Cause or merely association? …..explain what is meant by a cause-effect relationship in an epidemiological context …..recognise that associations may be present in the absence of a true cause- effect relationship …..describe why it is important to distinguish causal from non-causal associations …..evaluate the strength of evidence in favour of a cause-effect relationship
42
Embed
Cause or merely association? …..explain what is meant by a cause-effect relationship in an epidemiological context …..recognise that associations may be.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Cause or merely association? …..explain what is meant by a cause-effect
relationship in an epidemiological context
…..recognise that associations may be present in the absence of a true cause-effect relationship
…..describe why it is important to distinguish causal from non-causal associations
…..evaluate the strength of evidence in favour of a cause-effect relationship
Causes of TB• Poor living conditions
– Overcrowding– Poverty
• Lowered immunity– Poor nutrition– Being debilitated in old age– HIV
• Mycobacterium Tuberculosis
Causes of Measles• Measles virus
Causality
• A cause is termed sufficient when it inevitably initiates or produces the disease.
• A cause is termed necessary when it must always precede a disease
• Any given cause may be necessary, sufficient neither or both!
Four conditions where X may cause Y:
X is necessary
X is sufficient Example
1 + + Measles and the Measles virus
2 + - Tuberculosis and the Tubercle Bacillus
3 - + Lung cancer and radon
4 - - Tuberculosis and poor living conditions
Exposures do not have to be necessary OR sufficient causes of disease to be important
1. Explain what is meant by a cause-effect relationship in an epidemiological context
• Disease results from the interplay of factors from Host, Environment & Agent.
• In epidemiology a cause is an exposure/factor which increases the probability of disease.
• Exposures do not have to be necessary OR sufficient to be important causes.
• The aim is to use the knowledge to remove, avoid or protect against harmful factors.
2. Recognise associations may be present in the absence of a true cause- effect relationship
Cohort Study
• Start with Disease free individuals(sometimes go back in time to do this)• Monitor exposures of interest• Measure frequency of occurrence of
disease in exposed and non-exposed individuals
• Incidence rate ratio• Is there an association between
exposure and developing the disease?
Case Control Study
• Start with cases of disease• Get controls (up to 5) for each case• Investigate exposures of interest in
the past• Odds ratios • Is there an association between
being a case and the exposure?
Epidemiological Reasoning:-
1. HypothesisResulting from observations in clinical practice /lab research/surveillance/previous studies/theorising
2. Analytical StudyTo test the hypothesis
3. Observed associationTest the validity of the observed association by excluding alternative explanations: chance/bias/confounding
Chance
• Any result could be due to chance • statisticians can estimate how big
a role chance might have played • the results are stated and qualified
according to how much might be due to chance
• 95% confidence intervals• P value
95% confidence interval• With the data from this study, THIS
observed value is the most likely estimate of the real underlying true odds ratio/incidence rate ratio AND
• We can be 95% sure that the real population value lies within THIS range
• If the null value lies within this range (and the study was a reasonable size) then it is more likely that there is no true difference between the groups we have studied and the observed result was just due to chance
P value• The P value states how likely the results you have
in your study would occur by chance if the null hypothesis were true
• P = 0.05 means that if there was no difference your results would occur completely by chance 5 studies in 100i.e. not that likely to be due to chance so there might well be a real difference
• if P< 0.05 it can be thought of as equivalent to the null value being outside the 95% confidence interval
Bias• Deviation of results or inferences from the
truth or processes leading to such deviation
• Any trend in the collection, analysis, interpretation, publication or review of data that can lead to conclusions that are systematically different from the truth
Bias can occur at any stage• Selection bias
– Volunteers– Healthy worker effect– Controls from the same clinic in a hospital
• Information bias– Cases who know the putative risk factor– Stigma attached to the true answer
Important to exclude bias at the design stage because you cannot do it later
Dealing with bias
• Care with selection of controls• Care with questions used to ask
about risk factors• Consider blinding investigators and
subjects to the hypothesis• Check data collected with
independent records made at the time
Confounding
• The illusory association between 2 variables when in fact no association exists
• It is caused by a third variable – the confounder - which is associated with the first 2 variables i.e. with both the exposure and the outcome
Are people who wet their bed at night more likely to use bifocals?
Nocturnal eneuresis……
Use of Bifocals
Present Absent
YES 17 83 100
NO 8 92 100
25 175 200
………………………………………………Odds Ratio 1.93
Dividing the subjects by age……..
Nocturnal eneuresis aged <60yrs
Nocturnal eneuresis aged >60yrs
bifocals
Present
Absent Present
Absent
yes
1 19 20 16 64 80
no 4 76 80 4 16 20
5 95 100
20 80 100
Odds ratio = 1
Odds ratio = 1
…………………………………………….no association
Smoking confounds associations of social class/deprivation as a risk factor with diseases
• Smoking strongly linked with lower social class/increasing deprivation (the exposure)
• Smoking causes many diseases (the outcomes)
Solution is to stratify or correct using other statistical methods for known confounders
BUT there are probably many unknown and as yet unsuspected confounders….
