4

Response to Arboviral Threats

The workshop’s third session, moderated by Ann M. Powers, associate director for science at the Centers for Disease Control and Prevention, was devoted to the various ways that researchers, public health officials, and others have responded to arbovirus threats. The types of responses in the session included disease and vector surveillance, vector control, vaccines, and modeling. Gabriela Paz-Bailey, chief of the Dengue Branch, Division of Vector-Borne Diseases at the Centers for Disease Control and Prevention, gave an overview of current arbovirus vaccines. Thomas W. Scott, a distinguished professor of mosquito-transmitted disease ecology and epidemiology at the University of California, Davis, spoke about types of vector control. Thais dos Santos, regional advisor for surveillance and control of arboviral diseases at the Pan American Health Organization (PAHO), discussed arboviral diseases surveillance and integrated surveillance in the Americas. And Oliver Brady, an associate professor at the London School of Hygiene and Tropical Medicine, described the various ways that models can be used in arbovirus surveillance and control. A question-and-answer session followed the four presentations.

ARBOVIRUS VACCINES

Paz-Bailey began by providing brief overviews of the vaccines that have been developed for chikungunya, Japanese encephalitis, Zika, West Nile virus, yellow fever, and dengue with a focus on the specific challenges

encountered in each of the vaccines’ development. Then she offered a list of lessons that could be gleaned from the experiences with these vaccines, focusing on the idea that these lessons could help the public health sector be better prepared to deal with the next arbovirus pandemic.

Chikungunya

VLA1553, the first-ever chikungunya vaccine to be licensed, is a single-dose vaccine manufactured by the company Valneva. It was licensed by the U.S. Food and Drug Administration under accelerated approval for adults aged 18 years and older; that approval is based on the persistence of neutralizing antibodies at up to 6 months after administration of the vaccine. Recent 2-year follow-up data show that 97 percent of those vaccinated maintained antibody titers, Paz-Bailey said. The vaccine is under review by the European Medicines Agency, while the World Health Organization (WHO) is at the initial stages of considering recommendations for global use. Importantly, she added, the Coalition for Epidemic Preparedness Innovations has provided funding for VLA1553 and other chikungunya vaccines to ensure that low- and middle-income countries have access to the vaccines.

Other chikungunya vaccines are in phase 3 trials, Paz-Bailey said, either completing the trials or in the middle of them. The first is a single-dose, virus-like vaccine manufactured by Bavarian Nordic, and the second is BBV87, manufactured in collaboration by International Vaccine Institute in South Korea and Bharat Biotech in India. It is an inactivated whole-virus vaccine, administered with a two-dose primary series.

The chikungunya vaccines face various challenges. The phase 3 trials provide data on immunogenicity and safety but not on efficacy, Paz-Bailey said. Since there is no way to predict where the outbreaks will occur, antibody titers are used as a correlate of protection; however, this endpoint has not been validated against efficacy data. Long-term protection for the vaccines is also still unknown, which hinders the accelerated approval process that requires data from post-licensure studies. Finally, the ideal implementation strategy remains unknown. Given that the outbreaks can be very localized and completed within a couple of months, it is challenging to respond quickly enough to make a major difference.

Japanese Encephalitis

Next Paz-Bailey spoke about vaccines for Japanese encephalitis, which generally affects humans in rural areas who live and work close to pigs. Vaccines for the disease have been available for decades, with most endemic countries now having vaccination programs. Gavi, the Vaccine Alliance, has expanded the use of WHO-prequalified vaccines for Japanese encephalitis.

SA 14-14-2 is a live attenuated vaccine developed in China that is available at low cost to low- and middle-income countries. Many manufacturers have developed inactivated Vero cell culture–based production and live attenuated vaccines, which are easier to manufacture. Ixiaro is the only Japanese encephalitis vaccine licensed and available in the United States (Vannice et al., 2021). The challenges facing the Japanese encephalitis vaccine are representative of those facing other arbovirus vaccines, Paz-Bailey said. They include predicting outbreaks, when transmission will occur, and producing enough vaccines to meet the global supply needs.

