Lecture #13 - Case-control studies (methods)

Question 1

What were two key limitations of cohort studies?

What are alternatives to go around these?

Answer 1

1. Can be insufficient with rare/slow to develop outcomes
2. Transient/acute exposures
   - not particularly good at measuring transient/acute exposures - things that come and go e.g. fatigue/intoxication - not always fatigued or intoxicated so cohort studied not v efficient for those
3. Use historical cohort study or looking at care control studies

Question 2

Case control studies dress what issues that come with cohort studies?

Answer 2

1. Designed for rare/slow to develop outcomes

2. Can examine acute/transient exposures

Question 3

What is the brief difference between cohort and case-control studied?

Answer 3

Cohort: ascertain exposure status, then find out outcomes

Case-control: ascertain status then find out exposures prior to outcome

Question 4

How to: Case-control studies: 5 steps

Answer 4

1. Identify source population (people you’re interested in studying)
2. Identify people with outcome (cases)
   - recruit another group of people from source population who have outcome (many if not all) = case group
3. Sample people w/o outcomes from same population (controls)
   - recruit another group of people from source population who don’t have outcome = control group
4. Measure exposure prior to outcome in case + control
   - how many cases have exposure and how many controls have exposure
5. Compare odds of exposure to calculate measure association (odds ratio)
   - (odds of exposure in cases)/(odds of exposure in controls)

Question 5

What is the logic of case-control studies?

1. If controls are a good representation of course population then they……
2. If the odds of exposure are different in cases (people who did develop outcome) compared to people who didn’t develop outcome then that suggests….

Answer 5

1. …..represent the odds of the exposure in that source population.
2. ……there’s some relationship - some association between exposure and outcome

Question 6

If odds of exposure are greater in the cases then…..

If odds of exposure are greater in the controls then…..

Answer 6

1. Exposure more common in cases (potential risk factor - greater likelihood of having exposure if case)
2. Exposure more common in controls (potential protective factor - having exposure protective factor against developing outcome)

Question 7

Why use the odds ratio and not the prevalence or incidence etc?

Answer 7

Have selected no. of people in study with and without outcome (go to book)

Question 8

Go to book and read over the example of case-control

Answer 8

-

Question 9

1. What is the odds ratio?
2. What is the null value?
3. What is the interpretation?

Answer 9

1. Ratio of odds instead of incidences (how many times as likely cases are to have exposure compared to controls)
2. Null value = 1 (no association)
3. Just like RR: “The EXPOSED GROUP were VALUE as likely to develop OUTCOME compared to COMPARISON GROUP”

Question 10

What is the index date?

Answer 10

• Pretend control had event on same date as case (or close to it)
• Apply to controls to say “that’s when they had their fake-crash”

Question 11

Case selection: important points

1. Normally, cases are defined by……
2. Need clear outcome….
3. Comprehensive….

Answer 11

1. By outcome so there’s only one outcome (very rare for case-control study to have more than one outcome)
2. Definition and identification
   e. g. death from Alzheimer’s or Schizophrenia (else will get false causes in study and that’ll cause bias)
3. Comprehensive case findings
   e. g. routine data collection, medical records
   - Wanna try get as many cases as you can find - if miss out people bc those ppl could be systematically different in some way so use routine data/medical records

Question 12

Control selection: important points

1. Need to represent exposure…..
2. Control must be capable of…..
3. Often select multiple controls per case…..

Answer 12

1. Need to represent exposure distribution in source population
   - this is very hard to do - if come up w/o representative - your comparison and association may not be as valid
2. Controls must be capable of becoming a case
   e. g. won’t include women as control for studying prostate cancer
3. Often select multiple controls per case for statistical power

Question 13

Controls selected where from?

Answer 13

Paraphrase what’s written in your book

Question 14

Exposure measurement: important points

1. Need to measure _____ period before _______
2. Differential recall - explain it
3. Exposure measurement must be _____

Answer 14

1. Exposure period before outcome (don’t measure exposure ages before or ages after - wanna know just before crash)
2. Cases trying to work out what made them sick (controls wouldn’t probe this much into it so odds of exposure can be artificially inflated)
   - Outcome may affect recall ability (e.g. head injury)
3. Exposure measurement must be comparable - collected in same way
   - Dead cases vs alive controls (can’t interview dead cases)
   - Interviewers may act differently for cases and controls (don’t probe cases more for answers else you’ll get difference in odd ratio just due to difference in method of asking)

Question 15

Briefly, what are the 4 good things about case-control and the 3 bad things?

Answer 15

Good:

1. Rare outcome, transient exposure (only look at people you need to look at)
2. Multiple exposures
3. Temporal sequencing
4. Often combatively quick and inexpensive (not having to follow up)

Bad:

1. Usually can only study one outcome
2. Difficult to select appropriate control group
3. Can be susceptible to selection and recall bias