Integrating infection intensity and aggregation into the dynamics of pathogens with within‐host replication Authors: Ruijiao Sun, Jason Cosens Walsman, Mark Wilber, Cheryl J Briggs Published: 2026 ABSTRACT Abstract Disease outcomes depend heavily on infection intensity which is often heterogeneous across and within host populations. Most individuals carry low pathogen loads and a few carry high loads, a pattern known as aggregation. Although well characterized in macroparasite systems, aggregation and infection intensity are rarely incorporated into microparasite models.This raises key questions: Do similar mechanisms underlie aggregation in macro‐ and microparasite systems? Moreover, how do aggregation and load‐dependent effects shape outcomes such as host suppression and virulence–transmission trade‐offs?To address these questions, we developed a series of differential equation models that allow the pathogen load distribution across hosts to evolve dynamically, shaped by both within‐ and between‐host processes. We applied this framework to the amphibian chytrid fungus system caused by Batrachochytrium dendrobatidis (Bd), a fungal pathogen threatening amphibian populations worldwide.Our results show that both stronger load‐dependent mortality and faster within‐host replication reduce aggregation. Aggregation, in turn, weakens host suppression and flattens virulence–transmission trade‐off, shifting peak transmission to higher replication rates.Overall, our models show that similar mechanisms of infection intensity and aggregation influence host–pathogen dynamics in microparasites as in macroparasites. This work offers a framework for advancing theoretical and data‐driven understanding of how within‐host processes scale to population‐level disease dynamics, advocating for a unified approach to disease modelling that bridges the macro‐ and microparasites. Keywords: chytrid fungus, host–pathogen interactions, infection intensity, load‐dependent mortality, microparasite, pathogen aggregation, pathogen load, virulence–transmission trade‐off INTRODUCTION The outcome of infectious disease often depends on the infection intensity of the host, as host survival (Wu, 2023), recovery (Wilber et al., 2016) and transmission potential (DiRenzo et al., 2014) are strongly influenced by the number of infectious agents carried by the host. Infections caused by microparasites (viruses, bacteria, fungi and protozoa) have mostly been treated as binary variables, with individuals classified as either infected or uninfected (but see Walsman et al. (2026), Wilber et al. (2021)). This is based on the assumption that within‐host replication of microparasites is much faster than the host lifecycle, so pathogen load immediately reaches a stable within‐host carrying capacity. This may obscure the biological reality that pathogen load varies significantly in pathogens that replicate within the host as well. Capturing these fluctuations may improve understanding of host population dynamics and refine predictions of disease spread and severity. Microparasite systems exhibit substantial variation in pathogen load within host populations (Bonneaud et al., 2019; de Jong et al., 2006), shaping host–pathogen dynamics. Heterogeneity in pathogen load can arise from host heterogeneity (e.g. immune response or nutritional state), variation in environmental conditions (Langwig et al., 2015) and within‐host processes. Following exposure to a pathogen, host individuals may exhibit a continuum of infection intensity, ranging from failed infections to extremely high pathogen loads. Many hosts carry low pathogen loads, while few hosts carry high loads, a phenomenon known as ‘pathogen aggregation’. Although the influence of aggregation on host populations has been primarily considered for macroparasites (e.g. helminths and arthropods) (Anderson, 1980), recent evidence indicates that microparasites exhibit similar patterns of overdispersion (Enriquez et al., 2022; Grogan et al., 2016; Schrock et al., 2025; Vale et al., 2013). This suggests that aggregation may similarly regulate host–pathogen dynamics in microparasite diseases. Many factors may drive pathogen aggregation. Studies of macroparasite systems reveal that host behaviour (Johnson & Hoverman, 2014), variation in host susceptibility (Wilson et al., 2002) and pathogen virulence can shape aggregation patterns (Anderson & Gordon, 1982). Macroparasite studies further suggest that within‐host pathogen replication can affect aggregation (Grear & Hudson, 2011). Recent models have made progress in incorporating pathogen load effects and within‐host processes in disease dynamics. Integral projection models (IPMs), which track heterogeneity in multiple continuously distributed variables within a population, have been adapted to monitor infection load in fungal pathogens (Walsman et al., 2026; Wilber et al., 2016, 2021). The reduced dimension IPMs allow for the distribution of pathogen loads to be represented by only a small number of moments, specifically the mean and variance. Initial reduced dimension IPMs tracked only the mean pathogen load (Wilber et al., 2021), and more recent work has extended these models to track both mean and variance (Walsman et al., 2026). Previous applications of this framework have shown that heterogeneity in pathogen load can reduce pathogen invasion probabilities, generate mutual feedbacks between load mean and variance that drive disease outcomes (Walsman et al., 2026) and dampen virulence–transmission trade‐offs (Wilber et al., 2021). While IPMs have provided valuable insight into load‐dependent dynamics, they have not yet been applied to explore how changes in continuous pathogen load distributions influence host population suppression effects which has important implications for community stability, conservation