Zika

Many Zika vaccine candidates are in clinical trials, including a purified inactivated vaccine that has shown to be immunogenic after two doses, a live attenuated virus, and a chimeric measles vaccine that expresses some proteins of Zika. A DNA vaccine for Zika is currently in phase 2B trials in several sites in the United States and Latin America. Among the challenges facing Zika vaccines, congenital Zika syndrome may develop from infection at any point during pregnancy, which means that a Zika vaccine must provide protection against infection, Paz-Bailey said. Many other challenges exist: Cases have declined, so it is impossible to carry out phase 3 trials to assess efficacy, and there is no validated immune correlate of protection. Furthermore, the small number of participants in the phase 2 trials is not enough to allow the detection of infrequent but severe outcomes, such as Guillain-Barré syndrome. Ethical issues arise in conducting efficacy trials in pregnant women. Paz-Bailey shared that there are concerns about the interaction between the Zika virus and dengue with the possibility that vaccinating seronegative individuals who have not been exposed to dengue could increase their chances of contracting severe disease if they are later exposed to dengue. Also, the efficacy of the Zika vaccine may depend on an individual’s dengue exposure or serostatus. Finally, Paz-Bailey said, alternative licensing pathways are needed that could consider efficacy from human challenge models and extrapolate protection from animal models to humans.

West Nile Virus

Concerning the West Nile virus, Paz-Bailey said that several veterinary vaccines have been licensed for horses, but no human vaccine has been authorized. Several vaccine candidates have been evaluated in human trials, but none have progressed beyond phase 1 or phase 2. All these trials found minimal adverse events, and most found favorable immunogenicity. Concerning the challenges facing West Nile vaccines, the sporadic

and unpredictable nature of the virus makes it hard to plan efficacy trials. Severe disease is more common in those older than 50 years of age and with comorbidities, and there are concerns about vaccines causing adverse events for those in this group. Also, there is no consensus on what the trial endpoints should be, whether preventing neuroinvasive disease, preventing all diseases, or preventing infection. The endpoint chosen will affect the feasibility of conducting phase 3 trials. West Nile virus vaccination programs are expensive, and their cost-effectiveness is not clear. Finally, as with Zika vaccines, Paz-Bailey said, alternative licensing pathways are needed for West Nile virus vaccines that would consider approval based on immune protection in animal models and immunological markers.

Yellow Fever

Turning to yellow fever vaccines, Paz-Bailey said that a live attenuated yellow fever vaccine, 17D, has been applied safely and effectively for more than 80 years. One dose of the vaccine can generate long-lasting antibodies, and WHO no longer recommends a booster dose after 10 years. There have been no formal efficacy studies to determine how well the vaccine protects against specific adverse health outcomes, she said, but its effectiveness has been demonstrated through its use in practice in endemic areas and in laboratory workers. Fractional doses have been shown to be safe and effective, with one-fifth of a dose resulting in comparable protection for 10 years, so fractional doses have been used as a temporary solution for vaccine shortages. Several vaccine candidates for yellow fever are under development, with three of them now in clinical trials, but these vaccines would require multiple doses to achieve the same level of protective, long-lasting immunity that the current vaccine achieves (Montalvo Zurbia-Flores et al., 2022).

Speaking of the vaccine’s challenges, Paz-Bailey said that although adverse events are rare, there is a higher risk of occurrence in persons 60 years of age and older. The vaccine is contraindicated for pregnant and lactating women, infants under 6 months, persons older than 60 years, and those with severe immunodeficiency or hypersensitivity to eggs. Vaccine manufacturing is difficult to scale up because the vaccine is produced using traditional chicken-embryo methods, and even with six manufacturers globally, shortages of the vaccine still occur.

