Insecticide-based vector control is one of the main tools in the fight against malaria ¹ and so insecticide resistance monitoring is essential to inform operational decisions in vector-control. To be able to select appropriate vector-control interventions the insecticide resistance status of the local mosquito population needs to be well understood. Although the range of insecticide classes available for mosquito control is limited, a crucial element in the management of resistance is the ability to be able to switch between insecticide classes where resistance is seen to have evolved to a specific insecticide is observed. This approach helps address the issue of cross-resistance, ensuring that control measures remain effective against mosquito populations that have developed resistance to certain insecticides and particular modes of action (MoA) ² . Where a novel insecticide is introduced, it is important to be able to monitor for the emergence of resistance in the target mosquito population.

The CDC bottle bioassay, recommended by the Centres for Disease Control (CDC), was developed by Brogdon and McAllister in 1998 ³ . It is a resistance monitoring assay which looks at time-to-kill of a population exposed to a specified concentration of an insecticide, with knockdown measured 10–90 minutes after exposure, dependent upon the tested insecticide ^{4, 5} . It was developed as an additional means of monitoring resistance to the original World Health Organisation (WHO) tube bioassay, which measures mortality 1- and 24-hours post exposure.

The WHO method offers distinct advantages over the CDC method by standardizing test materials using centrally prepared, quality-assured impregnated papers, which ensures uniformity and reliability of results across various locations. This eliminates the inconsistencies associated with local preparation of bottle coatings. Moreover, the method minimizes hazardous exposure to insecticides for technicians by avoiding manual bottle coating processes. Additionally, the WHO method's logistical efficiency is enhanced through the use of easily transportable impregnated papers and plastic tubes, making it less cumbersome and more practical for field deployment compared to the more bulky and hazardous materials required by the CDC method ^{3, 6} . Data generated using the assay is widely used for the routine monitoring of resistance to commonly used insecticides in field populations and for the characterisation of laboratory populations of mosquito ⁵ .

The original CDC bottle bioassay test procedures ³ have been updated several times to reflect an evolving understanding of mosquito resistance phenotypes. Standardized guidelines published by the CDC offer detail on some parameters for conducting these assays, including recommendations on diagnostic dose (s), exposure time, and mosquito age ⁷ , but there are some experimental parameters in the method which are open to interpretation as to how exactly to perform the bioassay such as number of mosquitoes per bottle, the environmental conditions and orientation of the bottles for testing. This introduces the potential for variability in how the bioassay is conducted between testing sites, and this potential for inconsistency is borne out by differences in methods used between studies as reported in publications and elsewhere.

If these methodological differences affect the results obtained from the bottle assay, data may not be comparable between sites, time points or insecticides, which could impact the precision and repeatability of the evidence on which operational decisions are made. The age of mosquitoes when exposed to insecticides can have an impact on measured susceptibility, with a decline in resistance observed in older mosquitoes in comparison to younger ones ^{8– 10} . The number of mosquitoes per bottle and agitation of the bottle during testing, for example during scoring knockdown/mortality, could disrupt resting behaviour, impacting contact time with the active ingredients and subsequent insecticide exposure, as observed in the WHO tube bioassay ¹¹ . Bottles used for bioassays are typically coated manually either where the tests are conducted or at a collaborating institute before being shipped. This manual coating process might introduce extra variability compared to the World Health Organization (WHO) tube tests. In the WHO tube tests, insecticide-impregnated papers are produced and distributed from a central location, where they are also hand-treated ¹² . Interpretation of the bottle coating technique and how evenly coated insecticide is could vary between users, and prolonged or frequent contact with a surface that has not been evenly coated could impact mortality results. Environmental factors can also affect the resistant phenotype. Humidity has been shown to potentially increase insecticide-induced mortality in mosquitoes ¹³ . On the other hand, higher temperatures have been correlated with varied effects: an increase in susceptibility to insecticides in some instances, while in others, they have been associated with improved mosquito longevity ^{14– 16} . Therefore, it is important to conduct susceptibility bioassays consistently, controlling conditions as closely as possible and reporting in detail the methodological parameters and environmental conditions.