and applied contexts such as pest control. Here, we develop ordinary differential equation (ODE) models to study the dynamics of pathogen load and aggregation patterns over time. The framework focuses on environmentally transmitted microparasite systems in which pathogen load changes due to within‐host pathogen replication. Focusing on heterogeneity in pathogen load arising from initial infectious dose and within‐host pathogen growth, these ODE models are derived using the moment closure method, which reduces the dimensionality of a partial differential equation (PDE) model by approximating the full pathogen load distribution with a small number of moments. Specifically, moment closure allows us to track only the mean and variance of the pathogen load distribution. This approach is more computationally efficient and easier to implement than a full PDE model. We introduce two ODE models. Model 1 tracks only the first moment, dynamically modelling the mean pathogen load while assuming a fixed level of aggregation. This allows aggregation to be treated as a controllable input parameter, enabling exploration of its effects on disease outcomes. Model 2 tracks both the first and second moments of the PDE model, simultaneously capturing changes in the mean and aggregation level of pathogen load distribution. We apply our model to a fungal pathogen disease system. Emerging fungal pathogens have posed significant threats to global biodiversity. Species such as Batrachochytrium dendrobatidis (Bd), Ophidiomyces ophiodiicola and Pseudogymnoascus destructans, which infect amphibians, snakes and bats, respectively, have led to population declines and extinctions in hundreds of wildlife species (Fisher & Garner, 2020; Hoyt et al., 2021; Lorch et al., 2016). Unlike most other microparasites characterized by short generation times, high within‐host replication rates and lasting immunity, or macroparasites with longer generation times and limited direct reproduction, fungal pathogens represent an intermediate case. We apply our models to Bd, the fungus that causes chytridiomycosis and poses a global threat to amphibian populations (Fisher & Garner, 2020), with a specific focus on transmission from an environmental pathogen pool (Fernández‐Beaskoetxea et al., 2016; Kolby et al., 2015). Recent evidence suggests that Bd load distributions are consistently aggregated across diverse host species, though the degree of aggregation varies widely (Schrock et al., 2025). Using our models, we address three key questions: (Q1) How do pathogen load and aggregation affect host population dynamics? (Q2) What are the implications of pathogen load for the virulence–transmission trade‐off? (Q3) What are the underlying mechanisms affecting pathogen aggregation levels? We used Model 1, with a fixed level of aggregation, to explore Q1 and Q2, and used Model 2, with changing aggregation level through time, to explore Q3. MATERIALS AND METHODS Modelling pathogen load dynamics and load‐dependent effects To incorporate pathogen load effects, we first develop a full PDE model (Section 2.1.1) that includes pathogen load as an additional dimension. From this, we derive first‐order (Section 2.1.2) and second‐order (Section 2.1.3) approximations, each tailored to specific research objectives regarding pathogen aggregation. Model 1 (first‐order approximation) treats the level of aggregation as a fixed input parameter. This model allows us to investigate the effects of aggregation on disease dynamics. Model 2 (second‐order approximation) allows the aggregation level to evolve over time. This model captures the biological realism of temporal changes in variance, enabling us to investigate the specific mechanisms driving aggregation. By presenting these three models, we provide a flexible toolkit that allows future users to select the level of complexity appropriate for their specific data and research needs. The full PDE model We propose an infectious disease model that captures the dynamics of a continuous pathogen load distribution in infected hosts, incorporating within‐host pathogen growth and load‐dependent mortality. The model tracks three state variables: susceptible individuals, St; infected individuals with pathogen load x, ix,t; and free‐living environmental zoospores, Zt (Figure 1). While the model follows Zt on the natural scale, pathogen load x is represented on the natural log scale, x∈−∞+∞, to account for the wide range of pathogen loads observed in empirical studies (Wu, 2023). In microparasite systems, pathogen load often spans several orders of magnitude on the natural scale and is commonly quantified using qPCR, which yields continuous measurements that are routinely reported and analysed on the log scale. Representing pathogen load as a continuous variable on the logarithmic scale therefore makes the model more interpretable and statistically tractable, while also aligning it directly with how infection intensity is measured and interpreted empirically. Recent work on Bd indicates that pathogen load distributions are best described by a log‐normal distribution (Schrock et al., 2025). Thus, working on the logarithmic scale allows us to approximate the distribution as normal, following similar applications in IPM frameworks (Wilber et al., 2016, 2021). This enables the use of the moment closure method to reduce the complex PDE model into a tractable set of ODEs, providing easier analysis and application, as detailed in the following sections. A schematic of the disease cycle. The infection begins with the encounter between pathogens and susceptible hosts, initiating within‐host pathogen replication. Infected individuals may recover or die and can continue shedding pathogens into the environment, leading to the infection of additional susceptible hosts. Mortality is load dependent, with