The authors have declared that no competing interests exist.
Conceived and designed the experiments: DMP WEG HHW. Performed the experiments: DMP WEG HHW. Analyzed the data: DMP WEG HHW. Contributed reagents/materials/analysis tools: DMP WEG HHW. Wrote the paper: DMP WEG HHW.
Current address: Genesis Laboratories, Inc., Wellington, Colorado, United States of America
Visceral leishmaniasis (VL) is a disease caused by two known vector-borne parasite species (
Simulation results indicated efficacy of fipronil-based control schemes in reducing sand fly abundance depended on timing of drug applications relative to seasonality of the sand fly life cycle. Taking into account cost-effectiveness and logistical feasibility, two of the most efficacious treatment schemes reduced population peaks occurring from April through August by ≈90% (applications 3 times per year at 2-month intervals initiated in March) and >95% (applications 6 times per year at 2-month intervals initiated in January) relative to no control, with the cumulative number of sand fly days occurring April-August reduced by ≈83% and ≈97%, respectively, and more specifically during the summer months of peak human exposure (June-August) by ≈85% and ≈97%, respectively.
Our model should prove useful in
Visceral leishmaniasis is a disease caused by a virulent vector-borne parasite transmitted to man by phlebotomine sand flies. Fipronil-based drugs, administered to cattle orally, provide a potential means of sand fly control by permeating in cattle blood and being excreted in cattle feces, targeting adult females feeding on cattle blood and larvae feeding on cattle feces, respectively. An agent-based, stochastic simulation model was developed to represent sand fly population dynamics in a village in Bihar, India, at all developmental stages, with the goal of predicting the impact of various vector control strategies, utilizing drug treated cattle, on vector population numbers. Results indicate that success of treatment is dependent on the number of treatments applied annually and the seasonality of the sand fly lifecycle. Results further suggest that treatment schemes are most effective in reducing vector populations when high drug efficacy is maintained in cattle feces during periods of high larval density. Our approach incorporates detailed representation of the vector population and provides an explicit representation of the effects of insecticide application on adult and larval sand flies. Hence, this model predicts treatment schemes that may have the greatest potential to reduce sand fly numbers.
The deadliest form of leishmaniasis, visceral leishmaniasis (VL), is vector-transmitted through the bite of phlebotomine sand flies in the
The known VL vector on the Indian subcontinent is the sand fly species
It has been suggested that further research regarding alternative or integrated vector control approaches should be examined to supplement the current practice [
Fipronil-based drugs, orally administered to cattle and rodents, have been successful in killing laboratory-reared sand flies under controlled conditions, targeting blood-feeding adults and larvae that feed on host feces [
A reduction in vector density should lead to a reduction in the transmission rate of VL as suggested by a recent VL model which predicted that either reducing vector density >67% through application of adulticides or >79% through breeding site destruction could eliminate the ability of the VL pathogen to persist [
Susceptible-Infected-Recovered (SIR) compartment models, and variants of this, have been developed in the past to represent VL epidemiology within human populations on the Indian subcontinent. The first such model examined three historical VL epidemic peaks in Assam, India which occurred between 1875 and 1950 and concluded that intrinsic processes related to host and vector dynamics, rather than extrinsic factors such as earthquakes or influenza outbreaks, provided the simplest explanation of the timing of the peaks [
The model originally published by [
In this paper we describe an individual-based, stochastic, stage-structured model that represents a temperature-driven sand fly vector population within a village in Bihar, India and simulates the effects of vector control through the use of fipronil-based drugs orally administered to cattle. The model does not include a human population or VL pathogen, but rather focuses on the effects of fipronil-induced mortality of larval and adult life stages on sand fly population dynamics. We first describe the model and evaluate its performance. We then use the model to simulate several fipronil-based control schemes in which we vary treatment frequency and timing of treatment application, focusing on resulting reductions in sand fly populations during spring/summer and especially during the period of peak human exposure (June-August). We also examine sensitivity of model predictions of treatment efficacy to parametric uncertainty.
