Impact of Late–Rainy Season Indoor Residual Spraying on Holoendemic Malaria Transmission: A Cohort Study in Northern Zambia

Abstract

Background

Indoor residual spraying (IRS) is a malaria control strategy implemented before the rainy season. Nchelenge District, Zambia, is a holoendemic setting where IRS has been conducted since 2008 with little impact on malaria incidence or parasite prevalence. Pre–rainy season IRS may not reduce the post–rainy season peak abundance of the major vector Anopheles funestus.

Methods

A controlled, pretest-posttest, prospective cohort study assessed the impact of late–rainy season IRS on malaria prevalence, incidence, hazard, and vector abundance. A total of 382 individuals were enrolled across 4 household clusters, of which 2 were sprayed in April 2022 toward the end of the rainy season. Monthly household and individual surveys and indoor overnight vector collections were conducted through August 2022. Multivariate regression and time-to-event analyses estimated the impact of IRS on outcomes measured by rapid diagnostic tests, microscopy, and quantitative polymerase chain reaction.

Results

Among participants, 72% tested positive by rapid diagnostic test at least once, and incidence by microscopy was 3.4 infections per person-year. Residing in a household in a sprayed area was associated with a 52% reduction in infection hazard (hazards ratio, 0.48; 95% CI, .29–.78) but not with changes in incidence, prevalence, or vector abundance. The study-wide entomologic inoculation rate was 34 infectious bites per person per year.

Conclusions

Monthly tracking of incidence and prevalence did not demonstrate meaningful changes in holoendemic transmission intensity. However, hazard of infection, which provides greater power for detecting changes in transmission, demonstrated that late–rainy season IRS reduced malaria risk.

In a cohort study in northern Zambia, individuals in areas sprayed at the end of the rainy season had a reduced hazard of infection but no difference in odds of parasite prevalence when compared with individuals in unsprayed areas.

Despite decreases in the global burden of malaria in the early 21st century, recent gains have stalled, and control efforts have remained stagnant in many high-transmission settings [1–7]. Vector control is a core component of malaria control, consisting primarily of the distribution of long-lasting insecticide-treated nets (LLINs) and indoor residual spraying (IRS). While LLINs are generally effective across low- to high-transmission settings, evidence in support of the effectiveness of IRS in high-transmission settings is less conclusive [2, 7–9]. In addition to large vector populations and parasite reservoirs in these areas, the impact of IRS is further reduced by varied implementation coverage and quality, insecticide resistance, and multiple ecologies supporting sympatric vector species [10, 11]. Understanding when and where to implement IRS is important in light of different transmission settings and seasonal patterns. IRS is conventionally implemented before the rainy season due to the relative ease of travel on dry roads, the assumption that vectors and malaria cases peak during and just after the rains, and the understanding that most insecticide formulations have residual efficacy for at least 5 months after spraying [12–15]. Pre–rainy season IRS remains a core strategy recommended by the World Health Organization (WHO), even in holoendemic perennial (nonseasonal) transmission settings [16].

Nchelenge District in Luapula Province, Zambia, has intense perennial transmission and a year-round parasite prevalence ≥30% by microscopy [17]. Plasmodium falciparum is the primary species causing infection, though Plasmodium malariae is found occasionally [18]. Despite increasing IRS coverage with different insecticides since 2008, ongoing cross-sectional surveillance has not detected sustained declines in epidemiologic measures of parasite prevalence [17, 19]. One possible reason is that spatiotemporal vector dynamics are known to be complex and pose challenges to malaria control [20, 21]. Anopheles gambiae s.s. peaks during the rainy season in Nchelenge District, whereas the much more abundant vector Anopheles funestus peaks several months after the rains (typically June or July) [20]. Given that vector exposure is perennial, one hypothesis for the minimal impact of annual pre–rainy season IRS is that the campaign timing does not adequately affect the A funestus population. This study evaluated the impact of an IRS campaign that took place at the end of the rainy season (April) by investigating whether the campaign led to a measurable and sustained reduction in vector abundance, malaria incidence, and parasite prevalence through a prospective longitudinal cohort study with a controlled pretest-posttest analytic design. The study aimed to test the hypothesis that timing IRS campaigns to target local vector systems in perennial transmission settings may reduce transmission and lower parasite prevalence. Reductions of noteworthy magnitude might motivate a policy and operational change in IRS timing.