higher pathogen loads associated with an increased risk of death. Pathogen loads typically vary among infected hosts, often following a normal distribution on a log scale, with most individuals carrying low loads and a few carrying high loads. The relationship between pathogen load on the logarithmic and natural scales can be made explicit through a change of variables. Let y=ex denotes pathogen load on the natural scale, and let jy,t denotes the corresponding density with respect to y. The densities on the two scales are related by which implies For clarity, we note that the full model can equivalently be formulated on the natural scale (see Section S1.5), and the two formulations are mathematically consistent under this transformation. We include the natural‐scale formulation in the Appendix Section S1.5 to aid interpretation, while conducting analysis on the logarithmic scale to align with empirical measurements of microparasite load and to maintain analytical tractability. Hereafter, we refer to pathogen load on the logarithmic scale unless otherwise specified. The disease cycle (Figure 1) begins when susceptible individuals encounter free‐living zoospores in the environment, leading to infection with an initial pathogen load. Within‐host pathogen growth drives changes in pathogen load, while host mortality rates depend on the pathogen load. Infected hosts shed zoospores into the environment at a rate proportional to their pathogen load on the natural scale, driving new infections. The full PDE model is presented below: The dynamics of the susceptible population St are influenced by recruitment, recovery from infection, new infections and background mortality. Recruitment is modelled as a density‐dependent function RecruitHt=rHte−γHt, where Ht=St+∫−∞∞ix,tdx is the total host population, r is the maximum recruitment rate, γ is a parameter that determines the strength of the density dependence. The background mortality rate is d0. βZtSt is the transmission function and describes the rate at which susceptible hosts become infected, where β is the transmission coefficient. Recovery due to immune‐mediated clearance is given by ∫−∞∞lix,tdx, representing the total number of individuals recovering from infection across the pathogen load distribution. Here, we assume a constant recovery rate l, independent of pathogen load, and that clearance is relatively rapid rather than occurring through a prolonged decline in load (we further relax this assumption and present a model version with a load‐dependent recovery rate in Appendix Section S1.4). To account for the patchy distribution of free‐living organisms often observed in environmental reservoirs due to aquatic mixing and spatiotemporally heterogeneous shedding events (Cassie, 1962; Wiebe, 1970), we assume that free‐living zoospores are also aggregated and that the initial infectious dose follows a log‐normal distribution, following methods in Wilber et al. (2016), Wilber et al. (2021). Specifically, the initial load follows a normal distribution on a natural log scale, defined by the probability density function f0x=1σ02πexp−x−μ022σ02, where μ0 and σ0 denote the mean and standard deviation, respectively. While our model is specifically tailored to environmental transmission, the same framework could also be adapted to direct transmission. In that case, rather than specifying f0 independently, the distribution of new infections could be derived from the infected‐host load distribution. We follow the Kermack and McKendrick epidemiology model in Kermack et al. (1997) to derive the changes in the infected population ix,t with respect to pathogen load x and time t, based on the conservation of host individuals. The time evolution of the distribution is driven by the advective force of within‐host pathogen growth, which moves individuals through the continuum of pathogen loads x. In the absence of demographic turnover, the local change in density would be determined solely by this advection. In our model, however, this process is balanced by new infections, recovery and mortality. Load‐dependent mortality rate dx increases exponentially with the increase in pathogen load and is given by dx=d0+abx, where d0 is the baseline mortality rate, a (a>0) and b (b>1) determine how rapidly mortality increases with pathogen load (Appendix Figure S1). The rate of within‐host pathogen growth is given by Gx=rmax1−exK, assuming logistic growth on the natural scale with a maximum growth rate of rmax and carrying capacity of K. We assume that the pathogen shedding rate is proportional to the pathogen load on natural scale with λ representing the shedding rate per unit pathogen load per unit time. The total amount of log pathogen load in the population is measured by integrating over all infected individuals, where ex converts log pathogen load to the natural scale. We include the natural loss rate of environmental zoospores, denoted by dz. We also explore the properties of the PDE model with a diffusion term to simulate the behaviour of the host–parasite disease system with variance in within‐host pathogen growth as observed in many empirical studies (Appendix Section S1.1). Model 1: First‐order approximation Macroparasite models frequently assume that the distribution of parasites across hosts follows a negative binomial distribution, in which the degree of aggregation is determined by the dispersion parameter k, where a smaller k means a higher level of aggregation. Aggregation occurs when the variance is greater than the mean. Similar to macroparasites, recent evidence suggests that Bd load distributions follow a lognormal distribution among infected hosts (Schrock et al., 2025). Therefore, in our microparasite model, we assume that the distribution of the pathogen across hosts is best represented by a normal distribution