EBM

Question 1

Name 2 conditions in which a case-control study design should be used

Answer 1

1. The outcome of interest is very rare
   
   • Cohort study would need too large of a sample size to be feasible
2. The latency period (time between exposure and outcome) is very long
   
   • Cohort study would need to follow people for too long a time period to be feasible

Question 2

When selecting cases for a case-control study, is it preferable to use incident cases (newly diagnosed) or older cases (diagnosed many years prior)?

Why?

Answer 2

Incident cases.

This ensures that the exposure occured BEFORE the outcome of interest

Question 3

Describe the ideal population of the controls versus population of the cases

Answer 3

The controls should be from the same population as the cases

ie. controls must have been eligible to develop the “case” outcome, and must have similar traits overall (no confounding)

Question 4

Compare the number of outcomes and exposures that can be assesed with

1. Case control studies
2. Cohort studies

Answer 4

• Case-control studies can only look at 1 outcome, but can look at many exposures
• Cohort studies can only look at 1 exposure, but many different outcomes

Question 5

Define information bias in the context of case-control studies

Answer 5

Information bias is when our determination of whether a person had the exposure is systematically flawed

Question 6

Describe 3 types of information bias

Answer 6

• Recall (people with outcome more likely to remember exposure)
• Reporting (people with outcome report exposure that they believe to be the cause and don’t report exposure that contradicts this belief)
• Observer (researcher asks questions differently b/c they believe case more likely to have exposure in past)

Question 7

Name two ways to reduce confounding in a case-control study

Answer 7

• Restrict study entry
• Match controls for cases

(*can’t use randomization)

Question 8

What form of multivariate analysis is used for a categorical outcome?

Does the exposure need to be categorical or continuous?

Answer 8

Logistic regression

• Categorical outcome
• Categorical or continuous exposure

Question 9

The results of a cohort study will be ______ (RR v OR)

Answer 9

The results of a cohort study will be RR!

*A cohort of pirates says aRRg!

Question 10

The results of a case/control study will be __ (RR v OR)

Why?

Answer 10

The results of a case/control study will be ODDS ratio

Because the number of people in our case and control groups is not representative of the prevalence in our population (we cherry picked our study group!)

Question 11

Our case group has 100 people, 25 of whom were smokers.

Our control group has 200 people, 30 of whom were smokers.

What is the odds ratio?

Answer 11

OR = (25/75) / (30/170)

OR = [# cases exposed / # cases not exposed] /

[# controls exposed / # controls not exposed]

Question 12

Interpret an OR > 1

Answer 12

If OR > 1, the odds of exposure were greater among the cases than among the controls

Question 13

Interpret OR = 4.3

Answer 13

The cases were 4.3 more likely to have been exposed than the controls

Question 14

Describe the effect of prevalence on our interpretation of an OR

Answer 14

OR is a good approximation of RR when the outcome is rare.

OR is a poor approximation of RR when the outcome is common.

Question 15

What is the formula for NNT

Answer 15

NNT = 1/Absolute Risk Reduction