------------------------------------------- --------------------------- --------------------------------------
*Correspondence to: Dr. Ronald B. Postuma, Department of Neurology,
L7-305 Montreal General Hospital, 1650 Cedar Avenue, Montreal,
Quebec, Canada H3G1A4; ron.postuma@mcgill.ca; or Dr. Daniela Berg,
Hertie Institute of Clinical Brain Research, Hoppe, Seyler-Straße 3, 72076
Tubingen, Germany; Daniela.berg@uni-tueb
Drs. Berg and Postuma contributed equally.
Relevant conflicts of interest/financial disclosures: Nothing to report.
Full financial disclosures and author roles may be found in the online ver-
sion of this article.
Received: 3 April 2015; Revised: 11 August 2015; Accepted: 12
August 2015
Published online in Wiley Online Library
(wileyonlinelibrary.com). DOI: 10.1002/mds.26431
REVIEW
CME
1600 Movement Disorders, Vol. 30, No. 12, 2015
Parkinson’s disease (PD), as with all neurodegenera-
tive diseases, does not start suddenly. This implies that
the disease progresses through early phases, during
which neurodegeneration has commenced but has not
yet progressed to the point at which PD can be defini-
tively diagnosed. The early phase of PD has both
motor and nonmotor aspects, and in many cases, cer-
tain nonmotor abnormalities are the first to emerge.
The International Parkinson and Movement Disorder
Society (MDS) Task Force on the definition of PD was
created to clarify challenges to our current definition of
PD. An introductory statement of this task force was
recently published,
1
in which we proposed that early
PD should be divided into three stages: preclinical PD
2
(neurodegenerative processes have commenced, but
there are no evident symptoms or signs); prodromal PD
(symptoms and signs are present, but are yet insuffi-
cient to define disease); and clinical PD (i.e., diagnosis
of PD based on presence of classical motor signs).
Diagnostic criteria for clinical PD have been devel-
oped (the MDS-PD criteria) and appear in this issue
of Movement Disorders.
3
This article presents the
MDS criteria for prodromal PD. These are intended
as a research tool; in the absence of a clear neuro-
protective/disease-modifying therapy against PD,
given the potential ethical issues of disclosure of dis-
ease risk in a nonmedical context, and especially
given the uncertainty inherent to this early-
development field, we do not yet recommend using
these prodromal criteria outside of a research setting.
Because new data are constantly being generated
from the fields of neurobiology, genetics, neuroimag-
ing, and other arenas, these criteria will require con-
tinuous reupdating.
Key Definition Features of
Prodromal PD
Several key definition aspects of prodromal PD
deserve particular emphasis
1
:
1. Whereas the diagnostic criteria for clinical PD
remain centered on a motor syndrome, prodro-
mal PD can be defined also based upon nonmo-
tor markers (“marker” here refers to any disease
indicator, whether a symptom, sign, or bio-
marker). The prodromal terminology makes no
assumptions about the order in which motor ver-
sus nonmotor markers develop.
2. The speed of progression from prodromal to full
clinical stages varies among patients and cannot
be reliably predicted on the individual level.
Therefore, the criteria center upon whether symp-
tomatic neurodegeneration is present, and not
when this will progress to full clinical PD. Note
that in many situations (e.g., clinical trials that
require conversion within a limited time win-
dow), further stratification may be required,
using markers that signal faster progression or
advanced prodromal stage.
3. The criteria incorporate estimates of risk, based
upon age, sex, and documented PD risk factors.
However, prodromal PD criteria cannot be met
with only risk markers; some markers of ongoing
neurodegeneration (i.e., prodromal markers) must
also be present.
4. The selection of criteria is primarily data driven.
Criteria required prospective studies documenting
their predictive value for clinical PD.
5. As outlined in our definition statement,
1
demen-
tia with Lewy bodies (DLB) is not considered an
exclusion criterion for PD. Therefore, if patients
with prodromal PD markers develop DLB, we do
not consider such cases as “false positives”; many
of these subjects will eventually also meet criteria
for PD. Some studies included in our analysis
(particularly rapid eye movement [REM] sleep
behavior disorder [RBD] studies) assessed DLB
and PD as a combined outcome. Because most
DLB patients in these studies also meet the MDS-
PD criteria,
4
we used the combined conversion
rates to calculate likelihood ratios. For studies of
prodromal DLB that did not clearly assess coex-
isting PD (e.g., studies of nonamnestic mild cog-
nitive impairment [MCI]
5
), we could not identify
how many DLB patients would meet MDS-PD
criteria, so we could not include these studies in
our calculations.
6. Given that there are currently no means to iden-
tify prodromal PD with 100% certainty, diagnos-
tic criteria for prodromal PD must be based upon
probability. We specified probable prodromal PD
as a high likelihood (i.e. 80%) that prodromal
PD is present. This category might be used to
select candidates for future disease modification
trials. However, in a specific research setting,
investigators may elect to select patients with dif-
ferent probability cutoffs. For example, for a
long, randomized trial of a very well-tolerated
agent, investigators may elect to include all those
who have a > 50% calculated probability of pro-
dromal PD.