Dengue

For dengue, the Dengvaxia vaccine produced by Sanofi is a tetravalent, live attenuated vaccine. The vaccine requires three doses administered 6 months apart, so people need a full year before they are fully vaccinated. One problem with the vaccine is that it increases the risk of hospitalizations

among children who have not been exposed to dengue virus before, i.e., those who are seronegative. Thus, it is recommended to be used only among seropositive individuals. A laboratory test to assess the dengue serostatus must be done before administering the vaccine, which makes it very challenging to implement.¹ A second vaccine, QDENGA, produced by Takeda is a tetravalent live attenuated vaccine. It is administered in two doses given 3 months apart, and its efficacy is 61 percent against disease and 84 percent against hospitalization. In those who are seropositive, it protects against all four serotypes, while in seronegative individuals it protects against dengue-1 and dengue-2 but not against dengue-3, and there is insufficient data for dengue-4. WHO recommends that it be given to children 6–16 years in high-transmission areas, irrespective of their serostatus. A third vaccine, TV003 from Merck and the Butantan Institute, is live attenuated and requires one dose. Phase 3 trial results in Brazil found that after 2 years of follow-up, the efficacy against symptomatic disease was 89 percent for seropositive and 73 percent for seronegative.

A major challenge for dengue vaccines, Paz-Bailey said, is that they need to be four vaccines in one since they must provide protection against all four of the serotypes to avoid antibody-dependent enhancement. There is also no clear correlate of protection, and a longer period of observation following vaccination is necessary to make sure there is no increased risk of severe disease and hospitalization.

Lessons

In closing, Paz-Bailey offered several lessons. An ideal vaccine for arboviruses would be a single-dose vaccine that provides long-lasting protection, has a high safety profile, and is produced with cutting-edge technology to avoid the challenges of scaling up with traditional manufacturing. Vaccine development acceleration technologies are available, but funding has been a major obstacle for the progress of novel vaccines. It will be important to bring together public health institutions, governments, pharmaceutical companies, and nongovernmental organizations to establish priorities and settle on a united purpose, “as we are hoping we can achieve during this meeting,” Paz-Bailey added. Vaccine hesitancy can be very challenging to the implementation of vaccine programs, and it is something that should be addressed, noted Paz-Bailey. Finally, alternative licensing pathways could be prioritized for some vaccines.

___________________

¹ Production of the Dengvaxia vaccine has been discontinued by the manufacturer, and available doses will expire in August 2026. See https://www.cdc.gov/dengue/vaccine/index.html (accessed June 14, 2024).

VECTOR CONTROL

When vector control is done properly, it prevents disease, Scott said at the beginning of his remarks. This has been shown many times over, as with yellow-fever prevention efforts during the construction of the Panama Canal and in the hemisphere-wide campaign to eradicate Aedes aegypti from the Americas. More recently, programs in the 1970s and 1980s in Singapore and in the 1980s and 1990s in Cuba used adult and larval control to reduce dengue.

Despite these successes, vector control has not always been effective. Several years ago, experts realized that the evidence base for vector control tools that have been used for decades is too often not strong. Rather than including epidemiological outcomes demonstrating the impact on human disease or infection, most trials in the past were limited to entomological outcomes. A 2015 paper by Anne Wilson et al. addressed the design of phase 3 clinical trials for the assessment of vector control and concluded that the public health information such trials produced could enhance innovation for vector-borne disease control. After that, the Vector Control Advisory Group at WHO developed a detailed document explaining how to carry out a vector-control phase 3 trial (WHO, 2017). Many of the tools that have been used for the last 50 years have not been evaluated, Scott noted. However, Scott noted that many new tools are being evaluated in well-designed phase 3 trials, which proves helpful in developing public health policy.

Scott described the current interventions used for vector control and the innovative methods now being developed. The current interventions can be grouped into two categories: interventions that target immature mosquitoes and those that target adult mosquitoes. Getting rid of immature mosquitoes involves a combination of source reduction: getting rid of the sites where mosquito larvae develop and treatment of the sites to kill the larva or prevent them from maturing into adults. These site treatments can be carried out with insect growth regulators, chemical insecticides, biological treatments, or predation. Interventions for adult mosquitoes include space-spraying from trucks and aircraft, personal protection with topical repellants, and indoor residual spraying. These interventions can be either reactive or proactive. Reactive interventions are generally carried out in response to an increase in cases and use insecticides aimed at adult mosquitoes to quickly lower the number of adult mosquitoes in an area. Proactive interventions are intended to block outbreaks before they begin and generally combine methods aimed at controlling both adult and immature mosquitoes. The proactive approaches, Scott said, deserve greater emphasis and are the direction that the field needs to go.

