Community effectiveness of pyriproxyfen as a dengue vector control method: A systematic review

Background Vector control is the only widely utilised method for primary prevention and control of dengue. The use of pyriproxyfen may be promising, and autodissemination approach may reach hard to reach breeding places. It offers a unique mode of action (juvenile hormone mimic) and as an additional tool for the management of insecticide resistance among Aedes vectors. However, evidence of efficacy and community effectiveness (CE) remains limited. Objective The aim of this systematic review is to compile and analyse the existing literature for evidence on the CE of pyriproxyfen as a vector control method for reducing Ae. aegypti and Ae. albopictus populations and thereby human dengue transmission. Methods Systematic search of PubMed, Embase, Lilacs, Cochrane library, WHOLIS, Web of Science, Google Scholar as well as reference lists of all identified studies. Removal of duplicates, screening of abstracts and assessment for eligibility of the remaining studies followed. Relevant data were extracted, and a quality assessment conducted. Results were classified into four main categories of how pyriproxyfen was applied: - 1) container treatment, 2) fumigation, 3) auto-dissemination or 4) combination treatments,–and analysed with a view to their public health implication. Results Out of 745 studies 17 studies were identified that fulfilled all eligibility criteria. The results show that pyriproxyfen can be effective in reducing the numbers of Aedes spp. immatures with different methods of application when targeting their main breeding sites. However, the combination of pyriproxyfen with a second product increases efficacy and/or persistence of the intervention and may also slow down the development of insecticide resistance. Open questions concern concentration and frequency of application in the various treatments. Area-wide ultra-low volume treatment with pyriproxyfen currently lacks evidence and cannot be recommended. Community participation and acceptance has not consistently been successful and needs to be further assessed. While all studies measured entomological endpoints, only two studies measured the reduction in human dengue cases, with inconclusive results. Conclusions Although pyriproxyfen is highly effective in controlling the immature stages of dengue transmitting mosquitoes, and–to a smaller degree–adult mosquitoes, there is weak evidence for a reduction of human dengue cases. More well designed larger studies with appropriate standardised outcome measures are needed before pyriproxyfen is incorporated in routine vector control programmes. Additionally, resistance to pyriproxyfen has been reported and needs investigation.

Introduction intervention. Larvicides have most often been implemented against Ae. aegypti as it breeds almost exclusively in domestic water containers. Ae. albopictus uses both, artificial and natural breeding sites [7] and is therefore more difficult to tackle.
Pyriproxyfen is an insect growth regulator (IGR) with a slow-acting larvicidal activity against a broad spectrum of public health insect pests [8] and it is being used extensively worldwide both in public and private settings. Acting on the endocrine system of insects by mimicking the juvenile hormone, pyriproxyfen hinders molting and subsequently inhibits reproduction. In addition, it causes morphological and functional aberrations in emerging adults, such as decreased fecundity and fertility. Due to its very low mammalian toxicity [9], pyriproxyfen is approved by the World Health Organization (WHO) for the treatment of potable water against mosquitoes [10].
Pyriproxyfen has been studied extensively in experimental, i.e. controlled laboratory or semi-field settings, with evidence of efficacy against immature Aedes spp. [11,12]. Yet, in field application studies, pyriproxyfen demonstrated mixed outcomes regarding its effectiveness as well as persistence. To date, no systematic review of the scientific literature has been undertaken to examine the evidence for the effectiveness of pyriproxyfen against dengue vectors. Therefore, the objective of this study is to review systematically the available literature for evidence on the community effectiveness (CE) of pyriproxyfen as a vector control method reducing Ae. aegypti and Ae. albopictus populations and dengue transmission.

Methods
This review follows the reporting guidelines set forth in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) Statement for systematic reviews and metaanalyses [13] (Fig 1).