on the log scale, where aggregation is determined by the variance of the distribution. A higher variance corresponds to a higher level of aggregation. Relying on this distribution assumption, we simplify the PDE model into a set of ODEs using the moment closure method. The first‐order approximation enhances both analytical tractability and computational efficiency by capturing changes in the mean pathogen load while assuming a fixed level of aggregation. By manipulating the level of aggregation in the model, we can explicitly explore the effects of pathogen load aggregation on the outcomes of infectious disease dynamics (Q1 and Q2). Here, rather than tracking the full distribution of pathogen load x for infected population ix,t, we track its moments. We track four key state variables: St, the population of susceptible individuals; It, the total infected population given by It=∫−∞∞ix,tdx; Pt, the total pathogen load in the infected population calculated as Pt=∫−∞∞xix,tdx on a natural log scale; Zt, the free‐living zoospores in the environments on the natural scale. Note that the governing ODEs are derived by multiplying the equation for ix,t by xj (where j=0 for the zeroth moment It, j=1 for the first moment Pt and j=2 for the second moment as detailed in Section 2.1.3) and integrating over the domain of x. We close the system by assuming a normal distribution of x with mean μt=PtIt and a fixed standard deviation σF. The final form of the first‐order approximation is: where μt=PtIt and αt=μt+lnbσF22. Detailed derivation is presented in Section S1.2. Model 2: Second‐order approximation To investigate the mechanisms contributing to pathogen aggregation (Q3), we also derive a second‐order approximation of our PDE model, which tracks both the first and second moment of pathogen load. The second moment of pathogen load is defined as Qt=∫−∞∞x2Ix,tdx. The variance of the lognormally distributed pathogen load, σt2, is given by σt2=Ex2−Ex2=QtIt−μt2. The final form of the second‐order approximation is: The mean of pathogen load is given by μt=PtIt and the variance is σt2=QtIt−μt2. Detailed derivation is presented in Section S1.3. Connection to load‐independent microparasite models Our model can be reduced to a load‐independent structure when the within‐host pathogen growth rate becomes infinitely fast (rmax→∞) such that pathogen load reaches its carrying capacity almost instantaneously after infection. This eliminates within‐host growth dynamics and variability, and the initial infection load approaches equilibrium load. As aggregation disappears, the pathogen load distribution becomes tightly concentrated around the equilibrium load, and all infected individuals carry approximately the same load. Consequently, all individuals experience the same mortality rate because they carry an identical pathogen load. Under these limiting conditions, pathogen load is effectively constant across hosts, and infection status alone determines transmission and mortality, consistent with the assumptions of standard load‐independent microparasite models. Application to the amphibian chytrid fungal disease system To demonstrate the usefulness of our approach and explore pathogen‐host dynamics in fungal diseases, we apply our model to the amphibian chytrid fungal disease system. Bd has caused severe declines and extinctions in many amphibian species worldwide (Fisher & Garner, 2020; Luedtke et al., 2023; Scheele et al., 2019). These declines are strongly influenced by pathogen load and load‐dependent mortality in infected frogs (Vredenburg et al., 2010). We parameterize our demographic parameters according to the California red‐legged frog (Rana draytonii), which is threatened by Bd and faces additional challenges due to its endangered status (Russell et al., 2019). Because limited experimental data on Bd exist for this species, we draw disease parameters from the closely related mountain yellow‐legged frog (Rana mucosa), which has similarly experienced declines driven by Bd and other stressors (Briggs et al., 2005, 2010). Detailed baseline parameter values are provided in Table S1. Using these biologically informed parameters, we apply our modelling framework to test whether mechanisms driving aggregation in macroparasite systems also apply to microparasites and to examine how aggregation and load‐dependent effects shape host population dynamics and the virulence–transmission trade‐off. RESULTS Analytical insights At equilibrium, the mean pathogen load, μ* (applicable to both Model 1 and 2), and variance, σ*2 (specific to Model 2), satisfy the following conditions: where i*x,t represents the density function of the pathogen load distribution at equilibrium, which follows a normal distribution with mean μ* and standard deviation σ*. The equilibrium of the first moment (i.e. mean pathogen load, Equation 4) and second moment of pathogen load (Equation 5) reflect the balance between new infections and pathogen growth within the host. In particular, Equation (4) shows that the equilibrium mean pathogen load is determined by the balance between pathogen input through initial infection and within‐host growth versus the removal of infected hosts through recovery and load‐dependent mortality. Similarly, the equilibrium of the second moment (Equation 5) reflects a balance between variance contributions from initial infection and pathogen growth, offset by reductions due to recovery and mortality. Load‐dependent mortality and within‐host pathogen growth affecting aggregation We used Model 2, which incorporates both changing mean and changing aggregation level of pathogen load, to explore the factors influencing pathogen aggregation (Figure 2, see results for full PDE model and Model 2 in Section S3). We quantify aggregation using the coefficient of variation (CV), defined as CV=σtμt, which