PHRM3031 - Case Control Studies

Question 1

case-control study

Answer 1

purpose: to determine whether there is an association between an event or exposure and an outcome (eg.disease)
method: observation, prospective: identify incident cases and matched controls, retrospective: identify previous cases and matched controls

Question 2

case definition

Answer 2

people with the disease of interest in a defined population

Question 3

controls definition

Answer 3

people without the disease of interest

matched to cases for some characteristics

Question 4

1.selection

Answer 4

• select cases from the population at risk where the disease is present
• select at least one control for each case from the population at risk

Question 5

cofounding
definition

latent

Answer 5

a characteristic (or variable) is correlated with both exposure (risk factor) and the outcome (disease)
-common confounders are age, gender, SES
Latent: a variable that is sometimes not measured that affects both the exposure and outcome

Question 6

matching controls

Answer 6

match the control to the case so that they have similar characteristics

• aim to eliminate variation on confounding variables by:
• • restrict sampling to certain levels of characteristics
• -sample the comparison population and try to adjust in the analysis

Question 7

under matching controls

Answer 7

• not many characteristics (variables) are matched between cases and controls
• may lead to confounding
• statistical analysis must account for every variable that is not matched –> termed ‘adjust’
• adjusting for the effects of variables after the fact is not as good as prior matching of the variable

Question 8

over matching controls

Answer 8

match on variables so closely related to exposure that exposure rates in cases and controls become more similar compared to source population –> observed estimate of relative risk approaches 1 (no effect)

• why would matching variable be related to exposure?
  1. could be in chain of events from exposure –> disease
  2. some variables might be highly correlated (similar root causes)
  3. matching on diseases with the same treatment

Question 9

2. exposure - recall bias

Answer 9

• may be asked to remember exposure (retrospective data collection) if have disease (cases) may remember ‘differently’ to controls
• exposure data recorded by drs and other staff in medical record (if suspect a risk)
• disease may lead to exposure (esp. if medical treatment is the exposure) i.e early disease leads to Rx but research question is :does Rx lead to disease?

Question 10

3. outcome - measurement bias

Answer 10

• explicit and clear criteria for disease recognition management
• may have been change in practice over time when cases occurred
• adjust for change in method of measurement
• confirm diagnosis in cases and excluded diagnosis in controls

Question 11

odds ratio

formulas
odds
odds ratio
odds of exposure in cases
odds of exposure in controls 
odds cases/odds controls

Answer 11

odds=probability event/probability of no event = Pr/1-Pr
odds ratio=odds of exposure in cases (with disease)/odds of exposure in controls (without disease)
odds of exposure in cases = [a/(a+c)]/[c/(a+c)] = a/c
odd of exposure in controls = [b/(b/c)]/[d/(b+d)]=b/d
odds cases/odds controls = [a/c]/[b/d] = ad/bc

Question 12

interpretation of OR

Answer 12

OR is an approximation of the relative risk (if disease incidence is low i.e <1%)
OR=1 no association (if the 95% confidence interval includes 1)
OR >1.0 odds of exposure among cases is higher than odds of exposure among controls –> exposure is potentially a cause of disease
Or <1.0 odds of exposure among cases is lower than odds of exposure among controls –> exposure is potentially protective of disease

Question 13

when case control studies should be used

Answer 13

• are a useful tool for assessing the association of exposure (risk factor) with an outcome)
• only useful is the incidence of disease is low
• potential of substantial bias
• used for pharmacovigilance