METHODS

Study Site

Nchelenge District (population, 233 697) abuts Lake Mweru and is bordered by the Democratic Republic of the Congo [22]. Commerce is dominated by fishing but increasingly complemented by farming in the inland areas [23]. There is a rainy season between November and April, a cool dry season from May to July, and a hot dry season from August to October. Annual IRS occurs between September and December and used pyrethroids from 2008 to 2010, carbamates from 2011 to 2012, pirimiphos-methyl (Actellic 300CS) from 2014 to 2018, deltamethrin and clothianidin (Fludora Fusion) from 2019 to 2020, and clothianidin (SumiShield 50WG) in 2021 [24–26]. Historically, only portions of the district were sprayed, but a shift to maximize coverage resulted in an increase in district-wide household IRS coverage from 29% in 2014 to 81% in 2020 [19]. Continuous LLIN distribution occurs at antenatal and under-5 health center visits. The last universal net distribution before this study was in 2017.

Study Design

In 2022, a controlled pretest-posttest study was conducted in a prospective longitudinal cohort of individuals from households in 4 settlements in Nchelenge District to evaluate the impact of IRS with SumiShield at the end of the rainy season. Four household clusters were identified, and 65 households were randomly selected across the 4 settlements (Figure 1) [27]. Household selection took place in 2 rounds: 50 households were enrolled in March 2022, and 15 additional households were enrolled in April 2022 to account for absent household residents observed in March and meet sample size requirements. Individuals were eligible if they resided in or were visiting a selected household at the time of the survey. Written informed consent was obtained from adults aged ≥16 years and guardians of individuals aged <16 years. Oral assent was obtained from children aged between 12 and 16 years.

Figure 1.

Households in the study area were digitally enumerated by WorldView-3 satellite imagery (2017; Maxar). Four household settlements were identified by manual review of satellite imagery and were selected by proximity to the lake (lakeside) or Kenani Stream (inland) and accessibility by vehicle. Intervention assignment (sprayed vs not sprayed) was stratified by cluster location. A buffer region with a radius of 500 m was applied to each settlement, and the outermost boundaries of the buffer regions for each pair of settlements were confirmed to be separated by at least 1.5 km to reduce spillover effects across settlements. All households in the core and buffer region in spray areas were eligible for indoor residual spraying. Households were eligible for enrollment in the cohort if located in the core, nonbuffer region of each settlement. The study area lies completely in Nchelenge District across several zones, but residents are primarily served by St Paul's Hospital, Kashikishi Clinic, and their affiliated health posts.

Data Collection

From March to August 2022, monthly surveys collected information on household structure and assets, individual demographic information, care-seeking behavior, malaria history and treatment, LLIN use and ownership, travel, and malaria knowledge [28]. Participants were screened with a forehead thermometer and SD Bioline P falciparum histidine-rich protein 2–based rapid diagnostic test (RDT). Participants with a positive RDT result were offered treatment with artemether-lumefantrine [29]. Blood samples were collected as dried blood spots on Whatman 3MM chromatography paper, and thin and thick smears were prepared and read according to WHO standards [30]. DNA was extracted from the dried blood spots per the Chelex procedure, and quantitative polymerase chain reaction (qPCR) was performed with SYBR Green PCR Master Mix (Applied Biosystems, Thermo Fisher Scientific Inc) and primers specific for P falciparum mitochondrial cytochrome b gene (cytb) [18].

IRS Intervention

From 4 to 9 April 2022, the US President’s Malaria Initiative VectorLink Project and the Ministry of Health conducted an IRS campaign with SumiShield in the 2 settlements in the intervention arm. Programmatic spray coverage was reported at the level of the zone, a subdivision of the health facility catchment area, and was calculated as the number of structures sprayed divided by the number of structures found. A study spray coverage survey was conducted in 150 households across the 2 settlements (30 study households and each of their 4 nearest neighbors), which verified spray status by examining a household spray card distributed by the program. Survey spray coverage was calculated as the number of households reported as sprayed divided by the number sampled in the survey.

