Greetings,

Linda Herman and I researched some of the literature on this
question, and we present our results in the following report,
which Jerry Finch has also been posted at the
www.parkinsonalliance.net web site.  Note that columns in tables
should line up if viewed with a courier font. Comments, especially
from anyone else who has researched this subject, are welcome.

Phil Tompkins
Hoboken NJ
60/9

===================================================================


    HOW MANY PEOPLE IN THE U.S. HAVE PARKINSON'S DISEASE?

                by Phil Tompkins and Linda Herman

                         July 5, 1998


                    I.  Introduction

If there were better epidemiological data on Parkinson's disease, and
if this data were disseminated to researchers, the media, government
officials and the general public, this could lead to a better
understanding of the causes of Parkinson's disease, the discovery of
preventative measures and improved treatments, more equitable
research funding from the government, and increased knowledge and
acceptance of persons with Parkinson's disease among the general
public and in the workplace.

Even the basic question of how many people in the U.S. have
Parkinson's disease lacks a satisfactory answer.  We see statements
like "About 50,000 Americans are diagnosed with Parkinson's disease
each year, with more than half a million Americans affected at any
one time" (1) and "It is estimated that Parkinson's disease affects 1
to 1.5 million people in the United States" (2) in books and articles
on Parkinson's disease written for the general public, on Internet
web pages, and in justifications for increased government funding.
That's a very broad range of numbers.

There is more precise data on the occurrence of many other chronic
diseases in the U.S.  Isn't it therefore possible to obtain better
data on Parkinson's disease?

With hopes of determining just where these various numbers come from,
what they mean, and what it might take to get a more precise count,
the two of us, a college librarian and a computer software analyst,
with no training in epidemiology but with curiosity motivated by the
fact that we have Parkinson's disease, undertook a review of some  of
the professional literature that reports and discusses the surveys of
Parkinson's disease in the U.S.  This document reports what we found
and attempts  to summarize what is known at present about how many
people in the  U.S. have Parkinson's disease.

We found that the epidemiological survey data collected thus far on
Parkinson's disease in the U.S. is too scant to serve as a basis for
any U.S. totals beyond a rather rough estimate.  The intensive
surveys which have been conducted are few in number and cover just a
few areas of the country and a relatively small number of cases. Some
of the surveys are over 30 years old.  And none of the total figures
from the surveys are in proportions of one million to the total U.S.
population.

In the course of our research we gained an appreciation of what is
involved in taking a disease census and making estimates.  We
realized that there is data on diagnosed patients that has already
been collected and recorded on electronic media and is merely
waiting to be tallied and analyzed.  And we learned of a
widely-quoted annual U.S. government survey that collects data on
chronic diseases but neglects to ask about Parkinson's disease.

Before looking in detail at all the numbers, what they mean, and
where they come from, we need to introduce some definitions. Then we
will discuss the survey data and conclude by offering some
suggestions.


              II.  Units of Disease Measurement

There are two basic measures of the occurrence of diseases.

1.  Incidence.  Incidence is the ratio of the number of new cases
which arise during a time period (usually one year) to the number of
people in the whole population.

b2.  Prevalence.  Prevalence is the ratio of the number of
existing cases to the number of people in the whole
population.

Incidence and prevalence are sometimes expressed as percents, and
sometimes as fractions with a numerator of 1, as in 1 person in 250.

By convention, epidemiologists convert the incidence and prevalence
ratios to fractions with a common denominator of 100,000.  This is
done for the purpose of making comparisons on a convenient scale.
Even if an actual population is much smaller than 100,000, incidence
and prevalence may still be expressed in terms of cases per 100,000.

Note that using 100,000 as a common denominator does not imply
anything at all about whether the incidence and prevalence which were
determined from a survey of a particular population sample are
reflected in like proportion in any real group of 100,000 people.

Incidence and prevalence are mathematically related through another
number, the average duration of the disease. Incidence is equal to
prevalence divided by the average disease duration.  The average
duration depends in turn on the ages of onset, the number of people
who are at each age, and for diseases like Parkinson's disease that
have no cure, the remaining years of life after onset.  Canadian
epidemiologist A. H. Rajput suggests that a "reliable method to
determine prevalence rate is to multiply the best available incidence
rates with the best available information on mean life expectancy
after onset. . . " (3).


