**4. Discussion**

It is clear from the details outlined in Table 2 that the WHO susceptibility bioassay has undergone numerous updates to its methodology since its inception. While at each stage these updates have been relatively minor, it is still possible that these could impact bioassay results, and so it is important to ensure that the most recent iteration of the test procedures is followed and referenced. However, the literature review shows that this is not always the case.

The methodological variability between the published test procedures and the way these test procedures were historically presented on the WHO website leaves the WHO tube assay for insecticide susceptibility in mosquitoes open to interpretation as to how to perform the bioassay, as well as being unclear as to what the most up to date iterations of the guidelines are. The WHO website has been updated since this review of the method began, and the relevant test procedures can now be found considerably easier (https://www.who.int/teams/global-malaria-programme/prevention/vector-control/ insecticide-resistance [Accessed: 4 February 2022]). Moreover, the lack of comparable data from the published literature is due to the populations being tested being either field strains

of unknown resistance status or a susceptible laboratory strain. For the few publications where the same resistant laboratory strains were able to be compared, the mortality data was wildly different. This could be because the same strains held in different labs might in fact be vastly different from each other. This could be because of a whole host of reasons, including laboratory adaptations, contamination, selection pressure applied rearing conditions, genetic bottlenecks, and genetic drift. As a result, to optimize this bioassay, we planned to investigate the effect of mosquito age in the range of 5–10 days old, as well as the number of mosquitoes per test unit and the use of cardboard discs to cover the mesh of the exposure tube during the test, which is specified in the most recent iteration of the test procedures. These factors were chosen for investigation, as there is a lack of published literature investigating their effect on the outcome of this bioassay.

The susceptible *A. gambiae* strain Kisumu was exposed to permethrin-treated papers of a concentration expected to provide moderate mortality in an experiment to explore the effect of varying the parameters of interest when conducting the WHO tube assay. Based on three replicate tests, there was no evidence that covering the top of the exposure tubes with a cardboard disc during the exposure period had any impact on either 1 h knockdown or 24 h mortality. The rationale for the covering of the exposure tubes using in the test procedures is that it will prevent light entering through the mesh and so should discourage mosquitoes from resting on the upper mesh of the test units during exposure, which reduces their contact with the insecticide. It was not possible to assess if there was a reduction in resting on the mesh, as it was not possible to observe mosquito behavior during the exposure period, as the exposure chamber was covered by the insecticide-treated filter paper and the cardboard disc. However, due to the lack of significant difference in mortality seen in this study, we would sugges<sup>t</sup> that this step appears to be unnecessary. So long as all test units are treated the same in terms of lighting, mosquitoes resting on the mesh should be consistent between test units and therefore there should have no impact on the final mortality scoring.

When varying the number of mosquitoes per test unit, mortality in this same experimental set up was unaffected by mosquito numbers between 15 and 30 mosquitoes per test unit. However, when only 10 mosquitoes were added per test unit, the 24 h mortality was significantly lower (7% compared to 50%). The same trend is not seen in the proportion of mosquitoes knocked down immediately post exposure, with knockdown being reduced in treatments with 15 and 20 mosquitoes per test unit compared to covered and uncovered treatments containing 25 or 30 mosquitoes. Knockdown thus appears to be positively correlated with the number of mosquitoes per test unit in this laboratory strain. This implies that mosquitoes are being differentially exposed during the bioassay, depending on the number of individuals within a single test unit. It is possible that, when using 10 mosquitoes in a test unit, there is enough space for all or most of the mosquitoes to rest on the door at the base or at the mesh at the top of the test unit and therefore avoid contact with the insecticide-treated paper. With more mosquitoes, there is more opportunity for this free flight to disturb resting mosquitoes within the bioassay and cause them to fly and resettle in a different part of the exposure chamber. This could then force the mosquitoes which were previously resting on a non-insecticide-treated surface to encounter the insecticide-treated filter paper instead and become intoxicated with the insecticide. The more mosquitoes in a test unit, the more likely this disturbance is to occur and, in turn, the more likely a mosquito is to become intoxicated with insecticide through more frequent contacts. We recommend that at least 15 mosquitoes are included per test unit when conducting WHO tube assays, and that, where knockdown is the entomological endpoint of interest, the number of mosquitoes per test unit is held constant between replicates and treatments. As well as this, due to the general increase in mortality seen with the addition of more mosquitoes, we would not recommend exceeding 30 mosquitoes per test unit. It is also worth noting that there was a substantial divergence in the mortality estimate for biological replicate test compared to the other three replicates. Despite the quality control measures in place, this may be due to some difference in the cohort of mosquitoes being used. It does

highlight the variability that can be introduced into the bioassay data by minor changes in parameters, even in highly controlled conditions. Such a divergent bioassay result could easily go unnoticed in the absence of technical and biological replicates.

