Making weed biological control predictable,safer and more effective: perspectives from New Zealand

A persistent problem in weed biocontrol is how to reliably predict whether a plant that supports development in laboratory host-specificity testing will be utilized in field conditions, and this is undoubtedly preventing releases of safe and effective agents. Moreover, the potential for unanticipated undesirable indirect effects of weed biocontrol on ecological networks has raised concerns by policy-makers and the general public. The key to minimizing risks of non-target impacts is prioritizing candidate agents that are both host-specific and effective, such that the number of agents required to bring the weed under control is minimized. As a consequence both the weed and its biocontrol agents become minor components of the local biota. Here we review recent attempts in New Zealand to improve the predictive ability of host-range testing, to avoid potentially safe and effective agents being rejected. Research in New Zealand aimed at predicting whether an agent is likely to experience enemy-release (i.e. reduced parasitism and predation) could assist agent prioritization, potentially making biocontrol both environmentally safer and more effective.     

The EVect of Release Size on the Probability of Establishment of Biological Control Agents: Gorse Thrips ( Sericothrips staphylinus ) Released Against Gorse ( Ulex europaeus ) in New Zealand

Many biocontrol agents released against alien weeds and pests fail to establish in the field. Here, we ask whether better release strategies could improve the likelihood of successful establishment. A manipulative field experiment was used to investigate the relationship between the probability of establishment and the number of individuals released for a weed biocontrol agent. In this experiment, replicated releases of 10, 30, 90, 270 and 810 gorse thrips, Sericothrips staphylinus Haliday, were made on to isolated gorse bushes in New Zealand. The sampling eY ciency was determined using a further experiment in which known numbers of thrips were released on to bushes. The data obtained showed that in approximately nine out of 10 releases of 10 thrips, at least one thrips would be found. The thrips in the size of release experiment were sampled 1 year after their release. A higher proportion of the small releases became extinct during this time period: thrips were recovered from 100% of the releases of 270 and 810 thrips, but only from 33% of releases of 10, 30 and 90 thrips. Using gorse thrips as an example, a protocol was developed to determine the optimal release size for biocontrol agents. It is suggested that for a fixed number of insects available for release, smaller releases may increase the overall establishment rate. W hile a single large release can easily become extinct by chance, it is extremely unlikely that a large number of small releases will do so over the same time period. The optimal release size (i.e. that which maximizes the average number of successful establishments) for gorse thrips in New Zealand might be fewer than 100 thrips/ release site. This contrasts with the current strategy in New Zealand of 1000 thrips/release and the suggestion in the biocontrol literature that large releases optimize establishment. Over 1 year of observation the thrips had no eVect on gorse growth rate. The EVect of Release Size on the Probability of Establishment of Biological Control Agents: Gorse Thrips ( Sericothrips staphylinus ) Released Against Gorse ( Ulex europaeus ) in New Zealand     

Success and safety in the biological control of environmental weeds in New Zealand

Biological control, using specialist insect herbivores and plant pathogens, can be a self‐sustaining, cost‐effective and low‐risk tool for the management of environmental weeds. Agents have been recorded attacking non‐target plants in New Zealand and elsewhere, but the effects are usually minor and/or transitory. It seems probable that only two cases, worldwide, will result in significant damage to non‐target plants (representing 0.5% of the nearly 400 insect, mite, or fungal species used in classical weed biocontrol). Both of these cases were predictable from host range testing. Negative indirect, or ‘downstream’, ecological effects from specific weed biocontrol agents are difficult to predict and measure. They are probably insignificant compared to the impacts of the invasive plants that the agents are introduced to control. However, it is necessary to balance the risks associated with any introduction against the environmental benefits from controlling a weed to a predicted level. Recent analyses suggest that success rates are better than generally perceived. For New Zealand programmes, where enough time has lapsed to allow assessment, we calculate a full/partial success rate of 83%. Many of the costs associated with environmental weeds are difficult to quantify. Detailed risk assessment will make biological control programmes more expensive and time‐consuming, so that reliance on non‐biological management methods for environmental weeds may actually increase. The costs of biocontrol programmes against some New Zealand weeds can be kept down by using research already carried out in Australia and other countries, and the process is reciprocal. Developing international consortia of sponsors is also a potential way to fund programmes against weeds shared by several countries.     

