Int J Clin Oncol (2008) 13:102–111
DOI 10.1007/s10147-008-0762-6
REVIEW ARTICLE
Donald Maxwell Parkin
The role of cancer registries in cancer control
Received: January 17, 2008
Abstract Cancer control aims to reduce the incidence, mor-
bidity, and mortality of cancer and to improve the quality
of life of cancer patients through the systematic implemen-
tation of evidence-based interventions in prevention, early
diagnosis, treatment, and palliative care. In the context of
a national cancer control program (NCCP), a cancer sur-
veillance program (CSP), built around a population-based
cancer registry, is an essential element. Data on the size and
evolution of the cancer burden in the population are essen-
tial to evaluation of the current situation, to setting objec-
tives for cancer control, and defining priorities. Cancer data
are essential in monitoring the progress of the implementa-
tion of an NCCP, as well as providing an evaluation of the
many individual cancer control activities. In the context of
an NCCP, the CSP should provide a focus of epidemiologi-
cal expertise, not only for providing statistical data on inci-
dence, mortality, stage distribution, treatment patterns, and
survival but also for conducting studies into the important
causes of cancer in the local situation, and for providing
information about the prevalence of exposure to these
factors in the population. Cancer surveillance via the popu-
lation-based registry therefore plays a crucial role in formu-
lating cancer control plans, as well as in monitoring their
success.
Key words Cancer registration · Cancer control ·
Surveillance
Introduction
A cancer control program is a public health program
designed to reduce the incidence and mortality of cancer
and improve the quality of life of cancer patients in a
particular country or state, through the systematic and
D.M. Parkin (*)
Clinical Trials Service Unit and Epidemiological Studies Unit,
University of Oxford, Richard Doll Building, Old Road Campus,
Oxford, OX3 7LF, UK
Tel. +44-1865-743743; Fax +44-1865-743985
e-mail: max.parkin@ctsu.ox.ac.uk
© The Japan Society of Clinical Oncology 2008
equitable implementation of evidence-based strategies
for prevention, early detection, treatment, and palliation,
making the best use of available resources.1 It is well-rec-
ognized that, as Muir et al.2 noted: “the cancer registry is
an essential part of any rational program of cancer control,
benefiting both the individual and society in which he lives”.
But it was Armstrong who, in 1992,3 reviewed the potential
range of activities that registries might undertake in this
context, what they were actually doing (by means of a
survey of the members of the International Association of
Cancer Registries), and then provided a blueprint for an
“expanded role” for registries, over and above their core
function of providing routine statistics on cancer occurrence
and outcome.
Although, as in much of epidemiology, terminology is by
no means standardized, and writers use the same terms in
different ways, many of the activities envisaged by Arm-
strong might be subsumed by the terms “surveillance” and
“monitoring”.
Surveillance is defined4 as:
1. Systematic measurement of health and environmental
parameters, recording, and transmission of data, and
2. Comparison and interpretation of data in order to detect
possible changes in the health and environmental status
of populations.
“Surveillance” may encompass the concept of “monitor-
ing” of disease – assessing the impact of change or inter-
vention. Monitoring also implies a constant adjustment of
performance in relation to results. The fundamental ele-
ments of a surveillance program are ongoing collection,
analysis, and feedback or dissemination of data.
Disease surveillance plays a critical role in the develop-
ment and implementation of health policy – indeed, rational
planning is impossible without a means of identifying the
main health problems, determining priorities for preventive
and curative programs, evaluating whether goals are reached
in the target groups, and determining what has been
achieved in relation to resources expended. Thus, according
to the World Health Organisation (WHO),1 the role of a
cancer surveillance program is to provide data on a continu-
ing basis:

• To assess the current magnitude of the cancer burden
and its likely future evolution
• To provide a basis for research on cancer causes and
prevention
• To provide information on prevalence and trends in
risk factors, and,
• To monitor the effects of early detection/screening,
treatment, and palliative care.
Research activities relevant to cancer control go beyond
what might legitimately be encompassed by the term
“surveillance”, however. They include not only epidemio-
logical research into cause (including the interplay of envi-
ronmental exposure and genetic determinants) but also
research into the effectiveness of different interventions in
changing behavior (in relation to personal exposures, or
use of services for early diagnosis and screening) and into
aspects of survivorship.5 Armstrong3 pointed out that these
activities go beyond the role of the cancer registry per se,
but are eminently appropriate when the registry forms part
of a “Cancer Control Unit” with a mandate for research
and access to databases other than that of the register
itself.
Planning cancer control programs
Priorities and targets. Projections and predictions
The most basic function of a cancer registry in relation to
cancer control is to assess the current magnitude of the
cancer burden and its likely future evolution. Various sta-
tistics are available for assessing the “burden” of cancer,
and of different types of cancer, in the population.
Incidence is clearly a fundamental measure of burden,
because it describes the stream of new cases that will require
some kind of medical attention. It is the relevant measure
when considering (primary) prevention, the objective of
which is to prevent disease occurrence. Measurement of
incidence is the most basic function of population-based
cancer registries (PBCRs).
Mortality rates are widely used as a “proxy” for inci-
dence in epidemiological studies comparing disease risk
between different populations and over time. They can
serve this function, provided that survival is the same in all
of the groups being compared (or over time). Of more
immediate relevance to cancer control, they provide a
measure of the most serious outcome of disease. Compari-
sons of numbers of lives lost are widely used in health ser-
vices research and health economics; for example, in
evaluating the effectiveness of programs of prevention,
early diagnosis, and treatment of cancer.
Person-years of life lost (PYLL) provide a dimension
additional to mortality, by differentially weighting lives lost
according to age at death. There are many variations in the
calculations used, depending upon the weights to be used
(the value of years of life at different ages), the “normal”
lifespan against which to compare premature death (fixed
103
upper limit, or life table expectations of life), and the “dis-
count rate” to apply to life years that would have been lived
in the future.