Inclusion and exclusion criteria
The following inclusion criteria were used: 1) Studies providing original research dealing with the CE of pyriproxyfen-alone or in combination with other chemical vector control products. 2) As for study types, included were any cRCTs or randomised controlled trials (RCT); non-RCTs (nRCT) only if they were relevant to the research question and using a control, e.g. quasi-randomised controlled trials (quasi-RCTs), intervention control trials, controlled before and after studies. Unlike in RCTs, allocation in quasi-RCTs is performed in a way that is open to systematic bias, i.e. chances of being in one group or another are not equal. In cRCTs, preexisting groups of participants are allocated to (or against) an intervention. Intervention control studies use methods designed to examine efficacy or effectiveness of an intervention in a group but do not use randomisation. 3) Any study that applied pyriproxyfen in the fielddefined as any community or environment where dengue vectors naturally occur-was considered CE and included in the analyses.
Efficacy studies, defined as trials performed under laboratory conditions were excluded. Of the studies that undertook both methods, only the CE component was considered. Inclusion criteria included the above-mentioned study types, to give a broader picture of existing studies, since vector-control studies have varying designs and information may be useful.

Search strategy
Two researchers independently carried out the literature search until 01 August 2016 with no starting time limit. The search was conducted in English, but articles were not excluded if the full text was not available in English. The search strategy was applied to the following seven were: Dengue, Dengue haemorrhagic fever, Dengue shock syndrome, along with the abbreviations DF (dengue fever), DHF (dengue haemorrhagic fever), DS (dengue syndrome), and DSS (dengue shock syndrome); for the vector category: Aedes, Aedes aegypti, Aedes albopictus; and for the intervention category: pyriproxifen, and insect growth regulator.
Once screened for duplicates by author, title, journal and publication date, eligible studies were screened against the inclusion criteria. At first, titles and abstracts were screened by two independent reviewers, in a second step articles were reviewed in full and relevant information was extracted into the evidence table (Table 1). A third reviewer was available for the potential case of disagreement between the two independent reviewers. Studies were assessed for quality using the assessment of multiple system reviews (AMSTAR) [14]. For the purpose of this review, the overall quality of studies was not used to exclude studies, but as a tool for evaluating the impact of the reported outcomes.
A comparative analysis of the main study outcomes was conducted, using the quality of each individual study as a weighing tool. The use of randomisation, the calculation of sample sizes and the size of the unit of allocation all impacted the weight individual studies were assigned. Finally, analytical categories were developed based on the method of pyriproxyfen application. These categories were: 1) container treatment, 2) fumigation, 3) auto-dissemination, and 4) combination of pyriproxyfen with adulticides. As for 1), 'container treatment' is for the purpose of this review defined as any intervention performed by using any kind of Aedes spp. infested containers ( Table 2).

