The concept of using a tiered approach is utilized in many mixtures risk assessment frameworks and mentioned by some panelist as a viable option for assessing PFAS. Please provide a feasibility ranking and confidence ranking for the proposed tiered approach below, based on currently available PFAS information and the problem formulation presented above (i.e., development of drinking water standards). Based on input received from the panel in Round 1, “Carbon chain length/chemical structure” ranked highest when both feasibility and science merit are considerd (see figure link below). The steps belows provide a "strawman" tiered approach. Please provide comments on acceptability and/or ways to improve. a. Step 1. Define PFAS group based on potential presence in drinking water based on production and use b. Step 2. Define subgroups of PFAS based on shared similar physical-chemical properties, andcarbon chain length/ chemical structures, including chain-length and (e.g., functional groups) c. Step 3. Assess potential risk (hazard and exposure) for each subgroup. d. Step 4. Determine scientific feasibility for assessing potential mixtures interactions between subgroups.

I disagree with the tiered approach outlined here. I do not think that production/use information is sufficient to to determine that a chemical is present in drinking water. This would need to be expanded to examine bioavailability, pathway to drinking water, and some measure showing actual presence in drinking water. Otherwise the list of chemicals will be very long and not scientifically defensible. I would start this exercise with PFAS that are known to be present in drinking water and then go through the steps outlined above. This will demonstrate the issues in defining subgroups and the feasibility of grouping any of them.

This is a sound approach. In step 1, one would probably use analyticaI data if available. Production and use data could probably be used to get an approximation of exposure if analytical data are missing. In step 2, one would also need to group the precursors with their corresponding perfluoroalkyl acids.

Step 1: This may not be highly feasible given lack of information on production and use due to CBI and lack of information on smaller users of PFAS. Therefore, on a scale of 1-10, this step ranks under a 5 for both feasibility and confidence unless reliable information on production and use is made available along with purified product for analytical standards. Step 2: This is highly feasible as PFAS are already generally categorized by chain length/chemical structures. Confidence in this approach also is high and this information, if not already available, can be deduced from analytical approaches. Therefore, on a scale of 1-10, this step ranks at about a 8 or 9 for feasibility and confidence. Step 3: This may be feasible for subgroups that have associated hazard and exposure information. It's challenging to assign a ranking to this step without knowing whether or not such information will be available. IF information on hazard and exposure is available, then this step will rank high for feasibility and confidence (8-9). Step 4: Potentially highly feasible with high confidence IF information is available. Overall, this approach seems rationale, but the first step is dependent on the availability of appropriate information. If the appropriate information is not available for the first step then the other steps will drop in feasibility and confidence.

This step-wise approach is OK, albeit very based in chemical structure. By step 3, the subgroups will have to be modified to group PFAS with the same MOA, similar dose-response curves, or same adverse outcome. This re-grouping may well cross over the chemically-constructed groups. Step 4 may then be out of line; one wouldn't expect interactions across adverse outcomes -- at least not in any way that would be feasible to assess.

I suggest a step zero -- application of a screening device such as Risk 21 (e.g Embry M, Bachman A, Bell D, Boobis A, Cohen, S, Dellarco M, Dewhurst I et al. The Risk Assessment in the 21st Century (RISK21): roadmap and matrix. Crit Rev Toxicol 2014;44 (Suppl 3):6–16; Douglas C Wolf, Ammie Bachman, Gordon Barrett, Cheryl Bellin, Jay I Goodman, Elke Jensen, Angelo Moretto, Tami McMullin, Timothy P Pastoor, Rita Schoeny, Brian Slezak, Korinna Wend, Michelle R Embry. Illustrative case using the RISK21 roadmap and matrix: prioritization for evaluation of chemicals found in drinking water. Critical Reviews in Toxicology, 1-11, 2015. )

The approach presented above has many limitations:

Step 1: The PFAS measured in drinking water from impacted communities often do not mirror current production/use. In many cases, drinking water has been associated with disposal - rather than use of - PFAS (e.g., landfills). Presence in drinking water should be based on targeted and non-targeted methods. Because produced PFAS change over time, the grouping method should be sustainable, i.e., it should not have to be changed when new PFAS are produced/released.
Step 2: The ideal approach, while not feasible at this point, would be to define groups based on mode/mechanism of action. An approach based on chemical structure is useful if it predicts activity for a given mode/mechanism of action. I am not sure we have good predictive models for PFAS activity, namely because predictive models are trained on existing data, and there is little data out there on emerging PFAS.
Step 3: This issue with this step is that there is no point of departure or drinking water guideline for most PFAS. We would probably have to rely on a guideline for one PFAS that is representative of the group.
Step 4: There is barely sufficient information to define groups, so I don't believe we can assess the mixtures interactions between subgroups.

In terms of improvement, with the data currently available in the literature, I think it's best to keep all PFAS within the same group, and to use a single drinking water guideline. That avoids having to make decision in the context of large uncertainty.

I am unable to draw lines / connectors from that figure

The wording in this question is a bit confused (e.g., step 2 repeats the same ideas using different terms). In addition, it might have been helpful to provide a method for the panel to provide a feasibility and confidence ranking (e.g., rank from 0 to 10). The following are my comments on the four steps.
Step 1.
Production should include the unintended production of PFAS substances that may occur (byproducts and contaminants). In addition, exclusion of PFAS in this step could be extended to substances that do not have the physical characteristic that will allow them to contaminate surface or ground water supplies. This would be low water solubility and tendency to bind to soils and suspended solids.
Step 2.
Need to define what the subgroups mean. Are they CAGs? Are they groups that have different ranges of toxicity? Are they groups where the masses of all members are summed and compared to a standard? Are they groups with similar potentials for removal by a common technology (activated carbon, or pH adjustment)? Once this is defined then the subgroups can be defined.
Step 3.
Potential risk (hazard and exposure) should be made on a chemical basis not on a sub-group basis. This could be explained better in the final tier.
Step 4.
Interactions between chemicals in different groups should be considered. Not clear if this should done on a subgroup basis.

The procedure presented represents a potentially workable initial approach, with a feasibility ranking of 7 and a confidence ranking of 5. Issues arise when defining subgroups according to physical chemical properties and chain length. This forces considerations of interactions downward in the approach, potentially restricting the assessment of interactions to the subgroup level, and perhaps precluding an assessment of potential interactions at the chemical level. An alternate approach would be to define subgroups according to observed health endpoints. Then assess potential risk (hazard and exposure) for each subgroup by testing for additivity and assess for the likelihood of interactions. Lacking evidence of interactions, dose additivity could be applied to assess risk for each subgroup, and Step 4 could be undertaken.

The approach described is pretty consistent with RA practice, though it is a bit out of order and incomplete.
Step 1 = problem formulation.
Step 2 = part of hazard assessment (assuming that structure etc. is a good proxy for toxicity).
Step 3 = part of toxicity assessment, and exposure assessment.
Step 4 = part of toxicity assessment; perhaps belongs in an uncertainty analysis.

What is missing is risk characterization (describing and quantifying the risk) and an uncertainty analysis

A tiered approach is a viable option for the risk assessment of PFAS mixtures. The proposed "strawman" reads logical and a way to move forward. The step 1 is based on production/use data, but for chemicals like PFOS/PFOA, there is enough analytical/monitoring data and when there is enough analytical data, that should be used in risk assessment.

It is unfortunate that the two highest ranked for "scientific merit" (toxicological endpoints and mode of action) did not rank well for "feasibility". That said, the proposed "strawman" tiered approach is a pragmatic solution as an interim approach. It could become more refined as more data emerges on mode of action and toxicological outcomes for a broader range of PFAS. My comment on Step 1 is that it is too narrowly focussed on drinking water exposures (see also, my response to question 2.1).