7. Although the task force is aware that a primary
use of these criteria will be for enrolling patients
in neuroprotective trials, we do not wish to
define any details about trial procedures, means
of stratification, assessment of conversion rates
to classical PD or dementia, and so on. The cri-
teria simply attempt to define the probability
that an individual patient has prodromal PD; the
ultimate use of the criteria depends upon
context.
MDS CRITERIA FOR PRODROMAL PD
Movement Disorders, Vol. 30, No. 12, 2015 1601
Methodology of Criteria Design
The primary methodology for sequentially adding
diagnostic information is termed the na
€
ıve Bayesian clas-
sifier. This methodology takes a baseline pretest (i.e.,
“prior”) probability of disease, then adds the results of a
new diagnostic test to arrive at a new (post-test) proba-
bility of disease. Results of additional diagnostic tests
can then be sequentially added, to further refine disease
probability. This analysis requires two types of data:
1. The prior: This is the baseline probability of dis-
ease, given no diagnostic testing information. In
this case, it is the estimated age-adjusted preva-
lence of prodromal PD.
2. Likelihood ratios: Likelihood ratios (LRs)
describe the strength of a diagnostic test. One
can define positive LRs, which indicate how
much a positive test result increases disease prob-
ability (expressed as LR
1
>1), and negative LRs,
which indicate how much a negative test
decreases the probability (LR
2
<1). LR
1
can be
calculated as sensitivity / 1-specificity, and LR
2
as 1-sensitivity / specificity, or they can inputted
directly from 2:2 tables into Bayesian calculators
(e.g., http://araw.mede.uic.edu/cgi-bin/testcalc.pl).
Calculations Used to Generate
Criteria
A. Estimating the Prior Probability
(i.e., Prevalence of Prodromal PD)
The true prevalence of prodromal PD is unknown. In
estimating the prior probability, the Task Force consid-
ered four complementary sources of information:
1. The prevalence of PD. Point prevalence of PD is
approximately 0.25% at age 60, 0.5% at age 65,
0.5–1.0% at age 70, 1.0–1.5% at age 75, 2.0–
2.5% at age 80, and 3.0–4.0% at age 85.
6,7
If
one posits that prodromal PD has approximately
a 10-year average duration, this would estimate a
prodromal prevalence of 0.5% at age 55, 1.0–
1.5% at 65, and 3.5% at 75.
2. Incidence of PD. The incidence of PD is approxi-
mately 50 in 100,000 at age 65, 150 in 100,000
at age 75, and 400 in 100,000 at 85.
6
Assuming
a 10-year prodromal period, this would translate
to a prodromal prevalence of 0.5% at age 55,
1.5% at age 65, and 4% at age 75.
3. Cumulative lifetime risk of PD. According to cumula-
tive life risk tables, the chance of a 60-year-old devel-
oping PD by age 80 approximates 2.5%.
8
Therefore,
with an average 10-year prodromal period, the esti-
mated prevalence is 1.25% (i.e., half of 2.5%). With
a 20-year prodromal period, prevalence is 2.5%
4. Prospective studies: In prospective, population-based
studies that directly assessed PD as a primary out-
come, the incidence was higher than any other esti-
mation procedure. For example, in the Honolulu
Asia Aging study, 37 of 1,865 (2.0%) developed PD
over 8 years
9
(average age at diagnosis 5 83), and
the PRIPS study had 21 of 1,282 (1.6%) over 5 years
(average age 5 76).
10
This rate may more closely
reflect the disease risk in disease-modification trials,
which will also follow patients systematically.
Assuming equal risk distribution over a 10-year pro-
dromal period, these two studies would estimate pro-
dromal prevalence as approximately 2.5% at age 73
and 3.2% at age 66.
These estimates are clearly age dependent. They will
likely also increase with changes in definition of PD,
particularly the removal of dementia as an exclusion cri-
terion for PD.
1
Based on current knowledge, we com-
bined estimates from these different sources together,
adding subjective weighting and smoothing, to create an
age-specific estimate of prodromal PD prevalence.
We Estimate the Prior Probability According to
5-Year Age Intervals:
0.4% from ages 50 to 54;
0.75% from ages 55 to 59;
1.25% from ages 60 to 64;
2.0% from ages 65 to 69;
2.5% from ages 70 to 74;
3.5% from ages 75 to 79; and
4.0% age 80 and over.
B. Estimating LRs of Markers
The second critical piece of information is the LR of
each marker.
1. LRs of Risk Markers
Identifying risk and protective factors can help stratify
risk. Conceptually, risk markers help refine the prior
probability, rather than diagnosing the stage of prodro-
mal PD. For simplicity of calculations, these can be com-
bined with prodromal markers; however, risk factors
cannot diagnose disease alone. Prodromal markers with
combined minimum LR of at least 14 (i.e., approximat-
ing one strong or two combined moderate-strength
markers) must also be present, to indicate that the neu-
rodegenerative process has likely started. We included
only those risk factors that have broad consensus as true
risk factors for PD, and which were sufficiently powerful
to contribute meaningfully to probability estimation. For
inclusion, we required consistent evidence from at least
two prospective cohort studies or meta-analyses. Estima-
tion of the strength of each risk factor is anchored in
meta-analyses wherever possible.