                    III.  Types of Data Sources

Data from which incidence and prevalence can be computed is obtained
from the following sources:

1.  Patient medical records.  Data may be extracted from physician or
medical institution records retrospectively, or it may be gathered as
patients appear for treatment.  The Mayo Clinic in Rochester,
Minnesota has been a principal source of medical records data for
Parkinson's disease studies.

The main disadvantage to using medical records data is that people
who have not yet been diagnosed and people who do not receive
professional medical treatment are not included.

2.  Household interviews.  A number of door-to-door surveys have been
performed throughout the world, often involving health care personnel
capable of making diagnoses.  Only one such survey was conducted in
the United States specifically for Parkinson's disease, in 1978 for
Copiah County, Mississippi.

3.  Reportable diseases.  Physicians are legally required to report
certain contagious diseases to government health officials.
Alzheimers, although not contagious, is now a reportable disease.
Parkinson's disease is not.  A registry of Parkinson's disease cases
would provide not only incidence and prevalence data, but much
valuable clinical data as well.

4. Death certificates.  Death certificates as an indicator of
incidence may be useful for such diseases as cancer and heart
disease, but not for Parkinson's disease.  Only a single cause of
death is recorded on death certificates, rather than all the medical
diagnoses of the deceased.  Parkinson's disease is sometimes
indicated as the cause of death, but this is rare.  Even if all the
diagnoses could be recorded, it is unlikely that they would be.  A
certifying physician with no knowledge of the medical history of the
deceased would not be able to determine the presence of Parkinson's
disease without an autopsy.

A. H. Rajput notes one study of death certificates where "[in] only
46% of the known cases, the Parkinson's disease diagnosis was listed
as the primary cause of death" (3).

5.  Drug sales.  We found studies (4, 5) which evaluated the use of
drug sale data as an indicator of prevalence and determined it to be
reliable, but we found no subsequent studies that put this source of
data into use.

6.  Insurance claims.  Insurance claims data has been used for
studies in countries where there is a national health plan, such  as
Canada (6).  In the U.S. such data would have to be collected  from
many sources.  We did not find any studies of Parkinson's disease in
the U.S. that used this data.


        IV.  U.S. Parkinson's disease Survey Results

In professional literature which reviews the surveys of Parkinson's
disease in the U.S. (7, 8, 9, 10, 11), those surveys which are
typically cited cover three areas of the country:  Rochester,
Minnesota; Baltimore, Maryland; and Copiah County, Mississippi.

Much less data exists on incidence than on prevalence.  The estimated
incidence for the U.S. of 20 or 20.5 per 100,000, which has been
cited for years, and from which the number 50,000 new cases per year
appears to be derived, seems to come from the Rochester data alone.
It is interesting that this number is generally accepted as the
incidence for the entire U.S. even though it is apparently based on
just one area of the country.  Could this acceptance be due to the
absence of any conflicting incidence figures?

Regarding prevalence, the results of the above-mentioned surveys are
summarized in the table below.  Also shown are figures from surveys
of part of Manhattan, N.Y., of a selected group of older males in
Hawaii, and, for comparison's sake, of Alberta, Canada.  We show
"crude" figures rather than adjusted ones.  (Survey figures from
various populations with different age, gender, and/or ethnic group
mixes are often adjusted to be in equivalent proportions.  If they
are made proportionate to a government census, the census population
is taken as a standard.  Such an adjustment is called
"standardization.") Prevalences are shown here to two decimal places
on a scale of cases per 1,000.  Although it makes no sense to speak
of fractions of people, the number 1,000 may be easier to imagine
concretely than 100,000 is.


