Procedural Review of Health Reference Values Established by enHealth for PFAS

Prepared by

Prof (Adj) Andrew Bartholomaeus

School of PharmacyFaculty of HealthUniversity of Canberra

&

Therapeutic Research CentreSchool of MedicineUniversity of Queensland

30 August 2016

1 | P a g e

Contents1 Terms of Reference.......................................................................................................................3

2 Background – The enHealth Process..............................................................................................4

3 Comments on Administrative Aspects of the enHealth Process....................................................8

4 EFSA and US EPA Assessments......................................................................................................9

4.1 Preliminary Observations.......................................................................................................9

4.1.1 What are HRVs...............................................................................................................9

4.1.2 Balancing Risk................................................................................................................9

4.1.3 Status of Current Human Health Reference Values (HRVs)..........................................10

4.1.4 The Basic Process for Conducting HHRAs.....................................................................11

4.1.5 Basic Issues in Assessment of Epidemiology Studies....................................................13

5 Sources of variation between enHealth Workshop, EFSA and US EPA Risk Assessments............16

5.1 Toxicology and Selection of the PoD....................................................................................16

5.1.1 PFOS.............................................................................................................................17

5.1.2 PFOA............................................................................................................................20

5.2 Toxicokinetics......................................................................................................................22

5.3 Mechanisms of Action.........................................................................................................24

5.3.1 Pharmacokinetics.........................................................................................................24

5.3.2 Toxicity.........................................................................................................................24

5.3.3 Comment.....................................................................................................................25

5.4 Epidemiology.......................................................................................................................26

5.4.1 Exposure......................................................................................................................26

5.4.2 Carcinogenicity............................................................................................................26

5.4.3 Reproductive effects....................................................................................................27

5.4.4 Other Effects................................................................................................................29

6 Conclusions..................................................................................................................................30

6.1 Use of International Risk Assessments and derived HRVs...................................................31

6.2 Sources of Differences Between US EPA and EFSA Risk Assessments..................................31

6.3 Potential Public Health Consequences of the Choice of HRVs.............................................31

6.4 Balancing Risk Mitigation with Risk Generation...................................................................32

6.5 Overall Conclusion...............................................................................................................32

7 Recommendations.......................................................................................................................32

8 References...................................................................................................................................33

2 | P a g e

1 TERMS OF REFERENCEIn early August 2016 the Department of Health commissioned this review in fulfilment of the Government’s commitment to review the interim human health reference values (HRVs) for per- and poly-fluorinated alkyl substances (PFAS) in drinking water. Delivery of the review was therefore considered urgent and a period of one month was allocated.

The Terms of Reference endorsed by the Government for this independent review are as follows:

“The independent review will consider:

(1) Approaches and assumptions used by the European Food Safety Authority (EFSA), as outlined in the reports Perfluorooctane sulfonate (PFOS), perfluorooctanoic acid (PFOA) and their salts, Scientific Opinion of the Panel on Contaminants in the Food Chain (EFSA, 2008) and Perfluoroalkylated substances in food: occurrence and dietary exposure (EFSA, 2012).

(2) Approaches and assumptions used by the United States Environmental Protection Agency (US EPA), as outlined in the 2016 Health Effects Support Document for Perfluorooctane Sulfonate (PFOS) (US EPA, 2016b) and the 2016 Health Effects Support Document for Perfluorooctanoic Acid (PFOA) (US EPA, 2016a).

(3) The applicability and relevance of these approaches and assumptions in the Australian context, having regard to existing Australian regulatory science policy as described in such guidance materials as: a. Australian Pesticide and Veterinary Medicines Authority (APVMA) Data guidelines

(http://apvma.gov.au/registrations-and-permits/data-guidelines) and Application of science to regulatory risk assessment (http://apvma.gov.au/node/15486)

b. the enHealth Environmental Health Risk Assessment, Guidelines for Assessing Human Health Risks from Environmental Hazards (enHealth, 2012);

c. the Food Standards Australia New Zealand (FSANZ) Risk Analysis in Food Regulation publication: (http://www.foodstandards.gov.au/publications/riskanalysisfood regulation/Pages/de fault.aspx (FSANZ)

d. the National Industrial Chemicals Notification and Assessment Scheme (NICNAS) Handbook for notifiers: https://www.nicnas.gov.au/regulation-and-compljance/nicnas-handbook (NICNAS)

e. the National Health and Medical Research (NHMRC) Guidelines for Managing Risks in Recreational Water (NHMRC, 2008) and NHMRC Australian Drinking Water Guidelines (NHMRC, 2016).

Given the limited time available, and the considerable body of documentation, the primary focus of this review is to identify the principle sources of variation between the US EPA and EFSA risk assessments of PFAS and the resultant guidance values, to comment on the consistency of the approaches taken with Australian guidance on, and practice of, health risk assessment and to form a view on the suitability of the EFSA values selected by enHealth as an interim measure pending more extensive consideration by FSANZ. In the time available for this review it is not possible to definitively identify one or other of the approaches as “correct” and the other not. Rather, the potential sources of strength and weakness in each assessment are examined together with a consideration of the nature and significance of methodological deviations from general regulatory approaches.

3 | P a g e

2 BACKGROUND – THE ENHEALTH PROCESSThe Standing Committee on Environmental Health (enHealth) under the guidance of the Australian Health Protection Principle Committee (AHPPC) provides nationally agreed environmental health policy advice, based on the best available evidence and expertise, to the Australian Health Ministers Advisory Council (AHMAC) through the AHPPC. The committee consists of representatives from key Commonwealth departments, each of the States and Territories, and New Zealand. In addressing specific public health issues enHealth draws on specialist scientific and medical expertise through the establishment of working groups and or the programming of workshops where the issue can be discussed in detail, applying a multidisciplinary approach. Unlike Australian regulators of chemicals in food, FSANZ and APVMA, and international agencies such as the US EPA, US FDA or EFSA, enHealth is not supported by a specialist scientific secretariat (that is a risk assessment group) and therefore relies on members of its scientific workshops and the agencies of the enHealth membership to prepare background papers for consideration by the medical and scientific experts of its scientific workshops.

On 15 March 2016, the AHPPC endorsed the enHealth Guidance Statements on Perfluorinated Chemicals (PFCs), which include an undertaking by enHealth to convene an expert group, in early 2016, to provide advice to the AHPPC on the development of an Australian interim HRV for perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA) for consistent use in the undertaking of human health risk assessments and the management of contaminated sites across Australia. In developing the interim HRV, the workshop was to consider relevant international guidelines, as well as contemporary scientific and technical issues. This workshop was convened in April 2016 and provided recommendations for the establishment of HRVs to support jurisdictional responses to incidents of environmental contamination with PFAS and to set drinking water guideline values for these substances. The enHealth committee was aware that a number of international regulatory bodies and the OECD had previously considered the HRVs for PFAS based on access to a larger data base of published and unpublished studies than was readily available in Australia at short notice. EnHealth was also aware that there was an immediate need for HRVs to be identified for use by State and Territory environment agencies in the management of PFAS contamination of ground water and of food produced in contaminated areas. EnHealth therefore undertook to review overseas Human Health Risk Assessments (HHRA) and standards rather than a de novo assessment, and based on a consideration of these, determine temporary/interim Australian HRVs and drinking water guideline values for PFOS, PFOA and related substances. This view was reinforced by the knowledge that the US EPA values were then in draft, that it was understood that EFSA intended that their values would be under review, and that a more detailed review of HRVs for PFAS was due to commence in FSANZ.

The workshops comprised of recognised experts from a range of scientific disciplines of direct relevance to the objectives of the workshop and involved a consideration of international assessments with a principle focus on those of EFSA and the US EPA. Attendees included toxicologists, members of the enHealth committee, representatives of the Cooperative Research Centre for Contamination Assessment and Remediation of the Environment (CRC CARE), FSANZ, the Australian Government Department of Health and Australian Government Department of the Environment. Scientific and medical experts on the workshop were tasked with preparing and presenting papers on the toxicology and health effects of PFAS, and the differences in approach of the US EPA and EFSA for discussion, reflecting their respective expertise.

4 | P a g e

The international PFOS and PFOA HRV reviews considered by the workshop included;

(1) European Food Safety Authority (EFSA, 2008);(2) United States Environmental Protection Agency (USEPA);

a. Health Effects Support Document for PFOS 2014 (draft) (US EPA, 2014a);b. Health Effects Support Document for PFOA 2014 (draft) (US EPA, 2014b);c. Provisional Health Advisories for PFOA and PFOS 2009 (US EPA, 2009b);d. Soil screening levels for PFOA and PFOS 2009 (US EPA , 2009a);

(3) United States Agency for Toxic Substances and Disease Registry (ATSDR, 2015) (draft);(4) Danish Ministry of the Environment (Perfluoralkylated substances: PFOA, PFOS and

PFOSA: Evaluation of health hazards and proposal of a health based quality criterion for drinking water, soil and groundwater., 2015);

(5) German Ministry of Health Drinking Water Commission and Federal Environment Agency, 2006 (GDWC, 2006);

(6) Swedish Environmental Protection Agency, 2012 and 2014 (cited in Danish Ministry of the Environment document);

(7) United Kingdom Committee on Toxicity of Chemicals in Food (COT) (COT, 2006a; COT, 2006b; COT, 2009; COT, 2014);

(8) Minnesota Department of Health (MDH, 2009a; MDH, 2009b); and(9) CRC CARE 2016 (draft).

During consideration of the available reviews and established HRVs (Table 1), and the approach and assumptions utilised by the various regulatory agencies, the workshop engaged in extensive discussion of the relative merits and weaknesses of those assessments. Individual experts of the workshop identified potential strengths and deficiencies in both the assessment by EFSA and that of the US EPA. Although the workshop identified a number of issues impinging on the establishment of appropriate HRVs and discussed specific recommendations for these values there was not a final consensus on the treatment of uncertainties and the relative merits or otherwise of the approaches of EFSA and the US EPA.

The time and resources available to the workshop to explore these issues was necessarily limited, and enHealth were aware of the pending thorough review of PFAS HRVs by FSANZ. In this context the workshop noted that;

Because of the exceptionally long half-life (time required for blood levels to decrease by half once dosing has ceased) of PFAS in humans the systemic (i.e. internal) exposure to PFAS is determined by oral (or other routes of) exposure over long periods of time. As lowering of HRVs and drinking water guideline values cannot therefore affect internal exposures meaningfully over the short to medium term, and given the steps already taken to reduce exposure in affected communities, lowering the HRVs established by EFSA would have no short term impact on public health.

The establishment of interim HRVs substantially lower than those of EFSA had the potential to greatly constrain the FSANZ review.

The EFSA HRVs were therefore concluded to be adequately protective for short to medium term exposures as a temporary measure.

The FSANZ review would have greater scope in terms of time and resources than the enHealth workshop and could draw on the deliberations of the workshop to inform that review.

5 | P a g e

The enHealth committee therefore determined to utilise the EFSA HRVs as temporary (i.e. interim) values pending the finalisation of the FSANZ review.

