Skip to main content
NCBI home page
PLOS Global Public Health logoLink to PLOS Global Public Health
. 2025 Dec 11;5(12):e0004436. doi: 10.1371/journal.pgph.0004436

Disentangling the roles of different vector species during a malaria resurgence in Eastern Uganda

Max McClure 1,*, Ambrose Oruni 2, Emmanuel Arinaitwe 3, Alex Musiime 4, Patrick Kyagamba 3, Geoffrey Otto 3, James Adiga 3, Jackson Asiimwe Rwatooro 3, Maxwell Kilama 3, Paul Krezanoski 1, Jessica Briggs 1, Philip J Rosenthal 1, Joaniter I Nankabirwa 3,5, Moses R Kamya 3,5, Grant Dorsey 1, Bryan Greenhouse 1, Isabel Rodriguez-Barraquer 1,6
Editor: Amy Kristine Bei7
PMCID: PMC12697997  PMID: 41379838

Abstract

In 2021–23, a resurgence of malaria occurred in the Tororo District of Uganda following a change in formulations used for indoor residual spraying of insecticide (IRS). Prior analyses showed that this increase was temporally associated with the replacement of Anopheles gambiae sensu lato by An. funestus as the dominant local vector. To investigate this association, we used data from a cohort of 422 children in 94 households from 2021-2023 in Tororo District and neighboring Busia District, where IRS was not implemented. Participants underwent passive and monthly active surveillance for infection with Plasmodium falciparum by quantitative PCR. Mosquitoes were collected in each sleeping room in cohort households every 2 weeks using CDC light traps. We assessed the association between spatiotemporally smoothed annualized household entomologic inoculation rates (aEIR) and individual P. falciparum infections using shared frailty models. Overall, each doubling of the aEIR was associated with a 29% increase in the hazard of P. falciparum (adjusted HR 1.29, 95% CI 1.25-1.33). Considering species-specific aEIRs, this effect was primarily driven by An. funestus: a doubling of An. funestus aEIR was associated with a 29% increase in hazard rate (1.29, 1.25-1.34), while the association was smaller for An. gambiae (1.04, 1.01-1.08). These relationships were stronger in Tororo than in Busia. These results support the inference that the replacement of An. gambiae with An. funestus was a driver of increased malaria incidence in Tororo District and demonstrates associations between household-level entomological data and risk of P. falciparum infection.

Introduction

In 2021–23, following a five-year decline, malaria incidence notably resurged in the Tororo District of Uganda, increasing from 0.36 to 2.97 episodes per year in one region. This trend coincided with a transition from organophosphate to clothianidin-based indoor residual spraying (IRS) insecticides in March 2020, accompanied by replacement of Anopheles gambiae sensu lato (s.l.) by An. funestus, a species found to be relatively tolerant to clothianidin, as the principal local malaria vector. Neighboring Busia District, which never implemented IRS, saw no corresponding change in vector species composition or malaria incidence. Based on these findings, we previously suggested that a rising An. funestus population led to the increased malaria burden observed in Tororo [1].

It is well established that the success of vector control measures is dependent on the local vector composition: beyond adulticide selection, the effectiveness of control measures including housing modification [2,3], zooprophylaxis [4], and larval source management [5] have been dependent on species-specific vector ecology. This explanation squares with growing recognition of the epidemiological importance of An. funestus in eastern and southern Africa [6], but it is based only on temporal correlation. It has been difficult to causally tie mosquito exposures to clinical outcomes or implicate specific vector species in outbreaks [7], but doing so is critical for assessing the value of targeted vector control.

To further characterize the relationship between vector composition and the recent resurgence of malaria in Tororo, we analyzed data from a detailed cohort study to investigate the association between household-level annualized entomological inoculation rate (aEIR) and time to P. falciparum infection, considering aEIR for each relevant malaria vector.

Methods

Study participation

The PRISM border cohort study was conducted from August 2020 to September 2023 in Tororo and Busia, both of which are historically high malaria transmission districts in eastern Uganda. The region has historically experienced two seasonal peaks in malaria transmission following the rainy seasons, with precipitation peaking in April and November (based on the 2000–2018 mean) [8].

Data was collected in three geographic zones across the two districts: northern Busia, where IRS has never been implemented; “Tororo, near border,” a zone located 0.7-3.5km from the border between Tororo and Busia; and “Tororo, away from border,” farther north (5.5-10.8km from the border) in Tororo (Fig 1).

Fig 1. Household location by zone.

Fig 1

Open in a new tab

Tororo District was subject to multiple vector control interventions since 2013, with concomitant declines in malaria transmission until the 2020 resurgence. Pyrethroid long-lasting insecticidal nets (LLINs) were distributed in 2013, 2017 and June 2020, and a mix of LLINs treated with deltamethrin (Yorkool) and deltamethrin plus piperonyl butoxide (PermaNet 3.0) in August 2023. IRS was conducted with the carbamate bendiocarb from 2014 to 2015, the organophosphate Actellic (pirimiphos-methyl) from 2017 to 2019, and with Fludora Fusion (clothianidin and deltamethrin) starting in 2020. During the study period, there were three additional rounds of IRS in Tororo: Fludora Fusion in March 2021, SumiShield (clothianidin) in March 2022, and a return to Actellic in March 2023, which was followed by a decline in malaria transmission [1]. In Busia, pyrethroid LLINs were distributed in 2013 and 2017, LLINs treated with deltamethrin plus piperonyl butoxide (PermaNet 3.0) in December 2020, and LLINs treated with alphacypermethrin and clorfenapyr (Interceptor G2) in November 2023, but no IRS was performed (S1 Fig). In the adjoining Kenyan county of Busia, which we did not survey, modeled annual incidence rates from the Malaria Atlas Project suggest that there was no similar rise in malaria incidence during the period covered in this study [9]. The President’s Malaria Initiative funded mass distribution of pyrethroid-piperonyl-butoxide ITNs in the county in 2021, but did not implement IRS during the study period [10].

Household and participant enrollment is described in detail elsewhere [1]. Briefly, households containing at least two children ages 5 or younger were randomly selected from within the study areas. Household locations were mapped using handheld global positioning system (GPS) units. Enrollment was dynamic over the course of the study – children from participating households joined the study when they were born or their households were enrolled.

Data collection

Routine clinical visits were conducted every 4 weeks. In addition, participants were encouraged to present to the study clinic (open 7 days a week) for evaluation of any illnesses. All routine visits and unscheduled visits when participants were febrile or reported fever in the last 24 hours included collection of thick blood smears and testing of DNA extracted from whole blood for P. falciparum DNA using highly sensitive quantitative PCR (qPCR) targeting the var gene acidic terminal sequence, with a limit of detection of 0.03 to 0.15 parasites/μl [11].

Participants with objective or reported fever who tested positive on the thick blood smear were treated for malaria according to national guidelines.

