Estimating Risk

Dr. Tarek Tawfik Amin

Public Health Department, Cairo University

[email protected]

9/30/2017 Dr. Tarek Tawfik 1

 Objectives
By the end of this session, attendees should
be able to:
1- Differentiate between probability, Odd, and
risk with it variants.
2- Recall the basics for calculating Odds and
relative risk and interpret the results.
3- Define the indications of applying the risk
parameters.
 Definitions of terms: Probability, risk and Odd
Probability: Is the proportion (%) of times an event
would occur if an observation was repeated many times.

Risk: Is the probability of an event among those

experiencing the event divided by the number who
could experience it (at risk).

Odds: Probability of an event divided by the

probability of the event not happening.
 Probability Odds (chance) Risk

Proportion (%) Ratio Ratio (rate)

0-1 (100%) Probability/(1- 0 to infinity
probaility)
0 to infinity
The 10 year The Odds for OP Incidence of OP hip
probability of OP hip fracture is fracture among
hip fracture among 0.23/(1.0.23)=0.30 those aged 70 years.
those aged 70 years
is 0.23 (23%).
No denominator Denominator

Attack rate is an
other example
Risk
 A-Absolute Risk
The incidence of a disease in a population is termed absolute risk.
Can indicate the magnitude of risk in a group of people with a
certain exposure, but:
It does not take into consideration the risk of disease in the non-
exposed individuals,
It does not indicate whether the exposure is associated with an
increased risk of disease.
Absolute risk doe not stipulate an explicit comparison.
Rubella in 1st trimester: what is the risk that my child will be
malformed? Abortion will be decided on the basis of this
information.

9/30/2017 Dr. Tarek Tawfik 6

B-Relative risk: Determination that a certain disease is
associated with a certain exposure.
By using the (case-control) and cohort studies we can assess
whether there is an excess risk of disease in persons who have
been exposed.

Comparing different risks among different groups to assess the

presence of excessive risk (the incidence rate attack rates and
the difference in the risks).

Estimation of relative risks are vital in determining who will be

at a higher risk following the exposure.

9/30/2017 Dr. Tarek Tawfik 7

 Relative Risk (concept)
o 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?
o In cohort study: what is the ratio of the risk of disease
in exposed individuals to the risk of disease in non-
exposed individuals? (RR=relative risk).

Risk in exposed (incidence in the exposed)

Relative risk (RR) =
Risk in non-exposed (incidence in the non-exposed)
 Basic Structure of cohort study
Population

Comparing the incidence of disease in each group

Diseased
Disease-free The Relative Risk is calculated for exposure

Develop
Disease (a)

Disease-free
Exposed
Sample

(b)
to factor

Disease- Develop
free Disease (c)
Unexposed Disease-free
to factor (d)
Present time Future time
Starting point Follow
 Basic structure of case-control design Population

Diseased
Disease-free
Exposed to factor
(a)
Is calculated between both groups
The Odds chance of exposure

Unexposed to factor
Diseased
(b)
(cases)

Sample
Exposed to factor
(c)

Unexposed to factor Disease-free

(d) (controls)
Trace Present time
Past time
Starting point
The following table depicts the outcomes of isoniazid/placebo trial among children
with HIV (death within 6 months).
What is the risk of dying?
Interventions Dead (within Alive Total
6 months)

Placebo 21 110 131 Risk=21/131=0.160

Isoniazid 11 121 132 Risk=11/132=0.083

Absolute risk difference (ARD)=risk in placebo-risk in isoniazid= 0.077

Net relative risk (NRR)=risk in placebo/risk in isoniazid= 1.928

Relative risk reduction (RRR)=risk in placebo-risk in isoniazid/risk in placebo= 0.48

Number needed to treat (NNT)=1/ARD=1/0.077=13

 Relative risk (RR)

Mammography Breast cancer No breast cancer Total

Positive a-10 b-90 100

Negative c-20 d-998980 100,100

In Cohort design

RR= a/(a+b)c/(c+d)
10/(100) 20(100,100)=0.1/0.0002= 500
 The relative risk (RR)
Lung No lung Total
cancer cancer
Smokers 18 582 600
Non 6 1194 1200

Risk for smokers=18/600=0.03

Risk for non-smokers=6/1200=0.005
RR=0.03/0.005=6
 Interpreting the Relative Risk
(measure the strength of the association)

If RR = 1 Risk in exposed equal to risk in non-

exposed (no association).
If RR > 1 Risk in exposed greater than risk in non-
exposed (positive association; possibly
causal).
Risk in exposed less than risk in non-
If RR < 1 exposed (negative association; possibly
protective).
 Calculating the Relative Risk in Cohort Studies

Then follow to see whether

Disease develops Disease does not Totals Incidence rate of
develop disease
a
a b a+b
a+b
First
Exposed
select
c d c
No exposed c+d
c+d

a = incidence in exposed c
= incidence in non-exposed
a+b c+d
 Hypothetical Cohort
3,000 smokers and 5,000 non-smokers to investigate the relation of smoking to
the development of coronary heart disease (CHD) over a 1-year period.

