The TIRS trial: protocol for a cluster randomized controlled trial assessing the efficacy of preventive targeted indoor residual spraying to reduce Aedes-borne viral illnesses in Merida, Mexico

Merida, the capital city of Yucatan State, is the largest urban center in the region with 892,000 inhabitants [35]. The city has a tropical climate characterized by a mean annual temperature of 25.9 °C and an annual precipitation of 1050 mm. Merida is endemic for ABVs, with DENV being persistently transmitted since 1979 and, more recently, co-circulating with CHIKV (since 2015) and ZIKV (since 2016) [36, 37]. ABV transmission in Merida is seasonal, beginning in July and peaking in October–November. Baseline serological information (captured by ELISA methods) on natural ABV infection rates has been collected from Merida in 2015–2016 through a school-based cohort that followed all family members living in the same household as the enrolled children [37‐39]. In 2015, DENV seroprevalence in the cohort was 70.2%, which increased with age from 31% in 0–8-year-olds to 79% in adults ≥ 20 years. In 2015–2016, the incidence of lab-confirmed ABV illness in the cohort was 14.6 per 1000 person-years (95% CI 10.8, 19.2) [37]. The incidence of symptomatic dengue infections observed during the same period was 3.5 cases per 1000 person-years (95% CI 1.9, 5.9). The majority of seroconversions occurred in the younger age groups (≤ 14 years old) [37‐39]. The incidence of symptomatic chikungunya illness was 8.6 per 1000 person-years (95% CI 5.8, 12.3) and the incidence rate of symptomatic Zika illness was 2.3 per 1000 person-years (95% CI 0.9, 4.5) [37]. Zika virus symptomatic attack rate in pregnant women from the cohort was 31% [40].

Data from ~ 40,000 geocoded DENV, 2273 ZIKV and 1101 CHIKV symptomatic cases captured by Mexico’s national passive surveillance system from 2008 to 2016 identified DENV transmission “hot-spots” in Merida (areas with higher-than-average numbers of cases), which overlapped with CHIKV and ZIKV hot-spots [36]. Combining these data with information from the cohort, we found that DENV seroprevalence rates are ~ 2× higher in hot-spot areas compared to other areas [36].

Merida also has entomological laboratory infrastructure and trained personnel to conduct and evaluate TIRS [28, 31]. The Collaborative Unit for Entomological Bioassays (UCBE) is a reference laboratory within the Autonomous University of Yucatan (UADY) and is currently a World Health Organization Good Laboratory Practice (GLP) site for evaluating insecticide products for vector control [41].

The two-arm CRCT will include a total of 50 clusters of 5 × 5 city blocks each, with 25 clusters randomly allocated to the intervention (TIRS) arm and 25 clusters allocated to the control arm (Fig. 2). Routine Ministry of Health (MOH) vector control actions performed in response to symptomatic ABV cases reported to the healthcare system will not be interrupted and could occur across both study arms. Upon detection of a suspected ABV case in the national epidemiological database, Yucatan MOH mobilizes its staff aiming at containing local transmission by focusing efforts on adult mosquito control. Truck-mounted ULV spraying with the organophosphate insecticides chlorpyrifos and malathion is widely implemented in Merida, despite scientific evidence of its poor efficacy [34]. MOH response also involves indoor space spraying (ISS) with pyrethroids (mainly deltamethrin) and organophosphates (malathion) in houses that allow entry. Limitations in personnel, geographic extent of outbreaks, and availability of resources (e.g., insecticides) commonly challenge MOH operations, reducing the coverage and effectiveness of their actions [34]. All MOH actions will be mapped and included in secondary analyses evaluating the impact of TIRS in addition to routine vector control. Participants in both arms will have access to any concomitant care they may choose to pursue, including cleaning their own yard and eliminating mosquito breeding habitats or using commercially available insecticide sprays or repellents (e.g., transfluthrin coils).