The development of indices such as quality adjusted
life years (lost)(QALYs) or disability adjusted life years
(lost)(DALYs) takes this approach further, by quantifying,
in addition, the morbidity that is suffered between onset of
a disease, and death or recovery.
Three elements are needed in calculating these indices—
the incidence of the disease, its mean duration (or, equiva-
lently, survival probability), and a measure of life “quality”
in between onset and end of disease. DALYs have been
extensively used by the World Health Organisation,
to evaluate the challenge to health systems of different
diseases.6
Prevalence of cancer is often advanced as a useful
measure of cancer burden,7 indicating the number of patients
alive who require medical care. There is no standard defini-
tion of a prevalent case of cancer. Theoretically, prevalence
would include everyone once diagnosed as having cancer
who is still alive, but then long survivors who are “cured”
are included, and this would not provide a useful indicator
of the need for treatment and care. A simple compromise
is to regard only patients alive between 0 and 5 (or 10) years
after diagnosis as “prevalent” cancers, because this approxi-
mates the period of active treatment and follow-up of cases.8
Prevalence can be estimated directly by some cancer regis-
tries from their files of cases registered who have not died.9
Alternatively, prevalence can be estimated from the inci-
dence of disease and survival curves, either for short-term
survivors (e.g., to 5 or 10 years), or, if incidence and survival
data are available for long periods of time, including long-
term survivors also.10
Registries record not only the site of the primary tumor
but also other important characteristics, such as histological
type, and size or stage, and may calculate at least the pro-
portional distribution of these characteristics, or, better,
their incidence rates. The histological subtype of cancers
occurring in the same organ has important implications for
the etiology (and hence the appropriate preventive strate-
gies), treatment, and outcome. The size (or stage) of a
cancer is related to the levels of awareness of cancer in the
population, the facilities available for diagnosis, and their
accessibility to the population. Again, these are important
in planning and evaluating appropriate cancer control
strategies.
Survival from cancer is the usual measure used to evalu-
ate cancer treatment. Its computation requires follow-up of
diagnosed cancer patients, and the calculation of numbers
surviving after different intervals of time.
Registries can compute rates of cancer incidence (and
mortality), stage distribution, and survival according to
population characteristics, such as area of residence, socio-
economic status, or ethnicity.11–13 Geocoding of cancer
patients in the registry database permits linkage with census
information from the same small area as the patient to
impute personal characteristics; it has been widely used to
derive indicators of social status (or “deprivation”). These
methods may identify population subgroups that might be
104
Fig. 1. Observed trends in age-standardized rates (world standard), of
colon cancer in men in the Nordic countries 1958–1997, and predictions
of colon cancer rates to 202021
the target of specific interventions within a cancer control
program (e.g.,14,15).
Appraisal of the current situation provides a framework
for action, while the setting of targets allows the success (or
otherwise) of these plans to be monitored. There have been
several attempts to set goals for cancer control, at the
national and the international level.16–19 Logically, target-
setting involves not only a sound knowledge of the present
cancer situation but also an evaluation of how it might be
expected to evolve in future, because interventions can only
impact upon future patterns. Accordingly, there has been
increasing use of projections and predictions of the future
cancer burden. Projections usually involve the assumption
that past trends in rates of incidence/mortality will be main-
tained (and apply to projected changes in the population).
Predictions may encompass the likely effects of changes in
etiological factors or proposed interventions.20 Examples of
cancer projections prepared as an aid to service planning
have been published for the Nordic countries (Fig. 1),21
Australia,22 and Scotland.23
Research as a component of cancer control: the role
of the cancer registry
The original function of the cancer registry was to calculate
rates of incidence, so that the risk of various cancers in dif-
ferent populations could be compared. Although this still
remains their most basic role, the activities of cancer regis-
tries have developed far beyond this, to include studies of
cancer cause and prevention. Ideally, when the cancer reg-
istry forms part of a research institute, or acts as a cancer
surveillance unit/program (CSP) it should provide expertise
to conduct epidemiological studies on cancers of impor-
tance in the local population, and to advise other research-
ers wishing to do so.3 (Various names for this function are
in vogue, such as “cancer intelligence unit” or “cancer
observatory.”) It should act as a source of information on
the importance of any risk factors relevant to the local
population. At the very least, local researchers should be
aware of the registry, and the potential for using its data-
base in approved research projects.
Descriptive studies
Descriptive studies use information from the registry data-
base to examine differences in the incidence (or survival)
of cancer, according to variables associated with place (of
residence, or of birth), time, and personal characteristics
(sex, ethnicity, social status, etc). Classically, such descrip-
tive studies are said to be “hypothesis-generating” – provid-
ing clues to etiology, to be followed up in studies that focus
on specific risk factors. Thus, the role of diet in colon cancer
was suggested by geographic variations,24,25 international
correlations,26 migrant studies,27 and time trends.28 The
detail available in registry data on, for example, histological
type, provides greater insights into etiology than are avail-
able using mortality data – for example, trends in subtypes
of lung29 and esophageal cancers30 suggest changes in the
prevalence of etiological factors.
