Mosquito feeding behavior and how it influences residual malaria transmission across Africa

Abstract

The antimalarial efficacy of the most important vector control interventions-long-lasting insecticidal nets (LLINs) and indoor residual spraying (IRS)-primarily protect against mosquitoes' biting people when they are in bed and indoors. Mosquito bites taken outside of these times contribute to residual transmission which determines the maximum effectiveness of current malaria prevention. The likelihood mosquitoes feed outside the time of day when LLINs and IRS can protect people is poorly understood, and the proportion of bites received outdoors may be higher after prolonged vector control. A systematic review of mosquito and human behavior is used to quantify and estimate the public health impact of outdoor biting across Africa. On average 79% of bites by the major malaria vectors occur during the time when people are in bed. This estimate is substantially lower than previous predictions, with results suggesting a nearly 10% lower proportion of bites taken at the time when people are beneath LLINs since the year 2000. Across Africa, this higher outdoor transmission is predicted to result in an estimated 10.6 million additional malaria cases annually if universal LLIN and IRS coverage was achieved. Higher outdoor biting diminishes the cases of malaria averted by vector control. This reduction in LLIN effectiveness appears to be exacerbated in areas where mosquito populations are resistant to insecticides used in bed nets, but no association was found between physiological resistance and outdoor biting. Substantial spatial heterogeneity in mosquito biting behavior between communities could contribute to differences in effectiveness of malaria control across Africa.

Keywords: Anopheles; LLIN efficacy; Plasmodium falciparum; malaria transmission; vector interventions.

Fig. 1.

The systematic review process for mosquito biting behavior and human activity.

Fig. 2.

A summary of the raw data from a systematic literature review and collation of mosquito activity data from country reports produced for the President’s Malaria Initiative, PMI. (A) Geographic location of data on hourly mosquito activity indoors and outdoors (literature review: purple open squares; PMI reports: blue closed squares) and times at which people went indoors (green circles) or to bed (yellow triangles). (B and C) The mean proportion of people who were either indoors (B) or in bed (C) over 24 h for each study, regardless of the presence of an intervention. Colors correspond to the country represented: Benin (light blue), Burkina Faso (light purple), Cote D’Ivoire (red), Equatorial Guinea (dark blue), Ghana (pink), Kenya (green), Mozambique (brown), Tanzania (gray), and Zambia (yellow). (D–F) The mean proportion (lines) and range (shaded area) of mosquito activity in the absence of personal vector control: (D) An. gambiae s.l., (E) An. arabiensis, and (F) An. funestus s.l. during the night either indoors (blue, darker shade, solid line) or outdoors (black, lighter shade, dashed line).

Fig. 3.

Temporal and spatial heterogeneity in the estimated proportion of mosquito bites taken when people are indoors or in bed. Combined data from the systematic review (square symbols) and country reports for the President’s Malaria Initiative (PMI) (triangles). In A and B points show individual point estimates and solid line represents the linear mixed-effects model estimate of how the proportion of bites has changed over time (country is included as a random effect and the trend in the mean estimate across all countries is shown; SI Appendix, Table S1). Point colors denote countries as per C and D. In C and D raw data are plotted with the median estimate as the black line and box-plot bodies and whiskers denote 25% and 95% ranges in the estimates. Asterisks mark countries with estimates significantly lower (blue) or higher (red) than Benin in the linear fixed effects model (SI Appendix, Table S1), and the number of samples for each country is noted at the bottom of each panel. ***P < 0.001, *P < 0.05, ^•P < 0.1 significance level. The upward-pointing arrow on these significance levels in red indicates the estimate is significantly above that of Benin.

Fig. 4.

Estimated impact of outdoor biting on the prevalence of malaria and residual transmission. (A) Illustration of the public health impact of LLINs and IRS when used at 100% coverage and how this depends on the proportion of bites taken when people are indoors. Lines show malaria prevalence in 2- to 10-y-old children in a high-transmission, perennial setting with a mixed mosquito species population (50% An. gambiae s.s., 25% An. arabiensis, 25% An. funestus). Universal use of LLIN and IRS at time 0 is shown for communities where a different percentage of mosquito bites is taken when people are indoors, be it 98% (historical value, solid line), 88% (approximate current estimation, dotted line; Table 1), 78% (dashed line), or 58% (dotted-dashed line). (B) Estimates of residual transmission if high proportions of mosquito bites were taken when people are indoors. Shaded region indicates the annual entomological inoculation rate (EIR) measured 3 y after the introduction of LLIN and IRS at 100% coverage (see color scale). (C) Residual transmission (EIR) if 10% fewer bites were taken when people are indoors (comparable to the drop estimated between 2003 and 2018; Fig. 3A). Such a difference in outdoor biting is predicted to have a substantial impact on malaria prevalence. (D) The absolute increase in malaria prevalence (in 2- to 10-y-old children) estimated from the higher outdoor biting (malaria prevalence resulting from the difference between B and C). Note that the level of residual transmission and malaria prevalence in B–D is intended to be illustrative of the variance seen across Africa. Results should not be overinterpreted as transmission is averaged at an administrative unit-1 scale and there will be substantial variability within these units.

Fig. 5.

The occurrence of outdoor biting and physiological resistance and its predicted joint public health impact. (A) Field data showing estimates of the proportion of mosquito bites taken indoors (in people without direct personal protection) and how this varies with the level of physiological resistance to pyrethroid insecticide observed in the area (assessed as the percentage of mosquito survival during discriminatory dose bioassay susceptibility testing). There was no significant association between the level of outdoor biting and physiological resistance observed in the field. Symbols and colors represent the country of data collection (see key). (B) Model predictions for the reduction in the number of clinical cases that can be achieved by indoor interventions given the level of indoor biting. Line color indicates coverage of LLIN or IRS and line type denotes the level of pyrethroid resistance (solid line = no resistance, dashed line = high resistance). The reduction in effectiveness is predicted to be nonlinear in sites where there is no physiological resistance to pyrethroids (effectiveness is greatest when the proportion of bites taken indoors is high). (C) Model predictions for the efficacy of indoor interventions with varying levels of physiological resistance. For this setting there is a critical point, where ∼60% of mosquitoes survive during bioassay testing, when the efficacy of indoor interventions falls at a faster rate (especially when there is moderate outdoor biting). Line color as in B, although type denotes level of indoor biting (solid = high, dotted = low). Using a nonpyrethroid long-lasting IRS (Actellic300CS, parameterized as per ref. 32) mitigates the lost efficacy of LLINs that is due to physiological resistance. (D) The relative efficacy against prevalence in 2- to 10-y-olds is affected by both reduced indoor biting and physiological resistance to pyrethroids when LLINs are used at 100% coverage. At low levels of pyrethroid physiological resistance, the reduction in indoor biting has a larger impact.