A seafood risk tool for assessing and mitigating chemical and pathogen hazards in the aquaculture supply chain

Intricate links between aquatic animals and their environment expose them to chemical and pathogenic hazards, which can disrupt seafood supply. Here we outline a risk schema for assessing potential impacts of chemical and microbial hazards on discrete subsectors of aquaculture—and control measures that may protect supply. As national governments develop strategies to achieve volumetric expansion in seafood production from aquaculture to meet increasing demand, we propose an urgent need for simultaneous focus on controlling those hazards that limit its production, harvesting, processing, trade and safe consumption. Policies aligning national and international water quality control measures for minimizing interaction with, and impact of, hazards on seafood supply will be critical as consumers increasingly rely on the aquaculture sector to supply safe, nutritious and healthy diets. Aquaculture sector expansion requires limiting the chemical and pathogen hazards that can disrupt seafood supply. A schema is presented here for mitigating these risks and informing policy.

Intricate links between aquatic animals and their environment expose them to chemical and pathogenic hazards, which can disrupt seafood supply. Here we outline a risk schema for assessing potential impacts of chemical and microbial hazards on discrete subsectors of aquaculture-and control measures that may protect supply. As national governments develop strategies to achieve volumetric expansion in seafood production from aquaculture to meet increasing demand, we propose an urgent need for simultaneous focus on controlling those hazards that limit its production, harvesting, processing, trade and safe consumption. Policies aligning national and international water quality control measures for minimizing interaction with, and impact of, hazards on seafood supply will be critical as consumers increasingly rely on the aquaculture sector to supply safe, nutritious and healthy diets.
hazard subcategories within CH, AH and HH. Further definition of hazards falling within these subcategories (for example, specific animal and human pathogen taxa, anthropogenic chemical species, natural biotoxins and allergens) then forms a customized hazard list relevant to specific aquaculture scenarios, taking account of farmed species, farm location and method, intended market and product end use. Hazard subcategories and empirical illustrations of interaction between specific hazards and aquatic animals used for seafood are presented in Table 1 and, for representative marine fish, fresh water fish, crustacean and mollusc seafood groups, in Supplementary Section 2.1.
Pathogen and chemical hazards interact differentially with discrete phases of the supply chain, through early life (for example, hatchery production of larvae), grow-out (where juvenile and adult animals are grown in farm settings), harvesting, processing (to products), trading (nationally or internationally) and, eventually, consumption (in different forms). While the impacts of pathogen and chemical hazards on production phases are usually economic (for example, slow growth or mortality of stock, poor animal welfare and costs associated with therapies), those affecting processing, trade and consumption phases may be economic (where they may limit processing efficiency and restrict trade or the capacity to place products on the market) or health related (where intake of hazards via seafood consumption has public health consequences) (Supplementary Section 2.2 and Supplementary Fig. 2.1). Regardless of where in the supply chain hazards interact and impact, collectively they translate to a loss of supply of (safe and sustainable) food-a crucial consideration to be factored into future production aspirations at national, regional and global levels. Estimating the impact of specific hazards at given phases of supply also facilitates focus on those phases where interventions for control may have the greatest impact. The SRT may therefore be applied in three control states: (1) when assessing the potential uncontrolled impact of hazards acting upon supply from a specific aquaculture scenario; (2) when assessing benefit of applying discrete phase-specific control measures for limiting impact of hazards acting at that phase of supply; (3) when assessing multi-phase (cumulative or stepwise) control measures in limiting impact of hazards acting upon supply from a specific aquaculture scenario. Where hazard impact can be mitigated by intervention (for example, biosecurity control plans, active monitoring, post-harvest processing and so on), either at single or multiple phases in supply, the SRT provides a basis to target measures most efficiently and to calculate ensuing benefits of intervention compared with the uncontrolled state. In situations where application of controls is unable to adequately limit the impact of specific hazards, the SRT may guide go/no-go decisions relating to the feasibility of a stated aquaculture scenario to fulfil its proposed consequences (that is, production of seafood for an intended market and use). Here, amendment of the scenario (for example, alternative farmed species, site, intended market and product use) may lead to improved outcomes where seafood can be safely produced and consumed (Fig. 1).