The enHealth committee, meeting on 6 April 2016, considered the outcome of the technical workshop from the day before and agreed to have a workshop report prepared and, based on a consideration of this, to make recommendations to AHPPC. On the 26th of May 2016 enHealth met again by teleconference to consider a draft workshop report and consider their next steps. The record of outcomes states:

“Members noted the draft Workshop report prepared by [one of the invited experts] and that comments and suggested edits were in the process of being incorporated.

“Members noted that the workshop report may take longer than was desirable to finalise, and so agreed that enHealth proceed instead with a short statement on recommended interim human health reference values (the “enHealth statement”). A first draft of the statement was provided by SA Health. The draft set out, as a summary, many of the issues discussed at the workshop.”

EnHealth made the following decisions;

1) Adoption of Tolerable Daily Intake (TDI) values derived by the European Food Safety Authority (2008).

Members agreed that the EFSA approach is acceptable. Members further agreed that the US EPA health advisories on perfluorooctane sulfonate (PFOS) and perfluorooctanoic acid (PFOA), released in May 2016, should be considered and a response included in the statement.

2) Adoption of the same TDI value proposed for PFOS for perfluorohexane sulfonate (PFHxS).Given the comparative toxicity, members agreed that the PFOS TDI value be adopted also for PFHxS. The practical effect of this is that in applying the TDI for PFOS, any PFHxS present will also need to be taken into account.

3) Adoption of interim drinking water guideline values for PFOS and PFOA.Members agreed to adopt interim drinking water guideline values based on the EFSA TDIs and application of the methodology used in the Australian Drinking Water Guidelines (ADWG). These would be used for site-specific assessments. Members discussed the desirability of the NHMRC undertaking a formal process for establishing guideline values in the ADWG and agreed that this should be considered further following the completion of the work FSANZ is undertaking to establish health based guidance values for promulgation in the Australia New Zealand Food Standards Code.

4) Adoption of interim guideline values for surface water (recreational water and fish consumption).

Members agreed to adopt interim water quality guideline values for recreational water based on the approach recommended by the NHMRC, effectively 10 times the value of the drinking water guideline values.

Members agreed that the assessment of any health risks from contaminated seafood would be based on the levels detected in the seafood. As such, there was no requirement to set a level in water from which seafood is taken.

6 | P a g e

5) Adoption of seafood screening guideline values. Members agreed that the interim TDIs would form the basis of site –specific risk assessments. As such, general seafood screening guideline values were not required.

6) Updating the enHealth Guidance Statements and Factsheet.Members agreed that the Guidance Statements and Factsheet would be updated to include: (i) a reference to the enHealth statement on interim TDIs and water quality guideline values; and (ii) information about PFHxS.

Members also agreed to develop a “Questions and Answers” guide to facilitate consistent responses by jurisdictions to media enquiries about the interim reference values and the inclusion of PFHxS.

7) Changing references from “perfluorinated chemicals (PFCs)” to “per- and poly-fluoroalkyl substances (PFAS)”.

Members agreed to adopt the change in nomenclature for this group of chemicals.

8) Seek AHPPC endorsement of the enHealth statement on interim TDIs and updated Guidance Statements and Factsheet.

Members discussed the next steps and agreed that the enHealth statement and updated Guidance Statements and Factsheet should be provided to the AHPPC for consideration and endorsement.

A finalised record of the outcomes of the workshop was not available at the time of this review, however an uncirculated draft outcomes document was provided. The draft record of outcomes indicates that the workshop gave detailed consideration to the principal studies, points of departure, toxicological endpoints and uncertainty factors used to derive the HRVs of the different agencies considered. As many agencies had based their own reviews on that of EFSA, ATSDR or US EPA the workshop gave most attention to these. The workshop also carefully considered the different modelling approaches used by the US EPA and EFSA, as discussed later in this report.

As the EFSA and US EPA HRVs represent the upper and lower bounds of the HRVs considered by enHealth the focus of this review (and the terms of reference) is a comparison of those assessments and the enHealth deliberations on them. The US EPA has subsequently finalised its assessment of PFAS and the resultant HRVs. The terms of reference for this review directs attention to the finalised US EPA reviews of 2016. Consequently, this review considers the finalised US EPA review in place of the Draft available to enHealth at the time it completed its considerations.

Subsequent to the completion of the enHealth consideration, FSANZ has commenced a separate review of PFAS HRVs and will consider whether Maximum Levels (i.e. permissible levels) or some other guidance, should be set for PFAS in food. Additional information available subsequent to the enHealth assessment, the outcomes of the enHealth workshop, and the recommendations of this review will be available for consideration in the FSANZ review. FSANZ will also have the discretion to utilise experts in specific aspects of the PFAS as required.

7 | P a g e

Table 1 Health Reference Values for PFOS and PFOA from International Regulatory AgenciesPFOS PFOA

PoDmg/kg/d

UF TDI/RfDng/kg/d

PoDmg/kg/d

UF TDI/RfDng/kg/d

EFSA 2008 0.03 200 150 0.3 200 1500USEPA 2016 0.00051# 30 20 0.0045# 300 20ATSDR 2015 0.00252# 90 30 0.00154# 90 20Danish EPA 2015 0.033 1230 30 0.003# 30 100USEPA DWG 2009 0.03 390 80 0.46 2430 190Minnesota 2009 0.0025# 30 80 0.0023# 30 77Germany 2006 0.025 300 100 0.1 1000 100# Based on the human equivalent dose derived from theoretical pharmacokinetic modelling and incorporating a variety of assumptions to compensate for data deficiencies.

3 COMMENTS ON ADMINISTRATIVE ASPECTS OF THE ENHEALTH PROCESS

EnHealth provides a highly valuable and appropriate consultative jurisdictional forum to support co-ordinated approaches to environmental health issues across the Commonwealth and New Zealand (NZ). The use by enHealth of expert working groups to provide specialist medical and scientific expertise to support its work is consistent with that of most major international regulatory agencies, including EFSA and the US EPA, with the Australian Therapeutic Goods Administration (TGA), and to a degree, that of FSANZ and the APVMA. Where the enHealth process deviates is that it does not include a dedicated, specialist scientific secretariat (a risk assessment group) to support its work. Most regulators utilise a scientific secretariat to prepare initial risk assessments that address the available data, draw conclusions, make recommendations and pose specific scientific questions for the expert committee to address. The expert committees then address the specific recommendations and questions, critique the overall assessment and provide recommendations for further work. This process ensures that expert committees have all the necessary information at their disposal, and can then focus their individual attention and their available time on the specific aspects of the assessment that their specialist expertise can most add value to. The advantage of this approach is that the decision making process is considerably more transparent, the available expert resources are utilised efficiently, the scientific line of reasoning is documented in detail, approaches to HHRA are consistent, and where appropriate the engagement of external stakeholders is facilitated through public consultation processes such as those of EFSA, US EPA and FSANZ.

The absence of a finalised, ratified outcome report from the expert workshop at the time the current review of the process commenced is notable. The draft report provided is also notably brief given the complexity of the issues and the detail in which they were addressed by the workshop participants, and clearly reflects the absence of the support of a suitably capable expert scientific secretariat. The approach taken by enHealth in this specific case however, recognised the clear need to provide urgent guidance to the various State and Territory environment agencies to support their ongoing remediation and mitigation efforts. The draft report indicates that the workshop gave careful consideration to the sources of the variations between EFSA and US EPA reference values and considered the strengths and weaknesses of the approach taken by each agency. Additionally, the principle and central outcome of immediate importance from the workshop, that the EFSA values were appropriate as a temporary measure, was agreed by the participants of the workshop on the 5 April. A clear basis for the decision of enHealth to utilise the EFSA HRVs was provided by

8 | P a g e

the experts and enHealth committee members constituting the workshop of 5th April 2016. Thus, although there are aspects of the process that could be improved these do not reflect on the suitability or otherwise of the EFSA HRVs for PFAS as a temporary/interim measure pending formal review by FSANZ.

4 EFSA AND US EPA ASSESSMENTS

4.1 PRELIMINARY OBSERVATIONS In order to understand the different approaches applied by the US EPA and EFSA to their risk assessment of PFAS, and more specifically the HRVs they established, some general principles and background information first need to be discussed.

4.1.1 What are HRVsHealth Reference Values, or HRVs, come in many forms and are intended to cover a range of exposure scenarios. In the context of the considerations by enHealth of PFAS, the most common HRVs are Tolerable Daily Intakes (TDIs), Acceptable Daily Intakes (ADIs) and Acute Reference Doses (ARfDs). ADIs and TDIs are established to represent the maximum intake of a substance, whether naturally occurring or synthetic, that can be ingested by the population every day of their entire lifetime without appreciable risk. An ADI is used for substances intentionally added to food and a TDI is used for contaminants that may be naturally present in the agricultural environment or water source, or are anthropogenic contaminants. Because the ADI or TDI are maximum average daily intakes an ARfD may be established for acutely toxic substances to represent the maximum amount of a substance that can be safely ingested in a single day or a single meal/drink.

These values are not, and are not intended to be, bright lines between safety and risk, but rather represent the limit of confidence in the safe intake level. The greater the long term exposure of an individual exceeds the TDI, the more likely that some risk will be associated with that exposure. Because TDIs specify safe daily intakes for a lifetime of exposure, even quite substantial exceedance of the value for short periods is not generally associated with increased risk. If realistically achievable short term exposures above the TDI are considered likely to present a risk, an ARfD is established to place an upper boundary on the safe daily intake.

In circumstances where data are incomplete or a significant degree of uncertainty applies to the derivation of a TDI, a provisional TDI or PTDI may be established to reflect this.

4.1.2 Balancing RiskHuman Health Risk Assessment (HHRA) and the establishment of HRVs involves elements of both science and policy. As a matter of policy the process assumes that any uncertainties between the available data and their relevance to human health result in the general human population being more sensitive than the species or population from which the data are derived unless there is strong evidence to the contrary. This is a precautionary approach, and there is no evidence that humans are routinely more sensitive than experimental animals, for example. Indeed, because of the extreme conditions of exposure and a range of physiological differences between experimental animals and the general human population, humans are frequently less sensitive to toxicological effects. As a consequence, the HRVs established through these processes are frequently likely to be lower than the true tolerable intake level. Nonetheless, for chemicals intentionally added to food either during production or processing, this cautious approach provides additional protection without generating appreciable collateral health risks. This is not always the case however.

9 | P a g e

There are circumstances where overly conservative or precautionary processes for establishing HRVs can generate health risks greater and more probable than those initially intended to be avoided. For human therapeutics an excessively cautious approach may preclude the availability of life saving medicines for example, and consequently the safety assessment adopts a risk benefit or Margin of Exposure approach. For natural contaminants in food, such as some heavy metals, an excessively cautious approach may impact food security and nutritional adequacy, and care is therefore taken to balance those risks within the HHRA through progressive refinement of underlying assumptions.

