NO1 Top Black Magic Specialist In Lahore Black magic In Pakistan Kala Ilam Ex...
Progress in Cardiovascular Diseases 53 (2010) 55–61www.onlin.docx
1. Progress in Cardiovascular Diseases 53 (2010) 55–61
www.onlinepcd.com
Framingham Heart Study: The First 20 Years
Gerald M. Oppenheimer⁎
Brooklyn College and the Graduate Center, City University of
New York
Center for the History and Ethics of Public Health, Mailman
School of Public Health, Columbia University
Abstract The Framingham Heart Study remains the most famous
and influential investigation in
Statement of Conf
⁎ Address reprint
Associate Professor, C
Mailman School of P
NY 10032.
E-mail address: g
0033-0620/$ – see fro
doi:10.1016/j.pcad.20
cardiovascular disease epidemiology. To generations of
epidemiologists, it is a model for the
cohort design. Here we revisit the origins of the Framingham
Study before it became an
accomplished and famous investigation whose existence and
success are taken for granted.
When in 1947 the Public Health Service initiated the study,
knowledge of the distribution and
determinants of coronary heart disease was sparse.
Epidemiology was primarily focused on
3. 10.03.003
of successful epidemiological research, productive of
insights and applications…[and] the prototype and model
of the cohort study.”5
The purpose of this article is to revisit the origins of
the Framingham Study before it became an accomplished
and famous investigation whose existence and success
are taken for granted. I will describe the problems that
physicians, statisticians, and public health officials
confronted in order to create an epidemiological study
and later to defend the initial clinical recommendations
drawn from the path-breaking epidemiological results.
Framingham, however, was but one of several CVD
cohort studies funded by the federal government in the
wake of World War II. It was preceded by the Minnesota
Business and Professional Men's Study, headed by
physiologist Ancel Keys, and quickly followed by
investigations in Los Angeles, Calif; San Francisco,
Calif; and Albany, NY. The United States Public Health
Service was anxious to elucidate the causes of cardiovas-
cular disease, to which it attributed 44% of total deaths
in 1948, 20% higher than what it had been in 1940.6
55
mailto:[email protected]
http://dx.doi.org/10.1016/j.pcad.2010.03.003
Abbreviations and Acronyms
FHS = Framingham Heart
Study
CVD = cardiovascular disease
4. PHS = Public Health Service
TB = tuberculosis
NHI =National Heart Institute
CHD = coronary heart disease
RCT = randomized controlled
trial
56 G.M. Oppenheimer / Progress in Cardiovascular Diseases 53
(2010) 55–61
What those causes were
remained amatter of spec-
ulation. Experts on CVD
were increasingly focused
on the etiological role of
arteriosclerosis and the
importance of cultural
factors.7 Unlike an alter-
native model, aging, and
degeneration as determi-
nants of the disease, the
new conceptualization
permitted a role for pri-
mary prevention.
Heart disease was part of a growing concern about
chronic disease within the Public Health Service (PHS). In
the first decades of the 20th century, the PHS, along with
insurance companies and the Milbank Memorial Fund,
had pioneered morbidity surveys whose results suggested
that the burden of chronic illness and disability was rising.
During the Depression, the PHS conducted a massive
National Health Survey. Over one sixth of the US
5. population suffered from a disabling chronic disease or
physical impairment, the PHS found, and half of these
individuals were younger than 45 years; chronic disease
and disability affected many in their most productive
years.8 The PHS recognized the rising prevalence of
chronic disorders and called for public health research
aimed at primary and secondary prevention of heart
disease, cancer, stroke, and diabetes. World War II
temporarily postponed that policy.
Launching the study
In 1947, epidemiology's primary focus remained the study
of infectious disease. At the PHS, JosephMountin, director of
the Bureau of State Services, had recently created the
Communicable Disease Center for that purpose.9 But
Mountin, a master of public health policy, guided the PHS
in a new direction. He championed the application of control
programs, initially developed for infectious disorders, to
chronic, non-infectious diseases.11 These generally included
community-based screening and diagnostic interventions.
In pressing for heart disease control efforts, Mountin
added an epidemiological component.10 He subsequently
selected Gilcin Meadors, a young PHS office, to initiate
that epidemiological research, an investigation that evolved
into the FHS.12
Precisely how epidemiology could be applied to CVD
was unclear. Wade Hampton Frost, a pioneer in the
epidemiological study of chronic disease, tuberculosis
(TB) in particular, had investigated the factors responsible
for TB transmission and susceptibility.13 But, in this
instance, the necessary cause, the bacillus, was known, as
was tuberculosis' pathophysiology. How were epidemiol-
ogists to capture any of the determinants of CVD in the
absence of such knowledge about heart disease? How
6. were they to map a disorder which appeared to have a
long, silent onset before it abruptly announced itself?
A number of issues made the launching of the
Framingham study particularly difficult. The first, a
unique factor, was the multiple authorities invested
in the nascent project.12 The PSC worked with the
Massachusetts Commissioner of Health, Vlado Getting,
and David Rutstein, newly appointed chair of Harvard's
Department of Preventive Medicine. Rutstein who had
previously been medical director of the American Heart
Association, was the prime mover behind Framingham's
selection as the study site. Getting supported the idea,
noting that the town, just west of Boston, possessed 28 000
inhabitants, sufficient for an epidemiological investiga-
tion, and had a history of community participation in
research, having been the site of an important TB study
conducted from 1917 to 1923.14,15 (In addition, Framing-
ham had a stable, locally employed and medically served
populace; in 1947, it boasted two hospitals and a group of
well-trained physicians interested in the possibility of a
scientific study in their town10).
David Rutstein, a powerful and willful figure, had
significant ties to the PHS leadership. After leaving the
American Heart Association, he retained links to the
academic elite that dominated the Association, drawing
upon its Boston members to form Framingham's
technical advisory group. Rutstein16 sought to bend
the new study to his own clinical purposes. To the
Surgeon General, he suggested an evaluation of
population screening tools like the electrokymograph,
which recorded the movement of the heart.10,16 Both he
and his advisory board of cardiologists particularly
wanted to focus the research on diagnostic criteria and
clinical performance.
7. Gilcin Meadors, by contrast, was a young Mississippi-
born physician who had only graduated from Tulane
Medical School in 1940 and just completed a masters
degree in public health in epidemiology at Johns
Hopkins.12 By July 1947, before the Framingham site
had been selected, he developed the outline of a plan that
proposed a goal recognizable to later generations of
Framingham leaders:
“This project is designed to study the expression of
coronary artery disease in a normal or unselected popula-
tion and to determine the factors predisposing it to the
development of the disease through clinical and laboratory
examination and long-term follow-up of such a group.”17
Meadors envisioned a study composed of volunteers
who would be followed for 5–10 years, periodically
interviewed and tested to identify incipient cases of CVD
among them. Subsequent analysis would measure the
57G.M. Oppenheimer / Progress in Cardiovascular Diseases 53
(2010) 55–61
effect of heredity and psychic status, constitutional
factors like body build and behaviors like dietary habits.