Scott divided innovations in arbovirus vector control into two groups: innovations that are in development and innovations that are in or have

completed phase 3 trials. The in-development techniques include larval control with fungi and autodissemination; population suppression with such methods as Wolbachia, sterile insects, toxic sugar baits, and the release of insects with a dominant lethal gene; population modification with homing endonuclease genes; CRISPR-Cas9 gene-driven systems to create virus-resistant mosquito strains; and population reduction with sterile females or reduced female survival. Techniques that have gone through phase 3 trials include a population modification approach by the World Mosquito Program, which involves introducing Wolbachia bacteria into a mosquito population. Wolbachia infection in mosquitoes confers resistance to infection and, therefore, reduces virus transmission and human disease. A phase 3 trial of the technique in Indonesia led to an almost 80 percent reduction in virologically confirmed dengue, “which is really a remarkable outcome,” Scott said. The program is now carrying out a second trial in Brazil.

A trial in Peru tested spatial repellants, which are chemicals released into the air that interfere with mosquitoes biting humans. The results were promising, according to Scott, with a 34 percent reduction in Zika and dengue infections in humans. A second trial of the repellant began in late 2023 in Sri Lanka. Recently, Scott added, S. C. Johnson, the manufacturer of the product, has come out with a new product called Guardian, which the company claims will be effective for 1 year. This would be a major improvement, Scott explained, as the product used in the Peru trials had to be reapplied every 2 weeks, while the one in the Sri Lanka trials must be swapped out once a month. Another effective current innovation, Scott said, is targeted indoor residual spraying. An initial trial in Australia, where the targeted indoor residual spraying was combined with contact tracing, reduced dengue infection by 90 percent. Results from a trial now going on in Mexico should be available within the next year.

Looking to the future, Scott identified Africa as one area of growing concern for arbovirus infections. In the next 50 to 70 years, there will be a huge increase in the construction of urban environments around the world, and much of that will take place in Africa. “There’s an opportunity there for Aedes mosquitoes to get into those environments and create a public health disaster,” he said. According to a recent WHO assessment, many African countries have limited programs dedicated to Aedes vector surveillance, a lack of surveillance for insecticide resistance in Aedes mosquitoes, and no regular training session for specialists in vector control and surveillance of Aedes vectors. Much of the entomological expertise in Africa today is related to malaria and not so much with Aedes, Scott said, and this could be a growing problem in the future. He pointed to multiple resources that could help address this issue in Africa, such as the Global Arbovirus Initiative and Global Vector Control Response 2017–2030. There is also the West African Aedes Surveillance Network, which was formed in 2017

(Dadzie et al., 2022). A related publication by Roiz et al. (2018), offers a systematic, step-by-step process for the comprehensive control of Aedes-borne infections.

Speaking briefly about tick-borne disease, Scott pointed to three publications that discussed the best ways to protect people from tick-borne diseases (Eisen, 2020; Keesing et al., 2022; Stafford, 2004). These publications indicate that like mosquito control, many methods of tick control have not been evaluated for their public health value. However, there is growing interest in assessing the effectiveness of techniques to better inform the public health policies that are being developed.

In closing, Scott offered his thoughts on what people in the field should focus on most going forward. First, he said, it will be important to figure out which of the long-standing methods are effective in preventing human infection and disease and which should not be recommended any more. Noting the difficulty of getting funding for clinical trials of such accepted methods, he suggested finding alternative ways to test their effectiveness. Second, the innovation in vector control needs to continue. Creators of these new products should get together with the regulators to get the products out in the field as quickly as possible. Next, he continued, “we need an evidence base for determining the most effective delivery and coverage to reach and then to sustain the public health goals that we aim for.” This will require increased effort in implementation science. The National Institutes of Health now has a study section that funds implementation science, he said, but other groups should support it as well. Finally, Scott said, there is a growing consensus that no one approach will solve the problem by itself; combinations of interventions will be required. Thus, it will be important to build an evidence base concerning the effectiveness of various interventions, such as different types of vector control or vector-control programs combined with vaccines. However, carrying out phase 3 trials on various combinations of interventions would be logistically challenging and expensive. He suggested that there might be alternative ways to look at such combinations and decide which ones should move forward.