For anthropogenic contaminants such as PFAS which cannot readily be removed from the environment the establishment of values that are, with respect to the overall weight of evidence, disproportionately low, has the potential to result in a range of adverse health outcomes which may be greater than the toxicological risks intended to be avoided. Such outcomes may include prolonged unwarranted stress in exposed populations, the recommendation, or seeking out, of unnecessary medical interventions with their attendant risks, or interventions in pregnancy and avoidance of breast feeding to the detriment of the foetus and neonate. Other, economic impacts although likely, are beyond the scope of a HHRA and of this review. Simplistic selection of the lowest international HRV is therefore not necessarily optimal for the overall protection of public health. Determination of suitable HRVs for PFAS requires a careful consideration of the strengths and weaknesses of the approaches taken by international agencies that have had access to the underlying data and a considered selection of the most appropriate approach/values within the context of the exposure patterns in Australia. A suitably precautionary approach to public health requires a balancing of risk prevention against the potential for risk generation.

4.1.3 Status of Current Human Health Reference Values (HRVs)For any substance, Human Health Risk Assessments (HHRAs) are an ongoing, iterative process. As new data become available they are incorporated into the risk assessment and may over time alter the established HRVs. These values are as likely to increase as to decrease depending on the nature of the data generated and their impact on the magnitude and direction of uncertainties. For PFAS the EFSA and US EPA reviews are now finalised but EFSA has indicated an intention to review their values in the near future and FSANZ is currently in the early stages of a review of the HRVs and the need for permitted levels appropriate for food and bottled water. FSANZ will have the opportunity to consider the recommendations of this review, the deliberations of the expert workshop convened by enHealth, together with any new data that become available, and any revisions of HRVs and standards by international regulators.

The importation and use of PFAS has been progressively reduced in Australia over the past decade and exposure of the general community will also progressively decrease. For communities around point sources of contamination, such as defence bases where PFAS containing fire-fighting foam has been used, an expanding plume of contaminated ground water may result in a transient increase in exposure where ground water is used domestically or for food production. As such plumes move progressively further from the point source, concentrations in ground water decreases through dilution and consequent exposure also decreases. Provision of uncontaminated water for domestic use in these areas will greatly decrease exposure.

Internationally where the use of PFASs has become restricted, a general trend towards progressively lower PFAS serum levels has been observed reflecting a progressive reduction in exposure and a slow but continued elimination of PFAS.

Because PFAS have an exceptionally long half-life in human blood, the primary determinant of ongoing exposure is the existing blood level and not the daily intake (other than the unlikely scenario

10 | P a g e

of intake of aberrantly high levels of PFAS through non environmental sources). Consequently, even quite substantial differences in permitted levels in drinking water for example, do not have a significant short term impact on systemic (i.e. internal) exposure.

4.1.4 The Basic Process for Conducting HHRAs The basic process for human health risk assessment is essentially the same across every major chemicals regulator in Australia and internationally. The process involves elements of policy, convention and science and is designed to be inherently risk averse or precautionary. All aspects of the process however, are subordinate to the best available science and new approaches to maximising the scientific basis for decision making, and new understandings of mechanisms affecting cross species extrapolation, in HHRAs are continuously being refined. Consequently, deviation from a predefined process in order to add new understandings or a more physiologically based method of reducing uncertainty is not inappropriate simply because it might deviate from that process, provided the validity of the process is sufficiently robust that it does not simply replace one source of uncertainty with another.

Toxicology data obtained in animals or in humans through direct experimental testing or epidemiology studies is analysed to identify the most sensitive toxicological effect in the most sensitive species. The dose immediately below the dose at which this effect is observed is then the Point of Departure (PoD), the dose point at which the process departs from analysis of the toxicology data to an estimation of safe exposure levels for the general human population. The study or studies producing the PoD is often referred to as the pivotal study or studies. The lowest dose producing an adverse effect is called the Lowest Observed Adverse Effect Level (LOAEL) and the dose immediately below that dose is called the No Observed Adverse Effect Level (NOAEL).

There are two key sources of variability in this portion of the process. The first of these is in determination of the adversity of an observed effect in an animal. And the second is identification of effects in animals that are species specific and therefore not, or unlikely to be, relevant to humans where the physiology is different and the mechanism of toxicity does not apply.

Various physiological adaptive mechanisms may be stimulated by exposures that are either not adverse, or are not adverse at low levels of stimulation. Similarly, although experimental animals share the majority of physiological processes with humans there are important biochemical and anatomical differences that can render an effect in an animal irrelevant for human risk assessment, or indicate that studies in a different species that is a better model for human responses, is more appropriate as the basis for identifying a PoD for that particular process. The conclusions from a consideration of these issues depend on a number of factors, both policy and science based, including the expertise and experience of the evaluator(s), the availability of data informing the consideration and the balancing of the weight of evidence. For natural and anthropogenic environmental contaminants integration of epidemiology and toxicology may provide indications of the likely relevance of observations in animals to the general population, or indicate that the animal studies have over predicted the likely human sensitivity to the contaminant.

Various agencies make more or less use of various modelling techniques such as Benchmark Dose Modelling (BMD) to assist in identification of the PoD, although this is generally confined to circumstances where a clear NOAEL is not obtained (uncommon for OECD test guideline compliant studies). Where a study does not provide a NOAEL, because effects are seen at every dose, and there are 2 or preferably 3 doses producing an effect and defining a well characterised dose response curve, the BMD approach may constitute a useful addition to the risk assessment process. In other circumstances the BMD approach uses predefined, somewhat arbitrary, dose response

11 | P a g e

curves, tends to give values not dis-similar to the NOAEL/LOAEL approach, and frequently adds little of value to the process, whilst reducing the transparency of the assessment for audiences other than experienced toxicologists. Differences in the value of the PoD, generally small, may arise where a BMD approach is used by one agency and the NOAEL by another.

Where a PoD is taken from a human epidemiology study the uncertainty inherent in extrapolating from experimental animals to humans is removed, but often greater uncertainty arises in the estimation of the exposures (doses) of the studied population, and the introduction of various confounding factors that can hinder or preclude meaningful interpretation. The estimates of exposures and the identification and management of confounding factors can vary substantially.

Once a PoD has been identified the risk assessment gives consideration to the sources and extent of uncertainties, inherent in any form of extrapolation, and attempts to quantify those uncertainties into Uncertainty Factors (also often called safety factors or adjustment factors). As a matter of precautionary policy, rather than science, the uncertainty is always assumed to work to make humans more sensitive than the most sensitive animal species and the most sensitive individuals to be substantially more sensitive than the average. The internationally accepted default values are 10 for differences between animals and humans and 10 for the difference between the average and most sensitive individuals. Additional uncertainty factors may be added to adjust for other sources of uncertainty such as specific study types that are not available. Similarly, the default uncertainty factors may be reduced where data are available that indicates that animals are more sensitive than humans for a specific toxicological effect or data that reduce the uncertainty of extrapolation such as comparative toxicokinetics between humans and animals. This component is a substantial potential source of variability in the determination of HRVs between regulatory agencies and is a significant contributor to the variation in HRVs for PFAS.

A further source of variance between PoDs identified by different agencies is the use of pharmacokinetic modelling to determine the Human Equivalent Dose (HED). This approach is common or routine for the US EPA but an exception for most other regulatory agencies. The HED is the dose that would need to be given to a human to achieve the same blood level as that in the experimental animals in the pivotal study used as the basis for the PoD. Where comprehensive data are available for both humans and the experimental animals, this approach can remove substantial uncertainty in the cross species extrapolation and has considerable potential for improved risk assessment outcomes. Unfortunately, data are generally incomplete and most modelling must incorporate a range of assumptions which may create uncertainties equal to or exceeding those initially intended to be reduced.

Having identified the PoD and the appropriate uncertainty factors, the HRV is determined by dividing the PoD by the uncertainty factors and expressing the HRV as a tolerable (or acceptable, permissible etc) daily intake value (TDI) in weight units per kg of body weight per day.

The final step is setting permissible levels of a substance in drinking water or food. This process involves an estimate of the daily intake of water and food for high consumers from the most sensitive portion of the population, and a calculation of the highest permissible level that would not result in an individual exceeding the TDI on an average daily basis over a period of a year. The principle source of variation in the determination of an acceptable residue level in water is the determination of the HRV. Estimates of intake for water can vary depending on the climate of the target population but generally has a small impact on the permitted residue levels.

12 | P a g e

Estimation of permitted levels of a contaminant in food can be substantially influenced by both the dietary patterns of the target population and by the HRV. For agricultural and veterinary chemicals, the permitted level (the Maximum Residue Level or MRL) is set at the lowest value consistent with good agricultural practice (GAP) and is essentially an “As Low As Reasonably Achievable Approach” (ALARA), where the chemical is used at the minimum level to achieve the required effect. If the residue level of the chemical when used in accordance with GAP is too high to be safe, then the use is not permitted.

Drinking water guidelines are generally viewed as limit values rather than target values and the principles of ALARA are equally applicable. Climate, water sources and other factors will affect what constitutes “Reasonably Achievable” at any given time or location.

As the principle source of variation in permitted levels in water derived by EFSA and the US EPA is the derivation of the HRVs, rather than the exposure assessments, the latter are not further addressed in this review.

4.1.5 Basic Issues in Assessment of Epidemiology StudiesEpidemiology studies provide a valuable source of information for use in HHRAs. Because data are derived from humans the uncertainty inherent in extrapolating results in experimental animals to humans is avoided. Conversely however, the experimental conditions of an epidemiological study cannot generally be controlled and manipulated to the extent possible in toxicology studies and a range of uncertainties related to the study design, estimation of exposures, confounding by co-incident exposures, life style and other factors frequently create uncertainty of similar magnitude to that in cross species extrapolation. Additionally, a large proportion of epidemiology studies involve a substantial number of comparisons which generates the probability that statistically significant associations between a presumed level of exposure and a disease outcome will arise purely by spontaneous random variation and not through a cause and effect mechanism. The generally accepted basis for interpreting the results of epidemiology studies in terms of causation are the criteria first put forward by Bradford-Hill (The Environment and Disease: Association or Causation, 1965).

Epidemiology studies fall in to two broad categories: observational and experimental, or intervention, studies. For the purposes of HHRA of environmental contaminants the study design most commonly available is the observational analytical study, because intervention studies where exposures are manipulated are generally not ethical. An observational study is one where the researcher has no control over circumstances within which events occur. These studies are further divided into descriptive and analytical study types:

a. In a descriptive study the researcher collects data to describe or characterise the disease, pathology, event or condition of interest in terms such as time, location, population, and progression;

b. In analytical epidemiology the researcher will seek to identify risk factors or causes of a particular pattern of outcomes, such as disease, by comparing different groups.

Although this review cannot provide a comprehensive discussion of the basis for interpretation and critique of epidemiology studies, a consideration of both the Bradford Hill criteria and the statistical insights of Ioannidis (2005), discussed further below, illustrates the key issues of relevance to PFAS risk assessment that underpin the EFSA and US EPA assessments.

