Confounding and Causation

Question 1

What is a confounding factor?

Answer 1

Something that distorts the relationship between an exposure and outcome

independently associated with both outcome and exposure
→ but not a factor in the causal pathway

e.g. smoking is a confounder in the association between coffee consumption and heart disease

Question 2

What determines how strong of an influence a confounder has

Answer 2

Strength of confounder - disease link or exposure link

Prevalence of confounder in study population

Question 3

When is it determined that a confounder is present?

Answer 3

If adjusting for the confound changes the OR or RR

- Typically around a change >10% is deemed significant

Question 4

What are methods of controlling for confounding in the DESIGN of a study?

Answer 4

Restriction
- Restrict study design or analysis to subjects without potential confounder

Matching
- Potential confounders distributed equally in each group

Randomisation
- Participants have an equal chance of being in the intervention or control group → 2 groups similar in all aspects except intervention

Question 5

What are methods of controlling for confounding in the ANALYSIS of a study?

Answer 5

Matching

• Look at pair of people
• Take exposed cases and non-exposed controls and compare them to non-exposed cases and exposed controls
• -> beware of ‘’Over-matching’’ → more matching = smaller sample size = less generalisability

Stratification
- Data divided into groups (strata) according to levels of suspected confounder

Multivariable Methods

• Logistic Regression –> Allows us to produce OR for each variable in the model
• -> OR are adjusted for all the other variables in the model

Question 6

What is causation?

Answer 6

Factor that plays an essential role in producing an occurrence of a health outcome

If identified can prevent outcome by interrupting or modifying the cause

Question 7

What is component cause

Answer 7

Contributes towards disease causation but not sufficient to cause on its own

Question 8

What is Necessary cause

Answer 8

Any component necessary to cause the disease

Question 9

What is Sufficient cause

Answer 9

Group of components/causes that will inevitably produce disease

Question 10

What is the Bradford Hill Criteria for?

Answer 10

Criteria for judging assessing the causality of a exposure

CANNOT prove causality only support it

Question 11

What are the 9 Criteria in the Bradford Hill Criteria

Answer 11

TSCDBSEAC

Temporality → cause precedes effect → length of time consistent with biological mechanism

Strength of Association = Estimated using RR / OR
→ Strong = reduce likelihood that it’s due to confounding, increases likelihood relationship is cause and effect

→ Weak = difficult to exclude alternative explanations (can still be causal)

Consistency → replicated results in various studies with different study designs and populations

Dose-response → change in level (or duration) of an exposure is related to a change in risk of developing the outcome/disease

→ gradient of risk associated with degree of exposure → some disease require threshold exposure to develop

Biological Plausibility → Biological mechanism by which exposure alters risk of outcome

Specificity → Association found between one exposure and one outcome, but not for other exposures / outcomes

exceptions to this criterion - e.g. smoking = cause many many disease may have >1 cause

Experiment → Association shown in experimental study carries greater weight (intervention studies)

Coherence → The association is similar to a known causal association

Analogy → A cause and effect interpretation should not conflict with known facts