Meadors17 hoped in time to develop “methods for
examination of large population groups for heart
disease,” but he was unclear as to how his variables
were to be measured and analyzed. Some baseline factors
he sought to capture, such as psychic status, were
difficult to measure validly. He also wrongly believed
that a self-selected cohort could give him an unbiased
estimate of incidence and prevalence. Analytically weak,
Meadors had problems substantiating his claims for the
importance of clinical epidemiology to those like
Rutstein who were skeptical of its usefulness or
8. committed to alternative scientific goals. The scientific
purpose and design of the Framingham study was thus
contested terrain, pitting personalities, specialties, and
research approaches.
Meador's17 strength lay in his ability to draw the
Framingham community into support of the study,
whatever its subsequent shape. Ultimately of national
and international importance, the FHS began very much as
a local experiment bounded by the strengths and
weaknesses of a single town (the first foreign publications
using Framingham data only date from the mid-
1960s18,19). Meadors drafted town leaders to form the
core of his Executive Committee. The latter, in turn,
recruited critical constituencies, including the business,
industrial and civic elite, and formed outreach and
publicity committees. Meadors recognized that, to be
successful, the study had to be adopted, if not truly
“owned,” by the town's citizenry. Only strong community
support would stimulate voluntary participation and
subsequent adherence. Meadors insisted that “the Heart
Disease Study belongs to the community of Framingham
[and] is a part of the local health program.”20 As such, he
had to gain the confidence and support of Framingham's
physicians, assuring them that any information on
participants would be sent to them and not divulged to
their patients by the study's staff.12
As a small, self-governing New England town with a
prosperous economic sector and an unusually stable
population, Framingham was an ideal site for a cohort
study. But these features also made it highly unrepresen-
tative. In particular, the Framingham Study sample
consisted almost entirely of white, middle class Euro-
Americans. Despite its impulse to scientific universality,
the study lacked Blacks, Latinos, or Asians.
9. Over Framingham's first year, Rutstein proved increas-
ingly rash and proprietary, confronting the PHS, for
example, over administrative control of the Framingham
program. Thereafter, he urged a separation of the
epidemiological and control projects, perhaps initiating
the move of the latter to Newton, Mass.12 This split left the
epidemiological study with greater autonomy, more so as
Rutstein's influence faded as he made more demands the
PHS could not accept.
Revamping the study
But another event proved more decisive. The
epidemiological sophistication and validity of the Fra-
mingham study was transformed in 1949, the year it
was transferred to the newly created National Heart
Institute (NHI). The first head of the NHI directed Felix
Moore, his chief of biometrics, to evaluate Framingham's
methodology.21 Moore, a quantitative sociologist with
considerable federal experience, brought to Framingham
a talent for applied statistics, expertise in writing and
scaling questionnaires, and the rigor of years of research
work.22 During Framingham's first year under the
NHI, he was probably the principal architect of its scien-
tific transformation.23
With Meadors' assistance, Moore redesigned Framing-
ham into a 20-year study of individuals, initially free
of arteriosclerotic disease, selected randomly from the
town's adult residents, both male and female, 30 to
59 years of age.23,24 Because Moore underestimated the
refusal rate among those randomly selected, he and
Meadors had to fall back on volunteers; ultimately, they
comprised 14% of the cohort study's 5127 participants,
adding a further bias to the study.10,12 Baseline variables
10. were also reexamined. After some debate, cholesterol was
included, as were weight and blood pressure and the
consumption of alcohol and tobacco. From the study's
start, clinical variables predominated, with little interest in
psychosomatic, constitutional, or sociological determi-
nants of heart disease.25
Moore was particularly concerned about selecting
those variables that could by validly and reliably
measured. But other issues also intruded. As the
Framingham study depended on the good will of its
participants, he and the director who succeeded Meadors
in spring 1950, Thomas Dawber, excluded those items
they thought would deeply disturb people, for example,
questions about psychiatric status, sexual dysfunction,
and (perhaps given the conservative Cold War era)
income and social class.10,26,27 Dawber, who led
Framingham over the next 2 decades, was also skeptical
about the value of the social sciences or public health in
what he regarded as the domain of medicine. To him,
epidemiology was “clinical investigation on a commu-
nity level.”28 He strongly believed that his study should
be defined by doctors, and that the outcomes it sought
should those useful to practicing physicians.29
Finally, the study was guided by a manual of operations
describing its “methods of examination and acceptable
criteria for diagnosis.”30 The FHS was to correlate clinical
and laboratory data and individual histories collected
before the onset of disease with “findings related to and
58 G.M. Oppenheimer / Progress in Cardiovascular Diseases 53
(2010) 55–61
diagnostic of…degenerative cardiovascular disease.”30 In
11. this way, it would “detect early signs pointing to probable
development of disease” and uncover “etiological
factors.”30 The manual also laid out 28 factor-specific
hypotheses. Each of these expressed a possible link
between CVD and degree of exposure to a factor, for
example, “[d]egenerative cardiovascular disease appears
early and progresses more rapidly in persons who
habitually use tobacco.”30 In the logic of these hypotheses,
one can already perceive the kernel of the “risk factor.”
Although the term first appeared in a Framingham study
article in 1961,31 the concept predates the heart study32
and pervaded its publications over the 1950s.
Risk factors, primary prevention, and the coronary
prone patient
With its supportive cadre of NHI statisticians, always a
source of Framingham's strength, results came surpris-
ingly fast. At the American Public Health annual meeting
in 1956, Dawber et al26 presented some of the central
findings of the FHS. They revealed that atherosclerotic
heart disease was significantly associated with age and
male sex. In men 45 to 62 years of age, for whom
sufficient data existed, heart disease was correlated with
hypertension, hypercholesterolemia, and obesity. Where
two factors were elevated, probability of CVD increased
greatly. The value of Framingham's findings were
enhanced by corroboratory results from the Los Angeles
and Albany heart studies, reported at the same
meeting.33,34 Over the next decade, these cohort studies
would extend their findings. By 1960, for example,
Framingham and Albany, by combining data to provide
sufficient statistical power, showed a significant associa-
tion and dose-response relationship between cigarette
smoking and myocardial infarctions and heart disease–
12. related deaths.35
The stature of epidemiology as a method for
investigating CVD was heightened by the rapidity with
which Framingham and other contemporary cohort
studies reported their results. So too were the mortality
data that emerged. Framingham found that through 1956,
the case fatality rate for heart attacks was nearly 50%;
almost a third of these were “sudden” fatalities (defined
as deaths within minutes).26 With such stunning evi-
dence, the test for epidemiologists became what to do
with their information. Given the high coronary death
rates and the absence of curative therapies, they took on
the task of convincing their clinical colleagues, many of
them skeptical of quantitative population-based data, to
educate their patients about the value of primary
prevention of CVD.36
The message that Dawber and others26 within
Framingham like William Kannel sent to physicians
focused on the notion of susceptibility. In this they were
joined by the heads of other studies like Ancel Keys and
Jerry Stamler (who initiated a heart disease prevention
trial in Chicago as early as 1957).36-38 Kannel and
Dawber,36 starting in the early 1960s, advanced the
concept of the “coronary-prone” individual, the patient
already at greater probability of developing heart disease.