The model represents the lifecycle of sand flies as they develop from eggs to larvae to pupae to pre-reproductive adults to pre-oviposition adults to reproductive adults to post-reproductive adults, as well as fipronil-induced larval and adult mortality (
Sand flies are represented as eggs, larvae, pupae, pre-reproductive adults, pre-oviposition adults, reproductive adults, and post reproductive adults. Pre-reproductive adults require a blood meal to proceed with oviposition. Fipronil increases the mortality rate of adults feeding on treated cattle and larvae feeding on feces excreted by treated cattle. All processes represented by eqs
Below we present the equations used in the model to represent the development, reproduction, natural mortality, and fipronil-induced mortality of sandflies.
To calculate rates of development of immature stages (eggs, larvae, pupae), we drew upon results of laboratory experiments conducted under constant temperatures [
(A) eggs (
After pupation, pre-reproductive adults must obtain a blood meal to advance to the pre-oviposition stage (
Red squares represent the data points used to generate the curves.
We estimated the temperature-dependent development of adults from the pre-oviposition stage to the reproductive stage based on laboratory data collected by [
Females lay eggs the day they advance from the pre-oviposition to the reproductive stage. We represented the number of eggs laid per reproductive female (Eqs
If
If
After oviposition, reproductive females have a 90% chance of becoming post-reproductive and a 10% chance of returning to the pre-reproductive stage [
Natural mortality of cohorts of eggs, larvae, and pupae depend on the temperature (
If
If
If
If
(A) eggs (
We also represented density-dependent natural mortality of larvae based on rates of cannibalism observed in laboratory experiments conducted with different larval densities [
In addition to depending on the frequency of treatment application and the proportion of the cattle treated, which we represented as management variables, rates of fipronil-induced mortality depend on (1) the proportion of adult sand flies that feed on cattle, (2) the proportion of larvae that feed in organic matter containing cattle feces, (3) the efficacy of fipronil contained in the blood of cattle, and (4) the efficacy of fipronil contained within cattle feces. We assumed that 50% of adult flies obtain their blood meal from cattle [
We represented the efficacy of fipronil within the blood of cattle as decreasing exponentially as a function of the number of days after fipronil application:
(A) the decline in fipronil efficacy (measured as daily probability of mortality of adults) in cattle blood as a function of days post-application (
We represented the proportion of larvae dying due to fipronil within cattle feces as decreasing exponentially as a function of the number of days post-defecation [
For example, fresh feces deposited 1 day after cattle are treated have a higher efficacy than fresh feces deposited 2 days after cattle are treated (
To evaluate the potential usefulness of the model in simulating the population-level response of sand flies to fipronil-induced mortality, we first verified that the model simulated adequately the rates of development, reproduction, natural mortality, and fipronil-induced mortality observed under laboratory conditions. That is, that the model code produced simulated data that mimicked the laboratory data from which it was parameterized when we simulated the laboratory experiments. We next calibrated the model to represent environmental conditions typical of Bihar, India such that the simulated population established a seasonally-varying, dynamic equilibrium under baseline conditions (without fipronil-induced mortality). We then evaluated performance of the baseline model by (1) assessing the ecological interpretability of seasonal trends in the simulated sand fly life cycle and (2) comparing simulated fluctuations in abundance of adult sand flies to fluctuations observed in each of three villages in Bihar over a 12-month period. We ran 10 replicate stochastic (Monte Carlo) simulations for each portion of the model evaluation procedure, except for the simulations required for verification of the temperature-dependent development and mortality rates of eggs, larvae, and pupae, which were deterministic.
We assessed the potential efficacy of various schemes using fipronil-treated cattle to control sand fly populations by running 20 sets of simulations (10 replicate stochastic (Monte Carlo) simulations per set) in which we varied (1) the frequency of treatment application and (2) the seasonality of treatment application (
No. treatments | Jan. | Feb. | Mar. | Apr. | May | Jun. | Jul. | Aug. | Sep. | Oct. | Nov. | Dec. |
---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
1 | X | |||||||||||
3 | X | X | X | |||||||||
3 | X | X | X | |||||||||
3 | X | X | X | |||||||||
3 | X | X | X | |||||||||
3 | X | X | X | |||||||||
6 | X | X | X | X | X | X | ||||||
6 | X | X | X | X | X | X | ||||||
12 | X | X | X | X | X | X | X | X | X | X | X | X |
X indicates the month of treatment application. Treatment occurs on the first day of the month.