INTEGRATED SURVEILLANCE IN THE AMERICAS

Dos Santos began by noting that arboviruses have historically been a “high-visibility, high-priority topic” in the Americas. In 2003, PAHO adopted an integrated management strategy for dengue, and Zika and chikungunya were added in 2016. Unfortunately, that strategy has had little impact on transmission, dos Santos said. Part of the reason for the strategy’s ineffectiveness is because so many of the drivers of transmission are outside of the health sector. However, he did note that there was a lull in transmission in the post-Zika years in 2017 and 2018, and then, as the

COVID-19 pandemic progressed and social distancing measures were put into place, there was a marked decrease in arboviral transmission in the latter part of 2020 and 2021.

For dengue alone, in 2023 up to the time of the workshop, there had been more than 4 million cases reported in the Americas through passive surveillance systems, dos Santos said, which was more annual cases of dengue since PAHO began keeping track in 1980. However, the case fatality rate remained below the regional target of 0.05 percent due to a targeted strategy of identifying “early predictors of severe disease at the primary health care setting,” she said.

In the case of chikungunya, the year 2014—immediately after the virus’s introduction into the Americas in 2013—recorded the largest number of cases by far (nearly 1.1 million), and in the several years following 2014 the total number of cases steadily dropped until it was less than 10 percent of that initial value. It has trended up somewhat since then, and Brazil, Paraguay, and other South American countries have experienced epidemics since 2022.

Concerning Zika, dos Santos said, there was an epidemic in 2016, but there has been relatively little activity since. “We are not seeing the sentinels, the canaries in the coal mine, if you will, of neurological disease clusters.” It is difficult to get accurate data on the number of Zika cases because so many cases are asymptomatic or cause mild disease. For dengue, all four different serotypes have been circulating in the region, and the distribution of the serotypes has varied from country to country. In the first half of 2023, it was mostly dengue 1 and dengue 2 circulating, but in the second half of the year, dengue 3 and dengue 4 began playing a greater role in places like Guatemala, Costa Rica, and Mexico.

Next dos Santos spoke about collaborative surveillance and the Arbovirus Information Platform. The basic goal in implementing collaborative surveillance, she said, is to improve the amount and quality of evidence available for decision making. Granular data facilitates research and equips countries to guide their own interventions. To assemble that information, PAHO collects data from different streams, including case-based data, epidemiological surveillance data, entomological surveillance data, and laboratory data. PAHO transforms the country-provided data, unifying the data into a single format and depositing them into a single database. These transformed and aggregated arboviral surveillance data end up in two places. The first is PLISA, the Spanish acronym for the Health Information Platform for the Americas, which is publicly available.² The second is a collection of virtual collaboration spaces, which are private.

___________________

² PLISA can be accessed at https://opendata.paho.org/en.

The private collaboration spaces are “the mechanism through which we operationalize the collaborative surveillance,” dos Santos said, and PAHO epidemiologists and systems engineers work with the country to tailor the virtual collaboration space to the country’s needs and according to the data they have available. The potential for analysis depends on the number of variables collected by a country’s surveillance strategy, she said, “but we try to tell the story that the data is trying to tell us in a compelling way.” They use visualizations, charts, maps, and other techniques to make the data more useful, depending on what a country wants and needs, while still ensuring the security of the data. PAHO also has its own virtual collaboration space used to collaborate among the different areas of PAHO and the headquarters in Geneva. Some countries have agreements to share more granular data with each other through the virtual collaboration spaces than they make available publicly. There may be various details that the countries prefer not to be made public but that are important for neighboring countries to know so they can coordinate a multi-country response.

The virtual collaboration spaces offer a wealth of valuable information in various formats, such as epidemiological curves and maps, and in some countries the data are available at the subnational level. PAHO can provide special analyses, forecasts, and dashboards that allow officials in a country to get a clearer idea of what is happening in an outbreak or epidemic. When countries that have provided less information see the value that other countries are getting from the virtual collaboration spaces, she said, sometimes they will decide to start providing or collecting more data so that they can see the same benefit.

