case-control studies Flashcards
cases
case-control studies
have disease of interest
controls
case-control studies
don’t have disease of interest but COULD develop it
pros
case-control studies
- cheap bc smaller sample size
- good for diseases where medical care is sought
- provides leads for follow-up
- good for disease with long-lattency (takes long time to develop, we don’t have to wait bc it is retrospective)
- rare disease
- multiple exposures
- FAST
- good for rapid onset disease (can sort out what caused vs what is result of something)
cons
case-control studies
- retrospective –> biase
- need records
- can’t establish, risk, prevalence, incidence
- can only study 1 disease
- can’t determine cause and effect
- info on confounders may not be available
- cases may search for cause of their disease and over-report exposure
- assembling casees may be hard
- identifying control might be hard
when do we use case-control studies
to examine a possible relationship btwn exposure and disease
start with
case-control studies
cases and controls, then measure exposure
odds ratio =
case-control studies
ad/bc
OR = 1.2
the odds of using artifical sweeteners is 1.02 times greater in bladder cancer cases than in controls
if exposure is associated with disease, we expect
more exposed in cases than controls
OR < 1
case-control studies
protective factor
OR = 1
case-control studies
no association
OR = 2
case-control studies
cases twice as likely as controls
95% confidence interbal
case-control studies
if 1 is included, OR is not significant
if 1 is NOT included, OR is significant
OR provides good estimate of risk when
case-control studies
- controls are representative of target population
- cases are representative of all cases
- frequency of disease in population is small, disease is rare
design: finding cases
case-control studies
- define case conceptually and enroll in a specific time period
- tumor registry or vital statistics registry
- medical facilities
- hospital cases (multicenter is best)
- doctor’s office
- community patient registry
sources of controls
case-control studies
non hospitalized controls:
* community: can be probability sample
* neighborhood controls
* random digit dialing to match neighborhood
* friend control
hospitalized controls:
* all other patients
* defined diagnosis
exposure
case-control studies
- exposure must be designed same in case and control
- if looking through records, don’t let researcher know hypothesis
- assure no routine recording of variable of interest in records
- avoid physical measures that may be influenced by diseased state
wording of questions
- need to be sure exposure preceded disease
- diet and colon cancer: ask about diet before symptoms
- overian cancer and weight: ask about weight several years before diagnosis
- may need to exclude ppl w/ symptoms >1 yr ago
recall of exposure limitations
case-control studies
- limitations in recall
- recall bias
limitations in recal
case-control studies
ppl might not know specific info or remember
recall bias
case-control studies
cases may recall info to ggreater extent than controls
differential recall may lead to artifactual relationship
use of multiple controls
case-control studies
2 or 3 same controls per case increases power of study
controls of diff types may help (if looking at risk for brain tumors have a normal control and other cancer control)
nested case control study
case-control studies
case control nested in cohort study
population in defined cohort with baseline surveys and samples followed over time
cases: those who develop disease over time
advantages of nested case control
data collected before disease develops (no recall bias)
cheaper than analyzing all samples for all cohort members
nested case control start with
defined cohort
overtime, categorize into develop disease (cases) and have not developed (controls)
what is coffee drinking incases is greater than that in controls?
this is a diff in exposure
matching will fix
matching
match controls and cases for confounders (age, sex, race, SES, job)
group matching
select controls so proprtion of controls have same characteristic as a proportion of cases
individual matching
for each selected case, a control is selected who is similar to the case in terms of the matching variables
concordant pairs
pairs in which both case and control are exposed
OR
neither exposed
discordant pairs
case exposed, control not
OR
control exposed, case not
analysis of case-control
logistical regression analysis
