Potentially hazardous agents in land-applied sewage sludge: human health risk assessment

2. General Methodology

This section provides an overview of the methodology used within this generalised Quantitative Risk Assessment (QRA). Specific details and assumptions for individual potentially hazardous agents are described further in Section 3 (Results).

The aim of this work was to undertake a quantitative risk assessment that establishes the potential for harm to human health or well-being, resulting from the use of sewage sludge products, including the manner in which they were processed and used as shown in Figure 2‑1.

Table 2‑1 Summary of sewage sludge products and uses covered by this report

Input Materials

• Thickened sludge from primary and secondary water treatment

Treatment Method

• Conventionally (anaerobically) digested & dewatered cake
• Dewatered raw cake with limed pasteurisation (lime caked)
• Thermal hydrolysis pasteurisation (THP) digested and dewatered cake

End Use Parameters

• Agriculture (grazing land)
• Agriculture (land used to grow grain crops for animal consumption)
• Agriculture (land used to grow root crops for animal consumption)
• Agriculture (land used to grow leaf crops for animal consumption)

It must be emphasised that this assessment only considered sewage sludge that has been produced under (regulatory) control. Activities outside of this specification, including unauthorized contamination of feedstocks and illegal use of sewage sludges have not been considered. This assessment examines potential risks associated with a specific product, and therefore does not make predictions about system failure, bypass of processing systems, or illegal activities.

The work was undertaken using the 'classical model' for QRA. This approach has been adopted by several agencies including Department of the Environment, Food & Rural Affairs (DEFRA) and the Institute of Environment and Health (Defra 2002).

The standard QRA model involves four key stages, namely hazard identification, dose-response assessment, exposure assessment, and risk characterisation (Figure 2‑1). Briefly, the hazard identification comprises a literature-based review to identify which hazards, if any, are of most concern/most likely to be a risk; the dose-response assessment to characterise the magnitude of effect caused by specific doses of specific hazards, the exposure assessment to determine to what extent receptors are exposed to the hazards of concern, and finally, the risk characterisation to quantify the level of risk, i.e. the probability that a specific hazard will result in a specific adverse outcome. The risk characterisation may then be used to inform 'risk management', i.e. management of risk factors in order to reduce impacts of causative agents.

Figure 2‑1 The four stages of the 'Classical Model of Risk Assessment'

Risk Management

• Risk Characterisation
  • Hazard Identification
  • Dose-Response Assessment
  • Exposure Assessment

For this project, six categories of potentially hazardous agents listed in were considered (Table 2‑2). The risks posed by each of these categories under different treatment methods, and end uses (listed in Table 2‑1) were investigated.

Table 2‑2 Categories of potentially hazardous agents that might be associated with sewage sludge

Categories of Potentially Hazardous Agents included in this assessment

• Odour associated with spreading/applying sewage sludge products to agricultural land
• Heavy metals and Inorganics including metals and metalloids
• Organic contaminants including Polyaromatic Hydrocarbons (PAHs), Polychlorinated Biphenyls (PCBs), Dioxins and Furans, Flame retardants, Plasticisers, Synthetic phenolic compounds, Siloxanes, Benzothiazoles
• Pharmaceutical and Personal Care Products (PPCPs) including anti-inflammatories, anti-epileptics, anti-histamines, selective serotonin reuptake inhibitors (SSRIs), antacids, antibiotics
• Microplastics and Fibres defined as synthetic polymers measuring less than 5 mm in diameter (i.e. largest dimension)
• Human/animal pathogens including bacteria, antibiotic resistant bacteria (ARB), viruses, protozoa, prions

2.1. Hazard identification and screening

The approach adopted for this stage of the QRA was adapted from Pollard et al. (2008). It was considered important by the Steering Committee (SC) that the assessment should demonstrate that all potentially hazardous agents had been considered where practicable. While the focus was on hazards not included in the original SNIFFER report (SNIFFER, 2008); it was considered neither feasible nor necessary to carry out a full QRA on each potentially hazardous agent identified. Instead, a series of filters was applied to the long list of hazards in order to produce a short list for further quantification. This filtering process has been used effectively in previous projects including those assessing risks associated with soil amendments (WRAP 2016a, b & c; Hough et al., 2012).

Initially, for each of the categories listed in Table 2‑2, a comprehensive set of potentially hazardous agents were identified. As stipulated by the Steering Committee, information derived from peer-reviewed literature was used as primary source material. However, it was necessary to use some grey literature where relevant information was limited and the applicability, relevance and quality of this was judged by the project team before use. Potentially hazardous agents were included in the list if:

• They were not covered by the SNIFFER (2008) report, or significant new information published since 2008 was apparent,
• They had been identified or measured in sewage sludge, or
• Evidence was available that specific agents could enter the waste water treatment process assuming 'typical practice' was adhered to.

