Disease cause and causal inference

Headlines
 Concept of cause in epidemiology
 Association and cause association

Measures of association and types of associ

ation
Spurious association
Noncausal association
Causal association
 Evaluation causal association

Hill's criteria
 Summary
 Discover the causes of a disease

 Analytical studies
 Described cohort studies
 Case-control studies
 Experimental studies
 Concept of causes

 A causal relationship or causes of di

sease may indicate that: the factors t
hat increases the probability of the oc
curence of one or more of these factor
s decreases the frequency of that dise
ase.
 Association (relationship): statistical dependence between two or
more events, characteristics or other variables. Positive association
implies a direct relationship, while negative association implies an
inverse one. The presence of a statistical association alone does not
necessarily imply a causal relationship.
 Causality (causation / cause-effect relationship): relating causes to
the effects they produce.
 Causes may be “genetic” and / or
“environmental” (e.g. many NCDs
including: diabetes, cancers, COPD, etc)
 Concept of causes
 Notice: The concept here is different fr
om that in philosophy. In public healt
h practice or in prevention, it is neces
sary to identify an exposure without n
ecessary identifying the ultimate caus
e of the disease. Epidemiology freque
ntly provides a basis for action despit
e ignorance of mechanism.
 Concept of causes

 Example 1
Cigarette smoke has been identified as t
he vehicle associated with increased rat
es of lung cancer and other cancers, and
heart respiratory disease. It's not necess
ary to identify precisely which compone
nt in the smoke is the prime offender bef
ore instituting preventive measures.
 Concept of causes
 Example 2
In John Snow's era, when the people did not k
now the causative agent of cholera. It was nec
essary to know that polluted water was a majo
r vector of cholera. Lack of knowledge that a s
pecific bacterium is the causative agent did no
t prevent authorities from introducing legislati
on mandating that all water companies in Lon
don filter their water, thus greatly controlling t
he disease.
 Concept of causes

 In both of these example, reducing the e

xposure to the cause, whether cigarette
smoke or polluted water, helps solve the
disease problem in the population.
Henle-Koch's postulates (1877,1882)

Koch stated that four postulates should be met before a causal

relationship can be accepted between a particular bacterial parasite
(or disease agent) and the disease in question. These are:
1. The agent must be shown to be present in every case of the
disease by isolation in pure culture.
2. The agent must not be found in cases of other disease.
3. Once isolated, the agent must be capable of reproducing
the disease in experimental animals.
4. The agent must be recovered from the experimental
disease produced.
 Henle-Koch's postulates (1877,1882)

Koch's postulates contributed greatly to understa

nding the concept of cause in medicine. The appli
cation of these postulates helped people identify
the relationship between an agent and disease. T
he logical model is helpful even today when looki
ng at some infectious diseases, such as Legionnai
re's disease and AIDs.
 Henle-Koch's postulates (1877,1882)

However, for most of the diseases, especially non-com

municable chronic diseases, cause or causes cannot b
e esdablished simply by Koch's criteria. Few diseases
are so simple that there is a single cause. A given disea
se can be caused by more than one causal mechanism,
and every causal mechanism involves the joint action
of a multitude of component causes.
 Henle-Koch's postulates (1877,1882)

For instance, smoking might be the cause of lung canc

er, coronary heart disease as well as other diseases. (O
ne cause, multiple diseases)
Heart disease is associated with multiple factors or cau
ses.(Multiple diseases, one cause)
 Levels / Types of causality

 Molecular / Physiological
 Personal / Social
 Deterministic / probabilistic
 What aspect of “environment” (broadly defined) if
removed / reduced / controlled would reduce
outcome / burden of disease
 Definitions