Vector Collection

Indoor mosquito collections were conducted by miniature CDC light traps the night preceding the household survey. Detailed entomologic capture, processing, and laboratory methods are described elsewhere [31–33]. Captured mosquitoes were identified morphologically and verified by polymerase chain reaction (PCR). The infection status of the heads and thoraces of a subsample of anopheline mosquitoes was assayed by enzyme-linked immunosorbent assay to detect P falciparum circumsporozoite protein, and positive samples were confirmed by qPCR for P falciparum lactate dehydrogenase [34–36]. Sporozoite infection rate was calculated as the percentage of circumsporozoite protein–positive female anopheline mosquitoes of those that were tested. The human biting rate was estimated by dividing household vector counts by the number of people sleeping in the house. The entomologic inoculation rate (EIR) was estimated by multiplying the sporozoite infection rate by the mean human biting rate. WHO bottle assays (90-µg clothianidin/250-mL bottle) were conducted on A funestus s.l. caught in 25 households in the study area in May 2023 [37].

Outcomes

The primary outcome was monthly parasite prevalence as measured by RDT. Secondary outcomes were monthly incidence by microscopy and qPCR, as well as hazard of infection by microscopy. Cases were considered incident if the individual tested negative by the diagnostic test at the previous visit or was treated at the previous visit. qPCR was used to measure patent and subpatent infections. While incidence and hazard derive from the same diagnostic test (microscopy), incidence estimates the rate during a defined time interval (here, monthly) from dichotomous infection status measured at the end of that interval, whereby hazard estimates the instantaneous rate. Hazards are derived from time-to-event analyses that incorporate all information on visit timing, typically have better power, and can differentiate the timing of infections between groups that have similar cumulative risk over time [38]. The primary entomologic outcome was indoor vector counts. Secondary entomologic outcomes were relative species abundance, positive sporozoite rate, and EIR.

Sample Size and Power Calculations

The study was designed at 80% power to detect an effect size of 0.40 (as drawn from a meta-analysis of IRS impact) in the difference in the change in RDT prevalence before and after IRS between arms [39]. Power calculations were performed in Stata SE15 (StataCorp LLC) via a multilevel clustering approach and a household intracluster correlation coefficient of 0.10 [40]. The final sample size provided sufficient power to detect an effect size as small as 0.21.

Statistical Analysis

A difference-in-difference–like approach was used for all analyses: the within-arm (sprayed or not sprayed) relative change in the outcome from the pre- to postspray period was compared across arms (sprayed vs not sprayed). For the primary outcome, data were fit to a multilevel mixed-effects logistic regression model of the log odds of a participant being RDT positive. The postspray period was lagged to begin 14 days after the first day of IRS. Fixed effects included the settlement and month. Univariable analysis was performed to identify potential confounders, which were included as additional covariates [41]. Random intercepts were assigned to households and individuals. To account for additional correlation in individual risk of the outcome, the lagged outcome of the participant’s previous visit was included. For secondary outcomes, monthly incidence was modeled by multilevel mixed-effects Poisson regression, and time to first infection was modeled by marginal multifailure Cox proportional hazard regression [42]. For vector abundance, a multilevel mixed-effects negative binomial regression model was used with random intercepts on household.

Subgroup and Sensitivity Analysis

Four subgroup analyses were conducted to explore if vector control interventions in combination had a measurable impact. One analysis was restricted to individuals using LLINs and another to those living in households with adequate LLIN coverage, defined as 1 net per 2 persons. These analyses aimed to assess whether IRS impact was measurable in a population with theoretically less vector exposure. The third compared malaria risk in March 2022 across IRS status during the annual pre–rainy season IRS campaign in October 2021 (some households were missed) for evidence that the risk was lower in households sprayed in October. The fourth included all households and compared their combined IRS history (only April 2022, only October 2021, both, or neither) to explore evidence of an additive effect.

RESULTS

Study Population

In total, 382 participants from 64 households were enrolled. Each household had a median 6 occupants (IQR, 4–8), who on average were available and participated in 3.4 of 6 total possible visits. Missing visits were associated with older age and male sex but were not differential across arms. The median age was 14 years (IQR, 6–24), and most participants were female (58%; Table 1). Bed net usage differed by age group and was 93%, 63%, 50%, and 71% in those aged <1, ≤5, 6 to 16, and ≥17 years, respectively. Baseline vector counts and parasite prevalence by all measures were balanced across arms, although there were cluster-level differences (ie, not sprayed inland, sprayed inland, not sprayed lakeside, sprayed lakeside) in parasite prevalence by RDT and microscopy (analysis of variance test, P = .012 and P < .01; Supplementary Figure 1) and household vector counts (P < .01; Figure 2).

Table 1.

