Estimating Risk: Is There

an Association?
 Learning objectives
 To revisit the concept of absolute risk.
 To introduce and compare the relative risk and odds ratio
as measures of association
 between an exposure and a disease.
 To calculate and interpret a relative risk in a cohort study.
 To calculate and interpret an odds ratio in a cohort study
and in a case-control study and to describe when the odds
ratio is a good estimate of the relative risk.
 To calculate and interpret an odds ratio in a matched-pairs
case-control study.
  
ABSOLUTE RISK

• The incidence of a disease in a population is

termed the absolute risk (baseline risk).
• Absolute risk can indicate the magnitude of the
risk in a group of people with a certain
exposure, but because it does not take into
consideration the risk of disease in non-
exposed individuals, it does not indicate
whether the exposure is associated with an
increased risk of the disease.
  
ABSOLUTE RISK

• Comparison is fundamental to epidemiology.

Nevertheless, absolute risk may have
important implications in both clinical and
public health policy
 How do we determine whether a certain disease is associated with a
certain exposure?

• To determine whether such an association

exists, we must determine, using data obtained
in case-control and cohort studies, whether
there is an excess risk of the disease in persons
who have been exposed to a certain agent
 How do we determine whether a certain disease is associated with a certain exposure?

• Let us consider the results of a hypothetical

investigation of a foodborne disease outbreak.
• The suspect foods were identified, and for
each food, the attack rate (or incidence rate) of
the disease was calculated for those who ate
the food (exposed) and for those who did not
eat the food (non-exposed).
 How do we determine whether a certain disease is
associated with a certain exposure?

• How can we determine whether an excess risk is

associated with each of the food items?
• One approach, shown in column C of Table 11-2, is to
calculate the ratio of the attack rate in those who ate each
food to the attack rate in those who did not eat the food.
• An alternate approach for identifying any excess risk in
exposed individuals is shown in column D. We can
subtract the risk in those who did not eat the food from
the risk in those who did eat the food. The difference
represents the excess risk in those who were exposed.
 How do we determine whether a certain disease is
associated with a certain exposure?

• Thus, as seen in this foodborne outbreak, to

determine whether a certain exposure is
associated with a certain disease, we must
determine whether there is an excess risk of
disease in exposed populations by comparing
the risk of disease in exposed populations to
the risk of disease in nonexposed populations.
We have just seen that such an excess risk can
be calculated in the two following ways:
How do we determine whether a certain disease is
associated with a certain exposure?
How do we determine whether a certain disease is
associated with a certain exposure?
 Excess risk
• Does the method that we choose to calculate
excess risk make any difference? Let us
consider a hypothetical example of two
communities, A and B, seen in Table 11-3.
 Excess risk
• In community A, the incidence of a disease in
exposed persons is 40% and the incidence in
nonexposed persons is 10%. Is there an excess
risk associated with exposure?
• As in the food poisoning example, we can
calculate the ratio of the rates or the difference
between the rates. The ratio of the incidence
rates is 4.0. If we calculate the difference in
incidence rates, it is 30%.
 Excess risk
• In community B, the incidence in exposed
persons is 90% and the incidence in non-
exposed persons is 60%.
• If we calculate the ratio of the incidence of
exposed to non-exposed persons in population
B, it is 90/60, or 1.5. If we calculate the
difference in the incidence in exposed and
nonexposed persons in community B it is,
again, 30%.
 Excess risk
• What do these two measures tell us? Is there a
difference in what we learn from the ratio of
the incidence rates compared to the difference
in the incidence rates?
• This question is the theme of this chapter and
of Chapter 12.
 RELATIVE RISK

The Concept of Relative Risk

• Both case-control and cohort studies are
designed to determine whether there is an
association between exposure to a factor and
development of a disease.
• If an association exists, how strong is it?
 RELATIVE RISK

• If we carry out a cohort study, we can put the

question another way: “What is the ratio of the
risk of disease in exposed individuals to the
risk of disease in nonexposed individuals?”
This ratio is called the relative risk:
 RELATIVE RISK