The SRT. In lieu of a single method to efficiently capture the combined impacts of diverse chemical and pathogen hazards on discrete phases of seafood supply, the SRT uses a two-step semi-quantitative risk assessment schema to calculate impact as a multiple of scores for severity of harm caused and the likelihood of harm occurring (Supplementary Section 2.2 and Supplementary Table 2.1). Application of the SRT requires initial definition of the aquaculture scenario under investigation, including data on specific taxonomy, geography, seasonality, production method, product type, proposed market and intended end use. Further, a supply phase-specific customized hazard list (within the hazard definitions of CH, AH and HH) should be tailored to the scenario and used as the basis for generation of impact scores via the SRT schema. Outputs from the SRT include cumulative impact scores for specific hazard categories acting through the whole supply chain and scores for specific hazards interacting at discrete phases of supply. The SRT can be applied to both 'uncontrolled' (no hazard mitigations applied) and 'controlled' (hazard mitigations applied) states to inform a biosecurity and seafood safety plan appropriate to the aquaculture scenario under investigation (Fig. 1).
Here we demonstrate application of the SRT to a hypothetical aquaculture scenario intending to produce farmed bivalve molluscs in coastal waters of a non-European Union (EU) marine state for intended live export (and raw consumption) within nations of the EU. The scenario was chosen to represent one in which multiple CH, AH and HH hazards are likely to interact with different phases in supply, and where recognized control measures are potentially available at state and sub-state levels to mitigate hazard impact. The filter-feeding behaviour of bivalve molluscs and propensity for some species groups (for example, oysters) to be consumed raw also represent a particularly good example of the intricate relationship between certain seafood types, hazards present in their growing environments and risk to human consumers of certain end products arising from the sector (Table 1 and Supplementary Section 2.3). Environmental pathogens (HH2), natural biological toxins (CH4), anthropogenic pathogens (HH2), heavy metals (CH1) and bacterial diseases (AH2) represented the top five cumulative uncontrolled risks over the whole supply chain (Fig. 2a). The top-ranking risks within each of these categories over the whole supply chain were Vibrio parahaemolyticus (in HH2); amnesic, paralytic and lipophilic biological toxins (in CH4); hepatitis A virus and Salmonella (in HH1); cadmium, mercury and lead (in CH1); and diseases caused by various Vibrio spp. (in AH2) ( Fig. 2a and Supplementary Section 2.3). When specific supply phases were considered, pronounced impacts of AH hazards were predicted for early-life and grow-out phases (for example, viral, bacterial and parasite-induced mortality of animals on farms), with further potential for impact during the international trading phase, where pathogens of concern (for example, Marteilia refringens) are listed in legislation ( Fig. 2b and Supplementary Section 2.3). Human health hazards had a less pronounced impact on production phases but presented a higher risk of impacting harvest and processing (for example, hepatitis A virus and norovirus), trading (for example, Salmonella spp. or high levels of indicator bacteria indicative of faecal contamination) and, particularly, consumption phases. In the latter, contamination by faecal-borne human pathogens such as hepatitis A virus, norovirus and Vibrio cholerae non-01/139 may elicit significant public health consequences via consumption, if not controlled ( Fig. 2b and Supplementary Section 2.3). Similarly, CH had less impact on early-life and grow-out phases but impacted harvest and processing (for example, where concentrations of natural biological toxins exceed safe limits), trading (for example, where presence of heavy metals exceeds safe limits) and consumption phases (for example, where contamination of bivalves by natural biotoxins directly impacts human health) ( Fig. 2b and Supplementary Section 2.3).