13 | P a g e

4.1.5.1 Assessing causality – the Bradford-Hill criteriaThe most common errors occurring when non-epidemiologists such as the media interpret or comment on epidemiology studies is confusion of association with causation and not considering the available studies collectively. In considering the extent to which a study or data set supports or indicates a causal relationship between an exposure (or activity or other factor) and a specific outcome of interest, the risk assessment first needs to exclude non-causal explanations for the apparent association such as investigator or other sources of bias, potential sources of confounding and the potential for random or chance occurrence. If no obvious non-causal explanations are identified, consideration of standardised guidelines for causal inference guides the assessment. The foundations of this approach were first put forward by Sir Austin Bradford Hill in a speech to the Occupational Medicine Section of the Royal Society for Medicine in 1965, which are now known as the Bradford Hill criteria for causation. These criteria, with some examples taken from the report of the original speech, are:

1) Strength of Association

a) Where an association is exceptionally strong such as the many hundred-fold increase in the incidence of scrotal cancer amongst chimney sweeps observed in the 18th century by Percival Pott, even the weakest epidemiology study design may be sufficient to reliably assign causation.

b) Similarly, the 10-fold to 30-fold increased incidence of lung cancer amongst moderate to heavy smokers does not require a particularly sophisticated study design in order to attribute causation. Conversely, the two-fold increase in coronary thrombosis in smokers is not sufficiently strong, in isolation, to be causally attributed to smoking from the weaker study designs such as cross sectional studies.

c) Where the strength of association is low, as is frequently the case in the current day, better designed studies and or other types of data that allow consideration of the remaining criteria will be required to form a judgement about causation.

2) Consistency of Associationa) Has the association been observed by other investigators in other locations and at other

times and under other circumstances. In essence this criterion is about reproducibility of the association by other researchers.

b) The reproducibility of an outcome may however simply reflect a methodological or confounding factor common to multiple studies by multiple investigators and again therefore is not sufficient grounds, in isolation, on which to conclude causation.

3) Specificity of association

a) Specificity refers to a one to one relationship between a single causal agent and a single effect. The relationship between chicken pox and the virus which causes it is an example of specificity. Although a lack of specificity does not necessarily negate a conclusion of causation, where a single agent causes effects at multiple sites or tissue types for example, where it is present, it provides solid evidence of causation.

4) Temporality

a) In essence this criterion is concerned with the sequence of exposure and effect. For this criterion to be satisfied, exposure must occur prior to the observed effects and in the case of diseases with long evolution times, it must occur sufficiently in advance of the disease, and for a sufficient period, to plausibly have resulted in the disease development.

14 | P a g e

5) Dose response

a) The increase in lung cancer death rates with increased cigarette use, for example, provides strong evidence in support of a cause and effect relationship.

b) The absence of a dose response in the presence of good evidence of substantial variation of exposure, or dosage, raises significant doubt over a causal relationship.

c) However, if the internal dose or the critical dose or concentration at the target affected by the substance is not directly related to the external or administered dose, a dose response relationship may not be readily apparent if the administered dose is the only source of information.

6) Plausibilitya) Based on the information we have available, is the relationship between exposure and

effect biologically plausible? Although a lack of knowledge about a plausible mechanism for a relationship does not necessarily prevent a conclusion of causation, where that relationship has had very limited investigation for example, if a substantial body of knowledge about the toxicological mechanism(s) of a substance does exist, the lack of a plausible mechanistic relationship between the proposed cause and the effect will weigh against a conclusion of causation. Thus, information sources other than toxicological and epidemiological sources may need to be sourced.

7) Coherence

a) Is the proposed causal relationship consistent with the broader knowledge of the natural history and biology of the disease?

8) Experimental evidencea) This criterion primarily refers to the availability of experimental/intervention studies

where the causal relationship has been investigated, and supported, experimentally.

b) Intervention studies that demonstrate a reduction of a disease outcome from removal or reduction of a postulated causal factor provide strong evidence of a causal relationship, for example, the reduction in the incidence of cholera after Dr John Snow removed the pump handle from the Broad Street pump in the St James’s parish in 1854 was strong evidence the contaminated water was causing the disease.

9) Analogya) What do we know about similar/related substances and similar disease or pathology

outcomes that might add or subtract support to the proposed causal relationship?

4.1.5.2 Design Limitations in EpidemiologyIoannidis (2005) has discussed the statistical basis for the poor reproducibility of many types of epidemiology studies in the literature and proposed a number of criteria on which to judge the likely reliability of the results of such studies. These criteria, in conjunction with those of Bradford Hill, also provide a useful basis for considering the suitability of an epidemiology study for inclusion in a HHRA, and the weight that should be given to the results of these studies in forming risk assessment conclusions. Ioannidis observes that the design factors that tend to lead to unreliable results include:

1) small study size,

2) small effect sizes in relation to background variability,

3) large numbers of variables being tested that are unrelated to a specific prior hypothesis,

15 | P a g e

4) a high degree of flexibility in study designs, definitions, outcomes, and analytical modes,

5) conflicts of interest including; financial, ideological, philosophical, reputational.

There is a degree of overlap between the criteria of Bradford Hill and those of Ioannidis. Strength of association for example is related to study sample sizes and the size of the effect in comparison to background variation. The larger the sample and effect sizes the stronger the statistical significance and apparent association.

Epidemiological studies conducted in the absence of a specific (that is a precise/targeted) scientific hypothesis will generally employ a wide range of investigative variables and conduct multiple comparisons in search of “significance”. Studies of this type will frequently yield random statistically significant findings that are not reproducible in subsequent studies.

5 SOURCES OF VARIATION BETWEEN ENHEALTH WORKSHOP, EFSA AND US EPA RISK ASSESSMENTS

Primary sources of variation between the EFSA and US EPA assessments include elements of toxicology, particularly the selection of the PoD, approaches to address the differences in toxicokinetics, variations in selection and use of uncertainty factors, considerations of the mechanism of action, conclusions on the epidemiology studies and the use of modelling techniques.

5.1 TOXICOLOGY AND SELECTION OF THE PODThe risk assessments of EFSA and the US EPA each conform to the general approach described above. As the enHealth process endorsed the EFSA values for temporary use in Australia rather than a de novo assessment of the data, pending the FSANZ review, the following discussion focuses on the EFSA and US EPA reviews. The principle sources of variation arise from various aspects of the interpretation of the available evidence and in the approach to managing uncertainty. The nature of both the variations and the consistencies between these assessments are discussed below in terms of the key data sets supporting the determination of an HRV;

animal toxicology studies toxicokinetics mechanisms of toxicity human epidemiology.

A reappraisal of all the individual toxicology studies reviewed by EFSA and the US EPA is beyond the scope of and time available for, this review. However, a general consideration of the studies selected for identification of the PoD is important in understanding the source of the different HRVs derived by the two agencies. Additional studies that investigate the potential for, and mechanism of, higher (or lower) sensitivity in experimental animals than humans inform the selection of uncertainty factors and provide some indication of the likely magnitude and direction of conservatism within the derivation of an HRV.

Importantly the differences between the EFSA and US EPA assessments are not due to new data or information available to the US EPA that was not available to EFSA. Although there are a small number of new studies reviewed in the US EPA assessment they have not affected the choice of pivotal studies for the determination of the PoD.

16 | P a g e

Hepatic and developmental toxicity are the most sensitive toxic effects of PFAS in animals and these effects dominate the pivotal data sets used by both the US EPA and EFSA to derive their respective PoDs for their HRVs.

In selecting which study or studies should provide the PoD there are two ways to think of dose, administered dose and internal or systemic dose. For practical reasons related to accessibility of tissue for sampling, the key comparative measure of systemic exposure is the Area Under the (plasma level versus time) Curve or AUC. The AUC provides the time weighted average plasma level achieved from repeat dosing. For most compounds the AUC is directly proportional to the daily administered dose once plasma steady state has been reached (that is after a period of stabilisation related to the balance between dose administered, frequency of administration and the rate at which the body clears the administered substance). For compounds such as PFAS which have exceptionally long half-lives, and great variability between species, comparing doses on the basis of the amount administered each day may provide a misleading picture, because plasma steady state may take many months or years to reach. If data on actual plasma/serum levels is available, a comparison across species may be more robust if these are used rather than the daily administered dose. The liver is a possible exception due to its more direct exposure to administered dose rather than average systemic blood levels, as discussed further later in this review. A further step is to convert the serum levels in experimental animals to the (theoretically determined) equivalent oral dose that would need to be given to humans to reach that blood level – called the Human Equivalent Dose or HED.

As illustrated for PFOS in Table 2, and PFOA in Table 4, the US EPA has used serum level determinations from experimental animals, in Physiologically Based Pharmacokinetic Modelling (PBPK) as the basis for a calculation of the HED at the NOAEL and LOAEL. The validity of the calculated HED is entirely dependent on the validity of the PBPK model and the assumptions built into those models as discussed later in this review under toxicokinetics. These issues are explored further for each of PFOS and PFOA below. Identification of the pivotal study or studies to derive the PoD is a separate and distinct step to the subsequent conversion to a HED. More specifically, conversion to the HED is not a necessary step in the identification of the PoD.

5.1.1 PFOSIn the Australian context, PFOS is the predominant contaminant of concern. In considering sources of variation between the HHRAs of EFSA and the US EPA the first step is a comparison of the studies considered for the derivation of the PoD.

For the purpose of identifying the most sensitive toxicological effect in the most sensitive species the US EPA ranking of toxicity studies by modelling average serum levels based on measured levels at termination, at the NOAEL/LOAEL is the correct approach (for effects other than those in the liver) as it compares studies on the basis of systemic (that is internal dose) rather than the administered dose, and thereby removes a substantial proportion of the uncertainty in cross species comparison derived from very different pharmacokinetic behaviours in different species. With the exception of the liver and GIT tract, the concentration of PFAS in target tissues will be proportional to the serum levels. For reproduction and developmental studies in particular the evidence indicates that the ratio between maternal serum and both milk and cord blood are comparable across species inclusive of humans. Consequently, the “dose” experienced by the target tissue (or foetus/neonate) is proportional to the maternal serum and not necessarily to the maternal oral dose (except over very long periods). In this respect the US EPA approach to ranking studies to identify the PoD may be the more appropriate. As noted by the enHealth workshop however

17 | P a g e

“ … to predict the area under the serum concentration time curve (AUC). AUC was divided by study duration (presumably the same as dosing duration) to provide an average serum concentration over the study period. The model is based on Andersen et al. (2006), but further developed by US EPA. It is an empirical model – i.e. a model that includes some understanding of the processes involved but the estimated values for the parameters which drive the model are derived by Bayesian statistical techniques (using Markov Chain Monte Carlo) to optimise the fit of predicted serum concentrations (model outputs) to measured concentrations in a specific study or set of studies, rather than being determined from a mechanistic understanding of the kinetic processes. This model has 3 compartments amongst which the PFOS entering the body can move. Time and concentration dependency for transfer between compartments was required to replicate experimental serum levels. This model was developed using data from a monkey study rather than a study in rats. The US EPA did consider the model parameters relevant for rats and used them during the modelling.

The Workshop noted it is difficult for an independent third party to replicate the US EPA PBPK modelling for estimating the average serum concentration in an animal experiment. Other jurisdictions use serum concentrations actually measured during the experiment. “

Additionally the US EPA might be considered to have given insufficient weight to evidence supporting the importance of PPAR alpha in mediating developmental effects in rodents, as discussed later in this review.

