Parasite Prevalence
A REM of prevalence effect sizes showed an overall effect that was not significantly different from zero (z = 1.818, p = 0.069; Figure 1a), with a large amount of heterogeneity between studies (I2 = 91.00%), and significant publication bias (𝞽 = 0.200, p = 0.005). The trim-and-fill method estimated 18 missing negative studies, but inclusion of these studies did not change the outcome, with the modified REM still showing no evidence of an effect (z = -0.795, p = 0.427). In our analysis of moderators, predator spreader identity was included in nearly all MEMs with non-zero weights (Importance = 0.993; Figure 2A, Table 2). Interaction type was also important (Importance = 0.771), but other main effects were less so (parasite type importance = 0.620; study design importance = 0.547). The most important interaction term was between interaction type and predator spreader identity (Importance = 0.665). In univariate analyses, only predator spreader identity significantly affected mean effect size (QM1 = 11.278, p = 0.0008), despite significant residual heterogeneity (QE87 = 630.561, p < 0.001). Predator spreaders had more positive effects than non-spreader predators (Figure 3).