As peer-reviewed data from Scottish Water produced sewage sludge are limited, the identification of potentially hazardous agents included information from wider UK, EU, and North American sludge. Therefore, it should be emphasised that not all of the data reviewed was derived from sludge which had been produced to Scottish specification. Where the use of data from non-Scottish sewage sludge may have significant bearings on the findings of the QRA, this has been highlighted.

The agents to be considered were organized into the major groupings outlined in Table 2‑2. A series of successive, defined, filters were then applied to each grouping of agents to identify those considered most likely to present a risk to humans (Figure 2‑2). These filters are discussed in more detail in the following sub-sections.

2.1.1. Filter 1

Filter 1 asks whether the agent under consideration has a potentially serious effect on human health. This filter does not consider whether exposure is likely to occur, or if exposure would occur at a dose of concern, these factors are considered in the subsequent filters. For the majority of (chemical) hazards, a potentially serious effect was defined according to the definition used by the European Commission Enterprise and Industry Directorate (European Commission 2005):

"'Serious' means a hazard that could result in death, could be life-threatening, could result in significant disability or incapacity, could be a congenital anomaly/birth defect, or which could result in hospitalisation or permanent or prolonged signs in exposed humans or animals, or which could realistically cause these effects where the product enters the environment."

This definition was adopted for most hazards potentially associated with 'traditional' health outcomes (i.e. cancer, cardiovascular disease, neurological conditions, etc.) as it has been successfully applied in previous projects that have focussed soil amendments (WRAP 2016a, b & c; Hough et al., 2012). However, this definition was widened to incorporate health outcomes associated with mental health and wellbeing, with this wider definition being applied specifically to outcomes associated with malodour. Therefore, when considering malodour and similar outcomes, we also adopted the following definition of health and wellbeing (World Health Organisation 1948 in Grad 2002):

"Health and well-being is a state of complete physical, mental and social wellbeing and not merely the absence of disease or infirmity"

All other effects were defined as being either 'mild' (i.e. readily reversible causing little/no short-term deleterious effects) or 'moderate' (i.e. reversible, but likely to cause some minor short-term deleterious effects). Where agents under consideration were associated with little or no effect, or where knowledge was insufficient, this was noted. No attempt was made to examine positive or protective effects of agents under consideration as this was considered outside the scope of this study. Only those agents considered to have a potentially serious effect were passed through Filter 1.

2.1.2. Filter 2

Filter 2 considers if each agent is likely to be present in sewage sludge produced by a licensed operator at a level or concentration likely to cause harm to humans. This filter is important when considering the water and sludge treatment processes and storage/stockpiling of sludge products. For example, a compound found to be present at a quantity of concern in sewage sludge, does not necessarily pose a risk to public health until the sludge has been spread. Further, people are not likely to be heavily exposed to the spread sewage sludge, when sludge has been spread according to current agricultural practice and other operational constraints. (The potential for the compound to pose a risk once the sewage sludge has been spread is considered in Filter 3). This filter does however highlight agents that could become an issue if good agricultural practice is not adhered to.

For some agents, numerous estimates of harmful levels are available. Where this was the case, the level of each agent considered to cause harm was determined using the concept of 'Principle Source Documents' adopted by the Environment Agency (Defra, 2002). These are set out below in descending order of priority:

1 Authoritative bodies in the UK (DEFRA), Scottish Government, Scottish Environment Protection Agency (SEPA), Environment Agency (EA))

2 European Commission Committees

3 Other national organisations (e.g. United States Environment Protection Agency (USEPA))

4 Reports produced by authoritative organisations, but for different purposes

Measured concentrations in sewage sludge were then compared to the 'harmful level' sourced using the 'Principle Source Documents' concept. Where measured concentrations exceeded the harmful level, these agents passed on to Filter 3 after validation (discussed below). Where a measured concentration in sewage sludge was not available, it was sometimes possible to find measured values for untreated sludge. In some cases, it was necessary to extrapolate from measured concentrations in waste water influent combined with knowledge of the chemical behaviour of the compound of interest, in order to estimate concentrations in sludge. In situations where little data were available describing degradation during the sludge treatment process a 'worst case' scenario of no degradation was assumed.

Any agent that reached Filter 2 and is considered to be present in quantities of concern by virtue of documentary evidence, or potentially present in quantities of concern (where documentary evidence is lacking), was then validated. As stated previously, not all of the literature was related to sewage sludge that had been produced to Scottish\UK\EU specification. Consequently, the validation process involved further examination of the reliability and appropriateness of the source of information. This included comparability with Scottish water and sludge treatment, experimental design, and analytical procedures (including provision for Quality Assurance/ Quality Control). Where information was considered unreliable or inappropriate, these concerns were presented to the SC and wider stakeholders to reach a consensus whether it would be appropriate to consider this particular agent further.