Studies of cause
Cancer registries are not involved in the majority of epide-
miological studies of cause, except as a source of incidence
rates for studies with an “ecological” design. In theory,
because it includes all cases of cancer in the population, a
cancer registry provides an unbiased source of cases for
case-control studies. In practice, the registry database is
only complete several years after the incident date of the
cases, so “rapid reporting” of cases to the registry is usually
a prerequisite,31 and random sampling of controls from the
general population is necessary.32 Cancer registries have,
however, been extensively used to follow up specific groups
of individuals (“cohorts”) to detect the occurrence of new
cases of cancer. They may involve linkage of pre-existing
databases with the cancer registry (e.g., registers of specific
occupations,33 or of HIV-AIDS34). A special example is the
study of the risk of second cancers (in relation to the initial
cancer and its treatment), where assembly of the cohort,
and its follow-up, make use uniquely of the registry data-
base. These studies are of interest in detecting commonality
of risk factors (or susceptibility to them), or the adverse
effects of treatment.35
Effectiveness of interventions in prevention, early
diagnosis/screening and therapy
It is generally accepted that the effectiveness of therapeutic
interventions must have been demonstrated by means of

a randomized controlled trial (RCT), before they are
implemented in clinical practice. Conceptually, this is a
reasonable condition for other costly (or potentially
harmful) interventions in early diagnosis/screening, and
in prevention. However, although the results of RCTs
were sometimes available before widespread acceptance
(e.g., vaccination against human papilloma virus (HPV)
in preventing precursors of cervical cancer, or mammo-
graphic screening for breast cancer), often the benefit was
felt to be so obvious that no RCT evidence was available
before widespread implementation (most anti-smoking
interventions, cytological screening for cervical cancer,
prostate-specific antigen [PSA] screening for prostate
cancer).
The cancer registry has had a limited role in most RCTs.
On the other hand, post-hoc evaluation of already imple-
mented programs has made wide use of cancer registration.
Conceptually, such research is the same as that involved
in monitoring a program of prevention or screening that
was implemented after a proper appraisal or effectiveness,
and we are concerned with the efficacy of the program in
practice, rather than its theoretical effectiveness in the
context of an RCT. Examples of the use of cancer registries
in the evaluation of implemented programs are provided
below.
Evaluation of cancer control programs
Primary prevention
The effectiveness of preventive interventions against cancer
has rarely been evaluated by RCTs; more usually, success
has to be inferred from observations after the introduction
of programs.20 This may involve comparing observed inci-
dence rates (allowing for a time lag for the effects to
emerge), with the rates expected based on a prediction
model of some kind. This approach has been widely used
to evaluate the success of interventions against lung
cancer.36,37 An even more striking example38 is the dramatic
effect on incidence of hepatocellular carcinoma recorded by
the cancer registry in Taiwan, following the introduction of
vaccination against hepatitis B in the 1980s, first to neonates
born of hepatitis B surface antigen (HBsAg)-positive
mothers, then, in 1984, for all newborns. By 1994, it was
possible to compare liver cancer incidence in children aged
6–9 years born before vaccination was introduced, and
those born after. There was a fourfold difference. Another
approach has been to compare incidence rates in areas with
or without preventive programs, or with different intensities
of intervention.39,40
Screening and early detection
Cancer registry data have been widely used for the evalua-
tion and monitoring of screening programs.41 The effec-
tiveness of screening can only be correctly judged by the
extent to which the objective of reduced mortality (or
105
reduced incidence, for cancer of the cervix) is achieved.
Thus, screening for cervical cancer aims to reduce the inci-
dence of invasive cancer – this is the aim of oral cancer
detection programs also. Other screening programs (e.g.,
breast, large bowel), which aim to detect cancers early,
do not reduce incidence. Their objective is to decrease
mortality.
Ideally, information on the screening status of individu-
als from a suitable database can be linked to that of the
cancer registry in order to study the outcome of individuals
with different intensities of screening exposure. The sim-
plest study design is the cohort study, in which occurrence
of disease (incidence or mortality, as appropriate) is com-
pared in screened vs unscreened individuals, with the objec-
tive of deciding the decrease in risk which might plausibly
be ascribed to screening. Studies such as these were used to
study the potential benefit of cytological screening in pre-
venting invasive cancer of the cervix42,43 or mortality from
gastric cancer,44–46 which have never been demonstrated in
an RCT. The problem with these studies, as in all nonran-
domized comparisons, is the comparability of the groups
being compared – it has been repeatedly shown that indi-
viduals participating in screening programs are not at the
same risk of disease as those who do not. Controlling for
such confounding in an observational study is extremely
difficult.
The introduction of breast cancer screening in Finland
was very systematic, involving certain municipalities, and,
within these, invitations to women in different (single year)
birth cohorts over a 5-year period. This population could
be followed up by the cancer registry and, for a short period
after screening (3–4 years), the mortality rates in screened,
invited but unscreened, and control women compared.47
Although technically a cohort study, women were allocated
to screened and unscreened categories by adjacent birth
cohorts, so there could have been little difference between
the groups in terms of confounders. A small but significant
reduction in mortality was present in women invited
for screening (versus unscreened controls) at 3–4 years
post-screening.
Case-control studies have been extensively used in the
evaluation of the effectiveness of cancer screening pro-
grams.48 These studies have the advantage that screening
histories need only be obtained for a limited number of
subjects. This may even be done by interviewing the sub-
jects (if deaths from cancer do not comprise the case-group),
without recourse to screening records, although interview
data may not always agree with results from more objective
sources.49 There are major difficulties in interpreting the
results of case-control studies. One is the inevitable pres-
ence of selection bias. The effect of this is to exaggerate the
benefits of screening when the individuals who accept
screening tests are a subset of the population likely to have
a lower mortality from the cancer in question than those
who do not.50,51 The other difficulty is the inability to sepa-
rate tests performed in truly asymptomatic individuals,
from those in subjects presenting themselves because of
symptoms (or early disease) or as follow up of individuals
with equivocal results.52
106
When it has been demonstrated that a particular inter-
vention is effective in reducing mortality, intermediate end-
points (such as tumor size, stage, and survival) may be used
to monitor the screening process. By themselves, they
provide no reassurance, as the lessons of screening for neu-
roblastoma53 clearly demonstrate.
