Causation in Epidemiology - PowerPoint PPT Presentation

Loading...

PPT – Causation in Epidemiology PowerPoint presentation | free to view - id: 61bd5-OTUzY



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Causation in Epidemiology

Description:

Ecologic study of per capita smoking and lung cancer incidence ... This is used by tobacco companies to argue that smoking is not causal in lung cancer. ... – PowerPoint PPT presentation

Number of Views:2260
Avg rating:5.0/5.0
Slides: 43
Provided by: sharon112
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Causation in Epidemiology


1
Causation in Epidemiology
Summer Academic Preview 2006
Aruna V. Sarma PhD, MHA Assistant Research
Professor, Urology Assistant Research Scientist,
Epidemiology
2
Exposure OR Genetic Background OR Combination of
Both
Association
? Causation ?
Disease or Other Outcome
Suppose we determine that an exposure is
associated with disease. How do we know if the
observed association reflects a causal
relationship?
3
First step in determining causation
Understanding disease etiology
  • Experimental studies
  • in vitro systems
  • animal studies in controlled environments
  • Allows for
  • control of precise dose
  • control of environmental conditions
  • loss to follow up kept to a minimum
  • Problems with
  • extrapolating data to human populations
  • human diseases with no good animal models
  • Clinical pathologies

4
Second step in determining causation Conducting
Studies in Human Populations Heres where
Epidemiology is important. Epidemiology
capitalizes on natural or unplanned
experiments. We take advantage of groups who
have been exposed for non-study purposes. All
of the study designs are important here and
provide different evidence for or against a
causal hypothesis.
5
Clinical observations
Available data (Ecological or Cross-sectional
Studies)
Case-control studies
Cohort studies
Randomized trials (only used for potentially
beneficial treatments)
6
Ecologic studies
  • A study in which the units of analysis are
    populations or groups of people, rather than
    individuals.
  • Usually takes advantage of pre-existing data
    collected for other purposes - an efficient and
    economical study design
  • No time element - a snapshot of populations -
    think cross-sectional studies of populations, not
    individuals.

7
Figure 13-4. Correlation between dietary fat
intake and breast cancer by country.
USA
250
Switzerland
Canada
Fed. Repub. Of Germany
Italy
200
UK
Denmark
Israel
France
Sweden
New Zealand
Australia
Norway
150
Finland
Yugoslavia
Spain
Incidence Ratio per 100,000 Women
Poland
100
Romania
Hungary
Hong Kong
50
Japan
Per Capita Supply of Fat Calories
0
0
1600
1400
1200
1000
800
600
Prentice RL, Kakar F, Hursting S, et al Aspects
of the rationale for the Womens Health Trial. J
Natl Cancer Inst 80802-814, 1988.)
8
Is there a relationship between breast cancer
incidence and dietary fat consumption by country?
  • From the graph, we see that as average dietary
    fat consumption increases, breast cancer
    incidence increases.
  • What is wrong with this data?
  • The problem is the ecologic fallacy!
  • Prentice et al. J Natl Cancer Institute 1988
    80802-814

9
Why do an ecologic study? HYPOTHESIS
BUILDING! The data is easy to obtain, no
follow-up or individual contact is needed. An
ecologic study can suggest avenues of research
that may cast light on an etiologic relationship
between exposure and disease. HOWEVER, an
ecologic study does not itself demonstrate that a
causal relationship exists.
10
Case-control and cohort studies
  • Unlike ecologic studies, in case-control and
    cohort studies we have information on both
    exposure and outcome for individual subjects.
  • Case-control and cohort studies have a time
    element - in each, we are assuming that exposure
    occurred before disease.

11
Two step process to carry out studies and
evaluate evidence
  • 1. Determine if an association is present
  • Ecologic studies studies of group
    characteristics
  • Cross-sectional studies studies at one
    particular time
  • Case-control or cohort studies studies of
    individual characteristics.
  • 2. If an association is demonstrated, determine
    whether the observed association is likely to be
    a causal one using pre-determined criteria.

