What are the key data gaps that prohibit or limit our ability to assess risks from PFAS mixtures?

The largest data gap is an adequate problem formulation statement. It is really not feasible to regulate all PFAS as a single group. The pchem properties, the half-life, the toxicology, the exposure, etc. differ immensely among the different subgroups of PFAS as defined by the OECD. If care is taken to identify the exposure scenario to be addressed and the relevant PFAS compounds, then a reasonable approach can be developed. The data gaps will differ if one is developing an assessment of drinking water versus furniture versus clothing versus food. Risk assessment is an iterative process. As the assessment matures, the key data gaps will become evident and studies can be done to fill those data gaps.

Toxic mode of action? We need to know the critical mode of toxicity and if that mode is common for all PFAS subgroups. Otherwise, make a subgroup and perform testing to see if subgroups have similar mode of toxicity.

It would be helpful to identify how a chemical that does not form covalent bonds with macro molecules in cells is able to exert its toxicity. Is it displacement of signaling compounds, is it changes in fatty acid metabolism? If these processes could be identified, then PBPK/PD models could generate predictions of the dose response to the MIE for large numbers of PFAS.

- Comprehensive toxicology information is lacking for most PFAS individually. The compositions of whole mixtures are highly variable, and rarely if ever evaluated.
- Available (and limited) human and animal toxicology information is challenging to interpret for risk assessment purposes due to inconsistent results and variable concordance between animals and humans.
- Limited or no toxicokinetic data for most PFAS severely limits understanding relationships between external exposure, and internal dose to the target tissues, and how different PFAS affect each other in terms of ADME

My understanding is that methods to measure total organic fluorine through AOF and EOF measurements are not very sensitive. Also, the mode of action varies across PFAS, and is unknown for most PFAS on the market.

Human elimination half-lives, detailed dose-response toxicity studies for multiple PFAS/endpoints, mode of action information, spatiotemporal measurements (i.e. concentrations) for multiple PFAS (at least PFAAs and perfluoroalkyl ether acids) in relevant exposure media.

Definition of the group and the objectives of the assessment. Who says that mixtures must be assessed and that not one or two compounds are sufficient to take action? What shall be achieved within which period of time?
I appreciate the initial set of literature provided; however, is biased towards U.S. situations and would not apply to Europe or to research. Question: need to amend with information from other constituencies? Proposal: going back to original literature and assess from there along the terms of reference(agreed criteria.
In general, I believe that there are not many hard data available and most is copy and paste and using the same models and assumptions or are conclusions from observations (monitoring).
In general, there are different actors at different stages.

The TEF approach developed for dioxins, PCBs and PAHs assumed a common MoA and additivity of contributions to mixtures risk assessment. There have been some attempts to develop Relative Potency Factors (RPFs) for some PFAS (e.g. RIVM 2018, outlined further in Bil 2020; Borg et al 2013; Cox & Naderbaum Ecoforum 2020 abstract (RPFs vs PFOS) - references can be supplied on request). An alternative approach, using potency for thyroid hormone effects in a TTR-TTbeta CALUX assay, has been proposed by Behnisch et al 2021 (Env Int 157 106791). However, it is not clear whether a single toxicological endpoint, or in vivo vs in vitro study design, or study duration, will provide sufficient comparative toxicological potency information to enable robust RPFs do be developed across the PFAS spectrum. The challenges associated with development of RPFs across the spectrum of relevant PFAS is well described by Cousins et al 2020.

Fundamentally, exposure and dose response data are needed.

Exposure seems to be a key data gap, comprising both qualitative measures of exposure (identification of the PFAS chemicals comprising the mixture) and quantitative measures of exposure (determining the concentration of the PFAS chemical in the exposure media, and estimating the absorbed dose and/or target tissue concentrations attained).

Toxicity data seem to be limiting to absent for many PFAS chemicals potentially exposed to humans. These data relate to the target organ/tissue system affected, and it should be noted that chemicals may affect multiple organs or functions. Data describing the toxicity organism-wide, should be sought. The seeking and prioritization of dose-response data should prioritize data from in vivo studies, with a fall-back position of reliance on in vitro, high throughput data or read-across approaches to identify affected organs. In vivo dose response data represent the gold standard, but can be augmented by quantitative in vitro derived data; the application of QSAR or read-across data to estimate dose response will be accompanied by a level of uncertainty likely to be deemed unacceptable for quantitative reliance. Dose response data should be expressed in terms of tissue dose or tissue concentration.

To facilitate a hazard index approach, toxicity values (risk values, Reference Dose values) should be available, or should be developed for PFAS chemicals. It may be envisioned that a prioritization scheme based initially on exposure data might be used to identify and prioritize chemicals for the development, estimation or assignment of toxicity values.

Because mode of action information is the basis to deviate from dose additive approaches, consideration of available MOA information should be prioritized. The relative impact of moving from the default position of dose additivity may be hypothesized and investigated on the basis of available exposure information. The results from such an activity might be used as the basis for decisions whether to seek or develop experimental evidence (e.g., from in vitro investigations) that would support divergent modes of action.

The development of toxicity pathways, modes of action or adverse outcome pathways for the PFAS-associated toxicities should be strongly considered. Such information may lay the groundwork for high throughput data to become quite useful in chemical grouping.

Lack of data sufficient to establish MOA for various endpoints; this includes the applicability of short term and in silico results. Lack of information sufficient to determine likelihood of hazard (association of several adverse outcomes with exposure in humans). Information on generalizability of half-life in rodents to humans. Although half-life in humans is much longer than in experimental animals, are results scalable for individual (or groups) of PFAS?

Based on traditional risk assessment approaches, there are data gaps for: 1) Hazard identification. Due to limited numbers of purified standards for analytical chemistry approaches, confirmation of PFAS identities in environmental mixtures also is limited. Structures can be deduced but not verified analytically. This is one data gap. 2) Dose-response assessment. Only a handful of PFAS have been thoroughly studied toxicologically (at least with respect to what is available in the peer-reviewed literature). This is another data gap. However, such gaps in data may not as extreme as anticipated. For example, many epidemiological studies that have been conducted for PFAS involve populations exposed to complex PFAS mixtures but only a handful of PFAS have been analyzed for these studies. A broader analysis of PFAS present in archived environmental and/or biological samples from these studies may indicate that we actually have high quality data on complex PFAS mixtures and therefore the data gaps mentioned here may be somewhat narrow.