Day et al.54 have described how information on incident
cancers in screened/unscreened individuals may be used to
monitor breast cancer screening programs. This informa-
tion includes:
(a) The incidence of interval cancers
(b) The size and stage distribution of screen-detected
cancers (compared to the expected distribution)
(c) The incidence rate of advanced cancers, compared with
the period pre-screening (or an unscreened comparison
group).
Cancer registry data are required in order to estimate
the “expected” distribution of cancers by stage, or the
“expected” incidence rates (overall, by age group, and/or
by stage), against which the results in the screened popula-
tion can be evaluated. Linkage of the cancer registry and
the screening program is required in order to calculate inci-
dence rates of interval cancers (detected between screen-
ings) and the incidence of advanced cancers. The incidence
of interval cancers is useful in decisions about the appro-
priate intervals between tests,55 and the incidence of
advanced cancers provides an early surrogate for cancer
mortality.56
More usually, there is no information on the screening
status of individuals, and population-level analyses are
used. The simplest are time trends in incidence, for cancers
where screening should prevent invasive cancer, such as
that in the cervix. Figure 2 shows perhaps the most cele-
brated example – the trends in incidence in the Nordic
countries, following the introduction of population screen-
ing (Iceland, Finland, Sweden), its partial implementation
(Denmark), and postponement (Norway).57 For programs
designed to detect early invasive cancers (for example,
breast, colon, and prostate), no reduction in incidence
should occur; indeed, the introduction of screening should
be followed by a rise in incidence (as prevalent, asymptom-
atic cases are detected), followed by a fall, with cumulative
incidence unchanged over what it would have been without
screening.58 However, the cancers detected by screening
must have a more favorable stage distribution, and be of
smaller size than those detected clinically (via symptoms) if
the program is effective.
Using registry data alone, it has been common to examine
the proportionate distribution of diagnosed cancer cases by
stage at diagnosis, in relation to the presence of screening
(or measures of screening intensity, as above). More mean-
ingful is calculation of rates of disease, by stage, in relation
to screening activity. In principle, screening should lead to
a reduction in the incidence of advanced cancers. This is
usually in the context of trends over time59–62 (Fig. 3).
An inevitable problem of studying temporal trends in stage
(without a true comparison group) is the phenomenon of
stage migration. This occurs because improvements of
CERVIX UTERI
INCIDENCE/106
300
200 DENMARK
NORWAY
SWEDEN
100
ICELAND
70
FINLAND
50
1945 1950 1955 1960 1965 1970 1975 1980 YEAR
Fig. 2. Trends in age-adjusted incidence rates: cancer of the cervix
uteri
investigative methodology over time result in tumors being
classified to higher stages than they would have been in an
earlier period (by detection of small metastases, for
example). Comparisons between sub-populations receiving
screening or not (or different intensities of screening)63
largely avoid problems of stage migration, but do raise
other questions concerning the comparability of the popula-
tions being compared.
Detection of cancers by screening will lead to improved
survival, whether this results in reduction in mortality (the
goal of screening), or is simply due to advancing the date
of diagnosis (lead-time bias) or differential detection of
slow-growing tumors (length bias). Survival trends, by age,
have been studied in relation to early diagnosis of
breast cancer due to screening62 or to improving breast-
awareness64 and in relation to early detection of colon
cancer;63 the favorable trends in survival were due to a shift
to earlier stage at diagnosis, as well as better survival within
stage.
Outcome: survival and quality of life
While mortality is an indispensable measure of the success
(if any) of cancer control activities in different populations,
trends in mortality rates are influenced both by incidence
and survival. The effect of cancer control activities on mor-
tality will often be quite delayed, and data on both detection
and incidence, and on cancer patient survival, can give a
more immediate insight into changes in outcome after diag-
nosis.65 Randomized controlled trials (RCTs) are the recog-
nized technique for measurement of the precise effects of
specific interventions on survival, but the outcomes demon-
strated in RCTs will rarely be evident at the population
level. For a start, the patients being treated are more
heterogeneous, and are generally elderly, have significant
co-morbidity, and are less likely to derive benefit from
the intervention.66 Many other factors influence survival.

140
Total Invasive Localized Regional
In Situ Distant
120
100
Age-Adjusted Rates/100,000
80
60
40
20
0 tively easy one to measure. Years of life are of little value
1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993
Year of Diagnosis
Fig. 3. Age-adjusted breast cancer incidence rates by extent of disease
at diagnosis for white females, standardized to the 1970 United States
population (from Chu et al.,62 with permission)
Methods of data collection and analysis, tumor characteris-
tics (histology, stage, method of detection), and patient
characteristics (age, sex, ethnicity, socioeconomic status,
lifestyle) affect survival at the population level, as measured
by cancer registries.67,68 In any case, the objective of measur-
ing population-level survival is to give an indication of the
possible role of the process of care, and not simply the
effectiveness of a specific treatment, as a determinant of
survival differences.
International comparisons of survival, such as those
coordinated by the EUROCARE group69 (Fig. 4) have had
a profound influence on policy-making for cancer treatment
services in several countries, e.g., the United Kingdom
and Denmark. Similarly, the black-white disparities in
cancer survival in the United States had been the subject
of comment for many years.70 Although there has been
some justified criticism of the fixation on improvement in
survival as the target for cancer services,71–73 observations
107
Fig. 4. Age-adjusted 5-year relative survival from cancer of the colon
and rectum (males and females), by country, with area-weighted mean
European relative survival. Horizontal bars show 95% confidence
intervals (from Berrino et al.,69 with permission)
of survival differences in studies such as EUROCARE
have focused attention on the reasons underlying these
differences.
