The study was conducted in Muhoroni sub-County, Kisumu County, Kenya, located in the Kano Plain, at an elevation of 1,120–1,140 meters above sea level. The area is characterized by extensive sugarcane plantations and rice fields, which are often flooded during the rainy season, creating suitable habitats for mosquito breeding. Muhoroni experiences perennial malaria transmission, with Anopheles arabiensis as the dominant vector and Plasmodium falciparum accounting for 95.2% of infections (19), and standard LLINs serving as the main malaria control strategy. The study monitored standard LLINs distributed through Ministry of Health mass campaigns and PBO LLINs provided by the ICEMR program as part of a cluster-randomized trial evaluating optimal combinations of vector control interventions for malaria control between March 2021 and August 2024 (20). Standard LLINs were distributed by community health promoters and health facility staff during routine mass campaigns led by the Ministry of Health. In clusters assigned to PBO LLINs, participating household received WHO-certified PermaNet^® 3.0 (PBO polyester) nets to replace standard LLINs, ensuring adequate coverage of one net per two people or one net per sleeping space.

Anopheles gambiae s.l. larvae were collected from at least four clusters per intervention arm and pooled in the laboratory. Larvae were reared to adulthood under controlled conditions (27 ± 1 °C, 70 ± 10%, 12:12 h light: dark cycle) and fed on a mixture of fish food and yeast until pupation. Emerging adults were maintained on 10% sugar solution. Similarly, the susceptible An. gambiae s.s. Kisumu strain used as control was raised and maintained under the same conditions.

Nets were randomly retrieved from clusters assigned to either PBO LLINs or standard LLINs, with cluster selection and randomization procedures detailed elsewhere (20). Nets of different operational ages were included: 0, 6, 18, and 36 months post-distribution. At each time point, 13 nets were randomly collected from households within each study arm (PBO and standard) and replaced with equivalent new nets. The standard LLINs included Permanent^® 2.0 (polyester, deltamethrin-coated fibers at 55 mg/m²), Yorkool^® (polyester 100-denier, deltamethrin 55 mg/m²), and Dawa Tana (polyester 100-denier, deltamethrin 2.0 g/kg ± 25%, corresponding to 80 mg/m²). The PBO LLIN was Permanent^® 3.0, with a polyethylene roof incorporating deltamethrin (2.8 g/kg ± 25%) and PBO (4.0 g/kg ± 25%) into the fibers, and polyester lateral sides coated with deltamethrin (2.8 g/kg ± 25%). Community Health Promoters (CHPs) supported the trial by sensitizing households on the importance of retaining nets and subsequent utilization.

Phenotypic resistance was assessed using the WHO standard tube test (21) with insecticides commonly used for LLINs and IRS. Anopheles gambiae s.l. females, 3–5 days old, reared from field-collected larvae, were tested against deltamethrin (0.05%) annually from 2021 (pre-trial) to 2024. In 2024, additional bioassays were conducted using permethrin (0.75%), PBO 4% + deltamethrin (0.05%), clothianidin (X1), and pirimiphos-methyl (0.25%). For each insecticide, four replicates of 20–25 unfed females were exposed to insecticide-impregnated papers for one hour following WHO protocols (Supplementary Table S1). For synergist assays, mosquitoes were pre-exposed to PBO-impregnated papers for one hour prior to exposure to deltamethrin to assess metabolic resistance. Anopheles gambiae Kisumu strain was included as a reference susceptible control in all bioassays. Mortality was recorded 24 h post-exposure, and resistance status was classified according to WHO criteria (21).

Bio-efficacy of sampled nets, including both standard LLINs and PBO LLINs, was evaluated using cone test to determine the ability of insecticides on the roof and sides to induce mosquito mortality. Five pieces (25 × 25 cm each) were cut from standardized positions (positions 1–5) on each net, as specified by the WHO Pesticide Evaluation Scheme (16). Adult female Anopheles gambiae s.l., 3–5 days old, reared from field-collected larvae, and susceptible Anopheles gambiae s.s. Kisumu strain were used for the assays. For each net piece, a WHO cone was fitted, and five non-blood-fed females were introduced and exposed for 3 minutes. Following exposure, mosquitoes were carefully transferred to separate plastic cups and maintained under standard conditions. Mortality was recorded 24 h post-exposure. The bio-efficacy of each net was determined as the average mortality across the five net pieces. Mosquitoes exposed to untreated nets served as negative control.

A sub-sample of surviving and susceptible An. gambiae s.l. mosquitoes were randomly selected for species identification after exposure to the WHO tube bioassay. Genomic DNA was extracted from individual mosquitoes following previous methods (22). Molecular identification of sibling species of respective An. gambiae s.l. were conducted based on PCR methods described by Scott et al. (23).

