You are on page 1of 16

RR & OR

Confounders
Outline
• Example on RR and OR

• Computation of CI

• Mantel-Haenzel

• Confounders
RR example
• Data from a cohort study of current oral
contraceptive (OC) use and myocardial
infarction in women

MYOCARDIAL INFARCTION
YES NO TOTAL
OC USE
23 304 327
NEVER
USE OC 133 2816 2949

TOTAL 156 3120 3276


Example
MYOCARDIAL INFARCTION
YES NO TOTAL
OC USE
23 304 327
NEVER
USE OC 133 2816 2949

TOTAL 156 3120 3276

• RR = 23 / 327 / 133/2949
• RR = 1.56
• Women who take OC have 56% more likely to
develop MI than non-user.
Find CI for RR
  a b   a b 
 explog(RR)  1.96    , explog(RR)  1.96    
 c(a  c) d (b  d )  c(a  c) d (b  d )  
  

  23 133   23 133 
 explog(1.56)  1.96    , exp  log(1.56)  1.96    
 304(23  304) 2816(133  2816)  304(23  304) 2816(133  2816)  
  

 exp 0.445  0.0366, exp 0.445  0.0366 


 exp 0.408, exp 0.4816 
[1.504;1.619]
OR example
• Hypothetical case-control study of
cigarette smoking and lung cancer among
100 cases and 100 controls

LUNG CANCER
Cigarette smoking CASES CONTROL TOTAL
YES
70 30 100
NO
30 70 100

TOTAL 100 100 200


Example
LUNG CANCER
Cigarette smoking CASES CONTROL TOTAL
YES
• OR =70*70 / 30*30 NO
70 30 100

30 70 100

• OR = 5.4 TOTAL 100 100 200

• Smokers are 5.4


times more likely to
develop lung cancer
than non-smokers
Find CI for OR
    
 exp log(OR )  1.96  1  1  1  1 , exp log(OR )  1.96  1  1  1  1  
 a b c d a b c d 
     

    
 explog(5.4)  1.96  1  1  1  1 , explog(5.4)  1.96  1  1  1  1  
 70 30 30 70  70 30 30 70  
  

[exp(1.69  0.605); exp(1.69  0.605)]


[exp(1.085); exp(2.295)]
[2.96;9.92]
How the DESIGN affect the
RR or OR ??????
LUNG CANCER • RR=4.6
Cigarette smoking CASES CONTROL TOTAL
YES • OR=5.4
70 300 370
NO
30 700 730

TOTAL 100 1000 1100

LUNG CANCER
Cigarette smoking CASES CONTROL TOTAL
YES
70 30 100
NO • RR=2.3
30 70 100
• OR=5.4
TOTAL 100 100 200
Adjustment
Disease + Disease - Total
Risk Factor + 160 80 240
Risk Factor - 440 320 760
Total 600 400 1000

Females:

Disease + Disease - Total


Risk Factor + 240 330 570
Risk Factor - 160 270 430
Total 400 600 1000

ORMH 
 ( all tables i )
ai d i / ni

 ( all tables i )
bi ci / ni

(160  320) / 1000  (240  270) / 1000


ORMH   1.32
(80  440) / 1000  (330  160) / 1000
Introduction to confounding
• Epidemiologic studies can be affected by
different types of biases, which include
confounding

• Definition:
• A confounder is an extraneous variable that
totally or partially accounts for the apparent
effect of the study exposure on the outcome. It
may even mask an underlying true association
or reverse it.
OLD Depression
+ -
vitamin deficiency+ 1200 800 2000 RR=1.0
- 60 40 100

YOUNG Depression
+ -
vitamin deficiency+ 20 80 100 RR=1.0
- 200 800 1000

ALL Depression
+ -
vitamin deficiency+ 1220 880 2100 RR=2.5
- 260 840 1100

No relationship between vitamin deficiency and depression.


The coufounfing effect of age TOTALLY accounts for the
relationship
MALES Heart Disease
+ -
Red meat + 1400 280 1680 RR= 2
- 200 280 480

FEMALES Heart Disease


+ -
Red meat + 44 220 264 RR= 2
- 20 220 240

ALL Heart Disease


+ -
Red meat + 1444 500 1944 RR= 2.431
- 220 500 720
• Sex is partially responsible for the association
between red meat diet and heart disease
Smokers Respiratory disease
+ -
Air pollution + 30 70 100 RR= 3.282
- 67 666 733

Non-Smokers Respiratory disease


+ -
Air pollution + 14 412 426 RR= 3.286
- 1 99 100

ALL Respiratory disease


+ -
Air pollution + 44 482 526 RR= 1.025
- 68 765 833

• This represent an example of a confounding effect


masking a true association
Severe asthma Death
+ -
DRUG A 200 800 1000 RR= 0.5
B 4 6 10

Severe asthma DEATH


+ -
DRUG A 2 98 100 RR= 0.5
B 4 96 100

ALL Patients DEATH


+ -
DRUG A 202 898 1100 RR= 2.525
B 8 102 110

• Example on how confounding can reverse


the true exposure-outcome relationship
Overview of control strategies for
confounding

Strategy Level of use


Randomization Design
Restriction Design
Stratification Analysis
Multivariate analysis Analysis

You might also like