Develop CHD Do not develop Totals Incidence per

CHD 1,000/year

Smoke cigarettes 84 2,916 3,000 28.0

Do not smoke 87 4,913 5,000 17.4

cigarettes

Incidence among the exposed= Relative risk =

84/3,000 = 28.0 per 1,000 Incidence in exposed

Incidence in non-exposed =
Incidence among the non-exposed
= 87/5000 =17.4 per 1,000 28.0/17.4 = 1.61
 Example: the British Heart Study

A large cohort study of 7735 men aged 40-59 years randomly

selected from general practices in 24 British towns, with the
aim of identifying risk factors for ischemic heart disease. At
recruitment to the study, the men were asked about a
number of demographic and lifestyle, including information
on cigarette smoking habits.
Of the 7718 men who provided information on smoking
status, 5899 (76.4 %) had smoked at some stage during their
lives (including those who were current smokers and those
who were ex-smokers).
Over subsequent 10 years, 650 of these 7718 men (8.4 %) had
a myocardial infarction (MI).

9/30/2017 Dr. Tarek Tawfik 17

 MI in subsequent 10 years

Yes No Total
Smoking status at baseline
Ever smoked 563 (9.5%) 5336 (90.5%) 5899

Never smoked 87 (4.8%) 1732 (95.2%) 1819

Total 650 (8.4%) 7068(71.6%) 7718

The estimated relative risk=

The middle aged man who has ever
(563/5899)
smoke is twice as likely to suffer a
(87/1819)
MI over the next 10 years period as
= 2.00
a man who has never smoked.
CI = 1.60-2.49
(does not include 1)
Odds ratio
 The Odds ratio (relative odds)
In order to calculate a relative risk, we must have values for the
incidence in the exposed and non-exposed, as can be obtained
in the 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).

Hence, we can not estimate the RR in case-control study

directly and we implement another measure of association
called Odds ratio.
 Defining the Odds ratio in Cohort and Case-control
studies.

Suppose we betting on a horse named Little Beauty, which has a 60%

probability of wining the race (P). Little Beauty, therefore has a 40 %
probability of losing (1-P). What are the odds that the horse will win
the race?
The odds is defined as: the ratio of the number of ways the event can
occur to the number of ways the event can not occur.
Probability that Little Beauty will win the race
Odds =
Probability that Little Beauty will lose the race

Odds = P/(1-P) or 60 %/40 % = 1.5:1 = 1.5

Probability of wining is 60 %, while the odds (chance) of wining is 1.5
times.

9/30/2017 Dr. Tarek Tawfik 21

 Odds ratio (OR)

o An odds ratio (OR) is a measure of association

between an exposure and an outcome.
o The OR represents the odds that an outcome will
occur given a particular exposure, compared to the
odds of the outcome occurring in the absence of
that exposure.
o Odds ratios are most commonly used in case-
control studies, however they can also be used in
cross-sectional and cohort study designs as well
(with some modifications and/or assumptions).
 OR
Rare disease assumption (prevalence < 10%).
Case control-design
Regression analysis
Meta-analysis
 Calculation

Case control Diseased None Total

study
Exposed Cases+ exposed Exposed+ not a+b
(a) diseased (b)
Non-exposed Cases-not Not exposed+ not c+d
exposed (c) diseased (d)

Odds ratio= a/cb/d= ad/bc

Prevalence among the diseased/prevalence among the non-diseased

OR=1 Exposure does not affect odds of outcome

OR>1 Exposure associated with higher odds of outcome
OR<1 Exposure associated with lower odds of outcome
 Odds ratio
Case control Lung cancer No lung cancer Total
study
Smoking a-80 b-30 110
None c-20 d-70 90

80x70=5600 Or 80/2030/70=9.3
30x20=600
5600/600=9.3
 The Odds ratio (OR)
Lung No lung Total
cancer cancer
Smokers 80 30 110
Non 20 70 90