Classical biological control of Cirsium arvense: Lessons from the past

Cirsium arvense (L.) Scop. is a perennial herb indigenous to Eurasia that is now present throughout temperate regions of the world where it is considered one of the worst weeds of pastoral and agricultural systems. Classical biological control has been attempted in both North America (NA) and New Zealand (NZ). However, nearly 50 years after the first agent releases there are no indications of successful control. We review the status of the five agents deliberately released for control of C. arvense in NA and NZ, plus the species unintentionally introduced, and the occurrence of insects native to NA on C. arvense. We retrospectively evaluate C. arvense as a target weed, critique the agents selected for release, and contrast the different situations in NA and NZ. In retrospect, we see justification for the agents released in NA, but it is evident that these agents would not meet the more stringent host specificity requirements necessary to be released today. The failure of the program in NA is attributed to compromised safety, and lack of impact. Non-target impacts by one of the released agents, Rhinocyllus conicus, have raised safety concerns for native thistle plants. The other released agents either failed to establish, or if established, had no impact on the weed. In contrast, the situation in NZ is quite different because there are no related native thistles (Cardueae), and thus little chance of non-target impacts. Thus far, failure in NZ is attributed to lack of effectiveness due to non-establishment, or no impact, of released agents. In the past, the same agents that were released in NA were subsequently released in NZ, without considering whether or not these were the best choices. Thus, the past failure in NZ might be due to the previous lack of a NZ-specific approach to biocontrol of thistles in general and C. arvense in particular. A new approach taking into consideration the absence of native Cardueae has resulted in the release of agents more likely to be effective, and has potentially set NZ on track towards successful biological control of C. arvense, and other thistles.     

Decomposition dynamics of invasive alligator weed compared with native sedges in a Northland lake

Invasive weeds have been shown to alter ecosystem processes such as decomposition and nutrient cycling. However, little is known about the effects of introduced biocontrol agents on these processes. This study examined the effects of alligator weed (Alternanthera philoxeroides) and its biocontrol agent, the alligator weed flea beetle (Agasicles hygrophila), on nutrient cycling in a northern New Zealand lake. Alligator weed litter decomposed significantly faster than either of two native sedge species (Schoenoplectus tabernaemontani, Isolepis prolifer) in a litterbag experiment. In addition, the presence of the alligator weed flea beetle resulted in large amounts of decaying alligator weed litter entering the lake in early summer. Both the timing and magnitude of this litter input were uncharacteristic of seasonal biomass dynamics of the native sedges. Combined with alligator weed?s rapid decomposition, this indicates altered patterns of nutrient cycling at the lake, with potential flow-on effects including facilitation of further weed invasion.     

Isolation and characterization of microsatellite loci from Lochmaea suturalis, the heather beetle (Coleoptera: Chrysomelidae), a pest in Europe and a biocontrol agent in New Zealand

The heather beetle Lochmaea suturalis which is native to northwest Europe has been released as a biocontrol agent for heather in New Zealand. We have isolated and optimized eight microsatellite loci from New Zealand beetles. These loci provide markers with high polymorphism ranging from four to 20 alleles per locus. Observed heterozygosity averaged 0.631 per locus. These results suggest the markers are useful for population studies that will contribute to assessment of L. suturalis as a biocontrol agent.     

Constraints in weed biological control: contrasting responses by implementing nations

The compilation of papers in this Special Issue (SI) derives from a Symposium at the 25th International Congress of Entomology, in 2016, entitled “Rise or demise? A global outlook on the future of classical biological weed control”. In the SI-opening-paper, a summary of the 5th edition of the world catalogue of weed biocontrol agents and their target weeds provides a comprehensive international perspective. Weed biocontrol implementation is beleaguered by perceptions of risk and restrictive regulatory procedures, notably in the USA, and less so in Canada. Thus, most of the papers in this SI comprise accounts of innovative responses to these challenges from scientists in the USA. Political and funding issues have inhibited weed biocontrol in Australia over the past decade, but there appears to be a gradual reversal of this trend in recent years. In contrast, in New Zealand and in South Africa, the practice is flourishing, and there are significant recent initiatives in Europe. Overall, the contributions in the SI suggest an optimistic prognosis for weed biological control.     