The development of the WHO bottle bioassay method ^{6, 17– 19} was influenced by the challenges encountered in establishing diagnostic concentrations (DCs) for new classes of insecticides, specifically the neonicotinoid clothianidin and the pyrrole chlorfenapyr. Susceptibility to these insecticides has been measured using various testing methods by separate groups. For instance, clothianidin has been tested by impregnating filter papers with SumiShield™ 50WG ²⁰ , and both insecticides have been tested in CDC bottle bioassays ²¹ . Comparative studies have used both CDC bottles and SumiShield™ 50WG impregnated filter papers side by side ²² . This variation in methodology, coupled with the use of different insecticidal products or technical grade insecticides, make it difficult to compare results from these early studies. Recognizing this challenge, the WHO included these and other chemistries of interest in their discriminating dose study, which has now led to their recommendation of the WHO bottle bioassay method for both insecticides ^{6, 18} . The CDC bottle bioassay similarly sought to overcome the issues of stability on papers by instead coating a glass surface with insecticide. The method measures the time to knockdown (or incapacitation) of mosquitoes exposed to a specified dose of insecticide, in contrast to the endpoint measured by the WHO bottle bioassay, mosquito mortality 24 hours after a 1-hour exposure. Measuring 24hr mortality allows closer comparison with susceptibility in the WHO tube assay and provides more consistency between methods across the WHO guidelines, streamlining testing procedures for country teams ⁶ . As novel insecticidal compounds continue to emerge, the WHO bottle bioassay will play a crucial role in monitoring susceptibility to ensure their effective deployment in public health initiatives ^{19, 21} .

To try to understand how consistently each bottle bioassay methodology is being applied, we set out to review the current literature for instances where the guidelines are being referenced, to see what data and information is being reported, and where there are gaps or inconsistencies in method reporting.

We then aimed to experimentally explore parameters of the bioassay we found not to be consistently reported between studies, to determine their impact on bioassay results, and where appropriate suggest tighter ranges of values for these parameters to minimize variability in data. We aimed to do this by looking for parameters of the guidelines which are open to interpretation and seeing if they can influence the results of the bioassay. In doing this we hope to suggest additional guidance on optimal performance of this bioassay and key information required for reporting of insecticide resistance data, thus proposing a model for more robust data generation and reporting using this methodology.

We performed a multifaceted study to enhance our understanding of the CDC and WHO bottle bioassay methods, common methods used to collect data on mosquito susceptibility to insecticides. Our research involved a) analysing published guidance to determine how specifics of the methodological details have changed through historical iterations, b) reviewing literature for bottle bioassay studies to understand the variation in the methods used and how they are reported, c) attempting to compare results between studies which reported testing results from the same combinations of mosquito strain and insecticide, and finally d) by conducting laboratory controlled experiments. The primary objectives of the experiments were to identify key experimental variables and understand the impact of changing parameters on the entomological endpoints of interest. This approach is valuable because the effects of these parameters on results are not well-understood, making it difficult to compare and interpret data from different studies effectively.

