Transcript Epidemiology

W507 – Epidemiology
Epidemiology
• Scientific process that attempts to link exposure to
hazardous substances to specific health effects.
• Statistical correlations developed to indicate degree of risk
associated with a particular exposure pattern
• Epidemiological studies tend to be complex and
outside the remit of most industrial hygienists
• However, important to understand the principles of
epidemiology
Epidemiology
•
Much information is derived from animal tests, in-vitro tests and
comparison with similar chemicals
• However, to establish actual risks in a population an epidemiological study is
often the only valid technique
•
Aim is to establish a link between exposure and development of disease or
ill-health e.g.
• Angiosarcoma of the liver and exposure to vinyl chloride monomer
• Mesothelioma of the lung pleura and exposure to asbestos
•
However, clear evidence may not be available and many studies have
limitations and deficiencies
• Historical data may be unclear or not valid
• Potential problems of bias or other confounding factors in selection or
monitoring of groups under study
Epidemiological terms
• Incidence rates
• Number of new cases of ill-health or disease that occur in a population
in a period of time
• Commonly expressed in number of new cases per 1000 people in group
under study per year
• Prevalence rates
• Total number of cases of ill-health or disease that exist in the population
at a particular point in time
• Affected by incidence of new cases but also the duration of the effects
• Findings may be expressed in terms of morbidity (number of
cases of illness) or of mortality (death) rates
Measures of frequency
• Incidence and prevalence rates need to be compared to a
reference (or control) group or population
• Incidence and prevalence rates can be expressed as a ratio or
‘Standardised Mortality Ratio’ (SMR) compared to reference
group
• If no difference between study and reference group the SMR would be 1
• If greater incidence or prevalence rate in study group compared to
reference group the SMR would be greater than 1
• (Note: some texts quote SMR’s as percentage of expected numbers of
deaths or events – i.e. if no difference SMR would be 100)
Measures of frequency
• Major potential difficulty is identifying a suitable
reference population which must be matched to the
study group in all aspects except exposure to the
substance.
• Potential areas of difference include
• Age distribution
• General health levels
• e.g. working population tends to be more healthy than the general
population as this includes people unable to work due to ill-health
Causation or association
• Important to differentiate between causation and association
• Study may show a link (association) between exposure to a substance and
a health effect
• This does not necessarily mean there is a direct causal link
• The association may be spurious or it may be indirect through other
known or unknown variables
• A number of criteria have been developed (Bradford-Hill criteria) that
can be examined to provide support for a causal link
Bradford-Hill Criteria
•
•
•
•
•
•
•
•
•
Strength of association
Temporality
Consistency
Biological plausibility
Coherence
Specificity
Dose-response relationship
Experimental evidence
Analogy
Bradford-Hill Criteria
• None of the nine criteria provide indisputable
evidence for or against a cause and effect hypothesis
• Rarely will all nine points be present – however the more
that are, the stronger the association
• They can to a greater or lesser extent support or refute the
hypothesis
• Quality of study should also be assessed
• Those with appropriate statistical analysis and those
published in peer-reviewed journals carry more weight
• Claims of causation should not be made lightly as poorly
justified claims can be very misleading
Causation or association – smoking
and lung cancer
• Strength of association
• Lung cancer rates far higher for smokers than non-smokers
• Temporality
• Smoking precedes onset of disease
• Consistency
• Different types of studies produce the same result
• Biological plausibility
• Theory that smoking causes tissue damage that over time could
cause cancer is highly plausible
• Coherence
• Theory ‘makes sense’ given current knowledge
Causation or association – smoking
and lung cancer
• Specificity
• Lung cancer is predicted by incidence of smoking
• Dose-response relationship
• Data showed linear relationship between amount smoked and
incidence
• Experimental evidence
• Animal experiments using tobacco tar produced cancer in tissue
• Analogy
• Induced smoking in animal experiments showed increased lung
cancer
Bias
All epidemiological studies are subject to bias:
• Selection bias in study group or control group
• Information bias related to the quality and accuracy of data
gathered as well as errors by the interviewer or interviewee
• Confounding factors influence both the exposure and the
outcome and therefore may show a spurious link
Many of these factors can be controlled at the planning