A scientific approach to agent selection

Abstract  The prioritisation of potential agents on the basis of likely efficacy is an important step in biological control because it can increase the probability of a successful biocontrol program, and reduce risks and costs. In this introductory paper we define success in biological control, review how agent selection has been approached historically, and outline the approach to agent selection that underpins the structure of this special issue on agent selection. Developing criteria by which to judge the success of a biocontrol agent (or program) provides the basis for agent selection decisions. Criteria will depend on the weed, on the ecological and management context in which that weed occurs, and on the negative impacts that biocontrol is seeking to redress. Predicting which potential agents are most likely to be successful poses enormous scientific challenges. 'Rules of thumb', 'scoring systems' and various conceptual and quantitative modelling approaches have been proposed to aid agent selection. However, most attempts have met with limited success due to the diversity and complexity of the systems in question. This special issue presents a series of papers that deconstruct the question of agent choice with the aim of progressively improving the success rate of biological control. Specifically they ask: (i) what potential agents are available and what should we know about them? (ii) what type, timing and degree of damage is required to achieve success? and (iii) which potential agent will reach the necessary density, at the right time, to exert the required damage in the target environment?     

Ecological, practical, and political inputs into selection of weed targets: What makes a good biological control target?

The topic of ecological, practical, and political considerations in the selection of weed targets for biological control has been widely discussed during the past two decades, mostly from the perspective of insect herbivores. For conceptual and practical purposes, plant pathogens have been treated in these discussions as if they are a subset of inoculative biocontrol agents, with little said about the inherent differences between pathogens and insects as biocontrol agents or the selection of weed targets for control by the inundative, bioherbicide strategy. Herein, I attempt to address the question of what makes a good biological control target for plant pathogens used as inoculative as well as inundative agents, basing my analysis on examples from the past three decades. Despite the small number of examples available for this analysis, the following generalizations can be made: (1) Weeds with robust capacity for vegetative regeneration are more difficult to control with pathogens than those that lack this trait. (2) A plant’s growth habit is not a reliable guide for target selection; weeds that have been successfully controlled include annual and biennial herbs, perennial shrubs, perennial vines, and trees, while numerous failures have been reported irrespective of the target’s growth habit or reproductive mode. (3) It is more challenging to control species with genetic heterogeneity and capacity for introgression than genetically homogeneous and reproductively conserved species. (4) Matching the target host’s susceptibility with the candidate pathogen’s virulence is of utmost importance for biocontrol success since host–pathogen interactions at the species and subspecies levels are often governed by single-gene differences (e.g., varietal specificity). (5) Practical and political considerations are central to the selection of targets for control with pathogens. (6) Demand from influential stakeholders for control and/or for a nonchemical or economically sustainable control typically drives the initiative as well as the continuance of biocontrol projects to their completion. (7) In the case of inundative, bioherbicide agents, the continuity and ultimate implementation of a project will be dictated by the prospects of economic returns from developing and using a pathogen. (8) The stakeholders’ perceptions of the effectiveness of a biocontrol program can be unpredictable, leading to conflicting views of “success.” In the final analysis, a good weed target for control by a pathogen is one that has strong stakeholder backing and the list of available pathogens for the target suggests a possibility of acceptable control at a cost that is competitive with those of other control options. While this conclusion is also applicable to target selection for insect biocontrol agents, it is more relevant for pathogens because of limited funding and personnel available for development of pathogens and the added cost and technological complexity of implementing bioherbicides compared to classical biocontrols.     

Dispersal of Galerucella pusilla and G. calmariensis via passive water transport in the Columbia River Estuary

Dispersal of biological control agents and their subsequent population growth can be a major determinant of the success of landscape-scale weed control programs. Biocontrol agents must be able to disperse across the distances between patches of host plants in order to colonize and control their targets. The presence of three species of biocontrol agents for purple loosestrife (Lythrum salicaria L.): Galerucella calmariensis L. (Coleoptera: Chrysomelidae), Galerucella pusilla Duftschmid (Coleoptera: Chrysomelidae), and Nanophyes marmoratus Goeze (Coleoptera: Brentidae), on relatively remote islands in the Columbia River Estuary (CRE) indicate that these organisms have the ability to disperse across large expanses of open flowing water to colonize remote sites. Previous studies suggest that colonization of these islands by active flight is highly unlikely; therefore, some other dispersal mechanism must be responsible for colonization. A spatial database of all known biocontrol agent release sites for purple loosestrife within 68 river kilometers of our CRE study area was developed and field surveys for biocontrol agents were conducted. A GIS was used to model dispersal distances between biocontrol agent recovery sites and the nearest conspecific release site. Tidal water flow within the CRE was assessed as a potential dispersal mechanism across the modeled distances. The ability of the biocontrol agents to withstand submersion was evaluated in field tests. Our results indicate that it is highly likely that passive water transport has been responsible for some of the long-distance open-water dispersal that would have been necessary for colonization of the remote islands where biocontrol agents were recovered.     

