Causation
HRISTINA VLAJINAC
Institute of Epidemiology, School of Medicine,
University of Belgrade, Belgrade, Serbia
kristiv@eunet.yu
Synonyms
Causality
Definition
A cause of a disease can be defined as an event, condition,
or characteristic that plays an essential role in producing
an occurrence of the disease (Rothman 1986).
The most important aim of epidemiology is to identify
the causes and the risk factors of a disease, and to
improve public health by reducing or eliminating exposure
to these factors.
Basic Characteristics
Epidemiological – EcologicalModels
A disease is a result of the interaction of host factors
and the environment. Several models have been developed
in order to depict the ways in which these interactions
influence the occurrence of the disease: the triangle,
the wheel, and the web of causation.
The triangle – Thismodel consists of three components,
agent, host, and environment, which are in a kind of
dynamic equilibrium. Change in any of these components
will alter an existing equilibrium and increase or
decrease the frequency of the disease. This model has
been the most frequently applied to infectious diseases
in which infectious organismswere separated from other
environmental factors and identified as agents.
For diseases that have not been linked to specific agents,
two other models have been developed.
The Wheel – In this model, the host, with its genetic
make-up as its core, is presented as the hub of the
wheel, surrounded with the environment, which is separated
into biological, social and physical components.
This separation is artificial since these three parts of the
environment are closely interrelated with one another
and with host factors.
The Web – The web model emphasizes the concept that
effects never depend on single causes but develop as the
result of causal chains, which make the web. Each link
in the web is the result of antecedents, and breaking
of the web at any level can prevent occurrence of the
disease. This means that full knowledge of etiology is
not needed for effective disease prevention and control.
(Mausner, Kramer 1985; Bhopal 2002)
Search for Causal Relationship
There are two approaches to testing hypotheses about
causes of disease, experimental and observational.
Experimental study can establish the causal relationship
of a factor with a disease more conclusively, but
since experiments in epidemiology are performed on
humans, for ethical reasons the effects of some possible
causal factors cannot be investigated directly.
Observational studies have therefore been providing the
major contribution to the understanding of many diseases.
The first step in an investigation of causal relationship
is to see whether there is an association between
a disease and a postulated causal factor. If an association
exists, it does not necessarily mean that it is
a causal one. It can be: a) spurious ( spurious association),
b) secondary ( secondary association), or c)
causal. Before an association is assessed for the possibility
that it is causal, other alternative explanations,
such as chance, selection bias, information bias, and
confounding, have to be excluded.
A Concept of Necessary and Sufficient Cause
“A causal factor whose presence is required for the
occurrence of the effect” (Last 2001), that is, without
which the disease never develops, is the necessary
cause. Sufficient cause is a “minimum set of conditions,
factors or events needed to produce a given outcome;
minimal implies that none of the conditions or events is
superfluous” (Rothman 1986). A sufficient cause is not
usually a single factor, but often comprises several components
– component causes or contributing causes.
A disease can have several sufficient causes and these
may have one or more contributing causes in common.
Types of Causal Relationship A causal factor can be
either necessary or sufficient, both, or neither:
1. Necessary and sufficient
A factor can be both necessary and sufficient, which
means that the disease never develops without that factor,
and that factor always produces the development
of the disease. This type of causal relationship occurs
rarely. For example, a person who has three copies of
chromosome 21 instead of two will inevitably be mentally
retarded – Down’s syndrome (Bhopal 2002).
2. Necessary but not sufficient
The factor, although necessary, cannot produce the disease
without the presence of some other factors, called
component or contributory causes. For clinically manifest
tuberculosis, in addition to the bacillus, which is
the necessary cause, contributing causes such as poor
nutritional and socio-economic conditions are needed.
3. Sufficient but not necessary
Although sufficient for producing the disease, the factor
(usually more than one) is not necessary because there
are some other factors that can also produce the disease.
Either radiation exposure or benzene exposure can
produce leukemia independently of each other (Gordis
2004).
4. Neither sufficient nor necessary
Smoking is a cause of lung cancer but not everyonewho
smokes develops this type of cancer and not everyone
who develops lung cancer has smoked.
Guidelines for Causal Reasoning in Epidemiology
Although epidemiologic evidence by itself is insufficient
to establish causality, Bradford Hill (Hill, 1965)
suggested that the following attributes (criteria) of an
association be considered in assessment of the possibility
that it is a causal one.
1. Strength of the association – The strength of association
is measured by the relative risk (odds ratio), that is
the ratio of disease rates for those exposed and those not
exposed to the hypothesized causal factor. The stronger
the association, the more likely it is that the relation is
causal. However, it does not mean that a weak association
cannot be judged to be a causal one. “The strength
of an association is not a biologically consistent feature,
but rather a characteristic that depends on the relative
prevalence of other causes” (Rothman, 1986).