Physicians were urged to intervene early, based on their
diagnosis of their patients' risk-laden signs and symp-
toms, before sudden death, a heart attack or angina
pectoris announced the disease's existence.
From its earliest decades, Framingham sought to
make itself indispensable to medical practice, allaying
the indifference with which many contemporary doctors
may have held it. Dawber and Kannel intimated that the
13. “coronary-prone” patient could easily be identified using
normal office procedures.36,39 Theoretically, that person
was an obese middle-aged man, a heavy smoker with
high serum cholesterol and blood pressure, low vital
capacity and electrocardiographic abnormalities.40 In the
office, the picture would be less clear, more nuanced, but
doctors were warned that a combination of risk factors
increased coronary heart disease (CHD) probability,
doubling, tripling, or even quadrupling the risk.31
Finally, from epidemiological data, Dawber and Kannel
warned that, contrary to physicians' training, there was
no cutoff in levels of blood pressure or serum
cholesterol, below which they were “risk-free” or
“normal.”31 In fact, given epidemiological observations,
almost every patient was susceptible to CVD, especially
in modern cultures like the United States. The art and
science of medicine was then to determine, given
patients' composite of signs, symptoms and history,
their degree of “risk.”
Causal uncertainty and epidemiological activism
By the early 1960s, the epidemiologists at Framingham
and their contemporaries had begun to develop a new
concept of chronic disease based on notions of probability
and disease prevention.41 This model did not require much
understanding of underlying disease mechanisms. It
assumed that the disorders that were killing most
Americans, heart disease, cancer, and stroke, were the
consequences of a long induction period and generally
multiple factors, some of which could be isolated. Often,
certainly in the case of heart disease, none of these factors
were necessary or sufficient. Frequently, they were of
small statistical effect (the instance of cigarette smoking
and lung cancer being an obvious exception). It was
14. therefore difficult to predict who would be afflicted by the
chronic disease, who would not. Although chronic
diseases might be the consequence of “modern life,”
there was little purchase, politically and socially, in
advocating broad environmental reforms. In the 1960s,
the response to heart disease, stroke, and cancer was
59G.M. Oppenheimer / Progress in Cardiovascular Diseases 53
(2010) 55–61
primarily a medical one, advocating individual behavioral
or “life-style” change, initially, at least, against the grain of
market and cultural forces.
In making such life-style recommendations, epidemiol-
ogists like those at Framingham were drawing inferences
not yet supported by their evidence. They based their
advocacy on observational research, their cohort studies,
and not on controlled experiment. With the exception of
cigarette smoking, there was no scientific confirmation at
the time that by reducing risk factors—high blood
pressure, cholesterol, or weight in patients—doctors
would be able to lower the risk of heart disease, producing
outcomes like those in persons with habitually lower
numbers.40 And there was no way of predicting whether
the correlations between heart disease and risk factors,
based on current studies, would hold true in future ones.
Most importantly, epidemiologists were still without
tested scientific knowledge of the underlying causal
chain of events that led to heart disease. If they knew
anything, they knew correlations.
Without evidence of causality, Framingham and other
clinical epidemiologists fell back on the need for good
clinical practice and for action. William Kannel argued, for
15. example: “If physicians were to wait for unquestioned proof
of efficacy of every therapeutic measure they are called
upon to employ, they would never treat cancer, arthritis, or
most of the degenerative and metabolic disorders. In short,
they would be reduced to therapeutic nihilists.”40
Certainly, any physician could read and act upon the
risk characteristics of his or her patients, even absent proof
of the efficacy of such actions. Doing so—reducing blood
pressure or weight—was consistent with standard medical
care. And the hazards involved here were small compared
with the potential benefits.
In 1964, the Surgeon General's Advisory Committee
on smoking and health added its considerable scientific
and political status to the position taken by Dawber,
Kannel, and their colleagues. Interestingly, in the famous
report, Smoking and Health, which authoritatively out-
lined for a generation of epidemiologists the criteria
necessary for judging causality, the committee addressed
the need for precautionary action against heart disease risk
factors, despite the absence of demonstrated causal
association. Referring to cigarette smoking, high blood
pressure, serum cholesterol, and obesity, it argued, “The
causative role of these factors in coronary disease, though
not proven, is suspected strongly enough to be a major
reason for taking countermeasures against them.”42 Here a
lower level of proof seemed apt, given the consistency of
evidence across epidemiological studies and the enormous
toll wrought by CHD.
Despite the advocacy of the Advisory Committee and
the persistence of CVD epidemiologists, many physicians
in the 1960s were reluctant to accept the need for primary
prevention. There was still controversy over whether to
treat essential hypertension or wait until a patient's
16. pressure became very high or complications had set in.43
Doctors were unsure whether serum cholesterol was a
factor in CHD.43,44 Such questions persisted, despite the
consistency of the results produced by the growing
number of CVD cohort studies.
Constructing a consensus
The inability of the epidemiological cohort studies to
allay doubts about the importance of most risk factors
spawned a drive to field randomized controlled trials
(RCTs), a relatively new technology. Beginning in the
early 1960s, researchers sought to show under controlled
conditions that, by reducing such factors, they could
measurably diminish CHD events. To do so required the
resources of the US government, the NHI in particular. At
that time, the NHI, unlike its peer, the National Cancer
Institute, had little experience with such trials. Yet, over
the 1960s, the NHI produced two significant double-blind
RCTs focusing on the possibly causal role of cholesterol:
the Diet-Heart study and the Coronary Drug Project.45 The
Diet-Heart study undertook the seemingly impossible
task of testing the “coronary-genic” effects of the
contemporary American diet. More conventional scientif-
ically, the Coronary Drug Project assessed the effect of
5 cholesterol-lowering medications on CHD. Although
neither, for different reasons, succeeded, they were
followed by additional controlled trials. These helped
establish a broad consensus within the NHI that, like
smoking cessation, reducing blood pressure and choles-
terol lowered rates of CHD events; RCTs in the United
States and Europe played a pivotal part in shifting the
position of many physicians and policy makers toward
favoring primary prevention as a weapon against cardio-
vascular disease.