 Necessary cause: The cause must be present

for the outcome to happen. However, the c
ause can be present without the outcome h
appening.
Deterministic causality (I)
Deterministic causality (II)
 Sufficient cause: If the cause is present the o
utcome must occur. However, the outcome
can occur without the cause being present.
Deterministic causality (III)
Deterministic causality (IV)
 Deterministic causality: cause closely related to
effect, as in “necessary” / “sufficient” causes
Deterministic causality (V)
Deterministic causality (VI)
 Probabilistic Causality: in epidemiology, most
associations are rather “weak” (e.g. relationship
between high serum cholesterol and IHD), which is
neither necessary nor sufficient
 Multiple causes result in what is known as “web of
causation”or “chain of causation”
which is very common for noncommunicable /
chronic diseases
 Component causes: together they constitute a
sufficient cause for the outcome in question. In
CDs, this may include the biological agent as well
as environmental conditions (e.g. TB, measles,
ARF/RHD). In NCDs, this may include a whole
range of genetic, environmental as well as
personal / psychosocial / behavioral characteristics
(e.g. diabetes, cancers, IHD)
 Causal inference is an intelligent way of ap
plying common sense and judgement to co
nnect an exposure with a disease and to inf
er that the exposure is likely a cause of the
disease.
The process of causal inference is usually increme
ntal. Much of the data used in interpreting relation
ships comes from observational studies and must
be interpreted with careful considerations of repre
sentativeness of the data, potantial biases, the role
of effect modifiers and unmeasured confounders.
Disease in human population usually not caused b
y a single exposure and the causes of a particular e
vent may differ in different circumstances. Someti
mes, there are disagreements on what constitutes
disease. Sometimes there are disagreements abou
t whether a strong relationship is causal.
The approach that we take is to evaluate the evide
nce at hand and to take the best assessment of the
evidence. The answers often are in the form of pro
bability statements. Our conclusions are most ofte
n based on increased incidence of disease in a gro
up with certain characteristics or exposures and w
e label those exposures as causal if it makes sense
biologically.
On the other hand, if a group has significantly less
disease, then we may label the exposures or chara
cteristics as protective.
Notice: the increased incidence in a group does no
t necessarily explain the outcome for any individua
l in the group.
For example, consider air pollution. In developing
public health programs, knowing that exposure to
a particular pollutant increases the probability of d
eveloping a disease is important. Unless the pollut
ant is absolutely caustic, the information does not
necessarily tell us the outcome for a particular pati
ent.
 Definitions
 Deduction: reasoned argument proceeding from the
general to the particular.
 Induction: any method of logical analysis that proceeds
from the particular to the general. Conceptually bright
ideas, breakthroughs and ordinary statistical inference
belong to the realm of induction.
 Induction period: the period required for a specific cause
to produce the disease (health-related outcome). Usually
longer with NCDs
 Effect Measures /
Impact Fractions
 Effect measures (e.g. odds ratio, risk ratio) and
impact fractions (e.g. population attributable risk)
are closely related to the strength of association
 The higher effect measures (away from unity) and
population attributable risk (closer to 100 %) the
more the exposure is predictive of the outcome in
question
 E.g. PAR of 100 % means that a factor is
“necessary”
However, if statistical significance exists, e.g. p<0.0
5, we can not simply make an inference that the ex
posure is the cause of the outcome.
First, you need to consider the size of the sampling error.
What does p<0.05 mean? Assume p=0.05. That means for
this study the probability of that result of the study is
reliable is 95% and 5% of probability of that is caused by
sampling error, or by chance. When making this
calculation, the precondition is that the design, conduct of
the study and analyses for thus study are perfect. Almost
no study is perfect in either design or conduct.
If the probablility of sampling error is very limited,
then we may put more consideration to the associ
ation.
 Definitions
 Spurious association (false association): The association is
not true. The association may result from various biases. One
possibility is that there a random error introduced into the fi
ndings and it is completely due to sampling probability, Anot
her is that there is a systematic error (nonrandom). A non-ran
dom error is called a bias. It may occur in the subjects selecti
on for the study or in the data collection during the impleme
ntation of the study.
 Definitions
 Noncausal association: the association between the exposu
re of interested and the outcome really exists, but it is not ca
usal. Change in exposure does not result in a change in outco
me. In these conditions, the association is usually due to con
founding factors.
 Definitions
 Predisposing factors: factors that prepare, sensitize, condition or
otherwise create a situation (such as level of immunity or state of
susceptibility) so that the host tends to react in a specific fashion to a
disease agent, personal interaction, environmental stimulus or specific
incentive. Examples: age, sex, marital status, family size, education,
etc. (necessary, rarely sufficient).
 Precipitating factors: those associated with the definitive onset of a
disease, illness, accident, behavioral response, or course of action.
Examples: exposure to specific disease, amount or level of an
infectious agent, drug, physical trauma, personal interaction,
occupational stimulus, etc. (usually necessary).
 Weighing Evidence
 At individual level: clinical judgment (which
management scheme)
 At population level: epidemiological judgment
(which intervention)
 When weighing evidence from epidemiological
studies, we use “causal criteria” (usually applied to a
group of articles, to deal with confounding) e.g.
Hill’s / Susser’s criteria, which were preceded by
Koch’s postulates (on infectious diseases)
 Hill's Criteria (1897 - 1991)
The first complete statement of the epidemiologic criteria of a
causality is attributed to Austin Hill (1897 - 1991). They are:
 Consistency (Temporal )
 Strength (of association)
 Specificity
 Dose response relationship
 Temporal relationship (directionality)
 Biological plausibility (evidence)
 Coherence
 Experiment
 Consistency