PAHO has been doing this for two years, working closely with the Centers for Disease Control and Prevention in Atlanta, during which time it has grown from 5 countries to 13. Speaking of lessons learned from this implementation, dos Santos noted the importance of trust, specifically that PAHO will not share or make data public unless a country authorizes it. A second lesson is that everyone should benefit. “This becomes sustainable when the country sees the benefit of this.” If a country finds what PAHO offers to be useful, it will continue to take part. A third lesson is that integrated analysis catalyzes functional integration. “Putting those databases from case counts with entomological indicators drives an integrated response,” she said, “because we are all looking at the same information, and we see our part in this map of telling the story of what’s happening with these diseases.” Fourth, PAHO has found its “come as you are” approach to be very useful. “We take the data however the country has it, in whatever format,” dos Santos said. Working with existing data contributes to improved data quality and timeliness. Finally, PAHO started in the places with the highest likelihood of success to gain experience before moving on to areas where implementation could be more challenging.

Looking to the future, dos Santos said that PAHO is hoping to modernize and automate surveillance processes that are currently still manual and obsolete. PAHO will continue to use all available collected data—epidemiological, clinical, laboratory, entomological—to better understand transmission dynamics and apply prevention and control techniques most effectively. PAHO will also strengthen its capacities to improve data quality, analysis, integration, and decision making in prevention and control within national and subnational technical teams.

PREDICTIVE TOOLS AND MODELING

Brady, in the final presentation of the session, identified four different purposes for models: to better understand the process of arbovirus outbreaks, to improve the precision of estimates, to make predictions about the performance of different intervention options, and to forecast what will happen to a system in the future. He offered a general observation that modelers should use the simplest model that will answer their question and that sometimes the best approach is to create a model designed to answer a specific question rather than a very complicated one that can answer a collection of questions.

Models for Improving Understanding

Brady first looked at models of key epidemiological parameters for improving understanding and, specifically, for testing one hypothesis versus another concerning such things as immunity, climatic drivers, different aspects of the environment affecting the dynamics and transmission of viruses, or the origins of epidemics. This is often applied to hypotheses that are difficult to test experimentally, so data are retrospectively analyzed. As an example, he described how a model was used to understand why the Zika epidemic that hit the city of Salvador in northeast Brazil in 2015–2016 ended when it did. There was a big jump in cases, with the peak hitting around July 2015, and then an equally sharp drop-off (Netto et al., 2017). The two competing hypotheses for the epidemic ending was that the seasons changed and caused the weather to become unsuitable or that the population reached herd immunity.

Using seroprevalence data, Brady’s group estimated the proportion of the population that had developed immunity, then fit and projected a model forward to 2016, assuming that the climate was similar to that of 2015. It predicted the epidemic would fizzle out because the percentage of the susceptible population had declined to a point where the epidemic could not continue, fitting what was subsequently seen in the real epidemic. The

hypothesis that the epidemic ended because the population had reached herd immunity was supported.

Brady’s group then took the understanding they got from that specific area in northeast Brazil to create a much more complicated model of 50 or so cities across Latin America to create projections of what would happen out to 2018. “We wouldn’t have been able to do that if we didn’t have a really clear understanding about what was driving the epidemic at a macro scale,” he said. Modern models for understanding have gotten much more complex, Brady said, specifically for dengue. A study analyzing dengue emergence in Vietnam found that while heavy rain and drought lead to more dengue, this relationship weakens in areas with higher provision of piped water, suggesting expanding access to piped water may protect against extreme weather events (Gibb et al., 2023).

Several factors have made such detailed models possible, Brady said. The biggest factor has been the rapidly growing amounts of publicly available data on disease cases that are presented at higher spatial and temporal resolutions. There are also many more datasets available on explanatory variables such as infrastructure, human movement, and climate and other environment variables. In the future, he predicted, the models will be improved by including causal inference methods, which should provide more of a theoretical basis to the models and perhaps lead to a more consistent understanding of the drivers of transmission.