Of course, survival in terms of average years lived post-
diagnosis is a very crude indicator of outcome, albeit a rela-
if they are accompanied by disability and pain. For this
reason, survival measures have been refined, either into
medico-surgical categories such as disease-free survival
(before any recurrence) or metastasis-free survival, or by
evaluating the quality of life lived between treatment and
death, in terms, for example, of disability or side effects of
treatment. Measurement of health-related quality of life is
not part of routine cancer registration; it requires quite
complex data collection, by interview or observation,
from individual patients.74 Quantification of the outcome of
cancer care in these various ways is important, when
considering alternative strategies of intervention, as are
the economic considerations with respect to inputs (costs)
required to achieve given improvements in outcome.
Clinical care
To date, much of the research on the quality of clinical
cancer care and consecutive guideline development has
been based on the study of clinical case series, often in
randomized trials. But such studies cannot always be gen-
eralized, based as they are on selected groups of patients.
As a result, population-level understanding of quality out-
comes has been limited, especially for elderly patients. In
recent years, increasing attention has been paid to examin-
108

Fig. 5. Treatment for early-stage

breast cancer, 1991–2002, by age
at diagnosis. United States SEER
registry populations: BCS, breast-
conserving surgery; RT, radiation
treatment. Source, survelliance,
Epidemiology, and End Results
(SEER) 11 registries (Connecti-
cut, Hawall, Iowa, Utah, and New
Mexico and the metropolitan
areas of San Francisco, Detroit,
Atlanta, Seattle-Puget Sound,
San Jose-Monterey, and Los
Angeles County) (from Edwards
et al.,76 with permission)

ing variations in the process of care for the cancer patient,
between different providers, or different groups of patients
(clients).75,76
The location (e.g., type of hospital) of treatment for
specific cancers has been found to vary considerably for
different groups of patients, and this probably reflects the
type of facilities available, and the level of expertise of the
therapists. The technical expertise of physicians and sur-
geons may be a function of their familiarity with the appro-
priate diagnostic and therapeutic methods, as measured in
terms of case load. Site-specialization is recommended in
guidelines to improve the quality of care, and this can be
evaluated in registry data.77 Although these are indirect
indicators of quality of care, they have been shown to be
important in determining outcome.78–80 Simple measures
such as delay in diagnosis, or delay in receiving appropriate
treatment, provide information on equity and access, as well
as potentially influencing outcome.81 Auditing the nature of
the diagnostic and therapeutic procedures actually per-
formed provides a more direct indication of the quality of
care; if the focus is effectiveness of care, then the choice of
indicators should be known to have an effect on its outcome.
The dataset collected by PBCRs is limited with respect to
variables suitable for clinical care studies; they generally
include hospital, specialty of treatment, nature of first
therapy (surgery, radiotherapy, chemotherapy, yes or no),
and outcome (dead or alive). Nevertheless, these variables
may be used to monitor whether clinical practice is evolving
in accordance with clinical consensus. Figure 5 shows how
treatment for early breast cancer in the populations served
by the Surveillance, Epidemiology, and End Results
(SEER) registries in the United States changed between
1991 and 2002 following the 1990 NIH Consensus Develop-
ment Panel conclusion that breast-conserving surgery fol-
lowed by radiation therapy is an appropriate method of
primary therapy for the majority of women with stage I and
II breast cancer and is preferable because it provides sur-
vival equivalent to total mastectomy and axillary dissection
while preserving the breast.76 It is difficult for PBCRs to
expand their dataset, like those of hospital database
systems,82 to capture information on comorbidity, diagnos-
tic and staging procedures, extent of disease at diagnosis
and at surgery, the nature and sequence of treatment, and
follow-up in terms of recurrence and metastasis. This is
more feasible by linkage with other files of clinical data,
such as the Medicare database in the United States,83 and
in studies that use samples of the registry database–confined
to selected cancers, for limited periods of time, and for
representative samples of cases (e.g.,84).The advantages of
the use of population-based data in these ways are that they
relate to the whole community of cancer patients, rather
than a single institution, or self-selected and often atypical
subgroups of patients (such as those reaching comprehen-
sive cancer centers).
Conclusions
For many countries of the world, vital statistics on deaths
by cause are unavailable, cover very restricted populations,
are incomplete, or are inaccurate (especially if cause of

death is certified by nonmedical personnel). Remedying this
situation is a matter of national policy extending beyond
cancer control, and generally, beyond the remit of the
health ministry. Similarly, many countries do not have avail-
able data on cancer incidence from a PBCR. Some informa-
tion on the cancer profile can be deduced from statistics
derived from other data sources, such as hospital discharge
statistics, or pathology department records. However, the
picture that emerges is often quite a biased one, and much
care is needed in the interpretation of the data. For this
reason, all countries introducing a cancer control program
should attempt to include a PBCR as part of the cancer
surveillance program of their NCCP.
For planning purposes, a PBCR does not need to cover
the entire national population, although for many epidemi-
ological studies, this is highly desirable. Coverage of a
sample of the population is adequate for many purposes.
Ideally, the sample would be a reasonably representative
one (as was done with the SEER program in the United
States,85 which currently covers about 26% of the national
population), but it is perfectly possible to prepare national
estimates from registry populations that are nonrandom.86,87
In most countries where medical facilities are limited, the
registry generally covers one or more major cities, where
treatment and diagnostic resources are concentrated, admit-
ting that it is from these that the best quality information
can be obtained. Although this permits the collection of
quite accurate information on incidence, stage, survival, and
deaths, this has the disadvantage that the data are for pre-
dominantly urban populations, and thought must be given
to this in the planning and evaluation of cancer control
programs. On the other hand, the resources required for a
PBCR of this size are quite modest, comprising, as well as
a medical director (part-time with their usual duties), one
or two trained data clerks, a personal computer, expenses
for office supplies, and transport.
