Gates Open Res Gates Open Research 2572-4754 F1000 Research Limited London, UK 10.12688/gatesopenres.14300.1 Research Article Articles Using models and maps to inform Target Product Profiles and Preferred Product Characteristics: the example of Wolbachia replacement

[version 1; peer review: 1 approved with reservations, 1 not approved]

 Tiley Katie Formal Analysis Investigation Methodology Project Administration Software Validation Visualization Writing – Original Draft Preparation Writing – Review & Editing 1 2 Entwistle Julian Investigation Writing – Review & Editing 3 Thomas Bruce Investigation Writing – Review & Editing 4 Yakob Laith Methodology Supervision Validation Writing – Review & Editing 5 Brady Oliver Conceptualization Formal Analysis Funding Acquisition Investigation Methodology Project Administration Software Supervision Validation Visualization Writing – Original Draft Preparation Writing – Review & Editing a 1 2 Centre for Mathematical Modelling of Infectious Diseases, London School of Hygiene & Tropical Medicine, London, UK Department of Infectious Disease Epidemiology, Faculty of Epidemiology and Population Health, London School of Hygiene & Tropical Medicine, London, UK IPM Focus Ltd., Rowland’s Castle, Hampshire, UK The Arcady Group, Richmond, VA, USA Department of Disease Control, London School of Hygiene & Tropical Medicine, London, UK oliver.brady@lshtm.ac.uk

No competing interests were disclosed.

 12 5 2023 2023 7 68 24 1 2023 Copyright: © 2023 Tiley K et al. 2023 This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Background 

The global prevalence of diseases transmitted by Aedes aegypti mosquitoes, such as dengue, Zika and Yellow Fever, is increasing, but development of promising new mosquito control technologies could reverse this trend. Target Product Profiles (TPPs) and Preferred Product Characteristics (PPCs) documents issued by the World Health Organization can guide the research and development pathways of new products and product combinations transitioning from proof of concept to operational use.

Methods 

We used high resolution global maps of the case and economic burden of dengue to derive programmatic cost targets to support a TPP for Wolbachia replacement. A compartmental entomological model was used to explore how release size, spacing and timing affect replacement speed and acceptability. To support a PPC for a hybrid suppress-then-replace approach we tested whether Wolbachia replacement could be achieved faster, more acceptably or at a lower cost if preceded by a mosquito suppression programme.

Results 

We show how models can reveal trade-offs, identify quantitative thresholds and prioritise areas and intervention strategies for further development. We estimate that for Wolbachia replacement to be deployable in enough areas to make major contributions to reducing global dengue burden by 25% (in line with 2030 WHO targets), cost must ultimately be reduced to between $7.63 and $0.24 (USD) per person protected or less. Suppression, particularly interventions that induce mosquito sterility, can reduce the number of Wolbachia mosquitoes necessary to achieve fixation by up to 80%. A hybrid approach can also achieve fixation faster and potentially improve acceptability, but may not justify their cost if they require major new investments in suppression technologies.

Conclusions 

Here we demonstrate the value dedicated modelling can provide for interdisciplinary groups of experts when developing TPPs and PPCs. These models could be used by product developers to prioritise and shape development decisions for new Wolbachia replacement products.

 mosquito dengue model arbovirus policy intervention Wolbachia cost Medical Research Council MR/V031112/1 Biotechnology and Biological Sciences Research Council BB/T008709/1 Gates Foundation [OPP1183567] and UK Medical Research Council Career Development Award [MR/V031112/1] assigned to OJB. Biotechnology & Biological Sciences Research Council [BB/T008709/1] assigned to KT. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Introduction

The Aedes aegypti mosquito is the principal vector of dengue, Zika, yellow fever and chikungunya viruses. Dengue incidence has been rising and the WHO Global Vector Control Response 2017 – 2030 reports an annual 96 million cases, 1.9 million DALYs and 9,110 deaths 1 . Vaccines are only available for yellow fever and are not currently widely used for dengue, though there are other dengue and chikungunya vaccine candidates in clinical trials 2, 3 . There are no drugs available to combat these infections and so there is a reliance on prevention through vector control. Effective control of this vector is difficult to achieve and sustain given the mosquito’s high reproductive rate and adaptation to urban habitats, with an egg stage that can survive desiccation and a larval phase that can develop in small, temporary water volumes (e.g., water containers and roof gutters). The rapid growth of cities has also favoured this mosquito 4 . As a result, existing vector control tools alone have generally been unable to sustainably control Ae. aegypti or the diseases it transmits over the long term. A range of novel technologies are under development 5 , including biocontrol through use of Wolbachia spp. for population replacement or reduction/suppression, the release of genetically modified mosquitoes (such as Oxitec's 1 st generation self-limiting technology (1gSLT) 6 ), and other forms of sterile insect technique (SIT).

Mosquitoes infected with Wolbachia show reduced rates of virus dissemination, making them less capable of transmitting arboviruses 7 . Wolbachia infection is also dominantly maternally inherited and leads to inviable progeny when Wolbachia males and wild-type females mate. This means that Wolbachia can be used to either replace the existing mosquito population with a lower competence phenotype by releasing females (or males and females) or suppress the existing population by releasing only males.

Wolbachia population replacement involves regular releases of Wolbachia-infected mosquitoes into a wild mosquito population over a period of several months. Once a critical proportion of mosquitoes in the population have Wolbachia, coverage will continue to increase to fixation without further releases, but below this threshold Wolbachia coverage may decline (possibly to zero) once releases stop due to fitness costs associated with released mosquito strains 8 . Operationally, the chance and speed of exceeding this threshold and achieving self-sustaining coverage can be achieved by: increasing the number of releases, decreasing the time gap between releases and increasing the ratio of Wolbachia-infected Ae. aegypti in relation to wild-type Ae. aegypti in each release. All of these options increases cost and can also lead to undesirable temporary increases in the Ae. aegypti mosquito population that could be a key barrier to community acceptability 9, 10 .

A growing range of entomological, epidemiological and modelling evidence supports the widespread, long-term effectiveness of Wolbachia replacement 11– 13 . This includes a randomised controlled trial (RCT) of wMel Wolbachia in Yogyakarta City, Indonesia which demonstrated a 77% reduction in dengue incidence and an 86% reduction in hospitalizations 14 . To date, however, Wolbachia replacement programmes have only been conducted in specific mid-sized cities or specific neighbourhoods of cities. Thirteen countries have implemented replacement programmes at various levels of scale, with 12 through the World Mosquito Program (WMP) and an independent programme in Malaysia 15, 16 . Meanwhile, China (with Ae. albopictus), Singapore, and the USA have so far chosen to use suppression-based programs due to perceived greater compatibility with their existing intensive and long-term efforts to suppress mosquito populations 17– 19 . To date, there have been no large-scale (national or regional), non-donor funded implementations of Wolbachia replacement or suppression approaches.

These novel technologies ( Wolbachia replacement, Wolbachia suppression , 1gSLT and SIT ) are subjects of ongoing development, evaluation, demonstration and scale-up in various high-burden programmatic and private settings. In addition, there is the potential to combine Wolbachia population replacement with other suppression technologies in ways that may be synergistic from a cost, community acceptance, or dengue incidence reduction perspective 9 . Specifically, a programme of suppression followed by replacement has the potential to increase the likelihood of successful Wolbachia establishment and reduce the cost of, and risk of mosquito exacerbation associated with, achieving establishment 20 .

Development and transition to scale of new products and strategies can be accelerated by the development of internationally recognised Target Product Profiles (TPPs) and Preferred Product Characteristics (PPCs) documents 21 . TPPs provide specific quantitative guidance on the key characteristics a product must (minimum target), or should ideally (preferred target), meet when developed into a deployable mass market product. PPCs identify broader areas of unmet need and aim to stimulate new products or product combinations that can address these needs. In early 2022 the WHO convened a Technical Advisory Group (TAG) to develop a draft TPP for Wolbachia replacement and a draft PPC for a hybrid mosquito suppression then Wolbachia replacement strategy. TAG members agreed that these WHO TPPs and PPCs should closely align with the WHO’s strategy and goals to control dengue globally. As such the WHO’s goal to reduce dengue incidence by 25% by 2030 (2010 – 2020 baseline 22 ) provided a basis to understand the scale and range of settings in which these TPPs, PPCs and the products they ultimately produce are relevant. Computational models can play a key role in the development of TPPs and PPCs due to their ability to generalise beyond areas where data have been collected and make predictions if aspects of the product were to change. Here we describe a dynamic compartmental entomological model and a global geospatial economic model that we developed and used to explore how operational and economic aspects of Wolbachia replacement are likely to change once the technology is used at scale.

Methods Global dengue cost model

The global dengue cost model aims to produce high spatial resolution estimates of the economic costs of dengue that would be averted by Wolbachia replacement. These were conservatively estimated to be composed of the direct medical cost of treatment of dengue patients and emergency (outbreak) vector control costs. In the absence of primary data on willingness to pay for Wolbachia replacement programmes, these averted costs were assumed to represent an appropriate proxy.

A high resolution (5km × 5km at the equator) map of symptomatic dengue case burden was obtained from Bhatt et al. 23 , which estimates the spatial distribution of the 96 (67 – 196) million episodes estimated to occur each year. An average direct medical cost per symptomatic case (2013 USD) was derived for each country from Shepard et al. 24 , considering the different costs of hospitalised and ambulatory cases and the country-specific distribution of symptomatic cases among these two different treatment settings. Direct medical costs include the costs of specific medicines and staff time required to treat a dengue patient and a portion of infrastructural costs and is the most relevant measure of what governments need to pay to treat cases of dengue illness each year. All costs were inflated from 2013 to 2020 USD using World Bank country GDP deflators with a maximum capped value of a two-fold increase 25 .

A literature review on the cost of vector control in dengue endemic countries was conducted and identified studies with national and subnational estimates of vector control costs for 17 countries. Twenty studies included costs of routine vector control activities and seven studies included costs of vector control during dengue outbreaks (supplementary file 1 in Data Availability). All vector control cost values were converted back to local currencies using the exchange rate at the time of the costing, inflated to 2020 using country GDP deflators from the World Bank 25 , and then converted to 2020 US dollars using 2020 exchange rates published by the World Bank 26 . To make predictions of per capita routine vector control costs for countries without costing data, a Poisson generalized mixed linear model was fit to the costing data with national GDP per capita (log scale) as a covariate and national-level random effects. Predictions were then made for all countries globally using World Bank GDP per capita figures from 2020. For countries where this data was missing (some small Caribbean and Pacific Island nations), global median GDP per capita was assumed. Of the seven studies identified that included costs of vector control during dengue outbreaks, five studies gathered information on both routine and outbreak vector control activities. We assume that implementing a Wolbachia release program will not avert routine (principally preventative) vector control costs because Wolbachia replacement is unlikely to eliminate dengue in most settings and additional vectors (e.g., Ae. albopictus) and nuisance biting mosquitoes will still drive a need for routine vector control activities. Instead, it was assumed that the implementation of Wolbachia replacement will significantly limit the size of outbreaks and their required vector control response and thus cost. These studies suggested that during outbreaks, the monthly cost of vector control increases by 20–50%. Three scenarios were explored where additional avertable outbreak costs composed 35% of routine monthly vector control costs for a duration of three months every year, with a sensitivity analysis exploring lower (20%) or higher (50%) values.

Total annual averted costs were estimated assuming Wolbachia replacement results in a 70% reduction in symptomatic cases (and their associated costs) and 100% of emergency (outbreak response) vector control costs. This is based on a conservative interpretation of the 77% effectiveness of wMel Wolbachia measured in the Yogyakarta trial 14 and the expectation of variable effectiveness across areas with different transmission intensities 13 . While wMel Wolbachia replacement has been shown to be stable in Aedes mosquito populations for over ten years in Australia 12 , it is unclear how many future years of averted dengue costs would be appropriate to consider when estimating government or other funder willingness to pay. We therefore estimate total averted costs for three-, five- and ten-year time horizons and assume that these costs represent the maximum price a government or funder would be willing to pay for Wolbachia replacement in a given setting. To quantify uncertainty around these thresholds, the analysis was repeated with values from the upper and lower bounds of the case burden 23 and economic burden 24 estimates and with 50% and 20% avertable vector control outbreak proportions respectively.

Wolbachia replacement is most cost effective if targeted to higher density, high dengue burden areas 27 . We therefore identified which areas (5km × 5km pixels) would need to be targeted to reach the WHO goal of reducing global dengue burden by 25% in the most net cost-efficient manner. To do this, each 5km × 5km pixel was ranked from highest to lowest based on a benefit (averted medical and outbreak costs) to cost (approximate Wolbachia programme cost estimate based on population density from Brady et al. 27 ) ratio. Cumulative averted cases were then calculated and pixel selection ended when averted cases first exceeded 25% of the global total. The averted costs in the last, least cost-efficient pixel included in this subset then gave the cost threshold for Wolbachia replacement programmes, i.e., if Wolbachia replacement can be achieved at this cost (or lower) it will be possible, from the cost-efficacy perspective, to implement the intervention in enough areas to reduce the global burden of dengue by 25%. An alternative scenario was also calculated where it was assumed that Wolbachia replacement is only required to account for half of this global target, i.e., a 12.5% global burden reduction. Because such a global targeting approach prioritises countries with higher GDP, we also calculated a scenario where 95% of dengue endemic countries (defined as >10,000 symptomatic dengue infections a year as estimated by Bhatt et al. 23 ) needed to achieve at least a 25% burden reduction through deploying Wolbachia replacement, to explore the cost threshold implication of a much wider deployment with improved equity between countries 28 .

 Entomological model overview

This compartmental mechanistic model follows Aedes aegypti population dynamics at egg, larvae, pupae, and adult stages, with pupae developing into female and male adults in equal proportion and each stage subject to a constant death rate:

d O d t = φ F − α O O − μ O O d L d t = α O O − α L L − μ L L d P d t = α L L 1 + ( γ L ) β − α P P − μ P P d M d t = 0.5 α P P − μ M M d F d t = 0.5 α P P − μ F F ( 1 )

O denotes the number of eggs, L larvae, P pupae, M adult males, and F adult females. φ is the daily egg-laying rate of adult females. α O denotes the rate at which eggs develop into larvae and the μ O death rate of eggs. Similarly, α L denotes the rate at which larvae develop into pupae and μ L the larval death rate and α P denotes the rate at which pupae develop into adults and μ P the pupal death rate. μ M and μ F are the adult male and female death rates, respectively. Survival of larvae to pupal stage is density dependent, and using the flexible formulation proposed by Maynard Smith and Slatkin 29 , includes the parameter γ which determines the density at which mortality remains proportionate and the parameter β the ‘abruptness’ of density-dependence.

These equations were then further developed to account for Wolbachia deployments, respectively impacting mating and larval survival:

d O d t = φ F M + c i M W M + M W − α O O − μ O O d P d t = α L L 1 + ( γ ( L + L W ) ) β − α P P − μ P P

Where, c i denotes the failure rate of cytoplasmic incompatibility for the Wolbachia-infected adult males ( M W ), and Wolbachia-infected larvae in the wild ( L W ) also contribute towards larval competition.

While it is important that all wild-hatched larvae are subject to the same density dependence as they are occupying the same habitat, the introduced Wolbachia-infected eggs will be released in their own distinct larval habitat (self-contained release containers), therefore their survival is not impacted by the densities of wild-hatched larvae. The equations of released Wolbachia ( rW) are as follows:

d O r W d t = R R ( F ˙ + M ˙ ) − α O O r W − μ O O r W d L r W d t = α O O r W − α L L r W − μ L L r W d P r W d t = α L L r W 1 + ( γ L r W ) β − α P P r W − μ P P r W ( 2 )

O rW denotes the number of released Wolbachia-infected eggs, which is the product of the release ratio ( RR) and the equilibrial adult population prior to control ( Ḟ + Ṁ). L rW denotes the number of Wolbachia-infected larvae resulting from released eggs, and P rW denotes the number of Wolbachia-infected pupae resulting from released eggs. The aquatic-stage Wolbachia-infected Ae. aegypti that hatch outside of the release containers (subscript ‘ W’ instead of ‘ rW’) are tracked separately from those which are newly released. The wild-hatching Wolbachia-infected mosquitoes follow these dynamics:

d O W d t = φ F W − α O O W − μ O O W d L W d t = α O O W − α L L W − μ L L W d P W d t = α L L W 1 + ( γ ( L + L W ) ) β − α P P W − μ P P W ( 3 )

Wolbachia-infected adult mosquitoes comprise those that have emerged from the wild combined with those emerging from release containers:

d M W d t = 0.5 α P ( P W + P r W ) − μ M ε M W d F W d t = 0.5 α P ( P W + P r W ) − μ F ε M W ( 4 )

ε denotes the relative mortality of Wolbachia-infected adult mosquitoes compared to uninfected. A sensitivity analysis explored the impact of Wolbachia infection fitness costs on mosquito population dynamics under Wolbachia release scenarios by varying ε (Supplementary Figure 1). Parameter definitions and values are shown in Table 1.

