**3. Results**

## *3.1. Experimental Validity*

In all the bioassays conducted, control M24 was <10% and at M72 was <13%. BFS was ≥50% in both the human arm and the rabbit controls and was <23% in the membrane control (Table 2).

**Table 2.** Impact of bait on mortality and blood-feeding adjusted for the net condition. The difference in the odds of mosquito mortality at 24-h (M24) or 72-h (M72) and blood feeding success (BFS) for 100 pyrethroid-resistant *Anopheles arabiensis* exposed to Interceptor® and Interceptor® G2 with either a rabbit, human arm or membrane feeder as bait \*.


\* Mosquitoes were exposed for 12 h. Data presented are mean proportion (%) with 95% confidence interval (95% CI) and odds ratios (OR) derived from regression analysis with 95% CI adjusted for net type and condition.

#### *3.2. Experiment 1: The Impact of Baits*

The bait used affected both the feeding and mortality endpoints measured. The membrane measured a similar mortality and a lower blood feeding success than the rabbit. The human arm measured a lower mortality and higher blood feeding success than the rabbit.

M24 in the intervention arms was not significantly different between the rabbit and membrane (OR: 0.90, 95% CI: 0.79–1.02, *p* = 0.086) and was significantly lower using the human arm (OR: 0.42, 95% CI: 0.37–0.48, *p* < 0.001) compared to the rabbit (Table 2). M72 in the intervention arms was not significantly different between the rabbit and membrane (OR: 1.07, 95% CI: 0.93–1.22, *p* = 0.352) and was significantly lower using the human arm (OR: 0.31, 95% CI: 0.27–0.35, *p* < 0.001) compared to the rabbit (Table 2). Control M24 was higher in the membrane and human arms; but control M72 was higher in the human arm (OR: 1.83, 95% CI: 1.22–2.75, *p* = 0.004) and was not different between rabbit and membrane (OR: 1.16, 95% CI: 0.84–1.59, *p* = 0.366). In the treatment arm, BFS was significantly lower using a membrane (OR: 0.34, 95% CI: 0.28–0.48, *p* < 0.001) and was significantly higher using a human arm (OR: 9.81, 95% CI: 8.25–11.67, *p* < 0.001) compared to the rabbit (Table 2). The same trend was observed in the control arm.

The same trends in mortality and blood feeding inhibition (BFI) were observed for both Interceptor® ITN and Interceptor® G2 (Figure 4). Higher blood feeding resulted in lower mortality (Figure S1), which will explain the lower mortality measured with the human arm, which also had substantially higher BFS. Therefore, the human arm could not be considered for further evaluation. Between the membrane and the rabbit with 100 mosquitoes per replicate, no systematic difference was observed for agreemen<sup>t</sup> by Bland and Altman methods (Figure S2). The mean difference was 6% (−10.81–23.01) for BFS and −1.09% (−72.91–70.73) for M72.

**Figure 4.** Mean percentage mortality and 95% Confidence Interval (CI) for mortality at (**A**) 24-h (M24), (**B**) 72-h (M72) post exposure and (**C**) blood feeding inhibition (BFI) for Interceptor® and Interceptor® G2 nets with 100 pyrethroid-resistant *Anopheles arabiensis* mosquitoes using rabbit, Hemotek® membrane feeders and human arm as bait in the WHO tunnel bioassay. Red dashed line depicts the WHO minimum bioefficacy criteria of ≥80% M24 and ≥95% BFI.

#### *3.3. Experiment 2: Impact of Exposure Time on Mortality and Blood Feeding*

Increasing the time that mosquitoes are left in the tunnel from 1-h to 12-h increased mortality with either the human arm or the membrane (Table 3). With the membrane bait, longer exposure significantly increased both the odds of M72 (OR: 2.30, 95% CI: 2.02–2.62, *p* = 0.001) and the odds of BFS (OR: 1.55, 95% CI: 1.08–2.22, *p* = 0.017). Similarly, in the human arm, the longer exposure significantly increased the odds of M72 (OR: 1.66, 95% CI: 1.45–1.90, *p* = 0.001), while the effect of exposure time on BFS could not be measured since the human arm was only available for one hour (Figure 5). The time that mosquitoes are left in the tunnel overnight is a significant factor in mosquito mortality and should always be recorded and reported.