For some studies the US EPA choice of NOAEL is questionable. As the enHealth workshop notes the NOAEL set for the Butenhoff et al study of 0.3 mg/kg bw/day is based on effects seen only on post-natal day 17 but not on days 13, 21 or 61 indicating the effect is unlikely to be treatment related and not suitable for setting a NOAEL. Although this study was used as the pivotal study in the initial US EPA draft assessment, in the final report it is only supportive of the Luebker et al study.

18 | P a g e

Table 2 EPA Human Equivalent PFOS Doses Derived from the Modelled Animal Average Serum Values

Table 3 Identification of the PoD for PFOS by EFSA and the USEPAStudy Species, route, Duration NOAEL, LOAEL

Mg/kg bw/dayAva Serum level NOAEL, LOAEL, µg/mL

EFSAPivotal studySeacat et al (2002) Monkey, gavage, 183 days 0.03, 0.15 7.8, 38US EPASeacat et al 2002 Monkey, gavage, 186 days 0.15, 0.75, 38, 157Seacat et al. (2003) Rat (m), 98 0.34, 1.33 16.5, 64.6Luebker et al (2005b) Rat reproduction, 84 0.1, 0.4 6.26, 25Luebker et al (2005a) Rat reproduction, 63 - , 0.4 - , 25Luebkaer et al 2005 a Rat Reproduction, 63 0.4, 0.8 19.9, 39.7Butenhoff et al (2009) Rat developmental, gavage,

410.3, 1.0 10.4, 34.6

Lau et al (2003) Rat developmental, gavage, 19

1.0, 2.0 17.5, 35.1

a the US EPA calculated the average serum level based on the final level measured at sacrifice and using PBPK modelling. EFSA used the final serum level. Values for the EFSA NOAEL for their choice of the PoD of 0.03 mg/kg bw/day in the Seacat et al 2002 monkey study is calculated from the US EPA modelled serum levels at 0.15 mg/kg bw/day (their choice of NOAEL for this study) and assuming dose proportionality between 0.03 and 0.15 mg/kg bw/day.

Importantly, EFSA have concluded that the NOAEL in a 183-day monkey study reviewed by both agencies and by the ATSDR, is lower than that identified by the US EPA and ATSDR for this study. If the serum level for the EFSA NOAEL of 0.03 mg/kg bw/day is calculated from the modelled US EPA values at 0.15 mg/kg bw/day in the same study, and dose proportionality is assumed across this range, then the EFSA NOAEL is arguably the appropriate PoD, because the average serum level is the lowest (or comparable to the lowest) and the monkey is likely to be a better model for human risk

19 | P a g e

assessment than the rat (due to more similar pharmacokinetics and biochemistry). The study cited by US EPA (Luebker, et al., 2005) which had a NOAEL of 6.26 µg/ml, was reviewed by EFSA but not considered in selection of a PoD. The EFSA discussion of this study is insufficiently detailed to identify their reasoning for this. Nevertheless, in terms of internal dose based on average serum levels of PFOS, there are no material differences in the PoD identified by EFSA compared to the US EPA. Differences between the two agencies assessments therefore hinge on the validity of the US EPA PBPK modelling of the human equivalent dose and of the Uncertainty Factors applied by each agency.

5.1.2 PFOAThe key studies considered by the US EPA and by EFSA for derivation of the PoD for PFOA are provided in tables 4 and 5.

Table 4. EPA Human Equivalent PFOA Doses Derived from the Modelled Animal Average Serum Values

The US EPA modelled the HED for each of the studies in Table 5 then applied the relevant uncertainty factors. Those studies without a NOAEL accrued an additional 10-fold safety factor. On the basis of these calculations the EPA determined that the pivotal studies were those of Dewitt, Lau & Butenhoff all of which yielded the same HRV. EFSA did not have the studies by Macon et al 2011 or DeWitt et al 2008 as those became available subsequent to completion of their assessment. Conversely the US EPA did not have access to the study by Sibinski (1987) reviewed by EFSA. These differences in data assessed do not affect the derivation of the PoD however, as those studies were not pivotal, in isolation, to the US EPA decision and in fact the US EPA did not use the Macon study in its PoD modelling. Rather than apply arbitrary uncertainty factors for studies without a NOAEL, EFSA utilised benchmark dose modelling to calculate a dose equivalent to a 10 % increase the incidence of

20 | P a g e

the effect driving the LOAEL, over control values, and then used a value equal to the lower bound of the 95% confidence interval. This value is then the BMDL10. For consistency a BMDL10 was

determined for each of the studies considered for derivation of the POD. Although this approach is not necessarily routine, it is entirely consistent with international practice, appropriate to the specific dataset available and suitably precautionary, giving values lower than the NOAELs where those were available.

Table 5 Summary of studies on PFOA considered by EFSA for the PoD compared to those of the EPAStudy Species, route, Duration NOAEL, LOAEL

mg/kg bw/dayAva Serum level NOAEL, LOAELµg/mL or BMDL10

EFSALau et al (2006) Mice Developmental, Gavage,

GD 1-17-, 1BMDL10 0.46

17.5

Perkins et al (2004) Rat, diet, 13 weeks, 0.06, 0.64BMDL10 0.44

26

Sibinsky (1987)b Rat carcinogenicity, diet, 104 weeks (lifetime)

1.3, 14.2BMDL10 0.74

Butenhoff et al (2004a) Rat reproduction, oral gavage, 2 generation

1, 10BMDL10 0.31

14

US EPADeWitt et al (DeWitt, Copeland, Strynar, & Luebke, 2008)

Mice, gavage, 15 days 1.88, 3.75 38.2, 61.9

Lau et al 2006 Mice developmental, oral gavage, GD 1-17

-, 1 -, 38

Perkins et al 2004 Rat, diet, 13 weeks, 0.64, 1.94 31.6, 1.94Wolf et al 2007 Mice developmental, oral

gavage, GD 1-17-, 3 77.9

Butenhoff et al 2004a Rat reproduction, oral gavage, 2 generation

-, 1 -, 45.9

Macon et al (2011) Mice developmental, oral gavage, GD 1-17

-,0.3 -,12.4

a the US EPA calculated the average serum level based on the final level measured at sacrifice and using PBPK modelling. Because all the studies considered for the PoD were in rodents EFSA compared the studies based on oral doses administered. EFSA however modelled the BMDL10 for each study and used these values in place of the NOAEL for comparisons. b this report was not available to the US EPA.

The lowest BMDL10 was 0.31 mg/kg bw/day from the Butenhoff study equivalent to an average serum level of 14 µg/mL based on the US EPA calculations for the same study. In terms of internal dose expressed as average serum levels of PFOA, EFSA therefore established a PoD comparable to that of the US EPA.

As for PFOS the differences between the US EPA and EFSA derived HRVs for PFOA is predominantly dependent on the use of PBPK modelling by the US EPA and the selection of uncertainty factors by each agency, but with the added complexity of the use of BMD modelling by EFSA to identify the dose for the PoD.

21 | P a g e

5.2 TOXICOKINETICSThe most substantial difference in the US EPA versus EFSA derivations of HRVs is in the use of Physiologically Based Pharmacokinetic Modelling (PBPK) by the US EPA to determine the HED for doses used in the animal studies. The ATSDR describe the key aspects of this process as follows:

“PBPK models for a particular substance require estimates of the chemical substance-specific physicochemical parameters, and species-specific physiological and biological parameters. The numerical estimates of these model parameters are incorporated within a set of differential and algebraic equations that describe the pharmacokinetic processes. Solving these differential and algebraic equations provides the predictions of tissue dose. Computers then provide process simulations based on these solutions.

The structure and mathematical expressions used in PBPK models significantly simplify the true complexities of biological systems. If the uptake and disposition of the chemical substance(s) are adequately described, however, this simplification is desirable because data are often unavailable for many biological processes. A simplified scheme reduces the magnitude of cumulative uncertainty. The adequacy of the model is, therefore, of great importance, and model validation is essential to the use of PBPK models in risk assessment.

Importantly, the accuracy and utility of these models are dependent on the validity of assumptions that are made and the quality of the data providing the values for key parameters used in the equations. In considering the utility of one of these models the ATSDR makes the observation that;

”The human model was calibrated to predict limitation half-times estimated for human populations (e.g. 2.3 or 3.8 years for PFOA, 5.4 years for PFOS). As a result, comparisons made between observed and predicted serum concentrations evaluate whether or not the populations actually exhibit the half-times to which the model was calibrated, and not the validity of the model to predict the internal distribution of PFOA or PFOS. It is not currently possible to assess with confidence whether the human model can accurately predict doses to liver or any other tissues”

The draft record of outcomes from the enHealth workshop also notes several difficulties in adopting reference values based on the US EPA PBPK modelling due to weaknesses or lack of transparency in those models.

“There are some issues with the level of understanding of the toxicokinetics of PFOS and PFOA within and between species that make the approaches adopted by USEPA and ATSDR somewhat problematic.

Volume of distribution – the volume of distribution is the apparent volume of the body within which a chemical distributes once it enters the body. Reaching steady state concentrations (those where intake and elimination are balanced) requires a large proportion of the storage locations to be filled.

For PFCs it is known they are highly bound to serum albumin, they are therefore confined primarily to extracellular fluid and have limited distribution into other tissues. However the distribution mechanisms are not fully understood, may be different in different organisms, and the volume of distribution is difficult to determine. Hence the reliability of values used for the volume of distribution in extrapolating no effect serum concentrations in animals to an equivalent human dose is uncertain.

22 | P a g e

The workshop attendees agreed the variability in the toxicokinetics for these chemicals mean that it would be appropriate to use toxicokinetic modelling in determining a tolerable daily intake if there was sufficient evidence that the parameter values used were robust and reliable. However the workshop attendees agreed there was not sufficient understanding of the parameterisation of the modelling approach taken by USEPA or ATSDR for it to be adopted prior to the assessment to be undertaken by FSANZ.

At plasma steady state, the ratio of tissue levels of PFAS to that in a given tissue for a specific species will remain relatively constant regardless of the species under consideration or the differences in other pharmacokinetic parameters between species. The comparative values for those relationships however will differ markedly between species if other pharmacokinetic parameters vary between the species being compared. Critically, where the HED approach is used to determine the PoD for liver effects seen in rats this approach is likely to greatly overestimate the dose to the liver in humans. Where dramatic differences in t½ (or more precisely Clearance and Volume of distribution) exist as for PFAS in humans compared to rats, the ratio between serum levels and liver levels, particularly during the absorption phase, will also be very different. As the liver receives 75% of its blood flow from the hepatic portal vein which drains the intestinal mesentery and only 25% from the systemic blood through the hepatic artery, the dominant determinant of the pattern and extent of liver exposure is the concentration in the hepatic venous blood which is proportional to the intestinal concentration during the absorption phase (i.e. the oral dose and the time course of absorption) rather than the systemic blood concentration. This is particularly true where rodents are dosed in the food and allowed to eat ad libitum. Because clearance in humans is markedly slower, resulting in a half-life of years compared to a few days in rats, the serum concentration in humans for a given oral dose will be higher than in rats but the liver concentration during the absorption phase is likely to be more similar to that in rats for the same dose. This is true even given the observation that post mortem liver to serum ratios in rats are similar to post mortem ratios in humans as the ratio post mortem reflects equilibration subsequent to the absorption phase. Rats are generally fasted prior to termination and equilibration between liver and serum may occur subsequent to the absorption phase.