Grasses as appropriate targets in weed biocontrol: is the common reed, <Emphasis Type="Italic">Phragmites australis</Emphasis>, an anomaly?

Despite their importance as invasive species, there has been a hesitation to target grasses in classical biocontrol. This historic bias appears to be changing with multiple active research and release programs. Similarly, biocontrol workers appear to avoid targeting species with native congeners. These biases appear inappropriate as the ecological and entomological literature provide abundant evidence for sub-genus specificity for many herbivores, including those attacking grasses. The biocontrol program targeting Phragmites australis (Cav.) Trin. ex Steud (Poaceae) provides an informative example with endemic subspecies in North America and many sub-genus specific herbivores, including potential European control agents. Grasses and target weeds with congeneric native species require rigorous host range testing, similar to all other targets in current weed biological control programs. Furthermore, it appears prudent to ask petition reviewers and regulatory agencies to abandon their focus on results of no-choice studies and to distinguish between trivial feeding and demographic impacts.     

After biocontrol: Assessing indirect effects of insect releases

Development of biological control agents for weeds has been motivated by the need to reduce the abundance and distribution of a pest plant where chemical and mechanical control were not cost effective. Primary objectives have been direct reduction in abundance of the target and, secondarily, the increase of desirable species. Recently, wildland weeds have become a focus of biological control projects. Here, desired outcomes include both reduction of the target and indirect effects of increased diversity and abundance of native species and restoration of ecosystem services. However, goals and benefits of biocontrol programs are not always well-articulated and direct and indirect impacts are not easily predicted. We evaluated the extent to which several successful biological control projects for weeds of rangelands and waterways measured indirect impacts on invaded ecosystems. We also examined biocontrol of a wildland pest tree for which the principal objective is restoration of ecosystem services. We found few quantitative assessments of the impacts of pest plant reduction on community composition or ecosystem processes. All examples documented variation in the impacts of agent(s) across the invasive range of the target plant as well as variation in impacts on the invaded ecosystem. However, without appropriate quantitative information, we cannot evaluate site characteristics that may influence vegetation responses. Most successful weed management programs integrated the use of biocontrol agents with other weed management strategies, especially modifications of disturbance and competing vegetation. Discussion and evaluation of responses of nontarget species would improve our understanding of the context-specificity of outcomes.     

Ecological,practical, and political inputs into selection of weed targets: What makes a good biological control target?

The topic of ecological, practical, and political considerations in the selection of weed targets for biological control has been widely discussed during the past two decades, mostly from the perspective of insect herbivores. For conceptual and practical purposes, plant pathogens have been treated in these discussions as if they are a subset of inoculative biocontrol agents, with little said about the inherent differences between pathogens and insects as biocontrol agents or the selection of weed targets for control by the inundative, bioherbicide strategy. Herein, I attempt to address the question of what makes a good biological control target for plant pathogens used as inoculative as well as inundative agents, basing my analysis on examples from the past three decades. Despite the small number of examples available for this analysis, the following generalizations can be made: (1) Weeds with robust capacity for vegetative regeneration are more difficult to control with pathogens than those that lack this trait. (2) A plant’s growth habit is not a reliable guide for target selection; weeds that have been successfully controlled include annual and biennial herbs, perennial shrubs, perennial vines, and trees, while numerous failures have been reported irrespective of the target’s growth habit or reproductive mode. (3) It is more challenging to control species with genetic heterogeneity and capacity for introgression than genetically homogeneous and reproductively conserved species. (4) Matching the target host’s susceptibility with the candidate pathogen’s virulence is of utmost importance for biocontrol success since host–pathogen interactions at the species and subspecies levels are often governed by single-gene differences (e.g., varietal specificity). (5) Practical and political considerations are central to the selection of targets for control with pathogens. (6) Demand from influential stakeholders for control and/or for a nonchemical or economically sustainable control typically drives the initiative as well as the continuance of biocontrol projects to their completion. (7) In the case of inundative, bioherbicide agents, the continuity and ultimate implementation of a project will be dictated by the prospects of economic returns from developing and using a pathogen. (8) The stakeholders’ perceptions of the effectiveness of a biocontrol program can be unpredictable, leading to conflicting views of “success.” In the final analysis, a good weed target for control by a pathogen is one that has strong stakeholder backing and the list of available pathogens for the target suggests a possibility of acceptable control at a cost that is competitive with those of other control options. While this conclusion is also applicable to target selection for insect biocontrol agents, it is more relevant for pathogens because of limited funding and personnel available for development of pathogens and the added cost and technological complexity of implementing bioherbicides compared to classical biocontrols.     