Odds for smokers=80/30=2.67

Odds for non-smokers=20/70=0.29
OR=80*70/30*20=9.33
 Odds Ratios in Case-Control and Cohort Studies

Cohort Develop Do not Case-control Cases Controls

disease develop
disease History of a b
exposure
Exposed a b
No history of c d
Not exposed c d exposure

Odds ratio= Odds ratio =

Odds that an exposed person Odds that a case was exposed
Develops disease Odds that a control was exposed
Odds that a non-exposed = a/c
Person develops disease
b/d
= a/b
c/d = ad
= ad bc
bc
 Example: HRT
A total of 1327 women aged 50 to 81 years with hip
fractures, who lived in a largely urban area in Sweden,
were investigated in this un-matched case-controls
study. They were compared with 3262 controls within
the same age range selected from the National register.
Interest was centered on determining whether
postmenopausal hormone replacement therapy (HRT)
substantially reduced the risk of hip fracture.
The results in the table show the number of women
who were current users of HRT and those who had
never used or formerly used HRT in cases and controls.
 Current users of HRT Never used HRT/
former user of HRT
Total
With hip fracture (cases) 40 (14%) 1287 (30%) 1327

Without hip fracture (controls) 239 3023 3262

Total 279 4310 4589

A postmenopausal woman
in this age range in Sweden
who was a current user of
The observed Odds ratio = HRT thus had 39 % of the
(40X3023) risk of hip fracture of a
(239X1287) woman who had never used
=0.39 or formerly used HRT
C.I = 0.28 to 0.56 Being current user of HRT
reduced the risk of hip
fracture by 61%.
 When is the Odds Ratio a Good Estimate of the
Relative Risk?
In case-control, only the odds ratio can be calculated as a measure of
association, whereas in a cohort, either the relative risk or the odds ratio is a
valid measure of association.

Nevertheless, estimate of RR can be used in interpreting case-

control study in the following occasions:
When the cases are representative, with regard to history of exposure, of all
people with disease in the population from which the cases are drawn.
When the controls are representative with regard to history of exposure, of all
people without the disease in the population from which the cases were drawn.
When the disease being studied dose not occur frequently.
 Odds Ratios and Relative risk

Disease Do not Total Develop Do not Total

develops develop disease develop
disease disease
Exposed 200 9800 10,000 Exposed 50 50 100

Not Not exposed 25 75 100

exposed 100 9900 10,000

Relative Risk =
Relative risk=
50/100
200/10,000
25/100
100/10,000 =2
=2
Odds Ratio=
Odds ratio =
200X9900
50X75
100X9800=
25X50
2.02
=3
In cohort, the discrepancy between RR and Odds Odds ratio inflated due to high
Is less (the denominator is always large) prevalence of the outcome (>10%)
 Remember

The relative odds (odds ratio) is a useful measure of

association in both case-control and prospective studies
Cohort.
In a cohort study, the relative risk can be calculated
directly.
In a case-control study, the relative risk cannot be
calculated directly, so that the relative odds or odds
ratio (cross-product ratio) is used as an estimate of the
relative risk when the risk of the disease is low.

9/30/2017 Dr. Tarek Tawfik 33

 Calculating the Odds ratio in a Matched Pairs Case-
Control Study.

According to the type of exposure, case-control study can be classified into four
groups:

- pairs in which both cases and controls

were exposed.
Concordant pairs - pairs in which neither the cases nor the
controls were exposed.

- pairs in which the case was exposed but

the control was not.
Discordant pairs - pairs in which the control was exposed
and the case was not.

9/30/2017 Dr. Tarek Tawfik 34

 2X2 table
Control

Cases Exposed Not exposed

Exposed a b
Both the case and control were The case was exposed and the control was
exposed not

Not exposed c d
The case was not exposed and the Neither the case nor the control was exposed
control was exposed

Calculation entail the discordant pairs only (b and c), we ignore

the concordant pairs, because they do not contribute to our
knowledge of how cases and controls differ in regard to past
history of exposure.
The odds ratio will then equals = b /c
 Case-control study of brain tumors in children.

o A number of studies have Normal control Total

suggested that children Cases 8+ lbs < 8lbs

with higher birth weights 8+ lbs 8 18 26
are at increased risk for
childhood cancer. < 8 lbs 7 38 45

o In the next analysis,

exposure is defined as birth Total 15 56 71
weight greater than 8 lbs.
Odds ratio =
18/7 = 2.57
2= 4.00
P = 0.046

9/30/2017 Dr. Tarek Tawfik 36

 Attributable Risk

How much of the disease that occurs can be attributed to a

certain exposure?
Attributable risk is defined as the amount or proportion of disease
incidence (or disease risk) that can be attributed to a specific
exposure.
How much of lung cancer risk experienced by smokers can be
attributed to smoking?
More important than RR as it addresses important clinical
practice and public health. How much of the risk (incidence) of
disease can we hope to prevent if we are able to eliminate
exposure to the agent in question?