• The relative risk can also be defined as the

probability of an event (developing a disease)
occurring in exposed people compared to the
probability of the event in non-exposed
people, or as the ratio of the two probabilities.
RELATIVE RISK
 RELATIVE RISK

• Interpreting the Relative Risk: How do we interpret

the value of a relative risk? These three possibilities
are summarized in Table 11-4.
 RELATIVE RISK

• Calculating the Relative Risk in Cohort

Studies: In a cohort study, the relative risk can
be calculated directly. Recall the design of a
cohort study seen in Table 11-5
RELATIVE RISK
 RELATIVE RISK
• Table 11-6 shows a hypothetical cohort study
of 3,000 smokers and 5,000 nonsmokers to
investigate the relation of smoking to the
development of coronary heart disease (CHD)
over a 1-year period.
RELATIVE RISK
RELATIVE RISK
 The odds ratio (relative odds)

• We have seen that in order to calculate a relative

risk, we must have values for the incidence of the
disease in the exposed and the incidence in the non-
exposed, as can be obtained from a cohort study.
• In a case-control study, however, we do not know
the incidence in the exposed population or the
incidence in the non-exposed population because
we start with diseased people (cases) and non-
diseased people (controls).
 The odds ratio (relative odds)
• Hence, in a case-control study we cannot calculate
the relative risk directly. In this section we shall see
how another measure of association, the odds ratio,
can be obtained from either a cohort or a case-control
study and can be used instead of the relative risk.
• We will also see that even though we cannot calculate
a relative risk from a case-control study, under many
conditions, we can obtain a very good estimate of the
relative risk from a case-control study using the odds
ratio.
The odds ratio (relative odds)
 Defining the Odds Ratio in Cohort and in Case-Control Studies

• In previous chapters we discussed the

proportion of the exposed population in whom
disease develops and the proportion of the non-
exposed population in whom disease develops
in a cohort study.
• Similarly, in case-control studies, we have
discussed the proportion of the cases who were
exposed and the proportion of the controls who
were exposed (Table 11-8).
 Defining the Odds Ratio in Cohort and in
Case-Control Studies
• An alternate approach is to use the concept of odds.
Suppose we are betting on a horse named Epi Beauty,
which has a 60% probability of winning the race (P).
• Epi Beauty therefore has a 40% probability of losing (1 −
P). If these are the probabilities, what are the odds that the
horse will win the race?
• To answer this we must keep in mind that the odds of an
event can be defined as the ratio of the number of ways the
event can occur to the number of ways the event cannot
occur. Consequently, the odds of Epi Beauty winning, as
defined above, are as follows:
Defining the Odds Ratio in Cohort and in
Case-Control Studies
Defining the Odds Ratio in Cohort and in
Case-Control Studies
Defining the Odds Ratio in Cohort and in
Case-Control Studies
 The Difference Between "Probability"
and "Odds"?
• The probability that an event will occur is the
fraction of times you expect to see that event
in many trials. Probabilities always range
between 0 and 1.
• The odds are defined as the probability that the
event will occur divided by the probability that
the event will not occur.
 The Difference Between "Probability" and "Odds"?

• If the probability of an event occurring is Y, then the

probability of the event not occurring is 1-Y.
(Example: If the probability of an event is 0.80
(80%), then the probability that the event will not
occur is 1-0.80 = 0.20, or 20%.
•  The odds of an event represent the ratio of the
(probability that the event will occur) / (probability
that the event will not occur). This could be
expressed as follows:
• Odds of event = Y / (1-Y)
A, Odds ratio (OR) in a cohort study. B, Odds
ratio (OR) in a case-control study.
 C, Cross-products ratio in both
a cohort and a case-control study
 The Odds Ratio in a Case-Control Study

• As just discussed, in a case-control study, we

cannot calculate the relative risk directly to
determine whether there is an association
between the exposure and the disease.
• This is because, having started with cases and
controls rather than with exposed and non-
exposed persons, we do not have information
about the incidence of disease in exposed
versus non-exposed persons.
 The Odds Ratio in a Case-Control Study