Clusters will be located within the areas previously identified as hot-spots of ABV transmission [36] (Fig. 2). Placing all clusters within areas of high ABV incidence will increase power because of higher event rates and decrease the potential for imbalance across trial arms. To reduce contamination and edge effects, while all households in TIRS clusters will be offered the intervention, epidemiological and entomological evaluations will occur in the center of each cluster, following a “fried egg” design (Fig. 2). Entomological interventions that are constrained to a given area suffer from immigration of mosquitoes from untreated neighboring areas, as observed in a recent study that released Wolbachia-infected mosquitoes in Fresno, CA, and quantified mosquito dispersal up to 200 m from their release point [42]. By focusing participant enrollment on the central 3 × 3 blocks of the 5 × 5 clusters, we will minimize any contamination in our primary and secondary endpoints emerging from mosquitoes flying into treatment areas (Fig. 2). This “fried egg” design is novel for vector-borne diseases and has been proposed as a rational approach to quantify the epidemiological impact of vector control [10]. To prevent selection bias, enrollment into the trial will occur in all clusters before TIRS allocation has been determined.

To assess power and sample size requirements, we analyzed historical passive surveillance data from the 192 hot-spot census tracts with population size of at least 1000 (from our previous work characterizing the ABV hot-spot area [36]). We used yearly data from 2008 to 2016 on the number of dengue, chikungunya, and Zika cases recorded in children 0–14 years each year by census tract [36]. Data were combined into pairs of adjacent years to mimic a 2-year trial period, and Table 2 summarizes the mean incidence (number of cases over 2-year period/number of children) and intracluster correlation coefficient (ICC) for a given 2-year period [43]. Assuming 4% incidence over a 2-year period, 70% TIRS efficacy, an ICC of 0.035, and 20% loss to follow-up, we will require 92 age-eligible children enrolled per cluster for an overall sample size of 50 clusters and 4600 children to have 80% power to detect a significant reduction in ABV incidence between arms (Table 3).

Clusters will be selected from the set of 190 census tracts within the ABV hot-spot area [36] that have a total population size of at least 1000 and at least 300 children aged 0–14 years, per the 2010 census (Fig. 2). Clusters are also selected to maximize the distance between the centroid of each cluster to the centroid of its nearest neighbor also in the trial. Given a set of 50 clusters, covariate-constrained randomization [44] will be used to limit imbalance across trial arms with respect to the following census tract-level variables: population size, per 2010 census; population density, per 2010 census; percent employed population, per 2010 census; and cumulative number of ABV cases between 2008 and 2016, per passive surveillance. These variables were selected because of their association with ABV transmission risk. For each balancing factor, only allocation patterns where the mean value of clusters in group A divided by the mean value of clusters in group B is within 1/1.1 to 1.1 are retained. Furthermore, we eliminate any allocation pattern with imbalance in the number of clusters per arm per sector greater than ± 1. To ensure randomization is not overly constrained, we only consider sets of 50 clusters that have many acceptable allocations into two groups of 25, satisfying validity criteria proposed by Moulton [44] (e.g., pairs of clusters always or never appearing in the same arm). Given the set of allocation patterns that meet the above balancing criteria, the biostatistics team at UF will use equal probability sampling to randomly select one allocation. A sample allocation pattern is plotted in Fig. 2. For participant enrollment, the study teams will be provided with a list of 50 census tracts for inclusion in the study, without a record of which census tracts are in group A or B. A random number generator produced by biostatisticians from UF will assign one group to TIRS and one group to control.

The trial will focus on the pediatric population, enrolling children aged 2–15 years in a longitudinal cohort to track their ABV illness and lab-confirmed seroconversion over two (and potentially three) transmission seasons (Fig. 2). The previously conducted cohort study in Merida indicated that the majority of dengue-naïve infections and seroconversions occurred in children ≤ 14 years old [37‐39]. By following children aged 2–15 years at enrollment, we will capture the segment of the population with the highest probability of ABV illness. We excluded younger children (< 2 years) because of the difficulties in obtaining blood specimens and potential for cross-reactivity with maternal antibodies [45].