Baseline Characteristics of Individual Participants

	Overall (n = 382)	Not Sprayed Area (n = 199)	Sprayed Area (n = 183)	P Valuea
Individual characteristics
Positive
Microscopy	48	52	44	.10
RDT	28	29	26	.54
qPCR	49	51	47	.37
Age, y	18.14 (15.77)	18.05 (16.36)	18.23 (15.15)	.91
Sex: male	42	42	42	.98
Reside elsewhere for part of year	11	14	8	.06
Knows mosquito bites cause malaria	30	25	36	.02
Been to health facility in last month	18	14	21	.06
Employed	21	21	22	.93
Slept under bed net last night	61	63	58	.33
Spends time outdoors after sundown	47	50	43	.14
Household characteristicsb
Received IRS October 2022	57	55	61	.38
House has bed net	87	90	85	.14
Total nets owned by household	2.05 (1.16)	2.16 (1.23)	1.94 (1.08)	.07
Time owning net				.05
<1 mo	0	0	1
1 mo–1 y	90	94	85
2–5 y	10	6	14
No. of people in household	6.47 (3.20)	6.50 (2.49)	6.44 (3.83)	.87
Household head has at least secondary education	27	22	32	.03
House has open eaves	84	88	77	.01
House wall type				.50
Mud	18	17	19
Concrete	82	83	81
Wood	0	0	0
Other	0	0	0
House roof type is metal sheets	28	29	28	.99
House water source				<.01
Piped water	1	3	0
Well	63	84	44
Other	36	13	56
Household owns animals	63	55	71	<.01
Distance from
Health center	3.11 (2.16)	2.45 (1.74)	3.82 (2.36)	<.01
Road, km	0.15 (0.11)	0.12 (0.10)	0.18 (0.11)	<.01
Stream, km: categoryc
1	0.41 (0.27)	0.54 (0.28)	0.27 (0.18)	<.01
2	0.87 (0.40)	0.91 (0.35)	0.82 (0.43)	.02
3	1.47 (1.10)	1.89 (1.14)	1.01 (0.86)	<.01
4	2.50 (1.57)	1.32 (1.16)	3.78 (0.74)	<.01
5	9.36 (1.37)	8.14 (0.41)	10.69 (0.59)	<.01
6	20.05 (1.65)	21.33 (0.75)	18.65 (1.17)	<.01

All measures are provided as percentage of the total sample or mean (SD) across the sample. There were differences across arms at baseline in the percentage of adults with knowledge of causes of malaria, health center utilization in the preceding month, employment status, bed net usage and ownership, time spent outdoors at night, head of house education level, and house structure. Time-varying variables were measured at each study visit and treated as such in analysis.

Abbreviations: IRS, indoor residual spray; qPCR, quantitative polymerase chain reaction; RDT, rapid diagnostic test.

^a t test and chi-square tests used for continuous and categorical variables, respectively.

^bHousehold characteristics apply to the household in which the individual resides.

^cStreams are classified according to the Strahler stream classification system, where category 1 streams are the smallest and category 6 the largest.

Figure 2.

Anopheline outcomes by month show cluster-level differences at baseline and throughout the study period: A, vector abundance; B, entomologic inoculation rate (EIR). C, Anopheles funestus was the most abundant species throughout the study period in sprayed and unsprayed areas, and the percentage of vectors that were A funestus increased from 81% in March to 99% by July and August. EIR differed by cluster and over time, but the study was not powered to detect changes in EIR by site or over time.

Spray Coverage

Programmatically reported spray coverage was 99% (2430 of 2461 found households sprayed) [43]. However, spray coverage estimated from the survey was only 75% (n = 70) and 86% (n = 80) in the lakeside and inland sprayed settlements, respectively. Of the 50 households enrolled in March, 49 verified their October 2021 IRS campaign spray status. Only 34% (8/26) of intervention households and 36% (9/23) of control households reported having received IRS in October 2021 when the annual IRS campaign was conducted. Residual efficacy of SumiShield was observed to last up to 10 months [44].