stage of the study
Statistical significance
• It is important to determine whether the apparent link between
exposure and outcome is statistically significant i.e. unlikely to
have occurred by chance
• Most studies define statistically significant as 95% or better
confidence that the results were not due to chance – i.e. less
than 5% (1 in 20) probability the association was by chance
• Sometimes written as a probability of ‘p <0.05’
• Lower the probability or significance the stronger the evidence
• Establishing statistical significance is important but is not in itself
enough to confirm causality. At this stage criteria developed by
Bradford-Hill need to be examined
Types of epidemiological studies
• Longitudinal studies
• Follow a group of people over period of time with repeated
observations of risk factors and health effects
• Vary widely in size and duration (even decades)
• Two common types of longitudinal study are cohort and
case-control studies
• Cross-sectional studies
Longitudinal studies
• Case-control studies
• Used to identify factors that may have caused a particular
effect or illness
• Tend to be retrospective
• Study a group (the ‘cases’) who have the effect or illness and
compare with another group who do not have the effect or illness (the
‘controls’)
• Gather information on past exposure to the factor or substance –
often difficult to get accurate data
• Often difficult to avoid possible confounding factors when choosing
the control group
• Usually not the most robust type of study but often used to
identify exposure / disease relationships that can then be
investigated more fully
Longitudinal studies
• Cohort studies
• Tend to be prospective
• Gathers data over period of time on the incidence of the effect
• Study a group of people who are exposed to a particular
substance
• Cohort group identified before appearance of the effect
• Control group identified either from the general population or from
another similar cohort who have had little or no exposure to the
substance in question
• Potentially the most robust type of study
• Tend to take a long time to complete and as such can be
complex and expensive
Cross-sectional studies
• Involves observation of a group of people at a particular
moment in time
• Can be considered as a ‘snap-shot’ of the prevalence of the effect
• Group compared with a similar matched control group
• Relatively quick and inexpensive
• However, study only deals with a particular moment in time and
cannot give information on exposure and outcome as a time dependant
relationship
• Not suited to an effect that is uncommon
W507 – Epidemiology case study
Epidemiology case study
• Concerns have been raised that a number of
people working at a plant producing orthotoluidine have developed bladder cancer
over the last few years
• What information would you initially
require if you were investigating the above?
Initial information required
• Other chemicals used
• Are all workers exposed to the ortho-toluidine – or just
some groups?
• What levels of exposure have occurred to ortho-toluidine?
• Number of people working at the plant
• Number of cases of bladder cancer
• When were the cases of bladder cancer first noted?
• When was ortho-toluidine first used?
• What problems may there be in obtaining accurate
information?
Epidemiological case study
• What type of epidemiological study would
you suggest for this scenario?
• To evaluate whether there is an increased
risk of bladder cancer a reference or control
group is needed
• Who would you select for this reference group?
Epidemiological case study
• What sources would you use to gather
information about exposure levels and
occurrence of bladder cancer for the case
and reference groups?
• What problems may there be in getting
accurate information?
Epidemiological case study
• 13 cases of bladder cancer out of 1749
people who worked at plant
• Expected number of cases of bladder cancer in
the general population – 2.1 per 1000
• What is the SMR (standardised mortality
ratio)?
Epidemiological case study
• Is this SMR (3.5) significant or simply a
random or chance occurrence?
• Statistical analysis required
• 95% confidence limits of SMR found to be 1.8 – 6.2
• Does this prove a causal link?
Epidemiological case study
• What possible confounding factors may need to
be considered when designing the study?
• Other chemicals used
• Smoking habits
• Age distribution of case and reference groups
Epidemiological case study
• Results of epidemiological study
• See tables 1 to 3
• Animal studies show increased rates of cancer in both
mice and rats
• Ortho-toluidine is similar chemical structure to other
aromatic amines which have shown an increased risk
of bladder cancer
• Use Bradford-Hill criteria to assess whether there is a
causal link
Epidemiological case study
• Bradford-Hill criteria
•
•
•
•
•
•
•
•
•
Strength of association
Temporality
Consistency
Biological plausibility
Coherence
Specificity
Dose-response relationship
Experimental evidence
Analogy