Biology Reference
In-Depth Information
vector control interventions that are strongly
infl uenced by the ecology of the setting or by the
biology of the vector.
Thus, if we wish to use the same methods
for meta-analysis of vector control trials, we
must give critical scrutiny to the assumption
that a set of trials are all asking the same
question, and should be expected to produce
the same results, if these trials were carried out
in dif erent settings, and with dif erent vectors.
What is needed to justify this assumption?
What indicators can we use to reassure
ourselves that the environment and the vectors
are sui ciently similar to justify retaining this
assumption?
One way to approach this issue is to con-
sider the causal chain, the step-by-step process,
by which environmental and vector control
interventions bring about their epidemiological
benefi ts. With individual medical interventions,
the primary causal chain is internal, and takes
place entirely within the body of the human
subject. Vector control interventions, by con-
trast, have direct ef ects on the environment.
The causal chain leading to health outcomes is
external and larger in scale, and is mediated by
environmental features that must be expected to
vary from place to place. For example in the case
of malaria vector control, these environmental
features would include the local mosquito
population and its breeding sites, host choice,
patterns of movement, longevity and biting
habits. Such characteristics are locally specifi c;
they are products of the interaction between the
intrinsic biology of the particular vector species
and the local physical and ecological environ-
ment.
This has important methodological con-
sequences: three of these are considered here.
The fi rst arises from the area-wide ef ects of a
vector control intervention. Many vector control
interventions have ef ects on the local vector
population that are spread over an area, the size
of which depends on how far mosquitoes tend to
fl y after being exposed to the intervention and
before transmitting infection (or how far they
would have fl own if they had not been killed or
otherwise rendered harmless). Hence, in order
to include such ef ects, the experimental unit for
a vector control trial must be an area, not an
individual, and this area must be large relative to
mosquito movement.
In practice, most modern vector control
trials use community-randomized designs, in
which a group of villages is identifi ed, and then
each village is randomly assigned to be given one
or other of the interventions to be compared.
These methods were fi rst developed for, and
proved successful in, the ITN trials of the 1990s
(Lengeler, 2004). The idea here is that each
village forms a 'transmission unit', so that there
is free movement of parasites between hosts and
vectors within each unit, but negligible amounts
of movement between units. In other words, the
transmission unit assumes the existence of a
clear boundary between the causal processes
inside the unit and those in neighbouring units,
and is therefore analogous to the individual
human subject in a clinical trial. The concept of
a transmission unit has proved to be very useful,
but it is of course a convenient fi ction. In fact,
there can be substantial mosquito movement
between communities, depending on the setting,
especially the relative distribution of breeding
sites and hosts (Gillies, 1961; Rawlings
et al
.,
1981). For example, in an early mark-recapture
study with female
Anopheles gambiae
s.l. in
Tanzania, most recaptures occurred within 3 or
4 days of release, and the mean distance between
release and recapture was about 1.5 km (Gillies,
1961). Presumably, in the 10 to 12 day interval
between such a mosquito picking up a malaria
infection and then becoming infective, the mean
distance travelled is likely to be even larger. In
some cases, inter-village movement of mos-
quitoes can be mediated by the relative position
of breeding sites and potential hosts, and the
female mosquitoes' need to commute repeatedly
between a place to lay her eggs and a place to
take a blood meal (Gillies, 1988). This is thought
to have prevented the detection of the mass-
ef ects of ITNs in village-level mosquito
populations in The Gambia (Quinones
et al
.,
1998). Thus, the assumption that a village is a
'transmission unit' has been more or less
vindicated in some trials but not all: there have
been exceptions where the results were diluted
by movement of mosquitoes between experi-
mental units. The risk of contamination between
experimental units is dii cult to measure, but it
must always be considered.
The fact that vector control interventions
work through mechanisms that are primarily
external and ecological, not internal and
Search WWH ::
Custom Search