Epidemiologic Results

Monthly prevalence by each outcome is shown in Figure 3. From March to August 2022, the RDT period prevalence was 71.7%, and the incidence was 3372 and 5422 infections per 1000 person-years by microscopy and qPCR, respectively. Individuals ever positive (ie, those who tested positive at least once) by RDT, microscopy, or qPCR were positive by the corresponding diagnostic test more than once across the study period—on average, 2.4, 2.0, and 2.3 times. If individuals tested positive by RDT at the previous visit, their odds of testing positive at the subsequent visit were 2.4 times higher than if they had tested negative at the previous visit (univariable odds ratio, 3.4; 95% CI, 2.4–4.7; Supplementary Table 2a). Individuals who were high risk and frequently positive (defined as those who had at least 3 visits and at least 2 positive microscopy results) were younger, had less knowledge of the cause of malaria, and had lower health care utilization than those who were rarely or less frequently positive (defined as those testing positive never or only once). Correct knowledge of at least 1 symptom of malaria and the cause of malaria were univariably associated with lower prevalence and incidence by RDT and microscopy, as were female sex, older age, farther distance from health center, closer distance to streams, metal household roof type, and closed household eaves (Supplementary Table 2a and 2b).

Figure 3.

Unadjusted epidemiologic outcomes by month, including prevalence as measured by RDT, microscopy, and qPCR across sprayed and unsprayed areas. Abbreviations: qPCR, quantitative polymerase chain reaction; RDT, rapid diagnostic test.

Residing in a household in a sprayed area was associated with a 52% reduction in Cox-estimated hazard over the study period (hazard ratio [HR], 0.48; 95% CI, .29–.78; Table 2), and the proportional hazard assumption was met. The risk trajectory over the study period differed by baseline microscopy status (Figure 4): the hazard was 3.1 times higher in individuals positive at baseline than negative at baseline (HR, 4.11; 95% CI, 3.11–5.42; Supplementary Table 3). After the analysis was stratified by baseline status, participants who were parasitemic at baseline had a 43% reduction in hazard over the study period when residing in a household in a sprayed area (HR, 0.57; 95% CI, .35–.94; supplementary material). Individuals who were not parasitemic at baseline had too few infections during the study period to estimate the hazard of infection. Residing in the intervention arm postspray was not associated with a significant change in parasite prevalence or incidence, although the adjusted post- to prespray period reduction in microscopy incidence rate in the sprayed area was 69% greater than the reduction in the unsprayed area (ratio of incidence rate ratio, 0.31; 95% CI, .05–1.93; Table 2, Figure 5).

Table 2.

Measures of the Effect of Post–rainy season IRS on the RDT Positivity Odds Ratio, Microscopy and qPCR Incidence, Hazard Rate, and Anopheline Counts

	Unadjusted Estimate	95% CI	P Value	Adjusted Estimate	95% CI	P Value
RDT	1.31	.67–2.55	.31	0.98	.14–6.95	.99
Microscopy incidence	1.04	.60–1.78	.97	0.31	.05–1.93	.21
qPCR incidence	1.19	.79–1.78	.41	1.22	.76–1.95	.41
Hazard	0.74	.47–1.18	.21	0.48	.29–.78	<.01
Anopheline count	1.79	.75–4.27	.19	2.30	.92–5.72	.07

The estimated effect is the covariate of interest in each regression (ie, the interaction between the study arm and the time of sampling) and is interpreted as the ratio of change in the sprayed arm divided by the ratio of change in the unsprayed arm. The full regression model results are in Supplementary Tables 2a–2c and 3.

Abbreviations: IRS, indoor residual spray; qPCR, quantitative polymerase chain reaction; RDT, rapid diagnostic test.

Figure 4.

Nelson-Aalen unadjusted cumulative expected number of Plasmodium falciparum infections differed when stratified by baseline microscopy status and spray status. At the time of indoor residual spraying (IRS), individuals who were positive by microscopy had 1 more expected infection than those who were negative at baseline by definition. These curves diverged such that those positive at baseline accumulated an average of 3 to 4 infections by day 200, while those negative at baseline accumulated an average of <1 infection. Those positive at baseline who lived in sprayed areas had fewer infections after IRS than those in unsprayed areas. The Cox regression (Supplementary Table 3), which adjusts for confounders, confirmed that their risk trajectories differed in the post-IRS period. No difference in expected infections by spray status was seen in those negative at baseline.

Figure 5.

Ratio between the pre- to posttest period change in sprayed vs unsprayed areas from multivariate mixed effects models (all outcomes with the exception of hazard) and Cox regression models (hazard). The ratios for outcomes are presented: entomologic (gray) and epidemiologic (black). Vertical lines indicate 95% CI. Absolute parasite prevalence, incidence, and hazard increased over the study period, and a ratio of 1 indicates that this increase was the same across sprayed and unsprayed areas. A ratio <1 suggests that there was less of an increase in sprayed areas than there was in unsprayed areas, consistent with a protective effect of indoor residual spraying. Abbreviations: qPCR, quantitative polymerase chain reaction; RDT, rapid diagnostic test.