Mosquito collections were conducted from 7pm to 7am every 2 weeks in all study participants’ bedrooms using CDC light traps. All mosquitoes underwent morphologic species identification. Up to 50 mosquitoes per CDC light trap collection were assessed for sporozoites using ELISA [12].

Data analysis

The objective of this analysis was to characterize associations between species-level entomological surveillance data and time from treatment or natural clearance of infection to incident P. falciparum infection. First, we developed spatiotemporal models of entomological data to obtain smoothed estimates of species-specific household aEIR, a standard measure of malaria exposure. The smoothing process was intended to minimize the effects of measurement error, zero inflation, and transient changes in entomological measures with little relevance to human exposure. [13] We then used the predictions of these models to assess the relationship between species-specific aEIR and time to infection.

EIR modeling

Mosquito counts for the region’s two major vectors, the An. gambiae complex and An. funestus subgroup, were modeled as negative binomial temporal generalized additive models (GAMs) with the mgcv package [14] in R Statistical Software (v4.4.1; R Core Team 2024), with date of collection as a smoothed predictor using thin plate splines and including household as a random effect.

Sporozoite rates within the subset of each morphologic species that underwent ELISA were modeled as temporal binomial GAMs, again using thin plate splines and including household as a random effect. For both vector counts and sporozoite rates, separate models were formulated for each geographic zone (i.e., three models) due to their distance and differing vector control intervention histories. No spatial component was included within zones due to the small number of households and the relatively small distances between households within each site; models including household random effects performed better than those with an independent spatial smooth by Akaike information criterion (AIC) for all zones and models (Tables A-D in S1 Text). Additional model types were considered as detailed in the supplementary materials and compared by Akaike information criterion (AIC). Residuals for each model were assessed for deviation from the expected distribution with the DHARMa package [15] in R. After model fitting, the estimated EIR for each species was calculated as the product of the modeled daily vector count and sporozoite rate and then annualized.

Survival analysis

To investigate the association between entomological metrics and time to incident P. falciparum infection, we fit shared frailty models clustered at the level of the individual. Incident P. falciparum infection was defined as the first occurrence of either positive qPCR or positive microscopy, regardless of the presence of symptoms, following either a prior documented malaria treatment course or a spontaneous clearance (i.e., asymptomatic individuals who clear parasitemia without treatment). Spontaneous clearances were defined as three negative qPCR results in succession (covering approximately 3 months, based on sampling every 28 days as previously prescribed [16]). An alternate, more conservative analysis in which we only considered incident infections occurring after documented treatment (intended to address the possibility that persistent low-density parasitemia temporarily falling below the threshold of detection by qPCR could be falsely labeled as a clearance and subsequent reinfection) was also conducted and is reported as a supplementary result. Instances of low parasite density infections detected only by qPCR within 30 days of a prior treated infection were not considered incident infections as these could reflect residual gametocytemia following treatment.

Recurrent events were handled with a gap time approach, assuming independence between all observed event times: day 0 for each period of interest was defined as day 15 following the patient’s most recent malaria treatment or day 1 after natural clearance. To deal with interval-censoring of the outcome times, the time of infection was imputed to the midpoint between the most recent negative clinical evaluation and the day of diagnosis.

To minimize the impact of variations in host susceptibility on infection risk, we limited this analysis to children up to 15 years of age. Participants who were positive at all visits, with no documented treatment for malaria or spontaneous clearance, as well as participants followed for fewer than three months were excluded from the analyses.

Models were fit using the frailtyEM package [17] in R. We conducted analyses both for all sites combined and separately by district (i.e., two models, combining the two Tororo zones due to their shared vector control intervention histories). Daily log2-transformed species-specific modeled aEIRs were included as time-dependent covariates, lagged by 14 days to account for the P. falciparum intrinsic incubation period; 28-day lags for entomological covariates were also tested and are reported as supplementary results. Models included fixed effects for each of three previously identified time periods representing distinct epidemiological moments: before, during and after malaria resurgence (September 2020-August 2021, September 2021-March 2023, and April-September 2023, respectively) [1], as well as age. We also fit models that considered interactions between aEIRs and the epidemiological time periods. Reported incidence rate ratios are adjusted for all listed covariates unless otherwise specified.

Ethics statement

Ethical approval was obtained from the Makerere University School of Medicine Research and Ethics Committee (REF 2019–134), the Uganda National Council for Science and Technology (HS 2700), and the University of California, San Francisco Committee on Human Research (257790). Written informed consent was obtained for all participants prior to enrollment into the cohort study. Written informed consent was obtained from the parent/guardian of each participant under 18 years of age. Participants were recruited from 10-08-2020 to 21-02-2023.

Results

Study population and entomological data

The study enrolled 472 participants under age 15 from 94 households, of which 45 were excluded from analysis because of an absence of treatment episodes or natural clearances, and an additional 5 were excluded due to less than 90 days of follow-up. The final dataset from this analysis included 422 participants from 94 households, 49% of whom were male (Table 1). The median age at enrollment was 4.6 years (IQR 2.5-9.0). Participants were followed for a median of 977 days (IQR 565–1125).

Table 1. Participant characteristics.

Busia Tororo, near Tororo, away Overall
No. individuals 111 167 144 422
No. households 23 40 31 94
% Male 49.5 52.1 45.8 49.3
Median age in years (IQR) 4.8 (2.4, 8.9) 4.6 (2.5, 8.7) 4.4 (2.2, 9.7) 4.6 (2.5, 9)
Median days followed (IQR) 787 (561, 1139) 1124 (438, 1126) 977 (614, 979) 977 (565, 1125)
Median infections per person-year (IQR) 4.6 (2.5, 8.2) 5.7 (3.3, 9.5) 4.9 (3.4, 7.7) 5.1 (3, 8.6)
Median cases malaria per person-year (IQR) 1.4 (0.7, 2.8) 2.1 (1, 3.3) 2 (1.2, 3.4) 2 (1, 3.3)
Open in a new tab

For the purposes of calculating infection incidence, person-time here omits all days between an incident infection and day 14 following the individual’s next malaria treatment (see Methods).

Changes in entomological metrics over the study period

In order to quantify entomological trends associated with the resurgence of malaria in Tororo following the switch in IRS from Actellic to Fludora Fusion and SumiShield in 2022–2023, we compared district-wide household-level collections from before, during and after the resurgence by dividing collections into previously defined time periods (Table 2; see Methods). In Tororo near the border, median An. funestus aEIR rose from a baseline of 0 (IQR 0-7.3) to 8.9 (0-20.8) during the resurgence and returned to 0 (0–0) afterwards. Median nightly An. funestus counts remained stable during the resurgence (1.4 [0.9-2.3] before to 1.2 [0.6-2.2] during), while SR rose (0 [0–1] before to 1.7 [0-3.1] during). An. funestus aEIR. An. gambiae aEIR fell from 14 (0-27.5) to 7.8 (0-11.1) to 0 (0–0) over the same time periods.