17. Population-based studies, particularly randomized con-
trolled trials and observational cohort studies of CHD,
helped effect a major change in American culture and
marketplace in the second half of the 20th century.
Throughout that period, as investigations waxed and
waned, one scientific venture ceaselessly continued on its
prescribed course, producing a massive, corroboratory
base of data on the natural history of CVD. Beyond
Meadors' imagining, the Framingham study he initiated in
1947 still seeks, in his words, “the factors predisposing
[a normal population] to the development of the disease
through clinical and laboratory examination and long-term
follow-up of such a group.”
The Framingham study and its longevity were only
possible because of federal commitment to the project,
financially, politically and intellectually. It was a product
of the purposeful US government that emerged from the
60 G.M. Oppenheimer / Progress in Cardiovascular Diseases 53
(2010) 55–61
Great Depression and World War II. Subsequently, this
intensely local study, initially of indefinite shape and
purpose, has become an icon of modern epidemiology,
both here and abroad. But it remains very much an
American story, a linkage of government, science, and
community, midway in time between the New Deal and
the Great Society.
Acknowledgments
This work was supported in part by a National Endow-
ment for the Humanities Fellowship Award (FA-53395)
and a Professional Staff Congress-City University of
18. New York Award (#610940039).
Statement of Conflict of Interest
The author declares that there are no conflicts of interest.
References
1. Garcia-Palmieri MR, Feliberti M, Costas R, et al: An
epidemiological
study on coronary heart disease in Puerto Rico: the Puerto Rico
Heart
Health Program. Bol Asoc Med PR 1969;61:174-179.
2. Kagan A, Rhoads GG, Zeegen PD, et al: Coronary heart
disease
among men of Japanese ancestry in Hawaii: the Honolulu Heart
Study. Israel J Med Sci 1971;7:1573-1577.
3. Christakis NA, Fowler JH: The spread of obesity in a large
social
network over 32 years. N Engl J Med 2007;357:370-379.
4. Christarkis NA, Fowler JH: The collective dynamics of
smoking in a
large social network. N Eng J Med 2008;358:2249-2258.
5. Susser M: Epidemiology in the United States after World War
II: the
evolution of a technique. Epidemiol Rev 1985;7:147-177.
6. Moore F. Minutes, October 14-15, 1949, in Records of the
National
Institute of Health, National Advisory Heart Council, Box 21,
National Archives, College Park, MD.
7. Keys A: Nutrition in relation to the etiology and the course
19. of
degenerative diseases. J Am Diet Assoc 1948;24:281-285.
8. Fox DM: Power and Illness: The Failure and Future of
American
Health Policy. Berkeley, CA: University of California Press;
1993.
9. Etheredge EW: Sentinel for Health: A History of the Centers
for Disease Control. Berkeley, CA: University of California
Press; 1992.
10. Dawber TR: The Framingham Study: The Epidemiology of
Atherosclerotic Disease. Cambridge, MA: Harvard University
Press; 1980.
11. Mountin JW: Changing concepts of local public health
services, in
Selected Papers of Joseph W. Mountin, M.D. Joseph Mountin
Memorial Committee n.p., 1956.
12. Oppenheimer GM: Becoming the Framingham Study, 1947-
1950.
Am J Public Health 2005;95:602-610.
13. Frost WH: Risk of persons in familial contact with
pulmonary
tuberculosis. Am J Public Health 1933;23:426-432.
14. Aronowitz R: The social construction of coronary heart
disease risk
factors. In: Aronowitz R, editor. Making sense of illness:
science,
society and disease. New York, NY: Cambridge University
Press;
1998. p. 111-144.
20. 15. Framingham Monograph No. 10: Final Summary Report,
1917-1923
Inclusive. Framingham, MA n.p., 1924.
16. Rutstein D: Correspondence to T. Parran, October 10, 1947,
in
Papers of the National Heart, Lung and Blood Institute,
Epidemi-
ology Correspondence Folder.
17. Meadors G: Correspondence to B. Boone, July 19, 1947, in
Papers of
the National Heart, Lung and Blood Institute, Epidemiology
Correspondence Folder.
18. Rose G: Familial patterns in ischaemic heart disease. Br J
Prev Soc
Med 1964;18:75-80.
19. Kannel WB, Widmer LK, Dawber TR: Gefahrdung durch
coronare
Herzkrankheit; Folgerungen fur die Praxis aus 10 Jahren
Framing-
ham-Studie. Susceptibility to coronary heart disease; preventive
implications of 10 years of observation in the Framingham
Study.
Schweiz Med Wochenschr 1965;95:18-24.
20. Meadors G: Correspondence to B. Boone, May 6, 1948, in
Papers of
the National Heart, Lung and Blood Institute, Monthly Report
Heart
Disease Epidemiology Folder.
21. Meadors G: Correspondence to E.W. Ford, March 1, 1949,
in Papers
21. of the National Heart, Lung and Blood Institute, Monthly
Report
Heart Disease Epidemiology Folder.
22. Gordon T: NHI statistician, oral communication; 2001.
23. Moore F: Memorandum, August 1, 1949, in Papers of the
National
Heart, Lung and Blood Institute, Framingham Heart Disease
Epidemiology (General) Folder.
24. Giroux E: Cohort study and multivariate analysis: An
epistemolog-
ical and historical analysis of the Framingham heart study.
Revue
d'Epidemiologie et de Santé Publique 2008;56:177-188.
25. Moore F: Correspondence to C.J. Van Slyke, August 26,
1949, in
Papers of the National Heart, Lung and Blood Institute,
Framingham
Heart Disease Epidemiology (General) Folder.
26. Dawber TR, Moore FE, Mann GV: Coronary heart disease in
the
Framingham Study. Am J Public Health 1957;47:4-24.
27. Schrecker E: No Ivory Tower: McCarthyism and the
Universities.
New York, NY: Oxford University Press; 1986.
28. Dawber T: Risk factors in young adults: the lessons from
epidemio-
logical studies of cardiovascular disease—Framingham,
Tecumseh,
and Evans County. J Am Coll Health Assoc 1973;22:84-95.
22. 29. Dawber TR, personal communication, November 2001.
30. National Heart Institute: Heart Disease Epidemiology Study,
manual
of operations, 1949, in Papers of the National Heart, Lung and
Blood
Institute, Historical Folder.
31. Kannel W, Dawber T, Kagen A, et al: Factors of risk in the
development of coronary heart disease—six-year follow-up
experience, the Framingham Study. Ann Intern Med 1961;55:
33-48.
32. Rothstein W: Public health and the risk factor: a history of
an uneven
medical revolution. Rochester, NY: University of Rochester
Press;
2003.
