6 - Case control studies

Question 1

Case-control study

Answer 1

Observational study on individuals, which compares past ‘exposures’ of those with and without disease

Question 2

Rare events - cohort or case-control?

Answer 2

Impractical to study rare outcomes n cohort study as would need to be huge - case-control is good for rare events as can select individuals know to have the disease already

Question 3

Considerations when selecting cases

Answer 3

Clear case definition needed
Potential sources - routine testing, notification, registration, case-finding

Question 4

Considerations when selecting controls

Answer 4

Supposed to represent the target population - give an estimate of the expected level of exposure
Must represent the same population as the cases
Must be at risk of being a case
Can select from the population, hospital, or rest of cohort (in nested designs)

Question 5

What measure of disease frequency in case-control study

Answer 5

None! Proportion of diseased individuals is set by the design

(unless nested case-control within cohort - could calculate incidence)

Question 6

Why do a Nested Case-Control?

Answer 6

Cost-effective approach for large cohorts
for example:
* Use of baseline biological material (Blood samples, DNA)
– finite resource and often expensive to analyse
* Further data needs to be gathered on those who
develop the disease of interest.
– Additional exposure data may need to be acquired
– May be costly in terms of time and money

Question 7

Case-cohort design

Answer 7

Cohort study
Pre-select a subset of the cohort - more detailed follow up from the outset
Use this group as the controls for multiple case sets

Question 8

Purpose of matching

Answer 8

Assures comparable distribution of known confounders among cases and controls
Multiple matched controls per case can increase power

Question 9

Cautions when matching

Answer 9

Effects of matched factors cannot be estimated
Can restrict control selection (not enough farms exist!)
Overmatching - can make cases and controls too similar so no differences

Question 10

What measures can be calculated from case-control?

Answer 10

Odds of exposure (prevalence of Exposure in each group)
Odds ratio measures strength of association
(read down columns of disease status - prevalence of disease is fixed by study design)

Question 11

What measures can be calculated from cohort study?

Answer 11

Cumulative incidence (risk), incidence rate, relative risk.
Read across rows - compare incidence - RR

Question 12

Odds calculation

Answer 12

Odds of exposure in cases = A/C (both over A+C- cancel out)
Odds of exposure in controls = B/D (both over B+D - cancel out)

Odds Ratio = A/C / B/D = AD/BC

Question 13

Interpreting odds ratio

Answer 13

Case control determines the rate of exposure among diseased / non diseased groups
Interpretation strictly is the odds of an individual with X (disease) having been exposed to Y (exposure) are A times higher than that of a control
Often read as the odds of an individual with Y (exposure) having X (disease) are A times higher than without

Question 14

Reverse causality bias

Answer 14

Ordering of events can be unclear
Can try to disregard exposure shortly before diagnosis eg T2DM diagnosis vs antibiotics

Question 15

Selection bias

Answer 15

Systematic difference in characteristics between study participants and the population from which they were selected, or between groups compared within the study (e.g. cases and controls)
eg deprived patients less likely to participate, more likely lost to follow up

Question 16

Recall (response) bias

Answer 16

A type of information bias
Systematic error caused by differences in accuracy or completeness of past events or experiences (interviews, relying on individual recall)
Remedy - obtain exposure data rom sources independent of the subjects own reporting
Can cause misclassification eg mothers recalling ear infections

Question 17

Case-control advantages

Answer 17

Relatively cheap, quick and easy to carry out
– No loss-to follow up
– Particularly suitable for rare events or those with long lag between exposure and outcome

Question 18

Case control disadvantages

Answer 18

Potential for recall bias
– Timing of events cannot always be reliably established
– Cannot assess incidence (proportion with disease is fixed as part of the study design)
- reliance on recall or records to determine exposure status;
- confounders;
- selection of control groups may be difficult