Evidence Required for Adoption of New Vector Control Methods in Public Health - Biological and Environmental Control of Disease Vectors - page 194

physiological, also has consequences for the task
of standardization between trials. Cochrane
methods already specify the need, when
summarizing across a series of clinical trials, to
defi ne and standardize the key variables of the
trial, especially the human subjects, the
interventions to be compared, and the outcome
measures in each case. When doing this for a set
of vector control trials, we must take into
account the fact that the causal chain from
intervention to health outcome is mediated by,
and dependent upon, the local environment.
Hence, an observation that an intervention is
ef ective in one setting does not necessarily mean
that it will be ef ective in another setting where
the environment (including the mosquitoes) is
dif erent. Therefore, for a systematic review of
vector control trials, it seems that additional
attention also needs to be paid to defi ning and
standardizing the ecological nature of the
environment, especially the vector population. If
a series of trials of a given vector control
intervention are conducted in settings where
there is substantial variation in the local ecology
(especially vector biology and behaviour), then it
is no longer safe to assume that they are all
asking the same question and, in the absence of
bias and sampling error, would all produce the
same answer. Thus, in order to summarize
across a series of vector control trials, it is not
enough to defi ne only the human study
population, the intervention and the outcome
measure; in addition, it is also necessary to
defi ne and standardize the local ecology of the
vector.
Finally, we must also consider how the
environmental circumstances can af ect the
intervention itself. The implementation methods
used for some vector control interventions - for
example LLINs and IRS - are highly standardized;
the technical operations are carried out in more
or less the same way whatever the setting. Other
vector control methods cannot be so
standardized. In particular, larval source
management (LSM) by targeting vector breeding
sites is generally recognized as an intervention
that must be carefully adapted to suit the local
ecological conditions (WHO, 2006). This is
because there is so much local variation in the
nature of the breeding sites, not only by season
and between species, but also from place to place
for a single species. In practice, this process of

adapting the intervention to suit the local
setting consists largely of identifying features of
the environment that present opportunities,
and obstacles to be overcome, for delivery of the
intervention. Thus, for some vector control
interventions, it is not really possible to defi ne
the intervention without defi ning the
environment in which it is to be delivered. For
example, larvivorous fi sh have been used in
attempts to control malaria in subterranean
water storage tanks in arid areas of Somalia,
rice fi elds in Tajikistan and foothill streams in
northern Thailand (Alio et al ., 1985; Mouchet
et al ., 2004). However, although these examples
all have fi sh in common, not much else is the
same: the operational procedures, the
landscapes and the local vectors all dif er from
each other completely. Hence, it is probably not
appropriate to treat them as replicates of the
same intervention, or to expect the same
outcomes.
These arguments all point to the conclusion
that, when summarizing trial data in order to
ask whether a vector control intervention is
ef ective, it is important to take environmental
and ecological circumstances into account. And
if this is correct, then presumably the same
consideration is equally necessary when
generalizing from trial data in order to defi ne
where and when it will be ef ective, i.e. the scope
of the usefulness of the intervention in question.

11.5 Factors to be Considered in
Generalizing from Vector Control
Trials

This section discusses the basic principles that
must be considered in order to make useful
generalizations on the basis of vector control
trial data.
The fi rst step, again, is to make sure we
have a good understanding of exactly how the
intervention works. We need a complete
account of the chain(s) of cause-and-ef ect all
the way from the intervention's primary ef ects
on individual mosquitoes, through the
resulting changes at the level of the vector
population, to the epidemiological benefi ts in
people. There may be more than one such
pathway, and these may be direct or indirect. It
is useful to refer to this overall process, whereby