12
Surgeon Alton Ochsner observed that almost all
lung cancer patients were smokers
Ecologic study of per capita smoking and lung
cancer incidence
Case-control study of lung cancer patients versus
those without lung cancer
13
Understanding Causality
  • Types of Association
  • causal
  • noncausal
  • Types of Causal relationships
  • direct
  • indirect
  • Types of causal factors
  • sufficient
  • necessary

14
Two Types of Association Real and Spurious
  • A real association is present if
  • the probability of occurrence of an event or
  • the quantity of a variable
  • depends upon the occurrence of one or more other
    events, characteristics or variables.
  • Spurious associations refer to non-causal
    associations due to chance, bias, failure to
    control for extraneous variables (confounding),
    etc.

15
Interpreting Associations - Causal and Non-Causal
Non-Causal (due to confounding)
Causal
Characteristic Under Study
Characteristic Under Study
Factor X
Disease
Disease
16
The relationship between coffee consumption and
pancreatic cancer
  • In 1981, MacMahon et al. reported results from a
    case-control study of cancer of the pancreas.
  • There was an apparent dose response relationship
    between coffee consumption and cancer of the
    pancreas, particularly in women.
  • Was the disease caused by coffee consumption or
    by some factor closely related to coffee
    consumption?

MacMahon B, et al. N Engl J Med 1981 304630 - 33
17
The relationship between coffee consumption and
pancreatic cancer
  • Smoking is closely associated with both
    pancreatic cancer and coffee consumption.
  • There were many issues with control selection and
    measurement of exposure levels in cases and
    controls.
  • Subsequent studies were unable to reproduce the
    result.

MacMahon B, et al. N Engl J Med 1981 304630 - 33
18
Interpreting Associations - Causal and Non-Causal
Non-Causal (due to confounding)
Causal
Coffee Consumption
Coffee Consumption
Real Association
Spurious Association
Smoking
Real Association
Pancreatic Cancer
Pancreatic Cancer
19
Why is it important to distinguish between causal
and non-causal associations?
  • Causal relationships are used to make public
    health decisions and design interventions.
  • In our example, if smoking was indeed causal, it
    would be irresponsible to target coffee drinking
    as an intervention.
  • Very important to consider all confounders.

20
Types of Causal Relationships Direct vs Indirect
Indirect
Direct
Factor 1
Factor
Factor 2
Factor 3
Factor 4
Disease
Disease
21
Types of Causal Relationships Direct vs Indirect
Indirect
Direct
?F508 Polymorphism
High cholesterol
Artery thickening
Hemostatic factors
Myocardial infarction
Cystic Fibrosis
22
Four types of causal factors
  • Necessary and sufficient
  • Without factor, disease does not develop
  • Example HIV
  • Necessary but not sufficient
  • Multiple factors, including main factor, required
  • Example Development of tuberculosis requires M.
    tuberculosis and other factors, such as
    immunosuppression, to cause disease
  • Bacteria still necessary, but not sufficient to
    cause the disease

23
Four types of causal factors
  • Sufficient but not necessary
  • Factor can produce disease, but not necessary
  • Example Both radiation exposure and exposure to
    benzene are sufficient to cause leukemia, but
    neither are necessary if the other present.
  • Neither sufficient nor necessary
  • Complex models of disease etiology
  • Example High fat diet and heart disease,
    hypertension, diabetes, certain kinds of cancer

24
Understanding Causality
  • Lets say you have determined
  • there is a real association,
  • you believe it to be causal (ruled out
    confounding),
  • figured out that it is a direct causal factor
  • sorted out the necessary vs. sufficient factor
    issue
  • NOW have your proven CAUSALITY?

25
Nine guidelines for judging whether an
association is causal
  • Consideration of alternate explanations
  • Cessation of exposure
  • Specificity of the association
  • Consistency with other knowledge
  • Temporal relationship
  • Strength of association
  • Dose response relationship
  • Replication of the findings
  • Biologic plausibility

26
Temporal Relationship
  • Exposure to the factor must have occurred before
    the disease developed.
  • Easiest to establish in a cohort study
  • Length of interval between exposure and disease
    very important
  • If the disease develops in a period of time too
    soon after exposure, the causal relationship is
    called into question.

27
Asbetos and Lung Cancer
Well - established temporal relationship
Lung Cancer
Latent period of 10 - 20 yrs
Asbestos
New Study
Asbestos
Lung Cancer
Latent period of 3 yrs
In this case, the latent period is not long
enough for lung cancer to develop if caused by
exposure.
28
Strength of Association
  • The larger the relative risk or odds ratio, the
    higher the likelihood that the relationship is
    causal.
  • However, care must be taken to examine confidence
    intervals and sample size.
  • For example, if the confidence interval is wide
    (e.g., 1.8 - 22.6), an OR of 12.0 is less strong
    because we are less confident of the strength of
    the odds ratio.