Models for Estimation

Models for estimation can increase precision concerning specific parameters that are useful for arbovirus control, such as reproduction number, burden, and effectiveness. “Models can fill the gap where experimental measurement is impractical or impossible,” Brady said. For example, reproduction number, the average number of new infections generated by a single infected individual, is a key factor in determining whether an outbreak is growing or contracting. However, it is a very difficult number to measure directly, particularly for vector-borne diseases, because chains of transmission cannot be directly observed, so it must be inferred from case data. Burden must also be estimated with models. It is impractical to measure disease burden globally because there are too many data gaps. As a result, many of the disease burden figures that are given for dengue, yellow fever, chikungunya, and Zika are model-based estimates.

Models are also used for estimating the efficacy of interventions in a trial. For things like vector control, an intervention will apply to an area rather than individuals, so it is not possible to do a randomized, controlled trial on individuals; it is instead necessary to work with clusters, and gen-

erally there will be some level of contamination between the control and intervention clusters. Modeling makes it possible to get around this constraint, Brady said. “If we have a good idea about how much movement of mosquitoes, of people, of transmission coupling there is between these areas, we can take the results of the trial and remove that contamination to get a more precise estimate of what the true efficacy is.”

Concerning future developments related to using modeling for estimation, Brady shared that a key step will be expanding to other disciplines, such as measuring the effectiveness of vector-control programs. He also suggested that modeling could be used to inform trial design and that it would be useful to target data collection to improve modeled estimates.

Models for Prediction

Brady explained that modeling for prediction is scenario-based. It is about predicting the consequences of different intervention choices, such as figuring out the best long-term strategy for a new intervention or how best to combine multiple interventions together at a scale that cannot be measured experimentally. Being able to predict the consequences of interventions is increasingly critical for obtaining investment in new types of interventions. An example of this was the Sanofi Pasteur dengue modeling comparison group, which had seven different modeling teams predicting effectiveness across different settings (Flasche et al., 2016). It is very important, Brady said, to have multiple modeling groups with different assumptions and different interpretations of the data involved in these big decisions. In another case, a team led by Brady looked at how the cost-effectiveness of using Wolbachia would vary spatially (Brady et al., 2020). The modeling showed that the intervention is most cost-effective in high-density, high-burden cities, and now Wolbachia interventions often prioritize such cities. A third example is using modeling to predict the effects of combining interventions, which is generally not practical to study in trials. One recent study modeled the effects of combining dengue control programs in the state of Yucatán, Mexico (Hladish et al., 2020).

The enabling factors for model predictions have been good collaborations with intervention developers, governments, and international health organizations along with more work between modelers to build consensus on the models. Future developments will include more realistic and locally relevant predictions, which should improve robustness.

Models for Forecasting

The idea underlying the use of models for forecasting is fairly simple, Brady said. “For vector-borne diseases, there’s a lag between the climate

you observe and the outcome that you [later] observe with dengue cases. Rainfall leads to high [numbers of] mosquito[es] leads to infection, and there’s a 2-week to 1-month gap between those.” Thus, vector-borne diseases are more amenable to forecasting than some other diseases. The Dengue Forecasting Model Satellite-Based System (D-MOSS) for Vietnam is a particularly advanced model that includes a super-ensemble of 50 different models he said (Colón-González et al., 2021). The Vietnamese government uses it to decide how to respond to epidemics. It has also expanded to Malaysia and is starting to be used in Sri Lanka. The model’s short-term forecasts are useful for immediate control, such as spraying of insecticides. Medium-term forecasts help public health officials plan for the coming months, and long-term forecasts inform such decisions as the size of the budget that will be needed in the coming year.

In summary, Brady shared that a growing understanding of the drivers of arboviruses, more efficient Bayesian inference methods, and more relevant approaches to validation developed with ministries of health have enabled the development of more advanced models. Co-development with ministries of health, he continued, has also been instrumental in identifying the facets of disease outbreaks that need to be predicted. In the future Brady would like to see more evaluation of forecasts. “We want to do a clustered, randomized, controlled trial to show [whether] forecasting actually has an impact.” Concluding, Brady said that the development of models is not just about building better models but also is about working better with other groups so that modeling is used in practical situations to help with the control of and response to arbovirus outbreaks.