Table 2. Household-level medians (IQR) for biweekly entomological measures.

Busia Tororo, near Tororo, away
Before During After Overall Before During After Overall Before During After Overall
An. gambiae count 8.4 (4.6,20.8) 1.6 (0.8,4.1) 2.5 (1.3,4.3) 2.5 (2,7.3) 7.2 (3.8,11.8) 0.9 (0.4,1.8) 0.9 (0.5,2) 3.5 (1.9,9.1) 1 (0.6,2.3) 0.4 (0.2,0.8) 0.3 (0.1,0.5) 0.6 (0.3,1)
An. funestus count 0.9 (0.4,1.8) 0.4 (0.2,0.9) 1.4 (0.6,3.3) 0.8 (0.3,1.8) 1.4 (0.9,2.3) 1.2 (0.6,2.2) 0.8 (0.4,2.2) 1.4 (0.8,2.7) 0.1 (0.1,0.2) 0.6 (0.2,1.1) 0 (0,0.1) 0.4 (0.1,0.7)
An. gambiae SR 0.8 (0.5,1.1) 1.3 (0,2.3) 0 (0,1.9) 1.2 (0.7,1.8) 0.5 (0,1) 1.5 (0,2.9) 0 (0,0) 0.7 (0.4,1.2) 0 (0,0.4) 0 (0,1.6) 0 (0,0) 0 (0,1.8)
An. funestus SR 1.3 (0,2.8) 0 (0,2.9) 0 (0,2.5) 1.8 (0,2.7) 0 (0,1) 1.7 (0,3.1) 0 (0,0) 0.7 (0,2) 0 (0,0) 1.5 (0,2.5) 0 (0,0) 1.8 (0,2.6)
An. gambiae aEIR 33.5 (13.2,64.4) 7.9 (0,13) 0 (0,32.4) 16.4 (5.2,40.5) 14 (0,27.5) 7.8 (0,11.1) 0 (0,0) 9.3 (3.7,19.8) 0 (0,5.7) 0 (0,4.2) 0 (0,0) 0 (0,2.7)
An. funestus aEIR 8.1 (0,14.5) 0 (0,7.7) 0 (0,28.1) 6 (0,13.8) 0 (0,7.3) 8.9 (0,20.8) 0 (0,0) 4.6 (0,12.3) 0 (0,0) 4.5 (0,11.2) 0 (0,0) 2.6 (0,7.1)
No. collections 34 (26,49) 78 (41,82) 26 (13,26) 122 (77,154) 50 (26,51) 82 (54,84) 26 (13,26) 142 (64,160) 32 (24,32) 82 (58,84) 26 (25,26) 140 (101,142)
Open in a new tab

For the two Tororo collection zones, measures are divided into periods of time “before,” “during,” and “after” the malaria resurgence described in the text. Species-specific SR reported as percentage. SR = sporozoite rate; aEIR = annualized entomological inoculation rate; No. collections = number of household-nights of mosquito collection.

In Tororo away from the border, median An. funestus aEIR rose from 0 (0–0) to 4.5 (0-11.2) during the resurgence and returned to 0 (0–0) afterwards. Median nightly An. funestus counts rose during the resurgence (0.1 [0.1-0.2] before to 0.6 [0.2-1.1] during), as did SR (0 [0–0] before to 1.5 [0-2.5] during).

An. gambiae aEIR remained similar over the three time periods (0 [0-5.7]; 0 [0-4.2]; 0 [0–0], respectively). In Busia, which did not experience a resurgence, An. gambiae remained dominant over the study period, with a median aEIR of 16.4 (5.2-40.5), compared to an An. funestus median aEIR of 6 (0-13.8).

Entomological model comparisons are summarized in Tables A-D in S1 Text. Predictions generated from the best vector density and sporozoite rate models are overlaid on Fig 2a–f. Consistent with descriptive analyses, this model suggests that relative contributions of An. funestus to vector composition increased during the resurgence in both Tororo zones, and sporozoite rates generally remained equal or higher in An. funestus when compared to those in An. gambiae throughout the study period in Tororo.

Fig 2. Temporal trends in entomological measures and infection incidence.

Fig 2

Open in a new tab

Species-specific vector counts (a-c), species-specific sporozoite rates expressed as a percentage (d-f), species-specific annualized entomological inoculation rate (aEIR) (g-i) and P. falciparum infection incidence (j-l) by zone (Busia: a, d, g, j; Tororo, near border: b, e, h, k; Tororo, away from border: c, f, i, l). Points with whiskers represent monthly mean crude data and associated uncertainty while lines with confidence bands represent mean model output and associated uncertainty. For vector counts, whiskers show 95% confidence intervals for counts modeled as a Poisson process. For sporozoite rate and malaria incidence, whiskers show 95% confidence intervals from the exact binomial test. For aEIR, whiskers represent 2.5% and 97.5% quantiles of collection-level aEIRs. For GAM output, 95% confidence bands for temporal smooths are shown, excluding household random effects. Note that plots h and i have y axes at a different scale than plot g to emphasize differences in species-specific aEIRs. The timing of indoor residual spraying campaigns is indicated with vertical lines (FF = Fludora Fusion, SS = SumiShield). The period of resurgence described in the main text is indicated with gray shading. For the purposes of calculating infection incidence, person-time here omits all days between an incident infection and day 14 following the individual’s next malaria treatment (see Methods).

Changes in risk of infection over the study period

Median incidence of P. falciparum infection was highest in Tororo near the border at 5.7 infections per person-year (IQR 3.3-9.5), followed by Tororo away from the border at 4.9 (3.4-7.7) and Busia at 4.6 (2.5-8.2). Consistent with this result, the median time to incident infection was longer in Busia (29 days [95% CI 22–42]) than in the combined Tororo zones (19 days [18–21]) (S2 Fig). Between the start of enrollment at Tororo away from the border and start of the malaria resurgence as defined above (January-August 2021), the percentage of individuals who were positive for P. falciparum by PCR at the first evaluation falling within the analysis period was 42.1% (32/76) in Busia, 53.9% (69/128) in Tororo near the border, and 12.8% (15/117) in Tororo away from the border.

As with entomological measures, infection incidence increased during the resurgence in both Tororo zones. In Tororo near the border, incidence increased from a median of 3.1 (IQR 1.1-5.8) infections per person-year to 11.8 (5.3-18.7) and then declined to 7 (2.8-20.3). Away from the border, incidence increased from a median of 1.6 (0.0-1.9) infections per person-year to 9.8 (6.2-15.6) and then declined to 2.5 (0-6.6). Combining both Tororo zones, which experienced the same IRS regimes, incidence increased from 1.9 (0.0-4.1) infections per person-year to 10.8 (5.7-17.8) before declining to 4.0 (0.0-12.0). 26.3% (163/620) of incident infections in our analysis were accompanied by fever in Busia, 28.8% (290/1006) in Tororo near the border, and 38.8% (357/921) in Tororo away from the border.