Application of the SRT to the uncontrolled state can directly support decisions to progress or amend the aquaculture scenario plan (Fig. 1). The uncontrolled SRT also provides a baseline to which a Risk Mitigation Matrix (RMM) can be applied-a bespoke inventory of measures aimed at reducing risk associated with specific hazards impacting discrete phases of supply. Figure 3 shows the application of the RMM to the bivalve mollusc scenario and compares SRT scores for the uncontrolled state in which no controls are applied with those where either standalone/non-accrued control measures are applied at discrete phases of supply (control 1) or where the benefit of controls applied at one phase are accrued in subsequent phases of supply (control 2) (details provided in Supplementary Section 2.3). For anthropogenically derived CH and HH hazards, benefits of controlling hazards through the supply chain are enhanced by siting of farms where comprehensive environmental characterization has already been performed 6,7 . Subsequently, interventions during harvest include suspension of harvest, transfer of live animals to cleaner sites ('relaying') or otherwise informing onward processing requirements. Processing interventions include purification through re-immersion in clean water (for example, depuration) or other mechanical interventions (for example, irradiation for denaturing potential human pathogens in final products) 8 . Further, product monitoring during the processing phase may either occur at the official services level and/or by the food business operator informed by the application of Hazard Analysis Critical Control Point (HACCP) plans (including batch release measures) 9 . Labelling and traceability, good hygiene  Progress 8 Decision support tool-application of the SRT and development of a phase-and hazard-specific biosecurity and seafood safety plan is utilized to support progression of the stated scenario or may be used to inform amendment of the plan (for example, alternative species, farm location, market, intended use and so on). Amend Stepwise progression requires a clear definition of the scenario to which the SrT is being applied (1) followed by the formation of a customized hazard list relating to the major CH, AH and HH hazard categories likely to interact with specific phases of supply (2 and 3). The SrT is initially applied to the uncontrolled state (4) where no mitigations are applied. By considering the role of phase-specific control options identified within the rMM (5), the SrT can be re-applied to this controlled state (6), repeating, if necessary, with different control combinations. The outturn is a biosecurity and seafood safety plan (7) that assists a decision to progress, amend or reject the aquaculture scenario in fulfilling its goal, as initially stated (8). CH1, heavy metals; CH2, persistent organic pollutants; CH3, radiological contaminants; CH4, natural biotoxins; CH5, veterinary, pharmaceutical and personal care chemicals; CH6, allergens; AH1, viral pathogens; AH2, bacterial pathogens; AH3, protistan pathogens; AH4, metazoan pathogens; AH5, syndromes; HH1, environmental pathogens; HH2, anthropogenically derived pathogens; HH3, zoonotic pathogens. See Table 1   Veterinary medicines and other chemicals widely used (including illegally) in aquaculture to treat disease, as anaesthetics, and to manipulate physiology and immunity of stock. residues can reside in edible components of seafood, with potential to impact human health 37 . Antibiotics use and misuse can drive emergence of antimicrobial resistant (AMr) microbes, some of which may impact health of seafood consumers 38 . Pharmaceutical and personal care chemicals enter waterways and accumulate in edible components of seafood 39 . Impacts are probably greatest where seafood arises from production in high-population-density urbanized waterways, including effects of human medicines and personal care chemicals on health of aquatic animals 40 . Complex mixture effects are likely, though understudied.
CH6: allergens Tropomyosin, troponin C, arginine kinase, β-parvalbumin, histamine and other natural allergens Seafood allergy is a hypersensitivity disorder caused by numerous natural and spoilage-related elements present in fish and shellfish. Prevalence is increasing due to increasing seafood consumption, though misdiagnosis is frequent 41 . Common allergens are parvalbumin, tropomyosin and other proteins/peptides in fish and shellfish muscles. Histamine fish poisoning is a common seafood-borne disease associated with consumption of spoiled oily fish (for example, tuna) where muscle histidine is converted to histamine by bacterial histidine decarboxylase. Cooking destroys the bacteria but not the histamine 42 . Allergens are natural components of fish and shellfish tissues; thus, impacts are not associated with production phases of seafood. Taxonomically diverse metazoan eukaryotic organisms infecting many wild and farmed seafood species. Crustacean parasites cause significant direct losses in grow-out and during grading/harvest phases for salmon 46 . Platyhelminthes impact grow-out and trading of salmonids and are listed by OIE owing to potential for impact on wild stocks 4 . Nematode, trematode and cestode infestations cause pathology in invertebrate and fish hosts. Pathology is usually limited but can cause marketing issues for products-some have zoonotic potential (covered under hazard category HH3).

Animal pathogens
AH5: syndromes red mark syndrome, proliferative gill inflammation, white faeces syndrome, epizootic shell disease and various pathobiome disorders Syndromes are groupings of clinical signs associated with a particular health condition but for which specific aetiology has not been elucidated.