• However, we can use the odds ratio as a measure of

the association between exposure and disease in a
case-control study, but we ask different questions:
“What are the odds that a case was exposed?”
• Looking at the lefthand column in Figure 11-5B, we
see that the odds of a case having been exposed are
a:c or a/c Next, we ask, “What are the odds that a
control was exposed?” Looking at the right-hand
column, we see that the odds of a control having been
exposed are b:d or b/d.
 The Odds Ratio in a Case-Control Study

• We can then calculate the odds ratio, which in a case-

control study, is defined as the ratio of the odds that
the cases were exposed to the odds that the controls
were exposed. This is calculated as follows:

• Thus, interestingly, ad / bc represents the odds ratio

 When Is the Odds Ratio a Good
Estimate of the Relative Risk?

• In a case-control study, only the odds ratio can be

calculated as a measure of association, whereas in a cohort
study, either the relative risk or the odds ratio is a valid
measure of association.
• However, many people are more comfortable using the
relative risk, and this is the most frequently used measure
of association reported in the literature when results of
cohort studies are published. Even when the odds ratio is
used, people are often interested in knowing how well it
approximates the relative risk. Even prestigious clinical
journals have
 When Is the Odds Ratio a Good
Estimate of the Relative Risk?
• When is the odds ratio (relative odds) obtained in a case-
control study a good approximation of the relative risk in the
population? When the following three conditions are met:
1. When the cases studied are representative, with regard to
history of exposure, of all people with the disease in the
population from which the cases were drawn.
2. When the controls studied are representative, with regard to
history of exposure, of all people without the disease in the
population from which the cases were drawn.
3. When the disease being studied does not occur frequently.
 When Is the Odds Ratio a Good
Estimate of the Relative Risk?
• The third condition (that the disease
occurrence is not frequent) can be intuitively
explained as follows:
When the Odds Ratio is Good
Estimate of the Relative Risk?
When the Odds Ratio is not Good
Estimate of the Relative Risk?
 Calculating the Odds Ratio in an
Unmatched Case-Control Study
• Let us assume that this case-control study is
done without any matching of controls to
cases, and that we obtain the results seen in
Figure 11-8. Thus, 6 of the 10 cases were
exposed and 3 of the 10 controls were
exposed. If we arrange these data in a 2 × 2
table, we obtain the following:
Calculating the Odds Ratio in an
Unmatched Case-Control Study
Calculating the Odds Ratio in an
Unmatched Case-Control Study
 Reading Assignment

Calculating the Odds Ratio in a Matched-

Pairs Case-Control Study
More on Risk: Estimating the Potential for
Prevention
• How much of the disease that occurs can be attributed
to a certain exposure?
• The attributable risk, which is defined as the amount or
proportion of disease incidence (or disease risk) that
can be attributed to a specific exposure.
• For example, how much of the lung cancer risk
experienced by smokers can be attributed to smoking?
Whereas the relative risk is important in establishing
etiologic relationships, the attributable risk is in many
ways more important in clinical practice and in public
health, because it addresses a different question:
The attributable risk,
The attributable risk,
The attributable risk,
The attributable risk and relative risk