Association between entomological metrics and time to incident infection

We fit time to event models to assess the association between entomological metrics and time to incident infection. In survival models adjusting for epidemiologic time period (before, during and after the malaria resurgence), we found a positive association between overall aEIR and P. falciparum infection when combining both districts, with a 29% increase in the hazard of infection for each doubling of aEIR (mean adjusted hazard ratio [aHR] 1.29 [95% CI 1.25-1.33], Table 3). When considering species-specific aEIRs, we found that most of this effect was driven by An. funestus: a doubling of An. funestus aEIR was also associated with a 29% increase in infection rate (1.29 [1.25-1.34]), while the association was smaller for An. gambiae (1.04 [1.01-1.08]). Fitting models to districts separately, we observed a stronger relationship between An. funestus and infection rate across both Tororo zones (1.27 [1.22-1.32]) than in Busia (1.17 [1.07-1.28]). An. gambiae was significantly associated with infection risk when districts were considered individually, though to a lesser extent than An. funestus (Tororo: 1.07 [1.02-1.12]; Busia: 1.11 [1.04-1.18]).

Table 3. aHRs for primary frailty model.

Busia Tororo Overall
Total aEIR
 Total aEIR 1.23 (1.17,1.30) 1.33 (1.270,1.38) 1.29 (1.25,1.33)
 Age (years) 1.05 (1.02,1.09) 1.02 (0.998,1.03) 1.02 (1.01,1.04)
 During 1.81 (1.43,2.29) 4.62 (4.010,5.32) 3.71 (3.30,4.19)
 After 1.69 (1.26,2.27) 2.66 (2.140,3.30) 2.31 (1.95,2.74)
Sp.-specific aEIRs
 An. funestus aEIR 1.17 (1.07,1.28) 1.27 (1.220,1.32) 1.29 (1.25,1.34)
 An. gambiae aEIR 1.11 (1.04,1.18) 1.07 (1.020,1.12) 1.04 (1.01,1.08)
 Age (years) 1.05 (1.02,1.09) 1.01 (0.996,1.03) 1.02 (1.00,1.04)
 During 1.70 (1.35,2.15) 3.72 (3.190,4.33) 2.97 (2.62,3.37)
 After 1.57 (1.13,2.17) 2.18 (1.760,2.71) 1.87 (1.57,2.22)
Open in a new tab

aHRs assuming a linear relationship between all covariates and the log hazard, using models fit to expected EIR for either Busia, both Tororo zones, or all zones combined. “Total aEIR” denotes models with summed An. gambiae and An. funestus aEIRs. “Sp.-specific aEIRs” denotes models treating species-specific aEIRs as independent predictors. All EIRs are log2-transformed.

When we considered interactions between time periods, the findings for An. funestus were present over the entire study. In Tororo, An. funestus was significantly associated with infection risk before the resurgence (aHR 1.14), and the association was even stronger during the resurgence (1.30), while the association between infection risk and An. gambiae decreased (1.12 before; 1.02 during) (S2 Table). The findings for An. funestus were robust when using restricted cubic splines instead of assuming a linear relationship between species-specific aEIR and log hazard. Results were also qualitatively similar when excluding natural clearances, using 28-day-lagged aEIR, or using household SR means rather than modeled SR, with the exception that in these analyses An. funestus was not significantly associated with infection risk in Busia (S3 Fig, S1 Table, S3–S5 Tables). In both districts, the fully parameterized model performed better by AIC than a corresponding model without An. funestus or An. gambiae aEIRs as covariates.

Discussion

In Tororo, during a period after a change in IRS insecticide when malaria incidence rose and An. funestus became the dominant local malaria vector [1], increased household-level An. funestus EIR was a stronger predictor of changes in P. falciparum infection rates than An. gambiae EIR as measured using CDC light trap collections. This finding supports the inference that the replacement of An. gambiae with An. funestus was a driver of increased malaria incidence from 2021-23 and demonstrates associations between household-level entomological data and individual infection risk.

We have previously provided evidence that the recent dominance of An. funestus in Tororo was associated with the development of clothianidin resistance [1,18]. This pattern is analogous to a 2000 KwaZulu-Natal malaria resurgence, which was associated with the invasion of a resistant An. funestus population [19]. The present analysis further clarifies the role played by An. funestus in driving malaria infections. Increasing household An. funestus aEIR was more strongly associated with risk of P. falciparum infection both in Tororo, where An. funestus populations increased, and in Busia, where they remained stable, and An. gambiae s.l. aEIR had smaller associations in both districts.

Our study’s strengths lay in the measurement of longitudinal rates of infectious bites and infections, which offered an unusual opportunity to characterize the fundamental relationship between infectious mosquito exposure and P. falciparum infection in a real-world setting. We assessed this association at fine spatial and temporal scales, avoiding the imprecision inherent in using proxies for incident infection, such as incident malaria. It is in part because of this precision that we were also able to identify differences in the risk of exposure associated with different vector species.

There were, however, weaknesses inherent to our approach to estimating EIR. First, the relationship between EIR as estimated from field measurements and human exposure is complex. An. funestus s.s., the primary malaria vector within the Funestus subgroup, is typically considered endophagic and anthropophilic, meaning overnight indoor CDC light trap captures should be plausible surrogates for its biting behavior [20,21]. In contrast, the An. gambiae species complex in eastern Uganda primarily comprises two species with distinct ecologies: An. gambiae s.s., another classically endophagic and anthropophilic species, and An. arabiensis, a classically exophagic and zoophilic species, though both demonstrate substantial behavioral heterogeneity [6,21,22]. In this study, we were unable to distinguish the two An. gambiae sub-species or identify relative changes in proportional representation within the species complex, potentially obscuring important differences in human exposure. The respective roles played by these sub-species may also vary by district: An. arabiensis vector densities were greater than equal to those of An. gambiae s.s. throughout the study period in both Tororo zones, while the opposite was true in Busia [1]. Our use of an ELISA sporozoite assay that can be less sensitive than molecular assays, inability to quantify mosquito salivary gland sporozoite loads, and focus on P. falciparum to the exclusion of other malaria parasites also limited our ability to fully characterize risk associated with exposure to the two vector species.

The effects of vector control measures further complicate these differences between vector species. An additional study from this cohort suggested that, during the period when clothianidin-based IRS formulations were used, the proportion of exposure to mosquitoes occurring outdoors increased for An. gambiae s.l. but not for An. funestus [23]. A shift towards outdoor biting by An. gambiae s.l. would diminish the utility of indoor captures as surveillance. LLIN use, which may vary between sites, may also have differential effects on the mortality and behavior of different species. An. funestus has exhibited both resistance to pyrethroid-treated LLINs [24] and increased day biting in response to LLIN use [25]. If it were present, this behavioral shift could either obscure or amplify an association between estimated EIR and human infection rates, depending on the correlation between the behavior and mosquito captures.