29
Strength of association
Which odds ratio would you be more likely to
infer causation from? OR1 OR 1.4 95 CI
(1.2 - 1.7) OR2 OR 9.8 95 CI (1.8 -
12.3) OR3 OR 6.6 95 CI (5.9 -
8.1)
30
Dose-Response Relationship
  • With increasing dose, there is increasing risk of
    disease.
  • This is not considered necessary for a causal
    relationship, but does provide additional
    evidence that a causal relationship exists.

31
Figure 13-18. Age-standardized death rates due
to well-established cases of bronchogenic
carcinoma
Adapted from Hammond EC, Horn D JAMA
1661294-1308, 1958.)
32
Replication of the Findings
  • If there is a true causal relationship between
    exposure and disease, the expectation is that we
    would see the association consistently in other
    (NOT necessarily all) subgroups of the population.

33
Biologic plausibility
  • Consistency of epidemiologic plausibility with
    existing biologic knowledge.
  • Requires knowledge of the biologic etiology of
    disease

34
Consideration of alternate explanations
  • Consider the example of coffee consumption,
    smoking and pancreatic cancer.
  • Did the investigators consider the associations
    between smoking, coffee consumption and
    pancreatic cancer?
  • If the investigators did not consider possible
    confounders and effect modifiers, the association
    is less likely to be causal.
  • Requires a knowledge of the literature and known
    risk factors for the disease

35
Cessation of exposure
  • Upon elimination or reduction of exposure to the
    factor, the risk of disease declines.
  • HOWEVER, in certain cases, the damage may be
    irreversible.
  • Example Emphysema is not reversed with the
    cessation of smoking, but its progression is
    reduced.

36
Specificity of the Association
  • The weakest of the criteria (should probably be
    eliminated)
  • Specific exposure is associated with only one
    disease.
  • This is used by tobacco companies to argue that
    smoking is not causal in lung cancer.
  • Smoking is associated with many diseases.
  • If anything, may provide support for a causal
    relationship, but does not negate if not present.

37
Consistency with other knowledge
  • If a relationship is causal, the findings should
    be consistent with other data.
  • If lung cancer incidence increased as cigarette
    use was on the decline, we would have to be able
    to explain how this was consistent with a causal
    relationship (How?!?)

38
  • Associations are observed
  • Causation is inferred
  • It is important to remember that these criteria
    provide evidence for causal relationships.
  • All of the evidence must be considered and the
    criteria weighed against each other to infer the
    causal relationship.

39
What impairs inferences regarding disease
causation?
  • A lengthy interval between the presence of the
    causal factor and the manifestation of its effect
  • head injury as a possible cause of Alzheimers
    Disease
  • When the same effect can occur due to the action
    of other causes
  • When the causal factor requires the presence of
    other factors to produce the effect

40
Under what circumstances are we unable to
identify an etiologic factor through
observational studies of individuals?
  • Magnitude of increased risk caused by the factor
    too small to be reliably identified
  • Magnitude of increased risk theoretically not too
    small, but
  • Insufficient variation among individuals
    regarding presence/levels of factor
  • Unable to distinguish effect of the factor from
    that of other factors that are highly correlated
    with it
  • Practical problems cannot be overcome when
    conducting studies
  • No valid measure of past presence or level of
    factor
  • Lengthy induction or latent period

41
Why was it relatively easy to determine that
smoking was a cause of lung cancer?
  • History of exposure to cigarettes can be assessed
    with reasonable accuracy.
  • Cigarette smoking is common and present in
    persons whose environment is otherwise similar to
    that of nonsmokers.
  • Lung cancer incidence in smokers is much greater
    than in nonsmokers.
  • Lung cancer is uncommon in nonsmokers.

42
Why will it be relatively hard to determine if
community air pollution is a cause of lung cancer?
  • Difficult to measure pertinent exposure
  • Long latent period of disease
  • Migration
  • Little variation in exposure among individuals
    within a community
  • Lung cancer is common, even among persons not
    exposed to pollution.
About PowerShow.com