Subgroup analyses showed no substantial differences in parasite prevalence by RDT, microscopy, or qPCR in the control vs intervention arms when restricting the analysis to participants who reported LLIN use (Supplementary Table 5a and 5b). Subgroup analyses of baseline prevalence by prior IRS in October 2021 did not provide evidence of any lasting effect of the October IRS campaign. There were no substantial differences in the pre- or postperiod changes in the odds of parasite prevalence when households that received 0, 1, or 2 rounds of IRS were compared (Supplementary Table 6b).

Entomologic Results

Monthly household vector counts ranged from 538 to 1741 female anophelines, with variation in abundance by cluster. All molecularly speciated A funestus complexes were A funestus s.s., and all A gambiae complexes were A gambiae s.s. The EIR was highest in March, and species-level annual EIR estimates from CDC light traps for A funestus and A gambiae were 16.7 and 17.2, respectively, for an overall EIR of 33.9 infectious bites per person per year (Figure 3, Supplementary Table 1a). In bottle assays, there was 95% mortality after exposure to clothianidin in 91 A funestus mosquitoes, and further resistance testing showed complete susceptibility to clothianidin (www.pmi.gov/resources) [43, 44]. Vector counts increased in the postspray period, but this increase was not significantly different in households in sprayed areas as compared with households in unsprayed areas (incidence rate ratio, 2.30; 95% CI, .92–5.72; P = .07).

DISCUSSION

In a setting with holoendemic malaria in northern Zambia, IRS at the end of the rainy season was associated with a reduction in malaria transmission, as measured by the hazard of infection, but not with a reduction in traditional metrics, including parasite prevalence or incidence. Individuals positive by microscopy at baseline remained at higher risk of infection throughout the study period, and baseline microscopy status modified the protective effect of IRS. There was no evidence of a residual effect of IRS from the 2021 pre–rainy season spray campaign or any combined effect in those households that received IRS in both campaigns. There was also no observed reduction in indoor anopheline abundance, refuting the hypothesis that IRS at the end of the rainy season would affect peak A funestus abundance indoors. This could reflect a limited efficacy of clothianidin, as seen in other IRS implementation settings, but given the history of ineffective IRS in Nchelenge District and the demonstrated susceptibility of local vectors to clothianidin, it more likely resultant of the highly abundant vector population [45]. While outdoor biting was not measured in this study, it offers a possible explanation for the absence of a dramatic impact of IRS: about half of this study population spends time outdoors where infectious vectors have been documented [46]. Notwithstanding, the reduction in hazard of infection suggests that IRS with clothianidin reduced individual exposure to infected vectors, which could be mediated by a change in indoor biting behavior.

The complexities of measuring malaria transmission in a holoendemic setting provide further context for the observed decrease in the hazard of infection in the absence of detectable changes in prevalence, incidence, or vector abundance. The most direct measure of malaria transmission is EIR, which estimates mosquito-to-human transmission events [47]. The EIR in this study, as estimated from CDC light trap catches, was 34 infectious bites per year—likely an underestimate relative to EIR estimates based on human landing catches [48]. EIR is also difficult to measure due to the relative rarity of capturing infectious mosquitoes [49]. While incidence is more practical to measure than EIR and more useful than prevalence given the potential for parasite clearance and reinfection, it loses that advantage when measured monthly in a high-transmission setting, where it can be invariable over a wide range of EIRs [47, 50]. The hazard of infection, when data allow for this calculation, does not calculate risk within fixed time intervals, which lends greater power for detecting differences in statistical comparisons when more frequent sampling (ie, weekly) is not possible.

Coverage with vector control interventions in Nchelenge District could be optimized to increase impact. Programmatic IRS coverage was overestimated by 20% to 24%, and a recent analysis of IRS in Nchelenge District showed that each 10–percentage point increase in district spray coverage was associated with a 5% reduction in parasite prevalence [19]. Furthermore, only 60% of study participants reported LLIN use. Other studies of IRS in holoendemic settings had higher net coverage due to local policies for combined LLIN-IRS mass deployment [2, 7]. At the time of this study, the Zambia National Malaria Elimination Centre deployed a “mosaic” strategy to universal vector control, targeting areas with either the IRS or LLINs [24].