9/30/2017 Dr. Tarek Tawfik 37

 Attributable Risk for the Exposed Group

Background
Level of risk

Exposed risk
Group
In non
Exposed
group

9/30/2017 Dr. Tarek Tawfik 38

 Incidence due
to exposure

Incidence not due

to exposure

In exposed In the non-

group exposed group
 Calculations

The incidence of a disease that is attributable to the exposure in the exposed

group can be calculated as follow:

(incidence in the exposed group) - (incidence in the non-exposed group)

Then, what proportion of the risk in exposed persons is due to the exposure?
(incidence in the exposed group) - (incidence in the non-exposed group)

incidence in the exposed group

9/30/2017 Dr. Tarek Tawfik 40

 Attributable Risk for the Total Population
What proportion of the disease incidence in a total population (both exposed
and non-exposed) can be attributable to a specific exposure?
What would be the total impact of a prevention program on the community?

Calculations entail:
(Incidence in the total population) (incidence in non-exposed group background risk).

In proportion:
(Incidence in the total population) (incidence in non-exposed group background risk).
Incidence in total population

9/30/2017 Dr. Tarek Tawfik 41

 Example for calculating the attributable risk in the exposed group

Smoking status Develop CHD Do not develop Total Incidence per 1,000 per
CHD year

Smoke cigarettes 84 2,916 3,000 28.0

Do not smoke 87 4,913 5,000 17.4

cigarettes

Incidence among smokers = 84/3,000 = 28.0 per 1,000

Incidence among non smokers = 87/5,000 = 17.4 per 1,000
The AR = (incidence in exposed group) (incidence in the non exposed group) =
28.0 17.4 /1,000 = 10.6 /1,000????
In proportion = The AR = (incidence in exposed group) (incidence in the non exposed
group) /( incidence in exposed group)
= 28.0 17.6/ 28.0 = 10.6/28.0 = 0.379 = 37.9 %?????
 What does this mean?
The attributable risk = 10.6 /1,000, it means that 10.6 of the
28.0/1,000 incident cases in smokers are attributable to the fact
that these people smoke.

Thus if we had an effective smoking cessation campaign, we

could prevent 10.6 of the 28/1,000 incident cases of CHD that
smokers experience.

In proportion, 37.9 % of the morbidity from CHD among smokers

may be attributable to smoking and could presumably be
prevented by eliminating smoking.

9/30/2017 Dr. Tarek Tawfik 43

 Attributable risk in total population

The incidence in the total population can be calculated by

subtracting the background risk.
(incidence in the total population) (incidence in the non-exposed group),
for
calculation we must know the incidence of the disease in the
total population (which we often do not know), or all of the
following three values, from which we can then calculate the
incidence in the total population:
The incidence among exposed.
The incidence among the non-exposed.
The proportion of the total population that exposed (frequently
assumed or judged).

9/30/2017 Dr. Tarek Tawfik 44

 AR in total population.

Assuming that the incidence in the total population of smoking

is 44% (and therefore the proportion of non-smokers is 56%).
The incidence in the total population can then be calculated as
follows:
(incidence in smokers)(% of smokers in the population) +
(incidence in non-smokers)(% of non-smokers in population).
= (28.0/1,000)(0.44)+(17.4/1,000)(0.56)= 22.1/1,000
Then the AR= 22.1/1,000 17.4/1,000 = 4.7/1,000.
It means that, if we an effective prevention program, how much
reduction in the incidence of the CHD could be anticipated.

9/30/2017 Dr. Tarek Tawfik 45

 AR in total population

Proportion of incidence in the total population =

(incidence in the total population) (incidence in the non-
exposed group)/ incidence in the total population = 22.1-
17.4/22.1= 21.3%.

Thus, 21.3 % of the incidence of CHD in this total population

can be attributed to smoking, and if an effective prevention
program eliminated smoking, the best we could hope to achieve
would be a reduction of 21.3 % in the incidence of CHD in the
total population which consisting of both smoking and non-
smoking.

9/30/2017 Dr. Tarek Tawfik 46

Thank you