**Table 3.** Impact of exposure time on mortality and blood-feeding adjusted for the net condition; The difference in the odds of mosquito mortality at 24-h (M24) or 72-h (M72) and blood-feeding success (BFS)) for 100 pyrethroid-resistant *Anopheles arabiensis* exposed to Interceptor® and Interceptor® G2 with either a human arm or a membrane feeder as bait \*.


\* Mosquitoes were exposed for either 1 h before being removed from the tunnel and placed in holding cups with access to sugar or left overnight in the tunnel for 12 h. Data presented are a mean proportion (%) with a 95% confidence interval (95% CI) and odds ratios (OR) derived from regression analysis with 95% CI adjusted for net conditions.

**Figure 5.** Mean and 95% Confidence Interval (CI) for (**A**) 24-h (M24); (**B**) 72-h (M72); and (**C**) blood feeding success (BFS) with 100 pyrethroid-resistant *Anopheles arabiensis* mosquitoes with 12 h or 1 h exposure time in the WHO tunnel bioassay using Hemotek® membrane or human arm as bait.

#### *3.4. Experiment 3: Effects of Mosquito Density on Tunnel Test Endpoints*

M24, M72, and BFS were very similar and were not statistically different when either 50 or 100 mosquitoes were used in the tunnel bioassay with rabbit bait for either the susceptible or resistant strains (Table 4). This was consistent for both Interceptor® and Interceptor® G2, unwashed and washed 20 times (Figure 6). No systematic difference in agreemen<sup>t</sup> between methods was observed by Bland and Altman methods (Figure S3). The mean difference was −4.54% (−31.62–22.54) in BFS and 1.71% (−28.71–32.12) in M72. Furthermore, when tested using the pyrethroid-resistant strain, the 50-rabbit bioassay predicted the superiority of Interceptor® G2 to Interceptor®, as did the 100-rabbit (Table 5).

**Table 4.** Effects of mosquito density on mortality and blood-feeding. The difference in the odds of mosquito mortality at 24 h (M24) or 72 h (M72) and blood feeding success (BFS) for resistant *Anopheles arabiensis* exposed to Interceptor® G2 or susceptible *Anopheles gambiae* to Interceptor® in the gold standard rabbit-100 and 50-rabbit mosquitoes \*.


\* Mosquitoes were exposed for 12 h in the tunnel. Data presented are a mean proportion (%) with 95% confidence interval (95% CI) and odds ratios (OR) derived from regression analysis with 95% CI adjusted for net type and condition.

**Figure 6.** Mean and 95% confidence Interval (CI) for (**A**) 24 h mortality (M24), (**B**) blood feeding inhibition (BFI) of Interceptor® ITN with 100 vs. 50 pyrethroid susceptible *Anopheles gambiae*; (**C**) 72-h mortality (M72); and (**D**) BFI of Interceptor® G2 ITN with 100 vs. 50 pyrethroid-resistant *Anopheles arabiensis* in the WHO tunnel test. Red dashed line depicts WHO minimum bioefficacy thresholds of ≥80% M24 and ≥95% BFI.

**Table 5. Superiority of Interceptor® G2 over Interceptor® using 100 versus 50 resistant mosquitoes:** The difference in the odds of mosquito at 24 h (M24) and 72 h (M72) and blood feeding success (BFS) for pyrethroid-resistant *Anopheles arabiensis* exposed to Interceptor® G2 and Interceptor® in the gold standard rabbit-100 and 50-rabbit mosquitoes \*.


\* Mosquitoes were exposed for 12 h in the tunnel. Data presented are a mean proportion (%) with a 95% confidence interval (95% CI) as well as odds ratios (OR) derived from regression analysis with 95% CI, adjusted for net conditions.