Model parameters values.
Parameter Description Value Reference
φ Daily egg laying rate of adult females 500*(1/14) Otero et al., 2006 30
M null Male uninfected adults M + c iM w -
F all Total female adults 1+ F + F w -
M all Total male adults 1 + M + M W + c MM s -
α O Daily rate eggs hatch into larvae 0.5 Marinho et al., 2016 31
α L Daily rate larvae develop into pupae 0.18 Marinho et al., 2016 31
α P Daily rate pupae develop into adults 1 Masters et al., 2020 32
μ O Daily mortality rate of eggs 0.01 Trpis, 1972 33
μ L Daily mortality rate of larvae 0.1* Couret et al., 2014 34
μ P Daily mortality rate of pupae 0.1* Couret et al., 2014 34
μ M Daily mortality rate of adult males 1/14 Yakob et al., 2008 10
μ F Daily mortality rate of adult females 1/14 Yakob et al., 2008 10
γ Determines the density at which mortality remains proportionate 1 Bellows, 1981 35
β Determines the ‘abruptness’ of density dependence 0.5 Bellows, 1981 35
RR Release ratio of Wolbachia-infected adults compared to total adult mosquitoes Variable Estimated
c i Proportion of cytoplasmic incompatibility that fails 0.012 Walker et al., 2011 7
c m Competitiveness of released sterilised males 0.5 Winskill et al., 2014 36
c v Proportion of adult population reached by adulticide 0.141 Estimated, described in adulticide section
ε Relative mortality of Wolbachia-infected adult mosquitoes compared to uninfected 1.2 Joubert et al. 2016 37
s Adjustment parameter which matches average seasonal mosquito population to non-seasonal equilibrium mosquito population 2.09 Estimated, described in seasonality section
Suppression

Wolbachia-infected egg release was also explored after first deploying suppression interventions. The suppression techniques analysed were the release of 1 st generation self-limiting technology (1gSLT), sterile insect technique (SIT), Male Wolbachia release, environmental management, larvicides, and adulticides. Each type of suppression was included as a function of time, t, so that a value which influences model dynamics is pulsed at specific times or maintained over a specific period. The efficacy of each method was based on evidence sourced from the literature, selected with a preference for large randomised-controlled trials, however, each suppression method works differently and trials to measure effectiveness vary with study design. Efficacy of a single burst of application was preferable but only found for adulticide. Studies measuring repeated concurrent applications of 1gSLT and Male Wolbachia release, environmental management and larvicides used an interrupted time series design. No suitable publications were found giving field estimates of effectiveness for SIT. The impact of variations in study design are discussed in more detail below.

1gSLT. Release of 1gSLT adult males produce offspring with wild females of which only the males develop to adulthood from the pupal stage. 1gSLT was included in the model by pulsing adult males into a sterile adult male compartment M S weekly, which then contributed to the production of sterile eggs, O S , which developed through sterile larval and pupal compartments, L S and P S , contributing to a density dependant survival function. RR supp denotes the release ratio for mosquito release suppression techniques. ε denotes the relative mortality of Wolbachia-infected adult mosquitoes compared to uninfected adult mosquitoes.

d O S d t = φ F a l l c M M S M a l l − α O O S − μ O O S d L S d t = α O O S − α L L S − μ L L S d P S d t = α L L S 1 + ( γ ( L + L w + L S ) ) β − α P P S − μ P P S d M S d t = R R s u p p f R I D L − μ M ε M S ( 5 )

Parameters for the fixed rate efficacy of 20%, 50%, and 80% were calculated by comparing the total adult population at model equilibrium with the minimum adult population reached after five weeks of application. The literature-derived efficacy values were 45% five weeks after the last suppression period and 70% ten weeks after the last suppression period 38 , calculated by comparing the total adult population at model equilibrium to the total adult population after five- or ten-weeks of suppression which achieved the desired efficacy (summarised in Supplementary Table 1). A caveat of this approach is that the resulting minimum adult population is reached later than five- or ten-weeks, therefore, the maximum efficacy calculated in these scenarios is marginally greater than the literature value stated (shown in Supplementary Figure 2).

SIT. SIT involves releasing sterile adult males which produce sterile eggs (in the same manner of equation 5) that do not develop further. SIT was included in the model by pulsing adult males into the sterile adult male compartment, M S , which then contributed to the production of sterile eggs which then do not develop further.

d M S d t = R R s u p p f S I T 0.5 α P P S − μ M ε M S ( 6 )

Parameters for the fixed rate efficacy of 20%, 50%, and 80% were calculated by comparing the total adult population at model equilibrium with the minimum adult population reached after five weeks of application. There are currently no published randomised controlled trials for suppression of Aedes mosquitoes using SIT, therefore 50% efficacy is shown for SIT in analyses where only literature-derived efficacy is used.

Male Wolbachia release. Male Wolbachia release involves releasing only Wolbachia-infected adult males, resulting in no offspring due to cytoplasmic incompatibility with the local non- Wolbachia-infected females. Male Wolbachia release was implemented in the model by pulsing Wolbachia-infected adult males into the Wolbachia-infected male compartment:

d M W d t = R R s u p p f I I T 0.5 α P ( P W + P r W ) − μ M ε M W ( 7 )

Parameters for the fixed rate efficacy of 20%, 50%, and 80% were calculated by comparing the total adult population at model equilibrium with the minimum adult population reached after five weeks of application. The literature-derived efficacy values were 60% five weeks after the last suppression period and 95% ten weeks after the last suppression period 39 , calculated by comparing the total adult population at model equilibrium to the total adult population after 5- or 10-weeks of suppression which achieved the desired efficacy (summarised in Supplementary Table 1). Similar to 1gSLT, a caveat of this approach is that the resulting minimum adult population is reached later than 5- or 10-weeks, therefore, the maximum efficacy calculated in these scenarios is marginally greater than the literature value stated (shown in Supplementary Figure 2).

Environmental management. Environmental management reduces the amount of egg-laying habitat which was simulated by manipulating the egg production rate:

d O d t = f E M φ F M n u l l M a l l − α O O − μ O O d O W d t = f E M φ F W M + M W M a l l − α O O W − μ O O W ( 8 )

Parameters for the fixed rate efficacy of 20%, 50%, and 80%, and the literature-derived efficacy of 47.4% 40 , were calculated by comparing the total adult population at model equilibrium with the new population equilibrium reached after applying this technique for the duration of the simulation; this emulates real-world scenarios in which application and outcome are typically long-term (summarised in Supplementary Table 1).

Larvicides. Larvicides were simulated using the example of guppies, reducing the number of larvae through increased mortality to all aquatic life stages of both uninfected and wild-hatching Wolbachia-infected mosquitoes.

d O d t = φ F M n u l l M a l l − α O O − f L V μ O O d O W d t = φ F W M + M W M a l l − α O O W − f L V μ O O W d L d t = α O O − α L L − f L V μ L L d L W d t = α O O W − α L L W − f L V μ L L W d P d t = α L L 1 + γ ( L + L W + L S ) β − α P P − f L V μ P P d P W d t = α L L w 1 + ( γ ( L + L w + L S ) ) β − α P P W − f L V μ P P W ( 9 )

Parameters for the fixed rate efficacy of 20%, 50%, and 80%, and the literature-derived efficacy of 44% 41 , were calculated by comparing the total pupae population at model equilibrium with the new pupae equilibrium reached after applying this technique for the duration of the simulation; similar to environmental management techniques this emulates real-world scenarios in which application and outcome are typically long-term (summarised in Supplementary Table 1).

Adulticides. Finally, the deployment of adulticides through fogging and chemical spraying were simulated by pulses which manipulate the mortality rates of adult male and female compartments:

d M d t = 0.5 α P P − μ M M − f A D c v M d M W d t = 0.5 α P ( P W + P r W ) − μ M ε M W − f A D c v M W d F d t = 0.5 α P P − μ F F − f A D c v F d F W d t = 0.5 α P ( P W + P r W ) − μ F ε F W − f A D c v F W ( 10 )

Parameters for the fixed rate efficacy of 20%, 50%, and 80% were calculated by comparing the total adult population at model equilibrium with the minimum adult population after one application. Mani et al. 42 reported an initial 94% reduction in mosquito resting density from application of deltacide, a synergized mixture of pyrethroids, after which the population completely recovered within seven days (summarised in Supplementary Table 1). This combination of great suppression and swift recovery could not be replicated in the model by only manipulating adult mortality; this may be because some portion of the reduction in resting density was due to a repellent effect, which has been noted as a possibility by the source paper 42 or because recovery was due to recolonisation by neighbouring populations which is not modelled here. To fit this literature efficacy a parameter for 94% mortality rate was first calculated by comparing the total adult population at model equilibrium with the minimum adult population after one application and subsequently a coverage parameter, denoted as c v , was fitted using the 94% efficacy parameter. The highest proportion of coverage was calculated which allowed 80% population recovery within three weeks of suppression using literature-derived efficacy; the assumption of this recovery speed was explored with a sensitivity analysis, shown in Supplementary Figure 3, and found to be minimally affected by changing the number of weeks taken for recovery.

 Seasonality

Seasonality is defined by using a normalised and smoothed lowess curve of average monthly precipitation (sourced from www.meteoblue.com) to create a score, bounded by 0 and 1, for a seasonality profile with a distinct wet and dry season each year, produced using data from Rio de Janeiro. Within the model, this score influences γ within the density dependent function. The density-dependent seasonal function of larval survival is:

L 1 + ( γ f K s ( L + L w + L S ) ) β ( 11 )

The seasonality function, f K , returns a precipitation score dependent on time, t, which affects the rate larvae develop and enter the pupal stage. s is a constant, calculated to ensure the average mosquito population in the seasonal model is within 0.5 of the non-seasonal model equilibrium which allows comparability to the non-seasonal analyses as suppression efficacy parameters and functions execute according to this average. This affects all wild-hatched model compartments; however, the released Wolbachia-infected eggs are introduced in containers which isolate them from the limits of rainfall dependent egg hatching and larval growth (seasonal population dynamics shown in Supplementary Figure 4). Placement of the seasonality function within the model was explored (Supplementary Table 2) in addition to the impact of temperature, rather than precipitation, on larval development (Supplementary Figure 5).

 Release scenario analyses

Wolbachia coverage data was extracted from the report by Utarini et al. 14 using WebPlotDigitzer 43 and the intervention cluster-level results used as reference to calibrate the model. Simulations of Ae. aegpti population dynamics were undertaken to investigate the intervention conditions which would produce the desired Wolbachia coverage levels (>95%) in the mosquito population. Specifically, a range of release ratios (0.03 to 0.1, in increments of 0.01) and number of releases (9 to 14, pulsed every 14 days) were explored and their influence on the number of days until target coverage was achieved.

The influence of suppression efficacy (20%, 50%, 80%, and a literature derived efficacy) and the week of switch from suppression to Wolbachia release (1 to 10) was investigated in terms of the minimum release ratio (0.0025 to 0.4, explored in increments of 0.0025) necessary to reach Wolbachia target coverage within six months of the first Wolbachia release.

The seasonality model was run for 18 months and the initial six months burn in period needed for model calibration was discarded; five weeks of suppression followed by five rounds of Wolbachia replacement release were simulated, exploring minimum RR (0.0025 to 0.4 in increments of 0.0025) required to reach target coverage within six months of first Wolbachia replacement release.

 Results Exploring the sensitivity of <italic toggle="yes">Wolbachia</italic> replacement to key release characteristics

To explore how self-sustaining Wolbachia replacement can best be achieved and its sensitivity to various operational parameters we formulate, fit and simulate from an entomological dynamic compartmental model. By calibrating the release ratio parameter, our model showed a good fit to the mosquito release data from the Yogyakarta RCT with replacement dynamics and coverage levels proceeding at a similar rate. The model reached 50% coverage after 121 days, compared to an average of 117 days observed in the RCT, and 90% coverage after 180 days, compared to an average of 239 days observed in the RCT (Supplementary Figure 6). As Wolbachia reaches fixation the model slightly overestimates final Wolbachia coverage, likely due to prevalence being suppressed in the RCT due to migration of uninfected adult mosquitoes from outside the release area which were not included in our model and would be reduced when implemented as a wide-scale blanket intervention.

During a Wolbachia replacement programme, our model predicts that the total adult mosquito population experiences a temporary exacerbation above baseline levels, followed by a decline before reaching a new equilibrium once Wolbachia fixation has been achieved ( Figure 1A, 1B). Due to the fitness cost of Wolbachia (conservatively assumed to be 20% 37 , but variable depending on environment), this new equilibrium mosquito population size is predicted to be lower than before Wolbachia release.

Mosquito dynamics of <italic toggle="yes">Wolbachia</italic> replacement.

Figures A and B show the total adult mosquito population size over time during Wolbachia replacement after a nine ( A) and fourteen ( B) release (dotted vertical lines) round programme. Red dots indicate the date at which target coverage (>= 95% Wolbachia coverage) was first achieved. The blue lines show the time points from first release at 0 days, 6 months, and 1 year from left to right. C) Days until Wolbachia-infected adult mosquitoes reach target coverage for different numbers of releases and release ratios (RR). The horizontal blue line indicates 6 months. D) The percentage increase in total mosquito population for different numbers of releases and release ratios. This is calculated as the sum of the peak increase after initial release compared to the pre-release population equilibrium and the peak increase after target coverage is achieved compared to post-fixation equilibrium population, the latter excludes lower RR instances where the population at fixation is lower than the new population equilibrium.

As expected, Wolbachia target coverage can be achieved faster by increasing RR and/or by increasing the number of releases ( Figure 1C). Our model shows that increasing RR of each release will reduce time to target coverage more than increasing the number of release rounds, particularly at higher RR values. Above a RR of 0.06, increasing the number of release rounds has little additional effect on time to target coverage. Our model predicts that achieving Wolbachia target coverage within six months of the first release is possible with RR >= 0.06 with >= 9 releases. Higher RR and number of release rounds lead to ever diminishing decreases in time to target coverage. Increasing the RR from 0.06 to 0.1 is only predicted to increase time to target coverage by 40 days (nine release programme, Figure 1C).

Higher RRs and higher release round numbers also lead to disproportionately undesirable temporary exacerbation issues, particularly at higher values ( Figure 1A,B,D). A doubling of RR from 0.05 to 0.1 could lead to an approximate doubling of exacerbation (6.96% to 14.11%) under a nine-release programme, but this could be up to 2.6 times more (7.20% to 18.80%) if the number of release rounds were increased to 14. This is because prolonged releases at high RR led to a secondary peak in mosquito abundance that prolongs the period of exacerbation ( Figure 1B); in the fourteen-release programme this is still less than the original population, however, increasing number of releases could expect this secondary peak to eventually exceed the prior population size.

Overall, these simulations suggest the importance of balancing speed of Wolbachia replacement with the potentially negative consequences of temporarily exacerbating the mosquito population. In combination with other field evidence, this work supported the TPP’s guidance on “time to achieve target coverage”. The models suggested that a time to achieve coverage of less than 12 months was highly feasible (minimum TPP standard) and that a goal of 6 months (preferred TPP standard) was achievable. To counterbalance the issue of exacerbation, the TPP included a criterion for “community acceptability” that states that any increase in nuisance biting through the chosen release characteristics is “acceptable to local residents”, recognising that the definition of “acceptable” is likely to be highly context specific.

 Global cost targets for <italic toggle="yes">Wolbachia</italic> replacement

The TAG identified cost as a key reason limiting wider adoption of Wolbachia replacement and therefore a “mature product cost once implemented at scale” criterion was a key feature of the TPP. This cost criterion needed to be low enough to drive innovation and ensure a significant proportion of the global population at risk of dengue can benefit, but not too low as to exclude promising products from further development.

Because detailed data on willingness to pay was unavailable at the time of analysis, we developed a range of scenarios that assume willingness to pay is approximated by the costs of treating dengue cases and of vector control in response to outbreaks over a range of years ( Table 2). Each scenario gave a theoretical cost each 5km × 5km area would be willing to pay for Wolbachia replacement. Areas that supported higher costs typically had higher dengue burden but were also heavily influenced by the cost of dengue treatment and prevention.

The predicted target cost per person for <italic toggle="yes">Wolbachia</italic> replacement based on different assumptions about desired global impact (rows) and averted medical and outbreak control costs (assumed proxy of willingness to pay, columns).

Wolbachia replacement would need to be at or below this cost to achieve each impact scenario in full. All values show median estimates in 2020 US dollars, brackets show model predicted uncertainty around the true value of this cost threshold at the 95% credible interval level).

Impact scenario Required cost per person covered (10 years benefit) Required cost per person covered (5 years benefit) Required cost per person covered (3 years benefit)
12.5% global burden reduction $7.63 (5.15 – 29.42) $4.10 (2.77 – 15.83) $2.54 (1.71 – 9.78)
25% global burden reduction $4.33 (2.73 – 18.95) $2.33 (1.47 – 10.20) $1.44 (0.91 – 6.30)
12.5% national burden reduction $0.98 (0.64 – 3.78) $0.53 (0.34 – 2.03) $0.33 (0.21 – 1.26)
25% national burden reduction $0.72 (0.26 – 1.66) $0.39 (0.14 – 0.89) $0.24 (0.09 – 0.55)

We estimate that to achieve a 25% reduction in the global burden of dengue, as per the WHO 2020-2030 goals, using only Wolbachia replacement targeted to the most cost-efficient areas, would require releases across 924,557km 2 in 73 countries ( Figure 3). This corresponds to 34.7% of the urban (> 300 people per km 2) area at risk and just 1.7% of the total area at risk of dengue. If Wolbachia only needs to achieve half of the global 25% reduction, with other interventions responsible for the remaining half, Wolbachia releases would only need to be targeted to 255,459km 2 over 47 countries ( Figure 2B and 2E). However, because these cost estimates are uncertain and because this approach prioritises high income countries where dengue treatment costs are high, we also include a third and fourth targeting scenario where 25% or 12.5% of the national burden must be reduced for the majority (95%) of dengue endemic countries ( Figure 2C and 2F). These scenarios improve equity over dengue-endemic regions. A full list of 2 nd administrative units included under each targeting scenario, the costs each will support and additional contextual information (population, density, etc) is included in the following repository: https://github.com/katietiley/Wolbachia_TPP_PPC.git.