The lack of measured population-level effect of IRS could also be due to limitations in study design. The study was not powered to detect subtle changes in prevalence, as only a large effect would motivate a policy change in IRS timing. Importantly, this study was not designed as a noninferiority trial, and the lack of measurable effect is not evidence of no effect. Another limitation is that IRS may have occurred too late to suppress the May peak in infections, although this seems unlikely, as infective bites for early May cases likely occurred in mid-April, after the conclusion of IRS. Overall, this study's design is a strength as it is one of the few observational studies of IRS that conducted longitudinal active surveillance in a cohort across intervention and control areas. The baseline preintervention measurements and the geographically matched control groups allowed the statistical models to isolate the associated effects of the intervention from settlement-level, time-invariant factors and reduce confounding by time-varying factors, which may be ecologic or seasonal. Finally, the multiple outcomes included in the design provide a deeper understanding of the impact of IRS in a holoendemic setting. IRS was associated with a reduction in the hazard of infection but did not have a large public health impact.

CONCLUSION

This study aimed to determine if timing IRS to peak vector abundance in the dry season would have a measurable impact on malaria transmission in an area where pre–rainy season IRS has historically had little or no impact. There were no qualitatively large reductions in parasite prevalence or incidence to support a recommendation to change the timing of IRS. However, the hazard of infection was halved in sprayed areas, providing evidence of a quantifiable impact of IRS on malaria transmission in an intractable holoendemic setting, even in the absence of impact on parasite prevalence. Two practical implications for impact assessments in high-transmission settings follow. First, study designs should include longitudinal measures to estimate the hazard of infection. Second, where malaria burden is intransigent, studies should focus on evaluating combinations of interventions, with or without new interventions such as malaria vaccines. Together these will provide critical input for planners to model the optimal choice and combination of interventions for cost-effective, sustainable malaria control.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. The authors gratefully acknowledge the Southern Africa ICEMR field teams in Nchelenge District, the VectorLink teams conducting and executing the spray campaign, the National Malaria Elimination Centre for its support, and all study participants. The authors also acknowledge Nelius Betner, who conducted the field collections and bottle assays for clothianidin resistance testing in 2023. The Southern and Central Africa ICEMR (International Centers of Excellence for Malaria Research) is a research center that studies barriers to malaria control and elimination and informs national strategies. It has been funded by the National Institute of Allergy and Infectious Diseases since 2010.

Author contributions. A. C. M. oversaw data collection and management, conducted all analysis, and drafted the manuscript. M. C., M. M., and J. L. coordinated and supervised enrolment and data collection. M. E. G. organized and supervised entomological laboratory work. S. B. performed initial sample size calculations. T. K. supervised molecular epidemiological work. B. c., P. P., and R. Z. ensured execution of the spray campaign. T. S., A. W., D. E. N., D. E. I., M. I., and W. J. M. participated in study conceptualization. All authors reviewed and contributed to the final manuscript.

Data availability. The data underlying this article cannot be shared publicly due to regulations of the National Health Research Act, the Government of Zambia. Data sets, codebooks, and analytic code supporting the conclusions of this article may be shared on reasonable request. Interested investigators are required to submit a written request to the Ministry of Health. Contact Dr William J. Moss (wmoss1@jhu.edu) to coordinate the request. Analytic code for the analysis can be found here: https://github.com/anniecmartin/IRSlaterainyZambia.

Disclaimer. The findings and conclusions in this article are those of the authors and do not necessarily represent the official views of the National Institutes of Health, the US Centers for Disease Control and Prevention, the US Agency for International Development, or the US President's Malaria Initiative. Use of trade names is for identification only and does not imply endorsement by the National Institutes of Health, US Centers for Disease Control and Prevention, US Agency for International Development, US President's Malaria Initiative, or US Department of Health and Human Services.

Ethics. This study was a part the study “Malaria Transmission and the Impact of Control Efforts in Southern Africa” (institutional review board approval 3467), which obtained ethical approval from the Johns Hopkins Bloomberg School of Public Health Institutional Review Board and the Tropical Diseases Research Centre Research Ethics Committee.

Financial support. This work was supported by funds from the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (grants U19AI089680 to W. J. M. and T32AI138953-03 to A. C. M. and M. E. G.); and by the President’s Malaria Initiative (contracts 200AA23C00012 and AID-OAA-17-00008).