Results
745 articles were identified from the different databases for assessment (Fig 1). After removing duplicates (n = 698), 47 were left for closer analysis. An additional two studies were identified from the reference lists of the above. Applying full inclusion and exclusion criteria on these 49 studies, 17 met the pre-specified eligibility criteria. Eight of the studies were classified as CE studies, while nine reported both efficacy as well as CE components.
Common reasons for the exclusion of potentially relevant studies included the following: 26 studies reported only efficacy; one study was a meta-analysis without original data [15]; and for one study [16] the pyriproxyfen component of the intervention was not accepted by the community and therefore not analysed. Another CE study [17] had multiple study arms, but the arm in which pyriproxyfen was tested against Aedes was a simple intervention design without a control and was therefore excluded. Similarly, another study [18] described the use of pyriproxyfen against Aedes only as a simple intervention and did not use it in its RCT part. One proved to be a simple intervention [19] and one combined efficacy and CE study [20] was excluded as it did not use pyriproxyfen in its CE part.
Information on potential confounding factors such as the socio-economic status of residents [23,25,28,31,33] or housing construction [22,28] was not systematically reported. Weather conditions, either historical or during the intervention period, were reported in six of the studies [26,28,31,32,33,35]. No study incorporated a specific economic analysis or provided cost estimates for consideration. The most common study design was nRCT, used in 12 studies. Three were RCTs, one quasi-RCT and one cRCT. Regarding the vectors, Ae. albopictus alone was studied in two of the included studies [33,36], one study [32] tackled both, Ae. aegypti and Ae. albopictus, both laboratory-reared. The remaining 14 studies looked at Ae. aegypti only.
Among the studies with both CE and efficacy arms, three used pyriproxifen either in combination with spinosad [34] or temephos [29] or compared its effects with other vector control methods (temephos, diflubenzuron, Bti, spinosad; [35]).
Sample sizes and units of allocation. Various methods of pyriproxyfen application were used and information on sampling and reporting of sample size was often limited, making comparisons between the studies difficult. The largest study included communities with a total of 97,262 people [25], followed by 64,591 people [21]. Both studies were conducted in urban settings, facilitating a high coverage. Tsunoda [31] performed their intervention in a town of 12,000 inhabitants, while Marcombe [35] and Darriet [34] chose smaller settings, described as 'three separate villages', though the exact size population-wise is not specified. Two studies [27,30] were conducted in rural areas and reported a sample size of two villages (65/71 houses with 160/171 inhabitants) and 16 households, respectively. Likewise, [29] was undertaken in four villages in Thailand.
The included studies also differed regarding the scale of the intervention; and, again, the reporting of this information varied widely. Sizes and makings of containers were most commonly reported as units of allocation. Others treated defined surface areas with different formulations and concentrations of pyriproxyfen. Some had more than one unit of allocation depending on method and study design. The smallest intervention was a cage with 50 free-flying Aedes mosquitoes [37], followed by [24] with 3 x 200 liter tanks and [34] with 15 containers of various sizes. The largest intervention [25] comprised the catch basins of a whole town (32.224 houses).
The follow up-periods-defined as the total intervention period from the first intervention day to the last day at which the endpoints were measured and reported-varied greatly. They ranged from 12 days [36] to nine months [35] with a median follow-up time of five months. Outcome measures. The reported outcome measures differed depending on the respective research question and intervention ( Table 3). The most widely used outcome measure, reported by 11 of the 17 studies, was per cent adult emergence inhibition (EI) [21,22,23,24,26,28,32,33,34,35,37].
Only two studies linked human transmission data. One study [31] used serological surveys (Immunoglobulin M), Ocampo [25] measured dengue case incidence, drawn from the national reporting system.
In six of the studies [21,23,24,28,32,35], pyriproxyfen had a significant (82-100% EI) and long lasting (up to 8 months) effect. However, the two RCTs reported less positive results with significant effects for 4 weeks only. However, it must be considered that one RCT [25] measured catch basin positivity, making comparison with the other studies difficult. It also had several additional constraints such as moderate Aedes indices, low pupae/person indices from the start, inability to reach the intended sample size, and the emergence of a dengue epidemic during the intervention. The other RCT [35] and one nRCT [34] reported treatment persistence of only 4 weeks, though initial larval densities were significantly reduced.
One study [21] found a variable larvicidal effect in 20 containers and attributed this to larvicide dilution, though none of the other container studies reported efficacy limitations with increasing container sizes. Notably, Sihuincha [27] treated different sizes of water tanks successfully for five months despite a high turnover of the treated water.
Regarding area-wide ULV application, two studies [23,32] recommend this method of application over a wide range of container sizes, while one [33] found it not suitable for either larval habitat treatment or auto-dissemination. The two studies that did report good results with ULV application [23,32] used pyriproxyfen combined with permethrin.
As for persistence in container treatment, slow-release formulations had the best results. Sihuincha [27] reported a mortality rate >80% over five months with application by gauze bags, Seng [28] reported similar effects over eight months with resin strands and Seccacini [24] reported 100% EI at six months using pyriproxifen-impregnated O-rings. A shorter persistence of four weeks was reported from a large field trial with granules in water in Martinique [35], while the longest duration of effective EI through ULV treatment was 35 days [23], and the shortest only two weeks [32].

Analysis of fumigation studies
Two studies [21,22] examined the use of fumigant canisters. Both used a combination of pyriproxyfen and permethrin and reported a significant inhibitory effect on adult emergence of Ae. aegypti as well as on BI. However, the size of the effect and persistence (less than 9 weeks) was limited. Adding outdoor ULV application of permethrin increased the effectiveness of the intervention.