Targeting areas for Wolbachia replacement to meet different global and national goals in Southeast Asia ( A– C) and Central America and the Caribbean D– F). Maps show the areas most cost efficient to target (red) to reduce the global burden of dengue by 25% ( A and D) or 12.5% ( B and E) or the national burden by 25% ( C and F) based on the cost of treatment and prevention of current dengue burden. Predictions for other areas and lists of municipalities to be targeted are included in the following repository https://github.com/katietiley/Wolbachia_TPP_PPC.git.

 The cumulative contribution of each dengue endemic country to the 25% global reduction in dengue burden and the averted costs (willingness to pay proxy) per person covered at which it can be achieved.

Horizontal dotted lines show the cost thresholds of $2.33 per person and $0.24 per person chosen for the TPP minimum and preferred criteria respectively. Only high burden countries are labelled.

For Wolbachia replacement to be implemented in enough areas to meet these impact targets, the cost of implementation must ultimately be reduced to between $7.63 and $0.24 per person covered depending on scenario ( Table 2). The cost thresholds identified in Table 2 represent the area with the lowest averted costs (assumed lowest willingness to pay) within the areas needed to reach each impact target. This means that many eligible areas, or even whole countries, could support higher programme costs, but ultimately Wolbachia replacement will need to be implemented at or below this cost threshold in order to reach the impact target. The distribution of these costs and benefits by country is shown in Figure 3 for the 25% global burden reduction impact target. This shows that while globally Wolbachia replacement will need to be achieved for $2.33 per person to meet the 25% impact goal, many countries could support higher costs with many high burden countries able to implement in a wide range of high burden areas above the $10 per person line.

Cost targets become lower (increasingly more ambitious for product development) as impact scenarios become more ambitious or, to a lesser extent, as the accepted duration of benefits becomes shorter (rows and columns in Table 2 respectively). To achieve a 25% national burden reduction in all dengue endemic countries would require a cost target ~ 10x lower than to achieve a 12.5% global dengue burden reduction, emphasising that even higher cost products could still have substantial global impact, but would be less equitable unless subsidised for countries with lower financial capacity for dengue prevention and treatment. Due to high uncertainty in estimates of the true burden of dengue 23 and its costs of treatment 24 and prevention, uncertainty around these cost thresholds is moderate with higher uncertainty around higher median cost thresholds.

Each of these scenarios and their respective cost targets were presented to the TAG for discussion and selection for the TPP. Recognising that the TPP minimum criteria should reflect the minimum cost for a product to be viable at substantial scale, TAG members selected the $2.33 per person covered target (corresponding to a 25% global burden reduction with five years of benefit, Table 2). This cost needs to include the programme of activities required to reach 90% coverage of Wolbachia in the release areas one year after starting releases. The TPP also makes allowances for a slower programme where 90% coverage is achieved over three years, but this must be achieved at a more stringent minimum TPP cost target of $1.44 per person covered (corresponding to a 25% global burden reduction with three years of benefit, Table 2). However, to challenge developers to meet the more equitable 25% national dengue burden reduction, the TPP preferred cost threshold was set at $0.24 per person covered (corresponding to three years of benefits in this scenario, Table 2). These decisions were also informed by evidence that the current World Mosquito Program cost base for wMel Wolbachia replacement is in the US$5-22 per person range 27 with a medium-term goal of achieving Wolbachia replacement for $1 per person 44 . Given that these TPP targets represent the lowest averted medical and outbreak control costs per person among all areas where releases are required, there are many areas that could support higher programme costs. Therefore, provided developers can demonstrate the prospect of achieving the TPP cost targets in future, there is scope to operate higher cost programmes before these targets are achieved.

 Exploring the development of a hybrid “suppress then replace” approach

In addition to the draft TPP for Wolbachia replacement, the TAG was also tasked to develop a draft PPC for a hybrid mosquito population suppression followed by Wolbachia replacement approach. Reducing the natural mosquito population size allows Wolbachia replacement programmes to achieve higher release ratios or achieve comparable release ratios by releasing fewer Wolbachia mosquitoes. To support the development of the PPC we developed a compartmental entomological model and simulated hybrid strategies with a range of suppression types to answer the questions: can hybrid strategies achieve coverage faster, improve community acceptance and reduce costs relative to a Wolbachia replacement programme alone?

Development of this model first involved fitting the model to the literature-derived efficacy estimates, which differ for each suppression method. Study design influences measurement of maximum suppression efficacy, time taken to reach maximum suppression, and time taken to recover to pre-suppression levels, which were all considered when fitting the model and making predictions for a standardised single application programme. When comparing a single suppression application with literature-derived efficacy (with SIT assumed to be 50%) adulticide achieves the greatest suppression but rapidly returns to pre-suppression levels ( Figure 4). Other methods take longer to reach peak effectiveness, but also have longer durations of effectiveness, particularly insect release methods. Male Wolbachia release is the most effective insect-release method but 1gSLT has a longer-lasting effect. Finally, environmental management and larvicides were estimates to result in the least suppression when applied over a short time period as they typically require long consistent periods of application to reach maximum suppression efficacy. With reductions in adult mosquito population size in the range of 8.18 – 43.51%, from a single application all methods of suppression were predicted to remove the ~1–10% mosquito population exacerbation seen in replacement only programmes.

Dynamics of adult mosquito population after a 1-week suppression programme with literature-derived efficacy.

Showing the total adult female mosquito population dynamics under 1-week suppression application with different methods.

Our model predicts that a prior suppression programme of five weekly rounds could reduce the number of Wolbachia mosquitoes required to reach target coverage within 6 months by 16-81% depending on suppression method used ( Figure 5A, comparing literature-derived estimates). All insect release-based suppression methods gave greater reductions in required Wolbachia mosquitoes than conventional methods. This superiority is maintained even if the peak effectiveness is standardised across different methods of suppression ( Figure 5A), suggesting longer-lasting suppression methods are preferable for hybrid approaches. It may seem counterintuitive to use mosquito killing methods at the same time as Wolbachia mosquitoes are being released, but if these mosquito killing methods do not disproportionately affect Wolbachia mosquitoes relative to the wildtype (as assumed in our model), suppression will still reduce the overall number of Wolbachia mosquitoes required for replacement. Among conventional suppression methods, adulticide outperformed environmental management and larvicide with reductions in required release ratio of 39%, 19% and 16% respectively, considering literature-derived efficacy.

Proportional release ratios for <italic toggle="yes">Wolbachia</italic> replacement programmes post suppression.

Replacement with no prior suppression requires a Wolbachia release ratio of 0.0775 to achieve target coverage (95%) within 6 months of first release. A) shows the proportional reduction in required Wolbachia release ratio following a 5-week suppression programme with different methods. B) shows how this proportional reduction in required Wolbachia release ratio declines with increasing rounds of suppression. All suppression methods use literature-derived efficacy except SIT which uses an assumed 50%. 1gSLT and Male Wolbachia release only show two data points because the literature calculated efficacy at 5- and 10-weeks.

Similar results were found when the number of Wolbachia release rounds were reduced (as opposed to reducing the release ratios per round) suggesting programmes could realise this benefit by reducing the number or density of Wolbachia releases. Conversely, programmes could choose to release the same number of Wolbachia mosquitoes, but now at much higher release ratios which would achieve Wolbachia target coverage faster. Insect release suppression methods could be used to decrease the time to target coverage by up to 80%, while conventional methods would only marginally improve speed, or not at all in the case of environmental management ( Table 3).

Time reductions and potential cost savings enabled by a hybrid approach.

* Based on a programme that achieves the TPP minimum target cost of $2.33 per person. Maximum cost savings in the replacement phase assumes suppression has no cost or is an in-kind contribution. ** relative to a baseline programme with nine releases at a relative ratio of 0.09 achieving target coverage (95%) in 215 days.

Method of suppression Maximum percentage cost reduction Maximum cost savings in the replacement phase * ($ per person covered) Percentage reduction in days to achieve target coverage **
1gSLT 18.1 – 50.2 $0.42 – 1.17 83.4
SIT 17.4 – 48.3 $0.41 – 1.13 82.5
Male Wolbachia release 17.4 – 48.3 $0.41 – 1.13 82.5
Environmental management 3.8 – 10.7 $0.09 – 0.25 0.0
Larvicide 6.4 – 17.8 $0.15 – 0.42 11.3
Adulticide 7.7 – 21.4 $0.18 – 0.50 18.5

All hybrid programmes had the potential to offer cost savings over replacement alone, depending on the cost of suppression. Mosquito release suppression methods could reduce costs by 18-50% and conventional methods by 4-21% ( Table 3). If the baseline replacement programme can meet the TPP minimum cost target of $2.33 per person, then these percentage reductions would translate to reductions of $0.41 – $1.17 per person for mosquito suppression methods and $0.09 – 0.50 per person for conventional methods. This, however, does not take into account the costs of suppression. A different interpretation of these results would be: if suppression can be achieved for less than these costs, then a hybrid programme will cost less than replacement alone. Even if the costs are equivalent, a hybrid programme may be considered more acceptable from a community perspective due to prevention of the temporary increase in numbers of biting female mosquitoes.

Longer suppression campaigns give diminishing returns when used as part of a hybrid approach ( Figure 5B). The greatest benefits, in terms of reducing Wolbachia release requirements, are seen within the first few weeks of suppression with decreasing benefits beyond five weekly rounds of suppression. This effect is more pronounced for insect release suppression methods that see most of their benefits delivered from a one- to three-week suppression programme, while for conventional methods there may still be some benefit in continuing suppression for up to eight weeks ( Figure 5B).

The primary role of these modelling results in developing the PPC was to clarify the likely benefits of the hybrid approach. Namely that a hybrid approach could achieve Wolbachia coverage faster, with higher community acceptance and/or at a lower cost. Citing this modelling input the TAG concluded in the PPC that a range of suppression methods can be considered for combination with Wolbachia replacement to achieve potential benefits of faster achievement of coverage by Wolbachia, higher community and programmatic acceptance and lower cost. The draft PPC states that “trials of a hybrid approach would test the expected benefits of conducting suppression followed by replacement and that modelling thus far suggests that suppression methods involving insect releases will generally reduce the intensity of the Wolbachia replacement programme more than other methods”. It was also agreed that modelling would be a useful tool for prioritising intervention combinations for field trials and could be used to directly inform trial design. Finally, the PPC recognises that hybrid approaches may require additional logistical and practical complexities over replacement alone, particularly for mosquito release suppression methods that may require additional regulatory approval. This may mean that rather than hybrid approaches superseding replacement-only approaches, their use may be restricted to areas where replacement-only cannot meet speed, acceptability, and cost goals.

Extension of this modelling work to account for seasonal variations in mosquito population size with the aim of optimising the seasonal timing of replacement-only and hybrid approaches was also requested. Since the original PPC meeting, our model has been expanded to include a typical seasonal profile. Our model assumes mosquito population sizes closely follow variation in precipitation with a 41-day lag and a peak population size ~ 3 times the dry season minimum, consistent with various field observations 45 (Supplementary Figure 4, seasonality function sensitivity analysis in Supplementary Table 2).

We predict that the optimal time to begin replacement-only or hybrid programmes is just before the seasonal lowest point in mosquito abundance ( Figure 6). The release ratio required for Wolbachia fixation in a replacement only programme mirrors precipitation (and thus wild type mosquito population) dynamics, with a short lag.

Seasonal variation in hybrid approach effectiveness.

Minimum release ratio for a Wolbachia replacement programme to reach target coverage (95%) within 6 months of first release after a 5-week suppression programme with different methods. Required release ratio is relative to the annual mean (long-term average) wild type mosquito population. Seasonality in mosquito abundance is simulated by precipitation-dependent larval development.

When mosquito populations fluctuate throughout the year, the timing of Wolbachia replacement has a large effect on the number of Wolbachia mosquitoes that need to be released to reach fixation. Starting the replacement programme at the optimum time can reduce the number of Wolbachia mosquitoes by 65.26% compared to the least optimal time. The seasonal scenario also follows the prioritisation of suppression methods observed in the non-seasonal analysis when using a hybrid suppression-then-release approach with male Wolbachia release most effective and larvicide least effective. Furthermore, at the optimal time, in the dry season, the hybrid approach reduces the required Wolbachia release ratio by up to 93.94% compared to replacement alone, whereas at the least optimal time, in the wet season, the hybrid approach only reduces the required Wolbachia release ratio by up to 64.21%. Finally, because of their delayed effects the insect-release suppression methods (1gSLT, SIT, Male Wolbachia release) allow a hybrid strategy to remain effective for longer in the early stages of the wet season, so may be a better choice for areas where the timing of mosquito seasonal cycles is less predictable.

Discussion

Mathematical and geostatistical models can make important quantitative and qualitative contributions when developing TPPs and PPCs. Here we show that models can: i) identify important trade-offs, such as the time taken for Wolbachia to reach target coverage and the temporary exacerbation in the mosquito population, ii) quantify threshold criteria, such as the $2.33 per person Wolbachia replacement cost target, iii) predict characteristics of a product in new areas and at broader scales than it is currently implemented, such as to meet the WHO 25% global burden reduction targets and iv) understand synergies and antagonisms between combinations of products that have not yet been tested, such as a hybrid suppress then replace approach.

Broad community acceptability of Wolbachia replacement will clearly be a critical aspect of achieving implementation at the scale envisioned by these TPP and PPC documents. The success of current replacement programmes has been underpinned by extensive community engagement activities 46, 47 and other countries (Singapore 18 and China 17 ) have chosen to use Wolbachia for suppression only, in part, due to concerns over any increases in mosquito abundance. Here we show temporary increases in mosquito abundance can be minimised or avoided entirely by using lower Wolbachia release ratios, timing releases to coincide with the dry season, or conducting a prior suppression campaign in a hybrid approach. These steps may, however, involve additional programmatic complexity and likely cost. More work is needed to better understand how mosquito abundance relates to community acceptability in different contexts and how such barriers can be overcome with different release intensities, timings, and hybrid approaches. One alternative use case, considered in the TPP, is to conduct longer lower intensity Wolbachia replacement releases to reach an initial target coverage of 50%, after which gaps in coverage are gradually infilled by movement of mosquitoes and auto-dissemination of Wolbachia. These lower density releases may have significant cost advantages and could be a more acceptable method of dissemination over broad areas where faster implementation is a lower priority. Such a strategy would, however, take longer, be dependent on patterns of mosquito movement and may be limited by environmental barriers to mosquito spread 48 . Development of spatial models of mosquito movement and dengue spread could help identify where additional release points may be necessary, target initial release points to high-risk areas and quantify the collateral benefit in disease reduction in neighbouring areas 49, 50 .

Cost continues to be a barrier to wider adoption of Wolbachia replacement when its high costs but long-term benefits are compared to lower cost but short acting suppression methods 27 . A key strength of our analysis was to link TPP cost targets to conservative estimates of averted costs based on direct medical costs and emergency vector control expenditure over limited timeframes. This was critical to identify geographic differences in cost targets between, but also within countries. Pairing this analysis with high resolution global burden and cost maps identified cost targets that are compatible with wider international goals and equitable across a range of settings 22 . Work is currently underway to validate our approximation of willingness to pay for Wolbachia replacement through surveys targeted to key stakeholders in state and federal governments. The maps and models generated in this work could be adapted for planning national Wolbachia replacement campaigns and, in particular, could inform how re-use of release resources, variable pricing models, financing and slower release campaigns could be used to meet the TPP cost targets in even the most challenging countries 27 . Some of this functionality is already available in the freely available Wolbachia Decision-Support tool ( https://wolbachia-tool.netlify.app/tool#map) which makes use of the outputs of this analysis along with other geospatial layers.

This analysis predicts that Wolbachia replacement releases over 924,557km 2 could lead to a 25% reduction in dengue burden globally, averting US$3.05 billion (2.62 – 3.96) worth of medical and outbreak response costs per annum. This may appear ambitious for a minimum product but can be understood in the following context. Although the lowest averted cost/person covered in this area is predicted to be US$2.33, all other release areas could support substantially higher costs ( Figure 3). This means that a Wolbachia replacement product would still meet the TPP targets if initial programme costs were higher and if the product has the potential to reduce costs down to the $2.33 target. It is important to note that the TPP only requires that the product is suitable and available everywhere within the target areas, not that it is necessarily implemented in all suitable areas. In practice, commercial considerations, including the need to build capacity, to access funding that is incremental to the current routine control budgets, and to compete and combine with other methods, will limit the rate of uptake and the ultimate scale of deployment achieved.

Hybrid approaches offer one promising option for increasing speed, reducing cost and increasing the acceptability of Wolbachia replacement. The models presented here and the wider evidence provided to the PPC support field trials of hybrid approaches as a next logical step. These models can guide the prioritisation of suppression methods, trial sample size calculations and suggest how effectiveness should be measured. We predict that insect-based suppression methods (1gSLT, SIT, Male Wolbachia release) will be more effective than conventional suppression tools, but also outline a limited cost window which may be challenging for insect-based suppression methods to achieve. Investment in new infrastructure to conduct insect-based suppression may not be justified for a one-off suppression, but between overlapping resource requirements for suppression and replacement, ongoing use post-replacement (e.g., outbreak control or to achieve dengue elimination) and a continued drive to lower costs of mosquito suppression 51 , this investment cost may be justified. Intervention developers and countries must ultimately decide how to balance cost and efficacy when considering hybrid approaches. Timing replacement to coincide with the seasonal low point of mosquito abundance is an alternative low-cost hybrid approach and would be a useful addition to trials of hybrid approaches. More generally, our results also suggest that suppression methods that have a longer residual effect are likely to be more beneficial in a hybrid approach. This would suggest some emerging vector control methods, including targeted indoor residual spraying (TIRS) 52, 53 and Oxitec’s second-generation Friendly™ mosquito technology 54 that allows male survival, would also be strong candidates for a hybrid approach and should also be considered for inclusion in modelling and potentially in hybrid field trials.

