Application of causality Flashcards
How to discuss causality?
Guidelines have been proposed for judging whether an association is causal or not
- Positive evidence: Bradford Hill’s criteria
- Negative evidence: mistakes, biases
- Alternative explanations
Bradford Hill’s criteria
Temporal sequence
Strength
Dose-response gradient
Consistency
Specificity
Biological plausibility
Coherence
Experimental evidence
Analogy
Temporal sequence
o The cause always precedes the effect (NO EXCEPTIONS)
o This is the only necessary condition
Strength of the association
Magnitude of the effect, not the p value
(a mayor RR, + probable es la asociación)
How many times is the disease more frequent in those exposed in relation to those who don’t have the exposure
The stronger an association, the harder it is to explain the association as artifacts of chance or biases. Exceptions: some weak associations are causal and some strong associations are not causal. Usually, strong associations are more likely to be causal.
Dose-response gradient
o As the dose of exposure changes, the risk of the outcome also changes
o The higher the exposure the higher the effect
o There may be a sharp increase in risk for specific outcomes at low-moderate doses of a given exposure that tapers off at higher doses / no adverse effect until saturation of detoxification mechanisms are overcome = increased risk only at higher doses.
Consistency
• Replication of findings in different studies and in different populations
• Lack of consistency does not preclude causation: some association may arise only under certain circumstances, or study designs
• Consistency does not warrant causation: biases may be present across studies
Meta-analysis
Specificity
o It is the weakest criteria.
o The relationship of the exposure with the disease
• Specificity of the outcome (a cause having a single effect)
• Specificity of the exposure (effect having a single cause)
• Specificity due to predisposition or susceptibility
Biologic plausibility
There is a physiopathological mechanism that explains the observed association
(Causal hypothesis is reasonable or realistic not only by epidemiological knowledge but also because other human, animal and tissue studies, and current understanding of the biology, pathology, toxicology, etc. ‘a priori’ support this causal effect.)
Coherence
Observed results should not contradict gº knowledge of the natural history and biology of a disease
Experimental evidence
• Experiments conducted in animals
• Clinical trials
• Cure of the disease after cessation of exposure
Analogy
Effect of similar factors is considered.
(Inferences are drawn based on what is known about other exposure-disease relations.)
Negative evidence
If there is bias: overestimation or underestimation?
Alternative explanations
Ex: a higher nº of storks in a city is associated with a higher nº of births. The alternative (non-causal) explanation is that the larger the size of the city, the more bell towers and storks there are, and the larger the size of the city, the more births.
How do we establish causal associations?
o If the association follows Bradford Hill’s criteria = positive evidence for a causal association
o We always need to search for an alternative non-causal explanation
o If we find any biases and errors, then we will have negative evidence
