Lecture 28: Confounding II

Question 1

“Risk factors party together”. What does this mean?

Answer 1

people who have one risk factor for an outcome also tend to have other risk factors for that outcome

Question 2

As well can controlling for confounding in the study design, when else can we control for confounding?

Answer 2

in the study analyses

Question 3

What are three ways to control for confounding in the study analyses?

Answer 3

• stratification
• multivariable analysis
• standardisation

Question 4

What is stratification?

Answer 4

calculating the measure of association for each stratum (level of potential confounder eg age strata could be 20-29, 30-39) of potential confounder and comparing them

Question 5

What are the steps to stratification?

Answer 5

1. calculate the measure of association between the exposure and the outcome (the crude, univariate, unadjusted)
2. divide the potential confounder into strata
3. for each stratum, calculate the measure of association between the exposure and the outcome (ie. calculate the stratum specific measures of association)
4. compare stratum specific measures of association

Question 6

What are some advantages of stratification?

Answer 6

• easy for a small number of potential confounders with limited strata
• can evaluate the impact of confounding
• can identify effect modification

Question 7

What are some disadvantages of stratification?

Answer 7

• can leave residual confounding (distortion that remains after controlling for confounding in the design and/or analysis of a study)
• not feasible when dealing with lots of potential confounders with many strata

Question 8

Define multivariate analysis

Answer 8

statistical method for estimating measure of association whilst controlling for multiple potential confounders

Question 9

Define standardisation?

Answer 9

this is done when age differs between the populations and the disease risk varies by age

Question 10

What are some potential issues with controlling for confounding in the study analysis?

Answer 10

• residual confounding (using strata that are too wide eg. 65+)
• you can only control what you have measured

Question 11

If you controlled for confounding in the design phase by matching, what do you need to do in the analysis phase? Why?

Answer 11

a matched analysis otherwise you will underestimate the measure of association

Question 12

How can you evaluate confounding?

Answer 12

you need to compare crude and adjusted measures of association

Question 13

How do we know if confounding has occurred? Which MoA should we use?

Answer 13

When we compare the crude and adjusted MoA, if they are not equal, there is confounding present. You should calculate and use the adjusted MoA

Question 14

How do we know if confounding has not occurred?

Answer 14

When we compare the crude and adjusted MoA, if they are equal, there has been no major confounding. We can therefore just use the crude MoA

Question 15

How much change indicated confounding?

Answer 15

The general guideline is if controlling for confounding changes the MoA by more than 10%, confounding has occurred

Question 16

What is effect modification? Give an example

Answer 16

it is all about stratification and occurs when an exposure has a different effect among different subgroups. Effect modification is associated with the outcome but not the exposure.
For example, imagine you are testing out a new treatment that has come onto the market, Drug X. If Drug X works in females but does not work in males, this is an example of effect modification.

Question 17

How can you identify effect modification?

Answer 17

if the stratum specific MoA are different to one another, this is an effect modifier (but if they are similar to one another, this is a confounder, and then you compare to the crude MoA)