These models and the results they generate are not without their limitations and clear communication of these limitations was an important part of their use for the TPP and PPC. Our models of Wolbachia replacement do not include any spatial, temporal (beyond seasonal) or stochastic heterogeneities that may mean our model overestimates the speed to achieve fixation and target coverage, particularly in the latter stages (Supplementary Figure 6). We also do not account for a natural wild type egg bank emergence which can dilute Wolbachia release ratios, possibly accounting for the low RR values estimated in these analyses. The Wolbachia replacement system targets Ae. aegypti, which is the major vector of dengue, but Ae. albopictus can be locally important and the need for a Wolbachia replacement product to control both species is not considered in this modelling. Ae. albopictus and other mosquito species are assumed to be controlled by routine vector control programmes which would be ongoing in parallel with any Wolbachia Ae. aegypti replacement or hybrid control programmes. There remain large gaps globally in data on the cost of dengue treatment and prevention and no comprehensive cost estimates for Zika, chikungunya or yellow fever, all of which Wolbachia will provide some efficacy against. Cost estimates are therefore generally conservative and should not replace primary data on willingness to pay or more detailed cost-benefit analyses when considering programmes in any one given country. We also recognise that only one modelling group was included for this TPP and PPC and that inclusion of multiple modelling teams can help better represent the structural uncertainty of models and their interpretation when deciding between policy options 55 . Fitting each suppression method to literature reported effectiveness estimates was challenging due to incomparable ways in which suppression was implemented and evaluated, and therefore the modelling outputs for each method may not be representative. In particular, our chosen source of evidence for adulticide suppression reported a 94% followed by a return to pre-suppression population within 7 days 42 , a rate of rebound that our model was unable to replicate from newly emerging adult mosquitoes alone, thus we had to assume that the 94% effectiveness was only achieved in a fraction of the overall mosquito population. Moreover, this efficacy is higher than generally expected for outdoor space-spraying in urban environments in practice 56 . Our hybrid approach results also assume that the suppression method acts independently of any other forms of vector control already in use in the area. Suppression will likely be reduced if the method of suppression (or similar methods that use the same modes of action or insecticides) is already routinely used.

In this paper we show the value dedicated modelling research can add to the development of TPPs and PPCs. By making the data, code and fitted models freely available to accompany the TPP for Wolbachia replacement and PPC for the hybrid approach, product developers are able to continue to use and adapt them to steer the development of a range of Wolbachia products to meet the rising challenge of global dengue control.

 Ethics

No ethical approval was necessary as all data used was publicly available.

 Data availability

GitHub. Using models and maps to inform Target Product Profiles and Preferred Product Characteristics: the example of Wolbachia replacement. DOI: https://github.com/katietiley/Wolbachia_TPP_PPC.git

This project contains the following data:

All data and code used for each of the models used in this analysis and the predictions made by these models

Supplementary file 1 and all supplementary figures referred to in the main text

Data is available under the MIT Licence.

 Acknowledgements

We acknowledge Marjorie Opuni-Akuamoa for her contributions towards the cost of dengue vector control systematic review. We also acknowledge Raman Velayudhan for his guidance on experimental design, communication of results and for facilitating the modelling inputs to the TPP and PPC processes.

 World Health Organization: Global vector control response 2017–2030.Geneva; 2017. Reference Source Izmirly AM Alturki SO Alturki SO : Challenges in Dengue Vaccines Development: Pre-existing Infections and Cross-Reactivity. Front Immunol. 2020;11:1055. 32655548 10.3389/fimmu.2020.01055 7325873 de Lima Cavalcanti TYV Pereira MR de Paula SO : A Review on Chikungunya Virus Epidemiology, Pathogenesis and Current Vaccine Development. Viruses. 2022;14(5):969. 35632709 10.3390/v14050969 9147731 Kolimenakis A Heinz S Wilson ML : The role of urbanisation in the spread of Aedes mosquitoes and the diseases they transmit—A systematic review. PLoS Negl Trop Dis. 2021;15(9):e0009631. 34499653 10.1371/journal.pntd.0009631 8428665 Brady OJ Hay SI : The Global Expansion of Dengue: How Aedes aegypti Mosquitoes Enabled the First Pandemic Arbovirus. Annu Rev Entomol. 2020;65:191–208. 31594415 10.1146/annurev-ento-011019-024918 Patil PB Dasgupta SK Gorman K : Elimination of a closed population of the yellow fever mosquito, Aedes aegypti, through releases of self-limiting male mosquitoes. PLoS Negl Trop Dis. 2022;16(5):e0010315. 35576193 10.1371/journal.pntd.0010315 9135344 Walker T Johnson PH Moreira LA : The wMel Wolbachia strain blocks dengue and invades caged Aedes aegypti populations. Nature. 2011;476(7361):450–3. 21866159 10.1038/nature10355 Caspari E Watson GS : On the Evolutionary Importance of Cytoplasmic Sterility in Mosquitoes. Evolution (N Y). 1959;13(4):568–570. 10.2307/2406138 Yakob L Funk S Camacho A : Aedes aegypti control through modernized, integrated vector management. PLoS Curr. 2017;9. 28286698 10.1371/currents.outbreaks.45deb8e03a438c4d088afb4fafae8747 5319873 Yakob L Alphey L Bonsall MB : Aedes aegypti Control: The Concomitant Role of Competition, Space and Transgenic Technologies. J Appl Ecol. 2008;45(4):1258–65. 10.1111/j.1365-2664.2008.01498.x Pinto SB Riback TIS Sylvestre G : Effectiveness of Wolbachia-infected mosquito deployments in reducing the incidence of dengue and other aedes-borne diseases in Niterói, Brazil: A quasi-experimental study. PLoS Negl Trop Dis. 2021;15(7):e0009556. 34252106 10.1371/journal.pntd.0009556 8297942 Ross PA Robinson KL Yang Q : A decade of stability for wMel Wolbachia in natural Aedes aegypti populations. PLoS Pathog. 2022;18(2):e1010256. 35196357 10.1371/journal.ppat.1010256 8901071 O’Reilly KM Hendrickx E Kharisma DD : Estimating the burden of dengue and the impact of release of wMel Wolbachia-infected mosquitoes in Indonesia: a modelling study. BMC Med. 2019;17(1):172. 31495336 10.1186/s12916-019-1396-4 6732838 Utarini A Indriani C Ahmad RA : Efficacy of Wolbachia-Infected Mosquito Deployments for the Control of Dengue. N Engl J Med. 2021;384(23):2177–86. 34107180 10.1056/NEJMoa2030243 8103655 World Mosquito Programme: World Mosquito Programme Global Progress. Nazni WA Hoffmann AA NoorAfizah A : Establishment of Wolbachia strain wAlbB in Malaysian populations of Aedes aegypti for dengue control. Curr Biol. 2019;29(24):4241–4248.e5. 31761702 10.1016/j.cub.2019.11.007 6926472 Zheng X Zhang D Li Y : Incompatible and sterile insect techniques combined eliminate mosquitoes. Nature. 2019;572(7767):56–61. 31316207 10.1038/s41586-019-1407-9 Singapore National Environment Agency: Project Wolbachia Singapore. 2019; [cited 2019 Sep 30]. Mains JW Kelly PH Dobson KL : Localized Control of Aedes aegypti (Diptera: Culicidae) in Miami, FL, via Inundative Releases of Wolbachia-Infected Male Mosquitoes. J Med Entomol. 2019;56(5):1296–303. 31008514 10.1093/jme/tjz051 Ritchie SA Staunton KM : Reflections from an old Queenslander: Can rear and release strategies be the next great era of vector control? Proc Biol Sci. Royal Society Publishing; 2019;286(1905):20190973. 31238839 10.1098/rspb.2019.0973 6599993 World Health Organization: Links to WHO target product profiles (TPPs) and product profile characteristics (PPCs). Reference Source World Health Organization: Ending the neglect to attain the Sustainable Development Goals: a road map for neglected tropical diseases 2021-2030. Geneva;2021. Reference Source Bhatt S Gething PW Brady OJ : The global distribution and burden of dengue. Nature. 2013;496(7446):504–7. 23563266 10.1038/nature12060 3651993 Shepard DS Undurraga EA Halasa YA : The global economic burden of dengue: a systematic analysis. Lancet Infect Dis. 2016;16(8):935–41. 27091092 10.1016/S1473-3099(16)00146-8 The World Bank: GDP deflator (base year varies by country).2022; [cited 2022 Jan 1]. Reference Source The World Bank: Official exchange rate (LCU per US$, period average).2022; [cited 2022 Jan 1]. Reference Source Brady OJ Kharisma DD Wilastonegoro NN : The cost-effectiveness of controlling dengue in Indonesia using wMel Wolbachia released at scale: a modelling study. BMC Med. 2020;18(1):186. 32641039 10.1186/s12916-020-01638-2 7346418 Ryan PA Turley AP Wilson G : Establishment of wMel Wolbachia in Aedes aegypti mosquitoes and reduction of local dengue transmission in Cairns and surrounding locations in northern Queensland, Australia [version 2; peer review: 2 approved]. Gates Open Res. 2020;3:1547. 31667465 10.12688/gatesopenres.13061.2 6801363 Smith JM Slatkin M : The Stability of Predator-Prey Systems. Ecology. 1973;54(2):384–91. 10.2307/1934346 Otero M Solari H Schweigmann N : A Stochastic Population Dynamics Model for Aedes aegypti: Formulation and Application to a City with Temperate Climate. Bull Math Biol. 2006;68(8):1945–74. 16832731 10.1007/s11538-006-9067-y Marinho R Beserra E Bezerra-Gusmão M : Effects of temperature on the life cycle, expansion, and dispersion of Aedes aegypti (Diptera: Culicidae) in three cities in Paraiba, Brazil. J Vector Ecol. 2016;41(1):1–10. 27232118 10.1111/jvec.12187 Masters SW Knapek KJ Kendall LV : Rearing Aedes aegypti Mosquitoes in a Laboratory Setting. Lab Animal Sci Prof. 2020;55(6):42–45. 33106772 7585283 Trpis M : Breeding of Aedes aegypti and A. simpsoni under the escarpment of the Tanzanian plateau. Bull World Health Organ. 1972;47(1):77–82. 4538907 2480807 Couret J Dotson E Benedict M : Temperature, Larval Diet, and Density Effects on Development Rate and Survival of Aedes aegypti (Diptera: Culicidae). PLoS One. 2014;9(2):e87468. 24498328 10.1371/journal.pone.0087468 3911954 Bellows T : The descriptive properties of some models for density dependence. J Anim Ecol. 1981;50(1):139–56. 10.2307/4037 Winskill P Harris AF Morgan SA : Genetic control of Aedes aegypti: Data-driven modelling to assess the effect of releasing different life stages and the potential for long-term suppression. Parasit Vectors. 2014;7(1):68. 24524678 10.1186/1756-3305-7-68 3944930 Joubert DA Walker T Carrington LB : Establishment of a Wolbachia Superinfection in Aedes aegypti Mosquitoes as a Potential Approach for Future Resistance Management. PLoS Pathog. 2016;12(2):e1005434. 26891349 10.1371/journal.ppat.1005434 4758728 Carvalho DO McKemey AR Garziera L : Suppression of a field population of Aedes aegypti in Brazil by sustained release of transgenic male mosquitoes. PLoS Negl Trop Dis. 2015;9(7):e0003864. 26135160 10.1371/journal.pntd.0003864 4489809 Ching NL : Wolbachia-mediated sterility suppresses Aedes aegypti populations in the urban tropics.2021. 10.1101/2021.06.16.21257922 Andersson N Nava-Aguilera E Arosteguí J : Evidence based community mobilization for dengue prevention in Nicaragua and Mexico ( Camino Verde, the Green Way): Cluster randomized controlled trial. BMJ. 2015;351:h3267. 26156323 10.1136/bmj.h3267 4495677 Hustedt JC Doum D Keo V : Field efficacy of larvivorous fish and pyriproxyfen combined with community engagement on dengue vectors in Cambodia: A randomized controlled trial. Am J Trop Med Hyg. 2021;105(5):1265–76. 34491225 10.4269/ajtmh.20-1088 8592206 Mani TR Arunachalam N Rajendran R : Efficacy of thermal fog application of deltacide, a synergized mixture of pyrethroids, against Aedes aegypti, the vector of dengue. Trop Med Int Health. 2005;10(12):1298–304. 16359411 10.1111/j.1365-3156.2005.01522.x Rohatgi A : WebPlotDigitizer V4.5. Pacifica, California, USA;2021. Reference Source World Mosquito Programme: Our Wolbachia method demonstrated at city-wide scale for the first time.2022; [cited 2022 Aug 25]. Reference Source Lana RM Carneiro TGS Honório NA : Seasonal and nonseasonal dynamics of Aedes aegypti in Rio de Janeiro, Brazil: fitting mathematical models to trap data. Acta Trop. 2014;129(1):25–32. 23933186 10.1016/j.actatropica.2013.07.025 Kolopack PA Parsons JA Lavery JV : What Makes Community Engagement Effective?: Lessons from the Eliminate Dengue Program in Queensland Australia. PLoS Negl Trop Dis. 2015;9(4):e0003713. 25875485 10.1371/journal.pntd.0003713 4395388 Costa GB Smithyman R O’Neill SL : How to engage communities on a large scale? Lessons from World Mosquito Program in Rio de Janeiro, Brazil [version 2; peer review: 2 approved, 1 approved with reservations]. Gates Open Res. 2020;4:109. 33103066 10.12688/gatesopenres.13153.2 7569240 Schmidt TL Filipović I Hoffmann AA : Fine-scale landscape genomics helps explain the slow spatial spread of Wolbachia through the Aedes aegypti population in Cairns, Australia. Heredity (Edinb). 2018;120(5):386–95. 29358725 10.1038/s41437-017-0039-9 5889405 Dorigatti I McCormack C Nedjati-Gilani G : Using Wolbachia for Dengue Control: Insights from Modelling. Trends Parasitol. 2018;34(2):102–13. 29183717 10.1016/j.pt.2017.11.002 5807169 Brady OJ Kucharski AJ Funk S : Case-area targeted interventions (CATI) for reactive dengue control: Modelling effectiveness of vector control and prophylactic drugs in Singapore. PLoS Negl Trop Dis. 2021;15(8):e0009562. 34379641 10.1371/journal.pntd.0009562 8357181 Gilbert JA Melton L : Verily project releases millions of factory-reared mosquitoes. Nat Biotechnol. 2018;36(9):781–782. 30188519 10.1038/nbt0918-781a Cavany SM España G Lloyd AL : Optimizing the deployment of ultra-low volume and targeted indoor residual spraying for dengue outbreak response. PLoS Comput Biol. 2020;16(4):e1007743. 32310958 10.1371/journal.pcbi.1007743 7200023 Vazquez-Prokopec GM Montgomery BL Horne P : Combining contact tracing with targeted indoor residual spraying significantly reduces dengue transmission. Sci Adv. 2017;3(2):e1602024. 28232955 10.1126/sciadv.1602024 5315446 Oxitec: Oxitec Completes Successful Scale-Up of Next-Generation Friendly TM Mosquito Technology in Brazil Designed to Accelerate Global Impact Against Disease-Spreading Invasive Pests.(Press release).2022; [cited 2022 Aug 30]. Reference Source Brady OJ Slater HC Pemberton-Ross P : Role of mass drug administration in elimination of Plasmodium falciparum malaria: a consensus modelling study. Lancet Glob Health. 2017;5(7):e680–e687. 28566213 10.1016/S2214-109X(17)30220-6 5469936 Esu E Lenhart A Smith L : Effectiveness of peridomestic space spraying with insecticide on dengue transmission; systematic review. Trop Med Int Health. 2010;15(5):619–31. 20214764 10.1111/j.1365-3156.2010.02489.x 10.21956/gatesopenres.15611.r34325 Reviewer response for version 1 Ross Perran 1 Referee https://orcid.org/0000-0001-7645-7523 The University of Melbourne, Melbourne, Victoria, Australia

Competing interests: No competing interests were disclosed.

 3 8 2023 Copyright: © 2023 Ross P 2023 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation reject

General comments

Mosquitoes infected with Wolbachia bacteria are being used to control the spread of arboviruses including dengue. These programmes involve the release of lab-reared mosquitoes at a large scale into the natural population. After reaching a high frequency, the Wolbachia infection can be self-sustaining in the mosquito population due to its maternal transmission and ability to induce cytoplasmic incompatibility, and this can provide ongoing protection against dengue. The World Health Organization (WHO) has recently drafted a Target Product Profile (TPP) which includes a set of desired characteristics for Wolbachia strains and release programmes that would be required to meet WHO targets of reducing the global burden of dengue by 25%

This manuscript addresses aspects of the TPP using models comparing variables that affect the success of Wolbachia population replacement and mapping of locations where Wolbachia releases are likely to provide the most benefit. While there is value in the latter, I have concerns with their Wolbachia replacement models and hybrid “suppress then replace” models. The Wolbachia replacement model does not provide information that isn’t already clear from previous modelling studies and analysis of the field trials, while the hybrid suppression then replace component makes inappropriate comparisons between different suppression interventions and does not consider the potentially substantial costs of these interventions.

In the first part of the paper, the authors model Wolbachia infection frequencies and mosquito population sizes when there are different numbers of releases. Unsurprisingly, more releases and higher release ratios equals faster population replacement and a temporarily higher number of adult mosquitoes. The importance of these variables has been taken into account when planning releases and they have also been addressed in other models (e.g. Hancock et al. 2011a and 2011b). Another outcome of the model is that complete Wolbachia coverage can be achieved in under a year, but this has already been demonstrated directly through several field trials, so it is unclear what value the model here adds.

