Evidence Required for Adoption of New Vector Control Methods in Public Health - Biological and Environmental Control of Disease Vectors - page 192

Thus, after the fi rst small-scale fi eld trials of
ITNs in the mid-1980s, there was an initial
interval of a decade before the fi rst WHO general
recommendation that ITNs should be used in
public health. Then it took another 10 years of
implementation experience and technological
evolution before LLINs were fi nally recognized as
the default method of malaria vector control in
many or most settings. This example illustrates a
very important aspect of the ef ectiveness
question: most forms of vector control have a
niche role, because they are ef ective for only a
few purposes and/or in only a few specifi c
settings. Only a small minority of new vector
control technologies are destined to become the
default intervention of choice across a very wide
range of settings, regardless of geographic
variations in the ecology and behaviour of local
vector species.

11.4 The Constraints on Meta-
analysis of Vector Control
Intervention Trials

For many public health interventions, it is now
regarded as standard good practice to summarize
evidence from trials by conducting a Cochrane
review. The original purpose of The Cochrane
Collaboration (2008) was to summarize the
results of randomized-controlled clinical trials
of medical interventions. In formal terms, these
summary methods are based on the assumption
that the various studies have all asked more or
less the same question, and in the absence of
sampling error and experimental bias, they
would have produced more or less the same
results.
In order to understand this assumption, it is
helpful to recall that Cochrane methods, like
other clinical trials methods, were originally
developed for interventions where the treatment
is administered to an individual patient, and the
primary clinical outcome is also measured in
the same individual. With such interventions,
the main causal pathway from intervention to
outcome is located entirely within the body of
that individual, and the individual human
therefore of ers a natural experimental unit of
replication. This allows study populations to be
standardized using well-established biomedical
defi nitions, and justifi es meta-analysis across
trials. The fact that the human body represents
such a convenient, consistent and defi nable
experimental unit makes it relatively straight-
forward to compare and summarize multiple
studies of the same medical intervention carried
out by dif erent investigators, at dif erent times
and in very dif erent settings. Hence, a series of
trials can all ask the same question, even if they
are carried out on dif erent continents and many
years apart.
How, then, can the same approach be
extended to large-scale vector control inter-
ventions? Over the last two decades, Cochrane
methods have been successfully adapted and
applied to a wide range of large-scale public
health interventions. The methodological issues
that can arise in this process have been given
careful attention, and principles and methods to
deal with them have been developed. However,
less attention has been given to the specifi c issues
raised when one applies Cochrane methods to

11.3 The Effectiveness Question as a
Process of Generalization

Ef ectiveness is not a simple yes or no once-and-
for-all question. Of course the fi rst step is to do a
trial, but this can only show whether or not the
new intervention is ef ective when deployed in a
specifi c setting for a particular purpose, and
using a defi ned method of delivery. However
successful, a single trial does not necessarily
indicate that this method of control will be
equally ef ective when deployed using a dif erent
mode of delivery or in another setting. Ideally, we
would like to ask the 'Is it ef ective?' question a
thousand times, covering every possible setting
and every possible delivery method. To put it
another way, having established in a trial whether
a new form of vector control is ef ective delivered
in a particular way and in a particular setting, we
then have to establish where and when it is likely
to be ef ective elsewhere, defi ning the scope and
limits of its ef ectiveness, before we can issue an
adequate set of public health recommendations.
Therefore, a large part of the ef ectiveness
question is a process of generalization,
interpolating between and extrapolating from a
limited set of trials to the much wider range of
circumstances in various settings in dif erent
parts of the world and operational modes of
delivery that the intervention will face when it is
implemented.