Association and causation Flashcards
What does a p value of 0.33 denote
33% chance that the result was due to chance.
What does the p value usually have to be for the results to be statistically significant
p < 0.05.
What is an association
Association refers to the statistical dependence between two variables, that is the degree to which the rate of disease in persons with a specific exposure is either higher or lower than the rate of disease without that exposure.
It is a link, relationship or correlation.
At what stage do we have to think about bias
We have to think about bias at the design stage, as it is hard to eliminate during the analysis.
When evaluating a statistical association what do we have to consider
Consider chance, bias, confounding, cause
How can we get around chance in a study
Larger sample size, so it is representative of the population, reduces the probability that the results are due to chance.
Power calculations
P values and statistical significance
How do we assess the role of chance
• Most studies based on an estimate from samples • The role of chance can be assessed by performing appropriate statistical significance tests and by calculating confidence intervals
Describe a confidence interval
• The range within which the ‘true’ value (e.g. the strength of an association) is expected to lie with a given degree of certainty (e.g. 95% or 99%) • 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%) of the intervals will include the true underlying population parameter
Describe the p value
- The probability that a result could simply be due to chance
- Threshold usually <0.05 = 1/20 – ie if p<0.05 we can be pretty sure (at least 95% certain) that result of a study is not due to chance – If p>0.05 then result could be due to chance
What is meant by bias
Bias is a systematic error leading to an incorrect estimate of the effect of an exposure on the development of a disease or outcome of interest. The observed effect will be either above or below the true value, depending on the nature of the systematic error.
There are two types:
Selection bias
Measurement bias
Describe bias
• Biasis a consequence of defects in the design or execution of an epidemiological study. • Biascannot be controlled in the analysis of a study, and it cannot be eliminated by increasing the sample size.
What is meant by selection bias
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.
What is meant by measurement bias
Measurement (or information) – occurs when measurements or classifications of disease or exposure are inaccurate
Give some examples of selection bias
– Non-response bias – Healthy entrant effect e.g. “healthy worker” – Loss to follow-up (attrition bias)
Give some examples of measurement bias
Recall bias- people with a condition are more likely to recall events around a diagnosis and are likely to associated these events with the disease. Common in case-control studies. Different accuracies of equipment is also an example of measurement bias.
How can we reduce selection bias
Take random samples from the population- no choice in selection.
What is meant by confounding
• A potential confounder is any factor which is believed to have a real effect on the risk of the disease under investigation and is also related to the risk factor under investigation. • This includes – factors that have a direct causal link with the disease (e.g. smoking and lung cancer) – factors that are good proxy measures of more direct unknown causes (e.g. age and social class).
Give some examples of confounding
• Age • Sex • Socio-economic status – Poorer people have rates of almost all diseases • Geography – Disease prevalence varies greatly by place – North and South
Old more likely to die earlier, women likely to die earlier, poorer people more likely to die earlier
Provide a scenario where confounding can interfere with causality
• Coffee consumption is associated with the risk of cancer of the pancreas. • Disputed because coffee consumption is correlated with cigarette smoking, and cigarette smoking was known to be a risk factor for pancreatic cancer.
How could you reduce confounding
Restrict analysis to individual groups- high deprivation, low deprivation
Standardise for age and sex in results
Regression techniques which look at effect of individual risk factors put in anlayis and calculate that independently from all other risk factors, dissociating confounding from the associations.
▪ Design Stage: • Restriction – Inclusion/exclusion criteria. • Randomisation. ▪ Analysis Stage: • Stratification – risks are calculated separately for each confounding variable. • Standardisation. • Regression analysis.
When do we consider cause
Once, bias, chance and confounding have been analysed/eliminated.
How do we judge causality
• Judgement based on a chain of logic that addresses two main areas: – Observed association between an exposure and a disease is valid – Totality of evidence taken from a number of sources supports a judgement of causality
What is the purpose of Bradford-hill criteria
It is a tool to help us judge causality. Not all the criterion are essential. TEMPORAL is essential however.
What are the factors to consider
Factors to consider • Temporal relationship • Plausibility • Consistency with other investigations • Strength of the association • Dose-response relationship • Specificity • Experimental evidence • Coherence • Analogy also consider reversibility
Describe the role of strength
The strength of an association is measured by the magnitude of the relative risk. A strong association is more likely to be causal than is a weak association, which could more easily be
the result of confounding or bias. However, a weak association does nor rule out a causal connection. For example, passive smoking and lung cancer.
Describe the role of consistency
If similar results have been found in different populations using different study designs then the association is more likely to be causal since it is unlikely that all studies were subject to the same type of errors. However, a lack of consistency does not exclude a causal association since different exposure levels and other conditions may reduce the impact of the causal factor in certain studies.
Describe the role of 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. However, one-to-one relationships between exposure and disease are rare and lack of specificity should not be used to refute a causal relationship; for example cigarette smoking causes many diseases
Describe the role of temporal relationship
This is an essential criterion. For a putative risk factor to be the cause of a disease it has to precede the disease. This is generally easier to establish from cohort studies but rather
difficult to establish from cross-sectional or case-control studies when measurements of the possible cause and the effect are made at the same time. However, it does not follow that a reverse time order is evidence against the hypothesis
Describe the role of dose-response relationship
Further evidence of a causal relationship is provided if increasing levels of exposure lead to increasing risks of disease. Some causal associations, however, show a single jump (threshold) rather than a monotonic trend.
Describe the role of plausibility
The association is more likely to be causal if consistent with other knowledge (e.g. animal experiments, biological mechanisms, etc.). However, this criterion should not be taken too seriously because lack of plausibility may simply reflect lack of scientific knowledge. The idea of microscopic animals or animalcules as cause of disease was distinctly implausible before Van Leeuwenhoek’s microscope
describe the role of coherence
Coherence implies that a cause and effect interpretation does not conflict with what is known of the natural history. However absence of coherent information as distinguished from the presence of conflicting information, should not be taken as evidence against an association being causal. quite vague
Describe the role of experimental evidence
Experimental evidence on humans or animals. Evidence from human experiments is seldom available and animal research relates to different species and different levels of exposure.
Describe the role of analogy
At best analogy provides a source of more elaborate hypotheses about the association in question. Absence of such analogies only reflects lack of imagination or experience, not falsity of the hypothesis (Bradford Hill 1965).
What is meant by reversibility
If you take away the exposure, does the risk of the outcome go down