The authors emphasize the importance of balancing the speed of Wolbachia replacement with the potentially negative outcomes of temporarily increasing the mosquito population. However, I am unsure if this is supported by the outcomes of their models, which show a maximum temporary increase in the population size of under 20%. Is this likely to be noticeable by the community? And even with this increase, won’t there be significantly more nuisance biting at other times of the year (assuming that Wolbachia releases take place when mosquito populations are low)? If this increase is likely to be a concern, then surely the use of a hybrid suppress then replace approach, which the authors discuss later in the paper, would also be a concern due to the much higher release ratios required for releases of incompatible or sterile males. While males don’t bite, they can also be regarded as a nuisance and many people will not be able to distinguish between the sexes.

The final set of results addresses the utility of a hybrid “suppress then replace” approach, where the mosquito population is suppressed through different tools (insecticides, sterile male releases etc.) prior to population replacement releases. This will make population replacement easier as there are fewer mosquitoes to replace. Previous studies have used this approach, for example, the very first releases of wMel involved prior suppression of the population through the removal of larval habitats (Hoffmann et al. 2011). But a key question is cost effectiveness. The authors perform a cost analysis and conclude that a hybrid approach could be cost saving, but they don’t include the cost of the suppression itself for any of the approaches being compared, making their conclusions baseless. Some of the approaches they compare can be quite expensive in their own right. The incompatible insect technique for instance requires sex sorting of mosquitoes which is very labour intensive if done mechanically or very expensive if using automated systems, and mosquitoes need to be reared at large scales to achieve suppression.

The authors then use models to test the effect of different suppression interventions on mosquito population sizes, and later, the effect on the release ratio of Wolbachia-infected mosquitoes required for successful Wolbachia establishment. These approaches have different effects (e.g. adulticides kill off adults quickly but the population bounces back quickly, while incompatible/sterile insect approaches have slower but longer term population suppression), which is a reasonable point to make. But the authors then compare these approaches directly using literature estimates of efficacy. Parameters were taken from a single study for each approach and these studies are in no way comparable to each other. They were in different environments with different populations of mosquitoes and were done at different geographic scales, time periods and intensities. There are also issues with using a single study to represent the expected efficacy for a typical suppression programme. I just don't see how it's reasonable to compare these approaches directly and conclude that  one is more effective when the approaches have been applied at different intensities and durations.

In summary, while the paper addresses an important issue, in my view the entomological models are problematic and/or uninformative, and the paper is not suitable for indexing without substantial fundamental changes.

Specific comments

Abstract - results – These quantitative thresholds are adjusted to data from the wMel releases in Wolbachia and they should not be extrapolated to Wolbachia releases in other locations, given that Wolbachia releases can have vastly different outcomes depending on the environment

Abstract – results - Suppression interventions will reduce the number of mosquitoes required for replacement, but this statement ignores the very high numbers of mosquitoes that are needed for the suppression itself (if using incompatible males, which the authors conclude is the most effective approach).

Intro paragraph 2 – Not all mosquitoes infected with Wolbachia show reduced virus dissemination- be more specific about the species and Wolbachia strain.

Intro paragraph 3  – The paragraph is a bit of an oversimplification- Wolbachia doesn’t always reach fixation even if very high frequencies are reached, for instance, due to maternal transmission failure. There is also now evidence from field trials showing that Wolbachia frequencies can fluctuate seasonally or even decline to zero even after reaching near-fixation depending on the environment.

Intro paragraph 4 – “widespread, long-term effectiveness” is true in some locations but there are also cases where Wolbachia releases have failed (see above), meaning that there are likely to be environmental constraints on where Wolbachia infections can successfully establish

Methods – entomological model paragraph 3 – this seems to only cover a scenario where Wolbachia-infected mosquitoes are released as eggs. Please provide some context in the introduction as egg releases are not mentioned before here. The models are built on the assumption of egg releases, and the fact that Wolbachia-infected larvae are initially separate from wild larvae is an important component, but there is no justification for this or acknowledgement of other types of release. The authors discuss population suppression through male releases, but this will require adult releases.

Methods- suppression – “The efficacy of each method was based on evidence sourced from the literature” - this section is quite subjective and there is no information about how the authors searched for studies or selected them aside from having a preference for large randomised controlled studies.

Methods – suppression – I would prefer if the authors didn’t use “larvicides” here- guppies are not larvicides – they are predators. “Larvicides” typically refers to chemical insecticides or bacterial pathogens like Bt.

Figure 1 – The model shows that the mosquito population size permanently decreases after Wolbachia establishment, but I’m not aware of any evidence for this from field release data. For instance, data from the releases in Yogyakarta shows that Wolbachia releases had minimal impact on the population size: https://journals.plos.org/plosntds/article?id=10.1371/journal.pntd.0010284

Supp figure 1 – I’m not sure why the authors only considered costs up to 20% - data from release programs have estimated costs of around 30% for wMel (Hoffmann et al. 2011) and this is likely to vary substantially depending on the environment.

Table S1 – Why is SIT included here if the authors found no studies to base its efficacy on? It is true that few SIT release programmes in Ae. aegypti have been published, but there is at least one (e.g. de Castro Poncio et al. 2021)

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

No

Are all the source data underlying the results available to ensure full reproducibility?

Yes

Is the study design appropriate and is the work technically sound?

No

Are the conclusions drawn adequately supported by the results?

No

Are sufficient details of methods and analysis provided to allow replication by others?

Yes

Reviewer Expertise:

I am actively involved in research on Wolbachia population replacement programmes in Aedes mosquitoes, including lab experiments and analysis of field data. My comments focus on Wolbachia and mosquito biology and the outcomes of release programmes. I am not a statistician and am not qualified to comment on the equations behind the models, only their biological relevance.

I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

References : Strategies for introducing Wolbachia to reduce transmission of mosquito-borne diseases. PLoS Negl Trop Dis .2011;5(4) : 10.1371/journal.pntd.0001024 e1024 21541357 10.1371/journal.pntd.0001024 : Population dynamic models of the spread of Wolbachia. Am Nat .2011;177(3) : 10.1086/658121 323-33 21460541 10.1086/658121 : Successful establishment of Wolbachia in Aedes populations to suppress dengue transmission. Nature .2011;476(7361) : 10.1038/nature10356 454-7 21866160 10.1038/nature10356 : Novel Sterile Insect Technology Program Results in Suppression of a Field Mosquito Population and Subsequently to Reduced Incidence of Dengue. J Infect Dis .2021;224(6) : 10.1093/infdis/jiab049 1005-1014 33507265 10.1093/infdis/jiab049 Tiley Katie

Competing interests: No competing interests were disclosed.

 18 4 2024

2.1. General comments:

2.1.1. Mosquitoes infected with Wolbachia bacteria are being used to control the spread of arboviruses including dengue. These programmes involve the release of lab-reared mosquitoes at a large scale into the natural population. After reaching a high frequency, the Wolbachia infection can be self-sustaining in the mosquito population due to its maternal transmission and ability to induce cytoplasmic incompatibility, and this can provide ongoing protection against dengue. The World Health Organization (WHO) has recently drafted a Target Product Profile (TPP) which includes a set of desired characteristics for Wolbachia strains and release programmes that would be required to meet WHO targets of reducing the global burden of dengue by 25%.

2.1.1. We thank the reviewer for the time taken to read and thoroughly respond to this work which informed the World Health Organisation (WHO) Target Product Profile (TPP) and Preferred Product Characteristics (PPC) documents for Wolbachia replacement technology contributing to the goal of reaching the WHO targets of reducing global burden of dengue by 25%.

2.1.2. This manuscript addresses aspects of the TPP using models comparing variables that affect the success of Wolbachia population replacement and mapping of locations where Wolbachia releases are likely to provide the most benefit. While there is value in the latter, I have concerns with their Wolbachia replacement models and hybrid “suppress then replace” models. The Wolbachia replacement model does not provide information that isn’t already clear from previous modelling studies and analysis of the field trials, while the hybrid suppression then replace component makes inappropriate comparisons between different suppression interventions and does not consider the potentially substantial costs of these interventions.

2.1.2. We agree with the reviewer that previous modelling studies and field trials have reported similar findings to our Wolbachia replacement model. We consider our model outcomes being consistent with these previous published findings to be a strength, not a weakness, as this provides supporting evidence for the appropriateness of our methodology, and this supported the decision of the Technical Advisory Group (TAG) to include these specific criteria in the TPP. Specifically, the value of our Wolbachia replacement model and hybrid suppress-then-replace model is in quantifying the differences in mosquito population dynamics compared across different variables, such as different levels of release ratio or release number. Quantifying these measures is important because it allows us to link different scenarios to cost estimates and enables more informed decisions when addressing trade-offs between different scenarios, both essential for producing reliable values for the TPP and PPC documents.

Furthermore, we would like to make clear that the hybrid suppress-then-replace scenarios are presented to enable public health departments and organisations to better integrate their current vector control efforts with a novel Wolbachia replacement programme. These scenarios demonstrate the potential benefits of leveraging vector control interventions that are already in use (or could be with minor adaptations) during the transition to Wolbachiareplacement and we acknowledge that the additional cost of building the capabilities to implement new methods of suppression solely for the purpose of reducing the cost of Wolbachia replacement is unlikely to reduce overall cost. Since current vector control practices are already budgeted for in routine vector control programmes, we assume no additional cost for using conventional vector control methods prior to implementing Wolbachia replacement. That the suppression scenarios outlined in our study are most applicable to circumstances where suppression is already in place has been further clarified in the fifth paragraph of the results section under the subheading “Exploring the dynamics of a hybrid ‘suppress then replace’ approach”: 

“…than a hybrid programme will cost less than replacement alone. This might refer to specific circumstances when suppression might be achieved at negligible cost, for example, if it’s already part of a vector control programme. These are ideal situations where the resources for suppression are already in place, however, in most cases the potential savings achieved by a hybrid suppress then replace programme won’t justify the additional cost of implementing a novel suppression programme.”

2.1.3. In the first part of the paper, the authors model Wolbachia infection frequencies and mosquito population sizes when there are different numbers of releases. Unsurprisingly, more releases and higher release ratios equals faster population replacement and a temporarily higher number of adult mosquitoes. The importance of these variables has been taken into account when planning releases and they have also been addressed in other models (e.g. Hancock et al. 2011a and 2011b). Another outcome of the model is that complete Wolbachia coverage can be achieved in under a year, but this has already been demonstrated directly through several field trials, so it is unclear what value the model here adds.

2.1.3. Again, we consider the consistency of our findings with the existent literature to be an advantage, especially when considering fundamental dynamics mentioned by the reviewer; specifically, that more releases and higher release ratios leading to faster population replacement and that population replacement is achievable in under a year (e.g. Hancock et al, 2011a, 2011b). This is also the case with our finding that Wolbachia replacement can cause mosquito population exacerbation, previously modelled by Yakob et al (2017). The strength of this evidence from a variety of methods has been further clarified in the manuscript discussion: 

“The population dynamics shown in our model are consistent with previous modelling and field studies showing temporary population exacerbation and successful fixation within one year (44-46); this evidence supports the credibility of our findings quantifying the difference between scenarios with varying release ratio and release number.”

This manuscript aims to highlight the value modelling specifically adapted to the questions raised by the TAG in the TPP process can bring. These questions inevitably span a range of field, laboratory and data analysis studies which are also cited as evidence for the criteria that are set. The additional value of the bespoke modelling we did for this showed the trade-offs between criteria, e.g. yes replacement is achievable within 6 months, but what release ratios are necessary to achieve this and what consequences do they have for cases and community acceptability?

2.1.4. The authors emphasise the importance of balancing the speed of Wolbachia replacement with the potentially negative outcomes of temporarily increasing the mosquito population. However, I am unsure if this is supported by the outcomes of their models, which show a maximum temporary increase in the population size of under 20%. Is this likely to be noticeable by the community? And even with this increase, won’t there be significantly more nuisance biting at other times of the year (assuming that Wolbachia releases take place when mosquito populations are low)? If this increase is likely to be a concern, then surely the use of a hybrid suppress then replace approach, which the authors discuss later in the paper, would also be a concern due to the much higher release ratios required for releases of incompatible or sterile males. While males don’t bite, they can also be regarded as a nuisance and many people will not be able to distinguish between the sexes.

2.1.4. Any temporary population increase has the potential to decrease rates of community acceptance and undermine community engagement. This was an issue specifically raised as a concern by the TAG with our models quantifying the trade-offs between faster time to fixation but with higher temporary exacerbation when release ratios are increased. The temporary population increase shown in our model could be considered small (maximum exacerbation 20%) but it is important to quantify these levels for obtaining informed consent from communities. Furthermore, despite not being modelled here, even temporary minor increases could be epidemiologically important if Wolbachia is implemented at times of high transmission intensity, e.g. as an outbreak response. We agree with the reviewer that some suppression techniques would require additional mosquito releases, such as the male Wolbachiarelease or sterile insect technique, however, these would be accompanied by substantial separate community engagement activities and, since we recommend the hybrid suppress-then-release scenarios to complement existing vector control activities, likely already have taken place ahead of implementing Wolbachia replacement. Specifically, building community support through education campaigns would involve emphasising that males do not bite, as the reviewer comments, and that the exacerbation will not increase risk of infection. This has been further clarified in the manuscript: 

“The mosquito-release suppression techniques (i.e., 1gSLT, SIT, Male Wolbachia release) may also lead to a temporary increase in mosquito population and would require separate community engagement activities to emphasise that the released males do not bite or increase the risk of infection.”

2.1.5. The final set of results addresses the utility of a hybrid “suppress then replace” approach, where the mosquito population is suppressed through different tools (insecticides, sterile male releases etc.) prior to population replacement releases. This will make population replacement easier as there are fewer mosquitoes to replace. Previous studies have used this approach, for example, the very first releases of wMel involved prior suppression of the population through the removal of larval habitats (Hoffmann et al. 2011). But a key question is cost effectiveness. The authors perform a cost analysis and conclude that a hybrid approach could be cost saving, but they don’t include the cost of the suppression itself for any of the approaches being compared, making their conclusions baseless. Some of the approaches they compare can be quite expensive in their own right. The incompatible insect technique for instance requires sex sorting of mosquitoes which is very labour intensive if done mechanically or very expensive if using automated systems, and mosquitoes need to be reared at large scales to achieve suppression.

2.1.5. We acknowledge the reviewers concern regarding the cost of suppression, and our conclusions around cost savings previously recognise this depends on the cost of suppression, for example in the results: 

”All hybrid programmes had the potential to offer cost savings over replacement alone, depending on the cost of suppression… This, however, does not take into account the costs of suppression. A different interpretation of these results would be: if suppression can be achieved for less than these costs, then a hybrid programme will cost less than replacement alone.” 

And in the discussion: 

“We predict that insect-based suppression methods (1gSLT, SIT, Male Wolbachia release) will be more effective than conventional suppression tools, but also outline a limited cost window which may be challenging for insect-based suppression methods to achieve. Investment in new infrastructure to conduct insect-based suppression may not be justified for a one-off suppression, but between overlapping resource requirements for suppression and replacement, ongoing use post-replacement (e.g., outbreak control or to achieve dengue elimination) and a continued drive to lower costs of mosquito suppression (52), this investment cost may be justified.” 

However, we understand that expressing the hybrid approach as cost saving has the potential to be misinterpreted and therefore, as mentioned in point 2.1.2, have further edited the results to better clarify our conclusions.

2.1.6. The authors then use models to test the effect of different suppression interventions on mosquito population sizes, and later, the effect on the release ratio of Wolbachia-infected mosquitoes required for successful Wolbachia establishment. These approaches have different effects (e.g. adulticides kill off adults quickly but the population bounces back quickly, while incompatible/sterile insect approaches have slower but longer term population suppression), which is a reasonable point to make. But the authors then compare these approaches directly using literature estimates of efficacy. Parameters were taken from a single study for each approach and these studies are in no way comparable to each other. They were in different environments with different populations of mosquitoes and were done at different geographic scales, time periods and intensities. There are also issues with using a single study to represent the expected efficacy for a typical suppression programme. I just don't see how it's reasonable to compare these approaches directly and conclude that  one is more effective when the approaches have been applied at different intensities and durations.

2.1.6. We appreciate the reviewer’s comments regarding the difficulties encountered when comparing different suppression interventions. A WHO Evidence Review Group is currently conducting a rigorous systematic review of Aedes mosquito control effectiveness that was unfortunately not available at the time of analysis and remains in development. In the absence of this we aimed to derive evidence-based scenarios for suppression effectiveness which allowed approximation of the differences between methods of suppression at 5- and 10-week periods. We acknowledge that building the evidence base for these scenarios is not always straightforward to extract from the literature due to different study designs and contexts. To mitigate this complexity in our models we also include a range of hypothetical fixed effectiveness values for direct comparison between different suppression methods, this has been clarified in the discussion of limitations: 

“Additionally, suppression efficacy is highly context-dependent; to mitigate this we modelled an hypothetical 20%, 50%, and 80% suppression efficacy to allow clearer comparison between suppression techniques.” 

Our main findings (ie that insect release strategies offer advantages over conventional suppression strategies) are the same whether using the literature-derived efficacy estimates or the fixed efficacies. Another advantage of using a range of fixed efficacies is that it allows readers or implementers to interpret these results in different contexts where higher or lower suppression effectiveness may be feasible.

2.1.7. In summary, while the paper addresses an important issue, in my view the entomological models are problematic and/or uninformative, and the paper is not suitable for indexing without substantial fundamental changes.