Simulated development times of eggs, larvae, and pupae were similar to those observed by [
Vertical bars in (A) represent ±1 standard deviation of development times at 20°C [
Simulated lengths of the pre-reproductive stage were similar to those observed by [
Vertical bars represent ±1 standard deviation. The number of eggs observed in the laboratory in part
Simulated numbers of eggs laid per female were similar to those observed by [
Simulated rates of natural mortality of eggs, larvae, and pupae were similar to those observed by [
Vertical bars represent the range of values.
Simulated probabilities of fipronil-induced mortality of adults were similar to those observed by [
We calibrated the model to represent environmental conditions typical of Bihar, India by representing annual fluctuations in (1) simulated air temperatures (
We evaluated the baseline model by (1) assessing the ecological interpretability of seasonal trends in the simulated sand fly life cycle and (2) comparing simulated fluctuations in relative abundance of adult sand flies to fluctuations in relative abundance of adults caught in light traps in each of three villages in Bihar over a 12-month period using a Sign Test. Simulated seasonal trends were representative of the general temperature-dependent trends characteristic of the sand fly life cycle in Bihar (
Red brackets indicate a generation of overwintering sand flies. Black brackets indicate the time between initial oviposition and the first post-winter peak in abundance of adults (P1). PL indicates the largest peak abundance of adults.
Simulated fluctuations in relative abundance of adults were not significantly different from the general trends in relative abundance of adults caught in the three villages in Bihar (sign test:
Field data were scaled (Rasulpur data x 500; Mahesia data x 280; Mohammadpur data x 450) to facilitate comparison of relative seasonal abundances.
Simulation results indicated that the efficacy of fipronil-based control schemes in reducing sand fly abundance depended not only on the frequency of treatment applications, but also on the timing of applications relative to the seasonality of the sand fly life cycle (Figs
Black brackets indicate April-August and the red brackets indicate the summer months of peak human exposure (June-August). Red boxes indicate the months of treatment application.
Black brackets indicate April-August and the red brackets indicate the summer months of peak human exposure (June-August). Red boxes indicate the months of treatment application.
Black brackets indicate April-August and the red brackets indicate the summer months of peak human exposure (June-August). Red boxes indicate the months of treatment application.
Black brackets indicate April-August and the red brackets indicate the summer months of peak human exposure (June-August). Red boxes indicate the months of treatment application.
Simulation results regarding the efficacy of fipronil-based control schemes in reducing the cumulative number of sand fly days (SFD) occurring annually, occurring during April-August, and occurring specifically during the summer period of peak human exposure (June-August) are summarized in
The x-axis labels indicate the number of treatments per year and the months in parentheses indicate the month of the first treatment each year. Bars represent means (±1 standard error) of 10 replications and vertical arrows of the same color indicate pairs of control schemes that were not significantly different from one another based on Fisher’s least significant difference (LSD) tests. All other pairwise LSD comparisons between control schemes indicated significant (
Closer examination of the simulation of the most efficacious scheme involving 3 treatments per year (3 treatments at 2-month intervals initiated in March) provides insight into the processes linking the frequency and timing of treatment applications to the seasonality of the sand fly life cycle (
Vertical red arrows indicate day of application and horizontal red arrows indicate duration of fipronil efficacy (~60 days). Black brackets indicate April-August and the red brackets indicate the summer months of peak human exposure (June-August).