Analysis of auto-dissemination studies
Six studies evaluated auto-dissemination. The only RCT examining auto-dissemination [30] found significant results only in their BGS trap counts. However, these are known to be the most sensitive when counts are low from the start [39]. The remaining studies [26,33,36,37], using intervention control designs, primarily demonstrated that this approach is efficacious and that it can be applied easily and at low costs. Caputo [36] performed field experiments with wild Ae. albopictus and reported an inhomogeneous product transfer to different sentinel sites. However, it was rightly pointed out that if the approach is applied to reduce Ae. albopictus adult densities, the mosquitoes themselves will disseminate the larvicide to the most attractive (i.e. most productive) natural breeding sites.
Devine [26] used 0.5% pyriproxyfen concentration in 1 l plastic containers and demonstrated overall reductions of Ae. aegypti adult emergence of 49-84%, as opposed to 7-8% in controls. In another experiment [36] with the same pyriproxyfen concentration, the overall reduction of Ae. albopictus adult emergence was 20.8%, as opposed to 2.4% mortality in controls. Given that the LC50 reported for Ae. albopictus (0,11 ppb; [39]) is about 10 times higher than that reported for Ae. aegypti, this is a promising result.

Analysis of combination product studies
Seven of the reviewed studies combined pyriproxyfen with interventions targeting adults as well. They did not stratify the effects by product but rather as combined results. Five studies evaluated pyriproxyfen in combination with permethrin [21,22,23,29,31], one in combination with the fungus B. bassiana [35], and one with Spinosad [34].
Two of these studies were RCTs [22,31]. Both found significant changes through their respective interventions. While both used a combination of pyriproxyfen with permethrin, their application differed (fumigation versus EcoBio-block S) and so did their measured outcome values (BI versus CI and seroprevalence studies). Direct comparison is difficult but it can be summarised that they had good results that lasted for 9 weeks (fumigation plus ULV) and 5 months (EcoBio-block S), respectively. Similar results were reported in the non-RCTs.
In summary, the combination of adulticidal and larvicidal products can increase effectiveness by simultaneously controlling adults and larvae, and by expanding persistence [22].

Resistance and combination treatments
Only one study found and reported on resistance of pyriproxyfen. Marcombe [35] demonstrated Ae. aegypti being tolerant against pyriproxyfen presumably due to cross-resistance with temephos, as pyriproxyfen had never been used on the island of Martinique before.

Community participation and acceptance
Three of the included studies [22,25,28] described community perceptions and uptake of the interventions. Seng [26;Cambodia] reported overwhelmingly positive perceptions after an initial period of concern. Harburguer [22; Argentina] found the community capable and ready to participate in a mosquito control programme by using non-professional control tools (fumigant tablet). However, the authors also highlight the community's reluctance to take part in training workshops, even though most applied the tablet while only 16% attended the workshop. The third study ( [25]; Colombia) describes the importance of engaging and empowering local field staff regarding design and operation of entomological surveillance activities.