BERG ET AL
1602 Movement Disorders, Vol. 30, No. 12, 2015
Calculation of each LR is based upon the relative risk
aswellastheprevalenceoftheriskfactor(LRsarecal-
culated independent of the disease prevalence). A prior
probability (i.e., estimated prodromal PD prevalence) of
2% is used for calculations. An example is provided for
caffeine, assuming relative risk (RR) 5 1.5 for nonuse
of caffeine, with 25% of the population having this risk
factor.
11,12
In a population of 10,000, 200 will have
prodromal PD. Furthermore, 2,500 will be nonusers of
caffeine. If RR 5 1.5, then 67 of 2,500 (2.68%) of non-
users would get PD compared to 133 of 7,500 (1.77%)
of caffeine users. A 2:2 table can be generated:
Test 1 Test 2
PD 1 67 133 200
PD 2 2,433 7,367 9,800
2,500 7,500 10,000
LR of 1 test (LR
1
) 5 1.35
LR of – test (LR
2
) 5 0.88
11
Using a prior (i.e., prodromal PD prevalence) of 1%
or 4% produces almost exactly the same LR
1
(1.36,
1.34). Note that when the prevalence of a risk factor is
less than 10%, the LR
2
is very close to 1 (i.e., absence of
the risk factor does not add much information). There-
fore, for risk factors with population prevalence <10%,
only LR
1
calculations are pertinent (LR
2
need not be
applied and are indicated as not applicable [N/A]).
2. LRs of Markers
LRs are estimated in three ways, depending upon
the type of evidence:
(a) Prospective studies with control groups: This
approach is the most reliable and is available for most
prodromal markers. These studies included control
groups, allowing generation of a 2:2 table, and a sim-
ple calculation of LR directly from the tables (LR
1
5
sensitivity / 1-specificity, LR
2
5 1-sensitivity / specific-
ity). An example is provided below:
Transcranial ultrasound: In a prospective study
10
of
1,175 individuals, 17 developed PD; 14 of these had
SN hyperechogenicity at baseline. Of the 1,158 with-
out PD, 203 had positive ultrasound at baseline.
Therefore, the resulting table is:
Test 1 Test 2
PD 1 14 3 17
PD 2 203 955 1,158
217 958 1,175
Sensitivity 5 82.4%; specificity 5 82.5%
LR1 4.7 (95% CI 5 3.7–6.1),
LR
2
0.21 (95% CI 5 0.08–0.60)
(b) Prospective studies without control groups: In certain
cases, prospective follow-up of an at-risk group is reported,
but without control data. In this case, we estimate LR using
2% prior probability of PD, with published prevalence esti-
mates of the prodromal marker in the general population
and the PD population. For example:
Polysomnogram-proven RBD: Prospective studies
show that more than 75% will develop disease (com-
bining five long-term cohort studies
13-17
). MSA patients
(8% of convertors, or 9 “Test 1” patients) are elimi-
nated from calculations (note that most patients diag-
nosed as DLB would also meet MDS-PD criteria for
PD over their lifetime
4
). Prevalence of RBD in general
population51.15%
18
; population comparison uses life-
time PD risk, assuming risk 5 2%:
Test 1 Test 2
PD 1 77 123 200
PD 2 29 9,771 9,800
106 9,894 10,000
Sensitivity 5 38.5%; specificity 5 99.7%
LR
1
130 (confidence interval not calculable)
LR
2
0.62
Selection of Markers
A literature review (Medline search, supplemented by
review of reference lists of articles and expert consulta-
tion) was conducted to identify prospective studies of
prodromal markers of PD. The results of this review
and resultant LRs are provided in Table 1 and detailed
calculations in the Supporting Methods section. Note
again that for uncommon markers, LR
2
are generally so
close to 1 that they need not be added to the calcula-
tions (delineated N/A). Two prodromal markers, color
vision (LR
1
1.9 before adjustment, LR
2
N/A)
19
and the
PD-related pattern of glucose utilization (LR
1
3.0 before
adjustment, LR
2
0.5)
20
have thus far been tested only in
idiopathic RBD patients; because of generalizability con-
cerns, they were not included in these criteria.
Risk Markers
1. Sex: men have approximately a 1.5-fold increased
risk of PD.
-LR5 1.2 for men, 0.8 for women
2. Regular occupational exposure to pesticides (or
very frequent [>100 episodes] of nonoccupational
exposure)—RR/odds ratio [OR] 5 1.5–1.8,
21,22
prevalence approximately 5%.
-LR
1
1.5, LR
2
N/A
3. Occupational exposure to solvents: OR 5 1.58
21
(prevalence <5%).
-LR
1
1.5, LR
2
N/A
4. Nonuse of caffeine, defined as <3 cups of coffee or <6
cups of tea per week. RR 5 1.5, prevalence 5 25%.
-LR
1
1.35, LR
2
0.88
11
5. Nonsmoking status: This has three possible catego-
ries—only one LR is applied.
MDS CRITERIA FOR PRODROMAL PD
Movement Disorders, Vol. 30, No. 12, 2015 1603
