Hazard Assessment and Contaminated Sites - Dealing with Contaminated Sites: From Theory Towards Practical Application - page 542

( 1997 ) note, apart from temporality (whereby a putative cause must precede the
effect) there are no necessary and sufficient criteria for determining whether an
observed association is causal. Thus the term “guidelines” is more appropriate
than the slightly more absolute “criteria”; and there is not necessarily an easy
epidemiological road-map to finally determine causation.
With environmental health in particular, much decision-making rests on a
“weight of evidence” approach rather than definitive proof of cause, which is com-
monly not available - hence the final concept, “Judging the evidence” in Table 12.2 ,
is particularly relevant.
These guidelines are ordered in a logical sequence for making judgements on
causality. They are not weighted equally, and their relative contribution to a final
judgement will vary from one situation to another (Thomas and Hrudey 1997 ).
Consistency can be demonstrated if several studies give the same result, espe-
cially if a variety of designs is used in different settings since this reduces the
likelihood that all studies are making the same mistake. However, other factors such
as different exposure levels or study conditions may need to be taken into account,
and the best-designed studies should be given the greatest weight. It is important
to note that in environmental epidemiology, reliance on a single pivotal study is the
exception rather than the rule.
The technique of meta-analysis grew out of the need to reduce random error in
clinical trials. Meta-analysis in the context of systematic reviews can be used to pool
the data from well-designed studies, each of which may deal with a relatively small
sample size, in order to obtain a better overall estimate of effect. Meta-analysis has
pitfalls if poor quality studies are included, and needs to be applied with caution
to observational studies - which are less able to control for confounding than ran-
domised trials. Standard methods for conducting and reporting systematic reviews
have been published (Greenhalgh 1997 ). The reader is also referred to an excellent
resource published by NHMRC ( 2000 ), “How to review the evidence: systematic
identification and review of the scientific literature”.
The strongest evidence comes from well-designed and competently conducted
randomised controlled trials. The National Health and Medical Research Council
(1999) places strongest emphasis on evidence obtained from systematic reviews of
all relevant (and well-conducted) randomised controlled trials (“level 1”). 2
However, there are relatively few such trials available for environmental health
hazards that could form the basis for a systematic review. Most apply to the effects of
treatment or prevention campaigns. A rare example is the Melbourne Water Quality
Study which was a blinded study involving real and sham domestic reverse osmosis
water filters and an assessment of acute gastrointestinal disease (Hellard 1999 ).
In practice, most evidence comes from observational studies (e.g., nearly all the
evidence on the health effects of smoking). In well-conducted cohort studies bias is
2 The NHMRC document is oriented towards clinical interventions and clinical practice guideline
development. More recently, advice to guide assessment of epidemiological evidence for envi-
ronmental health practice has been provided in NHMRC ( 2006 ). Ambient Air Quality Standards
Setting: An Approach to Health-Based Hazard Assessment.