**3. Underutilization of existing methodology for informing malaria vector control programmes**

While the methodological limitations described above merit investment, a far bigger limitation is underutilization of long-established and widely-accepted methodologies to inform vector control product development, deployment and assessment. It is remarkable just how few study sites are available with consistently-collected, long-term legacy data that capture longitudinal trends for coverage with important interventions along with both entomological (human biting rate, sporozoite infection prevalence, entomologic inoculation rate) and epidemiological (parasitological incidence and prevalence, as well as disease burden) outcomes to enable satisfactory analysis. Dielmo is a rare exception on the vast continent of Africa, where the same entomological methods for monitoring vector population densities and malaria inoculation rates have been continuously applied in a consistent manner for more than two decades [33]. More recently, these vector population dynamics surveys have been supplemented with repeated characterisations of behavioural interactions between humans and mosquitoes [34]. These additional measurements of human exposure distribution across indoor and outdoor environments at different times of the day certainly help explain why robust residual transmission persists, and reveals worrying signs of a worsening situation [34, 35].

two of the four villages for which a full minimum set of parameters for modelling malaria transmission based on local measurements (**Figure 3D**) relate to a single species, neither of which relied on disaggregation of species-specific data, because only one sibling species from

**Figure 2.** A schematic illustration of how the different trajectories of intervention limitations and failures can be captured and distinguished through longitudinal surveillance of vector population and transmission dynamics (adapted from [6]).

Entomological Surveillance as a Cornerstone of Malaria Elimination: A Critical Appraisal

http://dx.doi.org/10.5772/intechopen.78007

407

There are also limitations to the quality of data collection, archiving and analysis that result in most available entomological measurements being reported at the level of species groups or complexes, rather than disaggregated on a species-by-species basis. For example, despite clear evidence for differences among species in both the mechanisms of insecticide resistance and in the prevalence of resistance phenotypes [41–44], the species-specific data presented in **Figure 3A** represents only 27% of the total available [45]. The remaining 73% of archived data represent aggregated mortality rates for mixtures of two or more undifferentiated species [45]. In some cases, species-level classification simply was not conducted. In others, species identification was conducted but the bioassay results were only provided for the pooled species. In other examples, the species data was compromised because only a subset of the mosquitoes assayed, for example only bioassay survivors, were identified to the species level. This lack of species-specific data reduces the power to investigate trends in insecticide resistance and to detect associations between resistance in wild vector populations and malaria transmission experienced by the human population. The data in **Figure 3B** represents only the small fraction (12%) of all available data on human blood indices that unambiguously relates to a single species rather than an undifferentiated mixture of two or more. Only three [46–48] of the small handful of estimates for the proportion of human biting exposure occurring indoors or while asleep (**Figure 3C**) relate to a single, clearly identified and

Generic (e.g., Microsoft Access®) or freely available (e.g., mySQL®) relational database applications have been adopted as standard tools and used ubiquitously by almost all epidemiologists and field biologists for decades, but medical entomologists generally lag far behind, especially in low income countries. If links between data fields are lost they cannot

within the relevant complex was abundant.

disaggregated species.

While this intensively studied village provides a valuable illustration of how informative such longitudinal surveillance can be, it is not necessarily representative of other parts of Senegal, much less any other country in Africa [35]. National malaria control or elimination programmes all need their own, nationally-representative set of surveillance sites like Dielmo, where malaria transmission is continually and indefinitely monitored using consistent methods. Such platforms are needed to reliably monitor the dynamics of malaria vector populations and transmission intensity across all major ecological and epidemiological strata, so that the limitations and failures of interventions can be identified, distinguished (**Figure 2**), investigated and responded to.

Looking more broadly at programmatically-relevant entomological measurements, insecticide susceptibility testing is now widespread, but measurements of important mosquito and human behaviours are remarkably sparse (**Figure 3**). The species identity of blood hosts that mosquitoes feed upon has long been recognized as a crucial determinant of malaria transmission intensity and an indicator of intervention impact [6, 36–39]. Although adequate field and laboratory methodology for surveying the blood meal choices of most vector species have been available for over 50 years, species-specific reports of this metric remain remarkably scarce for all but a few key vector species (**Figure 3B**). The principles of how to weight indoor and outdoor human landing catch data in proportion to survey results for where people spend each time of the night were first outlined by Garrett-Jones in 1964 [40], yet today less than a dozen such estimates of how human exposure is distributed have been reported, in most cases for undifferentiated mixtures of vector complexes or groups (**Figure 3C**). Only Entomological Surveillance as a Cornerstone of Malaria Elimination: A Critical Appraisal http://dx.doi.org/10.5772/intechopen.78007 407