Reviewing published literature for studies which reported results from bottle bioassay experiments, either CDC or WHO bottle bioassays, identified 76 papers. Among these publications there was a lot of variability in the level of detail reported about the methods used for testing, but the general trend was for the specific methodology not to be reported in much detail at all. For 1 in 10 publications guidelines were not cited at all for these standard protocols. In many cases authors report not the actual method used, but rather quote the published guidance, for example reporting that ’20–30 mosquitoes’ were exposed per bottle as per the cited guidelines rather than reporting the real ‘n’. When details of testing parameters such as mosquito age or the number of mosquitoes exposed per bottle were reported the values varied between publications and even between the reported methods and the guidelines cited in the paper. The fact that guidelines have evolved, and specific details changed over time no doubt contributes to this confusion, as might the fact that some elements of the available guidelines are open to interpretation or details are not specified. For example, Brogdon and McAllister ³ describe the CDC bottle assay method with 25 mosquitoes per bottle. The WHO tube bioassay is then cited as more suited for lab work rather than fieldwork due to the difficulty in collecting large numbers of mosquitoes in the field and that the CDC bottle bioassay is cited as more practical, as might the fact that some elements of the available guidelines are open to interpretation or details are not specified. For example, Brogdon and McAllister (3,3) describe the CDC bottle assay method with 25 mosquitoes per bottle. The WHO tube bioassay is then cited as more suited for lab work rather than fieldwork due to the difficulty in collecting a large number of mosquitoes in the field and that the CDC bottle bioassay is cited as more practical as it allows for testing with as few as 5 mosquitoes.

Variability in reported methodological detail makes it difficult to compare or contrast results between published studies, even when the testing method used is the same. In this case the only studies which tested the same strain and insecticide combination and could be compared all reported over 90% mortality, unsurprising since this data consisted of insecticide-susceptible lab populations exposed to discriminating doses designed to identify insecticide-resistance. From this review it was therefore difficult to determine whether discrepancies in methodology interpretation between studies was influencing the mortality data generated when using this bioassay. Reporting of more granular discriminating times generated through analysis of time-response data would facilitate comparison between studies and strains.

Even when precisely the same method is used to test the same insecticide against the same mosquito population under consistent testing conditions, a bioassay will have some level of inherent variability. Our investigation set out to quantify the WHO bottle bioassay method’s variability under controlled conditions, yielding significant insights for understanding and conducting these bioassays. A previous study showed that variability within World Health Organization (WHO) bottle bioassays was consistent for specific insecticides like chlorfenapyr and clothianidin across various mosquito species. However, for other insecticides such as flupyradifurone and transfluthrin, the variability significantly increased. Notably, with flupyradifurone, the variability in mortality rates for Ae. aegypti was as high as 12%, compared to other insecticides where the mean variability remained below 10%. The study also highlighted that pyriproxyfen exhibited a within-assay variability of 2.5% for An. gambiae s.s. and 3.4% for An. stephensi, highlighting differences in response to insecticides among species ¹⁹ . As a result, it is critical to understand the inherent variability in bioassay results to perform meaningful power calculations and then to interpret results properly, i.e. determining what is a real difference between treatments or populations and what might be noise. It is also interesting to compare the level of variability between bioassays to help in selection of the appropriate method. Even before the planned experiments were performed, we found evidence of variability in results between studies or over time – the LC ₅₀ previously established in the same laboratory with the same protocol for the same mosquito strain did not hold true at the time of this study. This may be due to changes in rearing conditions over time, test material variability such as the insecticidal stock, which was previously synthesised by the manufacturer with higher purity, and although test concentrations remained the same its possible additional impurities impacted biological efficacy, or between cohorts of mosquitoes reared at different times or variability or ‘noise’ in data affecting the result returned by the log probit analysis of dose response data. Having re-established an LC ₅₀ we then controlled as many variables as possible in testing 9,130 mosquitoes across 365 bottles. Our results revealed an average mortality rate of 67.18%, with a confidence interval ranging from 38.29% to 84.47%. This study is notable for its scale compared to similar studies in existing literature, which often involve lab-reared mosquito colonies and the same insecticide in bottle bioassays. Importantly, to our knowledge, this is the first study of its kind to consistently control and monitor a range of factors that could influence the results over such an extended period within a single lab. Replicates where control mortality exceeded 20% were excluded from the analysis, and control mortality in the analysed data ranged between 0 and 20%. This decision underscores the nuanced understanding that, while mean control mortality provides a baseline for comparison, the variability within these controls—ranging theoretically from homogeneity to significant heterogeneity—can profoundly impact the interpretation of experimental outcomes. Such variability gives some insight into the potential influence of operator handling and experimental conditions, emphasizing the importance of examining not only mean mortality but also the distribution of outcomes across replicates to ensure the reliability and reproducibility of our findings.