represents the relative standard deviation of pathogen load. Higher CV values indicate greater aggregation. Our results show that both load‐dependent mortality and within‐host pathogen growth affect pathogen aggregation in the infected population (Figure 2). The rate at which mortality increases with pathogen load varies depending on pathogen pathogenicity and host tolerance. Here, we interpret host tolerance as the degree to which mortality increases with rising pathogen load. This relationship can also be interpreted in terms of pathogenicity, with steeper mortality increases corresponding to greater pathogen‐induced harm at a given load. Hosts with lower tolerance (higher b value) exhibit a steeper increase in mortality as pathogen load rises, whereas more tolerant hosts (lower b value) maintain lower mortality even at higher pathogen loads, resulting in a flatter load‐dependent mortality curve (Figure S1). As host tolerance decreases (i.e. the load‐dependent mortality curve steepens), the CV in pathogen load also decreases (i.e. decreasing aggregation, Figure 2a). This occurs because lower tolerance leads to fewer hosts with high pathogen loads, as these individuals experience high mortality. Consequently, the distribution shifts towards the initial pathogen load, with more individuals having similar, lower loads (Figure 2b). In contrast, when host tolerance increases, hosts can survive with relatively higher pathogen loads, leading to a broader distribution with heavier tails (Figure 2b). Impact of infection processes on pathogen aggregation derived from stable equilibrium solutions of Model 2. (a) Joint influence of load‐dependent mortality steepness (b, reflecting host tolerance) and within‐host pathogen growth rate (rmax) on pathogen aggregation, measured by the coefficient of variation [CV]; (b) Equilibrium pathogen load distributions under varying levels of load‐dependent mortality; (c) equilibrium pathogen load distributions under different within‐host growth rates (rmax). In (b, c), the plotted pathogen load distributions are lognormal distributions parameterized by the equilibrium mean and variance predicted by Model 2. The black dashed line represents initial load distribution of new infection shown in (b, c) for comparison. Within‐host pathogen growth also influences aggregation. Increasing within‐host replication increases both variance in load and mean load (Figure 2c), resulting in a slight reduction in CV (Figure 2a), and therefore a reduction in our measure of aggregation. This occurs because rapid within‐host pathogen replication allows hosts to reach higher pathogen loads in a shorter time without dying, while load‐dependent mortality selectively removes high‐load individuals, effectively truncating the upper tail and compressing the distribution. Host population dynamics In Model 1, where the aggregation level is fixed, we change σF to investigate how pathogen aggregation, load‐dependent mortality and within‐host replication affect stable equilibrium host abundance (Figure 3, see Section S3 for analysis using the full PDE model and Model 2). We show that, as pathogen load aggregation increases (i.e. increasing σF2), the pathogen's ability to suppress the host population declines (Figure 3a) because heavily infected individuals are removed from the population before contributing significantly to onward transmission. Correspondingly, the basic reproductive number R0 (see Section S1.2.3 for its derivation using the next‐generation matrix method following Diekmann et al. (1990, 2010)) decreases with increased pathogen aggregation (Figure 3c). Effects of infection processes on stable equilibrium host population dynamics and disease transmission potential (R0). Panels show the effects of (a–c) pathogen aggregation, (d–f) load‐dependent mortality and (g–i) within‐host pathogen growth on host population suppression, disease prevalence and the basic reproduction number (R0). An interesting result of our model is the breakdown of the relationship between R0 and prevalence when infection intensity and aggregation are incorporated. In traditional susceptible–infected–recovered (SIR) load‐independent models, equilibrium prevalence and R0 are correlated (Diekmann et al., 1990). In our model, with increasing pathogen aggregation (higher σF), prevalence initially increases even as mean pathogen load declines due to the selective removal of heavily infected hosts via load‐dependent mortality (Figure 3b). This occurs because the transmission driven by shedding increases initially as aggregation rises from low levels, as the term eμt+σF22 increases with σF. As a result, the infected population carries lower average pathogen loads, yet contributes to a larger environmental zoospore pool, leading to a rise in prevalence. Eventually, at high levels of aggregation, this trend reverses as elevated mortality severely reduces the infected host pool, causing prevalence to decline. Therefore, even as R0 decreases, prevalence can initially rise due to increased shedding, then decline as mortality becomes the dominant force. When load‐dependent mortality steepens, this initial increase in prevalence disappears because the effects of mortality dominate as aggregation increases (Figure S7). Furthermore, higher b value and a steeper load‐dependent mortality curve leads to reduced host suppression effects (Figure 3d), accompanied by decreased prevalence (Figure 3e) and R0 (Figure 3f) due to increased mortality among infected individuals. Faster pathogen replication rates enhance host suppression effects (Figure 3g). Prevalence decreases with increasing within‐host pathogen growth (Figure 3h), likely because individuals rapidly reach higher pathogen loads, leading to higher mortality rates of infected individuals. R0 exhibits a virulence–transmission