Week 8 - Association vs Causation Flashcards
Who first thought of causality and what did he believe?
- Aristotle
2.Scientific knowledge requires explanations to state
the causes of outcomes - Cause —> effect
What criteria does an exposure-outcome association have to fulfil to be seen as valid?
- Chance has to be excluded as a possible explanation for the association
(p-value and 95% CIs provide this information) - Bias has to be excluded as a possible explanation for the association
(thorough check for errors during sample selection and assessment of
variables) - Confounding has to be excluded as a possible explanation for the
association (identify and adjust for all potential confounders)
Does valid association mean causation?
-Even a valid association, CANNOT automatically prove causation
-Association ≠ causation
What does causal association mean?
-A causal association implies that the exposure directly or indirectly causes (or
contributes to the development) of the outcome of interest
-Specific criteria exist in order to help decide about the presence of a causal
association
What are the Bradford Hill criteria for causality?
- Temporality
- Strength
- Biological gradient
- Reversibility
- Consistency
- Plausibility
(First 6 are the most important!) - Analogy
- Coherence
- Specificity
Explain Temporality as a Hill criteria
Exposure must precede outcome
Example 1: A study found an association between early life pesticide
exposure and Parkinson’s disease in old age temporal association
Example 2: A study found an association between current BMI and current
blood pressure non-temporal association
Explain Strength as a Hill criteria
The stronger the association between exposure and outcome, the
more likely it is to be causal. Strong associations are more likely
(though not necessarily) to be causal.
-Example 1: A study found a Risk Ratio (95% CIs) for the association
between smoking and lung cancer of 4.00 (3.70; 4.35) strong association
-Example 2: A study found a Risk Ratio (95% CIs) for the association between
sunbathing and lung cancer of 1.10 (1.05; 1.15) weak association
Explain biological Gradient as a Hill criteria
Dose-response association between exposure and outcome
Explain Reversibility as a Hill criteria
- Removing (or reducing) the exposure reduces the incidence (or severity)
of the outcome
-Example 1: U.V radiation has been found to be associated with skin cancer.
Applying sunscreen (i.e. reducing the exposure to U.V radiation), reduces the
incidence of skin cancer reversibility existsin association
-Example 2: Sugar consumption has been found to be associated with type 2 diabetes.
Reducing the amount of sugar in the diet does not seem (so far) to reduce the incidence of type
2 diabetes reversibility does not exist in association
Explain Consistency as a Hill criteria
An association is observed in multiple occasions (i.e. replication of
results in different populations and using different study designs)
-Example 1: A systematic review on the association between physical activity
and type 2 diabetes found 30 recent studies on the topic, from which 28
showed a statistically significant inverse association consistent association
-Example 2: A systematic review on the association between red meat intake
and type 2 diabetes found 10 recent studies on the topic, from which 3 showed
a statistically significant direct association not consistent association
Explain Plausability as a Hill criteria
-There is a theoretically sound (pathophysiological) explanation for the
observed association (i.e. can we identify a biologically plausible mechanism
underlying the observed association?)
-Example 1: Observed direct association between salted fish and stroke. The high amounts of
salt in salted fish increase blood pressure, which in turn increases the risk of stroke
biologically plausible association
-Example 2: Observed inverse association between smoking and Parkinson’s Disease. Up to
now not a single substance in tobacco smoke has been found to be involved in any pathway in
the pathophysiology of PD not biologically plausible association (note: always based on
current evidence)
Explain Analogy as a Hill criteria
There are other analogous observed findings in the literature
NOT USUALLY USED
Example: There is a clear association between smoking and Cardiovascular
Disease in the literature.
A study identifies an association between smoking and type 2 diabetes. We
know that type 2 diabetes and CVD share common causal pathways, therefore
this analogy strengthens the validity of the association analogy existsfor
this association.
Explain Coherence as a Hill Criteria
The association fits with the known facts of the natural history and
pathophysiology of the disease primarily coming from in vitro
experiments
NOT USUALLY USED
Example: Observed association between asbestos exposure and
mesothelioma.
In vitro experiments using transmission electron microscopy reveal that
amphibole fibres in asbestos alter the structure and function of lung tissue
coherent association
Explain Specificity as a Hill criteria
The exposure of interest is associated only with the outcome of
interest and no other outcome (Note: this criterion does not apply for
non-infectious diseases!)
Example: The HIV virus only causes AIDS and no other disease highly
specific association
What is important to remember when using Bradford Hill Criteria?
-The Bradford Hill criteria should not be used as definite proof
or disproof of a causal association!
They should just help in deciding on
the likelihood of causality.
-The Bradford Hill criteria were published 50 years ago. As indicated
before, some of these criteria are not at all relevant in modern
epidemiology, therefore not at all considered!
What are the Rothman’s causality pies?
Any given outcome has a combination of specific causes.
Each individual case of the outcome is caused by a different combination of
these causes.