Exploring the variability and excluding other explanatory variables, we observed a within-day variance that corresponds to a standard deviation of 0.331 (numeric value: 0.109). In contrast, the between-day variance was more pronounced, with a standard deviation pegged at 2.069 (numeric value: 4.282). Furthermore, the data underscores a significant aspect of long run precision, characterized by a Coefficient of Variation (CoV) standing at 0.69. It is essential to design studies with enough statistical power and relevant temporal spread to consider variations in data. Doing so would enable a more nuanced approach to detection of any changes in resistance, allowing for better-informed decisions in monitoring and managing resistance.

With regards to precision, our analyses showed that the inter-assay and intra-assay precision of the bottle bioassay method was much higher than the WHO tube method (unpublished data). This might be partially explained by the fact that the bottle method was less impacted by the number of mosquitoes per bottle, even though it was more susceptible to relative humidity changes. This difference in intra assay precision was reflected in lower mortality and higher mosquito dry weights. Though standard mosquito counts were used for bioassays conducted during this period. One specific observation from assays conducted in 2022 was a minor disparity in the mosquito count between bioassays set up by each of the two operators. On average, both operators aspirated around 25 mosquitoes per bottle. However, Operator 1 tended to aspirate slightly more than this average, whereas Operator 2 showed a tendency to aspirate slightly less. This trend, however, did not extend into the 2023 assays. In analysing the number of mosquitoes per bottle. There was a trend towards increased mortality with the highest numbers of mosquitoes tested, and a general positive correlation. Though this trend was not statistically significant it is supported by a previous study which demonstrated that the number of mosquitoes per test unit influences mortality and knockdown rates, with higher numbers resulting in greater insecticide exposure and effectiveness ¹¹ . However, it should be noted that upon further analysis with additional variables like wing length accounted for, the previously significant association of relative humidity with mortality disappeared (either as measured at the start (p=0.531) or the end (p=0.332) of the exposure period).

While initially differences were noticed there were no significant time or operator trends observed once core variables were accounted for, indicating that these bioassays were consistent in terms of these factors. Operator differences are not accounted for in the current guidance however it is encouraging to see that at least under controlled conditions of this study that this did not significantly impact the bioassay outcomes - this is a highly controlled study and so this may not be a surprise. We did observe a substantial difference between different 'experimental periods,' which highlighted the need for careful consideration of temporal factors in planning and interpreting these bioassays. Mosquitoes reared for the bioassays we conducted in the summer of 2022 were larger measured as dry weight than other testing periods, and wing lengths were on average about 10% bigger. Our modelling indicated that this was a key predictor of reduced mortality, though our analysis would only predict mortality to be about 5–10% lower and in fact the mortality showed a significant 52% drop from 73%±5% in Spring 2022 to 20%±4% in Summer 2023 before returning to 67%%±5% in Spring 2023. Additionally control mortality during this period was also lower. It is unclear why the mosquitoes during this period showed such an increase in size, but it could be due to subtle differences in rearing conditions.

By controlling as many variables as possible, we were able to examine the effect of changes in some parameters which varied beyond our control, to find those which were most influential, and which should therefore be controlled as far as possible in standardised protocols. Where they cannot be controlled, they should be recorded and used in interpreting the test data. By analysing the data and using the meta-data generated in this study, we were able to examine the effect of four parameters on observed mortality: testing conditions (temperature and relative humidity), age of mosquitoes, number of mosquitoes exposed per bottle and size of mosquito.