In the context of a national cancer control program
(NCCP), a cancer surveillance program (CSP), built around
a population-based cancer registry, is an essential element
with a major role in monitoring the progress of the imple-
mentation of an NCCP, as well as in providing an evaluation
of the many individual cancer control activities. In the
context of an NCCP, the CSP should provide a focus of
epidemiological expertise with the responsibility of provid-
ing data on a continuing basis on incidence, prevalence,
mortality, methods of diagnosis, stage distribution, treat-
ment patterns, and survival. The CSP will provide informa-
tion on the risk factors for cancers of importance locally,
and information about the prevalence of exposure to these
factors in the population. It will therefore play a crucial role
in formulating the cancer control plan, as well as in monitor-
ing its success. It is prudent to have administrative and
financial arrangements within the scope of an NCCP to
ensure the establishment and maintenance of such a unit,
to provide the epidemiological, surveillance, and evaluation
base for the NCCP.
109
References
1. World Health Organisation (2002) National cancer control pro-
grammes. Policies and managerial guidelines, 2nd edn., WHO,
Geneva
2. Muir CS, Demaret E, Boyle P (1985) The cancer registry in cancer
control: an overview. In: Parkin DM, Wagner G, Muir CS (eds)
The role of the registry in cancer control. IARC Scientific Publica-
tions no. 66., IARC, Lyon, pp 13–26
3. Armstrong BK (1992). The role of the registry in cancer control.
Cancer Causes Control 3:569–579
4. World Health Organisation (1973) Environmental and health sur-
veillance in occupational medicine. Technical report series no. 535.
WHO, Geneva, p 7
5. Hiatt RA, Rimer BK (1999) A new strategy for cancer control
research. Cancer Epidemiol Biomarkers Prev 8:957–964
6. World Health Organisation (2003) World health report 2003
“Shaping the future”. WHO, Geneva (also available at: http://
www.who.int/whr/2003/en/)
7. Hakama M, Hakulinen T, Teppo L, et al. (1975) Incidence, mortal-
ity or prevalence as indicators of the cancer problem. Cancer
36:2227–2231
8. Pisani P, Bray F, Parkin DM (2002) Estimates of the world-wide
prevalence of cancer for 25 sites in the adult population. Int J
Cancer 97:72–81
9. Adami HO, Gunnarson T, Spaŕen P, et al. (1989) The prevalence
of cancer in Sweden. Acta Oncol 4:463–470
10. Capocaccia R, Colonna M, Corazziari I, et al. (2002) Measuring
cancer prevalence in Europe: the EUROPREVAL project. Ann
Oncol 13:831–839
11. Jensen OM, Carstensen B, Glattre E, et al. (1988) Atlas of cancer
incidence in the Nordic countries. Nordic Cancer Union,
Helsinki
12. Miller BA, Kolonel LN, Bernstein L, et al. (eds) (1996) Racial/
Ethnic patterns of cancer in the United States 1988–1992. NIH
Publication no. 96–4104. National Cancer Institute, Bethesda,
MD
13. Singh GK, Miller BA, Hankey BF, et al. (2003) Area socioeco-
nomic variations in US cancer incidence, mortality, stage, treat-
ment, and survival, 1975–1999. NCI Cancer Surveillance Monograph
Series, number 4. NIH Publication no. 03–5417. National Cancer
Institute, Bethesda, MD
14. Roche LM, Skinner R, Weinstein RB (2002) Use of a geographic
information system to identify and characterize areas with high
proportions of distant stage breast cancer. J Public Health Manag
Pract 8:26–32
15. Adams J, White M, Forman D (2004) Are there socioeconomic
gradients in stage and grade of breast cancer at diagnosis?
Cross sectional analysis of UK cancer registry data. BMJ 329:
142 [Epub]
16. Greenwald P, Cullen JW, McKenna JW (1987) Cancer prevention
and control, from research through applications. J Natl Cancer Inst
79:389–400
17. Scottish Executive (2003) NHS National targets, related to
NHS national priorities. http,//www.scotland.gov.uk/Publications/
2003/10/18432/28415
18. World Health Organisation, Regional Office for Europe (1985)
Targets for health for all. European Health for all Series, no. 1.
WHO, Copenhagen
19. European Union (European Parliament and Council Decision)
Europe Against Cancer, action plan (1996–2002). Official Journal L
95,16.04.1996 [http://europa.eu/scadplus/leg/en/cha/c11505c.htm]
20. Bray F, Moller B (2006) Predicting the future burden of cancer.
Nat Rev Cancer 6:63–74
21. Møller B, Fekjær H, Hakulinen T, et al. (2002) Prediction
of cancer incidence in the Nordic countries up to the year 2020.
Eur J Cancer Prev 11(Suppl): S1–S96
22. Coory M, Armstrong BK (1998)Cancer Incidence projections for
area and rural health services in New South Wales. NSW Cancer
Council, Sydney
23. Stockton D (2004) Cancer in Scotland, Sustaining change. Cancer
incidence projections for Scotland (2001–2020) An aid to planning
cancer services. NHS Scotland, Edinburgh, http://www.scotland.