             Table 1.  U.S. Survey Results

  _Location________ Date_____Population____Cases___Prev/1000_
 |                                                           |
 | Rochester MN     1955        29,885        56       1.87  |
 | Rochester MN     1965        47,797        75       1.57  |
 |                                                           |
 | Baltimore MD     1969                                     |
 |   Estimated                2,070,000     1,630       .79  |
 |     Black                                                 |
 |       Female                   -           -         .09  |
 |       Male                     -           -         .31  |
 |     White                                                 |
 |       Female                   -           -        1.21  |
 |       Male                     -           -        1.28  |
 |   Sample                                  228             |
 |     Black                      -           17         -   |
 |     White                      -          211         -   |
 |                                                           |
 | Copiah Co. MS    1978         23,597       31       1.31  |
 |   Black                       11,666       12       1.03  |
 |   White                       11,931       19       1.59  |
 |                                                           |
 | Oahu Hawaii (older males only) ****                       |
 |  60 & older      1982         6,603        39      [5.91] |
 |  65 & older      1987         5,753        38      [6.61] |
 |  70 & older      1992         4,668        42      [9.40] |
 |                                                           |
 | Manhattan NY     1991       213,302       228       1.07  |
 |                                                           |
 | Alberta Ca       1984-89   2,400,000      5903      2.44  |
 |    Female                                           2.40  |
 |    Male                                             2.49  |
 |___________________________________________________________|

Note: numbers in brackets were computed by the authors.


In the Baltimore study (12) the number 1,630 cases in 2,070,000 was
estimated based on a smaller sample.  The number of Parkinson's
disease cases in the sample was 228. The size of the sample
population is not given in the report which we reviewed.  The main
purpose of the Baltimore study was not to determine prevalence, but
to identify causal factors by comparing people with Parkinson's
disease to a matching group of people without Parkinson's disease.

The prevalence differences by ethnic group reported for Baltimore may
be due to the sampling method and/or to differences in accessibility
of health care.  The patient population was not selected directly.
What was first selected was a group of 112 physicians whose patient
records were then studied.  In the physician sample were 28 of the 30
neurologists and neurosurgeons in Baltimore, all of whom were white.
127 of the patients in the sample came from these physicians.  There
were 5 black physicians in the sample, all general practitioners.  No
hospital clinic patients were surveyed.  We did not see a discussion
of the significance of the race of the doctors.

In the Copiah County study (13), where 97 percent of the county's
population was included in the sample, at least some of the
difference by ethnic group may be due to longevity. The proportion
of people over age 50 was 23 percent of the black population (2670 of
11,666) and 33 percent of the white population (4092 of 11,931).  It
is noteworthy that 40% of the cases were newly diagnosed as a result
of the survey. Thus another benefit of more comprehensive
epidemiological studies of Parkinson's disease could be an increase
in the number of people diagnosed, and earlier treatment for those
individuals.

Data for a study of Parkinson's disease in Hawaii (14) was obtained
from the Honolulu Heart Study, an ongoing study of 8,006 men of
Japanese or Okinawan ancestry born between 1900-1919 who resided in
Oahu Hawaii in 1995. Parkinson's disease cases were identified by 1)
review of hospital records for the cohort subjects, 2) ongoing review
of death certificates, 3) review of records of 8 neurologists, 4)
rescreening of the group in 1991-1994.  The age of the group studied
accounts for the apparent high rates, which are age-specific.

Although in the Manhattan study the report claims that "nearly
complete case ascertainment [was] achieved," an apparent problem with
the data is that the patients were self-selected. ." . . we
advertised in periodicals and on radio and television" (15).

The Alberta study (16) used the claims administrative data from the
Alberta Health Care Insurance Plan. "A cohort [study sample observed
over a period of time] of [i.e., from] all registered individuals
(2.4  million) was extracted and followed for a five year period,
April 1, 1984 - March 31, 1989. The overall crude prevalence rate of
248.9 and 239.8 per 100,000 were noted for males and females
respectively."  Registration with the AHCIP is required of all
residents of the province and is virtually complete.

In a review of mainly Canadian studies, epidemiologist A. H. Rajput
provided an estimated prevalence rate of 300/100,000 for Canada.
Rajput describes the methods of determining prevalence and their
shortcomings.

"In our large study of all inclusive PS [Parkinson's syndrome] cases
in Saskatoon, the mean survival was 12.3 years. If we do not make
allowance for the underdiagnosis, the prevalence rate would be
252/100,000. On the other hand, if we take the underdiagnosis into
consideration, the prevalence rate would be 333/100,000 - a rate
surprisingly close to 328/100,000 reported by Bharucha in a Parsi
community in India.... [The Parsi community had the highest recorded
prevalence of any study.] In my opinion the PS prevalence rate today
in Canada is close to 300/100,000" (3). Yet one cannot assume that
prevalence as determined for Canada applies to the U.S.  Disease
prevalences usually differ for different regions.