Association and Causation Flashcards
What is association?
statistical dependence between 2 variables - the degree to which rate of disease in persons with specific exposure is higher/lower than rate without exposure
What are the possibilities for an association?
chance
bias
confounding
THEN CONSIDER CAUSAL RELATIONSHIP
How is the role of chance assessed?
Perform statistical significance test by calculating CIs
If p<0.05 then result of study not due to chance
What are CIs?
range within which the true value is expected to lie within a given degree of certainty
What is bias?
systematic error leading to incorrect estimate of effect of exposure on development of disease/outcome of interest
What changes depending on the nature of systematic error?
whether the observed effect is above or below the true value
Does increasing sample size reduce bias?
no
What are the two types of bias?
selection - occurs when there is a systematic difference between the characteristics of those who were selected for the study and those not
- non response bias
- healthy entrant effect
- lose to follow up (attrition bias)
measurement/information - when measurements/classifications of disease/exposure are inaccurate
- recall bias
What is a confounder?
any factor which is believed to have a real effect on the risk of disease under investigation and is also related to the risk factor under investigation
- factors with direct causal link to disease (smoking/lung cancer)
- factors the are good proxy measure of more direct unknown causes (age and social class)
How is judgment of causation made?
- observed association between exposure and disease is valid
- totality of evidence taken from several sources supports a judgment of causality
What are the Bradford Hill criteria to consider?
STRENGTH
- strength of association measured by magnitude of relative risk
- strong association more likely causal than weak one
- but weak doesnt mean non-causal
CONSISTENCY
-more likely to be causal if similar results in different populations using different study designs as it is unlikely that studies will be subject to same type of errors
SPECIFICITY
- if particular exposure increases the risk of a certain disease but not the risk of others strong evidence in favour of causality
(e. g. Mesothelioma on asbestos) - However one-to-one relationships are rare and lack of specificity should not be used to refute causality
(e. g. smoking and many diseases)
TEMPORAL RELATIONSHIP
ABSOLUTELY NECESSARY
- for a putative risk factor to be the cause of disease it must precede the disease
- easy to establish this from cohort studies, difficult from cross sectional/case control studies when measurements of cause and effect made at same time
- Reverse time order is not evidence against hypothesis
DOSE RESPONSE RELATIONSHIP
- further evidence if increasing levels of exposure lead to increasing risks of disease
- but some causal relationships show a single threshold jump
PLAUSIBILITY
- association more likely to be causal if consistent with if consistent with other knowledge (animal experiments, biological mechanisms)
- but take lightly as lack of plausibility may be lack of scientific knowledge
EXPERIMENTAL EVIDENCE
- evidence on humans or animals
- human evidence seldom available and animal research relates to different species and different levels of exposure
COHERENCE (*)
- implies cause and effect does not conflict with what is known of natural history
- but absence of coherent info due to conflicting info is not evidence against causality
ANALOGY (*)
- source of more elaborate hypotheses about association
- absence of this is lack of imagination/experience not falsity of hypothesis
What is a factor not considered by BH?
reversibility
- if cause is removed is consequence affected?