Often associated with disorders in major organ systems, including skin, gills, carapace and gut. Emerging molecular diagnostic tools are augmenting pathology studies to identify cryptic pathogens or multi-agent dysbioses 47 . Development of syndromes may be driven by influence of wider stressors (including climate, feed quality, host genetics, exposure to chemicals and so on). Increased focus is required due to their impact on yield in numerous aquaculture sectors. Autochthonous constituents of aquatic environments, often favouring warm/brackish conditions. responsible for human illness associated with seafood contact and consumption, particularly of filter-feeding molluscs. Clinical manifestations range from mild-to-severe gastroenteritis to primary septicaemia and death (the latter from wounding following contact with contaminated shellfish) 48 . Vibrios are acknowledged as important sources of seafood-associated illness, but global surveillance is lacking. Climate change offers opportunities for further emergence and potential pandemic spread 49    (bacterial pathogens). b, relative impact of hazards belonging to hazard categories CH, AH and HH at the six phases in supply; animal health hazards impact predominantly during production phases (and during trade) while human health and chemical hazards impact more greatly during harvest and post-harvest phases. c, Hazard-specific relative impact following application of control measures as detailed in the rMM for bivalve molluscs (Fig. 3). Control 1 (non-accrued scores) and control 2 (accrued scores) are compared with the uncontrolled state in which no phase-specific controls are applied. See Table 1 Fig. 3 | RMM applied to bivalve mollusc aquaculture scenario where live animals are destined for export market to be consumed raw. Control measures for specific hazards can be applied to given supply phases. The rMM informs re-application of the SrT to the uncontrolled scenario (no mitigations applied) for potential de-risking of supply using standalone/non-accrued benefits of applying controls at specific supply phases supply (control 1) or to cumulative/accrued benefits of applying controls at subsequent supply phases (control 2). resultant scores are represented in red, yellow and grey columns (Fig. 2c). Data relating to calculation of the SrT for these control options are provided in Supplementary Section 2.3. a Site pre-selection (covering CH, AH and HH hazards) offers the best risk mitigation measure that may be accrued during all subsequent supply phases. b Actions include suspension of harvest, 'relaying' animals at clean sites or otherwise informing onward processing requirements. c Purification through re-immersion of molluscs in clean water (for example, depuration and relay) or other mechanical interventions where criteria for efficacy of intervention are measurable (for example, irradiation). d Product monitoring either by official services or food businesses informed by application of HACCP plans (including batch release measures). e Good hygiene practices and education of workers to avoid cold chain breach, contamination of seafood by staff and consumption by 'at risk' groups; labelling and traceability are critical. f Application of Progressive Management Pathway, supported by appropriate national biosecurity tools, on-farm biosecurity plans, application of BAP or similar, application of measures in OIE Code for listed pathogens and generic chapters (surveillance and biosecurity) for other pathogens. g Application of OIE standards for international trade as recognized by the WTO, including more stringent national/ regional controls where justified by risk assessment, and meeting other criteria (equivalence) set out in the WTO SPS agreement. NA, not applicable.

Human pathogens
(determined by government biosecurity policy/practice) and application of best aquaculture practices (BAP) approaches from organizations such as the Global Aquaculture Alliance 12 . During the trading phase, application of the Office International des Epizooties (OIE) Code is relevant for listed pathogens, with generic chapters (surveillance and biosecurity) also contributing to de-risking of disease outbreaks from non-listed taxa. Most producer and trading countries are OIE members, with standards for international trade recog nized by the World Trade Organization (WTO). More stringent national/regional controls can also be implemented if justified by risk assessment and meeting other criteria (equivalence) set out in the WTO SPS (sanitary and phytosanitary measures) agreement 13 . For the bivalve mollusc scenario, benefits of application of control measures are set out in Fig. 3 and summarized in Fig. 2c (detailed in Supplementary Sections 2.1 and 2.3). The most pronounced reductions in risk were observed where controls were applied in early phases, and accrued at subsequent phases, of supply. For some hazards (for example, CH6), the application of available controls did not materially reduce risk; for CH6 hazards, avoidance of a product by susceptible consumer groups was the most relevant measure to reduce risk (Supplementary Section 2.3). The SRT is widely applicable to other aquaculture scenarios, including for marine fish, freshwater fish and crustaceans, using the schema presented here-although in each scenario, the impacts of hazards associated with discrete CH, AH and HH hazards acting at specific phases in supply are expected to differ significantly (Supplementary Section 2.1).