There will be two levels of participation: at the household level and at the individual child level. Table 4 shows the inclusion/exclusion criteria for each level. For each participation level, consent (and assent) will be obtained, as follows. On August 2020, after being given time to review information about the intervention, one adult household decision-maker will be asked for written consent to have their house included in the trial (at the time of consent, neither study personnel nor householders will know to which arm of the trial the house will be allocated). In consenting houses with children meeting the inclusion criteria (Table 4), individual consent/assent will be obtained during December 2020–January 2021. Parental informed consent will be obtained for children aged 2–10 years, and both assent to participate from children and a parental informed consent will be obtained for 11–15-year-olds (Additional file 2). Enrollment of children will be focused in the central 3 × 3 city blocks of each cluster and will extend beyond if not enough children are enrolled in the core. Consent will be obtained in participants’ homes. Study explanations will be provided to small groups of adults present in the household, whereas written consent and assent will be obtained from each individual participant.

Engaging communities early in the trial will be essential for maximizing participant acceptance and retention [46, 47]. An experienced team of 10 social workers, who will interact directly with study participants (through informal conversations, games, and other educational activities with children), will ensure they remain engaged throughout the duration of the study [47]. Several factors may lead one household to withdraw from the intervention. Householders may sell their home and move to a different location, and we will consider them lost to follow-up. Householders may refuse to receive the intervention on a second or third opportunity, meaning they will not be subject to treatment (and therefore excluded from any future analysis). Our team will document voluntary withdrawals and communicate them as part of the trial reporting.

Table 5 shows our proposed milestones for the trial, following the SPIRIT checklist, and sections below provide information on each step (see Additional file 3). They can be divided into (a) trial planning, (b) TIRS evaluation, and (c) trial analysis and reporting. Trial design will be finished during the first year. Enrollment is expected to last up to 3 months, when all 4600 children will enter follow-up. Trial evaluation will occur for two transmission seasons, with the possibility of adding a third season should incidence of the primary endpoint be lower than assumed. Trial analysis will include a projection of TIRS impact, based on results from the CRCT, using our stochastic simulation model fitted to our study population.

Venipuncture procedures will be performed using standard aseptic techniques. An experienced phlebotomist will take the blood sample from an antecubital vein. Blood will be collected into Vacutainer® collection tubes or by a needle and syringe. A 22-gauge needle will be used for 5–15-year-olds, and a 23-gauge needle for children < 5 years. Blood specimens will be immediately taken to Yucatan State Diagnostics laboratory, dependent of the Ministry of Health for immediate molecular diagnostics (acute samples) or serum separation, followed by ELISA tests (convalescent samples and annual blood draws). Aliquots of all specimens will be stored at − 70 °C in labeled polypropylene cryogenic vials at UADY, and then transported to Emory University for advanced diagnostics. Long-term specimen storage will occur at Emory University. Specimens from individuals who did not sign the “future use” clause of the consent will be discarded after diagnostics, following sample processing procedures established by Yucatan State laboratory.

Figure 4 shows all lab testing components of the trial, which will occur at SSY, UADY and Emory University. Acute samples from active surveillance will be tested at the Yucatan State Laboratory using a multiplex reverse transcriptase-polymerase chain reaction (RT-PCR) [55] and virus-specific IgM ELISAs. Annual serologic samples will be tested at Yucatan State Laboratory by antigen capture ELISA for human IgG [56], and positive samples will be taken to Emory University for focus reduction neutralization testing (FRNT) [57‐59]. Natural ABV infection rates in Ae. aegypti will be detected by RT-PCR [55] at UADY. Standard CDC bottle bioassays [60] will assess phenotypic resistance of adult Ae. aegypti from treatment and control clusters pre-intervention and at 3 and 9 months post-intervention every year. F0, or F1 progeny, from field-collected eggs will be screened for susceptibility to pirimiphos-methyl [60]. If resistance is detected, both DNA and RNA will be analyzed from a subset of the phenotyped mosquitoes to calculate the frequencies of known resistance alleles as well as expression of resistance-associated genes.

