This secondary analysis of entomological surveillance data collected during the ATSB trial in Western Province Zambia indicates that in children 90% of infectious An. funestus bites can be prevented by perfect use of mosquito nets, since biting pressure remains highest indoors and during usual sleeping hours. Children who do not sleep under ITNs receive the majority of bites while indoors (91%) and are estimated to receive 13.6 infectious An. funestus bites over a six-month transmission season. Among children using ITNs, the number of bites received is much lower, but bites are primarily experienced while outside (84%), and children consistently using ITNs are estimated to have received on average 1.3 infectious An. funestus bites over a six-month season. Consequently, while the protective effect of ITNs is high in this setting among children, persistent malaria transmission largely occurs outdoors and represents a substantial risk to the population and public health burden, considering the high level of An. funestus biting and the high sporozoite positivity of the vector.
Directly-measured EIR, generated directly from the mean of indoor and outdoor landing rates reported by HLC collectors and sporozoite-positivity, was estimated at 0.07 infectious bites per person per night across the study site. However, use of human behaviour-adjusted EIR, which combines sporozoite-positivity, each child’s reported location (outside, inside but not under net, or inside and under a net) by hour and cluster- and month-specific hourly biting rates to estimate the actual bites that would be received by an individual, reveals further nuances. Among children who do not use nets, mean human-adjusted EIR was estimated at 0.08 infectious bites per person per night, slightly higher than the directly-measured EIR (0.07), as a result of children being exposed to the relatively higher indoor biting rates throughout the night. However human-adjusted EIR among net users is much lower, at 0.007 infectious bites/person/night, reflecting the personal protection offered by nets while indoors and sleeping under a net.
Substantial heterogeneity was observed in landing rates and sporozoite positivity between clusters across the study site, such that net users in the highest biting rate clusters were estimated to receive more infectious bites per night than children not using nets in relatively lower biting rate clusters. This emphasises that even in a setting with relatively high overall ITN use (83% among participating children, and > 70% among the general population), children may still experience substantial exposure to biting before entering their net for the night and that even in high transmission settings there remains heterogeneity in malaria risk over time and space.
Persistent, or residual, transmission in this setting is estimated as a human behaviour-adjusted EIR of 1.3 infectious bites per person per six-month season, despite very high ITN household possession (> 90%) and use (> 70%). Malaria prevalence has been shown to have an approximately linear relationship with the log of annual EIR [25], highlighting that EIR needs to be reduced well below 1 per year to have any observable impact on malaria prevalence. At our study site, malaria prevalence was 51.9% among individuals aged six months and older in the baseline year [7] (prior to top-up ITN distributions and assignment of ATSB to the intervention arm) and remained just over 50% throughout the trial [5]. Corresponding to this high infection prevalence in the human population, 3.3% of all tested An. funestus were CS-positive. While ITN use is high in this setting, it is clear that persistent malaria transmission contributes to maintaining a high parasite load and public health burden in the population.
Our estimate that 9% of all bites received by individuals who don’t use an ITN occur outside is consistent with the findings of a meta-analysis describing the role of outdoor biting [11]. However, among individuals using ITNs, outdoor biting contributes most of the exposure to biting since our study population, consisting only of children, usually go under their nets immediately after entering their sleeping structure. Our behavioural data describing net use and locations during the evening is limited to children, and it is likely that adults may go to bed later and spend additional time outside in the evenings, extending the period when they are exposed to persistent infectious biting, and playing a key role in sustaining very high malaria transmission across this population. The proportion of bites that can be prevented by using a net is likely to be substantially lower for adults than the estimate of 90% in children, whose use of nets is more effective because they spend more time under them than adults during the night.
Our analysis method combines indicator estimates generated at different levels, and while we have attempted to handle uncertainty in landing rates, sporozoite positivity, and resulting EIR using delta-method approximations, the uncertainties presented should be interpreted cautiously due to challenges in estimating and propagating covariance structures through the various levels of analysis. Considering the growing interest in estimating human behaviour-adjusted EIR [26–33], which joins estimates of biting rates with estimates of behaviour, often from slightly different individuals or populations, there is a need for further methodological guidance describing best practices for uncertainty generation in these calculations.
The current analysis focusses on personal protection provided by sleeping under ITNs, and community protection effects have not been explicitly estimated. However, since ITN ownership at the study site was > 90% and ITN use > 70%, the mosquito landing rates captured at sampling locations inherently reflect any impact of ITN on overall vector abundance. An. funestus at the study site has high levels of pyrethroid resistance, but resistance has not been detected against the chemical used for IRS in this location, clothianidin, or other IRS chemicals such as pirimiphos-methyl [7]. There was little difference in estimating biting rates when stratifying clusters according to whether they were targeted for IRS, suggesting a limited effectiveness of IRS at the study site that may be attributable to challenges in achieving high coverage in a rural setting with dispersed settlements, optimal timing of IRS campaigns, or other factors. ITNs used at the trial site were a mixture of deltamethrin (Permanet 2.0) and alpha cypermethrin plus piperonyl butoxide (PBO, Veeralin) types [7]. It is possible that the ITNs used had insufficient insecticidal effectiveness to reduce the vector population, resulting in the relatively high observed proportion of An. funestus that were sporozoite positive, and high EIR experienced by non-net users as a result of limited mass effect of ITNs. However, modelling studies indicate that even in settings with high pyrethroid resistance, there are still likely to be community-protection effects from high levels of pyrethroid ITN use [34]. Further studies are required to assess if the extent of persistent malaria transmission in this site could be reduced by use of bioefficacious ITNs or IRS, or combinations of these interventions.
Our estimates of directly-measured and human behaviour-adjusted EIR were limited by conducting mosquito trapping between 18:00 and 06:00 only. Hourly biting rates in the period 05:00–06:00 had not yet dropped to zero, suggesting that our data underestimated biting occurring in the early morning. Recent 24-hour mosquito trapping activities at the study site indicate that biting does occur beyond the usual sampling period of 18:00–06:00, and that expanding the trapping period beyond this period increases measured nightly biting rates by ~ 5% (B. Chanda, personal communication).
Nightly biting rates and sporozoite positivity in our study are broadly similar to those reported from other locations where An. funestus is the primary vector [17, 18, 35, 36], but much lower than biting rates experienced in west African settings where An. coluzzi is the primary vector [33]. In many settings in southern and eastern Africa, An. funestus has become the dominant malaria vector, replacing An. gambiae s.l. [37, 38]. Furthermore, the decreases in sporozoite positivity and corresponding EIR by An. gambiae s.l. since 2000 have not been observed in An. funestus [37]. Recent evidence of indoor morning An. funestus biting after people rise for the day [15, 31], and during school hours [16], further emphasise the challenge of addressing persistent transmission in settings where An. funestus is the primary vector.