Literature survey and study selection
To identify studies that evaluated the effects of environmental stressors on infectivity and host fitness traits in host-parasite systems, we conducted a systematic literature search in Web of Science using the search terms: (parasit* OR pathogen* OR disease) AND (environment* OR temperature OR pollution OR resource OR provision* OR toxi* OR contamination) AND (infection OR load OR yield OR resistance) AND (”birth rate” OR ”death rate” OR surviv* OR mortality OR reproduct* OR fecundity). We limited our search to journal articles published in English between 2010 and 2020 and scanned titles and, if relevant, abstracts of all 20,684 hits. This initial screening was carried out by two experienced reviewers working independently. In addition, we identified ten more studies from the references of the selected studies. One experienced reviewer or two student reviewers further examined articles documenting effects of environmental stressors on infectivity and host fitness.
We classified stressors into three groups: 1) environmental factors, which can vary naturally but are also subject to human-induced perturbation (hereafter “endogenous environment”); 2) presence or quantity of chemical pollutants (hereafter “chemical pollution”), that lead to negative expected outcomes for hosts; and 3) resource availability for hosts (hereafter “resource limitation”). We excluded studies in which stressful and control environments differed by the presence of additional antagonistic biotic interactions (e.g., the presence of predators or competitors) or by the presence of substances purposely used as therapeutic interventions on infected hosts (e.g., chlorine added to water as a treatment against ectoparasites). Furthermore, we limited our search to studies with animal hosts and excluded studies on parasitoid infections (Fig. 1).
We included only experimental studies with hosts exposed to or infected by parasites under laboratory conditions. Infected hosts were either exposed to pathogens or infected in laboratory settings. Yet, they could have cleared the infection during the experiment and had undetectable pathogen loads by the time fitness traits were quantified. We included studies only if infected hosts were exposed to stressful and control treatments and if the authors reported data of both host fitness (fecundity and/or survivorship) and pathogen infectivity (prevalence and/or intensity) from the same experiment (i.e., for the same pool of individuals divided between stressful and control treatments) at matched timepoint(s) (Fig. 1). For example, if a study reported infection intensity at 24 h and 72 h post-infection (hpi), but survivorship was only recorded at 72 hpi, we used the 72h data exclusively. If a study recorded both fitness and infectivity at multiple time intervals, we included all matched intervals in data collection. We accounted for the non-independence of these effects and their sampling errors in the random structure of our statistical models (see below). We excluded studies in which parasite shedding was recorded as a proxy for infection intensity. Studies were further excluded for pseudoreplication, missing sample size information, or when estimates were reported without associated errors (Fig. 1).