Given cross-reactivity and variable sensitivity of assay methods, we will use a composite approach to diagnose ABV infections (Fig. 4). For active surveillance, two diagnoses are used: preliminary diagnosis—suspected cases are confirmed if RT-PCR is positive for any ABV. If RT-PCR is negative, the acute specimen IgM result is considered and any positive IgM result indicates a preliminary diagnosis of ABV infection. If both ZIKV and DENV IgM assays are positive, the case is designated as a case of flavivirus infection. Final diagnosis—paired acute and convalescent specimens will be tested for IgM and IgG seroconversion. These results will refine the case designation. A case with laboratory evidence of ABV infection in the acute testing must also demonstrate seroconversion or increasing levels of IgG or IgM in the convalescent specimen. RT-PCR+ suspected cases that do not exhibit seroconversion or increase in IgG or IgM levels will be designated recent infections, but not ABV cases (that is, this result is most consistent with an etiology other than ABV infection as the cause of the symptomatic illness). Additionally, IgG or IgM seroconversion or increasing IgG that is observed when RT-PCR and acute specimen IgM are negative will be considered confirmation of an ABV case. This approach may increase the sensitivity to detect ABV cases that have false negative PCR testing and have not yet mounted an IgM response at the time of presentation. Finally, if convalescent serology does not distinguish between DENV and ZIKV infection, the annual surveillance sample for that subject will be considered. If it is clear from neutralization testing on the annual surveillance specimen what the intervening viral infection was, that will become the designation of the ABV case captured during active surveillance.

The annual serologic surveillance takes into account that the majority of ABV infections are inapparent. It will also account for the known cross-reactivity among ABVs. CHIKV is an alphavirus, and serologic assays for CHIKV perform with high sensitivity and specificity. ELISA is likely sufficient for annual CHIKV serosurveillance. DENV and ZIKV are related flaviviruses, and conventional approaches to serologic diagnosis of flavivirus cases can exhibit reduced specificity. However, the antibody response to DENV and ZIKV is dynamic, and cross-reactive antibody levels are greatest in the first few months after infection. Thus, cross-reactivity is present but less intense in late convalescence, which is one reason for performing serosurveillance in the low-transmission season. For flavivirus surveillance, neutralizing antibody titers will be compared using the FRNT50 (inverse of serum dilution that exhibits 50% of maximum neutralization). Conversion of neutralization assays from negative to positive in subsequent years is strongly supportive of interval infection. The precise infecting virus (DENV1–4 serotype or ZIKV) can often be identified by comparing relative FRNT50 values for each virus. A ≥ 4-fold difference in the FRNT50 is considered a significant difference. Once an individual has high titers to multiple DENV serotypes, detection of additional DENV infection is challenging by serosurveillance alone. The details of interpreting all possible flavivirus neutralizing antibody profiles are beyond the scope of the article. We have reviewed the key concepts recently [61].

Emory University will coordinate all aspects related to data storage, management, and sharing. A data management core (DMC) provides timely and efficient curation and dissemination of study data from multiple sources (e.g., clinical, laboratory, passive surveillance, entomology, demographic, Ministry of Health interventions), all essential to the success of the trial (Fig. 5). Information from the trial including consent forms, surveys, active surveillance forms, laboratory diagnostics, entomological surveys, mobility surveys, withdrawal forms, intervention acceptability, and annual blood draws will be collected in paper form and digitally recorded into our REDcap database (see below) by the data entry staff at UADY. Staff will enter information in a private dedicated space at UADY-UCBE. Laboratory results at Emory University will be entered directly into the REDcap database by laboratory staff using an online form. All forms were developed by our team specifically for this study.

