Research designs
Hein Stigum
Presentation, data and programs at:
http://folk.uio.no/heins/
May-16
H.S.
1
Designs
• Aims
– Disease occurrence
– Exposure-Disease association
• Designs
– Cross-sectional studies
– Cohort studies
– Case-control studies
• Case-Cohort
• Nested Case Control
• Traditional Case Control
May-16
H.S.
2
The 2 by 2 table
May-16
H.S.
3
True or false?
It takes 2 to tango
It takes 3 chords to play the blues
It takes 4 numbers to be an epidemiologist
May-16
H.S.
4
The 2 by 2 table
Exposure
+
-
Disease
+
100
100
10
100
OR=
10,0
1 1 1 1
se(ln( OR )) 
  
a b c d
.01+.01+.1 +.01 =.13
Add 100 to 100 cell: .01+.01+.1 +.005=.125
Add 10 to 10 cell: .01+.01+.05 +.01=.08
The information is in the contrasts
The lowest number sets the precision
May-16
H.S.
5
3 examples
• Gender and Smoking
– Do girls smoke more than boys?
• Exercise and coronary heart disease (CHD)
– Does exercise reduce the risk?
• Genes and Diabetes type 1
– Does gene-type increase the risk?
What design should we use?
May-16
H.S.
6
Cross-section
May-16
H.S.
7
Time
Existing
Cases
Healthy
time
May-16
H.S.
8
Cross-sectional example
Girls
Boys
Smoking
+
300
700
140
860
Disease freq
0.22
Exposure freq
0.5
1000
1000
2000
Pro:
Con:
Ex
1
2
3
Exposure
Girl
Exercise
Gene
May-16
Disease
Smoking
CHD
Diabetes 1
Frequency
Risk
Odds
0.30
0.43
0.14
0.16
Association
RD
RR
OR
0.16
2.1
2.6
0
1
1
fast and inexpensive
reversed causality
Crosssection
Dis freq
22 %
OK
9 % Rev. Caus.
1%
H.S.
CaseCohort Control
9
Cohort
May-16
H.S.
10
Disease frequency depicted
100
Existing
Cases
t
a New
Healthy
Healthy*time
0
50
Cases
time
May-16
H.S.
11
Cohort example
CHD
Exercise
Inactive
+
100
800
Disease freq
0.09
Exposure freq
0.2
1900
7200
2000
8000
10000
Pro:
Con:
Ex
1
2
3
Exposure
Girl
Exercise
Gene
May-16
Disease
Smoking
CHD
Diabetes 1
Frequency
Risk
Odds
0.05
0.05
0.10
0.11
Association
RD
RR
OR
-0.05 0.50 0.47
0
1
1
reliable
costly, time consuming
CrossCaseDis freq section Cohort Control
22 %
OK
OK
9 % Rev. Caus.
100 000
1%
H.S.
12
Case Control studies
May-16
H.S.
13
Cohort vs Case-Control
May-16
Not disease
Controls
Cases
Disease
H.S.
14
Case-Control studies
• Cohort studies
– Measure the exposure experiance of the entire
population
• Case-Control studies
– Measure the exposure experiance of a sample
of the source population of cases (=base)
– Key assumption
• Sample controls independent of exposure
(same k)
– Prospective or retrospective
May-16
H.S.
15
Case-Cohort
Existing
cases
t
.
.
.
.
New
cases
a+b
controls
Healthy
N1+N0
Healthy*time
py
start
c+d
end
time
May-16
H.S.
16
Nested Case-Control
Existing
cases
.
case
.
case
controls
Healthy
N1+N0
.
.
t
New
cases
a+b
controls
Healthy*time
py
.
risk set
c+d
.
risk set
start
end
time
May-16
H.S.
17
Traditional Case-Control
Existing
cases
Healthy
N1+N0
t
Healthy*time
py
start
.
.
.
.
New
cases
a+b
controls
c+d
end
time
May-16
H.S.
18
Traditional Case-Control example
Gene
Wildtype
Odds
OR
Ex
1
2
3
Diabetes 1
+
50
20
350
380
400
400
0.14 0.05
2.7
1
Exposure
Girl
Exercise
Gene
May-16
Disease
Smoking
CHD
Diabetes 1
Frequency
Association
OR
2.7
1
Pro:
Con:
small, efficient
recall bias
CrossCaseDis freq section Cohort Control
22 %
OK
9 % Rev. Caus.
OK recall bias
OK
1%
100 000
H.S.
19
Calculations
May-16
H.S.
20
Data layout
Cases
Noncases
Persons
PersonTime
Exposed
a
c
N1
T1
Nonexposed
b
d
N0
T0
OBS, table transposed, b and c different
from earlier tables
May-16
H.S.
21
Cross-sectional Data
Exposed
Nonexposed
Cases
e1
e0
Population
N1
N0
Prevalence
e1 / N1
e0 / N 0
Controls 2
kT1
kT0
Pseudo rates
a / kT1
Pro: fast
Controls
3 and inexpensive
kc
Con:
reversed causality
Pseudo odds
May-16
a/kc
b / kT0
RR :
e1 /N 1
e0 /N 0
Rate Ratio
a/T1
b/T0
Odds ratio
ab
cd
kd
b / kd
H.S.
22
Cohort Data
Cases
Persons 1
Controls
Exposed
a
kNN11
Nonexposed
b
kNN00
Pseudo
Risksrisks
aa//kNN11
bb//kNN00
Person
Time
PersonTime
T1
T0
Rates
a / T1
b / T0
Controls
3
kcin the 2x2kd
• Unbalance
table
– Exposure
Pseudo odds
a/kcis rare b / kd
– Disease is rare
May-16
H.S.
a/N 1
Risk Ratio
Ratio :
Risk
b/N 0
Rate
Rate Ratio
Ratio :
a/T1
b/T0
Odds ratio
ab
cd
23
1) Case Cohort
2) Nested Case Control
3) Traditional Case Control
Cases
Controls 1
Exposed
a
Nonexposed
b
kN1
kN00
Pseudo risks
a / kN1
b / kN 0
Controls 2
kT1
kT0
Pseudo rates
a / kT1
b / kT0
Controls 3
kc
kd
Pseudo odds
May-16
a/kc
b / kd
H.S.
k=sampling fraction
a/N 1
Risk Ratio :
b/N 0
Rate Ratio :
a/T1
b/T0
Odds ratio :
ad
ab
cb
cd
24