 Does the exposure always precede the outcome. If fac

tor A is believed to cause a disease, then factor A must
necessaily always precede the occurence of the disea
se.
 Among all of the criteria for judgment of the causal rel
ationship, this one is essential when time of exposure
and occurence of outcome can be determined. This is
why the result of a cohort study are relatively more po
werful than those from a cross-sectional study.
Consistency (I)
 Consistency (II)
 Meta-analysis is an good method for testing
consistency. It summarizes odds ratios from
various studies, excludes bias
 Consistency could either mean:
 Exact replication (as in lab sciences, impossible in
epidemiological studies)
 Replication under similar circumstances (possible)
Strength of Association
 Expressions of Strength of Association
 Quantitatively:
 Effect measure (OR, RR): away from unity (the
higher, the stronger the association)
 P-value (at 95% confidence level): less than 0.05
(the smaller, the stronger the association)
 Qualitatively:
 Accept alternative hypothesis: an association
between the studied exposure and outcome exists
 Reject null hypothesis: no association exists
Dose-response relationship (I)
Dose-response relationship (II)
Time-order (temporality, directionality)
Time order
Specificity of Outcome
Specificity of Exposure
 Coherence
 Theoretical: compatible with pre-existing theory
 Factual: compatible with pre-existing knowledge
 Biological: compatible with current biological
knowledge from other species or other levels of
organization
 Statistical: compatible with a reasonable statistical
model (e.g. dose-response)
Biological Coherence (I)
Biological Coherence (II)
 Analogy

 If an exposure similar to A causes an outcome

similar to B, then this is evidence supporting t
hat A causes B. Sometimes a commonly accep
ted phenomenon in one area can be applied t
o another area.
 Susser's criteria (I)
 Mervyn Susser (1988) used similar criteria
to judge causal relationships.
 In agreement with previous authors, he
mentioned that two criteria have to be
present for any association that has a claim
to be causal: i.e. time order (X precedes Y);
and direction (X leads to Y).
 Susser’s Criteria (II)
 Rejection of a hypothesis can accomplished with
confidence by only three criteria: time order,
consistency, factual incompatibility or incoherence.
 Acceptance or affirmation can be achieved by only
four, namely: strength, consistency, predictive
performance, and statistical coherence in the form of
regular exposure/effect relation.
Comparison of Causal Criteria
 References
1. Porta M. A dictionary of epidemiology. New York, Oxford:
Oxford University Press, 2008.
2. Rothman KJ (editor). Causal inference. Chestnut Hill:
Epidemiology Resources Inc., 1988.
3. Hill AB. The environment and disease: Association or causation.
Proceedings of the Royal Society of Medicine 1965; 58:
295-300.
4. Susser MW. What is a cause and how do we know one ? A
grammar for pragmatic epidemiology. American Journal of
Epidemiology 1991; 133: 635- 648.
5. Paneth N. Causal inference. Michigan State University.
6. Rothman J, Greenland S. Modern epidemiology. Second edition.
Lippincott - Raven Publishers, 1998.
Thank you for your kind attention