Environmental Engineering Reference
In-Depth Information
associations are inappropriate. However, ecological studies are relatively inexpen-
sive for linking available health data sets and environmental information and are
useful for hypothesis-generation (Yassi et al. 2001 ). Examples of ecological studies
are the assessments of the relationship between tobacco sales in different countries
and lung cancer rates, and fluoride in water supplies and dental caries.
A subset of ecological studies, known as time series studies, is regarded as very
helpful in understanding the influence of short-term fluctuations in air contaminants
on day-to-day changes in population morbidity and mortality after controlling for
factors such as season and air temperature. However, disentangling the effects of
individual contaminants as measured in a mixture such as urban air pollution can be
quite difficult.
To strengthen the design of ecological studies, Nurminen ( 1995 ) recommended
the selection of areas with populations that:
are homogeneously exposed (to minimise within-area exposure variation);
represent different extremes of exposure distribution (to maximise between-area
exposure variations);
are comparable with respect to co-variate distributions (e.g., socio-economic
status, demography); and
use the smallest possible sampling units for ecological analysis.
The largest number of environmental epidemiology studies found in the literature
is of the ecological or cross-sectional type, because they are easier to carry out
and cost less (Thomas and Hrudey 1997 ). However, as noted above and discussed
further below in relation to assessment of causality, such studies may be useful for
identifying potential hazards or hypothesis generation, but they cannot determine
cause and effect.
Characteristics of the various study types are summarised in Table 12.1 .
Epidemiological studies are rarely definitive and a single epidemiological study
cannot establish causality. A “weight of evidence” approach is generally required,
involving the interpretation of integrated information.
Unfortunately experimental interventions such as randomised controlled trials
are rarely available to assist Human Health Risk Assessment.
Table 12.1 Advantages and disadvantages of different observational study designs
Ecological
Cross-sectional
Case-control
Cohort
Probability of:
selection bias
N/A
Medium
High
Low
recall bias
N/A
High
High
Low
loss to
follow-up
N/A
N/A
Low
High
confounding
High
Medium
Medium
Low
Time required
Low
Medium
Medium
High
Cost
Low
Medium
Medium
High
Beaglehole et al. ( 1993 , p. 42)
 
Search WWH ::




Custom Search