Mosquitoes were considered susceptible if mortality was ≥98% and resistant if mortality was <90%. Mortality rates between 90% and 97% indicated possible resistance, requiring confirmation through additional tests (21). If the mortality in controls was < 10%, the results were adjusted by Abbott’s formula. If the mortality in controls were > 10%, the results were considered invalid and were discarded. The allele frequencies for resistant genotypes were calculated using the Hardy-Weinberg equilibrium equation. Mosquito populations were checked to determine whether they were in Hardy-Weinberg equilibrium using χ² test. For bio-efficacy, mortality rate was first transformed using arcsine transformation. Data was analyzed using analysis of variance (ANOVA) with repeated measures and Tukey-Kramer HSD test used to compare mean mortalities among different time points. Data analysis was performed using the open-source R programming language software R version 4.2.3.

All 600 randomly selected An. gambiae s.l. mosquitoes (survivors and susceptible) were confirmed by PCR as Anopheles arabiensis, confirming its predominance in the study area throughout the monitoring period. In early 2021, prior to the distribution of PBO nets, mortality of An. arabiensis exposed to 0.05% deltamethrin was 96.3 ± 4.3% (mean ± SD, n = 12 replicate tests) (Figure 1A). However, resistance increased sharply following subsequent years of LLIN use: mortality declined to 53.0 ± 6.0% in 2022 (n = 13), followed by 31.5 ± 4.5% in 2023 (n = 12) and 22.7 ± 4.2% in 2024 (n = 15). No significant difference in mortality was observed between 2023 and 2024 (p = 0.08, Figure 1A). Remarkably, pre-exposure to the synergist PBO in 2024, restored deltamethrin susceptibility with mortality increasing to 98.9 ± 0.4%, confirming role of oxidase mediated resistance mechanism (Figure 1B).

Resistance to other pyrethroid was evident in 2024, with An. arabiensis showing reduced susceptibility to permethrin with mortality of 35.2 ± 7.3% against the WHO standard diagnostic concentration (Figure 1B). In contrast, exposure to non-pyrethroid insecticides used for indoor residual spraying showed full susceptibility: mortality was 100% against clothianidin (neonicotinoid) and 99.7 ± 0.3% against pirimiphos-methyl (organophosphate) (Figure 1B).

A total of 350 Anopheles arabiensis specimens were successfully genotyped for the three target mutations in 2022 and 2024. Distinct temporal trends in the frequency of kdr mutations were observed between the two intervention periods. In 2022, the frequency of the 1014S allele was low (allelic frequency = 0.04), and the 1014F allele frequency was similarly low (0.09). In 2024, the 1014S allele showed a slight increase (0.06), while the 1014F allele exhibited a substantial rise to 0.17 (Table 1). Overall, both loci demonstrated an increase in resistant allele frequencies between 2022 and 2024, with a more pronounced change at the 1014F locus.

Hardy-Weinberg equilibrium analysis showed that none of the An. arabiensis populations tested for 1014S were in equilibrium in either year (P < 0.001), likely reflecting heterozygote deficiency. In contrast, the increased heterozygosity observed at the 1014F locus, suggesting active selection and ongoing spread of the resistance allele (Table 2). No Ace-1 (G119S) mutations associated with altered acetylcholinesterase were detected in any of the mosquitoes screened.

According to World Health organization Pesticide Evaluation Scheme (WHOPES) guidelines, LLIN bio-efficacy is defined by (i) ≥80% mortality after 24 hours and (ii) ≥95% knockdown (KD) after 60 minutes of exposure (16). For standard LLINs, the mean knockdown rate against the An. gambiae s.s. susceptible strain remained high during the first 18 months post-distribution (>95%) but declined below the WHOPES threshold (≥95% knockdown) by 36 months (86.3 ± 2.0%) (Table 3). When tested against field mosquitoes, the nets failed to achieve optimal knockdown efficacy even when new (Mean knockdown rate: 50.2 ± 2.4%), with a progressive reduction to 6.7 ± 1.5% at 36 months post-distribution. Piperonyl butoxide LLINs, knockdown rates for susceptible mosquitoes remained optimal (100%) for up to 18 months but declined (mean knockdown rate: 84.3 ± 5.5%) at 36 months post-distribution. For field mosquitoes, baseline knockdown rate was higher (95.1 ± 4.0%) but decreased markedly to 45 ± 7.8% at 6 months and 25 ± 7.5% at 36 months (Table 1).

In terms of mortality, standard LLINs maintained optimal operational efficacy (>80% mortality) for up to 18 months (91.6%) but dropped below the threshold by 36 months (74.7%) when tested with susceptible mosquitoes (Figure 2A, top-left panel). However, efficacy against field mosquitoes was consistently below the optimal level, with <1-month post-distribution mortality at 64.4%, declining sharply to 13% at 6 months and 10% at 36 months (Figure 2C, bottom-left panel). Similarly, PBO LLINs induced >99% mortality in susceptible mosquitoes up to 18 months post-distribution, followed by a decline to 68.1% at 36 months (Figure 2B, top-right panel). Against field populations, initial efficacy (100%) decreased substantially to 24% at 6 months, 18% at 18 months, and 15% at 36 months post-distribution, well below the 80% threshold (Figure 2D, bottom-left panel). Overall, both net types exhibited a significant decline in killing efficacy over time, with average mosquito mortality dropping below 20% within six months of field use (P < 0.001) (Figures 2C, D, bottom panel).