Professional Documents
Culture Documents
1
Association between number of TV sets per
10 persons versus cancer incidence per
Case series 100,000
Study of a larger group of
patients (e.g. 10 or more) with a
particular health condition
Often used to delineate the
clinical picture of a disease
• Rephrasing
p g the hypothesis
yp for a cross-sectional study:
y
In a general population with smokers and non-
time Descriptive smokers, as well as subjects with or without heart
Point of contact Component disease, is heart disease more common among
smokers?
A population is studied at a
particular “cross-section” in time
2
The cross-sectional study Cross-sectional study: salient features
• Measures frequency of a disease / health state
Smoker Non smoker
• Studies a specific / well-defined population
Heart disease patients A B
• Studies a specified point / period (i.e., cross-section) in
Healthy subjects C D time
• May have both descriptive and analytical components
• At any one time-point, 4 groups of people can be found in a
population. • ‘Risk factor’ (exposure) and ‘Disease’ (outcome) are
measured concurrently
• Prevalence of smokers among heart disease patients
• Descriptive use of a cross-sectional study provides valid
• = A/ (A+B)
results (i.e. what we estimate is the “truth”)
• Prevalence of smokers among healthy subjects = C/ (C/D)
• Analytical use of a cross-sectional study provides clues
• Ratio ([A/(A+B) / C/(C+D)] is the prevalence ratio and regarding a “true” causal association (i.e. suggestive, not
measures association between smoking and heart disease. conclusive evidence)
Time= x years
Study Determine Follow-up and compare
Time=0
population exposure outcome between
that is free status at exposed and unexposed
of outcome baseline Which group has more
di
disease or has
h di
disease
at a faster rate?
Free of lung cancer
Smoking Occurrence of
(exposed) lung cancer
Non-smoking Occurrence of
Slowly progressing “surviving” cases observed in a cross-section in time may (not exposed)
represent “healthy” cases and may not be suitable for studying putative risk factors lung cancer
3
Cohort studies Cohort studies
4
Selection bias
Smokers Occurrence of Case-control studies
(exposed) lung cancer cases
Occurrence of Compare frequency of exposure between
lung cancer persons with disease (cases) and persons
without disease (controls)
Non-smokers
(not exposed)
Without lung controls
cancer
Selection of all incident cases within a
Free of cancer Without lung specific period of recruitment: clinics,
cancer hospitals registries,
hospitals, registries etc.
etc
STUDY DESIGNS
Case-control studies
COHORT CASE-CONTROL
Start with X; end with Y Start with Y; end with X
Disadvantages
Advantages Temporal relationship
Quick, less resource
Quick resource- more difficult to
establish
Y = f (X)
intensive
Practically the only Prone to bias (e.g. recall
study design possible bias or selection bias)
for rare diseases Does not measure “risk”
directly (calculates CROSS-SECTIONAL
“odds”) Start with X and Y
together
References
• Medical
Epidemiology
• Raymond SS. Greenberg et al
• Chapters 8 and 9