33. Doyle JT, Haslin AS, Hilleboe HE, et al: A prospective
study of
degenerative cardiovascular disease in Albany: report of three
years
experience. Am J Public Health 1957;4:25-32.
34. Chapman JM, Goerke LS, Dixon W, et al: Measuring the
risk of
coronary heart disease in adult population groups. The clinical
status
of a population group in Los Angeles under observation for two
to
three years. Am J Public Health 1957;4:33-42.
35. Doyle JT, Dawber TR, Kannel WB, et al: Cigarette smoking
and
23. coronary heart disease: combined experience of the Framingham
and
Albany studies. N Eng J Med 1962;266:796-801.
36. Dawber TR, Kannel WB: Susceptibility to coronary heart
disease.
Mod Concepts Cardiovasc Dis 1961;30:671-677.
37. Keys A, Taylor HL, Blackburn H, et al: Coronary heart
disease
among Minnesota business and professional men followed 15
years.
Circulation 1963;28:381-395.
38. Stamler J, Berkson DM, Young Q, et al: Approaches to the
primary
prevention of clinical coronary heart disease in high-risk,
middle-
aged men. Ann N Y Acad Sci 1963;97:932-951.
61G.M. Oppenheimer / Progress in Cardiovascular Diseases 53
(2010) 55–61
39. Kannel WB, Dawber TR, McNamara PM: Detection of the
coronary-
prone adult: the Framingham Study. IowaMed Soc 1966;56:26-
34.
40. Kannel WB, Kagan A, Dawber TR, et al: Epidemiology of
coronary
heart disease. Implications for the practicing physician.
Geriatrics
1962;17:675-690.
41. Oppenheimer GM: Emergence of coronary heart disease
24. epidemio-
logy in the United States (1947-70). Int J Epidemiol
2006;35:720-730.
42. U.S. Department of Health, Education and Welfare. Surgeon
General's Report, Smoking and Health: Report of the Advisory
Committee to the Surgeon General of the Public Health Service.
PHS Publication No. 1103. Washington, DC: Government
Printing
Office; 1964.
43. Dawber TR, Kannel WB: Application of epidemiology of
coronary
heart disease to medical practice. Mod Med 1962;30:85-101.
44. Kannel WB: personal communication, July 11, 2002.
45. Marks HM: The Progress of Experiment, Scientific and
Therapeutic
Reform in the United States, 1900-1990. Cambridge, England:
Cambridge University Press; 1997.
Framingham Heart Study: The First 20 Yearssection1Launching
the studyRevamping the studyRisk factors, primary prevention,
and the coronary �prone patientCausal uncertainty and
epidemiological activismConstructing a
consensusAcknowledgmentssection8References
ARTICLES
The epidemiology of chronic pain in the community
Alison M Elliott, Blair H Smith, Kay I Penny, W Cairns Smith,
W Alastair Chambers
25. Summary
Background Chronic pain is recognised as an important
problem in the community but our understanding of the
epidemiology of chronic pain remains limited. We
undertook a study designed to quantify and describe the
prevalence and distribution of chronic pain in the
community.
Methods A random sample of 5036 patients, aged 25 and
over, was drawn from 29 general practices in the Grampian
region of the UK and surveyed by a postal self-completion
questionnaire. The questionnaire included case-screening
questions, a question on the cause of the pain, the chronic
pain grade questionnaire, the level of expressed needs
questionnaire, and sociodemographic questions.
Findings 3605 questionnaires were returned completed.
1817 (50-4%) of patients self reported chronic pain,
equivalent to 46-5% of the general population. 576
reported back pain and 570 reported arthritis; these were
the most common complaints and accounted for a third of
all complaints. Backward stepwise logistic-regression
modelling identified age, sex, housing tenure, and
employment status as significant predictors of the
presence of chronic pain in the community. 703 (48-7%)
individuals with chronic pain had the least severe grade of
pain, and 228 (15-8%) the most severe grade. Of those
who reported chronic pain, 312 (17-2%) reported no
expressed need, and 509 (28-0%) reported the highest
expressed need.
Interpretation Chronic pain is a major problem in the
community and certain groups within the population are
more likely to have chronic pain. A detailed understanding
26. of the epidemiology of chronic pain is essential for efficient
management of chronic pain in primary care.
Lancet 1999; 354: 1248-52
Department of General Practice (A M Elliott BSC, B H Smith
MRCGP)
and Department of Public Health (Prof W C Smith MFPHM).
University of Aberdeen Medicai School, Aberdeen;
Department of Anaesthetics, Aberdeen Royai infirmary,
Aberdeen
(W A Chambers FRCP); and iVIedical Statistics Unit,
University of Edinburgh iVIedical School, Edinburgh (K I
Penny phD),
UK
Correspondence to: Mrs Alison M Elliott, Department of
General
Practice and Primary Care, University of Aberdeen, Foresterhill
Health
Centre, Aberdeen AB25 2AY, UK
Introduction
Chronic pain is recognised as a common problem within
the community, and is known to affect general health,'
psychological health," and social and economic well-
being.'' Chronic pain is one of the most common reasons .
why people seek medical care; patients with chronic pain
use health services up to five times more frequently than
the rest of the population.'' Chronic pain also results in
many days lost from work.' Nevertheless, our
understanding of the epidemiology of chronic pain
remains limited. Several previous studies of the
prevalence of chronic pain have been based on patients
attending pain clinics,'"" but these patients represent one
extreme of the population with chronic pain and are not
27. likely to be representative of the general population.'^'"
Good community-based epidemiological studies have
been difficult to do. Many approaches used to estimate
the prevalence of chronic pain have had limitations, such
as the use of non-specific measurement instruments.'''*
Some studies have concentrated on patients with specific
pain-related diagnoses, such as back pain, fibromyalgia,
and chronic musculoskeletal pain,^"" and others have
been small.'"' Previous estimates of the prevalence of
chronic pain in the general population have ranged from
2% to 45%.^'"''^'"-"'
We present results from an epidemiological study of
chronic pain, designed to quantify and describe the
prevalence of chronic pain in the community, according
to age, sex, and sociodemographic status.
Methods
Participants and procedures
The International Association for the Study of Pain defines
chronic pain as "pain that persists beyond normal tissue
healing time, which is assumed to be 3 months"." This
definition does not allow for intermittent pain, so for the
purposes of this study we defined chronic pain, after pilot
testing, as "pain or discomfort, that persisted continuously or
intermittently for longer than 3 months".
The study was done in the Grampian region of northeast
Scotland. All practices that used the General Practice
Administrative System for Scotland (GPASS) were invited to
collaborate. We had to limit the study to these practices
because GPASS was required for a related study. Of the 67
practices eligible to participate in the study, 29 (43%) were
keen to be included. These practices accounted for about 35%
of the total Grampian population and were representative of
28. the practices in Grampian in terms of size of practice,
geographical location, and socioeconomic status of those
attending. There was no follow-up of the remaining practices.
