Lecture 12 Cause and Effect

Question 1

Some key points about cause and effect in physics:
1. Time between cause and effect is short
2. Cause is necessary and sufficient for effect
3. Replication easy
How does cause and effect in disease differ?

Answer 1

1. Time between cause and effect is unpredictable
2. Cause is necessary (e.g. infection by M. tuberculosis to get clinical tuberculosis) but not sufficient
3. Cause increases frequency of effect
4. Replication difficult/impossible

Question 2

All causes of disease are really risk factors.

Define risk factors

Answer 2

Risk factors is anything that increases the likelihood of the disease

Question 3

What is a randomised controlled trial?

Answer 3

Participants are randomly allocated to either placebo/control or treatment condition, then measure the effect. Random/blind allocation reduces selection bias.

Question 4

What are the criteria for a causal association in biology?

Answer 4

1. Strength of causal association: quantitative strength, relative risk >2 (risk has to at least double when compared to background).
2. Consistency of causal association: same results when use different study designs, groups, places - REPRODUCIBILITY
3. Specificity of the association: cause is necessary and sufficient for effect, never met in biology
4. Time sequence: exposure before disease
5. Biological gradient: dose-response association
6. Plausibility: mechanism may not be understood, but if strength of association great enough can be accepted
7. Coherence: with existing knowledge of disease
8. Experiment: can changing variables or using preventative action reduce rate of disease
9. Analogy: examples of previous scenarios can be put forward to support hypothesis

Question 5

What are the possible types of association?

Answer 5

No association.
Artifactual: by chance or bias
Indirect association: confounded variables
True causal association

Question 6

If the association is independent, do we accept or reject the null hypothesis?

Answer 6

No association so we accept null hypothesis.

Question 7

If association is independent, what is the next step we consider in statistics to ensure this independent association is true?

Answer 7

Did the experiment have enough statistical power.

Question 8

What is artifactual association?

Answer 8

Statistical significance shown but is irrelevant/not true. May be caused by chance, this can be eliminated by statistical testing. Or may be caused by bias in study design e.g. interview bias, recall bias

Question 9

What is indirect association?

Answer 9

A variable confounds a relationship if it is itself associated with both the exposure and disease
E.g. age, education, socio-eco
E.g. age confounds a relationship as it is associated with both exercise and myocardial infarction

Question 10

How can we deal with confounding variables?

Answer 10

1. Randomisation: achieve similar composition in both groups
2. Matching: males to males
3. Statistical methods: ANCOVA

Question 11

Explain how ANCOVA works.

Answer 11

Effect of age is not included by performing analysis separately at each level of continuous variable (confounder) but simultaneous analysis at each level.