MEASURES OF STRENGTH OF ASSOCIATION

Measures of Association
Exposure Outcome

Is there a relationship between the exposure and outcome of interest?

Types of Measures of Association

• Absolute (Measure Public Health Impact)

– Attributable Risk

– Population Attributable Risk

• Relative (Measures strength of association)

– Relative Risk

– Odds Ratio 1
 P-values indicate the likely hood that chance variation
can be excluded as an explanation for an association – and
thus is a measure of the degree of statistical significance.
 However, it doesn’t tell us about the strength of an
association i.e how strong is the effect of exposure on the
disease.
 Measures of association between risk factors and disease
are often calculated from data presented in 2x2

tables.

2
 A 2x2 table showing association
Disease
Yes No
Exposure

Yes A B A+B
No C D C+D
A+C B+D A+B+C+D
A= # of individuals who are exposed and have the disease
B = exposed and do not have the disease
C = not exposed and have the disease
D = non exposed and do not have the disease
A + B = total # of individuals exposed
C + D = total # of individuals non exposed
A + C = total # of individuals with the disease
B+ D = total # of individuals with out the disease 3
Relative Risk (RR)
 Indicates the likelihood of developing the disease in the
exposed group relative to those who are not exposed.
 Expresses the risk of developing a certain disease in
people exposed to a certain factor as compared to the
risk of disease in people not exposed to the factor.
 It estimates the magnitude of the association between
exposure and disease.
RR = IR among exposed
IR among non-exposed 4
 A C
 
A B CD

 A  C  D
  
=  A  B   C 

The RR as all other measures of association can also be

used to compare risks of death, accident or other possible
outcome of an exposure.
Example:
In a study of lung cancer mortality in male physicians in
relation to smoking; the annual age-standardized mortality
rate for lung cancer was 96/100,000 in smokers, and
7/100,000 in non-smokers.
RR = 96/100000 = 13.7
7/100000 5
 i.e. Smokers were 13.7 times as likely to die from lung
cancer as non-smokers.
 Strength of association in relation to RR

RR Strength of association

1.2-1.4 weak

1.5-2.9 Moderate

3.0-10 or more Strong

6
Example: Data from a cohort study of oral contraceptive
(OC) use and bacteriuria among women aged 16-49 years

Bacteriuria
OC use Yes No

Yes 27 455 482

No 77 1831 1908
104 2286 2390

Solution:
RR = Ie = a/(a + b)
= 27/482 = 1.4
Io c/(c + d) 77/1908
Conclusion: This value tells that women who used OCs had 1.4
times risk of developing bacteriuria as compared with none users.
7
Odds Ratio (OR)
 In case control studies where participants are
selected on the basis of disease status, It is not
possible to determine IR.
 Therefore, direct computations of RR is not
possible.
 Here, an indirect estimation of the RR is given by
the OR.
8
 OR = Odds of having the disease if exposed
Odds of having the disease if non- exposed

= ratio of diseased to non-diseased in exposed

ratio of diseased to non-diseased in non- exposed
 Odds of disease is a simple ratio(not proportion)
and it indicates the odds of diseased relative to
exposure status.

9
Odds of disease in exposed = a/b or a:b

Odds of disease in non-exposed = c/d or c:d

OR = a/b ÷ c/d = ad/bc

Derivation – in most chronic disease studies, cases

represent only a small fraction of the exposed
and unexposed population.

Therefore- B is about equal to A + B &

D is about equal to C+ D 10
Hence, A/A+B = A/B = AD
C/C+D C/D BC
Example:

The relationship between the deaths from lung cancer &

exposure to asbestos from a case control study
Asbestos Lung cancer death
exposure Yes No

Yes 90 270 360

No 60 360 420
150 630 780

11
OR = AD/BC = 90 × 360 = 2
60 ×270

i.e. People exposed to asbestos have 2x higher risk of developing

lung cancer as compared to those without the exposure.
Attributable Risk (AR)

As has been discussed, the RR represents the likelihood of

disease in exposed individuals relative to those who are none
exposed.
 It indicates how much of the risk is due to ( attributable) to the
exposure

12
The AR also called the risk difference is a measure of
association that provides information about the
absolute effect of the exposure or the excess risk of
disease in those exposed compared with those none
exposed. This measure is defined as the difference
between the incidence rates in the exposed and none
exposed groups and can be calculated as follows:

13
 It guaranties the excess risk in the exposed that can
be attributed to the exposure by removing the risk
of disease that could have occurred due to other
causes.
AR  Ie  Io
= risk in exposed - risk in non exposed

In cohort studies, AR is calculated as the difference of CIs or

incidence densities.
AR  CIe  CIo
a c
= 
ab cd 14
 The interpretation is dependent on the assumption that a
cause-effect relationship exists between exposure and disease.
 If there is no association, no difference between the IRe & IRo
so AR = 0

If AR >0 Causal association

When AR>0, the value indicates the number of cases of the
disease among the exposed, that could be eliminated if that
exposure were eliminated.

15
 Thus AR can be useful as a measure of public health impact of a particular
exposure.

Example: AR of ( The study of the relation between smoking & lung cancer)

AR = IR among smokers - IR among non-smokers

= 96/100000/year - 7/100000/year

= 89/100000/year

This means- in a population of 100,000 smokers similar to those studied,

96 would be expected to die of lung cancer per year (89 of those deaths
being related to the smoking and the remainder 7 to other factors.