5036 patients aged 25 years and older who attended
participating practices were selected randomly from the
community health index (a list of all patients registered with a
general practitioner). The sample was stratified for sex and age
(six age-groups ranging from 25-34 years to 75 years and over)
and was weighted to ensure an equal representation of patients
in each of the subgroups. Each list of patients was screened by
the relevant general practitioner so that inappropriate or
insensitive inquiries could be prevented (for example, in the
1248 THE LANCET • Vol 354 • October 9, 1999
ARTICLES
388543 people in
GPASS practices
185747 people in
participating practices
5036 randomly
selected
425 excluded by
general practitioner
4611 sent
questionnaires
232 questionnaires not delivered
29. 176 incorrect or unknown
addresses
56 not contactable
4379 questionnaires
delivered
774 did not reply
3605 respondents
Study profile
case of terminal illness). No specified criteria for the exclusion
of patients were provided. General practitioners were simply
asked to score through the name of anyone whose inclusion
they felt would be unsuitable or inappropriate.
Questionnaire
Participants were surveyed by a postal self-completion
questionnaire, developed through piloting and validation. This
questionnaire included case-screening questions, a question on
the cause of the pain, the chronic pain grade questionnaire, the
level of expressed needs questionnaire, and sociodemographic
questions.
Chronic pain cases were identified through two questions:
one question to assess whether pain or discomfort was present,
and a second to establish whether this pain or discomfort had
started longer than 3 months ago. Patients were asked to say
whether they knew what caused their pain and were given a
choice of responses (based on pilot studies) of angina, arthritis,
back pain, injury, women's problems, don't know, and other.
The chronic pain grade questionnaire'™ is a seven-item
30. questionnaire that measures severity of chronic pain in three
dimensions: persistence, intensity, and disability. The results of
the questionnaire enable patients to be classified into five
hierarchical grades: grade 0 (pain free), grade I (low disability,
low intensity), grade II (low disability, high intensity), grade III
(high disability, moderately limiting), and grade IV (high
disability, severely limiting). This questionnaire has been tested
Education
No educational qualifications
Secondary schooi qualifications
Higher education quaiifications
Marital status
Never married
Currently married
No longer married
Housing tenure
Owned or mortgaged
Council rented
Privately rented and other
Number of occupants
Living alone
Living with others
Employment status
Empioyed
Retired
Not able to work owing to
iilness/disabiiity
Others not empioyed
Social class (occupation based)
Highly skilled
31. Skilled
Part or non-skilled
Number of
respondents
1637
960
735
376
2528
659
2436
898
208
751
2639
1809
1086
161
428
708
719
351
Odds ratio*
(95% CI)
32. 1-OOt
0-93 {0-78-110)
0-99 (0-82-119)
l o o t
1-23 (0 98-1-54)
1-20 (0-91-1-57)
1-OOt
1-29 (1-10-1-51)
0-97 (0-72-1-29)
1-OOt
0-99 (0-83-1-17)
1-OOt
1-33 (1-03-1-72)
8-15 (5-03-13-21)
0-92 (0-73-1-16)
1-OOt
1-02 (0-82-1-26)
1-07 (0-82-1-40)
P
0-40
0-89
0-08
0-20
0-01
0-82
33. 0-87
0-03
<0-001
0-50
0-87
0-61
•Adjusted for age and sex. tReference category.
Table 2 : Association of demograpiiic factors witii ciironic pain
for use in a UK population,^' and was found to be acceptable,
valid, and reliable for use as a self-completion postal
questionnaire. Only those patients who responded positively to
the case-screening questions, which identified whether chronic
pain was present, were asked to complete the chronic pain
grade questionnaire.
The level of expressed needs questionnaire is based on
previous work.̂ ^ It represents a measure of patients' response to
chronic pain in a way that reflects demand for and uptake of
health-service resources and has meant that a series of
increasing amounts of help-seeking behaviour can be identified.
The questionnaire enquires about the treatment sought by
patients in an attempt to relieve their pain, as measured by how
recently and frequently patients used painkillers or visited
health professionals. For the purposes of this study, this
questionnaire consisted of four questions: Have you sought
treatment for your pain or discomfort recently? Have you
sought treatment for your pain or discomfort ofren? Have you
taken painkillers for your pain or discomfort recently? Have
you taken painkillers for your pain or discomfort ofren?
35. Rate of chronic pain (range*)
31-8 (19-1-58-9)
38-3 (25-6-58-8)
47-6 (35-8-60-6)
53-9 (44-5-62-0)
56-6 (49-6-61-9)
59-9 (49-6-66-8)
48-9 (37-0-61-4)
Women
n
94
140
170
198
190
173
965
Rate of chronic pain (range*)
31-6 (23-3-49-6)
43-6 (35-0-54-8)
53-1 (42-9-62-1)
60-2 (50-4-66-7)
57-9 (50-0-63-7)
64-3 (51-3-71-5)
51-8 (41-8-61-1)
•Minimum rates assume no non-responders had chronic pain;
36. maximum rates assume ail non-responders had chronic pain.
Table 1 : Prevaience of self-reported ciironic pain
All respondents
n
170
242
309
371
379
346
1817
Rate of chronic pain (range*).
31-7 (21-3-54-3)
41-2 (30-3-56-8)
50-5 (39-4-61-4)
57-1 (47-4-64-3)
57-3 (49-8-62-8)
62-0 (50-4-69-1)
50-4 (39-4-61-2)
THE LANCET • Vol 354 • October 9, 1999 1249
ARTICLES
Sex
Men
37. Women
Number of
respondents
1511
1610
Odds ratio*
(95% CI)
Total with Grade I
chronic pain
Grade II Grade III Grade IV
1-00*
1-24 (1-07-1-43) 0-003
Sex
Men 693
Women 752
348 (50-2%) 176 (25-4%)
355 (47-2%) 177 (23-5%)
70 (10-1%) 99 (14-3%)
9 1 (12-1%) 129 (17-2%)
Age-group (years)
25-34
35-44
45-54
55-«4
65-74
375
40. 165
1-00*
1-23 (1-03-1-47)
0-92 (0-66-1-29)
0-02
0-63
Employment type
Empioyed
Retired
Unable to work owing to
iiiness/disabiiity
Others not empioyed
1672
917
136
396
1-00*
1-32 (1-00-1-74)
7-75 (4-59-13-09)
0-96 (0-76-1-22)
0-047
<0-001
0-73
* Reference category.