DISCUSSION

In response to a question as to whether PAHO might take unilateral action based on data from its monitoring network, dos Santos first noted that since PAHO is monitoring the data with the countries, the organization generally does not see things a country has not seen. However, she said, sometimes PAHO might notice a pattern across countries, such as more and more countries crossing their epidemic thresholds, at which point PAHO issues an “epi-alert,” which warns countries that they are anticipating an intense season for arboviral disease. PAHO issued six epi-alerts for arboviral diseases in 2023 alone, she said, but PAHO does not act in opposition to a country’s wishes.

Next, Scott was asked a question about how climate change may be affecting populations of vectors such as mosquitoes and ticks. Scott said that weather certainly affects vector populations, but connections to climate change are less clear. For example, Scott noted that laboratory research has shown that temperature fluctuations affect how viruses interact with the

vectors. Scott also mentioned a study he was involved in that concluded that climate change could affect the distribution of mosquitoes, although not necessarily their overall abundance “because in some places they would go away, and in other places they would show up.”

Climate change is important, Gubler commented, and it influences arboviral diseases, but the current focus on the effects of climate change on arboviral disease are misleading and taking away attention from other important issues. For instance, increased flooding is one effect associated with climate change, but in the case of most arboviruses, flooding decreases transmission instead of increasing it, “because it floods all of the larval habitats away, and it takes weeks to sometimes months for those mosquito populations to come back.” Furthermore, he continued, the focus on climate change can take money away from public health projects and research on vector-borne diseases, which instead goes to climate change research.

A participant asked Scott about indoor residual sprays leading to pesticide resistance in the mosquitoes, noting that residual spraying had not worked well in Singapore because of such resistance. It is a major problem, Scott acknowledged, but strategies such as targeted indoor residual spraying cuts down the amount of insecticide that is delivered and the amount of time it takes to deliver it. Lenhart expanded on that by discussing a large clustered, randomized trial in Mexico that is monitoring for insecticide resistance. After 2 years of targeted application, mosquitoes have retained susceptibility. Although the emergence of resistance is always going to be a risk when an insecticide is being used, Lenhart said, the approach being used in the trial is a “much more logical way of applying the insecticide” than typical schemes.

Paz-Bailey then offered a comment about new vector-control strategies. In particular, she pointed to the use of Wolbachia replacement methods, in which large numbers of Aedes mosquitoes infected with Wolbachia bacteria are released into an area to compete with and interact with the other mosquitoes there. Studies have shown that it greatly reduces the rates of transmission of such arboviruses as dengue, Zika, and chikungunya, and Paz-Bailey indicated that “in the field, Wolbachia replacement seems to show a lot of hope.”

Next, a question was posed about the need for a national or international clearinghouse structure among federal agencies and countries that could announce the detection of certain invasive species that may spread disease and could help speed up detection and response. Scott answered that effective detection efforts for Aedes-borne viruses must start at the local level, but ultimately “you would want to have something that was structured across countries within a region.” Brady added that such a system does exist in the European Union, and it has been important for tracking

arbovirus vectors through France, Germany, and the UK. However, he added, it is only useful in Western Europe because of large gaps in surveillance in Eastern Europe.

The session’s final question concerned the types of data that will be needed to get insight into the complex interactions that affect vaccine success. Dos Santos answered that more information is needed about what interventions are being done, where they are located, and the vaccine status of people living in the areas of interest. Brady added that the “ability to pick apart interactions is really boosted by the amount of follow-up,” even if just from regular collection of case data. Serological data are also important, he added.

Scott then spoke about the interactions between vaccine and vector control. Vaccines elevate herd immunity, and vector control lowers the force of infection. “The vaccine would make vector control sustainable.” However, experience in Singapore has shown that it is possible to have very low numbers of mosquitoes and still have an outbreak if herd immunity is low enough, he added. “So, the real question here is what sort of protection from infection do the vaccines give?” Scott continued. “If the vaccines don’t protect against infection, then this whole scenario doesn’t work.”

REFERENCES