Population genetic structure of a colonising, triploid weed, Hieracium lepidulum

Understanding the breeding system and population genetic structure of invasive weed species is important for biocontrol, and contributes to our understanding of the evolutionary processes associated with invasions. Hieracium lepidulum is an invasive weed in New Zealand, colonising a diverse range of habitats including native Nothofagus forest, pine plantations, scrubland and tussock grassland. It is competing with native subalpine and alpine grassland and herbfield vegetation. H. lepidulum is a triploid, diplosporous apomict, so theoretically all seed is clonal, and there is limited potential for the creation of variation through recombination. We used intersimple sequence repeats (ISSRs) to determine the population genetic structure of New Zealand populations of H. lepidulum. ISSR analysis of five populations from two regions in the South Island demonstrated high intrapopulation genotypic diversity, and high interpopulation genetic structuring; PhiST = 0.54 over all five populations. No private alleles were found in any of the five populations, and allelic differentiation was correlated to geographic distance. Cladistic compatibility analysis indicated that both recombination and mutation were important in the creation of genotypic diversity. Our data will contribute to any biocontrol program developed for H. lepidulum. It will also be a baseline data set for future comparisons of genetic structure during the course of H. lepidulum invasions.     

Melding ecology,classical weed biocontrol,and plant microbial ecology can inform improved practices in controlling invasive plant species

Early research leading to the successful biological control of invasive species such as Opuntia spp., and Hypericum perforatum set examples and provided data useful for research programs that would follow. However, this early work failed to become established as a source of applicable principles for later workers in weed biocontrol. Recently, retrospective and parallel studies have been suggested as a means to reengage with earlier work to derive useful ideas and data to enhance future programs in weed biocontrol. Parallel studies by workers in plant community ecology on the nature of feedback elicited by plant species in their invaded and native range have shown the importance of soil microbial communities in effecting feedback. Retrospective reexamination of previous studies would likely provide clues to other insect–plant pathogen interactions in addition to those described by the author and others. The effects of invasive species in profoundly altering soil microbial communities point to the need for further studies on key microbial species contributing to or driving the impact of biocontrol. These collective data suggest that the desired goal of selecting for and utilizing stronger biocontrol agents to reduce nontarget effects and to increase the impact of biological control programs would be best served by prerelease studies that assess the propensity of a candidate agent for direct or indirect interaction with other agents. This could be assessed through the use of survival analysis. Overall, parallel empirical and retrospective studies should be a necessary part of how biological control is practiced.     

Melding ecology, classical weed biocontrol, and plant microbial ecology can inform improved practices in controlling invasive plant species

Early research leading to the successful biological control of invasive species such as Opuntia spp., and Hypericum perforatum set examples and provided data useful for research programs that would follow. However, this early work failed to become established as a source of applicable principles for later workers in weed biocontrol. Recently, retrospective and parallel studies have been suggested as a means to reengage with earlier work to derive useful ideas and data to enhance future programs in weed biocontrol. Parallel studies by workers in plant community ecology on the nature of feedback elicited by plant species in their invaded and native range have shown the importance of soil microbial communities in effecting feedback. Retrospective reexamination of previous studies would likely provide clues to other insect–plant pathogen interactions in addition to those described by the author and others. The effects of invasive species in profoundly altering soil microbial communities point to the need for further studies on key microbial species contributing to or driving the impact of biocontrol. These collective data suggest that the desired goal of selecting for and utilizing stronger biocontrol agents to reduce nontarget effects and to increase the impact of biological control programs would be best served by prerelease studies that assess the propensity of a candidate agent for direct or indirect interaction with other agents. This could be assessed through the use of survival analysis. Overall, parallel empirical and retrospective studies should be a necessary part of how biological control is practiced.     