More generally, entomological surveillance at small scales is invariably limited by noise and by the zero inflation that results from estimating sporozoite rates using small sample sizes. Although we attempted to address these issues through smoothing, this difficulty remains a fundamental issue for EIR estimation, particularly with the low nightly mosquito captures we described. Relatively small changes in vector density or sporozoite rates can translate to large changes in estimated EIR, meaning our results should be interpreted with caution. For instance, the changes in An. funestus aEIR shown in Table 2 are driven in part by a rise in An. funestus SR during the study period. While potential contributors include ecological factors such as increased feeding success or longevity in An. funestus or temporarily decreased rates of An. funestus emergence with a consequently aging, sporozoite-enriched population, this change could also reflect an artifact of the larger sample size available for sporozoite rate estimation during the resurgence.

It has historically been difficult to correlate household-level measures of a particular malaria vector species, or even anopheline mosquitoes in general, with increased P. falciparum infection risk in members of that household. Our results lend further credence to the idea that An. funestus was a key driver of the regional malaria resurgence in Tororo, Uganda, while demonstrating associations between fine-scale entomological data with individual risk of P. falciparum infection in cohorts with frequent follow-up and longitudinal assessment of parasitemia. This finding supports the role of detailed surveillance in identifying drivers of outbreaks, potentially allowing for targeted control measures.

Supporting information

S1 Text. Comparison of candidate entomological generalized additive models by geographic zone (Tables A-D).

X’s in columns indicate whether a model contained a spatial smooth or a household random effect. Models are compared using Akaike information criterion (AIC) and percentage of deviance explained. The models selected for further analysis are bolded.

(DOCX)

pgph.0004436.s001.docx (20.4KB, docx)
S1 Fig. Timeline of interventions.

Timeline of recent indoor residual spraying (IRS) and long-lasting insecticidal net (LLIN) distribution campaigns by district, including names of insecticide formulations and LLIN models.

(PNG)

pgph.0004436.s002.png (8.4KB, png)
S2 Fig. Distribution of times to infection, comparing Busia to combined Tororo zones.

Each boxplot represents the distribution of gap times in days for an individual participant. The boxplot includes marks indicating the median and first and third quartiles, two whiskers, and outlying points, with color varying by site.

(TIFF)

pgph.0004436.s003.tiff (185KB, tiff)
S3 Fig. Disregarding natural clearances: distribution of times to infection, comparing Busia to combined Tororo zones.

Each boxplot represents the distribution of gap times in days for an individual participant. The boxplot includes marks indicating the median and first and third quartiles, two whiskers, and outlying points, with color varying by site.

(TIFF)

pgph.0004436.s004.tiff (184.3KB, tiff)
S1 Table. Disregarding natural clearances: participant characteristics.

For the purposes of calculating infection incidence, person-time here omits all days between an incident infection and day 14 following the individual’s next malaria treatment (see Methods).

(DOCX)

pgph.0004436.s005.docx (14.2KB, docx)
S2 Table. Reanalysis including interactions between time period relative to malaria resurgence and aEIR.

All aEIRs are log2-transformed.

(DOCX)

pgph.0004436.s006.docx (15KB, docx)
S3 Table. Disregarding natural clearances: aHRs assuming a linear relationship between all covariates and the log hazard.

Models fit to expected EIRs. All aEIRs are log2-transformed.

(DOCX)

pgph.0004436.s007.docx (14.6KB, docx)
S4 Table. 28-day lagged aEIR: aHRs assuming a linear relationship between all covariates and the log hazard.

Models fit to expected EIRs. All aEIRs are log2-transformed.

(DOCX)

pgph.0004436.s008.docx (14.6KB, docx)
S5 Table. Alternative aEIR: aHRs assuming a linear relationship between all covariates and the log hazard.

aEIR estimated using modeled vector counts and household mean SR. All aEIRs are log2-transformed.

(DOCX)

pgph.0004436.s009.docx (14.5KB, docx)

Acknowledgments

We would like to thank the study participants and their families, as well as the study team, Makerere University-UCSF Research Collaboration, and Infectious Diseases Research Collaboration. Portions of this work were performed on the Wynton HPC Co-Op cluster, which is supported by UCSF research faculty and UCSF institutional funds. IRB is a Chan Zuckerberg Biohub Investigator.

Data Availability

The dataset underlying this study is available in the ClinEpiDB database: https://clinepidb.org/ce/app/workspace/analyses/DS_17191d35b9.

Funding Statement

Funding was provided by the National Institutes of Health as part of the International Centers of Excellence in Malaria Research (ICEMR) program (U19AI089674 to GD) and the UCSF Biology of Infectious Diseases Training Program (2T32AI007641 to GD). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