2.1.7. We are glad the reviewer agrees that this modelling work informs decisions in an important area of research. We hope that the above general responses and below specific responses clarify the value of the entomological models and their importance for this specific application of informing the Wolbachia replacement TPP and hybrid suppress then replace PPC.

2.2. Specific comments:

2.2.1. Abstract - results – These quantitative thresholds are adjusted to data from the wMel releases in Wolbachia and they should not be extrapolated to Wolbachia releases in other locations, given that Wolbachia releases can have vastly different outcomes depending on the environment

2.2.1. The statement: 

“We estimate that for Wolbachia replacement to be deployable in enough areas to make major contributions to reducing global dengue burden by 25% (in line with 2030 WHO targets), it must have the potential for cost be reduced to between $7.63 and $0.24 (USD) per person protected or less”

Does not depend on any data from wMel releases. These figures come directly from table 2 with a legend:

“The predicted target cost per person for Wolbachia replacement based on different assumptions about desired global impact (rows) and averted medical and outbreak control costs (assumed proxy of willingness to pay, columns).”

These figures are based on the cost of illness and conventional control alone, while effectiveness was set by the TAG for the TPP at an assumed 70%, neither of which depend on data from the wMel releases.

2.2.2. Abstract – results - Suppression interventions will reduce the number of mosquitoes required for replacement, but this statement ignores the very high numbers of mosquitoes that are needed for the suppression itself (if using incompatible males, which the authors conclude is the most effective approach).

2.2.2. This statement refers to how suppression reduces the number of mosquitoes required during Wolbachiareplacement. This was identified as a key parameter by the TAG because it requires release of biting, potentially dengue-transmitting, mosquitoes and therefore has a different set of community concerns, engagement strategies, costs and regulatory implications to the release of male mosquitoes. We also clarify in various sections of the manuscript that the suppression in hybrid suppress-then-replace strategies is recommended to complement existing vector control strategies, therefore, releasing mosquitoes for insect-release suppression technologies (i.e., Male Wolrelease, SIT, and 1gSLT) would already be in effect, meaning that there would be a net reduction in released mosquitoes (both male and female) in such a hybrid strategy.

2.2.3. Intro paragraph 2 – Not all mosquitoes infected with Wolbachia show reduced virus dissemination- be more specific about the species and Wolbachia strain.

2.2.3. We agree it was not clear from the text that not all Wolbachia strains reduce virus dissemination, therefore this has been clarified in text: 

“Ae. aegypti mosquitoes infected with certain strains of the bacterium Wolbachia, such as wMel, wMelPop, and wAlbB (7) show reduced rates of virus dissemination, making them less capable of transmitting arboviruses (8).”

2.2.4. Intro paragraph 3  – The paragraph is a bit of an oversimplification- Wolbachia doesn’t always reach fixation even if very high frequencies are reached, for instance, due to maternal transmission failure. There is also now evidence from field trials showing that Wolbachia frequencies can fluctuate seasonally or even decline to zero even after reaching near-fixation depending on the environment.

2.2.4. We agree that Wolbachia is not always successful reaching fixation and has been known to fluctuate or decline. That environmental variation affects Wolbachia replacement dynamics has been clarified in the text: 

“Modelling has shown that once a critical proportion of mosquitoes in the population have Wolbachia, coverage should continue to increase to fixation without further releases, but below this threshold Wolbachia coverage may decline (possibly to zero) once releases stop due to fitness costs associated with released mosquito strains 8 … It should be noted that in practice, Wolbachia frequencies may fluctuate seasonally and still decline to zero after reaching fixation depending on environmental variables such as temperature, rainfall, and physical barriers (12, 13).”

2.2.5. Intro paragraph 4 – “widespread, long-term effectiveness” is true in some locations but there are also cases where Wolbachia releases have failed (see above), meaning that there are likely to be environmental constraints on where Wolbachia infections can successfully establish

2.2.5. We further acknowledge the need to emphasise there is heterogeneity in the success of Wolbachiareplacement, and have edited the text to underscore these considerations: 

“A growing range of entomological, epidemiological and modelling evidence supports the widespread, long-term effectiveness of Wolbachia replacement (14-16), and research continues to identify environmental conditions associated with spatially and temporally heterogeneous Wolbachia establishment (13).”

2.2.6. Methods – entomological model paragraph 3 – this seems to only cover a scenario where Wolbachia-infected mosquitoes are released as eggs. Please provide some context in the introduction as egg releases are not mentioned before here. The models are built on the assumption of egg releases, and the fact that Wolbachia-infected larvae are initially separate from wild larvae is an important component, but there is no justification for this or acknowledgement of other types of release. The authors discuss population suppression through male releases, but this will require adult releases.

2.2.6. We appreciate the reviewers feedback that the text was unclear about the different methods of mosquito release during Wolbachia replacement. Our work modelled egg releases in release cups only; text has been added to the entomological model section of the methods for additional context: 

“Since this model was calibrated using data of Wolbachia fixation dynamics from the Yogyakarta RCT (16), any features of this RCT influenced the methodology, such as delivery mode (egg releases in release cups) and the number of releases (ranging from nine to 14), therefore, other delivery modes such as releasing Wolbachia-infected adult Ae. aegypti were not simulated.” 

Each of the insect-release suppression technologies (Male Wol release, SIT, and 1gSLT) were simulated with adult releases.

2.2.7. Methods- suppression – “The efficacy of each method was based on evidence sourced from the literature” - this section is quite subjective and there is no information about how the authors searched for studies or selected them aside from having a preference for large randomised controlled studies.

2.2.7. We acknowledge the literature-derived efficacies have limitations, such as estimates of effectiveness being context specific, variations in implementation and ambiguous reporting of effectiveness values. It is precisely because of this that we did not pursue a systematic search for studies with specific inclusion or exclusion criteria. Instead, we chose studies that we judged best evaluated the effectiveness of short-term pulses of the intervention. We clearly cite the sources of these assumed literature-derived effectiveness values and acknowledge that readers may expect alternative values in different settings. To address this, we chose to also include a range of hypothetical values (20%, 50%, and 80%) for suppression effectiveness which allow a more standardised comparison of the dynamics of different methods of suppression and how such dynamics affect requirements for successful replacement programmes. As mentioned earlier, this has been further emphasised in the discussion of limitations.

“Additionally, suppression efficacy is highly context-dependent; to mitigate this we modelled an hypothetical 20%, 50%, and 80% suppression efficacy to allow clearer comparison between suppression techniques.” 

2.2.8. Methods – suppression – I would prefer if the authors didn’t use “larvicides” here- guppies are not larvicides – they are predators. “Larvicides” typically refers to chemical insecticides or bacterial pathogens like Bt.

2.2.8. Since both chemical larvicides and guppies primary mode of action is to kill eggs, larvae and pupae, their effects on the model are identical, therefore we believe the hypothetical range of values (20%, 50%, and 80%) can refer to either larvicides, which are more widely used, or guppies. However, in response to this comment the text has been edited to clarify this labelling specifically: 

“Larvicides were simulated by equally reducing the number of eggs, larvae, and pupae. Because the effect is similar and because the best measurement of effectiveness of an intervention that targets the aquatic stages of the mosquito came from a trial of predatory guppies, this effectiveness as measured by Hustedt et al (37) was chosen to represent the effectiveness of larvicides.”

2.2.9. Figure 1 – The model shows that the mosquito population size permanently decreases after Wolbachia establishment, but I’m not aware of any evidence for this from field release data. For instance, data from the releases in Yogyakarta shows that Wolbachia releases had minimal impact on the population size: https://journals.plos.org/plosntds/article?id=10.1371/journal.pntd.0010284

2.2.9. We assumed that Wolbachia exerts a fitness cost, following the reported evidence (e.g., Joubert et al, 2016; Ross & Hoffmann, 2022), and as a consequence of this, the total population size was reduced following replacement with Wolbachia-infected mosquitoes. The fitness cost we selected was intended as a conservative estimate to reduce risk of under-estimation in field deployments, so the projected reduction in total mosquito numbers is thus likely a concomitant over-estimation.

2.2.10. Supp figure 1 – I’m not sure why the authors only considered costs up to 20% - data from release programs have estimated costs of around 30% for wMel (Hoffmann et al. 2011) and this is likely to vary substantially depending on the environment.

2.2.10. We agree that the fitness cost of Wolbachia infection is very dependent on the environment. We edited the manuscript to emphasise this:

“Due to the fitness cost of Wolbachia (conservatively modelled to be 20% (39), but highly variable depending on environment (41))...”

Furthermore, we have extended the range of values included in the sensitivity analysis in Supplementary Figure 1 to encompass the 30% value highlighted by the reviewer.

2.2.11. Table S1 – Why is SIT included here if the authors found no studies to base its efficacy on? It is true that few SIT release programmes in Ae. aegypti have been published, but there is at least one (e.g. de Castro Poncio et al. 2021)

2.2.11. The lack of a literature-derived efficacy value for SIT suppression technique was a genuine oversight and we’re grateful for the reviewer’s feedback; the manuscript has been amended to include a literature-derived efficacy for SIT. A literature-derived efficacy of approximately 49% at 5-weeks and 77% at 10-weeks has been extracted from the field trial by De Castro Poncio et al (2021) Figure 2b. The analyses which previously used an hypothetical 50% efficacy for this suppression technique have been updated to use the literature-derived 49% efficacy, correspondingly there has been only marginal changes in resulting population dynamics and no change in the overall trends or findings discussed. The methods section for SIT suppression have been updated to include this development:

“Parameters for the hypothetical fixed rate efficacy of 20%, 50%, and 80% were calculated by comparing the total adult population at model equilibrium with the minimum adult population reached after five weeks of application. The literature-derived efficacy values were 49% five weeks after the last suppression period and 77% ten weeks after the last suppression period (33), calculated by comparing the total adult population at model equilibrium to the total adult population after five- or ten-weeks of suppression which achieved the desired efficacy (summarised in Supplementary Table 1). A caveat of this approach is that the resulting minimum adult population is reached later than five- or ten-weeks, therefore, the maximum efficacy calculated in these scenarios is marginally greater than the literature value stated (shown in Supplementary Figure 2).”

In addition, SIT values have been updated in Figures 4, 5, and 6, added to Supplementary Figure 2, and the SIT literature source has been added to Supplementary Table 1.

References

De Castro Poncio, L., Dos Anjos, F. A., De Oliveira, D. A., Rebechi, D., De Oliveira, R. N., Chitolina, R. F., Fermino, M. L., Bernardes, L. G., Guimarães, D., Lemos, P. A., Silva, M. N. E., Silvestre, R. G. M., Bernardes, E. S., & Paldi, N. (2021). Novel Sterile Insect Technology Program Results in Suppression of a Field Mosquito Population and Subsequently to Reduced Incidence of Dengue. Journal of Infectious Diseases, 224(6), 1005–1014. https://doi.org/10.1093/infdis/jiab049

Hancock, P. A., Sinkins, S. P., & Godfray, H. C. J. (2011a). Population dynamic models of the spread of Wolbachia. American Naturalist, 177(3), 323–333. https://doi.org/10.1086/658121

Hancock, P. A., Sinkins, S. P., & Godfray, H. C. J. (2011b). Strategies for introducing Wolbachia to reduce transmission of mosquito-borne diseases. PLoS Neglected Tropical Diseases, 5(4). https://doi.org/10.1371/journal.pntd.0001024

Joubert, D. A., Walker, T., Carrington, L. B., De Bruyne, J. T., Kien, D. H. T., Hoang, N. L. T., Chau, N. V. V., Iturbe-Ormaetxe, I., Simmons, C. P., & O’Neill, S. L. (2016). Establishment of a Wolbachia Superinfection in Aedes aegypti Mosquitoes as a Potential Approach for Future Resistance Management. PLoS Pathogens, 12(2). https://doi.org/10.1371/journal.ppat.1005434

Ross, P. A., & Hoffmann, A. A. (2022). Fitness costs of Wolbachia shift in locally-adapted Aedes aegypti mosquitoes. Environmental Microbiology, 24(12), 5749–5759. https://doi.org/10.1111/1462-2920.16235

Yakob, L., Funk, S., Camacho, A., Brady, O., & Edmunds, W. J. (2017). Aedes aegypti Control Through Modernized, Integrated Vector Management. PLoS Currents, 1–34. https://doi.org/10.1371/currents.outbreaks.45deb8e03a438c4d088afb4fafae8747

 10.21956/gatesopenres.15611.r34088 Reviewer response for version 1 O'neill Scott 1 Referee https://orcid.org/0000-0002-4131-3615 World Mosquito Program, Monash University, Clayton, Victoria, Australia

Competing interests: I am the CEO of the World Mosquito Program, a not-for-profit that has developed and implements Wolbachia replacement technologies. My conflict is non-financial but relevant given that the paper describes the use of our technology

 3 8 2023 Copyright: © 2023 O'neill S 2023 This is an open access peer review report distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. recommendation approve-with-reservations

This paper uses entomological and economic modeling approaches to derive programmatic cost targets to support a WHO defined TPP for Wolbachia replacement. The starting premise for the modeling approach is that Wolbachia replacement needs to be deployable on a global scale at a sufficiently low cost to achieve the WHO’s goal of a 25% reduction in global dengue incidence by 25% (relative to a 2010-2020 baseline). This premise, and the consequent epidemiologic and economic assumptions for how Wolbachia could be most cost-effectively deployed at a global scale, lead to overly simplistic conclusions around a target global cost of deployment of Wolbachia replacement. Although the authors have qualified this by stating that “TPP targets represent the lowest averted medical and outbreak control costs per person among all areas where releases are required, there are many areas that could support higher programme costs”, the unrealistic target of reducing 25% of national dengue burden and averting these costs against the lowest medical and outbreak control costs over three years provides an overly optimistic global PCC of $0.24 per person. The linkage in this instance has a substantial impact on the total cost of coverage. The current modeling approach does not adequately address this issue. The entomological modelling of the hybrid “suppress then replace” is based largely on theoretical estimates of suppression efficacy (Figure 4) or extrapolation of data from small scale pilot studies. The assumptions behind some of these approaches are flawed, and present an unrealistic scenario around the feasibility of undertaking hybrid suppress and then replace methods. This section is highly speculative, adds little to the paper and should be removed.

Specific comments:

1. In order to frame the impact targets in the TPP in terms of the WHO NTDs roadmap target of a 25% reduction in global dengue burden by 2030, the authors identify the priority settings for achievement of that target by ranking 5x5km pixels globally (Page 4). There are several issues with this approach to ranking target geographies:

1a. The authors state on page 4 that they “identified which areas (5km x 5km pixels) would need to be targeted to reach the WHO goal of reducing global dengue burden by 25% in the most net cost-efficient manner. To do this, each 5km x 5km pixel was ranked from highest to lowest based on a benefit (averted medical and outbreak costs) to cost (approximate Wolbachia program cost estimate based on population density from Brady et al 2020) ratio.” A critical input into this calculation of the pixel-level benefit-to-cost ratio, which forms the basis for identifying the target geographies considered in the rest of the modeling, is the assumed pixel-level cost of Wolbachia deployment, yet insufficient detail is provided about that assumed intervention cost. The paper cited reports an economic analysis of Wolbachia deployment in Indonesia, in which the predicted Wolbachia program cost was modeled as a cost per km2, as a function of human population density. Presumably the current paper also used cost per km2 as the input for Wolbachia program cost in calculating pixel-level benefit-cost ratios, and this should be described in more detail. More fundamentally, it seems circular to identify target geographies based on a benefit-to-cost ratio that requires an assumption of programmatic cost per km2 / per person as an input, and then use the predicted avertable cost of dengue in those target geographies to define a target programmatic cost of Wolbachia per person. Please clarify the logic of this approach, and why the inclusion of a Wolbachia cost assumption in the initial step of ranking the target geographies doesn’t compromise the approach.

1b. The ranking of 5x5 km pixels to identify target geographies for Wolbachia deployment  produces unrealistic scenarios of fragmented small release areas (totalling 924,557 km2 in size) that are distributed across 73 different countries. This undermines the cost-per-person assumptions, as it would clearly be more costly to deploy across these fragmented areas than to cover large contiguous urban populations.To achieve the most net cost-efficient implementation of Wolbachia at a global scale, ranking would be better done at e.g. administrative 2 level to identify the contiguous highest-burden areas in which the global reduction targets could be achieved or, if the pixel-based approach is to be used, at least apply an additional criterion that filters out identified target release areas below a minimum km2 (e.g. 10 km2) or population size (e.g. 50,000 population). It is evident from the datasets of target release area km2 and population at 2nd level administrative units, provided in the supplementary materials, that many of the target geographies are far too small to support Wolbachia replacement deployments in reality.

Despite this being one of the major limitations of the modeling approach, the authors give no consideration at all in the discussion to the fact that in real-world programmatic implementation of Wolbachia replacement, it will always be most efficient from operational, entomological, and cost-efficiency perspectives to deploy across a large contiguous urban area rather than small fragmented release areas. This needs to be addressed in the discussion.

2. The 'Required cost per person covered' thresholds in Table 2 are presented at a global level, based on the projected costs averted in the least cost effective setting in any given scenario. Although regional and national cost targets are included in the supplementary materials, and the authors acknowledge in the discussion that many countries could support higher costs and ‘that a Wolbachia replacement product would still meet the TPP targets if initial program costs were higher and if the product has the potential to reduce costs down to the $2.33 target, this should be presented more explicitly within the primary results. It is not meaningful to report the primary cost targets at a global level without including also the range of national- and regional-level target costs, as that's not the spatial scale at which economies of scale in production and implementation costs are achieved, or at which customers' willingness-to-pay or funding decisions operate.