There is a large body of evidence of the impact of temperature on mosquito however relative humidity less often directly examined. Multiple Anopheles mosquito species have been shown to be less susceptible to pyrethroids at higher temperatures ³⁵ , which could be attributed to enhanced enzymatic activity in the mosquitoes at elevated temperatures leading to an increased rate of insecticide detoxification. Temperatures below the recommended 27 ± 2°C have been seen to influence the tolerance of An. arabiensis and An. funestus to both bendiocarb and deltamethrin ³⁶ . In a Ugandan field insectary without environmental controls, but with monitored temperature and humidity, there was a pronounced and statistically significant decrease in the mortality of A. gambiae mosquitoes as humidity levels rose ³⁷ , though others have seen a high degree of tolerance to a wide range of relative humidity (RH) levels in Anopheles, with their survival not significantly impacted between 60% to 100% RH ¹³ , though RH levels below 10% prove to be lethal within hours. Some studies suggest that mosquitoes, especially those from arid regions, might develop resistance to desiccation, potentially surviving low relative humidity conditions slightly longer than laboratory-bred counterparts ³⁸ . Even if relative humidity has a limited impact on bioassay outcomes it is plausible that a reduction in longevity due to low relative humidity would compound the effects of bioassay mortality and mosquitoes may appear less resistant to insecticides. Temperature and RH were not reported in the published studies we reviewed, and so the recent addition of a RH range to the new CDC guidance for the bottle bioassay method is welcome ⁵ . In situations where environmental conditions cannot be fully controlled, it is crucial to report the environmental conditions in detail to ensure that the data can be correctly interpreted.

In the published literature most studies reported testing with mosquitoes 2–5- or 3–5-day old. The recommendation of 2–5-days was introduced in the CDC 2023 guidance; however, this age range was potentially commonly used prior to this due to guidelines already in place for the WHO cone test and the WHO tube bioassay ^{17, 34} . Studies which reported mosquitoes in the testing age range of 3–5-days all dated from 2015 onwards. WHO 2013 guidance was published 2 years prior to this recommending this age range, which could explain this shift. The wider age range category used in the CDC guidance could be due to the challenges of collecting field mosquitoes within a small window of time. Mosquito age has been shown to have an impact on resistance status, with larvae and adults that are older having increased susceptibility ³⁹ . It is therefore important to have consistency between published guidelines to ensure standardisation of testing ages and comparability of data.

In terms of the number of mosquitoes per bottle used in the bioassays, insecticide delivery relies on mosquitoes contacting the treated surface of the bottle and the quantity of mosquitoes per bottle could influence resting and flight behaviour. An increase in flight behaviour could cause some mosquitoes to lose contact with the insecticide before receiving a lethal dose resulting in fluctuations in the mortality data generated. We found that there was no significant relationship with either knockdown or 24-hour mortality within the studied range of 20–30 mosquitoes, though a trend for reduced mortality below 20 and increased above 30. This possible reduction in mortality below 20 is in agreement with previous work which has shown a decrease in mortality in the WHO tube bioassay with number of mosquitoes ¹¹ . In the 2022 assays, there was a slight difference in the average number of mosquitoes added to each bottle by the two operators, with one operator tending to use numbers slightly above 25 and the second slightly lower, though in the 2023 assays no such difference was observed. The Brogdon and McAllister (1998) CDC bottle method states that 5 or fewer mosquitoes can be tested, but only a small number (3%) of studies we reviewed reported using 10 or fewer mosquitoes per bottle ³ . Although no justification was provided for the low number, the mosquitoes tested were F ₁ field mosquitoes and obtaining sufficient field-derived mosquitoes can be challenging. Indeed, Brogdon and McAllister ³ suggest that results from multiple small-scale tests can be combined, though they do not provide data on the effect of numbers this low on mortality. More specific guidance on the quantity of mosquitoes exposed per bottle would help consistency between studies. As well as the quantity of mosquitoes per bottle, agitation of the bottle during testing could also have an impact on resting and flight behaviour. The published CDC guidance documents all suggest gently rotating the bottle while scoring mortality ^{3, 5, 7} , though only 4% of published studies mentioned bottle rotation. Inconsistent agitation of bottles may create variability in results so either clearly specifying a suggested frequency for agitation or reporting methods accurately in studies would be beneficial.