  • 1.Kamya MR, Nankabirwa JI, Arinaitwe E, Rek J, Zedi M, Maiteki-Sebuguzi C, et al. Dramatic resurgence of malaria after 7 years of intensive vector control interventions in Eastern Uganda. PLOS Glob Public Health. 2024;4(8):e0003254. doi: 10.1371/journal.pgph.0003254 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Charlwood JD, Hall T, Nenhep S, Rippon E, Branca-Lopes A, Steen K, et al. Spatial repellents and malaria transmission in an endemic area of Cambodia with high mosquito net usage. Malariaworld J. 2017;8:11. [PMC free article] [PubMed] [Google Scholar]
  • 3.Charlwood JD, Pinto J, Ferrara PR, Sousa CA, Ferreira C, Gil V, et al. Raised houses reduce mosquito bites. Malar J. 2003;2(1):45. doi: 10.1186/1475-2875-2-45 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Donnelly B, Berrang-Ford L, Ross NA, Michel P. A systematic, realist review of zooprophylaxis for malaria control. Malar J. 2015;14:313. doi: 10.1186/s12936-015-0822-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Konradsen F, van der Hoek W, Amerasinghe FP, Mutero C, Boelee E. Engineering and malaria control: learning from the past 100 years. Acta Trop. 2004;89(2):99–108. doi: 10.1016/j.actatropica.2003.09.013 [DOI] [PubMed] [Google Scholar]
  • 6.Kahamba NF, Finda M, Ngowo HS, Msugupakulya BJ, Baldini F, Koekemoer LL, et al. Using ecological observations to improve malaria control in areas where Anopheles funestus is the dominant vector. Malar J. 2022;21(1):158. doi: 10.1186/s12936-022-04198-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Charlwood JD, Smith T, Lyimo E, Kitua AY, Masanja H, Booth M, et al. Incidence of Plasmodium falciparum infection in infants in relation to exposure to sporozoite-infected anophelines. Am J Trop Med Hyg. 1998;59(2):243–51. doi: 10.4269/ajtmh.1998.59.243 [DOI] [PubMed] [Google Scholar]
  • 8.Climate Hazards Center Infrared Precipitation with Stations version 3. CHIRPS3 Data Repository. 2025. doi: 10.15780/G2JQ0P [DOI] [Google Scholar]
  • 9.Weiss DJ, Dzianach PA, Saddler A, Lubinda J, Browne A, McPhail M, et al. Mapping the global prevalence, incidence, and mortality of Plasmodium falciparum and Plasmodium vivax malaria, 2000-22: a spatial and temporal modelling study. Lancet. 2025;405(10483):979–90. doi: 10.1016/S0140-6736(25)00038-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.U.S. President’s Malaria Initiative. Kenya Malaria Operational Plan FY 2023. 2023. [Google Scholar]
  • 11.Hofmann N, Mwingira F, Shekalaghe S, Robinson LJ, Mueller I, Felger I. Ultra-sensitive detection of Plasmodium falciparum by amplification of multi-copy subtelomeric targets. PLoS Med. 2015;12(3):e1001788. doi: 10.1371/journal.pmed.1001788 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Wirtz RA, Duncan JF, Njelesani EK, Schneider I, Brown AE, Oster CN, et al. ELISA method for detecting Plasmodium falciparum circumsporozoite antibody. Bull World Health Organ. 1989;67(5):535–42. [PMC free article] [PubMed] [Google Scholar]
  • 13.McClure M, Arinaitwe E, Kamya MR, Rosenthal PJ, Nankabirwa J, Maxwell K, et al. Relating household entomological measures to individual malaria risk. Malar J. 2025;24(1):261. https://doi:10.1186/s12936-025-05504-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Wood SN. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. J R Stat Soc Series B Stat Methodol. 2010;73(1):3–36. doi: 10.1111/j.1467-9868.2010.00749.x [DOI] [Google Scholar]
  • 15.Hartig F. DHARMa: residual diagnostics for hierarchical (multi-level/mixed) regression models. R Package version 020. 2018. [Google Scholar]
  • 16.Briggs J, Teyssier N, Nankabirwa JI, Rek J, Jagannathan P, Arinaitwe E, et al. Sex-based differences in clearance of chronic Plasmodium falciparum infection. Elife. 2020;9:e59872. doi: 10.7554/eLife.59872 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Balan TA, Putter H. frailtyEM: An R package for estimating semiparametric shared frailty models. J Stat Soft. 2019;90(7). doi: 10.18637/jss.v090.i07 [DOI] [Google Scholar]
  • 18.Oruni A, Arinaitwe E, Adiga J, Otto G, Kyagamba P, Okoth J, et al. Significant variations in tolerance to clothianidin and pirimiphos-methyl in Anopheles gambiae and Anopheles funestus populations during a dramatic malaria resurgence despite sustained indoor residual spraying in Uganda. bioRxiv. 2025:2025.02.13.638152. doi: 10.1101/2025.02.13.638152 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Hargreaves K, Koekemoer LL, Brooke BD, Hunt RH, Mthembu J, Coetzee M. Anopheles funestus resistant to pyrethroid insecticides in South Africa. Med Vet Entomol. 2000;14(2):181–9. doi: 10.1046/j.1365-2915.2000.00234.x [DOI] [PubMed] [Google Scholar]
  • 20.Gillies M, Coetzee M. A supplement to the Anophelinae of Africa south of the Sahara (Afrotropical region). Johannesburg: South African Institute for Medical Research; 1987. [Google Scholar]
  • 21.Takken W, Verhulst NO. Host preferences of blood-feeding mosquitoes. Annu Rev Entomol. 2013;58:433–53. doi: 10.1146/annurev-ento-120811-153618 [DOI] [PubMed] [Google Scholar]
  • 22.Takken W, Charlwood D, Lindsay SW. The behaviour of adult Anopheles gambiae, sub-Saharan Africa’s principal malaria vector, and its relevance to malaria control: a review. Malar J. 2024;23(1):161. doi: 10.1186/s12936-024-04982-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Krezanoski P, Musiime A, Oruni A, McClure M, Kyagamba P, Otto G, et al. Adjusting vector surveillance for human behaviors reveals Anopheles funestus drove a resurgence in malaria despite IRS with clothianidin in Uganda. Sci Rep. 2025;15(1):17728. doi: 10.1038/s41598-025-00623-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Morgan JC, Irving H, Okedi LM, Steven A, Wondji CS. Pyrethroid resistance in an Anopheles funestus population from Uganda. PLoS One. 2010;5(7):e11872. doi: 10.1371/journal.pone.0011872 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Sougoufara S, Diédhiou SM, Doucouré S, Diagne N, Sembène PM, Harry M, et al. Biting by Anopheles funestus in broad daylight after use of long-lasting insecticidal nets: a new challenge to malaria elimination. Malar J. 2014;13:125. doi: 10.1186/1475-2875-13-125 [DOI] [PMC free article] [PubMed] [Google Scholar]
PLOS Glob Public Health. doi: 10.1371/journal.pgph.0004436.r001

Decision Letter 0

Amy Kristine Bei

4 Jun 2025

PGPH-D-25-00552

Disentangling the roles of different vector species during a malaria resurgence in Eastern Uganda

PLOS Global Public Health

Dear Dr. McClure,

Thank you for submitting your manuscript to PLOS Global Public Health. After careful consideration, we feel that your manuscript has significant merit but does not fully meet PLOS Global Public Health’s publication criteria as it currently stands. Therefore, we would encourage you to submit a revised version of the manuscript that addresses the points raised during the review process.

Please carefully consider the comments and constructive suggestions by both reviewers especially regarding the methodological aspects and the interpretation of the data given the limitations of the study. 

Please submit your revised manuscript by Jul 04 2025 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at globalpubhealth@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pgph/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

  • A rebuttal letter that responds to each point raised by the editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

  • A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

  • An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

We look forward to receiving your revised manuscript.

Kind regards,

Amy Kristine Bei

Academic Editor

PLOS Global Public Health

Journal Requirements:

1. Please include a complete copy of PLOS’ questionnaire on inclusivity in global research in your revised manuscript. Our policy for research in this area aims to improve transparency in the reporting of research performed outside of researchers’ own country or community. The policy applies to researchers who have travelled to a different country to conduct research, research with Indigenous populations or their lands, and research on cultural artefacts. The questionnaire can also be requested at the journal’s discretion for any other submissions, even if these conditions are not met.  Please find more information on the policy and a link to download a blank copy of the questionnaire here: https://journals.plos.org/globalpublichealth/s/best-practices-in-research-reporting. Please upload a completed version of your questionnaire as Supporting Information when you resubmit your manuscript.

2. Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Does this manuscript meet PLOS Global Public Health’s publication criteria?>

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?-->?>

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available (please refer to the Data Availability Statement at the start of the manuscript PDF file)??>

The PLOS Data policy

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English??>

Reviewer #1: Yes

Reviewer #2: Yes

**********

Reviewer #1: This is a complicated statistical study of Anopheles vectors captured in different Ugandan districts over the course of malaria resurgence in one district, that the authors are trying to casually link to the change in IRS chemistries from predominately organophosphates, to clothianidin-based, and then back to organophosphates. The foundation of the paper was already laid in the more associative analysis of case and parasitological data at the sites relative to the timeline of the IRS events, vector density and mosquito phenotypic insecticide resistance data published last year (Kamya et al, 2024, PLOS GPH). The paper reads very well and I saw no grammar or confusing text that needed editing. The strengths of the paper are in using a frailty model analysis connecting the time to incident infection of the children studied (adjusted Hazard Ratios) with the annualized EIR data overall, and by species at the two sites, which points to Anopheles funestus as driving the malaria resurgence in Tororo during the implementation of the two clothianidin IRS and then driven back down when OP (Actellic) was used for IRS. It is interesting that these entomological data trends in are not very apparent in simply examining the raw plotted data in Figure 1 compared to the child infection data. I would have liked to know if the entomological data were analyzed by quarter or semi-annually, and if so, were the associations made even stronger? - because there seem to be a couple of peaks each year and maybe different depending on species. The modeling performed seems solid, and the collection of time to incident infection data is particularly strong and analyzed in a couple of different ways to address biases. Overall, the claims made by the paper are backed up by the data presented, and considering the prior published paper. A major weakness of the assumptions is the collection of indoor host-seeking mosquitoes using light trap catches when IRS is the primary control tool and should be acting as a major repellent for those indoor catches in addition to the overall lethal effect, although they appropriately speak to this weakness in the discussion. Another weakness they don’t address that they only tested/account for P. falciparum, and only using an ELISA assay that can be less sensitive than molecular assays. Lastly, the data are lumped together as FF+SS are both clothianidin-based, but infection incidence doesn’t seem to rise until the SS is implemented rather than FF, and the latter contains added pyrethoid and I’m sure the formulations/long-lasting efficacies are different – can this all be more solidly connected to the SS spray alone?

Reviewer #2: Summary

The authors have provided an interesting manuscript detailing further attempts at elucidating the role of different Anopheles vector species in contributing to a malaria resurgence that occurred in Tororo District, Uganda that followed district level changes in insecticide use for Indoor Residual Spraying (IRS) in 2020. Previous work from the same group had suggested this resurgence also coincided with replacement of Anopheles gambiae sensu lato by Anopheles funestus as the primary local vector. Specifically, these same authors have published extensively on this data cohort in this same journal (Kamya et al., 2024) where they initially did all the work to analyze changes in case rates, parasite prevalence, IRS insecticide use, and local vector characteristics between Turoro and Busia Districts.

Here the authors also used data from their prior cohort of 422 children followed from 2021-2023 in two neighboring districts, Turoro (where IRS is implemented) and Busia (no IRS) with passive and monthly active surveillance for Plasmodium falciparum (Pf), and also within the 94 households where these children resided, mosquitoes were collected every 2 weeks using CDC light traps, speciated, and assessed for malaria sporozoite presence in order to calculate annual household entomologic inoculation rates (aEIR). The authors focused in this manuscript on modeling analyses at the household and species level using aEIR and time to infection data. Specifically, they analyzed changes in entomologic metrics (including species-specific aEIR) and changes in infection risk (time to incident infection) over the study period and used this data to model associations, namely the relationship between aEIR and Pf infection risk. They found aEIR increases to be associated with increases in Pf (two-fold increase resulting in 29% HR increase), and that An. funestus aEIR increase was associated with a larger increase in Pf infection HR than for An. gambiae aEIR. Additionally, this An. funestus aEIR association was larger in Tororo than Busia, leading them to conclude this modeling data supports the idea that replacement of An. gambiae with An. funestus drove the malaria resurgence seen in Turoro in 2021-2023. They highlight their unique ability to use longitudinal household-level and species-level EIR data to determine Pf infection risk, which has historically been lacking and is important.

I overall agree with their data presentation, findings, and interpretations. My biggest question has to do with the ability to make significant conclusions based on at times relatively small differences in their model outputs (although the modeling here certainly supports their extensive prior work). In addition, I would ask for some further clarifications (see comments below).

Comments to address

• The manuscript is clear, well-organized, and well written, however there are a few figure legend issues that could be clarified.

o Figure 1 - no legend included? I understand it is a simple figure so may not need it; just want to verify it wasn’t missed.

o Supplemental Figures S1 and S2 – would be helpful to include in the legend the description of the line in the plot. Presumably based on its color it represents the Busia trends of time to infection across individuals. I would want to see the same line for Turoro as well, to better depict comparisons you make in the text (specifically stating time to next infection is longer in Busia).

• Given the proximity of Busia households and the also those in the Turoro border region to Kenya (as evidenced by Figure 1), I think you should also include in the intro or discussion the basics about what is known about malaria cases in this Northwestern Kenya region during the same time frame and their infection control measures.

• Not absolutely necessary, but I would consider the benefit of another component to Figure 1 being a longitudinal depiction of the study timeframe with different IRS insecticide and LLIN distribution interventions highlighted (e.g. arrows above the line for Turoro interventions, arrows below the timeline for Busia interventions)

• Would appreciate better/more explicit description of malaria cases (rates of symptomatic vs asymptomatic) at least in the methods. What percentage of people at baseline enrolment time had asymptomatic parasitemia? You refer to spontaneous clearances a lot without explicitly defining them as presumably untreated asymptomatic people who naturally clear – it would be better to state that (pg 8 line 151-152), as well as to clearly state that 3 consecutive negative screening periods would cover 3 months (based on q28 day sampling?) – which I agree is a reasonable time to assume that newly detected parasitemia is more likely from a new infection. Of course it is possible, as you allude to, that if a person had asymptomatic infection with no treatment required, their subsequent parasitemia or symptomatic case could be from prior parasite versus from new infection from new mosquito bite. I know that you also included data in your supplemental info about a more “conservative” analysis with incident infections only considered after treatment, presumably to account for this scenario. I think it would just overall help to be more clear about in your manuscript about these nuances, but why you would ultimately make the same conclusions – could even just add a sentence in the discussion.

• I also think somewhere in your manuscript you should describe seasonality of malaria cases in this region. Were there any significant weather pattern or case seasonality changes during the resurgence (i.e. changes in rainfall) that should be considered – given An. funestus is slightly more adapted to dryer conditions than An. gambiae?

• Similarly, for more context regarding regional intervention differences – why doesn’t Busia get have IRS implemented? Are there other significant differences in the Busia landscape? What happened with bednet distribution in 2023 (presumably every 3 year planned distribution schedule) – did this occur and could this have affected the subsequent case decline in Turoro again?

