Epidemiology and Human Health (Moeller Chapter 3)

1
Epidemiology and Human Health(Moeller Chapter 3)
Geography 3432Environment and Health

• Context
• What is epidemiology?
• Limits to epidemiology
• Assessing epidemiological evidence
• Epidemiologic study designs
• Exposure Assessment
• Epidemiologic successes
• Epidemiology scare of the week

2
Definition Environmental Epidemiology

• The study of the effect on human health of
  physical, biologic, and chemical factors in the
  external environment, broadly conceived. By
  examining specific populations or communities
  exposed to different ambient environments, it
  seeks to clarify the relationship between
  physical, biological or chemical factors and
  human health. National Research Council (1991)
  (emphasis added)

3
Context

• Less than 1 of chemicals assessed using
  epidemiologic studies
• Why?
• Implications?

4
Limits Environmental Epidemiology

• not causal, or proof
• association, relationship, link,
  connection
• outcomes are not toxin-specific
• exposures are not health-impact-specific
• Which is better for establishing connections
  between toxic exposures and human health
  outcomes toxicology or epidemiology?

spurious relationships difficult to avoid
5
Assessing Epidemiologic Evidence

• Strength and specificity of association (RRgt 4)
• Consistency with other studies
• Existence of dose-response relationship
• Biological plausibility
• Coherence with natural history of disease
  temporal precedence
• Supporting experimental (toxicological) evidence

6
Epidemiologic Study Designs

• experimental
• strictly speaking not epidemiological see below
• observational
• cohort
• case-control
• ecological

7
Question You live in a small community whose
primary industry is a pesticides manufacturing
plant. Some families in your neighbourhood begin
to notice their children suffer serious illness -
e.g. asthma, leukemia. How would you design a
study to determine if toxins from the plant are
to blame? For example What would you measure?
How would you measure it?
8

• features
• random assignment exposed unexposed
• best form of control, strongest design
• major limitations
• unethical
• example
• unethical, remember?

9

• features
• track exposed unexposed (control) over time
• false exposure reporting less likely
• major limitations
• large sample sizes to detect rare health outcomes
• lengthy tracking
• high risk false negative results

10
Design Limitation Issues

• false negatives
• negative study does not find statistically
  significant relationship between contamination
  exposure and a negative health outcome
• false result should actually be statistically
  significant effect, but is missed because of
  some design flaw
• false positives
• positive study does find a statistically
  significant effect between contamination
  exposure and a negative health outcome
• false result should actually be no effect but
  is missed because of some design flaw
• Which is worse?
• C depends who wants to know...

11
Cohort Case ExampleLove Canal, New York
Lasalle housing development (eventually
demolished), directly above the Love Canal proper

• example Love Canal (Heath et al., 1984)
• topic organic chemicals and chromosomal
  aberrations
• description comparison near dump (exposed)
  away from the dump (unexposed)
• finding true negative (BUT only single marker
  measured)

12
Cohort DesignHeath et al 1984 Love Canal Study

• Lack of significant differences (i.e., pgt 0.05)
  between cases and controls

13
Cohort DesignSample size problem

• The rarer the outcome the larger the sample
  needed to detect effects

14

• features
• health outcome already present (cluster
  busting)
• track backwards health outcome (cases)
  non-health outcome (controls) to look for
  environmental suspects
• best for rare health outcomes
• major limitations
• biased exposure reporting
• high risk false positives

15
Case Control Case Example Woburn Massachusetts

• example Woburn Mass. (Lagakos et al., 1984)
• topic organochlorines in well water and
  childhood leukemia
• description amount of water ingested
• finding (false?) positive - highly publicized

16

• features
• large spatial areas both exposure outcome
• cheapest, quickest, most common, geographers use
  most
• major limitations
• within spatial-unit differences
• ecological fallacy
• high risk false positives/negative

17
Ecological Fallacy

• rates of health outcome and level of
  environmental contamination exposure are
  aggregates for a geographical (spatial) area
• rates relate at aggregate level, but do not know
  whether people who have health outcome also
  exposed to the environmental contaminant (i.e.,
  when disaggregated)

18
Ecological Case ExampleMichigan, PBB in Milk

example Michigan (Humble and Speizer, 1984) C
topic PBBs in meat and milk and late foetal
mortality C description quarantined counties
non-quarantined counties C finding (false?)
negative - mortality lower in quart.
19
Summary of Epidemiologic Evidence

• C inherent weaknesses in any design
• C no definitive study to link environment and
  health
• C understand likely biases when interpreting
  results
• precautionary principle or prudence vs proof -
  costs of taking action
• C epi studies only one component of
  decision-making

20
Challenges

• exposure assessment
• average or peak? (e.g. noise)
• likely pathways, body entry route?
• indoors or outdoors? (90 of time indoors)
• eating habits, cooking facilities, water
  source(s)
• retrospective (e.g., see case-control design)
  self-report bias, data unavailable

21
Exposure Assessment
22
Exposure Assessment

• More on exposure assessment in risk lecture

23
Challenges

• health outcomes
• move beyond carcinogenesis
• duration of effect
• onset of effects
• number of persons affected
• standardized measures e.g. Ravens progressive
  matrices
• biological markers e.g., blood lead phenols in
  urine to detect benzene exposure

24
Standardized Outcome Measures Ravens
Progressive Matrices
25
Epidemiologic Success Stories

• fluoride in water and dental caries (cavities)
• smoking and lung cancer
• ionizing radiation and cancer

26
Epidemiologic Scare of the Week

• proliferation of epidemiologic studies
• often contradictory
• often weak designs
• caveat emptor see assessing epidemiologic
  evidence above

27
Review

• You are the environmental specialist for a
  company who owns a large sour gas plant. They
  have been accused of causing, or possibly causing
  in the future, adverse health impacts in the
  local community. The president of the your
  company wants you to do a study to determine if
  any health effects can be linked to the plant.
  What study design will you choose to maximize the
  likelihood that the plant will not show an
  impact?
• a) ecological
• b) cohort
• c) case-control
• d) experimental

28
Review

• You are an environmental consultant hired to
  advise a community on the design for a study to
  detect health effects from chemical leaks from a
  local shoe factory. If the community wants to
  maximize the likelihood of detecting positive
  cause-effect links, which design will you
  recommend?
• a) ecological
• b) cohort
• c) case-control
• d) experimental