Consequently, the use of HED for liver effects in experimental animals with markedly shorter t½ than humans, particularly rats, may greatly overestimate the potential liver exposure, and therefore may not necessarily decrease uncertainty in dose extrapolation, and is arguably inappropriate at least for liver effects.

The PBPK approach does however have potential value for modelling the HED for other toxicological effects. The ratio between maternal serum levels of PFAS and that of breast milk and placental cord blood are similar between humans and rats. A cross species dose comparison based on serum PFAS rather than oral dose administered to rats, does therefore provide a potentially more robust basis for identifying the PoD provided assumptions used in the PBPK modelling are robust and grounded in adequate data for each species and humans modelled.

In this last respect a recent review by Dong et al (2016), of the PBPK approach of the US EPA is pertinent and reflects the concerns of the enHealth workshop (attended by at least two of the authors of the paper). These authors, while acknowledging the potential benefit of PBPK modelling and HEDs, have concluded that the approach of the US EPA to toxicokinetic modelling may be compromised by “systematic fitting residual errors across species”, that the current approaches may lead to unnecessarily conservative reference doses and that the basis of modelling used by the US EPA may benefit from further refinement. A reconciliation of the various viewpoints on the validity

23 | P a g e

and utility of PBPK modelling in the specific instance of PFAS risk assessment is beyond the scope of this review but will need to be a key consideration in the FSANZ review of HRVs for PFAS.

5.3 MECHANISMS OF ACTIONWhere the data will allow, a careful consideration of the mechanisms of pharmacokinetic variability and toxicological action are essential components of any risk assessment. Interspecies variations in responses to a toxicant are determined by one or both of these. How and to what extent these mechanisms are incorporated into a risk assessment and the extent and direction to which they moderate the degree of uncertainty and the uncertainty factors applied, is a common source of variation between risk assessments of the same toxicant.

For PFAS there are two distinct sources of interspecies variation in sensitivity. Firstly, there is a clear and, very unusually, dramatic difference in pharmacokinetics between humans and all experimental animals so far examined, with non-human primates having the most similar kinetics. This variance acts to increase the blood levels of PFAS in humans for any given dose (in mg/kg bw/d) compared to experimental animals, because although humans and rats appear to absorb PFAS to a similar extent, humans excrete the compounds extremely slowly by comparison (Clearance in humans 0.03 mL/kg bw/day, female rats 666 mL/kg bw/day). Regardless of the risk assessment process applied, the derivation of HRVs must incorporate adjustments to accommodate this known difference.

5.3.1 PharmacokineticsIngested PFOA and PFOS are essentially fully absorbed over approximately an hour. The mechanism of absorption has not been fully elucidated but, as PFAS do not have the characteristics necessary for ready passive absorption, is likely to involve elements of active transport by organic anion transporters (OAT), and/or absorption in conjunction with lipids.

PFAS bind to serum albumin and various other plasma proteins including gamma-globulin, alpha-globulin, alpha-2-macroglobulin, transferrin, and beta-lipoproteins in both rat and human plasma but the affinity in rats for albumin binding is an order of magnitude greater than for humans. The albumin -PFOA dissociation constant is 0.4 mM for human serum albumin and 0.36 nM for rat serum albumin and involves 6–9 binding sites with noncovalent binding apparently at the same sites as fatty acids.

Absorbed PFAS distribute widely from plasma into the soft tissues. The highest concentrations are found in the liver. As for absorption from the GIT the mechanism of absorption into the liver is not fully understood but, again, is likely to involve active transport by OAT involved in movement of fatty acids and other organic anions. There is evidence that PFOA is a ligand for OAT in the luminal and basolateral membranes of renal tubular epithelial cells, which transport PFOA in the glomerular filtrate back into the tubular cells. The difference in plasma half-lives of the PFAS between species is therefore likely to be attributable to differences in the nature, density and activity of renal tubular OAT between species and differences in their affinity for PFAS.

5.3.2 Toxicity

The most sensitive toxic effects of PFAS in animals are hepatic and developmental effects which are both mediated to some extent by activation of the peroxisome proliferator-activated receptor-α (PPARα). PPARα is a member of the nuclear receptor superfamily that mediates a broad range of biological responses including lipid metabolism, energy homeostasis, and cell differentiation. Marked differences in the sensitivity of species to PPARα activating toxicants have been observed.

24 | P a g e

Humans, non-human primates and guinea pigs substantially resistant to PPARα mediated toxicity and rodents (rats and mice) the most sensitive.

Differences in response to PPARα agonists across species are likely to be due to a combination of; differences in the ability of PPARα to be activated by peroxisome proliferators, differences in the inducibility (increase in production) of PPARα after exposure to peroxisome proliferators and differences in pattern and levels of tissue-specific expression of PPARα. Notably the level of expression of PPARα in h uman liver is about 1–10% of that in the rat and mouse liver .

Although the critical biological target pathways leading to developmental effects in rodents have not been established, the observation in a number of studies that developmental toxicity of PFOA/PFOS and other ligands for PPARα, is significantly dependent on the expression of PPARα, indicates that rodents are more likely to over-predict human developmental risk for such substances than under-predict. A study of PFOA in pregnant mice for example, used wild-type, PPARα-null, and PPARα-humanized (expressing human PPARα) mice and demonstrated lower postnatal survival in wild-type mice, as predicted by rat studies, but no effect in null or humanized mice. Results indicate that PPARα mediates at least some of the developmental effects in mice, and that species differences exist between mice and humans.

The normal role of PPARα is the regulation of lipid homeostasis through the modulation of expression of genes involved in fatty acid uptake, activation, and oxidation. PPARα receptor activation by toxicants in rats and mice initiates a sequence of biochemical events that include marked hepatocellular hypertrophy due to an increase in number and size of peroxisomes, a large increase in peroxisomal fatty acid β-oxidation, CYP450 induction and alterations in lipid metabolism. Both PFOA and PFOS alter the expression in rats of genes associated with lipid homeostasis and down- regulate genes that control cholesterol biosynthesis. In comparison with naturally occurring long-chain fatty acids such as linoleic and α-linoleic acids, PFOA and PFOS are relatively weak ligands for PPARα.

5.3.3 CommentThe marked differences between the pharmacokinetics of PFAS in experimental animals and humans has been inadequately addressed in the EFSA assessment for effects other than liver toxicity (where a MOE approach was used for PFOS based on a monkey study). EFSA have used standardised uncertainty factors to account for the differences in pharmacokinetics whereas the US EPA has taken a different approach to managing this uncertainty, using PBPK modelling incorporating many assumptions. The use of HED or high uncertainty factors for liver toxicity may not be appropriate as it may greatly over-predict risk. For developmental toxicity use of average serum levels as the dose comparator across species may be the most appropriate approach as the ratio between maternal serum levels and foetal/neonate exposure is sufficiently similar in rats and humans. The extrapolation may need to be tempered by the likely higher sensitivity of rodents to developmental effects due to differences in PPARα between species.

The available evidence indicates PPARα plays a key role in the toxicology of PFAS and that rodents in particular are likely to be considerably more sensitive than humans. This area requires further consideration in terms of the appropriate uncertainty factor to be applied for interspecies extrapolation.

25 | P a g e

5.4 EPIDEMIOLOGYEpidemiology studies have been conducted on populations exposed occupationally, those exposed to, relatively, high levels in contaminated drinking water through leakage from nearby manufacturing facilities, and the general population. Blood levels in occupationally exposed populations are 2 to 3 orders of magnitude (100 – 1000 times) higher than the general population and provide a worst case exposure. As the occupationally exposed population is predominantly male it can only provide limited evidence regarding potential female reproductive toxicity. The following discussion compares the conclusions of EFSA and the US EPA/ATSDR reviews regarding the findings from these studies to highlight consistencies and differences in their conclusions. A detailed consideration of individual studies is beyond the scope of, and time available for this review.

Epidemiology studies available for PFAS exposed populations have found a range of associations between exposure and various diseases outcomes including cancer. When analysed as outlined earlier, these studies have, except as noted, generally failed to provide substantive evidence of adverse outcomes in exposed human populations, including those occupationally exposed where serum levels are 2-3 orders of magnitude greater than that likely in exposed populations in Australia. The conclusions of EFSA, the US EPA and the ATSDR reflect the considerations discussed above and for the most part are largely concordant, with greater or lesser degrees of caution and nuancing of those conclusions. There is a trend for the US EPA to conclude evidence of association where both the ATSDR and EFSA (together with most other international regulatory assessments) have concluded the deficiencies in or inconsistencies between studies preclude such conclusions. These differences are likely to reflect the input of the various review participants rather than a fundamental difference in interpretation by the expert epidemiology evaluators.

A reconciliation of these differences is beyond the scope of this review and should be included in the FSANZ HHRA.

5.4.1 ExposureSerum levels of PFOA and PFOS in occupationally exposed subjects are of the order of 1 to 2 µg/mL for each. In the highly exposed population around one US facility serum levels for PFOA averaged 0.423 µg/mL in 2004-2005. For the US population as a whole serum levels were 0.00392 µg/mL in 2004-2005.

5.4.2 CarcinogenicityStudies in occupationally exposed populations who have the highest blood levels of PFAS and highly exposed populations around PFAS manufacturing facilities have not, collectively, provided a basis for concluding that PFAS cause cancer. The overall conclusions of the ATSDR, EFSA and US EPA are largely concordant, with the exception of the US EPA conclusion regarding PFOA.

The ATSDR summarised the evidence as follows:

Although several studies have found significant increases in cancer risk, the results should be interpreted cautiously since most studies did not control for potential confounding variables (particularly smoking), the number of cancer cases was low, and a causal relationship between perfluoroalkyls and cancer cannot be established from these studies. Additionally, the lack of consistency across facilities may be suggestive of a causative agent other than PFOA or PFOS.

EFSA conclusions on epidemiology are:

Epidemiological studies in PFOA-exposed workers do not indicate an increased cancer risk.

26 | P a g e

Epidemiological studies in PFOS exposed workers have not shown convincing evidence of increased cancer risk.

US EPA conclusions on Epidemiology are:

PFOSHuman epidemiology studies did not find a direct correlation between PFOS exposure and the incidence of carcinogenicity in worker-based populations. Although one worker cohort found an increase in bladder cancer, smoking was a major confounding factor, and the standardized incidence ratios were not significantly different from the general population. Other worker and general population studies found no statistically-significant trends for any cancer type.

PFOAUnder EPA’s Guidelines for Carcinogen Risk Assessment (USEPA 2005a), there is “suggestive evidence of carcinogenic potential” for PFOA. Epidemiology studies demonstrate an association of serum PFOA with kidney and testicular tumors among highly exposed members of the general population.