**3. Underutilization of existing methodology for informing malaria** 

While the methodological limitations described above merit investment, a far bigger limitation is underutilization of long-established and widely-accepted methodologies to inform vector control product development, deployment and assessment. It is remarkable just how few study sites are available with consistently-collected, long-term legacy data that capture longitudinal trends for coverage with important interventions along with both entomological (human biting rate, sporozoite infection prevalence, entomologic inoculation rate) and epidemiological (parasitological incidence and prevalence, as well as disease burden) outcomes to enable satisfactory analysis. Dielmo is a rare exception on the vast continent of Africa, where the same entomological methods for monitoring vector population densities and malaria inoculation rates have been continuously applied in a consistent manner for more than two decades [33]. More recently, these vector population dynamics surveys have been supplemented with repeated characterisations of behavioural interactions between humans and mosquitoes [34]. These additional measurements of human exposure distribution across indoor and outdoor environments at different times of the day certainly help explain why robust residual transmission persists, and reveals worrying signs of a worsen-

While this intensively studied village provides a valuable illustration of how informative such longitudinal surveillance can be, it is not necessarily representative of other parts of Senegal, much less any other country in Africa [35]. National malaria control or elimination programmes all need their own, nationally-representative set of surveillance sites like Dielmo, where malaria transmission is continually and indefinitely monitored using consistent methods. Such platforms are needed to reliably monitor the dynamics of malaria vector populations and transmission intensity across all major ecological and epidemiological strata, so that the limitations and failures of interventions can be identified, distinguished (**Figure 2**), investi-

Looking more broadly at programmatically-relevant entomological measurements, insecticide susceptibility testing is now widespread, but measurements of important mosquito and human behaviours are remarkably sparse (**Figure 3**). The species identity of blood hosts that mosquitoes feed upon has long been recognized as a crucial determinant of malaria transmission intensity and an indicator of intervention impact [6, 36–39]. Although adequate field and laboratory methodology for surveying the blood meal choices of most vector species have been available for over 50 years, species-specific reports of this metric remain remarkably scarce for all but a few key vector species (**Figure 3B**). The principles of how to weight indoor and outdoor human landing catch data in proportion to survey results for where people spend each time of the night were first outlined by Garrett-Jones in 1964 [40], yet today less than a dozen such estimates of how human exposure is distributed have been reported, in most cases for undifferentiated mixtures of vector complexes or groups (**Figure 3C**). Only

**vector control programmes**

406 Towards Malaria Elimination - A Leap Forward

ing situation [34, 35].

gated and responded to.

**Figure 2.** A schematic illustration of how the different trajectories of intervention limitations and failures can be captured and distinguished through longitudinal surveillance of vector population and transmission dynamics (adapted from [6]).

two of the four villages for which a full minimum set of parameters for modelling malaria transmission based on local measurements (**Figure 3D**) relate to a single species, neither of which relied on disaggregation of species-specific data, because only one sibling species from within the relevant complex was abundant.

There are also limitations to the quality of data collection, archiving and analysis that result in most available entomological measurements being reported at the level of species groups or complexes, rather than disaggregated on a species-by-species basis. For example, despite clear evidence for differences among species in both the mechanisms of insecticide resistance and in the prevalence of resistance phenotypes [41–44], the species-specific data presented in **Figure 3A** represents only 27% of the total available [45]. The remaining 73% of archived data represent aggregated mortality rates for mixtures of two or more undifferentiated species [45]. In some cases, species-level classification simply was not conducted. In others, species identification was conducted but the bioassay results were only provided for the pooled species. In other examples, the species data was compromised because only a subset of the mosquitoes assayed, for example only bioassay survivors, were identified to the species level. This lack of species-specific data reduces the power to investigate trends in insecticide resistance and to detect associations between resistance in wild vector populations and malaria transmission experienced by the human population. The data in **Figure 3B** represents only the small fraction (12%) of all available data on human blood indices that unambiguously relates to a single species rather than an undifferentiated mixture of two or more. Only three [46–48] of the small handful of estimates for the proportion of human biting exposure occurring indoors or while asleep (**Figure 3C**) relate to a single, clearly identified and disaggregated species.

