Flashcards in Association and Causation Deck (19):
List possible explanations for observed associations
Chance, bias, confounding and causation (always consider first 3 before assuming causation)
What is chance?
Most studies are based on one estimate from samples than whole populations. Role of chance assessed by statistical tests. If independent samples are taken repeatedly from the same population, and a confidence interval calculated for each sample, then a certain percentage (e.g. 95%0 of the intervals will include the true underlying population parameter
What is bias?
A systematic error leading to incorrect estimate of effect of an exposure of disease development. Can be due to defects in study design
Can bias be controlled by analysis of study or increasing sample size
What are the 2 broad types of bias?
Selection- occurs when there is a systematic difference between the characteristics of the people selected for a study and the characteristics of those who were not
Measurement- occurs when measurements or classifications of disease/ exposure are inaccurate
What does confounding mean?
Any factor which believed to have a real effect on the risk of the disease under investigation and is related to the risk factor under investigation.
What are common confounders?
Age, sex, socioeconomic status, geography
What is causation?
Judgement based on a chain of logic that addresses 2 main areas: association is valid and evidence from several sources supports there being causality
List the Bradford hill criteria
Factors to consider are: strength, consistency, specificity, temporal relationship, dose-response relationship, plausibility, experimental evidence, coherence and analogy (last 2 not very important in assessing causation) + reversibility (if cause is removed consequence affected)
Bradford Hill: strength
Strength of association measured by magnitude of relative risk. EG lung cancer and passive smoking (example of weak association can still mean causality)
Bradford Hill: consistency
More likely to be causal if similar results in different populations using different study designs. Lack of consistency doesn't exclude causal association as other conditions might reduce impact of causal factor
Bradford Hill: specificity
If a particular exposure increases the risk of a certain disease but not the risk of other diseases then this is strong evidence in favour of a cause-effect relationship e.g. Mesothelioma an asbestos.
BUT one-to-one relationships between exposure and disease are rare and lack of specificity should not be used to refute a causal relationship e.g. cigarette smoking causes many diseases.
Bradford Hill: temporal relationship
This is the only criteria which is completely essential. FOR A PUTATIVE RISK FACTOR TO BE CAUSE OF DISEASE IT MUST PRECEDE THE DISEASE. This can be more easily established from cohort studies than cross-sectional. Reverse time-order isn't evidence against hypothesis.
Bradford Hill: Dose-response relationship
Increasing levels of exposure causes increase risk of disease. Some causal associations can show single jump rather than monotonic trend
Bradford Hill: plausibility
Causation more likely if consistent w/ other knowledge. But lack of plausibility might be due to lack of knowledge. - The idea of microscopic animals or animalcules as cause of disease was distinctly implausible before Van Leeuwenhoek's microscope
Bradford Hill: coherence
Cause and effect relationship doesn't conflict with knowledge of the natural history. Absence of coherence doesn't provide evidence against causality.
Bradford Hill: analogy
Best analogy provides a source of more elaborate hypotheses about the association in question.
What is matching?
A method for controlling the effect of confounding at the design stage- controls selected to have similar distribution of potentially confounding variables