However, when considering the superiority of Interceptor® and Interceptor® G2, the lower mosquito density (50) resulted in a higher BFS in the Interceptor® G2 arm (Table 5). This indicates that mosquitoes at a high density are either interacting with each other to disturb each other from feeding, or discomfort from high biting rates is making the host more defensive. This increased blood feeding success is likely translating into the lower odds of mortality observed for washed Interceptor® G2 relative to Interceptor® using 50 mosquitoes (OR: 1.07, 95% CI: 0.85–1.34, *p* = 0.579) compared to 100 mosquitoes (OR: 1.31, 95% CI: 1.12–1.54, *p* = 0.001) (Table 5). This observation underlines the importance of consistent control blood feeding success on mortality estimates from the WHO tunnel test and this should always be recorded and reported.

#### *3.5. Experiment 4: Possibility to Replace Standard Bait with the Membrane Feeding*

The membrane assay with 50 mosquitoes (membrane-50) did not measure statistically different M24 or M72 compared to the rabbit with 100 mosquitoes (rabbit-100) (Table 6) when testing pyrethroid only Interceptor® or Interceptor® G2 against pyrethroid-resistant *An. arabiensis.* Again, BFS was different, with a far higher BFS in the rabbit-100 assay than in the membrane 50-assay.

**Table 6.** Comparison of the membrane assay to the gold standard with rabbit assay. The difference in the odds of mosquito mortality at 24-h (M24) and 72-h (M72) and blood feeding success (BFS) for resistant *Anopheles arabiensis* was measured between the gold standard rabbit assay with 100 mosquitoes and the membrane assay with 50 mosquitoes \*.


\* Mosquitoes were exposed for 12 h in the tunnel. Data presented are a mean proportion (%) with a 95% confidence interval (95% CI) as well as odds ratios (OR) derived from regression analysis with a 95% CI adjusted for net type.

However, when used for predicting the difference in bio-efficacy between Interceptor® and Interceptor® G2, both assays were measured in the same way (Figure 7) and both predicted superior odds of M72 for Interceptor® G2 (100-rabbit OR: 1.23 (95% CI: 1.10–1.38), *p* < 0.0001; 50-membrane 1.79 (95% CI: 1.50–2.14) *p* < 0.0001) and inferior reduction in blood feeding (100-rabbit OR: 1.76 (95% CI: 1.47–2.10), *p* < 0.0001; 50-membrane 1.87 (95% CI: 1.05–3.33) *p* = 0.033) with Interceptor® G2 relative to Interceptor® (Table 7). No systematic difference was observed in agreemen<sup>t</sup> for membrane-50 and rabbit-100 by Bland and Altman methods, with a mean difference (and limits of agreement) of 9.06 % (−11.42–29.54) on BFS and −5.43 % (−50.3–39.45) on M72 (Figure S4).

**Figure 7.** Mean percentage mortality and 95% confidence interval (CI) for (**A**) 24-h (M24); (**B**) 72-h (M72); and (**C**) blood feeding inhibition (BFI) for Interceptor® and Interceptor® G2 nets against pyrethroid-resistant *Anopheles arabiensis* with 100-rabbit (rabbit bait and density of 100 mosquitoes) and 50-membrane (Hemotek® membrane bait and density of 50 mosquitoes) in the WHO tunnel test. The red dashed line depicts the WHO minimum bio-efficacy thresholds of ≥80% M24 and ≥95% BFI.

**Table 7.** Superiority of Interceptor® G2 over Interceptor® was estimated by comparing the membrane assay to the gold standard assay with pyrethroid-resistant mosquitoes. The difference in the odds of mosquito at 72-h (M72) and blood feeding success (BFS) for resistant *Anopheles arabiensis* measuring the superiority of Interceptor® G2 and Interceptor® with the gold standard with 100-rabbit compared to 50-membrane bioassays \*.


\* For the gold standard, 100-mosquitoes with rabbit and 50-mosquito with 2 Hemotek® membrane feeders augmented with worn socks were used in the WHO tunnel bioassay, adjusted for net type and condition.