Generic (e.g., Microsoft Access®) or freely available (e.g., mySQL®) relational database applications have been adopted as standard tools and used ubiquitously by almost all epidemiologists and field biologists for decades, but medical entomologists generally lag far behind, especially in low income countries. If links between data fields are lost they cannot

be reinstated and the utility of the unlinked data is severely reduced. The most obvious and common example is a collection of morphologically indistinguishable mosquitoes that are used in an insecticide bioassay, then separated into live and dead mosquitoes, and tested for species and sporozoite infection. If unique identifiers are not assigned to each mosquito and linked to data on bioassay survival, species, sporozoite presence and blood meal for that individual mosquito, then any analysis of the relationships among species, insecticide susceptibility, infection and blood meals is severely hampered. Too often, all that is reported is the sporozoite rate, species composition, etc., for the whole sample, i.e., the aggregate value for all the mosquitoes in the original collection even though they comprise a mixture of two or more different species. This is probably the single most important limiting factor when it comes to species-specific measurements for the variables that matter most, or the interactions

Entomological Surveillance as a Cornerstone of Malaria Elimination: A Critical Appraisal

http://dx.doi.org/10.5772/intechopen.78007

409

In addition to being so limited in quantity and quality, the utility of existing vector bionomic data is also compromised by the fact that it has mostly been collected haphazardly and opportunistically with project-based research funding. As a result, it has typically been collected on scales varying from villages to districts, over only a few years at a time. Different research studies in any given setting typically have different objectives that often necessitate different sampling and trapping approaches, so collating data sets from multiple projects often yields a patchwork of data with substantial temporal gaps and methodological inconsistencies that

**4. Addressing data deficits through programmatic entomological** 

The latest guidelines from WHO for entomological surveillance, monitoring and evaluation offer an excellent new framework for comprehensively applying existing field and analytical methods, and for conducting operational research into improving their use practices in the future [53]. In order for national programmes to make evidence-based decisions about what vector control measures to deploy, and evaluate their ongoing impact, the remarkably diverse arsenal of entomological methods already at our disposal [54] now need to be adapted to programmatic surveillance platforms that are nationally representative of all major ecological and epidemiological strata in the country [6–8, 11]. Such programmatic platforms should emphasize the absolute minimum number of essential entomological metrics, with strong data quality control and assurance processes to maximize confidence and minimize ambigu-

Data quantity and quality obviously trade off against each other, especially when working across very large geographic areas, so it is often necessary to select the smallest number of surveillance sites required to adequately represent all major epidemiological and ecological strata in the country. Based on our experience, a minimum of five surveillance sites per stratum is suggested to adequately capture variation within each stratum. The recent WHO guideline of one sentinel site per million people [53] also represents a good benchmark that

between them.

confound unambiguous interpretation [6–8, 11].

**surveillance platforms**

ity of interpretation (**Figure 4**).

**Figure 3.** The global distribution of reported measurements for (A) insecticide resistance, (B) the proportions of bloodmeals obtained from humans, (C) the proportion of human exposure to bites occurring indoors, for *Anopheles* vectors of malaria, and (D) the only four locations, we are aware of, where estimates for the mean adult biting density, survival and human blood index for even a single vector, as well as human population size and infectiousness are all available, so that malaria transmission and control can be modelled in a site-specific manner based on local estimates of these parameters. Panels A and B respectively represent the species-specific subset of all insecticide susceptibility bioassay [45] and human blood index [7, 49] data collated by the time they were most recently published as dataset summary reports. The studies represented in Panel C include only three reports of species-specific estimates [46–48] for the proportion of human biting exposure occurring indoors, with the remaining handful all relating to undifferentiated mixtures of species in a complex or group [34, 50, 51]. Panel D represents only four locations in only three countries (Nigeria, Tanzania, Papua New Guinea), all of which were small rural villages with intense transmission and anthropophagic vectors (otherwise at least some of these parameters would probably have been impossible to measure), but nevertheless yielded remarkably different vector-parasite demography patterns and suitability for various complementary vector control measures beyond LLINs and IRS [26, 52].

be reinstated and the utility of the unlinked data is severely reduced. The most obvious and common example is a collection of morphologically indistinguishable mosquitoes that are used in an insecticide bioassay, then separated into live and dead mosquitoes, and tested for species and sporozoite infection. If unique identifiers are not assigned to each mosquito and linked to data on bioassay survival, species, sporozoite presence and blood meal for that individual mosquito, then any analysis of the relationships among species, insecticide susceptibility, infection and blood meals is severely hampered. Too often, all that is reported is the sporozoite rate, species composition, etc., for the whole sample, i.e., the aggregate value for all the mosquitoes in the original collection even though they comprise a mixture of two or more different species. This is probably the single most important limiting factor when it comes to species-specific measurements for the variables that matter most, or the interactions between them.

In addition to being so limited in quantity and quality, the utility of existing vector bionomic data is also compromised by the fact that it has mostly been collected haphazardly and opportunistically with project-based research funding. As a result, it has typically been collected on scales varying from villages to districts, over only a few years at a time. Different research studies in any given setting typically have different objectives that often necessitate different sampling and trapping approaches, so collating data sets from multiple projects often yields a patchwork of data with substantial temporal gaps and methodological inconsistencies that confound unambiguous interpretation [6–8, 11].
