Confounding - I

Question 1

what is confounding?

Answer 1

a mixing or muddling of effects when the relationship we are interested in is confused by the effect of something else - the confounder

risk factors party together - if someone has one risk factor, they are likely to have multiple

Question 2

what are the properties of a potential confounder?

Answer 2

1. independently associated with the outcome
   - a risk/protective factor for the outcome by itself (regardless of the exposure status)
2. independently associated with the exposure
   - different proportions of people with potential confounder across exposure groups
3. not on the causal pathway
   - not the mechanism by which the exposure affects the risk of the outcome (not how the exposure affects the outcome)

Question 3

what can confounding do?

Answer 3

• over-estimation of a true association
• under-estimation of a true association
• change direction of true association (Simpsons paradox) - risk factor becomes protective factor and vice versa
• give appearance of an association when not one - go from null to something else

Question 4

describe the individual impacts of confounding for harmful exposures and beneficial exposures

Answer 4

Harmful exposures:
1. study RR is higher than true RR
- confounding resulted in over-estimating the true harmful effect of the exposure (association appears stronger than it really is, RR is ‘further away from the null’)
2. study RR is lower than true RR
- confiding resulted in under-estimating the true harmful effect of the exposure (association appears weaker than it really is, RR is ‘closer to the null’)

Beneficial exposures:
1. study RR is lower than true RR
- confounding resulted in over-estimating the true protective effect of the exposure (association appears stronger than it really is, RR is ‘further away from the null’)
2. study RR is higher than true RR
- confounding resulted in under-estimating the true protective effect of the exposure (association appears weaker than it really is, RR is ‘closer to the null’)

Question 5

how do we identify potential confounders?

Answer 5

plan ahead:
- collect information on all potential confounders
- use literature to identify known and suspected risk factors for outcome
- collect information on factors strongly associated with exposure, regardless if known factor
looking for potential confounders in a study:
- look for imbalance in potential confounder between groups
- if there is a difference between cases and controls then it means that the risk/protective factor is independent of the exposure and therefore fits the criteria for confounding

• if you don’t measure it, difficult to do anything about it later

Question 6

what are the mechanisms of controlling confounding in the study design?

Answer 6

• randomisation
• restriction
• matching

all attempt to make groups being compared alike with regard to potential confounder(s)

Question 7

describe randomisation as a way of controlling confounding

Answer 7

Only used in RCTs
- the strength of randomisation is that this applies to known and unknown confounders

BUT
- works best with large sample size
- need equipoise
- need intention to treat analysis (to maintain comparability, analyse as randomised)

Question 8

describe restriction as a way of controlling confounding

Answer 8

restrict to one stratum of a potential confounder (only recruit people within a particular bracket, like age, make it within 10 years or something)
- easy and applies to all study designs

BUT
- can reduce generalisability
- reduces number of potential participants
- potential for residual confounding with imprecisely measured (or broadly defined) confounders
- usually only one potential confounder

Question 9

describe matching as a way of controlling confounding

Answer 9

choose people to make the control.comparison group have the same composition as the case/exposed group regarding the potential confounder
- usually used in case control studies

individual vs. frequency:
individual:
- each case matched with one or more controls having the same confounding variable charactaristsics
frequency:
- matching at an aggregated level
- eg. 20% of cases are males between 45-50 years, so get 20% of controls that are males 45-50 years old

positives:
- useful for difficult to measure/complex potential confounders
- can improve efficiency of case-control studies with small numbers (need fewer people to detect an association if there really is one

BUT:
- individual matching can be difficult and limit the umber of potential participants
- need special matched analysis for individual matching
- otherwise will under-estimate the measure of association
- NOT a form of controlling confounding in the analyses

Question 10

what is the problem with controlling confounding in the study design

Answer 10

• can’t assess association between potential confounder and outcome
• can’t assess whether truly a confounder
  (because you have already taken it into account beforehand?? so can’t measure its actual effect??)

Question 11

need to be able to:

Answer 11

• describe the criteria for a potential confounder
• identify potential confounders and explain how they might fit these criteria
• explain what impact confounding might have on a study’s findings
• explain what they are, when hey are used, and their strengths and weaknesses
• recognise when they are used