gov.uk/Publications/2004/12/20257/46696
110
24. Boyle P, Zaridze DG (1983) Colorectal cancer as a disease of the
environment. Ecol Dis 2:241–248
25. Cummings JH, Bingham SA (1998) Diet and the prevention of
cancer. Br Med. J 317:1636–1640
26. Armstrong B, Doll R (1975) Environmental factors and cancer
incidence and mortality in different countries, with special refer-
ence to dietary practices. Int J Cancer 15:617–631
27. McMichael AJ, McCall MG, Hartshorne JM, et al. (1980) Patterns
of gastrointestinal cancer in European migrants to Australia; the
role of dietary change. Int J Cancer 5:431–437
28. Key TJ, Allen NE, Spencer EA, et al. (2002) The effect of diet on
the risk of cancer. Lancet 360:861–868
29. Yoshimi I, Ohshima A, Ajiki W, et al. (2003) A comparison of
trends in the incidence rate of lung cancer by histological type in the
Osaka Cancer Registry, Japan and in the Surveillance, Epidemiol-
ogy and End Results Program, USA. Jpn J Clin Oncol 33:98–
104
30. Forman D (2004) Review article: oesophago-gastric adenocarci-
noma – an epidemiological perspective. Aliment Pharmacol Ther
20(Suppl 5):55–60
31. Chatterjee N, Hartge P, Cerhan JR, et al. (2004) Risk of non-
Hodgkin’s lymphoma and family history of lymphatic, hemato-
logic, and other cancers. Cancer Epidemiol Biomarkers Prev
13:1415–1421
32. Breslow NE (2005) Case control studies. In: Ahrens W, Pigeot I.
(eds) Handbook of epidemiology. Springer, Berlin Heidelberg
New York Tokyo, pp 287–319
33. Kjaerheim K (1999) Occupational cancer research in the Nordic
countries. Environ Health Perspect 107(Suppl 2):233–238
34. Mbulaiteye SM, Katabira ET, Wabinga H, et al. (2006) Spectrum
of cancers among HIV-infected persons in Africa: the Uganda
AIDS-Cancer Registry Match Study. Int J Cancer 118:985–990
35. Travis LB (2002) Therapy-associated solid tumors. Acta Oncol
41:323–333
36. Hakulinen T, Pukkala E, Kenward M, et al. (1990) Changes in
cancer incidence in North Karelia, an area with a comprehensive
preventive cardiovasular programme. In: Hakama M, Beral V,
Cullen V, Parkin DM (eds) Evaluating effectiveness of primary
prevention of cancer. International Agency for Research on Cancer
Scientific Publications, Lyon, pp 133–148
37. Jemal A, Cokkinides VE, Shafey O, Thun MJ (2003) Lung
cancer trends in young adults: an early indicator of progress in
tobacco control (United States). Cancer Causes Control 14:
579–585
38. Chang MH, Chen CJ, Lai MS, et al. (1997) Universal hepatitis B
vaccination in Taiwan and the incidence of hepatocellular carci-
noma in children. Taiwan Childhood Hepatoma Study Group.
N Engl J Med 336:1855–1859
39. Weir HK, Thun M., Hankey BF, et al. (2003) Annual report to the
nation on the status of cancer 1975–2000, featuring the uses of
surveillance data for cancer prevention and control. J Natl Cancer
Inst 95:1276–1299
40. Barnoya J, Glantz S (2004) Association of the California tobacco
control program with declines in lung cancer incidence. Cancer
Causes Control 15:689–695
41. Sankila R, Démaret E, Hakama M, et al. (eds) (2001) Evaluation
and monitoring of cancer screening programmes. European Com-
mission, Brussels-Luxembourg
42. Fidler HK, Boyes DA, Worth AJ (1968) Cervical cancer detection
in British Columbia. A progress report. J Obstet Gynaecol Br
Commonw 75:392–404
43. Pettersson F, Näslund I, Malker B (1986) Evaluation of the effect
of Papanicolaou screening in Sweden: record linkage between a
control screening registry and the National Cancer Registry. In:
Hakama M, Miller AB, Day NE (eds) Screening for cancer of the
uterine cervix. IARC Scientific Publication no. 76. IARC, Lyon
44. Oshima A, Hanai A, Fujimoto l (1979) Evaluation of a mass
screening program for stomach cancer. NCI Monogr 53:181–186
45. Tsubono Y, Nishino Y, Tsuji I, et al. (2000) Screening for gastric
cancer in Miyagi, Japan: evaluation with a population-based cancer
registry. Asian Pac J Cancer Prev 1:57–60
46. Lee KJ, Inoue M, Otani T, et al. (2006) Gastric cancer screening
and subsequent risk of gastric cancer: a large-scale population-
based cohort study, with a 13-year follow-up in Japan. Int J Cancer
118:2315–2321
47. Hakama M, Pukkala E, Heikkila M, et al. (1997) Effectiveness of
the public health policy for breast cancer screening in Finland:
population based cohort study. BMJ 314:864–867
48. Morrison AS (1985) Screening in chronic disease. Oxford Univer-
sity Press, New York
49. Gordon NP, Hiatt RA, Lampert DI (1993) Concordance of self-
reported data and medical record audit for six cancer screening
procedures. J Natl Cancer Inst 85:566–570
50. Moss SM (1991) Case-control studies of screening. Int J Epidemiol
20:1–6
51. Friedman DR, Dubin N (1991) Case-control evaluation of
breast cancer screening efficacy. Am J Epidemiol 133:974–
984
52. Weiss NS (1998) Analysis of case-control studies of the efficacy of
screening for cancer: How should we deal with tests done in
persons with symptoms? Am J Epidemiol 147:1099–1102
53. Morris JK (2002) Screening for neuroblastoma in children. J Med
Screen. 9:56
54. Day NE, Williams DR, Khaw KT (1989) Breast cancer screening
programmes, the development of a monitoring and evaluation
system. Br J Cancer 59:954–958
55. Schouten LJ, de-Rijke JM, Schlangen JT, et al. (1998) Evaluation
of the effect of breast cancer screening by record linkage with the
cancer registry, the Netherlands. J. Med. Screen 5:37–41
56. McCann J, Stockton D, Day N (1998) Breast cancer in East Anglia,
the impact of the breast screening programme on stage at diagno-