Many studies have shown that mosquito body size is inversely related to insecticide susceptibility ⁴⁰ , often by artificially altering mosquito size by adjusting larval nutrition. Our study did not artificially adjust the mosquito size but instead allowed it to vary naturally. Although the size range is small because of standardised rearing protocols it still adds to this body of evidence on the impact of size on susceptibility ²⁵ . We measured both dry weight and wing length as a proxy for body size and found that with every 1 standard deviation increment in dry weight (from an average of 0.48mg to 0.54mg) there was an associated 13.35% reduction in mortality, and an increase in average of wing length from 2.94mm to 3.00mm led to a 4.54% reduction in mortality. Tight control of rearing conditions when generating cohorts of mosquitoes for testing is critical to ensure minimal variance in body size ²⁵ , and should ideally be monitored through a quality control process clearly defining minimum and maximum size thresholds for testing, using an appropriate size assessment methodology (be it wing length or dry weight) that is powered to detect significant differences in your test population over time. When testing field-derived mosquitoes it is important to record size alongside test results so that seasonal differences, for example, can be considered when interpreting results.

In addition to these four key parameters, there are some others which were not investigated during this study, but which may influence the measured endpoint. Physiological parameters of the mosquitoes other than size and age might be a source of variation and should be either kept consistent where possible or reported alongside data. All bioassays in this study were conducted during the same time period in the afternoon to minimize variability due to circadian differences in mosquitoes. However, time of day could influence mosquito activity and is a parameter worth investigating, particularly for insecticides with target sites or modes of action affected by metabolic activity of the mosquito. The current CDC 2023 guidelines specify testing no further than F ₂ generations in bottle bioassays ⁵ , and CDC guidelines are to report the physiological status of the mosquitoes i.e., unfed, blood fed, semi-gravid, gravid on the results sheet.

As well as characteristics of the mosquitoes, testing parameters may also affect results. The CDC guidelines suggest that the orientation of bottles during exposure does not significantly impact results, provided that the method is consistently applied though we know of no experimental evidence. The orientation of a bottle during exposure could potentially influence the behaviour of mosquitoes within the bioassay, an idea supported by evidence from the WHO cone bioassay where angle of testing affects contact with a treated surface ^{34, 41} . The World Health Organization (WHO) cone bioassay plays an integral role in the evaluation of the efficacy of long-lasting insecticidal nets as well as insecticides used in indoor residual spraying. The test is used on a variety of treated substrates, such as pieces of bed nets, mud, cement and wood. The cone setup assumes a wide variety of angles under different settings in which it is applied. However, the guidelines provided for the performance of the assay do not specify the angle at which the test must be performed ^{34, 41} . We are also unaware of any validation of show how evenly bottles are coated with insecticides using the methodology outlined in the CDC guidelines (2010 and 2023). It is possible that temperature and humidity could impact this drying process especially when surfactants are involved. Issues were seen with control mortality across sites within the WHO multicentre study when using RME ¹⁸ . Mosquitoes show avoidance behaviour and irritancy to insecticides and mosquitoes may choose to rest in areas less coated with insecticide ^{42, 43} impacting insecticide uptake and in turn mortality. It would be interesting to explore behavioural resistance, mosquito behaviour that enables reduced contact with insecticides ⁶ . The period which mosquitoes are held for after exposure for mortality scoring was not reported in all reviewed publications. Although the relevant holding time may be chemistry specific, standardising and reporting holding time is important to control for the mosquito recovery period, particularly in the case of metabolic resistance as the mosquito detoxifies the insecticide over time. The longer the holding period the greater the variability which might be introduced by factors such as desiccation of mosquitoes, or death of mosquitoes from spilling or drying out of sugar solution.