Although the overall conclusion of the US EPA regarding the association between PFOA exposure and cancer varies somewhat from the other two assessments, in their discussion of the overall conclusions from the human cancer epidemiology studies they note that:

A group of independent toxicologists and epidemiologists critically reviewed the epidemiological evidence for cancer based on 18 studies of occupational exposure to PFOA and general population exposure with or without co-exposure to PFOS. The project was funded by 3M, but the company was not involved in the preparation or approval of the report. The authors evaluated the published studies based on the study design, subjects, exposure assessment, outcome assessment, control for confounding, and sources of bias. They followed the Bradford Hill guidelines on the strength of the association, consistency, plausibility, and biological gradient in reaching their conclusion. They found a lack of concordance between community exposures and occupational exposures one or two magnitudes higher than those for the general population. The discrepant findings across the study populations were described as likely due to chance, confounding, and/or bias (Chang et al. 2014).

The discordance within the US EPA assessment likely reflects the nature of the agencies guidance rather than an issue of scientific interpretation. Notably the available epidemiology studies do not appear to have had a bearing on the selection of a PoD for human health risk assessment by any of the agencies whose reviews were considered by enHealth.

A more detailed review of the epidemiological assessments regarding carcinogenesis than is possible in this review is an appropriate inclusion in the FSANZ consideration of HRVs for PFAS for Australia.

5.4.3 Reproductive effectsA number of studies have found an association between PFAS exposure and lower, but not low, birth weight. “Low birth weight” is a clinically defined condition associated with adverse outcomes in neonates. A lower average birth weight, that remains within the clinically normal range, does not constitute “low birth weight” in the clinical context.

27 | P a g e

This observation however does not equal a conclusion that lower average birth weights in exposed populations are acceptable or appropriate, a consideration requiring further attention in the FSANZ review preferably with clinical input.

ATSDR

There is evidence to suggest that high serum PFOA or PFOS levels are associated with lower birth weights. The significant associations have come from general population studies and a study of highly exposed residents. Studies of populations with lower serum PFOA or PFOS levels have not found significant associations for birth weight. Although significant associations were found, decreases in birth weight were small and may not be biologically relevant. No studies found an increased risk of low birth weight in infants (<2,500 g) in highly exposed residents.

EFSA

PFOA

In two recent studies, PFOA exposure of pregnant women, measured by maternal and/or cord serum levels was associated with reduced birth weight. The Panel noted that these observations could be due to chance, or to factors other than PFOA.

PFOS

The very few epidemiological data available for the general population do not indicate a risk of reduced birth weight or gestational age.

US EPA

PFOA

Studies in the high-exposure community reported an association between serum PFOA and risk of pregnancy-related hypertension or preeclampsia, conditions related to renal function during pregnancy; this outcome has not been examined in other populations. An inverse association between maternal PFOA (measured during the second or third trimester) or cord blood PFOA concentrations and birth weight was seen in several studies.

The epidemiology studies did not find associations between PFOA and neurodevelopmental effects, or preterm birth and other complications of pregnancy.

PFOS

….the available information indicates that the association between PFOS exposure and birth weight for the general population cannot be ruled out.

Although there remains some concern over the possibility of reverse causation explaining some previous study results, these collective findings indicate a consistent association with fertility and fecundity measures and PFOS exposures.

28 | P a g e

5.4.4 Other Effects

ATSDR

… consistent findings were found for association of serum PFOA and PFOS with increases in serum lipid levels, increases in uric acid levels, and alterations in biomarkers of liver damage. Although other effects have been reported, they have not been consistently found in similar types of studies, have only been examined in a single study, or were only found in general population studies.

EFSA

PFOA

Epidemiological studies in PFOA-exposed workers do not indicate an increased cancer risk. Some have shown associations with elevated cholesterol and triglycerides, or with changes in thyroid hormones, but overall there is no consistent pattern of changes.

PFOS

Epidemiological studies in PFOS exposed workers have not shown convincing evidence of increased cancer risk. An increase in serum T3 and triglyceride levels was observed, which is the opposite direction to the findings in rodents and monkeys.

US EPA

PFOA

….epidemiology studies have generally found positive associations between serum PFOA concentration and total cholesterol (TC) in the PFOA-exposed workers and the high-exposure community (i.e. increasing lipid level with increasing PFOA); similar patterns are seen with low-density lipoproteins (LDLs) but not with high-density lipoproteins (HDLs). These associations were seen in most of the general population studies, but similar results also were seen with PFOS, and the studies did not always adjust for these correlations. Associations between serum PFOA concentrations and elevations in serum levels of alanine aminotransferase (ALT) and gamma-glutamyl transpeptidase (GGT) were consistently observed in occupational cohorts, the high-exposure community, and the U.S. general population. The associations are not large in magnitude, but indicate the potential for PFOA to affect liver function. Diagnosed thyroid disease in females and female children was increased both in the high-exposure C8 study population and in females with background exposure; thyroid hormones are not consistently associated with PFOA concentration. Associations between PFOA exposure and risk of infectious diseases (as a marker of immune suppression) were not identified, but a decreased response to vaccines in relation to PFOA exposure was reported in studies in adults in the high-exposure community population and in studies of children in the general population; in the latter studies, it is difficult to distinguish associations with PFOA from those of other correlated PFASs.

29 | P a g e

PFOS

The strongest associations are related to serum lipids with increased total cholesterol and high density lipoproteins (HDLs). The associations for most epidemiology endpoints are mixed. While mean serum values are presented in the human studies, actual estimates of PFOS exposure (i.e., doses/duration) are not currently available. Thus, the serum level at which the effects were first manifest and whether the serum had achieved steady state at the point the effect occurred cannot be determined. It is likely that some of the human exposures that contribute to serum PFOS values come from PFOS derivatives or precursors that break down metabolically to PFOS. These compounds may originate from PFOS in diet and materials used in the home, thus, there is potential for confounding. Additionally, most of the subjects of the epidemiology studies have many perfluoroalkyl substances (PFAS), other contaminants, or both in their blood. Taken together, the weight of evidence for human studies supports the conclusion that PFOS exposure is a human health hazard.

6 CONCLUSIONS The primary focus of this review has been to identify the principle sources of variation between the US EPA and EFSA risk assessments of PFAS and the resultant guidance values, and to form a view on the suitability of the EFSA values selected by enHealth as an interim measure pending more extensive consideration by FSANZ. In the time available for this review it is not possible to definitively state that either one of the reviews is “correct” and the other not. There are potential strengths and potential weaknesses in each assessment and both contain value judgements that are as much policy or convention based as they are science based. Deviations by each agency from what are the general routine approaches to risk assessment are not inappropriate simply because they are not routine. A judgement to that effect requires a more detailed analysis of the reasoning and of the science underlying the respective approaches. Equally, while there is considerable room for improvement in the enHealth process and particularly the level of documentation supporting the decision making process, that observation does not impact the validity of the decisions themselves. The draft enHealth workshop report, and background documents utilised by the workshop, indicate that the workshop gave careful consideration to the sources of the variations between EFSA and US EPA reference values and considered the strengths and weaknesses of the approach taken by each agency. In particular, the workshop discussed the use of PBPK modelling by the US EPA in some detail and noted significant concerns around the nature and range of assumptions required to support the model.

Some weaknesses in the enHealth process have been identified, most particularly around the lack of scientific support provided to expert working groups and the enHealth committee, and a consequent lack of adequate detail and transparency in records of the decision making process and rationale. The draft record of the deliberations of the expert workshop is notably brief given the extent of the work conducted by the experts attending, and the extensive discussion of the various points raised. That the draft workshop report was not finalised, and had not been circulated to the attendees for comment, as at the completion of this review is notable.

30 | P a g e

6.1 USE OF INTERNATIONAL RISK ASSESSMENTS AND DERIVED HRVS

The use by Australian regulatory agencies of human health reference values, and the associated risk assessments, derived by international agencies such as the US EPA or EFSA, as interim measures to support immediate actions or deliberations is not unusual and is an appropriate mechanism for responding to environmental contamination issues in a timely manner.

6.2 SOURCES OF DIFFERENCES BETWEEN US EPA AND EFSA RISK ASSESSMENTS

As with all toxicological risk assessments, there are various sources of uncertainties, strengths and weaknesses in both the EFSA and US EPA derivations of their respective HRVs.

Although the US EPA assessment is more recent, the differences between the two assessments and the resultant HRVs are not due to differences in the studies available for assessment by the two agencies. The key sources of variation relate to the use of PBPK modelling by the US EPA and ATSDR uniquely, and differences in selection of uncertainty factors by EFSA and the US EPA. PBPK modelling is not a routine aspect of risk assessment methodologies for major international agencies other than the US EPA. That PBPK modelling is not a normal aspect of risk assessment in Australia or internationally is not however, of itself, a basis for rejecting the approach, particularly where pharmacokinetics is a pivotal source of interspecies differences in response to PFAS. The core consideration, as identified by the enHealth workshop, is the validity and utility of the model. Thus although the US EPA assessment utilised ostensibly sophisticated PBPK modelling the necessary range of assumptions, between humans and the respective experimental animals, means that the use of PBPK has potentially replaced one set of uncertainties with another.

The US EPA use of HEDs based on PBPK modelling of serum levels of PFAS is not likely to be appropriate for liver effects because liver exposure for the same serum PFAS levels will be higher in rats than in humans, at the least during the absorption phase. Actual administered dose in mg/kg bw is likely to be a better basis for determining the PoD for liver effects.

The use of average or actual final serum levels, by both the US EPA and EFSA, to compare internal exposures/doses across experimental animals to determine the PoD is appropriate for compounds with high variation in pharmacokinetic parameters. The limitations of the model used to convert final serum level to average levels by the US EPA PBPK modelling, which is not a routine aspect of risk assessment methodologies for major international agencies other than the US EPA (and ATSDR), is noted however.

There are indications in the data set that, at least for a number of toxicological effects used as the basis for determination of the PoD, humans are likely to be less sensitive than animals due to both pharmacokinetic (liver) and pharmacodynamics (developmental effects) differences. Further consideration of the appropriate adjustment (reduction) of uncertainty factors to recognise this observation are warranted.

6.3 POTENTIAL PUBLIC HEALTH CONSEQUENCES OF THE CHOICE OF HRVS

Because of the exceptionally long half-life of PFAS in humans, the systemic (i.e. internal) exposure to PFAS is determined by oral (or other routes of) exposure over long periods of time. As a consequence, even an order of magnitude reduction in levels in drinking water would not

31 | P a g e

significantly impact blood levels for a protracted period and the issue of whether the EFSA or US EPA values are more appropriate is largely esoteric over the short to medium term.

As the phasing out of PFAS for most uses has resulted in declining exposures and a progressive decline in serum levels where this has been monitored, the choice of EFSA HRVs over those of the US EPA has no substantive impact on public health in the short to medium term.