A European pest to control a new zealand weed: investigating the safety of heather beetle, Lochmaea suturalis (Coleoptera: Chrysomelidae) for biological control of heather, Calluna vulgaris

Heather, Calluna vulgaris (L.) Hull, is a serious invasive weed in the central North Island of New Zealand, especially in Tongariro National Park, a World Heritage Area. Heather beetle, Lochmaea suturalis (Thomson), is a foliage-feeding pest of Calluna in Europe, that was selected as the most promising biological control agent for introduction into New Zealand, because it causes high levels of damage to Calluna in Europe. Host-range tests indicated that L. suturalis poses a negligible threat to native New Zealand plants. Cultivars of Calluna grown as ornamentals are suitable food plants, but are unlikely to be severely affected because L. suturalis requires a damp understorey of moss or litter for successful oviposition and pupation, which is rarely present in gardens. However, mosses and litter occurring under Calluna stands in Tongariro National Park are suitable substrates for eggs and pupae. Lochmaea suturalis released in New Zealand has been freed of parasitoids and a microsporidian disease that attack the beetles in Europe.     

Plant/pathogen interactions observed during host range testing of the rust fungus Uromyces pencanus,a classical biological control agent for Chilean needle grass (Nassella neesiana) in Australia and New Zealand

Nassella neesiana (Chilean needle grass) is a South American grass species that is a serious weed in Australia and New Zealand. The rust fungus Uromyces pencanus is a promising biocontrol agent that could be used to control the weed in both countries. Extensive host range testing has been conducted to explore the specificity of the rust. In this paper we discuss the different degrees of invasion by the rust of the tissues of target and non-target species; the plant defences elicited by such invasion at the cellular level; and their relevance to the biological control of Chilean needle grass.     

Biological control: Premises,ecological input and Mimosa pigra in the wetlands of Australia's Top End

The adverse impacts of weeds on natural ecosystems, together with the inadequate outcomes from treating weeds as a symptom, have escalated interest in finding efficacious control methods. With the aim of protecting wetlands from invasive weeds, this contribution uses the woody shrub Mimosa pigra L. (mimosa) as a case history to examine the methodology of classical biological control and the reasons for the widely accepted 75% failure rate. Overall objectives are for all biocontrol agents to have the opportunity to fully express their potential and to insure that limited resources are spent wisely on attainable weed control. The three main conclusions were that (1) the premises on which biocontrol is based has restricted advancement of this method; (2) monitoring is the logical first step to improving the selection of agents and release sites; and (3) it may be more cost-effective to introduce fewer agents that have undergone agent/plant and host/home range pre- and post-release ecological studies. Weed control may remain elusive unless advantage is taken of every beneficial result. Innovative assistance to agents and the integration of different control methods may preserve a role for weed biological control and has the potential to be of great importance for future weed management. It is proposed that the climate in the Top End of the Northern Territory and the lack of competition on the floodplains has contributed to mimosa's invasiveness. Climate may also underly the difficulties faced by agents. Agents appear unable to impart effective control in the dry season because of low numbers which relates to mimosa's poor growth; nor in the wet season, when the impact from high numbers of agents is outstripped by mimosa's growth.     

What makes a successful biocontrol agent? A meta-analysis of biological control agent performance

We qualitatively reviewed the biocontrol literature in two major journals, Biological Control and Environmental Entomology, over the past 10 years by scoring 878 studies into 11 biocontrol-oriented questions. Quantitative meta-analyses were then used on data from 145 studies to examine the effects of different types of biocontrol agents (parasitoids, predators, and pathogens) on several attributes of weed and pest populations. Results for our qualitative review showed that most biocontrol studies were focused on lepidopteran pests, and that parasitoids were the most common biocontrol agents used. Our quantitative review showed that, for weeds, biocontrol agents significantly reduced weed biomass (−82.0%), flower (−98.9%), and seed production (−89.4%). For pests, our quantitative review showed that biocontrol agents significantly reduced pest abundance by 130% compared to control groups, increased parasitism (+139.0%) and increased overall pest mortality (+159.0%) compared to targets not exposed to biocontrol agents. Effects on pest mortality tended to be stronger for parasitoids than predators, although reductions caused in pest abundance were much stronger when predators were used as biocontrol agents. Addition of two or more biocontrol agents increased mortality by 12.97% and decreased pest abundance by 27.17% compared to single releases. Separate sets of meta-analyses demonstrated that the negative impacts of biocontrol on non-target species were much smaller than those for target species, although adverse effects of biocontrol on non-target organisms are based on small sample sizes and should be interpreted with caution. Our results also showed that biocontrol efficacy tended to be higher when agents were generalists than when they were specialists. Large fail–safe numbers found for most of the estimated effects indicate the robustness of the results found for the efficacy of biological control programs.