case-control study 1: design and odds ratio preben aavitsland (partly based on epiet 2004)
TRANSCRIPT
Case-control study 1:Design and odds ratio
Preben Aavitsland
(partly based on Epiet 2004)
Contents
• Monday 1– Design: Case-control study as a smarter cohort study– The odds ratio
• Tuesday 2– Choosing cases and controls– Matching– Power calculation
• Wednesday– Case-control studies in outbreaks
• Thursday 3– Bias and confounding– Analysis
Why case-control study
• Best way to acquire knowledge about causes and protective factors of disease
• Both for outbreaks and endemic diseases
• Easy to perform and analyse
• Needs thorough planning
• Next step after surveillance and outbreak investigations
Source population
The cohort study
unexposed
exposed
The cohort study
unexposed
exposed
Occurrence amongexposed (I1 or R1)
Occurrence amongunexposed (I0 or R0)
a
b
The cohort study: calculation
Exposed I1 = a / Nt1 = 16 / 125 pyar
Unexposed I0 = b / Nt0 = 8 / 120 pyar
IRR = I1 = a / Nt1 = 16/125 pyar = 1.92
I0 b / Nt0 8/120 pyar
Problems of cohort
• Large sample size required– At least if disease is rare
• Latency period• Time consuming• Loss to follow up• Changing exposure over time• Only one exposure• Ethical considerations• Cost
That is whythe case-control study is smarter
• Gives the same measure of causal effect as the cohort study (risk ratio RR or incidence rate ratio IRR)
• It is called odds ratio (OR)
• Easier, quicker, cheaper, smarter
Exposed
Unexposed
Source population
The case-control study
CasesExposed
Unexposed
Source population
CasesExposed
Unexposed
Source population
Sample
Controls
CasesExposed
Unexposed
Source population
Sample
Controls
Cases = the same as in cohort study
Controls = sample of the source population, with representative distribution of exposed and unexposed persons (or person-time)
CasesExposed
Unexposed
Source population
Sample
Controls
a
d
c
b
CasesExposed
Unexposed
Source population
Sample
Controls
a
d
c
b
d / Nt0 = c / Nt1
because sampled independent of exposure
24 / 120 = 25 / 125
The case-control study: calculation
IRR = I1 = a / Nt1
I0 b / Nt0
= a . Nt0
b Nt1
= a . d b c
d = c
Nt0 Nt1
Nt0 = d
Nt1 c
= 16 . 24 = 1.9225 8
Saving resources with case-control study
• In stead of following a cohort of 245 people for one year to wait for the 24 cases
• We investigated the 24 cases in order to divide them between– exposed a = 16 and– unexposed b = 8
• We chose 49 controls and investigated them in order to divide them between– exposed c = 25 and– unexposed d = 24
• The result– exactly the same as cohort study, but much easier
The odds ratio (OR)
• OR = Incidence rate ratio (IRR)
• OR = Risk ratio (RR)
• Cross product ratio: ad / bc
Exposed Unexposed
Cases a b
Controls c d
Summary of the case-control study
• Imagine a source population– Consists of exposed and unexposed people– Gives rise to cases (same as if cohort study)
• Control group is a sample from this source population– Independent of exposure status– Same distribution of exposed persons (person-time) as
in source population• Determine exposure status of cases and controls• Calculate odds ratio
– = risk ratio or incidence rate ratio if a cohort had been done
Challenges in case control study
• No measure of disease occurrence– Not risk R or incidence rate IR
• Difficult to define source population
• Difficult to sample controls correctly– Independently from exposure
• Recall bias– Cases remember differently from controls