Causality Or Association? Flashcards
Why is it questionable that any exposures satisfy Koch’s Postulates?
- One exposure can cause several diseases e.g. SMOKING
- Most diseases are multi-factorial with many contributing causes
- Disease is usually the result of interaction between host, agent and environment
When is a cause ‘necessary’ and when is it ‘sufficient’?
- Necessary when the cause must ALWAYS precede the disease
- Sufficient when the cause can initiate the disease ON ITS OWN
- Causes can be necessary, sufficient, neither or both
Define ‘cause’ in an epidemiological context
Cause is an exposure or factor that increases the probability for disease
State Koch’s Postulates (1882)
- Agent must be present in every case of the disease by isolation in pure culture
- Agent must not be present in cases of any other disease
- Once isolated, agent must be capable of reproducing the disease in experimental animals and must be recovered from the experimental disease produced
Why is it important to differentiate between causal and non-causal factors?
- Attention to the reduction in risk factors that cause disease will have an effect on population health
- Attention to non causal factors will waste time, money and effort
Describe the steps of epidemiological reasoning
- Hypothesis
- Design an analytical study to test the hypothesis
- Test for validity by excluding association factors such as change, bias and confounding
- DOES THE STATISTICAL ASSOCIATION SHOW A CAUSE-EFFECT RELATIONSHIP???
What are the key assumptions in Epidemiology?
- That the disease does not occur at random
- That the disease have causal and preventable factors that can be identified through systematic investigation
What is meant by association?
- STATISTICAL DEPENDANCE between two or more events, characteristics or other variables
- The presence of association DOES NOT always imply a causal relationship
What is meant by reverse causality?
- Cause-Effect relationship exists in the opposite direction
- Believe X causes Y when in fact Y causes X
- This is a common problem in CASE CONTROL STUDIES because you cannot be certain that the exposure precedes the outcome
Define information bias
Error due to systematic differences in the measurement or classification of subjects in the groups being studied
- RECALL BIAS
- PUBLICATION BIAS
Define selection bias
Error due to systematic differences in the characteristics of the groups being studied due to differences in the way they were selected
- HEALTHY WORKER EFFECT
- ALLOCATION BIAS
What is the effect of reverse causality?
May result in true associations which occur other than the proposed causal link
What factors may cause apparent associations to arise?
- Chance
- Bias
- Confounding
What is meant by bias?
Deviation of the results from the truth due to systematic errors
How could you test for validity of an analytical study?
- Remove the assumption that the results were due to chance using p-value and 95% CI
- Excluding possible bias and confounding factors
What are the main Bradford Hills criteria for association factors?
- STRENGTH
- SPECIFICITY
- CONSISTENCY
What is meant by STRENGTH OF ASSOCIATION?
- A STRONG ASSOCIATION IS MORE LIKELY TO BE CAUSAL
- Strong associations are less likely to be explained by undetected bias or confounding factors
- However weak associations may still be causal
What is meant by SPECIFICITY OF ASSOCIATION?
- ASSOCIATION THAT IS SPECIFIC TO THE EXPOSURE-DISEASE ASSOCIATION UNDER INVESTIGATION IS MORE LIKELY TO BE CAUSAL
- However most diseases are multifactorial and several exposures can cause several different diseases
What is meant by CONSISTENCY OF ASSOCIATION?
- ASSOCIATION THAT HAS BEEN DEMONSTRATED IN MULTIPLE DIFFERENT STUDIES ON DIFFERENT GROUPS OF PEOPLE AT DIFFERENT TIMES IS MORE LIKELY TO BE CAUSAL
- Unlikely to explained by the same bias or confounding
- Lack of consistency may be due to differences in study designs or other causal factors
What Bradford Hulls criteria are specific to the exposure/outcome relationship?
- Temporal sequence
- Dose response
- Reversibilty
What is meant by TEMPORAL SEQUENCE?
- ASSOCIATION WHERE THE EXPOSURE IS DEMONSTRATED TO PRECEDE THE OUTCOME IS MORE LIKELY TO BE CAUSAL
- Strength of this criterion is in its corollary (if the outcome precedes the exposure then the exposure cannot be said to cause the outcome)
What is meant by DOSE RESPONSE?
- IF VARYING AMOUNTS OF EXPOSURE LEADS TO VARYING STRENGTHS OF ASSOCIATION WITH THE OUTCOME OF INTEREST THIS IS MORE LIKELY TO BE CAUSAL
- Unknown confounders and bias are unlikely to be operating to the same degree in the varying associations of the exposure and outcome
What is meant by REVERSIBILITY?
- ASSOCIATION IN WHICH REMOVAL OF THE EXPOSURE OR PUTATIVE FACTOR LEADS TO A REDUCED OR NON-EXISTENT RISK OF ACQUIRING OUTCOME IS HIGHLY LIKELY TO BE CAUSAL
- Strongest criterion for inferring causality but is sometimes difficult to demonstrate
What other evidence in Bradford Hills is used for inferring causality?
- Coherence of theory
- Biological plausibility
- Analogy
What is meant by COHERENCE OF THEORY?
- ASSOCIATION THAT CONFIRMS WITH CURRENT KNOWLEDGE AND THEORY IS MORE LIKELY TO BE CAUSAL
- However there are examples in biomedical science where causal links have been established with did not conform to contemperous thinking
What is meant by BIOLOGICAL PLAUSIBILITY?
- ASSOCIATION FOR WHICH THERE IS A BIOLOGICALLY PLAUSIBLE MECHANISM IS MORE LIKELY TO BE CAUSAL
- However this is limited by current knowledge of biological mechanisms
- Presence of a biological mechanism does not guarantee that an association or effect will exist
What is meant by ANALOGY?
- ASSOCIATION FOR WHICH AN ANALOGY OR INFERENCE EXISTS IS MORE LIKELY TO BE CAUSAL
- This is the weakest argument for inferring causality
How can results be validated externally?
Comparison to reference population and general population
How can results be validated internally?
- Assess results for non causal associations by excluding bias, confounding factors, chance and reverse causality
- Test features of causality to determine whether there is statistical evidence for a cause-effect relationship (STRENGTH, DOSE RESPONSE, SPECIFICITY, REVERSE CAUSALITY)
What is an ecological comparison?
Exposure and outcome are not linked on an individual basis
