Lecture #23 -Association and Causation

Question 1

What’s the importance of measuring potential confounders?

Answer 1

If don’t measure then can’t find out it was a confounder by doing the measurement (do this through stratification, multivariate analysis and standardisation?)

Question 2

(In stratification) What do we mean by “similar MoA”?

Answer 2

Use judgement

Question 3

Do we need to know how to calculate MoA in stratification?

Answer 3

Nope - just use the stratum-specific MoA (if similar MoA)

Question 4

Are effect modifiers on the causal pathway?

Answer 4

No and these are not things that distinguish the two e.g. people who have this AND the exposure are at a greater risk

Question 5

What is cause?

Answer 5

“If I do this, then this will happen”

Question 6

1. What’s the causal pie model and how does it arise?
2. What’s the whole pie called?
3. Each exposure is a _____ of the sufficient cause
4. So we call each of the exposures a ____ _____
5. Have a read of the health example of smoking and lung cancer on page 35
6. Can you have lots of pies for one outcome? Do we know them all?

Answer 6

1. Certain things need to happen together for an outcome to occur and together these exposures are SUFFICIENT to cause the outcome (will inevitably get outcome if have all three)
2. Sufficient cause of outcome
3. Component
4. Component cause
5. Yes for multiple and no, we don’t know all. If someone gets a disease, not random chance - they just had a sufficient cause we didn’t know about

Question 7

Why bother with causal pies?

Answer 7

Can use them as a guide to prevention

Find the proportion of cases that are due to each sufficient cause so we can focus on what to intervene

Question 8

Causal pathways and PAFs

1. Why can PAFs sometimes be over 100%?
2. What kind of PAFs do you aim to target with interventions?

Answer 8

1. Because things can overlap and cross over and so you’re sometimes adding things twice if the exposure is part of 2 different sufficient causes
2. Aim for exposures with high PAFs and the ones that are modifiable

Question 9

What is the prevention dream? What’s the most common example?

Answer 9

• When all sufficient causes share a component cause
• This is a necessary cause - the outcome can never occur without it
• Most common example is infectious disease (have to actually be exposed to the infectious agent to get disease)

Question 10

Complexity - what can you say about this?

Answer 10

We’re being told about simple examples but reality is much more complex. The cause of a disease isn’t just the last thing that precedes it.

Question 11

Summing up terms:

1. Sufficient cause
2. Component cause
3. Necessary cause

Answer 11

1. A cluster of component cause which are sufficient to produce the disease (whole pie)
2. A factor/exposure which in combo with other component causes makes up a sufficient cause
3. A component cause which is necessary for the outcome to occur. It must be part of every sufficient cause

Question 12

Can a sufficient cause be just one component cause?

Answer 12

Rarely ever the case

Question 13

When can we assess if the association is causal?

Answer 13

After considering internal validity - chance, bias, confounding (if these aren’t the reasons for results then look at causal)

Question 14

Bradford Hill Criteria

1. “What”?
2. Set of guide lines to aid what?
3. “Positive list” - what does this mean?
4. Which criteria is absolutely vital for assessing causal association?
5. Name the 7 guidelines (BESTCDS)
6. Which study gives the best evidence for causal relationship?

Answer 14

-