Parasitoids
An REM of parasitoid effect sizes detected a statistically significant, negative, overall effect of predation on parasitoid abundance in prey (z = -6.919, p < 0.001; Figure 1c), with a smaller amount of heterogeneity between studies as compared to the analyses of parasite responses (I2 = 35.47%). While there was evidence of significant publication bias (đťž˝ = -0.227, p = 0.032), the inclusion of 9 missing positive effect sizes estimated by the trim and fill method did not eliminate the overall significant negative effect of predators on parasitoids (z = -4.630, p < 0.001).
DISCUSSION
The healthy herds hypothesis (HHH) (Packer et al.2003) predicts that predators should have negative effects on parasites in their prey, but empirical studies testing this hypothesis have reported a variety of different effects. We hypothesized that this variation is a result of nuances in predator-prey-parasite interactions, including transmission strategy of the parasite studied and the type of predator interaction manipulated. Specifically, we hypothesized that the negative effect predicted by the HHH would be larger for macroparasites and parasitoids than for microparasites and would only hold when consumptive interactions are manipulated and when those predators are not “predator-spreaders”. Using a meta-analytic approach that accounted for potential sources of variation in observed predator-prey-parasite interaction outcomes, we found that the main effect of predators on parasites in prey differed between parasites and parasitoids but not between conventional macro- and microparasites, with a net negative effect only present for parasitoids. Additionally, we found that interaction type (all vs. non-consumptive), and its subset of predator-spreader interactions, were most important in predicting the effect of predators on parasites in prey. These findings provide clear evidence that the HHH prediction is not universal. The degree to which it holds in a given system is both parasite- and context-dependent, but also predictable with limited information.
We observed significant heterogeneity across studies of the HHH resulting from substantial variation in the magnitude and direction of the main effect of predators on parasites in prey. We, therefore, sought to determine if there were factors that explained this variation in effects. First, we found that the difference between consumptive and non-consumptive interactions can explain variation in the effect of predators on parasites, but specific mechanisms of those interactions are also very important. In studies that measured intensity variables, the effect size significantly differed between interactions involving consumptive and non-consumptive interactions, with non-consumptive interactions having generally more positive effects. This result aligns with our prediction that consumptive interactions will have more negative effects on parasites compared with non-consumptive interactions. We note that our studies involving consumptive interactions typically were open to all sorts of interactions including non-consumptive, suggesting that this result may, in fact, be conservative. Our result for studies measuring prevalence variables contradicts this finding as consumptive and non-consumptive interactions were estimated to be nearly identical on average. We suggest that the difference between these two response variables is an artifact of the significant residual heterogeneity even in our best fit models. Most of this variation is likely hidden in unexplored mechanisms within these studies. Duffyet al. (2019) outlined 7 independent mechanisms whereby consumption can directly or indirectly impact disease in prey. For example, predators can selectively prey on uninfected individuals, shift host population structure toward more susceptible or heavily infected classes, and suppress competition between hosts allowing them to support more parasites. Unfortunately few studies provide sufficient information to assess which mechanisms are at play. Nonetheless, we were able to directly test this idea by including one of these mechanisms (predator-spreaders; (Cáceres et al.2009)) as a moderator variable since researchers typically identified this attribute of predators in their studies. As expected, predator-spreader identity was highly important for predicting the parasite outcome in the prevalence dataset, generally increasing parasite prevalence. The difference in the number of predator-spreader effect sizes between prevalence (n = 25) and intensity (n= 0) responses explains why we saw this effect emerge in the prevalence but not intensity dataset. Ultimately, the lack of universal support for the HHH is a result of the conflicting negative effects in studies of typical consumptive interactions versus positive effects in studies of consumptive predator-spreader interactions and certain non-consumptive interactions,
Second, unlike predator interaction type, we failed to detect an effect of parasite type in our analysis. We hypothesized that differences in the aggregation patterns of micro- and macroparasites would result in macroparasites having a stronger and more negative response to predator pressure than microparasites, but found no evidence for a difference between parasite types in either intensity or prevalence effect sizes and this variable was generally of less importance for explaining variation. This lack of an effect may be due to a number of factors. While one might expect random predation, or predation on infected individuals, to decrease parasitism more when parasites are aggregated (Packer et al.2003), the opposite is also true. Gape limited predators, such as many piscivorous fish and carnivorous snakes (Nilsson & Brönmark 2000; King 2002) that selectively prey on smaller and younger individuals may cause population demographics to shift towards larger, older and more heavily infected hosts (Dobson 1989; Nilsson & Brönmark 2000; Byers et al. 2015; Duffy et al. 2019). Alternatively, our assumption that high aggregation among macroparasites makes them more vulnerable to predation may be countered by the existence of significant aggregation in microparasite systems as well (Lord et al. 1999; Grogan et al. 2016).
Third, while there may not be a significant difference between micro- and macroparasites we saw a clear difference between parasites and parasitoids. Even when controlling for publication bias, predators had a significant negative effect on parasitoids as compared to the lack of any overall effect on parasites. Our ability to detect a strong directional effect for parasitoids is perhaps partly due to the uniformity across the studies in the parasitoid analysis, also supported by the more limited heterogeneity in the parasitoid REM. The negative direction of the effect may be due to the fact that consumptive effects of predators on parasitoids rarely include mechanisms that could produce positive effects. Predators rarely act in a “spreader” role for parasitoids in their prey because the larval life-cycle of the parasitoid is typically interrupted by predation (Naselli et al.2017). Perhaps most non-consumptive effects of predators on parasitoids concern free-living adult life stages, which may avoid areas with predators due to direct intraguild predation of predators on adult parasitoids (Heimpelet al. 1997; Brodeur & Rosenheim 2000). As a result, it is conceivable that parasitoids would display the a stronger negative response to predator addition than other parasitic organisms.
One of the main limitations of this study, as with all quantitative synthesis, is the selection bias in the field being synthesized. We detected significant publication bias in the literature in multiple directions. Particularly, our analysis of prevalence showed a significant bias towards publication of positive effect sizes, probably due to the abundance of predator-spreader associated effect sizes. In the case of parasitoids, however, there was significant evidence of publication bias for negative effect sizes. While correction for these biases did not influence qualitative conclusions, their presence does suggest the need for additional attention to the types of results published. Besides publication bias in effect sizes, we noted a number of important imbalances in study characteristics, particularly the lack of observational studies that inspected non-consumptive effects. We also found that studies which identified predators as predator-spreaders were largely limited to studies of microparasite prevalence. This finding suggests that the empirical dissection of consumptive effect mechanisms is not only limited to cases that are easy to characterize (like predator-spreaders), but also limited in taxonomic coverage. Given these gaps in the literature, we suggest the following priorities for future work: (i) examining the effect of non-consumptive predator interactions on parasites in non-manipulative field observations and (ii) further dissecting the effect of predator-spreaders and other types of consumptive interactions on both micro- and macroparasites.
Overall, we found that the healthy herds hypothesis is not broadly supported by the current literature. Instead, the average effect of predators on parasites in prey varies significantly according to the type of interaction being studied and whether the focus is on parasites or parasitoids. Our findings provide the first quantitative analysis supporting the growing consensus (Hethcote et al. 2004; Choisy & Rohani 2006; Holt & Roy 2007; Roy & Holt 2008; Duffy et al.2019) that predator effects on parasites are context dependent. Our results further suggest that the mechanistic basis of predator-prey interactions strongly influences parasite outcomes and that these effects are predictable.