2.1.3. Filter 3

Filter 3 assesses only those agents that have remained after the first two filters have been applied. This filter is concerned with exposure once the product has been spread in accordance with current agricultural practice. This process is further described in the following sections.

2.2. Exposure assessment

One aim of the exposure assessment was to quantify as much as possible potential exposure of individuals to the various hazardous agents. The level of quantification achieved by the exposure assessment was driven by data availability and accuracy and was different for the different potentially hazardous agents. The exposure modelling was particularly challenging due to the focus of this study on 'emerging' contaminants that had not been previously included in the SNIFFER (2008) report. As a result, different exposure modelling techniques were adopted for different hazardous agents/types of exposure. Two main modelling techniques were implemented:

• Bayesian Belief Network (BBN) models were adopted for those exposures were evidence was particularly uncertain or where information was limited/missing. BBNs have been used for a number of land-based risk assessments where information has been too incomplete to undertake a fully quantitative assessment (Troldborg et al., 2013; Aalders et al., 2011; Hough et al., 2010a)
• Multi-media fugacity modelling was implemented for the majority of organic contaminants and PPCPs. Fugacity modelling relies on partition coefficients and these have usually been derived for most commercially available chemicals/compounds. Fugacity modelling has been used successfully in a number of studies looking at sewage sludge application to land, more recently with respect to exposure to Bisphenol A (Zhang et al., 2015)

These methodologies are described in detail in Section 6 - Appendix A.

2.3. Dose-Response assessment

Dose-response data describe the magnitude of an outcome (response) in relation to the magnitude of a dose (exposure) of a specific agent. Dose-response data in the literature are in several different formats and it may therefore be necessary to convert data to a standard form. Most data are derived from laboratory experiments where discrete groups of animals, e.g. mice, are exposed to a specific dose. A number of groups of animals are used so that several exposures of different magnitude can be administered. Some data are presented as percentage of animals showing a specific response, while other studies present continuous biochemical data.

The majority of toxicologic dose-response data relate to exposures far greater than environmental levels in order to get an observable response in a limited number of experimental animals. Hence care must be taken in extrapolating such data to environmentally relevant concentrations. There are many methodological approaches to carry out such extrapolations, including various mathematical curve-fitting models. Since 1995, many agencies have started to use the benchmark dose method to estimate the no observed adverse effect level (NOAEL) and/or the expected dose (ED) (Crump, 1984). The benchmark dose is based on the lower 95 % confidence interval of the fitted dose-response model, resulting in a response in 10 % of the study animals. The rationale being that a 10 % response is at or just below the limit of sensitivity in most animal studies. The use of the lower confidence interval, rather than the model fit itself, accounts for experimental uncertainty. Overall, the benchmark dose approach improves certainty in estimates of NOAEL. However, choice of linear or curve-linear (etc.) models to extrapolate from high to low dose is still, in many instances, reliant on expert judgement and is associated with significant uncertainty.

For non-cancer end points, it is standard practice to assume that a threshold of effect exists, while no threshold is assumed with carcinogenic endpoints. Although carcinogenic, a threshold effect was also assumed to exist for dioxins and dioxin-like PCBs. This is in line with the Committees on Toxicity, Mutagenicity and Carcinogenicity of Chemicals in Food, Consumer Products and the Environment (COT COM 2001). The COT COM (2001) agreed there was sufficient information to assume a threshold existed for the effects of dioxins and hence a tolerable daily intake could be established. There were two critical components to this decision:

• There is considerable evidence that dioxins do not directly damage the genetic material.
• There is considerable understanding of the biological reactions by which dioxins cause harmful effects, and evidence that these reactions will not occur at sufficiently low levels of exposure.

For the majority of organic pollutants and PPCPs, estimates of exposure (ADD, mg kg^-1 d^-1) could be compared directly to 'safe' reference doses (RfD, mg kg^-1 d^-1) published in the literature. However, for a number of agents, RfD values had to be estimated from reported NOAELs and other points of departure as there were no published RfDs. This was done following the method of the United States Environment Protection Agency (Equation 2.1; USEPA, 1996):

The RfD is considered to be a daily dose to which the receptor can be exposed without experiencing any deleterious effects. The RfD is determined by applying Uncertainty Factors (UF) to the NOAEL or other point of departure (Barnes & Dourson, 1988; Clegg et al., 1986). In this study, a maximum of two uncertainty factors were applied to the lower 95 % confidence interval of the NOAEL (NOAEL₅). The first UF (UF_L) was used to account for uncertainties associated with extrapolating from the experimental population to the population at risk. This UF was applied where species differences existed, e.g. extrapolating from an experimental rat population to a human population. The second factor (UF_H) was used to account for variability within receptor populations, e.g. differences in the amount of exposure medium consumed, differences in the inherent susceptibility of different members of the population (Barnes and Dourson, 1988). Following standard procedures, each UF is usually assigned a value of 10 but can be much greater depending on the uncertainties inherent in the toxicological data or the suitability of the toxicological data for extrapolation to human receptors (Barnes and Dourson, 1988). The reference doses for the different potentially hazardous agents are given in Table 3‑5 & 3‑9 alongside their associated uncertainty factors.