• The relative risk is a measure of the strength of

the association and the possibility of a causal
relationship, but the attributable risk indicates
the potential for prevention if the exposure
could be eliminated.
The concept of attributable risk.
 Attributable Risk for the Total Population—
Population Attributable Risk (PAR)
• We need to calculate the attributable risk for the
total population: What proportion of the disease
incidence in a total population (including both
exposed and nonexposed people) can be
attributed to a specific exposure?
 Attributable Risk for the Total Population—
Population Attributable Risk (PAR)
• What would be the total impact of a prevention program
on the community?
• If we want to calculate the attributable risk in the total
population, the calculation is similar to that for exposed
people, but we begin with the incidence in the total
population and again subtract the background risk, or
the incidence in the non-exposed population.
• The incidence in the total population that is due to the
exposure can be calculated as shown in Formula 12-3:
Attributable Risk for the Total Population—
Population Attributable Risk (PAR)
 Attributable Risk for the Total Population—
Population Attributable Risk (PAR)
• The question addressed is: What proportion of lung
cancer in the total population can be attributed to
smoking?
• This question could be reworded as follows: If
smoking were eliminated, what proportion of the
incidence of lung cancer in the total population (which
consists of both smokers and nonsmokers) would be
prevented? The answer is: the attributable risk in the
total population also called the population
attributable risk—or the PAR (as discussed earlier).
An Example of an Attributable Risk
Calculation for the Exposed Group
An Example of an Attributable Risk
Calculation for the Exposed Group
An Example of an Attributable Risk
Calculation for the Exposed Group
An Example of an Attributable Risk
Calculation for the Exposed Group
From Association to Causation: Deriving
Inferences from Epidemiologic Studies
APPROACHES FOR STUDYING
DISEASE ETIOLOGY
APPROACHES FOR STUDYING
DISEASE ETIOLOGY
TYPES OF CAUSAL RELATIONSHIPS
• A causal pathway can be either direct or in-
direct (Fig. 14-11).
• In direct causation, a factor directly causes a
disease without any intermediate step.
• In indirect causation, a factor causes a disease,
but only through an intermediate step or steps.
In human biology, intermediate steps are
virtually always present in any causal process.
TYPES OF CAUSAL RELATIONSHIPS
 Causal relationships
If a relationship is causal, four types of causal
relationships are possible:
(1) Necessary and sufficient;
(2) Necessary, but not sufficient;
(3) Sufficient, but not necessary; and
(4) Neither sufficient nor necessary.
 Necessary and sufficient
• In the first type of causal relationship, a factor
is both necessary and sufficient for producing
the disease. Without that factor, the disease
never develops (the factor is necessary), and in
the presence of that factor, the disease always
develops (the factor is sufficient) (Fig. 14-12).
Necessary and sufficient
 Necessary, But Not Sufficient
• In another model, each factor is necessary, but
not, in itself, sufficient to cause the disease
(Fig. 14-13). Thus, multiple factors are
required, often in a specific temporal
sequence. For example,
 Necessary, But Not Sufficient
• Carcinogenesis is considered to be a multistage
process involving both initiation and promotion.
For cancer to result, a promoter must act after an
initiator has acted. Action of an initiator or a
promoter alone will not produce a cancer.
• Again, in tuberculosis, the tubercle bacillus is
clearly a necessary factor, even though its
presence may not be sufficient to produce the
disease in every infected individual.
Necessary, But Not Sufficient
 Sufficient, But Not Necessary
• In this model, the factor alone can produce the disease,
but so can other factors that are acting alone (Fig. 14-14).
• Thus, either radiation exposure or benzene exposure can
each produce leukemia without the presence of the other.
• Even in this situation, however, cancer does not develop
in everyone who has experienced radiation or benzene
exposure, so although both factors are not needed, other
cofactors probably are. Thus, the criterion of sufficient is
rarely met by a single factor.
Sufficient, But Not Necessary
 Neither Sufficient Nor Necessary
• In the fourth model, a factor, by itself, is
neither sufficient nor necessary to produce
disease (Fig. 14-15). This is a more complex
model, which probably most accurately
represents the causal relationships that operate
in most chronic diseases.
Neither Sufficient Nor Necessary
EVIDENCE FOR A CAUSAL RELATIONSHIP
Guideline for judging whether association
is causal
 GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL

1. Temporal Relationship. It is clear that if a

factor is believed to be the cause of a disease,
exposure to the factor must have occurred before
the disease developed.
GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL
GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL
GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL
GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL
GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL
 GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL

7. Cessation of Exposure. If a factor is a cause

of a disease, we would expect the risk of the
disease to decline when exposure to the factor is
reduced or eliminated.
GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL
GUIDELINES FOR JUDGING WHETHER AN
OBSERVED ASSOCIATION IS CAUSAL
 Reference
• Gordis, L. (2014). Epidemiology (Fifth
edition.). Philadelphia, PA: Elsevier Saunders.