Since this data was generated from three biological replicates with the same number of technical replicates per treatment, the sample size for each treatment differs, with three times as many mosquitoes tested in the 30-mosquitoes-per-test-unit treatment as in the treatment with 10 mosquitoes. To account for this difference, an additional biological replicate was carried out to equalize the sample size for each treatment to ~180. With this additional replication, the trends seen did not change. The only significant change was that the 24 h mortality for the 10-per-test-unit treatment increased from 7 to 15%. The variability of the results within each treatment was either unaffected or reduced with increased replication, showing the value of maximizing both test unit replicates and mosquito 'n' values when generating data using the WHO tube assay. For the Tiassalé 13 strain in the 15-per-test-unit treatment, the variability was significantly higher (Figure 7); however, this is due to the large intraspecific variation between the technical replicates in one of the biological replicates. It is possible that using more test units with fewer mosquitoes could result in a lower variability than fewer test units with larger numbers, and so, in settings where mosquitoes are less available, it may be beneficial to divide the cohort up into multiple smaller batches with more test units. Where possible, we would recommend increased replication with different cohorts of the same mosquito population to increase the sample size to at least the WHO recommended 100 mosquitoes per insecticide treatment and 50 mosquitoes per control. We also would not recommend using any more than 30 mosquitoes per test unit, as there is no data available, that we are aware of, to support it, and we can see no logistical reason for using more than this number per test unit.

Repeating the experiment with the insecticide-resistant Tiassalé 13 exposed to permethrintreated DC papers with results based on three replicates showed similar trends to those using Kisumu. Covering the exposure tube during the period of exposure again had no significant effect on knockdown or mortality relative to the uncovered test units. Mosquito numbers between 15 and 30 mosquitoes per test unit did not affect knockdown or mortality, but again the treatment with only 10 mosquitoes per test unit resulted in lower mortality. The effect size of the number of mosquitoes per test unit seemed to be smaller than in the susceptible Kisumu strain, and a larger sample size was needed to be able to detect a difference. Further replicates of the experiment to ensure equal '*n*' values for the number of mosquitoes tested per treatment did not affect the trends of results; the variability was reduced, though not to the same extent as it was in the Kisumu experiment. This replicated result, even with a DC assay with insecticide-resistant mosquitoes, which are more field relevant than an old laboratory colony, supports the recommendation to use a minimum of 15 mosquitoes per test unit but that covering the exposure tube does not have an effect.

When investigating the effect of mosquito age at the time of testing, we found that mosquitoes both 2 and 4 days older than the recommended testing age (2–5 days) show an increased susceptibility to permethrin. This increased susceptibility is seen at 6–9 days old for Kisumu and 4–7 days old for Tiassalé 13. This difference could be due to the increased fitness cost caused by resistance mechanisms in the Tiassalé 13 strain compared with the susceptible Kisumu strain. This supports previous findings that mosquitoes aged 10 days and above show an increased susceptibility to insecticides [8,9], but there has previously been little data on mosquitoes aged between 5 and 10 days post-eclosion. While investigations of the effect of insecticides on malaria transmission focus on older female *Anopheles* which are the vectors of malaria, it is useful for other testing purposes to know across what age range mosquitoes can be used for testing and still produce the same result. When monitoring a population for the emergence of resistance, it is important that variables including ag, and mosquito density, as discussed above, are held constant to allow robust comparisons between test replicates and to allow true changes in test results over time to be identified. The data from this study sugges<sup>t</sup> that, when performing a WHO tube assay, the recommended testing age of 2–5 days should be adhered to.

Intertest variability (between biological replicates) was generally larger than or similar to intra-test variation (between test units within a biological replicate) for both strains. It is possible that using more test units with fewer mosquitoes could result in a lower variability than fewer test units with larger numbers, and so, in settings where mosquitoes are less available, it may be beneficial to divide the cohort up into multiple smaller batches with more test units, within the limits of 15–30 mosquitoes discussed above.