2. Dose-response relationship – A dose-response is
established when, with increasing level of exposure
(“dose” or duration), the risk of disease also increases.
The absence of a dose-response relationship does not
rule out the possibility of causal association since, for
some causes, a threshold may exist and a disease may
not develop unless a certain level of exposure is present.
3. Consistency of the association – A cause-effect relationship
is supported when similar results are obtained
in a number of studies performed in various populations
or population groups, by different investigators,
and with different methodology. The causal relationship
might not be found in some studies because “the
effect of a causal agent cannot occur unless the component
causes act, or have already acted, to complete
a sufficient cause” (Rothman 1986).
4. Temporality – Exposure to the postulated causal factor
must precede the onset of disease by a period of
time consistent with the proposed biologic mechanism
( induction, incubation, latency). In some diseases,
especially chronic and those with a long period
of latency, temporality cannot always be easy to establish.
Although the only indispensable attribute among
all Hill’s conditions, a temporally correct association
between two events does not necessarily mean that it is
that of cause and effect. They could both be generated
by the same factor.
5. Biologic plausibility – The existence of a causeeffect
relationship is enhanced if it is coherent with the
current body of biologic knowledge. This, of course,
depends on the state of scientific information at a given
time. An association that is biologically implausible at
one time may eventually prove to be plausible.
6. Experimental evidence – Causal understanding can
be greatly advanced by “in-vivo” and animal experiments,
but data obtained in that way must be integrated
with observations in the human population. Because of
ethical reasons, experimental evidence is seldom from
the human population. However, evidence for a causal
relationship is supported if reduction or elimination of
exposure to a certain factor (postulated causal factor) is
related to decline of disease frequency.
7. Coherence – Coherence implies that a cause-effect
interpretation of an association does not conflict with
the generally known facts of the natural history and
biology of the disease.
8. Specificity of the association – An association is
specific when a certain exposure is associated with
only one disease. Taking into account the multifactorial
nature of disease and the fact that one factor can
cause more than one disease, the specificity is the least
important criterion to satisfy, and “should be probably
deleted from the list” (Gordis 2004).
In making decisions about causation, the list of criteria
presented above should be considered only as guidelines.
If temporality is viewed as part of the definition
of causation, “there is no necessary or sufficient criterion
for determining whether an observed association is
causal” (Rothman, Greenland 1998; Rothman, Greenland
2005). Decisions about causation must always
remain a matter of judgment based on all available evidence
“achievable through hypothesis generation and
testing, with data interpreted using a logical framework
of analysis, which draws on multidisciplinary perspectives”
(Bhopal 2002).
Hill himself pointed out that these “viewpoints” cannot
be used as criteria for causal inference, but can help
to make a judgment, and to act on the premise that
a causal relationship exists rather than awaiting further
evidence: “All scientific work is incomplete – whether it
be observational or experimental. All scientific work is
liable to be upset or modified by advancing knowledge.
That does not confer upon us a freedom to ignore the
knowledge we already have, or to postpone the action
that it appears to demand at a given time”.
Most definitions are taken from the last edition of Last’s
Dictionary of Epidemiology (Last, 2001).We are much
obliged to Professor Last for his kind consent.
Cross-References
Agent (of Disease)
Association
Incubation
Induction
Interaction
Latency
Risk Factor
Secondary Association
Spurious Association
References
Bhopal R (2002) Concepts of Epidemiology: An integrated introduction
to the ideas, theories, principles and methods of epidemiology.
Oxford University Press, Oxford
Gordis L (2004) Epidemiology. Elsevier Saunders, Philadelphia
Hill B (1965) The environment and disease: Association or causation.
Proc Roy Soc Med 58:295–300
Last J (2001) A Dictionary of Epidemiology, 4th edn. Oxford
University Press, New York
Mausner J, Kramer S (1985) Epidemiology. WB Saunders,
Philadelphia
Rothman K (1986) Modern Epidemiology. Little Brown, Boston
Rothman K, Greenland S (1998) Modern Epidemiology, 2nd edn.
Lippincott – Raven Publishers, Philadelphia
Rothman K, Greenland S (2005) Basic Concept. In: Ahrens W,
Pigeot I (ed) Handbook of Epidemiology. Springer, Berlin,
pp 45–
Tags: antecedents, belgrade serbia, causal chains, dynamic equilibrium, environmental factors, epidemiological, host factors, hristina, infectious diseases, medicine university, occurrence, physical components, Public Health, risk factors, rothman, school of medicine, synonyms, triangle, web model, web of causation