Lecture 14 - Association & Causality Flashcards
3 types of associations btw exposure and outcome
- Artifactual
- Non-causal
- Causal
How do artifactual associations arise?
Can arise from significant bias and/or extensive confounding
Koch’s 4 postulates for implicating a causal relationship
- Must be present in every case of disease
- Must not be found in cases of other diseases or healthy individuals
- Must be capable of isolation, culture, and reproducing disease in experimental animals.
- Must be recovered from experimentally-induced diseased animals
Koch’s postulate limitations
- Disease production may require cofactors that postulates don’t address
- Viruses can’t be cultured similar to bacteria
- Not all viruses/bacteria induce clinical disease (carriers & sub clinical disease - unable to be detected)
Mill’s Canons
The cause of any effect must consist of a constellation of components that act in concert
Synergism
Interaction of 2 or more presumably-causal variables so that the combined effect is clearly greater that the sum of the individual effects.
2 ways non-causal associations can occur
- The disease may cause the exposure
2. Disease and exposure both associated with a third factor (confounding)
Examples of 2 ways non-causal associations occur
- Physical inactivity thought to lead to Arthritis. But actually Arthritis is what causes people to be physically inactive.
- 3rd or 4th born said to be more likely to have down syndrome. However, not becasue of birth order. It is because mothers are older when 3rd or 4th born and older increases risk of down’s syndrome.
Sufficient Cause
Relationship sufficient enough that when cause is present you will always get disease. RARE
(T/F) Sufficient causes always act alone in causing a disease.
False, may have multiple required components that collectively act to induce disease
Necessary Cause
Necessary for disease to occur, BUT the cause may also be present without disease occurring.
Component cause (Risk Factor)
If present, increases probability of a particular disease. Helps prime people for disease. Some patients must be primed or suceptible to disease before component cause can induce disease.
Temporality
The necessity that the cause precede the effect/outcome
Proximate cause
Short term interval btw cause and disease occurrence in individual
Distant cause
Long term interval btw cause and disease occurrence in individual
Induction period
Component cause to disease onset
Latent period
Disease onset to diagnosis/clinical presentation
Which is easier to see association (Proximate or distant cause)?
Proximate, distant may take a long time for disease to show up so hard to determine relationship btw cause and disease becasue exposed to so much more in that long induction period
What do you collectively call it when multiple component causes work in concert to cause disease?
Component causes collectively become sufficient cause
3 ways to control/adjust for other variables to examine effect of a single factor
- Restriction - keeps other factors out
- Matching - similar characteristics in each group
- Stratification - categorize by disease severity or exposure level
Hill’s guidelines/criteria for causal inference
- Strength
- Consistency
- Temporality
- Biologic gradient
- Plausibility
The higher the # of Hill’s criteria met….
The more likely association MAY be causal
Strength
Greater the association the more convincing it is that association MIGHT be causal
Consistency
The repeated observation of association in diff populations under diff circumstances in diff studies. Increases strength of likelihood of potential causal relationship.
Menopausal hormone therapy
- Benefits of MHT very consistent
- So did official clinical study
- combo estrogen/progesterone group has more heart attacks than non-combo
- Estrogen alone group not same rate of heart attack.
- Watched for another year and saw estrogen alone group also increased heart attacks
Why can consistency show wrong effects in some situations like MHT?
Biases
Temporality
necessity that disease precede the effect/outcome
Explain temporality in higher cancer rates in former smokers during first yr after quit
Substantial portion often quit because have started to see early signs and symptoms of already existing but not yet diagnosed lung cancer
Biologic gradient
Observation of gradient of risk (dose response) associated with degree of exposure. Smoke more per day, more likely to have lung cancer
Threshold effewct
No effect until certain level of exposure (Alcohol consumption and death)
Plausability
Biological feasability that associated can be understood and explained. Is event/exposure biologically plausible if really true?
Issue with plausability
Plausability decision may be based on prior beliefs that can be flawed or incomplete. Example: Didn’t know organism (H.pylori) could cause ulcers.
Cause
A precursor event/condition/characteristic that is REQUIRED for the occurrence of a disease.