Discussion
This systematic review of the CE of the juvenile hormone mimic pyriproxyfen against Aedes spp. presents evidence suggesting that pyriproxyfen can effectively control the adult emergence of immature stages of dengue vector mosquitoes in a variety of real world habitats. If the most productive breeding sites are identifiable and accessible-e.g. catch basins, water storage containers-direct treatment by monthly application appears to be the most effective and feasible with controlled-release formulations having strong and long-lasting effects.
With regards to efficacy, inconsistent results were presented [11,27]. These are most probably the result of differences between strains, formulations, and experimental conditions. There is a clear evidence that pyriproxyfen effectively inhibits Aedes adult emergence at concentrations of <1 ppb.
As to the methods of application, the evidence is highly variable. For container treatment, the effectiveness of pyriproxyfen seems to depend on factors such as the material of which the individual containers are made and the local environmental conditions. For example, Vythilingam [40] reported much higher levels of sustained residual activity in plastic tubs than in earthen jars, while Schaefer [41] and Glare [42] demonstrated a lower stability of pyriproxyfen at higher temperatures. In addition, there seems to be a natural inter-individual variation between containers [35]. There are contradictory reports on whether different container sizes play a relevant role, probably to be addressed by the amount of pyriproxyfen used.
Overall, the available evidence shows that a targeted treatment with a slow-release (e.g. granular) or a long-lasting (e.g. resin strands) formulation can yield an adequate EI for up to 34 weeks. However, more research is needed to define the lowest effective concentrations for each formulation and the frequency by which the treated containers should be replenished. Studies should be performed with standard containers in field trials to avoid accurate estimates of mosquito density being obscured by random variation among individual containers. Also, further work is needed to determine the impact of environmental conditions (UV light, dilution, temperature, etc.) on effectiveness and persistence.
Regarding ULV application, only interventions with a product combination (e.g. permethrin) showed a significant effect, suggesting that ULV treatment with pyriproxyfen alone cannot be recommended. Some authors discussed the potential reasons for the failure of ULV and hypothesised that a) container openings were too small for pyriproxyfen to enter; b) treatment surfaces were too small to cover a significant number of breeding sites; c) extreme weather conditions (first year very dry, second year higher than usual precipitation) adversely impacted the larvicidal effect; d) cryptic habitats of Ae. albopictus not accessed or e) the slow mode of action of pyriproxifen was not considered in the follow up evaluation [32,33].
Space spraying-here as fumigation-has shown its effect only indoors, and it should be considered as part of a multi-intervention approach.
According to Devine [26], auto-dissemination could be an interesting approach to reach elusive breeding sites. Sihunicha [27] showed that an exposure of 30 min to water containing 0.003 g a.i./m 2 pyriproxyfen allowed for horizontal transfer of effective larvicidal pyriproxyfen doses to untreated environments. The same study reports that subsequent eclosion of eggs was decreased by 70-90%.
Different ovitraps have been designed and their effectiveness tested. Extraordinarily low doses of pyriproxifen (Ae. aegypti: LC 50 = 0.011 ppb (11), 0.012 [27], 0.0039 ppb [43]; Ae. albopictus: LC 50 = 0.11 ppb [45]) are needed for this approach [27,44]. The herein reviewed studies reported maximum distances of 150 m (28 2013) and 200 m [45] travelled by the mosquito from the treatment sites. These are in line with Marini et al. [46] who demonstrated that gravid female Aedes spp. travel 50-200 m from the release sites. Kaufmann et al. [47] reported longer distances of up to 3 km. Further studies are needed to prove and improve the auto-dissemination strategy under field conditions. Specifically, methods of application, concentrations and frequencies of treatment need to be clarified.
Controlled-release/long-lasting formulations had a good effect with much longer persistence than other treatments. In order to improve adherence in community based approaches, such interventions should be preferred. They may even be more efficient as re-treatment would have to be less often and, therefore, the overall operation is more cost-effective as less professional personnel is involved.
Those studies that combined the application of pyriproxyfen with different adulticidal products clearly showed that effectiveness can be increased by simultaneously controlling adults and larvae, and by expanding persistence [22].
From the presented studies, there is insufficient evidence to determine what impact the level of motivation in a community could have on vector control. This question warrants further investigation in larger prospective studies. Such trials are also needed to assess whether and how communities can be motivated and control efforts sustained.
The majority of the included studies were from South and Central America, yet, Bhatt [1] postulates that Africa's dengue burden is nearly equivalent to that of the Americas (i.e. 16 (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22) million infections annually, representing, 16% of the global total). Also, India is estimated to contribute 34% (i.e. 33  million infections per year) of the global total dengue infections. Further evidence is required on the effectiveness of pyriproxyfen in Africa and India to understand the influence of local environmental and societal factors.
Concluding, although pyriproxyfen is highly effective in killing the larvae of dengue transmitting vectors, and-to a smaller degree-also adult mosquito stages, evidence for the reduction of human disease transmission is weak. Lack of evidence is primarily due to small sample sizes, inappropriate study designs and lack of relevant, standardised outcome measures. Before issuing specific recommendations for the routine use of pyriproxyfen as a larvicide in dengue control programmes these research gaps must be addressed. Additionally, cross-resistance to pyriproxyfen has previously been reported by Macoris-Andrighetti [48] and, should it be verified, would have operational consequences for future dengue vector control. It therefore requires further attention in future studies as well as public health programmes.
Supporting information S1 Checklist. PRISMA checklist for "Community effectiveness of Pyriproxyfen as a Dengue Vector Control Method: A Systematic Review". (TIF)