Transcript Evaluating Environmental/Occupational `Clusters` of Disease

Evaluating
environmental/occupational
“clusters” of disease
Robert J. McCunney, MD
Investigating Clusters of Disease
Definition: "cluster" is an unusual
aggregation, real or perceived, of health
events that are grouped together in time
and space
A cluster may be useful for generating
hypotheses but is not likely to be useful for
testing hypotheses
Disease Clusters
Focus: usually on increased rate or risk
of disease
Breast cancer and pesticides
Leukemia and contaminated water
Increased risk of lung cancer from
hexavalent chromium
Cancer Cluster Activities at the
CDCP
Cancer cluster investigations occasionally have led to the
discovery of important pathways in the etiology of
specific cancers, such as with
angiosarcoma
lung cancer
Kaposi sarcoma
vaginal clear-cell carcinoma
bladder cancer
scrotal cancer
NOTE: The majority of studies that yielded etiologic
information were of occupational, drug-induced, or
infectious pathogenic exposure rather than studies
of environmental exposure
Media survey of cancer
cluster reports
For the period 1977 through 2001, media
reports of 1,440 records of approximately 175
suspected cancer cluster reports
These data reflect the breadth of popular
concern and awareness regarding issues of
exposure types, pollution sites, and specific
environmental chemicals.
Clusters: case study
7 cases of kidney cancer at a
manufacturing plant
Is there a cluster?
Criteria to determine whether to
proceed to investigate a cluster
identification of a single cancer type;
biological plausibility
adequate latency
political pressure;
identification of a common cancer in an unusual
age group;
identification of a rare cancer;
identification of exposure to a carcinogenic
substance;
elevated ratio of observed/expected confirmed
cancer cases
Guidelines for Investigating
Clusters of Health Events
CDC
GUIDELINES FOR A
SYSTEMATIC APPROACH
The issue of increased frequency of occurrence
should be separated from the issue of potential
etiologies
Stage 1. Collect information.
A variety of diagnoses speaks against a
common origin
Stage 2. Assessment
Separate two concurrent issues:
1. whether an excess has occurred and
2. whether the excess can be linked
etiologically to some exposure
Stage 3. Feasibility Study
Conduct detailed literature search with
particular attention to known and putative
causes of the health effect of concern.
Consider an appropriate study design
Determine what data should be collected on
cases and controls, including laboratory
measurements. Determine the nature, extent,
and frequency of methods used for
environmental measurements.
Determine hypotheses to be tested and power
to detect differences
Stage 4. Etiologic Investigation
Purpose: Perform an etiologic investigation of a
potential disease- exposure relationship.
Types of studies:
Retrospective cohort mortality-most common
Case/control-good for rare diseases
Cross sectional-good for morbidity assessments
Cluster Investigations
A number of problems are encountered in
the study of clusters.
The health events being investigated
(often morbidity or mortality) are usually
rare, and increases of these events tend to
be small and may occur over a long
period.
A major complication is that most clusters
are chance events
CDC
a) provides a centralized, coordinated response
system for cancer cluster inquiries,
b) supports an electronic cancer cluster list
server,
c) maintains an informative web page, and
d) provides support to states, ranging from
laboratory analysis to epidemiologic assistance
and expertise
Occupational/Environmental
epidemiology
Goal: Evaluate exposure- disease
relationships
Strengths: risks in humans –no need to
extrapolate from animal studies
Evaluate consequences of exposure in
which it actually occurs
Occupational epidemiology
Limitations:
Low level risks difficult to identify
Small increases in risk may be affected
by bias, confounding and chance
Long latency with cancer
Inadequate exposure categorization
Measures of effect
RATES are the central metric used to assess disease
occurrence in occupational cohort studies
Comparison of a study group to a reference groupusually done by assessing the ratio of their respective
standardized rates
one frequently used outcome measure is the
Standardized Mortality Ratio (SMR)
SMR: ratio of the sum of the observed events in the
study group to the sum of the expected numbers in the
study group; expected numbers are based on
standardized rates in the reference group
Example of use of Standardized
Mortality Ratio (SMR)
Null hypothesis (H0): no effect of exposure
SMR for lung cancer of 1.8
80% excess compared to reference
population
P<.05; 95 % Confidence Intervals; (1.22.3); thus, statistically significant
? Role of confounding and bias
Interpreting SMR results
Chance: P values; p<.05 and 95%
confidence intervals (reflect uncertainty to
random error-not confounding
Bias: selection, recall
Confounding (understates the uncertainty
about a true “effect”: smoking and lung
cancer (biases SMR upward); alcohol and
liver disease, diabetes and neuropathy
Occupational epidemiology
Reference population – of paramount
importance- ideally should match study group in
all important demographics characteristics aside
from “exposure” of interest
Type of reference populations: General
(national, state, county) population disease rates
categorized by age, gender, race, geographical
area, calendar time
Key point: reference and study populations
should be identical as much as possible
aside from the exposure of interest
Reference population
Lung cancer and carbon black: German
study
SMR: 2.2 (national rates)
SMR: 1.8 (local rates)
PRINCIPLES IN EVALUATING WHETHER A
DISEASE IS ENVIRONMENTALLY RELATED*
Strength of association
Consistency of results; does the
association hold in different settings and
among different study groups?
Specificity: How closely are the exposure
factors and health outcome associated?
Temporality: Does exposure precede
disease outcome. Is latency involved?
* From Hill AB. The environment and disease:
association or causation? Proc R Soc Med 1965:
58: 295-300
PRINCIPLES IN EVALUATING WHETHER
A DISEASE IS WORK RELATED
Biological gradient: does a dose- response
relationship exist?
Plausibility: Does the association make
sense biologically?
Coherence: Is the association consistent
with the natural history and biology of the
disease?
Experimental Evidence: Does experimental
evidence support the hypothesis of an
association?
Analogy: Are there other examples with
similar risk factors and outcome?
Interpreting occupational
epidemiological literature
Study design- is the hypothesis clearly defined?
Methods: are they adequate to evaluate the hypothesis?
How was exposure assessed?
How was the cohort defined? Is there adequate
ascertainment of vital status?
What is the reference population? Is it an appropriate
comparison group for the cohort being studied?
Results
What statistical methods were used?
How was chance assessed?
How was confounding controlled?
How were potential biases addressed in the analysis, such as
selection bias and confounding?
Interpreting occupational
epidemiological literature
Discussion
Have the authors contrasted their results with
previous scientific literature?
Have the authors discussed the limitations of their
study?
What further work can be done to more fully define
the results?
Conclusion
Have the authors properly assessed the results
based on their own analysis ands limitations in
light of previous literature?
Renal cell carcinoma
Renal cell carcinoma
Incidence: increased diagnostic accuracy
:CT, ultrasound, MRI
Obesity, hypertension, smoking:
established risk factors
Occupational: ? TCE, solvents
Plant
1200 employees
Produces plastics, TDI, benzene,
phosgene
Built in 1950s
Product manufacturing and research
Is there an excess ?
Expected rate of renal cancer: 12/100,000
Preliminary analysis:
From 1990-2006: 1200 employees for 16
years :~20,000 person years: 7 observed
35 observed over 100,000 person years
12 expected
Thus: ~ 3 fold excess???
Disease clusters and students
Be critical of media reports
Systematic approach to evaluations
Recognize role of health departments
Most clusters are statistical artifacts
Interpret epidemiology studies cautiously
Be attentive to possible new links between
exposure and disease ) eg. nanoparticles,
new materials etc
Discussion
Next step?