An association is statistical dependence between 2 or more events, characteristics or other variables
The presence of an association does not necessarily imply a causal relationship
Association between factor X and factor Y• Unknown confounder making it look as
though X causes Y i.e. not a true association
• Causal association X does cause Y• Reverse causality Y causes X
– can be a problem in case control studies• Factor A causes both X and Y
– smoking causes chronic bronchitis and lung cancer – but it might look as though chronic bronchitis causes lung cancer
2. Recognise that associations may be present in the absence of a true cause-effect relationship• Hypothesis• Study to test the hypothesis• Validate any association found by
excluding possible alternative explanations– Chance– Bias– Confounding
• Could the statistical associations represent a cause-effect relationship between exposure and disease?
4. Evaluate the strength of evidence in favour of a cause-
effect relationship
How do epidemiologists attempt to establish causation – decide whether factor A could possibly be the cause of disorder B?
Koch’s Postulates (1877) to determine if an infectious agent is
the cause of a disease
• The organism occurs in every case of the disease
• It occurs in no other disease • On removal from the body and
growing in pure culture it can induce the disease anew
very exacting…
Bradford Hill proposed criteria
• Strength of association• Time sequence• Consistency• Gradient• Specificity• Biological Plausibility• Experimental Models in Animals• Preventive Trials
Strength of association
– Individuals who smoke heavily have a risk of mortality from laryngeal cancer that is 20 times that of non-smokers
this strong association increases the likelihood of it being cause and effect
Time sequence
– The exposure of interest would HAVE to precede onset of disease for it to be a cause effect relationship, the existence of an appropriate time-sequence can be difficult to establish
– Does low activity predispose to CHD OR do individuals with symptoms of CHD find it difficult to exercise?
Difficulty in case-control studies …possible strength of cohort studies
ConsistencyIf a number of studies; conducted by different
investigators; using alternative methodologies; in different time frames and amongst different populations, all show similar results…..
Cause-effect between smoking and risk of CHD: many studies; case-control and cohort; millions of person-years of observation All demonstrated increased risk
Artificial sweeteners and bladder cancer….majority of studies no effectthose which have shown an effect have not been consistent in findings of who is at risk….
Gradient (dose response)
The presence of a clear dose/response relationship strengthens the evidence for a cause-effect relationship
Specificity
• The exposure is specific to the disease (not always the case e.g. smoking)
• Asbestos and mesotheliomaMalignant mesothelioma 3 cases per million for men; 1.4 cases per million in women
• Mesothelioma in asbestos workers is 100 to 200 times higher
Specificity strengthens the case for causality but lack of it does not weaken the case
Biological Plausibility
Credible explanation of the mechanism by which the exposure could cause the disease
e.g. association between reduction of cardiac risk and moderate amounts of alcohol; cause-effect relationship enhanced by knowledge that alcohol raises HDL cholesterol
Biological plausibility depends on current knowledge
Useful cause-effect relationship may be demonstrated before mechanisms are known
e.g. John Snow & cholera… Scurvy and vitamin C
Preventive Trials
• If removal of the putative risk factor results in reduction of disease this is strong evidence to support cause and effect
Animal Models
• Experimental exposure in animals to reproduce the disease
• Exposure of an agent in animals CAN produce a disease similar to humans
• BUT NOT ALWAYS• So can be helpful but failure does
not mean much
Epidemiology is the study of the distribution & determinants of disease frequency in human populations
2 fundemental assumptions1. That human disease does NOT
occur at random2. That human disease has causal
and preventive factors that can be identified through systematic investigation
Epidemiological Reasoning:-1.Hypothesis
Resulting from observations in clinical practice /research /surveillance/previous studies/theorising
2. Analytical Study - To test the hypothesis
3. Observed associationTest the validity of the observed association by excluding
alternative explanations: chance/bias/confounding
4. Does the statistical association represent a cause-effect relationshipJudge whether the statistical association represents a cause-effect relationship – requires inferences beyond the data from any single study and is done in the light of current knowledge
• Disease results from the interplay of factors from Host, Environment & Agent.
• In epidemiology a cause is an exposure/factor which increases the probability of disease
• Exposures do not have to be necessary OR sufficient to be important causes.
But they do have to be REAL causes
• The aim is to use the knowledge to remove, avoid or protect against harmful factors and so reduce disease
Toxic shock syndrome
• 1978 “new disease” in young women in North America
• Fever, Rash & Desquamation• Hypotension and multi-organ failure • In a very short time 50 cases and 3