sis. J Med Screen 5:42–48
57. Hakama M, Magnus K, Pettersson F, et al. (1991) Effect of orga-
nized screening on the risk of cervical cancer in the Nordic coun-
tries. In: Miller A, Chamberlain J, Day N, et al. (eds) Cancer
screening. Cambridge University Press, Cambridge, UK, pp
153–162
58. Walter SD, Day NE (1983) Estimation of the duration of a pre-
clinical disease state using screening data. Am J. Epidemiol 118:
865–886
59. Johannesson G, Geirsson G, Day NE (1978) The effect of mass
screening in Iceland 1965–1974, on the incidence and mortality of
cervical cancer. Int J Cancer 21:418–425
60. Hisamichi S, Fukao A, Sugawara N, et al. (1991) Evaluaution of
mass screening programme for stomach cancer in Japan. In: Miller
AB, et al. (eds) Cancer screening. Cambridge University Press,
Cambridge, UK
61. Chu KC, Tarone RE, Chow WH, et al. (1994) Temporal patterns
in colorectal cancer incidence, survival, and mortality from 1950
through 1990. J Natl Cancer Inst 86:997–1006
62. Chu KC, Tarone RE, Kessler LG, et al. (1996) Recent trends in
U.S. breast cancer incidence, survival, and mortality rates. J Natl
Cancer Inst 88:1571–1579
63. Taplin SH, Mandelson MT, Anderman C, et al. (1997) Mammog-
raphy diffusion and trends in late stage breast cancer, evaluating
outcomes in a population. Cancer Epidemiol Biomarkers Prev
6:625–631
64. Stockton D, Davies T, Day N, et al. (1997) Retrospective study of
reasons for improved survival in patients with breast cancer in east
Anglia, earlier diagnosis or better treatment. Br Med J 314:472–
475
65. Brenner H, Soderman B, Hakulinen T (2002) Use of period analy-
sis for providing more up-to-date estimates of long-term survival
rates, empirical evaluation among 370 000 cancer patients in
Finland. Int J Epidemiol 31:456–462
66. Trimble EL, Carter CL, Cain D, et al. (1994) Representation
of older patients in cancer treatment trials. Cancer 74(7 Suppl):
2208–2214
67. Black RJ, Sankaranarayanan R, Parkin DM (1998) Interpretation
of population-based cancer survival data. In: Sankaranarayanan R,
Black RJ, Parkin DM (eds) Cancer survival in developing coun-
tries. IARC Scientific Publication, no. 145, IARC, Lyon, pp
13–17
68. Erridge SC, Møller H, Price A, Brewster D (2007) International
comparisons of survival from lung cancer: pitfalls and warnings.
Nat Clin Pract Oncol 4:570–577
69. Berrino F, De Angelis R, Sant M, et al. (2007) EUROCARE
Working Group. Survival for eight major cancers and all cancers
combined for European adults diagnosed in 1995–99: results of the
EUROCARE-4 study. Lancet Oncol 8:773–783

70. Young JL, Ries LG, Pollack ES (1984) Cancer patient survival
among ethnic groups in the United States, 1973–79. J Natl Cancer
Inst 73:341–352
71. Welch HG, Schwartz LM, Woloshin S (2000) Are increasing 5-year
survival rates evidence of success against cancer? JAMA 283:
2975–2978
72. Peto R, Boreham J, Clarke M, et al. (2000) UK and USA breast
cancer deaths down 25% in year 2000 at ages 20–69 years. Lancet
355:1822–1823
73. Peto R (2000) Author’s reply, Lancet 356:593
74. Clauser SB (2004) Use of cancer performance measures in popula-
tion health, a macro-level perspective. J Natl Cancer Inst Mono-
graphs 33:142–154
75. Sankila R, Black R, Coebergh JWC, et al. (2003) Evaluation of
clinical care by cancer registries. IARC Technical Report no. 37.
IARC, Lyon
76. Edwards BK, Brown ML, Wingo PA, et al. (2005) Annual report
to the nation on the status of cancer, 1975–2002, featuring popula-
tion-based trends in cancer treatment. J Natl Cancer Inst 97:
1407–1427
77. Morris E, Forman D, Haward B (2007) The impact of the Calman-
Hine report: analysis of breast and colorectal cancer surgical work-
loads and the degree of surgical site specialization in the Yorkshire
region of the UK, 1990–2000. Eur J Cancer Care (Engl) 16:
150–155
78. Schrag D, Panageas KS, Riedel E, et al. (2002) Hospital and
surgeon procedure volume as predictors of outcome following
rectal cancer resection. Ann Surg 236:583–592
111
79. Hodgson DC, Zhang W, Zaslavsky AM, et al. (2003) Relation of
hospital volume to colostomy rates and survival for patients with
rectal cancer. J Natl Cancer Inst 95:708–716
80. Parks RW, Bettschart V, Frame S, et al. (2004) Benefits of speciali-
sation in the management of pancreatic cancer: results of a Scottish
population-based study. Br J Cancer 91:459–465
81. Jones AP, Haynes R, Sauerzapf V, et al. (2007) Travel times to
health care and survival from cancers in Northern England. Eur J
Cancer Sep 19; [Epub ahead of print]
82. Menck HR, Cunningham MP, Jessup JM, et al. (1987) The growth
and maturation of the National Cancer Data Base. Cancer 80:
2296–2304
83. Warren JL, Klabunde CN, Schrag D, et al. (2002) Overview of the
SEER-Medicare data: content, research applications, and general-
izability to the United States elderly population. Med Care 40(8
Suppl):IV-3–18
84. Harlan LC, Clegg LX, Trimble EL (2003) Trends in surgery and
chemotherapy for women diagnosed with ovarian cancer in the
United States. J Clin Oncol 21:3488–3494
85. Surveilllance, Epidemiology and End Results Program. National
Cancer Institute. http://seer.cancer.gov/
86. Remontet L, Estève J, Bouvier AM, et al. (2003) Cancer incidence
and mortality in France over the period 1978–2000. Rev Epidemiol
Sante Publique 51:3–30
87. Kamo K, Kaneko S, Satoh K, et al. (2007) A mathematical estima-
tion of true cancer incidence using data from population-based
cancer registries. Jpn J Clin Oncol 37:150–155