Policy implications and outlook. Diverse hazards interacting with seafood supply undermine sustainability via lost yield (food and profit) relative to the human, organism and environmental capital inputs required to create it 2 . Aquatic animal health and seafood safety are public goods, given that they cannot be easily purchased in the marketplace and thus require government intervention to ensure they are enacted 9,14 . Nationally, state-level responsible authorities designated to oversee aquaculture production and trade must be supported by official control laboratories able to apply quality-assured surveillance, analytical and diagnostic tools with respect to animal health (for example, OIE, the Progressive Management Pathway for Aquaculture Biosecurity (PMP-AB) and National Biosecurity Plan), anthropogenic and natural contaminants, and pathogens threatening seafood safety (for example, Codex Alimentarius codes of practice and standards). Known hazards (where regulatory requirements exist) can also be controlled by industry (for example, by farm-level best management practice and application of HACCPs to production and processing), supported by formal responsible authority monitoring, and surveillance activities and audit functions. Individual and societal preferences for, say, cooked seafood may confer additional protection against the impact of microbial hazards present within some seafoods, though they may have less effect at mitigating the risk of chemical threats. Where seafood is exported, regulations spanning primary production and final product are frequently in force with audit by importing countries or by trading blocs (for example, the EU) helping to mitigate risks of identified hazards in final products for consumers within those markets. The desire to trade often becomes a primary motive for deployment of hazard controls in producer nations. However, understanding hazards at each stage of the supply chain in the country or region of production, which may vary geographically, is considered vital irrespective of whether the product is destined for export or domestic markets. For all seafood production, quality and safety standards should be designed to control risks extant within that region and intended use of the product, with export regulatory requirements applied in addition. Increased reliance on protein arising from aquaculture in global diets 3 coupled with significant potential for blue foods to support development of a 'low stressor' global food system 15 must now be placed in context with the impact of mass global human migrations to coastal zones 16 , substantial pressures on water supply and quality, and the widespread use of water systems to dispose of human, agricultural and industrial wastes containing diverse pollutants 17 . Special focus must be applied to low-and middle-income nations where >90% of current aquaculture production occurs, where the most future growth and altered blue food consumption is predicted 3 and where the majority of wastewater from land-based sources is currently discharged without treatment 17 . While predominant scientific, policy and public discourse has focused on the potential impact of aquaculture on aquatic systems-outlined and discussed in ref. 2 -much less consideration has been paid to the impact of land-based human activities on contamination of those aquatic habitats that will be increasingly relied upon to provide human dietary protein in the coming decades 18 . The SRT considers those hazards with potential for greatest impact on supply of seafood from different aquaculture sectors, and the transnational-, state-, farmand societal-level interventions that may be required to mitigate them. It also provides a flexible framework to which novel emerging chemical and pathogen hazards may be added, potentially including those hazards (exemplified by severe acute respiratory syndrome coronavirus 2) that although not directly impacting aquatic animal health or seafood safety may nevertheless significantly impact supply chains 19 . For enactment, national strategies for aquaculture growth must therefore include (or interact with) comprehensive policies aimed at protecting aquatic habitats from diverse pollution sources, not least to protect the biodiversity upon which future aquaculture and its diversification will inevitably rely 19 . Initiatives such as the Global Burden of Animal Diseases approach aimed at identifying baseline metrics for supply chain losses (to disease) and justifying resource allocation for interventions may provide a logical methodology for extension to justify investment in the control of wider chemical and microbial hazards in food systems 20 . Further, proportional investment in state infrastructures that minimize the release of hazards to aquatic systems and increase the capability to detect known and emerging hazards where they occur and to apply appropriate controls to ensure the blue food revolution is a safe one should be considered a multi-faceted public good, where benefits extend beyond food and wealth to protection of biodiversity and climate change mitigation relative to food systems.

Methods
The SRT scores were generated for farmed bivalve molluscs in coastal waters of a non-EU marine state for live export (and raw consumption) within nations of the EU through small expert group elicitation (subgroups of authors of this paper) for each hazard category or subcategory, according to the framework provided in Fig. 1 (and detailed in Supplementary Sections 2.2 and 2.3). Impact and likelihood scores (with supporting evidence) for discrete hazard categories acting at specific phases in the supply chain for bivalve molluscs were provided by each subgroup to a coordinator (R.H.). The coordinator (an expert in the scenario under consideration), working with representatives of each subgroup, then agreed a final score for each hazard (at each phase) on the basis of evidence presented, using a probabilistic approach. Subgroups were asked to assess three states: (1) where there is uncontrolled impact of hazards on supply; (2) where application of phase-specific control measures is used to limit impact on supply; (3) where application of multi-phase (cumulative or stepwise) control measures is used to limit impact on supply. The evidence used was a mixture of peer review, grey literature and expert opinion generated within subgroups and was represented as the RMM provided in Fig. 3. The semi-quantitative process broadly followed the expert knowledge elicitation method, a structured approach to collate opinions from expert groups in a transparent manner focusing primarily on probabilistic methods to elicit expert judgement on quantitative parameters whilst minimizing bias 21 . Other formal expert elicitation processes (for example, the IDEA protocol) 22 have previously been used to calculate impact of discrete aquatic animal diseases in aquaculture 23 and may also be suitable to SRT application.