Table 3: Logistic-regression modei of factors significantiy
41. associated witii ciironic pain
Data analysis
Each patient was sent a copy of the questionnaire accompanied
by a letter, signed by his or her own general practitioner and
the general practitioner in the research group. Up to two
reminders were sent after intervals of 2 weeks to maximise the
response rate.
Questionnaire responses were entered into the Microsoft
Access database (version 2) and the data were analysed with
SPSS (version 6). x̂ tests were used to test for associations
between sociodemographic variables and questionnaire
responses, and Mantel-Haenszel tests were used to investigate
trends across age-groups. To identify the sociodemographic
factors that were associated with chronic pain, odds ratios
(adjusted for age and sex), 95% CI, and p values for a two-
sided test were calculated. We considered for inclusion in a
model all factors for which the odds ratio had a p value below
0-2,̂ '' that were deemed clinically important, or that had had a
significant association with chronic pain in previous studies.
Backward stepwise logistic-regression modelling was then used
to obtain a subset of sociodemographic factors that were
associated with chronic pain. We used likelihood ratio tests to
judge whether or not to remove factors.
Results
Of the 5036 individuals randomly selected, 425 were
excluded from the survey by general practitioners
(figure). The reason for exclusion was not always given.
267 exclusions were explained: 72 patients had died; 99
were no longer registered with the practice; 18 had
terminal illness; 43 were demented, frail, or very elderly;
nine had mental illness; 12 were in hospital; and 14
Back
43. 55-64
65-74
S 7 5
Total 576 (16-0%) 570 (15-8%) 213 (5-9%) 161 (4-5%) 139 (3-
9%) 154 (4-3%)
Percentages are of totai in each subgroup.
Table 4: Seif-reported cause of pain
Total 1445 703(48-7%) 353(24-4%) 161(11-1%) 228(15-8%)
Percentages are of total with chronic pain in each subgroup who
completed the chronic
pain grade questionnaire.
Table 5: Seif-reported ciironic pain severity
were excluded for other reasons. Of 4611 questionnaires
mailed, 176 questionnaires were not delivered because
the address was unknown or incorrect, and were
returned to us. 56 people did not receive their
questionnaires; this group consisted of people who had
died recently, who were in hospital, or who were
working overseas. The patients' families notified us of
these individuals.
3605 questionnaires were completed and returned—a
response rate of 82-3%. The response rate varied with
age and sex. 1864 (83-9%) of 2223 women responded,
compared with 1741 (80-8%) of 2156 men (p<0-01).
The response rate increased with age from 72-3% (536
of 741) in the age-group 25-34 years to 88-9% (558 of
628) in the age-group 65 years and older (p<0-001).
44. The proportion of respondents reporting chronic pain
was 50-4% (table 1). There was no significant difference
between men and women in the proportion reporting
chronic pain (48-9 vs 51-8%) but the proportion
significantly increased with age from 31-7% for the
youngest age-group to 62-0% for the oldest age-group
(p<0-001). By standardising the sample to the age and
sex distribution of the total population of patients
registered with participating practices, we estimated the
prevalence of chronic pain in the general population as
46-5% (95% CI 44-8-48-2).
To assess the possible effect of differences in response
in the various age-sex groups, we calculated the
minimum possible rate of chronic pain (based on the
assumption that none of the non-responders had
chronic pain), and the maximum possible rate (based on
the assumption that all of the non-responders had
chronic pain).
Table 2 gives odds ratios for pain associated with
various sociodemographic factors. Individuals living in
rented council accommodation were more likely to have
chronic pain than those who lived in owned or
mortgaged property. Individuals who were retired or not
able to work were also more likely to report chronic pain
than those who were employed. To obtain a subset of
sociodemographic variables associated with chronic
pain, backward stepwise logistic-regression modelling
was done (table 3). Age, sex, housing tenure, and
employment status were identified as factors
independently and significantly associated with chronic
pain. Although sex was not a significant factor in the
univariate analysis, it was included in the final model as
a significant factor, with women more likely than men to
report chronic pain. Chronic pain was also associated
45. with older age, living in rented council accommodation,
and being retired or unable to work.
1250 THE LANCET • Vol 354 • October 9, 1999
A R T I C L E S
Sex
Men
Women
Age-group
25-34
35-44
45-54
55-64
65-74
a 7 5
Total
Total with
chronic pain
852
965
170
242
309
371
379
346
48. Percentages are of total with chronic pain in each subgroup.
Table 6: Expressed need of patients with chronic pain
Back pain and arthritis were the two most commonly
reported causes of chronic pain, accounting for about a
third of all reported causes (table 4). Back pain was the
most common problem in men and in the younger age-
groups, and arthritis was the most frequent cause of
pain in women and in the older age-groups. A large
number of other causes of pain were given. Some people
identified the actual complaint and others reported an
anatomical site for their symptoms. The most frequently
reported complaints in this category were leg pain
(1-7%) and headaches (1-0%). All other causes given
were reported by less than 1% of the sample.
Table 5 shows the pain severity as measured by the
chronic pain grade questionnaire. Of 1445 individuals
who reported chronic pain, 703 (48-7%) had pain
graded as least severe (grade I), and 228 (15 8%) had
pain graded as most severe (grade IV). There was no
significant difference in the distribution of grades of
pain between men and women but the frequency of the
more severe pain grades (grades III and IV) increased
with age (p<0-001).
Of those patients with chronic pain, 312 (17-2%)
showed no expressed need and 509 (28 0%) showed the
highest level of expressed need (table 6). The small
percentage of individuals with chronic pain who had no
apparent expressed need suggests that patients with
chronic pain are likely to make extensive use of health
services. Women were more likely to report high
expressed need than were men (p<0-001). The level of
49. expressed need also varied significantly with age
(p<0-001).
Discussion
This study shows that chronic pain is a major problem
in the community, with about half of the sample
reporting chronic pain; after standardisation we estimate
that 46-5% of the general population have chronic pain.
This figure is higher than previous estimates of chronic-
pain prevalence in general population studies."*'''"*""'
This discrepancy is likely to reflect differences in the
research methods and case definitions used; differences
that previously accounted for the majority of variation in
general-population prevalences." Prevalence of chronic
pain also varies widely across different geographical
locations.^
These findings expand our understanding of chronic
pain in two important ways. Unlike many previous
studies, our study was done in a community setting and
used a definition of chronic pain based on the most
commonly used standard definition, that of the
International Association for the Study of Pain. Also,
our study is the first general population study of all
chronic pain to use validated measures of the severity
and significance of chronic pain. This approach allows
stratification of chronic pain, which means that in
addition to providing prevalence rates for any chronic
pain we have established the prevalence of severe and
significant chronic pain. These prevalence rates are
likely to be of greatest clinical significance, and of more
use, to physicians and health-service planners.
