What Modeling Parasites, Transmission, and Resistance Can Teach Us

Abstract

Veterinarians and farmers must contend with the development of drug resistance and climate variability, which threaten the sustainability of current parasite control practices. Field trials evaluating competing strategies for controlling parasites while simultaneously slowing the development of resistance are time consuming and expensive. In contrast, modelling studies can rapidly explore a wide range of scenarios and have generated an array of decision support tools for veterinarians and farmers such as real-time weather-dependent infection risk alerts. Models have also been valuable for predicting the development of anthelmintic resistance, evaluating the sustainability of current parasite control practices and promoting the responsible use of novel anthelmintics.

Keywords: Anthelmintic resistance; Climate change; Decision support; Disease; Model; Modeling; Parasite; Ruminant.

Fig. 1

Comparison of typical empirical (A) and mechanistic (B) modeling processes. Data inputs are shown in grey boxes (italicized font if viewing in grayscale) and key steps in the modeling process are shown in green boxes. The processes are described in detail in Box 1.

Fig. 2

Mechanistic model development is aided by conceptual frameworks, which visualize the current understanding of the system, help formulate mathematical equations, and identify key parameters. This conceptual framework details a model developed for the population dynamics of trichostrongylid gastrointestinal nematodes infecting ruminants. Based on previous research, it is known that eggs are deposited in dung and develop (transition) through 2 larval stages (L1 ad L2) to reach the third, infective, larval stage (L3). L3 then migrate (transition) out of the dung onto pasture, where the total L3 on pasture is partitioned between the soil and the herbage. Data in the literature were available to estimate death rates for each life-cycle stage, development rates from egg to L3, and bidirectional migration between the soil and herbage, based on temperature. Further controlled observations were required to estimate the influence of moisture on the rate of migration between dung and pasture. Because trichostrongylid nematodes share the same life cycle, the model can be adapted for different species by simply adapting the death and transition rates.

Fig. 3

One of the factors limiting the application of models developed to inform parasite control strategies for ruminant livestock is the availability of the data that can be used as model input. For example, weather stations may be some distance from the farm, and gridded weather data are often low resolution (eg, several km²) in comparison with the scale at which transmission takes place. Comparing output of an empirical model (“Ollerenshaw Index”; A) and a mechanistic model (“HELF: Hydro-Epidemiological Liver Fluke model”; B) for risk of F hepatica infection, for a river catchment area in Wales, UK, demonstrates how mechanistic modeling may provide a solution. High risk of infection is shown in orange, moderate risk is shown in gray, and low risk is shown in white. The empirical model output (A) may be useful to highlight larger regions at high risk of fasciolosis caused by high rainfall. However, a moderate to high risk was predicted throughout the catchment despite the fact that spatial risk often varies between and within fields owing to heterogeneity in suitable habitats for the snail intermediate host, which is determined in part by hydrologic processes. Beltrame and colleagues coupled a mechanistic model of hydrologic processes with a simple mechanistic model of the population dynamics of F hepatica to predict metacercariae abundance depending on rainfall runoff and soil moisture (B; high abundance is shown as high risk). This mechanistic model used the same low-resolution weather data as the empirical model (left) but was able to predict risk at a finer spatial scale (25 m) by coupling this with high-resolution topography (elevation) data. The model predicted that much of the area predicted to be moderate to high risk using the empirical Ollerenshaw index (A) was actually likely to be low risk (white, right). These results could be used to plan grazing strategies to avoid infection.

Fig. 4

Using models it is possible to simulate processes that are not easily measurable in the field (such as the development of anthelmintic resistance [AR]) over extended time scales. For example, Dobson and colleagues simulated the population dynamics of multiple trichostrongylid nematode species infecting sheep in Australia, and the development of AR in these populations in response to a range of treatment strategies. The efficacy of each strategy was expressed as a percentage delay in the development of AR over a 20-year period compared with control scenarios whereby flocks were left untreated or treated exclusively with monepantel (MPL), moxidectin (MOX), or a combination (COM). Simulations varied the percentage of adult stock left untreated and the anthelmintic products used, and were replicated using weather data from 3 regions in Australia. Data shown were extracted from Tables S2–S4 of the original publication. Points represent the output of model simulations. Simulations suggest that leaving even a small proportion of the flock untreated delays the development of resistance (A). However, how effective this strategy is in delaying AR was variable (eg, leaving 1% untreated results in approximately 60%–100% delay in AR; (A), depending on the treatments used (B) and regional weather/climatic conditions (C). Crucially, with the exception of MPL + COM combination treatment, which was always 96% to 100% effective, the optimal treatment strategy varied by region, highlighting the importance of considering environmental conditions in the development of sustainable parasite control strategies. −, treatments applied in rotation; +, treatments applied in combination; COM, combination treatment of benzimidazoles + imidazothiazoles + abamectin; MOX, moxidectin; MPL, monepantel.