3. The cost targets are very conservative in requiring cost-neutrality from a health system perspective (i.e. implementation cost ≤ direct medical costs + outbreak vector control costs averted) and assuming only 3-5 years of benefits for the preferred TPP. Although a scenario of 10 years of benefits is also included in Table 2, the authors present the cost target of $2.33 as the minimum TPP, based on a scenario of achieving a 25% global burden reduction and only 5 years of benefit, with a PPC cost target of $0.24 (25% national burden reduction in all countries and only 3 years of benefit). These durations of benefit are too pessimistic given the evidence from field studies of Wolbachia durability in Ae. aegypti populations for at least 10 years.     

4. The entomological modelling of the hybrid “suppress then replace” is based largely on theoretical estimates of suppression efficacy (Figure 4), or extrapolation of population suppression results from small scale pilot interventions. This reviewer questions whether the efficacy of these suppression methods can ever be achieved at the scale required to avert 25% of the global dengue burden. The assumptions around the suppressive effects of a single round of intervention seem overly simplistic, and are generally not consistent with field implementations that take many rounds of intervention to produce consistent and reproducible suppression (see Supplementary Table 1 references below). “With reductions in adult mosquito population size in the range of 8.18 – 43.51%, from a single application all methods of suppression were predicted to remove the ~1–10% mosquito population exacerbation seen in replacement only programmes.” (Page 12) Further, “Our model predicts that a prior suppression programme of five weekly rounds could reduce the number of Wolbachia mosquitoes required to reach target coverage within 6 months by 16-81% depending on suppression method used.” (Pages 12-13)

Based on empirical data from a Male Wolbachia release program that targeted Ae. aegypti, there is questionable evidence that a 5-week suppression program would substantially impact mosquito population abundance. In the Supplementary Table 1, it indicates a 5-week suppression efficacy of 60%, and 10-week suppression efficacy of 90% (sourced from Ching et al. 2021). This is incorrect. Figure 2, Panel G indicates <10% suppression out to week 7 in the Yishun Core Area where releases were concentrated. In the Tampines Core area, suppression efficacy varied significantly over the first 12-13 weeks, ranging from 0-40%. Ching et al. (2021) presented a statistical analysis of the suppression effect (Figure 3, Panels A and C) and found no significant suppression in Yishun until the fourth month of releases, and variable results in Tampines (statistically significant suppression in first month, but not significant in the second month of releases).

Similarly for the “1gSLT” suppression rates of 45% and 75% at weeks 5 and 10, respectively (Supplementary Table 1. Suppression Efficacy Carvalho et al. 2015). These data were extracted from Figure 2, panel D, from the 5- and 10-week timepoints in release area A, the authors have failed to state that there were 7-months of releases undertaken across these combined Areas A and B prior to the 5- and 10-week suppression period over which the above rates were calculated (for Area A). Also of note was the very small size of the suppression area in this trial (Area A = 0.055km2). Again, the assumptions drawn from these studies are inaccurate and should be withdrawn.

How have the SIT suppression efficacy estimates been validated? In Supplementary Table 1, it states there is no currently published trials for suppression using SIT, and also in the Methods it states “There are currently no published randomised controlled trials for suppression of Aedes mosquitoes using SIT, therefore 50% efficacy is shown for SIT in analyses where only literature-derived efficacy is used. What is the “literature derived efficacy”?

The assumptions underpinning the modelling of the suppress then replace section of the paper appear to be erroneous, or at least very optimistic of their effectiveness in terms of suppression of the mosquito populations. This section is highly speculative, adds little to the paper and should be removed.

5. “All hybrid programmes had the potential to offer cost savings over replacement alone, depending on the cost of suppression” (Page 14). This statement is not supported by presentation of any costing data for suppression. From an economic perspective, there is no evidence that a hybrid approach of suppression followed by replacement would be more cost effective than Wolbachia replacement alone, and a recent study based on data from Singapore’s Wolbachia suppression program suggests that a hybrid suppression-followed-by-replacement program could be significantly more expensive (Soh et al. 2021 1 ). The current annual per capita cost of suppression in Singapore is approximately USD $5 per person reached. This cost is roughly equivalent to the one-time, all-inclusive cost required to achieve Wolbachia replacement across some target settings (Brady et al BMC Med 2020; Turner et al PLoS Negl Trop Dis 2023). The ‘potential cost savings enabled by a hybrid approach’ presented in Table 3 “assumes suppression has no cost or is an in-kind contribution”, which is entirely implausible in any setting, let alone in a scenario of large-scale global implementation across the 73 countries (>900,000 km2), which would be required to achieve the TPP minimum targets. Without a more extensive analysis of the efficacy of suppression programs and their cost, the current analyses are highly speculative and add little to the paper. This reviewer feels the whole section on hybrid “suppress then replace” should be removed.

Minor comments:

 

Page 3, Introduction, second paragraph: The virus-blocking phenotype is specific to the species of Wolbachia, mosquito, and virus - the authors should use more specific language here.

Page 3, Introduction, second paragraph: There is no empirical evidence that temporary increases in the Ae. aegypti mosquito population have been a key barrier to community acceptability of Wolbachia releases; this should be expressed as theoretical consideration.

Page 6, Results, paragraph 1 (and throughout): the terminology ‘coverage’ is misleading as it implies the % coverage of a geographic area, whereas the authors are referring to % Wolbachia prevalence in the local Ae. aegypti population. The proportion of Wolbachia-infected mosquitoes in the population should be considered the “percentage prevalence,” not the target coverage.

Page 11, Figure 3: The axes titles and figure legend needs to make explicit what time period this data relates to, in terms of the cumulative cases and costs averted . Presumably 5 years of benefit, since the $2.33 cost threshold is shown, however this is inconsistent with the $0.24 preferred criteria referred to in the figure legend, which relates to 3 years of benefit. And the lower horizontal dotted line appears to be at $0.39, not $0.24.

Page 16, Discussion: “Cost continues to be a barrier to wider adoption of Wolbachia replacement when its high costs but long-term benefits are compared to lower cost but short acting suppression methods”. This statement needs to reflect the fact that Wolbachia replacement has an evidence base for epidemiological impact while the short acting suppression methods do not.

Is the work clearly and accurately presented and does it cite the current literature?

Partly

If applicable, is the statistical analysis and its interpretation appropriate?

I cannot comment. A qualified statistician is required.

Are all the source data underlying the results available to ensure full reproducibility?

Yes

Is the study design appropriate and is the work technically sound?

Partly

Are the conclusions drawn adequately supported by the results?

Partly

Are sufficient details of methods and analysis provided to allow replication by others?

No

Reviewer Expertise:

Wolbachia biology and its use for disease control

I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard, however I have significant reservations, as outlined above.

References : Economic impact of dengue in Singapore from 2010 to 2020 and the cost-effectiveness of Wolbachia interventions. PLOS Glob Public Health .2021;1(10) : 10.1371/journal.pgph.0000024 e0000024 36962069 10.1371/journal.pgph.0000024 Tiley Katie

Competing interests: No competing interests were disclosed.

 18 4 2024

1.1 Summary comments:

1.1.1. This paper uses entomological and economic modeling approaches to derive programmatic cost targets to support a WHO defined TPP for Wolbachia replacement. The starting premise for the modeling approach is that Wolbachia replacement needs to be deployable on a global scale at a sufficiently low cost to achieve the WHO’s goal of a 25% reduction in global dengue incidence by 25% (relative to a 2010-2020 baseline). This premise, and the consequent epidemiologic and economic assumptions for how Wolbachia could be most cost-effectively deployed at a global scale, lead to overly simplistic conclusions around a target global cost of deployment of Wolbachia replacement. Although the authors have qualified this by stating that “TPP targets represent the lowest averted medical and outbreak control costs per person among all areas where releases are required, there are many areas that could support higher programme costs”, the unrealistic target of reducing 25% of national dengue burden and averting these costs against the lowest medical and outbreak control costs over three years provides an overly optimistic global PCC of $0.24 per person. The linkage in this instance has a substantial impact on the total cost of coverage. The current modeling approach does not adequately address this issue. 

1.1.1. We are grateful that the reviewer read and thoroughly responded to this work which informed the World Health Organisation (WHO) Target Product Profile (TPP) and Preferred Product Characteristics (PPC) documents for Wolbachia replacement technology contributing to the goal of reaching the WHO targets of reducing global burden of dengue by 25%. We have carefully considered the reviewers feedback and responded point-by-point in detail below.

1.1.2. The entomological modelling of the hybrid “suppress then replace” is based largely on theoretical estimates of suppression efficacy (Figure 4) or extrapolation of data from small scale pilot studies. The assumptions behind some of these approaches are flawed, and present an unrealistic scenario around the feasibility of undertaking hybrid suppress and then replace methods. This section is highly speculative, adds little to the paper and should be removed.

1.1.2. The criticism of the hybrid suppress-then-replace modelling justifiably covers some specific areas where the evidence supporting the effectiveness of several suppression tools could have been improved (which has now been implemented, see points 1.2.4a – e), but also covers wider conceptual issues of the value of the hybrid approach. One key aspect here is to clarify that modelling the hybrid approach was for the PPC process and not the TPP. This PPC process is designed to guide the development of novel products or strategies and are both more forward looking and less prescriptive than the TPP. The results from the modelling for these two sections should, therefore, also be interpreted within these different use contexts. We agree that the current evidence for hybrid suppress then replace strategies is limited, the modelling is intended to guide the (limited) use cases in which future improvements to suppression interventions may enable hybrid strategies.

1.2 Specific comments:

1.2.1. In order to frame the impact targets in the TPP in terms of the WHO NTDs roadmap target of a 25% reduction in global dengue burden by 2030, the authors identify the priority settings for achievement of that target by ranking 5x5km pixels globally (Page 4). There are several issues with this approach to ranking target geographies:

1.2.1a. The authors state on page 4 that they “identified which areas (5km x 5km pixels) would need to be targeted to reach the WHO goal of reducing global dengue burden by 25% in the most net cost-efficient manner. To do this, each 5km x 5km pixel was ranked from highest to lowest based on a benefit (averted medical and outbreak costs) to cost (approximate Wolbachia program cost estimate based on population density from Brady et al 2020) ratio.” A critical input into this calculation of the pixel-level benefit-to-cost ratio, which forms the basis for identifying the target geographies considered in the rest of the modeling, is the assumed pixel-level cost of Wolbachia deployment, yet insufficient detail is provided about that assumed intervention cost. The paper cited reports an economic analysis of Wolbachia deployment in Indonesia, in which the predicted Wolbachia program cost was modeled as a cost per km2, as a function of human population density. Presumably the current paper also used cost per km2 as the input for Wolbachia program cost in calculating pixel-level benefit-cost ratios, and this should be described in more detail. More fundamentally, it seems circular to identify target geographies based on a benefit-to-cost ratio that requires an assumption of programmatic cost per km2 / per person as an input, and then use the predicted avertable cost of dengue in those target geographies to define a target programmatic cost of Wolbachia per person. Please clarify the logic of this approach, and why the inclusion of a Wolbachia cost assumption in the initial step of ranking the target geographies doesn’t compromise the approach.

1.2.1a. We thank the reviewer for raising this lack of clarity and concur the current description may be misinterpreted to be circular. This section has been edited to address this issue:

“Next we identified which areas (5km x 5km pixels) would need to be targeted to reach the WHO goal of reducing global dengue burden by 25% in the most net cost efficient manner. For a generic environmental intervention, where cost of the intervention only depends on area covered, this would involve targeting areas with the highest density of dengue costs. However, because the cost of Wolbachia programmes have been shown to depend on the human population density and per capita GDP in the release area (27), this can change which areas are most important to prioritise from an optimal net cost perspective. To account for these variable implementation cost factors Wolbachia replacement is most cost effective if targeted to higher density, high dengue burden areas (27). We therefore identified which areas (5km x 5km pixels) would need to be targeted to reach the WHO goal of reducing global dengue burden by 25% in the most net cost-efficient manner. To do this, each 5km x 5km pixel was ranked from highest to lowest based on a benefit (averted medical and outbreak costs) to cost (approximate Wolbachia programme cost estimate based on population density and per capita GDP from Brady et al. (27)) ratio. For clarity, the approximate Wolbachia programme cost from Brady et al. (27) only affects the ranking of pixels (i.e. targeting), not the TPP target cost estimates. Cumulative averted cases were then calculated and pixel selection ended when averted cases first exceeded 25% of the global total….”

1.2.1b. The ranking of 5x5 km pixels to identify target geographies for Wolbachia deployment  produces unrealistic scenarios of fragmented small release areas (totalling 924,557 km2 in size) that are distributed across 73 different countries. This undermines the cost-per-person assumptions, as it would clearly be more costly to deploy across these fragmented areas than to cover large contiguous urban populations. To achieve the most net cost-efficient implementation of Wolbachia at a global scale, ranking would be better done at e.g. administrative 2 level to identify the contiguous highest-burden areas in which the global reduction targets could be achieved or, if the pixel-based approach is to be used, at least apply an additional criterion that filters out identified target release areas below a minimum km2 (e.g. 10 km2) or population size (e.g. 50,000 population). It is evident from the datasets of target release area km2 and population at 2nd level administrative units, provided in the supplementary materials, that many of the target geographies are far too small to support Wolbachia replacement deployments in reality. Despite this being one of the major limitations of the modeling approach, the authors give no consideration at all in the discussion to the fact that in real-world programmatic implementation of Wolbachia replacement, it will always be most efficient from operational, entomological, and cost-efficiency perspectives to deploy across a large contiguous urban area rather than small fragmented release areas. This needs to be addressed in the discussion.

1.2.1b. We recognise and agree that many additional considerations will go into which areas will ultimately be targeted for Wolbachia implementation and the resulting map is not intended to be prescriptive, but rather give an indication of the kinds of areas that would be most sensible to prioritise from a net cost efficiency perspective. Even if the decision to implement occurs at administrative unit-level, not everywhere within an administrative unit will see Wolbachia releases and the pixel-level maps can be useful for informing within-administrative unit targeting subject to a set of criteria that should be locally determined. It should also be emphasised that the TPP is, by definition, for a hypothetical Wolbachia replacement product, not the current most widely-used wMel programme. This means that such considerations may differ in their importance and, as such, are beyond the scope of the TPP and the accompanying modelling analysis. We have added the following sentence to third paragraph of the discussion to convey this:

“It is important to clarify that these maps should not be used prescriptively, but rather give an indication of the kinds of areas that are likely to be most cost efficient to target, with the final decision on which areas are targeted for release subject to additional operational, entomological, financial and political considerations.”

The administrative-unit summary provided in the supplement shows that 85.5% of 2nd administrative units (admin2s) have predicted release areas at or above 10km 2. Among the remaining admin2s with smaller release areas all of them are in countries that also contain admin2s with larger (>10km 2) release areas and 66% of smaller release admin2s occur in Vietnam, Colombia, Thailand, Brazil and Mexico which would all require large release areas across their respective countries. These smaller areas could thus be considered an extension of programmes in nearby areas. Introducing restrictions on contiguity or release area will, therefore, only affect a very small number of areas and have minimal impact on the overall findings.

1.2.2. The 'Required cost per person covered' thresholds in Table 2 are presented at a global level, based on the projected costs averted in the least cost effective setting in any given scenario. Although regional and national cost targets are included in the supplementary materials, and the authors acknowledge in the discussion that many countries could support higher costs and ‘that a Wolbachia replacement product would still meet the TPP targets if initial program costs were higher and if the product has the potential to reduce costs down to the $2.33 target, this should be presented more explicitly within the primary results. It is not meaningful to report the primary cost targets at a global level without including also the range of national- and regional-level target costs, as that's not the spatial scale at which economies of scale in production and implementation costs are achieved, or at which customers' willingness-to-pay or funding decisions operate.

1.2.2. The primary purpose of this work was to support the TPP. TPPs are used to decide if particular products should or should not be developed. For this TPP the Technical Advisory Group (TAG) decided that these criteria should be set to be globally relevant – i.e. the product should only be developed if it can be globally relevant – for equity reasons. The TPP uses the specific wording that a product “must have the potential to reach” these thresholds. We have now updated the abstract and the results sections to use this specific phrase for consistency. The pathway to achieving this with initially higher costs is mentioned repeatedly in the results and discussion sections with Figure 3 and the detailed municipality-level summaries in the supplement giving precise figures.

1.2.3. The cost targets are very conservative in requiring cost-neutrality from a health system perspective (i.e. implementation cost ≤ direct medical costs + outbreak vector control costs averted) and assuming only 3-5 years of benefits for the preferred TPP. Although a scenario of 10 years of benefits is also included in Table 2, the authors present the cost target of $2.33 as the minimum TPP, based on a scenario of achieving a 25% global burden reduction and only 5 years of benefit, with a PPC cost target of $0.24 (25% national burden reduction in all countries and only 3 years of benefit). These durations of benefit are too pessimistic given the evidence from field studies of Wolbachia durability in Ae. aegypti populations for at least 10 years.

1.2.3. It is worth reiterating that these estimates should not be interpreted as predictions of the likely costs and benefits of Wolbachia programmes, but as a proxy for willingness to pay for Wolbachia implementation. We agree that there is good evidence that Wolbachia persists for at least 10 years and specifically included a scenario with 10 years of benefits in our analysis because of this (Table 2). However, these longer duration benefits were judged by members of the TAG (which included programme managers) as being less relevant when estimating willingness to pay for a new programme and hence why the 5-year option was chosen for the minimum criteria.

