Partial indoor residual spraying with pirimiphos-methyl as an effective and cost-saving measure for the control of Anopheles gambiae s.l. in northern Ghana

Abstract

The scale up of indoor residual spraying (IRS) and insecticide treated nets have contributed significantly to global reductions in malaria prevalence over the last two decades. However, widespread pyrethroid resistance has necessitated the use of new and more expensive insecticides for IRS. Partial IRS with pirimiphos-methyl in experimental huts and houses in a village-wide trial was evaluated against Anopheles gambiae s.l. in northern Ghana. Four different scenarios in which either only the top or bottom half of the walls of experimental huts were sprayed, with or without also spraying the ceiling were compared. Mortality of An. gambiae s.l. on partially sprayed walls was compared with the standard procedures in which all walls and ceiling surfaces are sprayed. A small-scale trial was then conducted to assess the effectiveness, feasibility, and cost of spraying only the upper walls and ceiling as compared to full IRS and no spraying in northern Ghana. Human landing catches were conducted to estimate entomological indices and determine the effectiveness of partial IRS. An established transmission dynamics model was parameterized by an analysis of the experimental hut data and used to predict the epidemiological impact and cost effectiveness of partial IRS for malaria control in northern Ghana. In the experimental huts, partial IRS of the top (IRR 0.89, p = 0.13) or bottom (IRR 0.90, p = 0.15) half of walls and the ceiling was not significantly less effective than full IRS in terms of mosquito mortality. In the village trial, the annual entomological inoculation rate was higher for the unsprayed control (217 infective bites/person/year (ib/p/yr)) compared with the fully and partially sprayed sites, with 28 and 38 ib/p/yr, respectively. The transmission model predicts that the efficacy of partial IRS against all-age prevalence of malaria after six months would be broadly equivalent to a full IRS campaign in which 40% reduction is expected relative to no spray campaign. At scale, partial IRS in northern Ghana would have resulted in a 33% cost savings ($496,426) that would enable spraying of 36,000 additional rooms. These findings suggest that partial IRS is an effective, feasible, and cost saving approach to IRS that could be adopted to sustain and expand implementation of this key malaria control intervention.

Figure 1

Partial IRS Study Sites in northern Ghana. Partial IRS was evaluated in an experimental hut study at one site (yellow star) in Tamale Metropolitan District and subsequently in a village-scale trial conducted in four districts in northern Ghana. In Kumbungu, Bunkpurugu Nakpanduri, and Gushegu districts one community was partially sprayed (red/pink semi-circles) and one was fully sprayed (red circles). The trial also included two unsprayed communities (green circles) in Tamale Metropolitan District.

Figure 2

Anopheles gambiae s.l. mean total mortality in experimental huts with different IRS scenarios up to 18 weeks post-intervention. Post-spray mosquito mortality decreased gradually for fully and partially sprayed treatment scenarios and the rate of decrease was consistent over time. Mosquito mortality in experimental huts where upper wall + ceiling (solid red line) or lower wall + ceiling (solid blue line) was not significantly different to that observed in fully sprayed huts (solid black line), whereas spraying only the upper walls (dashed red line) or lower walls (dashed blue line) resulted in significantly lower mosquito mortality. Mosquito mortality in untreated huts (solid gray line) was consistently low throughout the study. A spike in mortality was observed across all treatments in weeks 15–16, during which the mean weekly temperature (gray bars) in experimental huts shown also increased.

Figure 3

Predicted outcomes of Anopheles mosquito feeding attempts following full or partial IRS with pirimiphos-methyl CS. (a) A logistic binomial function was fit to the mosquito mortality measured in experimental hut trials over time. The observed data from the experimental huts are overlaid as points to show the capacity of the statistical model to fit the data. (b) The relative deterrence of mosquitoes was fit to the initial experimental hut data (overlaid points) and the decay of the deterrence effect was then assumed to mirror that of mortality given that it is infeasible to rotate the sprayed huts and account for local mosquito populations being reduced by effective insecticide (see also Supplementary Fig. S7 and Sherrard-Smith et al.). These trials were completed in northern Ghana where either all walls and ceilings (full IRS) were sprayed (orange), or partial IRS was conducted with pirimiphos-methyl CS spraying the upper walls and ceiling (green), lower walls and ceiling (dark blue), upper walls (red) or lower walls (blue) only. Supplementary Fig. S7 shows how these data are combined, for each spray strategy, to estimate the probable outcome of mosquito feeding attempt including that the mosquito is killed, deterred, blood fed successfully, or repelled without being killed or blood-feeding.

Figure 4

Mean daily spray operator outputs and insecticide consumption in full IRS and partial IRS communities during the 2019 spray campaign in northern Ghana. (a) Mean number of eligible rooms sprayed by a spray operator (SOP) in a day, (b) Time spent spraying a room, and (c) Insecticide consumption expressed as the mean number of rooms sprayed with one bottle of pirimiphos-methyl CS. Different letters on error bars denote statistically significant differences between means at p = 0.05.

Figure 5

Entomological indices of malaria transmission estimated in full and partial IRS and unsprayed control communities in a village scale trial in northern Ghana, recorded from March–December 2019. (a) Human biting rates (HBRs), expressed as mean bites per person per night, of An. gambiae s.l. by treatment. (b) Mean parity rates of An. gambiae s.l. recorded from study sites. (c) Entomological inoculation rates (EIR) of An. gambiae s.l., expressed as total number of infective bites per person per year, by treatment. The annual EIR was estimated from the sum of monthly EIRs between March and December. Different letters on error bars denote statistically significant differences between means of biting and parity rates at p = 0.05. Statistical analysis for EIR were not done.

Figure 6

Model predicted epidemiological impact of partial versus full IRS with pirimiphos-methyl CS. In all panels, the different partial IRS scenarios are shown as: lower walls only (dark blue), upper walls only (light blue), lower walls and ceiling (light green), upper walls and ceiling (dark green), or full spray (red). The model uncertainty (95% uncertainty intervals) is shown in panel (a) and (c) around the mean (line, panel (a)). Panel (a) shows the model simulated all-age malaria prevalence (mean, lines, and 95% uncertainty intervals (shaded polygons)) over time with the spray campaign taking place annually from May 2019 onward. The no spray counterfactual (grey dashed line and shaded polygon) is used to compare with all spray strategies to estimate the efficacy against prevalence (as shown in Table 1 and panel (c) at 6 months after the first spray campaign). (b) The corresponding model estimates of all-age clinical incidence of malaria (mean estimate over time for clinical cases per person per year) for the same trial arms and counterfactual simulation. The vertical dotted lines on panel (a) and (b) in May 2019 indicate the time when spray was deployed. (c) The model predicted efficacy against all-age malaria prevalence relative to no spray campaign (grey dashed lines on panels (a)) at six months post-IRS for each spray strategy respectively. The uncertainty is carried through from the model simulations and parameter uncertainty in IRS efficacy estimates (Supplementary File 2) and shown by the width of the histogram for each spray strategy.