trade‐off, maximized at an intermediate within‐host pathogen growth rate (Figure 3i). Virulence transmission trade‐offs We also used Model 1 to explore virulence–transmission trade‐offs (Figure 4, see results for full PDE model and Model 2 in Section S3). In our framework, increased within‐host pathogen replication raises pathogen load, which in turn increases both transmission and infection‐induced host mortality. We therefore treat within‐host replication as the trait under selection and use it as the explanatory variable in this analysis. A virulence–transmission trade‐off arises when increasing replication enhances transmission over the course of infection up to a point, beyond which further increases in replication shorten host lifespan sufficiently to reduce overall transmission. Factors influencing virulence–transmission trade‐offs at stable equilibrium. Pathogen aggregation (a), load‐dependent mortality (b) and the loss rate of environmental zoospores (c) shape the relationship between virulence and transmission. The dashed line indicates the level of within‐host pathogen growth that maximizes the basic reproduction number (R0). We find that the virulence–transmission trade‐off occurs only when the rate at which mortality increases with pathogen load exceeds the within‐host pathogen growth. Otherwise, the effects of pathogen replication remains dominant, leading to consistently higher transmission with increasing replication rates, and no virulence‐transmission trade‐off occurs. Since our model assumes logistic within‐host pathogen growth, we explore virulence–transmission trade‐offs for b values greater than e. We find that increased pathogen aggregation dampens the virulence–transmission trade‐off curve and reduces the maximum R0 (Figure 4a), as mean pathogen load decreases with higher aggregation (Figure S6). R0 is maximized at a higher within‐host pathogen growth rate under greater aggregation because replication does not uniformly raise individual pathogen loads, and the associated mortality costs are delayed. When host tolerance decreases (i.e. higher b values and a steeper load‐dependent mortality curve), the maximum R0 also decreases, but the reduction in fitness at high aggregation becomes steeper (Figure 4b). R0 is then maximized at a lower within‐host pathogen growth rate, as a lower growth rate is needed to offset the increased mortality of infected individuals. We also examined the effects of environmental zoospore loss rates (dz) on virulence–transmission trade‐offs (Figure 4c). Higher environmental loss rates reduce the maximum R0. Our model assumes that infection occurs exclusively from the environmental zoospore pool, with no direct transmission from infected individuals, so changes in environmental persistence do not impact the evolution of virulence. Consequently, the equilibrium mean pathogen load remains independent of the environmental loss rate, and R0 is always maximized at the same pathogen growth rate. DISCUSSION Infection intensity is a key source of individual infection heterogeneity. Recent evidence suggests that aggregation levels vary across populations (Schrock et al., 2025) and that both within‐host and between‐host processes may contribute to different aggregation levels. Progress has been made in addressing theoretical gaps using IPM models, particularly in understanding pathogen invasion and virulence evolution (Walsman et al., 2026; Wilber et al., 2021). To further investigate the factors driving aggregation and host suppression, we developed differential equation models that are both analytically tractable and easy to implement. Our model accounts for (1) pathogen load‐dependent mortality rates across a range of function behaviours, capturing different levels of host tolerance or virulence; (2) logistic within‐host pathogen growth; and (3) a continuous pathogen load distribution, approximated using a lognormal distribution that is both mathematically valid and ecologically meaningful. Using the amphibian‐chytrid fungal disease system as a case study, our results align closely with findings from macroparasite models. This highlights the importance of incorporating pathogen load and infection intensity heterogeneity to accurately infer host–pathogen dynamics. Different levels of pathogen load aggregation have been documented across many microparasite systems, both between populations of different host species and within the same host population under varying environmental conditions (Enriquez et al., 2022; Schrock et al., 2025; Vale et al., 2013). Our model predicts that both stronger load‐dependent mortality and faster within‐host pathogen replication reduce pathogen load aggregation. Empirical evidence from microparasite systems supports these predictions. In the Daphnia magna and the bacterial parasite Pasteuria ramosa systems, improved environmental conditions, which increase host tolerance and reduce pathogen virulence (Vale et al., 2011), are associated with higher levels of parasite load aggregation (Vale et al., 2013). In the context of our model, this environmental shift corresponds to a reduction in host mortality, allowing heavily infected hosts to survive longer, thereby generating the higher levels of parasite load aggregation observed in these improved environments (Vale et al., 2013). Thus, environmental factors that modulate key demographic parameters (e.g. host mortality rates, within‐host pathogen growth rates) can play an important role in determining aggregation patterns, offering insights into underlying epidemiological dynamics. Our findings align with previous theoretical predictions (Anderson, 1980), emphasizing the key roles of pathogen aggregation, load‐dependent mortality and within‐host pathogen growth in shaping host suppression effects (Figure 3). These dynamics also help explain