1.2.4a. The entomological modelling of the hybrid “suppress then replace” is based largely on theoretical estimates of suppression efficacy (Figure 4), or extrapolation of population suppression results from small scale pilot interventions. This reviewer questions whether the efficacy of these suppression methods can ever be achieved at the scale required to avert 25% of the global dengue burden. The assumptions around the suppressive effects of a single round of intervention seem overly simplistic, and are generally not consistent with field implementations that take many rounds of intervention to produce consistent and reproducible suppression (see Supplementary Table 1 references below). “With reductions in adult mosquito population size in the range of 8.18 – 43.51%, from a single application all methods of suppression were predicted to remove the ~1–10% mosquito population exacerbation seen in replacement only programmes.” (Page 12) Further, “Our model predicts that a prior suppression programme of five weekly rounds could reduce the number of Wolbachia mosquitoes required to reach target coverage within 6 months by 16-81% depending on suppression method used.” (Pages 12-13)

1.2.4a. We agree that the evidence base for hybrid suppress and replace strategies is not as developed as that for Wolbachia replacement alone. It is this very fact that led WHO to develop a TPP for replacement but only a PPC for hybrid strategies. PPCs are not subject to the same specific criteria (like the 25% mentioned by the reviewer) as TPPs and come with clear guidelines that they should be tested using a range of methods and experimental designs before a TPP can be developed. We believe our modelling work is proportionate to the differing needs of these two distinct policy objectives.

We also agree that achieving the efficacy of the hybrid suppress-then-release scenarios at scale would be challenging and generally only an option that might be considered if cost, community acceptability or speed considerations could not be met with a replacement-only approach. We state this explicitly in the relevant section of the results:

“This may mean that rather than hybrid approaches superseding replacement-only approaches, their use may be restricted to areas where replacement-only cannot meet speed, acceptability, and cost goals.”

We also agree that it takes time and multiple rounds of application to consistently achieve suppression. Areas most likely to consider hybrid programmes will likely already have a large suppression component to their vector control programmes, so a hybrid programme could be considered more of an extension (e.g. continuing suppression for longer or at times of the year when suppression is limited) than a new application. Under such circumstances experienced personnel with a high degree of familiarity with the local vector ecology, supported by existing programme infrastructure would be more likely to achieve the levels of vector control suggested by these studies than if starting from scratch. Because we agree that it is difficult to make comparisons between different study designs for suppression effectiveness we also include fixed efficacies that readers can use instead that give a more standardised comparison between different suppression methods and can be used to explore lower effectiveness suppression options if they believe these are more realistic in their own contexts:

“Additionally, suppression efficacy is highly context-dependent; to mitigate this we modelled an hypothetical 20%, 50%, and 80% suppression efficacy to allow clearer comparison between suppression techniques.”

1.2.4b. Based on empirical data from a Male Wolbachia release program that targeted Ae. aegypti, there is questionable evidence that a 5-week suppression program would substantially impact mosquito population abundance. In the Supplementary Table 1, it indicates a 5-week suppression efficacy of 60%, and 10-week suppression efficacy of 90% (sourced from Ching et al. 2021). This is incorrect. Figure 2, Panel G indicates <10% suppression out to week 7 in the Yishun Core Area where releases were concentrated. In the Tampines Core area, suppression efficacy varied significantly over the first 12-13 weeks, ranging from 0-40%. Ching et al. (2021) presented a statistical analysis of the suppression effect (Figure 3, Panels A and C) and found no significant suppression in Yishun until the fourth month of releases, and variable results in Tampines (statistically significant suppression in first month, but not significant in the second month of releases).

1.2.4b. We acknowledge the reviewers concerns regarding the source for the literature-derived efficacy values for Male Wolbachia release, which we cite as Ching (2021). Having revisited the paper and our extraction process we agree that the time interval over which these levels of suppression can be achieved is overoptimistic in this example. We have since searched for alternative studies that give more appropriate estimates of short-term effectiveness. The analyses have been updated to include literature-derived efficacy values for Male Wolbachia release sourced from Crawford et al (2020), a field trial which reported suppression efficacy values reaching approximately 65% and 92% at 5- and 10-weeks, respectively; values extracted using WebPlotDigitizer (Rohatgi, 2021) from Figure 6d. The previous analyses used 60% at 5-weeks and 95% 10-weeks and this update has only caused small changes in the population dynamics and have not altered the overall trends or our findings. In addition, Male Wolbachia values have been updated in Figures 4, 5, and 6, added to Supplementary Figure 2, and the literature source for Male Wolbachiarelease has been updated in Supplementary Table 1.

1.2.4c. Similarly for the “1gSLT” suppression rates of 45% and 75% at weeks 5 and 10, respectively (Supplementary Table 1. Suppression Efficacy Carvalho et al. 2015). These data were extracted from Figure 2, panel D, from the 5- and 10-week timepoints in release area A, the authors have failed to state that there were 7-months of releases undertaken across these combined Areas A and B prior to the 5- and 10-week suppression period over which the above rates were calculated (for Area A). Also of note was the very small size of the suppression area in this trial (Area A = 0.055km2). Again, the assumptions drawn from these studies are inaccurate and should be withdrawn.

1.2.4c. We acknowledge the reviewers concerns that the literature-derived value extracted for 1gSLT was from the period of releases in Area A only (Carvalho et al, 2015), while previously there were seven months of releases across the combined Areas A and B. However, the prior suppression efforts concerned the refinement and scale up of technology and did not consistently impact vector populations until after the effort to area ratio was increased. This is confirmed within the paper: “Up to 11th February 2012 we released into areas A and B (Fig 1), comprising 11 ha in total. However, despite improvements in rearing over the period, in this highly infested area we were unable to produce enough OX513A males with the available resources to consistently maintain a mating fraction of 50%, as judged by the percentage of fluorescent larvae. We therefore reduced the release area to an area of 5.5 ha (Fig 1A). As expected, the fluorescence ratio increased correspondingly” and further supported by the observations from Figure 2, panel B, that the suppression effort was increased approximately 3-fold from levels prior to release in Area A only. Since these hybrid scenarios are most informative to circumstances where suppression technology is already in use as part of the vector control programme, it is appropriate for us to assume the region already has the necessary resources for suppression and therefore use values from more mature examples where initial training and testing phases are omitted. This makes our assumptions about using such methods for hybrid programmes more comparable with our assumptions about replacement, for which we also assume relatively rapid implementations can be achieved (< 6 months) when compared to the currently published literature based on the same argument of product maturity and familiarity. Furthermore, the extracted values for this technology used in the analysis are conservative in the short-term when compared with another field trial reported by Harris et al (2012), which shows an estimated 82% and 68% efficacy after 5- and 10-weeks of intervention, respectively (extracted from Figure 2c). Finally, we agree that the size of the suppression area is not extensive, and this detail has been added to the Supplementary Table 1:

“Extracted from Figure 2, panel D, from the 5- and 10-week timepoints of release in area A only, a concentrated area of 0.055km 2.”

1.2.4d. How have the SIT suppression efficacy estimates been validated? In Supplementary Table 1, it states there is no currently published trials for suppression using SIT, and also in the Methods it states “There are currently no published randomised controlled trials for suppression of Aedes mosquitoes using SIT, therefore 50% efficacy is shown for SIT in analyses where only literature-derived efficacy is used. What is the “literature derived efficacy”?

1.2.4d. Previously, we had no literature reference for SIT and therefore this statement from the manuscript refers to the fact that an hypothetical 50% efficacy value was used instead of a literature-derived efficacy in figures which otherwise presented only literature-derived efficacy values; this was stated in the legend. However, the lack of a SIT literature source has now been highlighted as an oversight by the other reviewer and a literature-derived efficacy of approximately 49% at 5-weeks and 77% at 10-weeks has been extracted from the field trial by De Castro Poncio et al (2021), Figure 2b. All analyses which previously used an hypothetical 50% efficacy for this suppression technique have been updated to use the literature-derived 49% efficacy, correspondingly there has been only marginal changes in resulting population dynamics and no change in the overall trends or findings discussed. The methods section for SIT suppression have been updated to include this development:

“Parameters for the hypothetical fixed rate efficacy of 20%, 50%, and 80% were calculated by comparing the total adult population at model equilibrium with the minimum adult population reached after five weeks of application. The literature-derived efficacy values were 49% five weeks after the last suppression period and 77% ten weeks after the last suppression period (33), calculated by comparing the total adult population at model equilibrium to the total adult population after five- or ten-weeks of suppression which achieved the desired efficacy (summarised in Supplementary Table 1). A caveat of this approach is that the resulting minimum adult population is reached later than five- or ten-weeks, therefore, the maximum efficacy calculated in these scenarios is marginally greater than the literature value stated (shown in Supplementary Figure 2).”

In addition, SIT values have been updated in Figures 4, 5, and 6, added to Supplementary Figure 2, and the SIT literature source has been added to Supplementary Table 1.

1.2.4e. The assumptions underpinning the modelling of the suppress then replace section of the paper appear to be erroneous, or at least very optimistic of their effectiveness in terms of suppression of the mosquito populations. This section is highly speculative, adds little to the paper and should be removed.

1.2.4e. We acknowledge the reviewer’s concerns about the measurement and comparability of effectiveness of suppression-based vector control for Aedes mosquitoes and agree that the values we use from the cited studies are context and application specific (and state so in the discussion). WHO is currently conducting a Cochrane review of Aedes mosquito control effectiveness that originally planned to inform these parameters with a more rigorous evidence base, however delays meant that we instead chose values based on a single well-designed study. Acknowledging this approach may introduce bias, we also simulate a range of fixed values for suppression effectiveness (20%, 50%, and 80%) which allowed us to better compare different suppression methods for the purpose of hybrid programmes and set expectations for areas where suppression efficacy is likely to be lower. We hope that the above improvements to the evidence we cite and further explanation of the differing objective of the modelling for the TPP and PPC more clearly clarifies the value of this section of the work.

1.2.5. “All hybrid programmes had the potential to offer cost savings over replacement alone, depending on the cost of suppression” (Page 14). This statement is not supported by presentation of any costing data for suppression. From an economic perspective, there is no evidence that a hybrid approach of suppression followed by replacement would be more cost effective than Wolbachia replacement alone, and a recent study based on data from Singapore’s Wolbachia suppression program suggests that a hybrid suppression-followed-by-replacement program could be significantly more expensive (Soh et al. 20211). The current annual per capita cost of suppression in Singapore is approximately USD $5 per person reached. This cost is roughly equivalent to the one-time, all-inclusive cost required to achieve Wolbachia replacement across some target settings (Brady et al BMC Med 2020; Turner et al PLoS Negl Trop Dis 2023). The ‘potential cost savings enabled by a hybrid approach’ presented in Table 3 “assumes suppression has no cost or is an in-kind contribution”, which is entirely implausible in any setting, let alone in a scenario of large-scale global implementation across the 73 countries (>900,000 km2), which would be required to achieve the TPP minimum targets. Without a more extensive analysis of the efficacy of suppression programs and their cost, the current analyses are highly speculative and add little to the paper. This reviewer feels the whole section on hybrid “suppress then replace” should be removed.

1.2.5. As mentioned above, the hybrid suppress-then-replace scenarios are specifically for circumstances where suppression is already part of an existing vector control programme. Since areas which would benefit from Wolbachia replacement are likely to already engage in vector control we included this model to quantify the potential gains from utilising existing suppression resources when transitioning to novel Wolbachia technology. In these specific contexts, suppression resources are likely to already have been purchased therefore the cost of suppression is assumed to be zero, however if this were not the case then the additional cost savings from implementing a new suppression technique are likely to be marginal. This has been further emphasised in the manuscript:

“…then a hybrid programme will cost less than replacement alone. This might refer to specific circumstances when suppression might be achieved at negligible cost, for example, if it’s already part of a vector control programme. These are ideal situations where the resources for suppression are already in place, however, in most cases the potential savings achieved by a hybrid suppress then replace programme won’t justify the additional cost of implementing a novel suppression programme.”

We acknowledge the reviewers concern regarding the cost of suppression, and our conclusions around cost savings previously recognise this depends on the cost of suppression, for example in the results:

”All hybrid programmes had the potential to offer cost savings over replacement alone, depending on the cost of suppression… This, however, does not take into account the costs of suppression. A different interpretation of these results would be: if suppression can be achieved for less than these costs, then a hybrid programme will cost less than replacement alone.”

And in the discussion:

“We predict that insect-based suppression methods (1gSLT, SIT, Male Wolbachia release) will be more effective than conventional suppression tools, but also outline a limited cost window which may be challenging for insect-based suppression methods to achieve. Investment in new infrastructure to conduct insect-based suppression may not be justified for a one-off suppression, but between overlapping resource requirements for suppression and replacement, ongoing use post-replacement (e.g., outbreak control or to achieve dengue elimination) and a continued drive to lower costs of mosquito suppression (51), this investment cost may be justified.”

Finally, it is worth repeating that the aim of the PPC is to motivate the development of new products and strategies. It is precisely because one-off suppression campaigns can’t meet these cost objectives that a PPC is needed, and these results give approximate targets that novel products or strategies could aim for to be useful for a hybrid suppress and replace approach.

1.3. Minor comments:

1.3.1. Page 3, Introduction, second paragraph: The virus-blocking phenotype is specific to the species of Wolbachia, mosquito, and virus - the authors should use more specific language here.

1.3.1. We agree that more specific language should be used when referring to the virus-blocking potential of wMel, therefore this has been clarified in text:

“Ae. aegypti mosquitoes infected with certain strains of the bacterium Wolbachia, such as wMel, wMelPop, and wAlbB (7) show reduced rates of virus dissemination, making them less capable of transmitting arboviruses (8).”

1.3.2. Page 3, Introduction, second paragraph: There is no empirical evidence that temporary increases in the Ae. aegypti mosquito population have been a key barrier to community acceptability of Wolbachia releases; this should be expressed as theoretical consideration.

1.3.2. We agree that temporary increases in Ae. aegypti mosquito population have not been a barrier to community acceptability in the past, therefore we have edited the text to reflect this nuance:

“All of these options increase cost and can also lead to undesirable temporary increases in the Ae. aegypti mosquito population which should be addressed during community engagement to avoid it becoming a barrier to community acceptability (9, 10).”

1.3.3. Page 6, Results, paragraph 1 (and throughout): the terminology ‘coverage’ is misleading as it implies the % coverage of a geographic area, whereas the authors are referring to % Wolbachia prevalence in the local Ae. aegypti population. The proportion of Wolbachia-infected mosquitoes in the population should be considered the “percentage prevalence,” not the target coverage.

1.3.3. We agree the terminology “coverage” could be a potential point of confusion, therefore we have clarified the phrasing of coverage in the introduction so that the usage is clear throughout:

“…self-sustaining coverage, defined as the percentage of Ae. aegypti population infected with Wolbachia, can be achieved by: increasing the number of releases…”

Our definition of coverage must be consistent with that used by the TAG for the TPP and was hence a decision outside of our control.

1.3.4. Page 11, Figure 3: The axes titles and figure legend needs to make explicit what time period this data relates to, in terms of the cumulative cases and costs averted. Presumably 5 years of benefit, since the $2.33 cost threshold is shown, however this is inconsistent with the $0.24 preferred criteria referred to in the figure legend, which relates to 3 years of benefit. And the lower horizontal dotted line appears to be at $0.39, not $0.24.

1.3.4. We thank the reviewer for their feedback and have changed the axis labels and figure legend to make it clear what the temporal denominator is for each measure.

1.3.5. Page 16, Discussion: “Cost continues to be a barrier to wider adoption of Wolbachia replacement when its high costs but long-term benefits are compared to lower cost but short acting suppression methods”. This statement needs to reflect the fact that Wolbachia replacement has an evidence base for epidemiological impact while the short acting suppression methods do not.

1.3.5. Agreed, this sentence now reads:

“Cost continues to be a barrier to wider adoption of Wolbachia replacement when its high costs but long-term benefits are compared to lower cost but short acting suppression methods, despite differences in the evidence base underpinning these benefits”

References

Carvalho, D. O., McKemey, A. R., Garziera, L., Lacroix, R., Donnelly, C. A., Alphey, L., Malavasi, A., & Capurro, M. L. (2015). Suppression of a field population of Aedes aegypti in Brazil by sustained release of transgenic male mosquitoes. PLoS Neglected Tropical Diseases, 9(7), 1–15. https://doi.org/10.1371/journal.pntd.0003864

Ching, N. L. (2021). Wolbachia-mediated sterility suppresses Aedes aegypti populations in the urban tropics. https://doi.org/10.1101/2021.06.16.21257922

Crawford, J. E., Clarke, D. W., Criswell, V., Desnoyer, M., Cornel, D., Deegan, B., Gong, K., Hopkins, K. C., Howell, P., Hyde, J. S., Livni, J., Behling, C., Benza, R., Chen, W., Dobson, K. L., Eldershaw, C., Greeley, D., Han, Y., Hughes, B., … White, B. J. (2020). Efficient production of male Wolbachia-infected Aedes aegypti mosquitoes enables large-scale suppression of wild populations. Nature Biotechnology, 38(4), 482–492. https://doi.org/10.1038/s41587-020-0471-x

De Castro Poncio, L., Dos Anjos, F. A., De Oliveira, D. A., Rebechi, D., De Oliveira, R. N., Chitolina, R. F., Fermino, M. L., Bernardes, L. G., Guimarães, D., Lemos, P. A., Silva, M. N. E., Silvestre, R. G. M., Bernardes, E. S., & Paldi, N. (2021). Novel Sterile Insect Technology Program Results in Suppression of a Field Mosquito Population and Subsequently to Reduced Incidence of Dengue. Journal of Infectious Diseases, 224(6), 1005–1014. https://doi.org/10.1093/infdis/jiab049

Rohatgi, A. (2021). WebPlotDigitizer V4.5. https://automeris.io/WebPlotDigitizer