16
 Interpretation: If we prevented the 100,000 from
smoking, we would have prevented about 89 deaths
from lung cancer per year.
• Unlike RR, AR retains the time unit used in the
calculation of the IR.

Attributable Risk percent (AR%)

The attributable risk percent (AR%), also referred to as

the attributable rate percent, attributable proportion, or
etiologic fraction is the proportion of the disease among
the exposed that is attributable to the exposure.
17
 It is the proportion of disease in that group ( study)
that could be eliminated by eliminating exposure.
AR
AR %   100
Ie
 Ie  Io 
   100
 Ie 
In the previous example:-

89 / 100,000 / year
AR%   100
96 / 100,000 / year
 93%

18
- In this study, 93% of the deaths from lung cancer in
smokers, can be attributed to their smoking habit.
- 93% of the deaths from lung cancer in smokers could
have been prevented if they did not smoke
- For most case control studies AR can not be
calculated because the IR of disease among exposed
& non-exposed is not available. However, AR% can
be calculated using the OR.
19
Example: OR =2  OR  1 
AR%     100
So,  OR 
 2 1
AR %     100
 2 
 50%
Interpretation:
50% of deaths from lung cancer in those exposed to
asbestos could be attributed to the exposure.
By avoiding the exposure to asbestos 50% of the
deaths from lung cancer would be prevented in that
group. 20
Example: From the cohort study of OC use and bacteriuria, the
attributable-risk percent would be calculated as follows:

AR% = (AR/ Ie) 100

= 1566/105 x 100

27/482

= 27.96%
• Thus, if OC use does cause bacteriuria, about 28 percent of
bacteriuria among women who use OCs can be attributed to
their OC use and could therefore be eliminated if they did not
use OCs 21
 Population Attributable Risk (PAR)

• Estimates the excess rate of disease in the total

study population of exposed and non-exposed
individuals that is attributable to the exposure.
• Population attributable risk (PAR), helps determine
which exposures have the most relevance to the
health of a community. The population attributable
is calculated as the rate of disease in the population
(IT) minus the rate in the unexposed group (Io): 22
 PAR = IT - Io

Example: Using the data of bacteriuria among the OC users,

calculate the population attributable risk.

Solution:

PAR = IT - Io

= 104/2390 – 77/1908

= 316/105/year

Thus, if OC use were stopped, the excess annual incidence rate

of bacteriuria that could be eliminated among women in this
study is 316 per 100,000. 23
 Also PAR = (AR) x (Pe)

Where Pe-is the proportion of exposed individuals in

the population

Example: If 20% of the total population are smokers

with an AR of 89/100,000/year then

PAR = AR x prevalence rate of smking in the

population.

= 89/100,000/year x 20%

= 17.8/100,000/year 24
 i.e. In a general population of 100,000 with a
prevalence rate for smoking of 20%, about 18 deaths
from lung cancer per year would be prevented by
eliminating cigarette smoking.
 AR among the exposed is always greater than the PAR
since the impact of removing the exposure on the
number of cases of the disease will always be greater
for those with the exposure than for a total population
which is a combination of exposed & non-exposed
25
Population Attributable Risk percent (PAR%)
The population attributable risk percent expresses
the proportion of disease in the study population
that is attributable to the exposure and thus could
be eliminated if the exposure were eliminated. The
population attributable risk is calculated by dividing
the population attributable risk by the rate of the
disease in the population:

26
PAR% = Risk in population – Risk in non exposed x 100
Risk in population
IT  I o
  100
IT
PAR
  100
IT
Example: in the cohort study of OC use and bacteriuria ,
the incidence rate of bacteriuria in the total study
population was 104 per 2390 0r 4351.5 per 100,000, and
the population attributable risk was 316 per 100,000. The
population attributable-risk percent can then be
calculated as follows: 27
 PAR% = (PAR/ IT) x 100

=(316/ 4351.5) x 100

= 7.3%
Thus, if OC use causes bacteriuria, about 7 percent of all
the bacteriuria in the study population (and about 28
percent of bacteriuria among women taking OCs) could be
prevented if they did not use OCs.
Interprétation of measures of association
Possible outcomes in studying the relationship between
disease and exposure 28
1. No association between Exposure and disease
AR = 0
RR = 1
2. Positive association between Exposure and disease
AR > 0
RR > 1
3. Negative association between Exposure and disease
AR < 0 (negative)
RR < 1 (a fraction)
Example: In a cohort study of the association between prolonged alcohol drinking and
chronic liver disease the figures in the 2x2 table were found.

29
 Alcohol Chronic liver disease
drinking Yes No

Yes 40 960 1000

No 60 2940 3000
100 3900 4000

Calculate
- Ie = 40/1000 = 0.04
- Io = 60/3000 = 0.02
- RR = Ie/Io = 40/1000 x 3000/60 = 2
- AR = Ie-Io = a/a+b- c/c+d = 40/1000-60/3000 = 0.02
- AR% = AR/Ie x100 = 0.02/0.04 x100 = 50%
- PAR = IT – Io = 100/400- 60/300 = 0.005
- PAR% = PAR/IT = 0.005/0.025 x100 = 20% 30