It should be noted that within this study, dioxins and dioxin-like PCBs were assessed both on an individual basis, and collectively using Toxic Equivalency Factors (TEFs) and Toxic Equivalents (TEQs). Toxic Equivalency Factors (TEFs) are toxicity potency factors that are used by the World Health Organization (WHO) and regulators as a consistent method to evaluate the toxicities of highly variable mixtures of dioxin compounds. In previous risk assessments of agricultural soil amendments, this approach has been favoured by some members of the Steering Committees, including the Food Standards Agency (WRAP 2016a, b & c; Hough et al., 2012). While TEQs are the standard approach, it was considered appropriate for this study to also assess each congener separately because: (i) published data on the levels of all congeners in sewage sludge were not available; (ii) there are differences in the extent to which different congeners move through the environment.

The issue of mixtures and their actions has been studied using laboratory rodents. Some data suggest that even when each component of a cocktail is present at concentrations that, individually, result in no observable biological effect, the mixture can exert significant biological effects (Payne et al., 2001; Rajapakse et al., 2002; Tindall and Ashby, 2004). Despite advances, this evidence is in its infancy, and comprehensive data are not available for all of the combinations of chemicals possibly present in all agricultural amendments, including sewage sludge. This is especially the case with respect to the 'emerging' contaminants that are the focus of this study. However, some work involving sheep exposed to mixtures of pollutants, through grazing pastures fertilised with sewage sludge, has shown that exposure to low concentrations of multiple pollutants can disrupt foetal ovarian and testis development (Paul et al, 2005; Fowler et al., 2008), offspring behaviour (Erhard and Rhind, 2004) and adult bone structure (Lind et al., 2009). It would be problematic to incorporate the issue of mixtures into the risk assessment approach used here. With thousands of pollutants known to have the potential to disrupt biological systems, understanding of mixture effects and meaningful assessment of risk will require the integration of observations from a wide range of empirical approaches together with the use of powerful, predictive computer models (Suk et al., 2002).

2.4. Risk characterisation

For the majority of agents, 'risk' was defined as the modelled probability that after spreading sewage sludge on agricultural land, an individual human receptor would experience deleterious health effects from either direct ingestion/inhalation or ingestion of food products produced on that land. This approach of calculating risk on an individual basis is the most appropriate because associated legislation, e.g. food safety, is based on individual products, rather than on the market as a whole.

Risk was calculated as the ratio (or Hazard Quotient, HQ) of the exposure (Average Daily Dose, ADD, mg kg^-1 d^-1) to the appropriate reference dose (RfD, mg kg^-1 d^-1) derived in Section 2.3; Equation 2‑2. If the ADD exceeds the RfD, we might expect to see deleterious effects to occur during the lifetime of the receptor.

Due to the significant uncertainties associated with estimating risks, an HQ greater than 1.0 indicates an issue that may require further investigation – but does not automatically imply a 'real' risk. An HQ less than or equal to 1.0 may be regarded as 'safe' (or negligible risk). For ease of interpretation, risk in this study was expressed either as 'negligible' (HQ 1.0) or potentially requiring further investigation (HQ > 1.0).

For agents modelled using the BBN approach, once the network structure has been optimised, health risk may be considered as an expectation value measuring the probability of the extent to which a specific (health) outcome is likely (Hough et al., 2010a):

Where V(x) is a numerical expression of the specific (health) outcome, p(x) is the probability of the (health) outcome arising and the integral is performed over all possible realizations (denoted by a variable x).

2.5. Sensitivity analysis

A simple point sensitivity analysis was conducted to identify which input parameters the risk assessment is most sensitive to and therefore are most important to characterise accurately in order to reduce the output uncertainty. A point sensitivity analysis investigates how the model output changes relative to the change in each input parameter while keeping all the other inputs at a fixed level. The sensitivity can be expressed in different ways. Here, the sensitivity of the model output, O, to a parameter i taking the value x_i is expressed through a normalised sensitivity index, SI, calculated as (Spitz and Moreno, 1996):

where |dO|is the absolute change in the model output following a change in the input parameter value dx_i , and x_i is the initial parameter value (i.e. in the base case).