Upon reviewing the literature for publications reporting results of bottle bioassays there was general inconsistency in the precise methods used for testing and in the way that methodologies were reported and the relationship between reported methods and published guidance documents which were cited. Some publications incorrectly cited the CDC guidelines, an issue which could be overcome by a clear referencing format for the CDC 2023 guidance being made available. The specific methodology was often not reported in detail or the parameters which were documented varied between studies, making it impossible to compare results between published studies. These findings highlight the importance of having a standardised set of guidelines to follow for conducting the assay consistently and reporting. Since we have shown experimentally that changes to testing parameters can have an influence on the entomological endpoint, it is important that precise methodological detail is published with results, and with reference to published guidelines as appropriate.

Our experiments investigating the inherent variability in results from the WHO bottle bioassay method has underscored the significant impact of variables like mosquito dry weight, relative humidity, and mosquito count on the entomological endpoint (24h mortality). Notably, we found a substantial negative correlation between wing length and mortality, indicating that larger mosquitoes were less likely to die in the bioassay (p<0.001). This finding points to the necessity of rigorous quality control and standardized rearing practices for laboratory mosquito strains. Additionally, it emphasizes the need for detailed documentation of metadata related to mosquitoes collected from the field. We discovered that the dry weight of mosquitoes and relative humidity were core predictors of outcomes in the bioassay. In particular, the starting relative humidity was a slightly better predictor than the relative humidity at the end of the experiment, though both were almost interchangeable in their predictive power. Reports of experimental data should include specifics about mosquito collection, testing, and rearing methods, as well as information about the conditions under which bioassays are conducted, like temperature and humidity. Such measures will enhance the clarity and consistency of bioassay data interpretation. Our study demonstrates how minor variations in experimental parameters can introduce significant variability into bioassay outcomes, even under tightly controlled conditions. The validation of methods to characterise and generate a methods claim is an important part of being confident that the method selected to address a given question is appropriate and that the data that is generated is reliable. Part of a formal validation will include investigating the effect of altering testing parameters and characterising and minimising variability ⁴⁴ . Where relevant, generation of consensus SOPs which are developed and agreed by key practitioners ⁴⁵ can help with the uptake of standardised methods which aid interpretation of results between studies.

The inclusion of technical and biological replication in resistance monitoring testing is vital for detecting and accounting for these variances in bioassay results, so that data can be interpreted meaningfully, strengthening evidence-based decision making in vector control. Currently CDC guideline recommendations of 100 per test concentration does not seem sufficient considering this level of variation in this method and it is possible that this guidance needs updating ⁵ . Another factor that should not be ignored are the differences in endpoint which are observed between testing days. Our analysis of data collected over repeated testing sessions highlights the importance of spreading testing out over more than a single day to be able to get a more representative data set and allow you to be more confident in your conclusion about the presence or absence of resistance in a mosquito population. Currently there is no guidance around spreading testing over numerous days, but this kind of could allow for smaller tests to be conducted over multiple days which could be more feasible in settings where large numbers of testing age mosquitoes are not necessarily available on the same day. We recognise that we only studied variability and testing parameters with a pyrethroid, and interesting further work might be to repeat elements of this study with other insecticides either using currently used pyrethroid insecticides such as alpha-cypermethrin or deltamethrin, or insecticides from other classes such as chlorfenapyr or clothianidin for which susceptibility is also assessed using bottle bioassay.

Generating more robust and consistent susceptibility data using the bottle bioassay is possible through tightening up the methodology to minimise variability and recording meta data alongside bioassay results to allow more informed interpretation. More accurate reporting of the precise methods used to generate bioassay data will facilitate better comparability of data sets, which should include reference to relevant published guidelines and reporting of raw data and actual rather than target values for parameters such as number of mosquitoes exposed per bottle or temperature in the testing room. Generation and reporting of better data in these ways will facilitate better decision-making in designing vector control strategies, contributing to more effective efforts to combat mosquito-borne diseases.