Interestingly, with a single annual application in May, as well as with some other treatments in which the drug was administered one or three times a year, the abundance of adult sand flies actually exceeded baseline (non-treated) levels later in the year, usually during October or November (Figs
An important caveat associated with these results is related to the uncertainty with which we estimated the two parameters representing the proportion of adult sand flies that feed on cattle and the proportion of larvae that feed in organic matter containing cattle feces. Obviously, avoidance of the drug nullifies its effect. Analysis of the sensitivity of predictions of SFD accumulated during June through August under the most efficacious control scheme (treatments applied 3 times per year at 2-month intervals initiated in March) indicated that decreasing the proportion (X) of adult sand flies that fed on cattle from 1 to 0 resulted in an approximately 5-fold increase in SFD from ≈61,000 to ≈306,000 (SFD = 429.95X2 + 19497X + 41258, R2 = 0.9997). Decreasing the proportion of larvae that fed in organic matter containing cattle feces from 1 to 0 resulted in an approximately 18-fold increase in SFD from ≈54,000 to ≈964,000 (SFD = -4830.4X2 + 149729X − 99269, R2 = 0.9997).
Simulation results suggest that the success of fipronil treatments in controlling sand flies depends not only on the frequency of applications but also on the timing of applications relative to the sand fly lifecycle. Synchronizing applications to maintain high efficacy of the drug in cattle feces during the period of high larval abundance seems particularly important. While more frequent applications obviously are more efficacious, they also are more expensive and more difficult logistically. Thus, the ability to assess not only efficacy of treatment schemes
Several previous studies of VL epidemiology have focused on villages in Bihar and have included models with detailed representations of disease dynamics within human populations [
By explicitly representing the effects of seasonally-varying temperatures on development and survival of the various sand fly life stages, our model has allowed initial assessment of a novel control scheme targeted specifically at both larvae and adults. By specifically examining the relationship among the timing and frequency of treatment applications, the duration of drug efficacy, and the seasonality of the sand fly lifecycle, we can make initial assessments not only in terms of reducing average sand fly abundance, but also in terms of cost-effective reduction of human exposure to sand flies given local social practice and availability of alternative hosts.
As a common social practice, some family members of the vast majority of Bihari villager households sleep outdoors, particularly during the months with the hottest evening temperatures (June, July, August) [
Economically, Bihar is the poorest state in India, with roughly $100 million gross domestic product compared to the national average of $274 million [
Potential environmental and human health impacts, as well as effects on non-target species, always are a concern when evaluating new vector control methods. In this regard, treating cattle orally with fipronil-based drugs may have benefits over conventional IRS. IRS often involves the application of insecticides to the walls of homes and cattle sheds, thus exposing human inhabitants as well as non-target species coming into contact with the walls. DDT has known environmental consequences, but chronic exposure also could potentially be linked to human health concerns such as pancreatic cancer [
Further evaluation of the effects of sand fly control through the use of fipronil-based drugs orally administered to cattle ideally would involve a field trial in Bihar. Among the most critical data obtained from such an experiment in terms of increasing confidence in our model predictions would be those shedding light on the proportion of adult sand flies that obtain their blood meal from cattle and the proportion of eggs oviposited in organic matter containing cattle feces. By far the most restrictive assumptions we have made in our model are that 50% of adult sand flies obtain their blood meal from cattle [
The interaction of vector feeding tendencies and host availability on the success of vector control based on systemic insecticides is a topic of current investigation. A recent study modeled the effect of assuming different hypothetical functional relationships between biting behavior of mosquitos (e.g., indiscriminate, anthropophilic, zoophilic) and human host availability on subsequent predictions of malarial infections [
Empirical evidence regarding the substrates in which oviposition occurs is sparse. Sensitivity analysis indicated the obvious importance of assumptions which directly affect the exposure of simulated sand flies to the drug. Immature sand flies are typically collected from the floors of cattle sheds and human dwellings in India, but often in small numbers [
Other data limitations affecting parameterization of the present model included the need to use developmental data from laboratory studies of another species of Phlebotomus rather than field data on our target species. The use of
Notwithstanding the inevitable parametric uncertainties associated with the current model, we would suggest that our model structure might be adapted for initial evaluation of fipronil-based sand fly control under a range of different environmental conditions involving a variety of potential hosts. For instance, the VL vector in East Africa,
While 90% of reported VL cases occur in six countries on the Indian Subcontinent: India, Bangladesh, South Sudan, Sudan, Ethiopia, and Brazil [
(DOCX)
Details of the sampling procedure can be found in [
(DOCX)
We would like to thank Richard Poché of Genesis Laboratories, Inc. for providing suggestions which helped to enrich this manuscript.