empirical observations in Bd‐host endemic systems. For instance, in a recovered amphibian community affected by Bd, Fleay's barred frog (Mixophyes fleayi) exhibits high pathogen load aggregation and lower prevalence compared to sympatric species, underscoring the role of aggregation in modulating endemic disease patterns (Hollanders et al., 2023). Variation in tolerance and resistance also explains the divergent population trajectories observed in Bd‐impacted amphibian populations (Figure 2). For example, Bd‐impacted boreal toad (Anaxyrus boreas boreas) populations in Colorado and Wyoming exhibit different levels of disease prevalence (Muths et al., 2011). The Wyoming population shows higher prevalence, likely associated with greater host tolerance and slower within‐host Bd growth (Hardy et al., 2025), which is consistent with our model predictions that both increased host tolerance (Figure 3e) and slower within‐host pathogen growth (Figure 3h) can lead to higher prevalence. Given the strong interactions among load‐dependent mortality, within‐host pathogen growth and aggregation as suggested in our results, integrating these factors into microparasite models is essential for predicting population‐level outcomes and understanding why some host populations persist despite chronic infections. Our model focuses on fixed host traits, and the dynamic changes in pathogen load distributions are driven primarily by the interplay between within‐host pathogen growth and load‐dependent effects. However, systems such as the Mountain Yellow‐Legged frog (Rana muscosa) recovering from Bd demonstrate that hosts may evolve diverse defence strategies (Knapp et al., 2024), including tolerance to high loads, resistance to infection or resistance to pathogen growth. Recent theoretical work suggests that different defence mechanisms correspond to distinct patterns in pathogen load and that the distribution of load can differentiate among hypotheses for defence evolution (Walsman et al., 2026). Future work incorporating dynamic host traits in modelling pathogen load distribution would enable researchers to leverage pathogen load distributions to inform mechanisms of host adaptation and population recovery. Empirical evidence from macroparasite systems generally shows that higher levels of aggregation are associated with lower prevalence (Shaw et al., 1998; Shaw & Dobson, 1995). When we assume a load‐dependent mortality function that increases slightly faster than exponentially with log pathogen load in our microparasite model, we find that prevalence initially rises at low aggregation levels but subsequently declines as aggregation continues to increase, driven by the non‐linearity of load‐dependent mortality on the log scale (Figure 3b). When load‐dependent mortality is even steeper, prevalence decreases monotonically with increasing aggregation (Figure S7). Thus, realistic load‐dependent mortality relationships may be considerably steeper than linear on the natural scale, as supported by empirical studies (Gupta & Vale, 2017; Louie et al., 2016; Wu, 2023). Strong pathogen aggregation dampens the virulence–transmission trade‐off, consistent with the findings in Wilber et al. (2021). Although virulence–transmission trade‐offs have been widely predicted in theoretical studies, empirical support has been limited (Acevedo et al., 2019; Ebert & Bull, 2003). This is potentially because changes in aggregation influence the emergence of these trade‐offs, while most previous studies have assumed homogeneous pathogen loads (Alizon & van Baalen, 2005; Day, 2002b). We show that as the pathogen aggregation increases, R0 is maximized at higher within‐host pathogen growth rates, suggesting that aggregation may select for increased pathogen virulence. Previous research has primarily focused on static heterogeneity in host populations, with mixed predictions on its impact on virulence evolution (Ganusov et al., 2002; Osnas & Dobson, 2012). For example, random heterogeneity in demographic parameters has been shown to increase virulence with rising heterogeneity (Ganusov et al., 2002), while heterogeneity involving trade‐offs in parasite fitness predicts reduced virulence (Ebert & Mangin, 1997; Regoes et al., 2000). However, dynamic heterogeneity, such as variability in pathogen load resulting from factors not fixed for a specific host has received less attention. Our results demonstrate that dynamic heterogeneity can dampen the strength of virulence transmission trade‐offs while also altering the overall value of R0. Given that variability in infection intensity is nearly universal across infections, incorporating this dynamic heterogeneity is crucial for accurately predicting for the evolution of virulence. Our results also show that virulence–transmission trade‐offs depend on mortality accelerating faster with load than transmission does, aligning with previous theoretical predictions assuming a homogeneous pathogen load (Alizon & van Baalen, 2005; Day, 2002b). Since we assume transmission is proportional to host–pathogen load on the natural scale, the trade‐off is dampened when mortality increases less steeply with pathogen load (i.e. when b is larger but close to e) even when load‐dependent mortality is present. Unlike the results of the IPM model in Wilber et al. (2021), where steeper declines in survival with increasing load lead to higher maximum R0, our model predicts lower maximum R0 under steeper mortality functions. This difference arises because Wilber et al. (2021) alter the shape of the survival curve directly in a way that also increases survival for individuals with lower pathogen loads when the survival curve is steeper. Our findings emphasize that R0 is highly sensitive to the functional form of load‐dependent mortality, highlighting the importance of using biologically