6.4 BALANCING RISK MITIGATION WITH RISK GENERATIONFor anthropogenic contaminants such as PFAS which cannot readily be removed from the environment, the establishment of values that are, with respect to the overall weight of evidence, disproportionately low, has the potential to result in a range of adverse health outcomes which may be greater than the toxicological risks intended to be avoided. Such outcomes may include prolonged unwarranted stress in exposed populations, the recommendation, or seeking out, of unnecessary medical interventions with their attendant risks, or interventions in pregnancy and avoidance of breast feeding to the detriment of the foetus and neonate. Other, economic impacts although likely, are beyond the scope of a HHRA and of this review. Simplistic selection of the lowest international HRV is therefore not necessarily optimal for the overall protection of public health, regardless of the time sequence of the available assessments. Determination of suitable HRVs for PFAS requires a careful consideration of the strengths and weaknesses of the approaches taken by international agencies that have had access to the underlying data, and a considered selection of the most appropriate approach/values within the context of the exposure patterns in Australia. A suitably precautionary approach to public health requires a balancing of risk prevention against the potential for risk generation.

6.5 OVERALL CONCLUSION The adoption of the EFSA HRVs as a temporary (i.e. interim) measure, pending a formal more extensive review by FSANZ, is appropriate and is protective of public health.

7 RECOMMENDATIONSBased on the findings of this review the following recommendations are made;

The adoption of the EFSA health reference values (TDI) and their use to derive Australian drinking water guideline values, as an interim measure pending FSANZ review, can be concluded to be appropriate, to be based on the expert consideration of the strengths and weaknesses of the available risk assessments from international agencies, and to be consistent with current risk assessment practices both in Australia and internationally.

Consideration should be given to the need for the responsibility for setting HRVs, and particularly for contaminants that are also present in food and water, to be formally supported in future by an existing agency with;

experience and expertise in setting, and documenting, these values,

appropriate consultation mechanisms in place to ensure the highest possible degree of transparency,

the capacity to provide expert scientific support to expert working groups and decision making committees,

32 | P a g e

The scientific literature on, and international regulatory assessments of the HRVs for PFAS should be monitored on an ongoing basis by FSANZ in conjunction with enHealth and adjusted up or down as indicated by the emerging data.

Australia, through FSANZ or another suitable agency, should consider whether there is value in seeking to initiate an international consultative review of HRVs for PFAS through the CODEX/JECFA mechanism to establish consistent international HRVs for these substances.

As identified in both this review and by the enHealth workshop, pivotal issues that FSANZ should address and consider seeking specialist expert advice on, include;

the strengths, weaknesses and validity of the PBPK approach to HED calculation for PFAS, the relative merits of the interpretation of the epidemiology data by the US EPA compared

to most other international agencies’ the clinical relevance of the observed lower birth weights and elevated cholesterol levels in

highly exposed populations, the significance of the recent US National Toxicology Program (NTP) review of the immune-

toxicity potential of PFAS.

8 REFERENCESAndersen, M., Clewell HJ, Tan, Y.-M., Butenhoff, J., & Olsen, G. (2006). Pharmacokinetic modeling of

saturable, renal resorption of perfluoroalkylacids in monkeys—Probing the determinants of long plasma half-lives. Toxicology, 227(1-2), 156-164.

ATSDR. (2015). Toxicological profile for perfluoroalkyls, Draft, . Agency for Toxic Substances and Disease Registry. Retrieved from http://www.atsdr.cdc.gov/ToxProfiles/tp200.pdf

Bradford-Hill, A. (1965). The Environment and Disease: Association or Causation. Proceedings of the Royal Society of Medicine - Section of Occupational Medicine, 58(5), 295-300.

Butenhoff, J., Ehresman, D., Chang, S., Parker, G., & Stump.D. (2009). Gestational and lactational exposure to potassium perfluorooctanesulfonate (K+PFOS) in rats:Developmental neurotoxicity. Reproductive Toxicology(27), 319-330.

Butenhoff, J., Kennedy, G., Frame, S., O’Conner, J., & York, R. (2004a). The reproductive toxicology of ammonium perfluorooctanoate (APFO) in the rat. Toxicology, 196, 95-116.

COT. (2006a). COT Statrement on the Tolerable Daily Intake for Perfluorooctane Sulfonate, . UK Committee on Toxicology of Chemicals in Food, Consumer Products and the Environment. Retrieved from http://cot.food.gov.uk/cotstatements/cotstatementsyrs/cotstatements2006/cotstatementpfos200609

COT. (2006b). COT Statement on the Tolerable Daily Intake for Perfluorooctanoic Acid, . UK Committee on Toxicology of Chemicals in Food, Consumer Products and the Environment. Retrieved from http://cot.food.gov.uk/cotstatements/cotstatementsyrs/cotstatements2006/cotstatementpfoa200610

33 | P a g e

COT. (2009). COT Update Statement on the tolerable daily intake for perfluorooctanoic acid. UK Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment. Retrieved from http://cot.food.gov.uk/cotstatements/cotstatementsyrs/cotstatements2009/cot200902

COT. (2014). COT Statement on the potential risks from perfluorooctane sulfonate in the infant diet, . UK Committee on Toxicity of Chemicals in Food, Consumer Products and the Environment. Retrieved from http://cot.food.gov.uk/committee/committee-on-toxicity/cotstatements/cotstatement2014/cotstmtonpfos

Danish Ministry of the Environment. (2015). Perfluoralkylated substances: PFOA, PFOS and PFOSA: Evaluation of health hazards and proposal of a health based quality criterion for drinking water, soil and groundwater.

DeWitt, J., Copeland, C., Strynar, M., & Luebke, R. (2008). Perfluorooctanoic acid-induced immunomodulation in adult C57BL/6J or C57BL/6N female mice. Environmental Health Perspectives, 116, 644-650.

Dong, Z., Bahar, M., Jit, J., Kennedy, B., Priestly, B., Ng, J., . . . Naidu, R. (2016). Critical Comments on Refrence Doses for Perfluorooctane Sulfonate and Perfluorooctanoic Acid. Pending - Personnal Communication.

EFSA. (2008). Perfluorooctane sulfonate (PFOS), perfluorooctanoic acid (PFOA) and their salts, . European Food Safety Authority. , Scientific Opinion of the Panel on Contaminants in the Food chain, .

EFSA. (2012). Perfluoroalkylated substances in food: occurrence and dietary exposure. EFSA Journal, 10(6), 2743 (1-55).

enHealth. (2012). the enHealth Environmental Health Risk Assessment, Guidelines for Assessing Human Health Risks from Environmental Hazards. Retrieved from http://www.eh.org.au/documents/item/916

FSANZ. (n.d.). Risk Analysis in Food Regulation . Food Standards Australia New Zealand. Retrieved from http://www.foodstandards.gov.au/publications/riskanalysisfood regulation/Pages/de fault.aspx

GDWC. (2006). Provisional evaluation of PFT in drinking water with the guide substances PFOS and PFOS as examples. German Ministry of Health and the Federal Environment Agency., Drinking Water Commission.

Ioannidis, J. (2005). Why most published research findings are false. PLoS Medicine,, 2(28), e124.

Lau, C., Thibodeaux, J., Hanson, R., Narotsky, M., Rogers, J., Lindstrom, A., & Strynar., M. (2006). Effects of perfluorooctanoic acid exposure during pregnancy in the mouse. Toxicological Sciences, 103, 510-518.

Lau, C., Thibodeaux, J., Hanson, R., Rogers, J., Grey, B., Stanton, M., . . . Stevenson, L. (2003). Exposure to perfluorooctane sulfonate during pregnancy in rat and mouse. II: Postnatal evaluation. Toxicological Sciences(74), 382-392.

Luebker, D., M, C., York,R, Moore, J., Hansen, K., & Butenhoff, J. (2005). Two-generation reproduction and cross-foster studies of perfluorooctanesulfonate (PFOS) in rats. Toxicology, 215, 126-148.

34 | P a g e

Luebker, D., York, R., Hansen, K., Moore, J., & Butenhoff.J. (2005a). Neonatal mortality from in utero exposure to perfluorooctanesulfonate (PFOS) in Sprague-Dawley rats; Dose-response and biochemical and pharmacokinetic parameters. Toxicology(215), 149-169.

Macon, M., Villanueva, L., Tatum-Gibbs, K., Zehr, R., Strynar, M., Stanko, J., . . . FentonS. (2011). Prenatal perfluorooctanoic acid exposure in CD-1 mice: Low dose developmental effects and internal dosimetry. Toxicological Sciences, 121(1), 134-145.

MDH. (2009a). Health Risk Limits for Groundwater 2008 Rule Revision – Perfluorooctane Sulfonate. Minnesota Department of Health.

MDH. (2009b). Health Risk Limits for Groundwater 2008 Rule Revision – Perfluorooctanoic Acid. Minnesota Department of Health.

NHMRC. (2008). Guidelines for Managing Risks in Recreational Water. National Health and Medical Research Council.

NHMRC. (2016). Australian Drinking Water Guidelines. National Health and Medical Research Council.

NICNAS. (n.d.). Handbook for notifiers. National Industrial Chemicals Notification and Assessment Scheme . Retrieved from https://www.nicnas.gov.au/regulation-and-compljance/nicnas-handbook

Perkins, R., Butenhoff, J., Kennedy, G., & Palazzolo, G. (2004). 13-Week dietary toxicity study of ammonium perfluorooctanoate (APFO) in male rats. Drug & Chemical Toxicology, 27, 361-378.

Seacat, A., Thomford, P., Hansen, K., Clemen, L., Eldridge, S., Elcombe, C., & Butenhoff, J. (2003). Sub-chronic dietary toxicity of potassium perfluorooctanesulfonate in rats. Toxicology, 117-131.

Seacat, A., Thomford, P., Hansen, K., Olsen, P., Case, M., & Butenhoff, J. (2002). Subchronic toxicity studies on perfluorooctanesulfonate potassium salt in cynomolgus monkeys. Toxicological Sciences, 68, 249-264.

Sibinsky, L. (1987). Final report of a two-year oral (diet) toxicity and carcinogenicity study of fluorochemical FC-143 (perfluorooctanane ammonium carboxylate) in rats.. 3M Company/RIKER.

Two-generation reproduction and cross-foster studies of perfluorooctanesulfonate (PFOS) in rats. (2005b). Toxicology, 215, 126-148.

US EPA . (2009a). Memorandum – Soil screening levels for perfluorooctanoic acid (PFOA) and perfluorooctyl sulfonate (PFOS). (soil screening.

US EPA. (2009b). Provisional health advisories for Perfluorooctanoic acid (PFOA) and Perfluorooctane sulfonate (PFOS). Retrieved from http://water.epa.gov/action/advisories/drinking/upload/2009_01_15_criteria_drinking_pha-PFOA_PFOS.pdf

US EPA. (2014a). Health Effects Document for Perfluorooctane Sulfonate (PFOS) (DRAFT). Office of Water.

35 | P a g e

US EPA. (2014b). Health Effects Document for Perfluorooctanoic Acid (PFOA) (DRAFT),. Office of Water.

US EPA. (2016a). Health Effects Support Document for Perfluorooctane Sulfonate (PFOA). Office of Water.

US EPA. (2016b). Health Effects Support Document for Perfluorooctane Sulfonate (PFOS. Office of Water.

36 | P a g e