Confounding

Question 1

What is the definition of confounding?

Answer 1

Distortion of the relationship between an exposure and outcome due to shared relationship with something else

Question 2

What are the potential effects of confounders?

Answer 2

Confounders can either increase association between exposure and outcome, or decrease association between exposure and outcome

Question 3

What is a cohort study?

Answer 3

Where you start with a group free of disease and measure exposure over a time period, and compare outcomes

Question 4

What is a case-control study?

Answer 4

Start with group who have the outcomes and look back to see exposures, and compare exposures

Question 5

How is confounding addressed in design and analysis?

Answer 5

In 4 ways:

1. Restriction: limit the participants of your study who have possible confounders
   - Means you have less data and difficult with multiple confounders
2. Matching: create a comparison group that is matched on the possible confounder
   - Make case and control group as similar as possible on the confounder and then ask about exposure status
   - Used for strong confounders like age and sex
3. Stratification: analyse exposure:outcome association in different subgroups of the confounder
   - E.g If you are looking at joggers vs non-joggers and risk of CHD, confounding element may be pies (non-joggers may eat more pies). So, you look at the relationship of jogging and CHD in pie eaters and non-pie eaters separately then look at the overall picture
   - Recombine data and use a weighted average of the strata
   - Limitations: to take into account all confounders would require lots of strata and you may run out of data to fill all possible options in your strata
4. Multiple variable regression: you can adjust for the effects of multiple confounders, try and produce a linear model between the outcome and different exposures
   - Allows time for adjustment of estimates for confounding

Question 6

If the confounder is on the casual pathway between exposure and outcomes, then what can be said?

Answer 6

It it NOT a confounder

Such as exposure: confounder: outcome being Number of sexual partners:HPV+ve: Cervical cancer

Question 7

What is standardisation? When is it used?

Answer 7

It is another way to limit confounding, often used to control differences in age groups when comparing rates of disease in two populations with different age structures