realistic functions. In our model, we assume transmission increases proportionally with pathogen load on the natural scale. However, in real systems, factors such as host behaviour and physiology can modify this relationship (VanderWaal & Ezenwa, 2016), potentially affecting the occurrence of virulence–transmission trade‐offs. Despite strong theoretical support for the virulence–transmission trade‐off, empirical evidence for this relationship remains limited (Froissart et al., 2010). Therefore, measuring the steepness of load‐dependent mortality curves and how transmission potential changes with pathogen load could help identify host–parasite systems where virulence–transmission trade‐offs are more likely to occur. When assuming that infection occurs exclusively from the environmental zoospore pool, we found that changes in environmental persistence do not impact the evolution of virulence. This finding is consistent with previous theoretical work showing that, for pathogens transmitted solely through free‐living propagules, the optimal virulence is independent of propagule persistence in the environment (Bonhoeffer et al., 1997; Day, 2002a). If additional transmission pathways (e.g. direct transmission or death‐mediated zoospore release) were present, dz would alter the relative importance of environmental transmission and act as a selective pressure on pathogen virulence strategies (Day, 2002a). An interesting future extension would be to adapt the framework to direct transmission, in which the distribution of new infections would be derived from the infected host load distribution and vary over time. Our results reveal a complex interplay between load‐dependent mortality and within‐host pathogen replication in shaping population outcomes and disease prevalence. Most previous models have treated microparasites and macroparasites as distinct, emphasizing within‐host multiplication for microparasites and external reproduction for macroparasites (Anderson, 1979). This has often led to pathogen load being tracked only in macroparasite models, while microparasite models only track infection status. However, this separation may oversimplify the continuum of parasite life‐history strategies observed in nature. Microparasite replication rates may vary widely across genotype of strains (Little et al., 2008), host populations (Hardy et al., 2025) and environmental conditions (Vale et al., 2011). Moreover, some macroparasites also exhibit direct replication on hosts (Stepek et al., 2006). Thus, there is a clear need for integrative models that bridge the microparasite–macroparasite continuum by incorporating pathogen aggregation, load‐dependent effects and within‐host dynamics to improve our understanding of host–parasite interactions and their demographic consequences for host species. Our differential equation models offer a computationally efficient, empirically relevant and analytically tractable framework for advancing our understanding of pathogen load dynamics. By incorporating load‐dependent effects and tracking changes in the distribution of pathogen loads among infected individuals, our results demonstrate that both pathogen load distributions critically influence host suppression, disease prevalence and invasion potential in microparasite systems. These findings highlight the importance of integrating load‐dependent processes and infection intensity heterogeneity into both theoretical and empirical studies of infectious disease. More broadly, our model provides a foundation for data‐driven inference using population‐level observations and offers a flexible tool for exploring host–parasite dynamics across a wide range of ecological contexts. AUTHOR CONTRIBUTIONS Ruijiao Sun: Conceptualization, methodology, investigation, interpretation and writing original draft. Jason Cosens Walsman: Methodology, interpretation, writing and editing. Mark Wilber: Methodology, interpretation, funding acquisition, writing and editing. Cheryl J. Briggs: Conceptualization, methodology, interpretation, funding acquisition, project administration, writing and editing. All authors provided final approval for publication and accepted responsibility for the work conducted. CONFLICT OF INTEREST STATEMENT The authors declare that they have no conflicts of interest. ETHICS STATEMENT All data used to inform model structure and parameterization were obtained from previously published studies and publicly available sources. No new field sampling, animal handling or laboratory experiments were conducted as part of this research. Supporting information Section S1. Model derivation and approximation using moment closure. Section S2. Parameters. Section S3. Results. ACKNOWLEDGEMENTS This work was supported by the University of California Climate Action 2023 Seed Award R02CP7252 to R.S. and C.B.; the Strategic Environmental Research and Development Program (SERDP) RC24‐4111 to R.S. and C.B.; USDA National Institute of Food and Agriculture Hatch Project 7001607 to M.W.; and National Science Foundation grant DBI‐2120084 to J.C.W., M.W. and C.B. Sun, R. , Walsman, J. C. , Wilber, M. , & Briggs, C. J. (2026). Integrating infection intensity and aggregation into the dynamics of pathogens with within‐host replication. Journal of Animal Ecology, 95, 1686–1698. 10.1111/1365-2656.70272 Handling Editor: Daniel Becker DATA AVAILABILITY STATEMENT Code used to generate the results in this study is available on Zenodo at: https://doi.org/10.5281/zenodo.19635241 (Sun et al., 2026). REFERENCES Associated Data Supplementary Materials Section S1. Model derivation and approximation using moment closure. Section S2. Parameters. Section S3. Results. Data Availability Statement Code used to generate the results in this study is available on Zenodo at: https://doi.org/10.5281/zenodo.19635241 (Sun et al., 2026).