• Some clarifications regarding Table 2:

o I understand your data implies there was not a malaria resurgence in Busia, nor changes in vector species, over the period of interest. Still, comparing segmented Turoro data (divided as before/during/after IRS insecticide change) to Busia data collectively over that whole period makes it harder to verify. The Busia data for Table 2 would be helpful to see in a similar before/during/after breakdown (at least in supplemental info) for better comparison.

o It was surprising and somewhat confusing to see aEIR of 0 at so many times – presumably driven by the SR with no infected mosquitoes collected for a given time. Normally would not expect the IQR to be 0 so often given the overall high malaria prevalence in this region? Please clarify your thoughts.

o I would better define the numbers scale in the legend. Presumably number of collections equates to households? What are the An. gambiae or An. funestus counts – an actual number of mosquitoes collected? If so, this is quite low for mosquitoes collected in each house. It is confusing because, looking at the before/during/after of An. funestus count in Turoro near, there is a decline over all times (like An. gambiae, though much less drastic) – which seemingly doesn’t align as well the idea of with An. funestus influx driving a malaria resurgence. This would make it seem the SR changes (increases) for An. funestus are driving aEIR data. It’s more intuitive to think about mosquito numbers changing with insecticide use – but why would An. funestus SR rates change over the study period? This is another reason it would be helpful to see Busia data also divided in time for comparison.

• This previous point brings up the general issue of how the relatively low overall number of sporozoite-infected mosquitoes detected makes it difficult to precisely evaluate temporal trends and make significant conclusions. I appreciate that your modeling work tries to overcome this limitation – but I think it is worth mentioning in your discussion/limitations. (Especially also considering we don’t have any data about quantitative sporozoite burden between An. gambiae and An. funestus here, another factor which can contribute to subsequent infection risk – though I recognize it’s not realistic to assess sporozoite loads with your sample set.)

• Likewise, regarding data in Fig 2 – An. gambiae and An. funestus data look quite similar regarding vector numbers in Turoro areas for much of the main period of interest as do the species-specific SR data especially for Turoro near and Busia. And yet aEIR increases more notably for An. funestus. Hard to see clear pattern in graphed mosquito capture and SR data, but the modeling implies the aEIR changes are bigger. I think this requires more nuanced interpretation of results – making clear that your conclusions are based on aEIR estimates and modeling and noting small differences in vector and SR detection can show up bigger in these EIR-based models, though obviously the infection rise over this period is striking.

o Also, following the data from Figure 2 input to generate the aHR data in Table 3 – it is important to contextualize and note that despite the story and data fitting nicely regarding the switch of IRS insecticide, rise in malaria cases, and noted relative increase in An. funestus mosquito presence, the differences in species-specific aEIR-based aHR calculations between Busia and Turoro are overall still small. Thus, I do especially appreciate the discussion acknowledgement about EIR measures being complex and hard, particularly with An. gambiae having 2 subspecies, and different behavioral characteristics of the various Anopheles vectors and how this could skew the modeling here based on your indoor capture data. It will be interesting to see the data from the publication you mentioned that is under review regarding outdoor exposures. (Nothing requested to change, just a comment.)

• Did you consider using PCR as the method for sporozoite identification rather than ELISA? Would be interesting to have simultaneously looked at presence of different malaria species in your cohort (given some others are finding potential associations between An. funestus and P. ovale, for example (https://www.medrxiv.org/content/10.1101/2024.10.09.24315124v1)).

**********

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

PLOS Glob Public Health. doi: 10.1371/journal.pgph.0004436.r003

Decision Letter 1

Amy Kristine Bei

18 Nov 2025

Disentangling the roles of different vector species during a malaria resurgence in Eastern Uganda

PGPH-D-25-00552R1

Dear Dr. McClure,

Thank you for submitting your revised manuscript and for thoughtfully incorporating the suggestions of reviewers. 

We are pleased to inform you that your manuscript 'Disentangling the roles of different vector species during a malaria resurgence in Eastern Uganda' has been provisionally accepted for publication in PLOS Global Public Health!

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they'll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact globalpubhealth@plos.org.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Global Public Health.

Best regards,

Amy Kristine Bei

Academic Editor

PLOS Global Public Health

***********************************************************

Reviewer Comments (if any, and for reference):

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    S1 Text. Comparison of candidate entomological generalized additive models by geographic zone (Tables A-D).

    X’s in columns indicate whether a model contained a spatial smooth or a household random effect. Models are compared using Akaike information criterion (AIC) and percentage of deviance explained. The models selected for further analysis are bolded.

    (DOCX)

    pgph.0004436.s001.docx (20.4KB, docx)
    S1 Fig. Timeline of interventions.

    Timeline of recent indoor residual spraying (IRS) and long-lasting insecticidal net (LLIN) distribution campaigns by district, including names of insecticide formulations and LLIN models.

    (PNG)

    pgph.0004436.s002.png (8.4KB, png)
    S2 Fig. Distribution of times to infection, comparing Busia to combined Tororo zones.

    Each boxplot represents the distribution of gap times in days for an individual participant. The boxplot includes marks indicating the median and first and third quartiles, two whiskers, and outlying points, with color varying by site.

    (TIFF)

    pgph.0004436.s003.tiff (185KB, tiff)
    S3 Fig. Disregarding natural clearances: distribution of times to infection, comparing Busia to combined Tororo zones.

    Each boxplot represents the distribution of gap times in days for an individual participant. The boxplot includes marks indicating the median and first and third quartiles, two whiskers, and outlying points, with color varying by site.

    (TIFF)

    pgph.0004436.s004.tiff (184.3KB, tiff)
    S1 Table. Disregarding natural clearances: participant characteristics.

    For the purposes of calculating infection incidence, person-time here omits all days between an incident infection and day 14 following the individual’s next malaria treatment (see Methods).

    (DOCX)

    pgph.0004436.s005.docx (14.2KB, docx)
    S2 Table. Reanalysis including interactions between time period relative to malaria resurgence and aEIR.

    All aEIRs are log2-transformed.

    (DOCX)

    pgph.0004436.s006.docx (15KB, docx)
    S3 Table. Disregarding natural clearances: aHRs assuming a linear relationship between all covariates and the log hazard.

    Models fit to expected EIRs. All aEIRs are log2-transformed.

    (DOCX)

    pgph.0004436.s007.docx (14.6KB, docx)
    S4 Table. 28-day lagged aEIR: aHRs assuming a linear relationship between all covariates and the log hazard.

    Models fit to expected EIRs. All aEIRs are log2-transformed.

    (DOCX)

    pgph.0004436.s008.docx (14.6KB, docx)
    S5 Table. Alternative aEIR: aHRs assuming a linear relationship between all covariates and the log hazard.

    aEIR estimated using modeled vector counts and household mean SR. All aEIRs are log2-transformed.

    (DOCX)

    pgph.0004436.s009.docx (14.5KB, docx)
    Attachment

    Submitted filename: reviewer_responses_IRB.docx

    pgph.0004436.s011.docx (36.9KB, docx)

    Data Availability Statement

    The dataset underlying this study is available in the ClinEpiDB database: https://clinepidb.org/ce/app/workspace/analyses/DS_17191d35b9.

    Articles from PLOS Global Public Health are provided here courtesy of PLOS

    RESOURCES