The response rate for the study was high, with slightly
higher response rates from women than men, and
50. increasing response rates with age. The increase in
response rate with age is likely to be related to the fact
that prevalence of chronic pain increases with age. To
explore the possible effects of non-response on the
prevalence of chronic pain, we calculated minimum and
maximum rates for each subgroup of age and sex. The
range between the minimum and maximum rates is
widest for the youngest age-group, reflecting the lower
response rate in these patients.
Analysis of the sociodemographic factors identified
age, sex, housing tenure, and employment status as a
subset of factors significantly associated with chronic
pain. These findings highlight subsets of the general
population who may be more likely to have chronic pain
and are important in consideration of the onset and
management of this important and common problem.
Further longitudinal studies are needed to establish the
direction of these sociodemographic associations, and to
differentiate causes from effects.
Although the proportion of our sample who had
chronic pain was high, most individuals (48-7%)
reported pain of the mildest chronic pain grade. Even
so, 17-2% of those with chronic pain reported no
expressed need. A fairly small proportion (15-8%)
reported the most severe grade of pain, yet 28-0% of all
those with chronic pain reported the highest expressed
need. These results suggest a disparity between reported
severity of chronic pain and the expressed need of
patients, but the disparity arises because the two
instruments measure different features of chronic pain.
The chronic pain grade assesses severity of chronic pain
in terms of intensity and disability, and the level of
expressed need questionnaire measures patients'
response to their chronic pain in terms of treatment and
51. analgesics. Our results suggest that response to chronic
pain, as measured by the level of expressed needs
questionnaire, is not solely dependent on pain intensity
and disability and may reflect the multifactorial and
subjective experience of, and response to, chronic pain.
The disparity between the two measures is important
since it may highlight areas of unmet need and highlight
THE LANCET • Vol 354 • October 9, 1999 1251
ARTICLES
those patients who display high expressed need yet have
mild chronic pain. Further research into the
complexities of this issue is in progress.
There is evidence that many patients with pain, in
many cases treated inadequately, could be relieved with
improved management by general practitioners."
Management strategies need to be practicable and
appropriate for use in primary care, where the majority
of patients who seek treatment will present. These
management strategies must be evaluated by
randomised control trials, the design and interpretation
of which will be informed by epidemiological studies of
chronic pain in primary care.
Contributors
Blair Smith, Cairns Smith, and Alastair Chambers designed the
study.
Alison Elliott collected the data. Alison Elliott and Kay Penny
analysed
the data. The paper was drafted by Alison Elliott and edited by
all other
52. investigators.
Acknowledgments
We acknowledge the support of the Scottish Home and Health
Department, grant reference K/RED/4/C286, and the patients
and
general practices for their support.
References
1 Becker N, Thomsen AB, Olsen AK, Sjorgren O, Bech P,
Eriksen J.
Pain epidemiology and health related quality of life in chronic
non-
malignant pain patients referred to a Danish multi-disciplinary
pain
center. Pain 1997; 73: 393-400.
2 Magni G, Marchetti M, Moreschi C, Merskey H, Luchini SR.
Chronic musculoskeletal pain and depressive symptoms in the
National Health and Nutrition Examination—I: epidemiological
follow-up stidy. Pain 1993; 53: 163-68.
3 Gureje O, Von Korff M, Simon GE, Gater R. Persistent pain
and
well-being: a WHO study in primary care. JAMA 1998; 280:
147-51.
4 Turk DC, Rudy TE. Toward an empirically derived taxonomy
of
chronic pain patients: integration of psychological assessment
data.
J Consult Clin Psychol 1988; 56: 233-38.
5 Latham J, Davis BD. The socioeconomic impact of chronic
pain.
53. DisabilRehabil 1994; 16: 39-44.
6 Locker D. Disability and disadvantage. London: Tavistock
Publications, 1983.
7 Von Korff M, Dworkin SE, Le Resche L. Graded chronic pain
status: an epidemiologic evaluation. Pain 1990; 279: 91.
8 Von Korff M, Wagner EH, Dworkin SF, Saunders KW.
Chronic
pain and use of ambulatory health care. Psychosom Med 1991;
53:
61-79.
9 Bowsher D, Rigge M, Sopp L. Prevalence of chronic pain in
the
British population: a telephone survey of 1037 households. Pain
Clinic 1991; 4: 223-30.
10 Davies HTO, Crombie IK, Macrae WA, Rogers KM. Pain
clinic
patients in North Britain. Pain Clinic 1992; 5: 129-35.
11 Crook J, Weir R, Tunks E. An epidemiological follow-up
survey of
persistent pain sufferers in group family practice and speciality
pain
clinic. Pain 1989; 36: 49-61.
12 Crombie IK, Davies HTO. Selection bias in pain research.
Pain
1998; 74: 1-3.
13 Smith BH, Smith WC, Chambers WA.Chronic pain-time for
epidemiology, j'5oi; AfeJ 1996; 89: 181-83.
54. 14 Crook J, Rideout E, Browne G. The prevalence of pain
complaints
in a general populaton. Pain 1984; 18: 299-314.
15 Croft PR, Macfarlane GJ, Papageorgiou AC, Thomas E,
Silman AJ.
Outcome of low back pain in general practice: a prospective
study.
BAfJ 1998; 316: 1356-59.
16 Croft P, Rigby AS, Boswell R, Scholium J, Silman A. The
prevalence
of chronic widespread pain in the general population. J
Rheumatol
1993; 20: 710-13.
17 Brattberg G, Thorslund M, Wikman A. The prevalence of
pain in a
gneneral population: the results of a postal survey in a county
of
Sweden. Pain 1989; 37: 215-22.
18 Verhaak PF, Kerssens JJ, Dekker J, Sorbi MJ, Bensing JM.
Prevalence of chronic benign pain disorder among adults: a
review of
the literature. Pain 1998; 77: 231-39.
19 International Association for the Study of Pain.
Classification of
chronic pain. Pain 1986; suppl 3: S1-S226.
20 Von Korff M, Ormel J, Keefe FJ, Dworkin SF. Grading the
severity
of chronic pain. Pain 1992; 50: 133-49.
21 Smith BH, Penny KI, Purves AM, et al. The Chronic Pain
55. Grade
Questionnaire: validation and reliability in postal research. Pain
1997; 71: 141-47.
22 Purves AM, Penny KI, Munro C, et al. Defining chronic pain
for
epidemiological research—assessing a subjective defmition.
Pain
Clinic 1998; 10: 139-47.
23 Blaxter M. Health and lifestyles. London:
Tavistock/Routledge,
1990.
24 Rothman KJ, Greenland S. Modern epidemiology, 2nd edn.
Philadelphia: Lippincott-Raven, 1998.
25 Diamond A. The future development of chronic pain relief.
Anaesthesia 1991; 46: 83-84.
1252 THE LANCET • Vol 354 • October 9, 1999