All data will be stored on secure data servers and kept strictly confidential (with participant identifiers blinded by using non-identifiable IDs). Households are assigned codes unique to the project database, which are then used to identify all subsequent data we will collect. Outside of the database, these codes will not be interpretable, rendering the data effectively unidentifiable without access to our servers. Blinding of identifiable data will occur in the analysis stage also. All diagnostics of specimens will be conducted using the sample ID, blinding laboratory personnel from any identifiable information or membership of samples to a given study arm.

Access to the database will be primarily administered through a custom, web-based interface with restricted access privileges and encrypted data transfer (REDcap, https://​www.​project-redcap.​org/​). Different data entry interfaces will be generated for each component. Access will be limited to certified project personnel and certified associates, who will be provided unique login and password combinations. Database servers will be protected by multiple layers of security. Databases will be shared electronically through secure servers among key project personnel for analyses, publications, oral presentations, and project development. Regular checks of the database for completeness and accuracy will be performed.

The heterogeneous nature of ABV transmission may dictate the need for a third transmission season to evaluate the epidemiological impact of TIRS. The decision to continue into a third season will follow an event-driven decision process. After the second season evaluating TIRS, the statistical team will quantify the number of total primary endpoints. We will pursue the following ranking in order to evaluate whether to stop or continue into a third season:

The choice of 90+ endpoints is based on our power calculations and represents the target number of events expected for a power of 80% and a TIRS efficacy of 70%. The choice of < 20 endpoints represents the target number of events needed for a power of 80% when TIRS efficacy is 90%.

Overall, the risks to study participants are minimal in all of our study procedures (Table 6). The most serious risk is related to potential intoxication with the insecticides used in TIRS (Table 6). Both Actellic 300CS® and Ficam® have been approved by the World Health Organization (WHO) for indoor control of mosquitoes [18, 19]. The WHO’s hazard assessments concluded that, when used for indoor residual spraying as instructed and at the recommended doses, both products do not pose undue hazards to the spray operators or residents of the treated dwellings [69‐71]. Provided that operational guidelines are followed, routine cholinesterase monitoring of spraying personnel during indoor residual spraying programs is not required [69‐71].

The study protocol and associated documents including informed consent forms are approved by the respective Institutional Review Boards (IRB) of all collaborating institutions as well the National Institutes of Health. The trial protocol was registered on clinicalTrials.gov (NCT04343521) on April 13, 2020. It will be made clear during the consent process that no information can be shared with anyone other than designated study personnel, the paper and computer files will be well protected, and we will ask that interviews be carried out one-on-one to prevent other family members listening in. Consent and assent forms include a separate section where participants give permission to the PI to keep their specimens for future tests or studies. We will take all necessary measures to ensure confidentiality. It will also be made clear to study personnel that any violation of confidentiality would be a fireable offense. All paper data forms will be stored in locked files or cabinets in UADY in a specified storage facility with limited access. Access to computer data files will be password protected to allow exclusive access to appropriate study personnel. The paper data forms associated with the project (e.g., consent forms, questionnaires, census) will be stored in accordance with IRB regulations. Should consent be given for future use, then serological samples will be stored indefinitely. The samples will not have any participant identifiers, beyond the participant’s code. If, however, consent for future use is not given, the blood samples will be destroyed immediately (using strict protocols at UADY for disposal of biological samples) following completion of the project. Monitor evaluations will occur once a year and will be timed to occur right after the epidemiological evaluation of TIRS (January–March). On every visit, Dr. Contreras-Capetillo will file a Monitoring Log and a Self-Monitoring Tool form. Self-monitoring will be performed on a random selection of 10% of study participants. The monitor will also review records of all adverse events as well as the information of any dropouts that occurred between monitoring periods. After the visit, the monitor will submit the Self-Monitoring Tool to the PI, together with any recommendations based on the visit. A phone call between the monitor and the PI will be scheduled, should corrective actions be required.