LinkedIn

Wednesday, December 12, 2018

Are REACH Exposure Models Good Enough?


The political will in the European Union to enact REACh was and is extraordinary.   The body politic in the EU wants this regulation and certainly needs it to be effective. It should be clear that it cannot be effective if the exposure assessment half of the risk equation used for REACh is faulty.  Underestimation of exposure and risk hurts people's health directly, over-estimations hurts people's well-being by unnecessary hurting of the economy.   The use of good modelling tools is critical or REACh, in my opinion, will ultimately be doomed to fail. 

I have always thought that first principle physical chemical models (FPModels) have been superior to models that are not based on first principles (NFPModels).  Now a thoughtful and talented Danish researcher (Dr. Antti Joonas Koivisto) is examining and demonstrating with logic and DATA exactly why first principle models are better and, most likely, even necessary to make good regulatory decisions.

An early question might be:  Why develop NFPModels when FPModels are available for development?   The easy and probably correct answer:  They can be developed relatively quickly and with less effort and expense. FPModels are available but need to be parameterized for critical exposure scenarios and that means research dollars.

NFPModels, for the most part, are based on dimensionless factors to calculate scores, which are then converted to exposure values.  They are conceptual models than do not have to conform to first-principles and are thus (using Joonas' word) somewhat vague.

While there are other NFPModels, the big hitter in the EU for modelling exposure via REACh appears to be Stoffenmanager® v.7.1 which as of last month:

·         is reportedly validated by 15 scientific studies based on more than 6000 measurements. 
·         has more than 33,000 users with 50 new users per week. 
·         used to make over 200,000 regulatory decisions

It is accepted by the Dutch Labour inspectorate as a validated method to evaluate exposure to hazardous substances in the workplace.   More important, the European Commission officially recognises Stoffenmanager as a instrument to comply with the REACh regulation.

Other REACh-recommended NFPModels include:

ECETOC TRA
MEASE
EMK-EXPO-TOOL
ART

Although somewhat varied in their approach, they all share the same feature that they are
all based on dimensionless factors to calculate scores, which are then converted to exposure values.  They are conceptual models than do not have to conform to first-principles (like the conservation of mass).  Thus, they are not scientifically formalized and that leaves them difficult to explain.

Dr. Koivisto asserts, and I agree, that there should be minimum requirements for regulatory exposure models and that those criteria should be no less than the Daubert criteria used in US Courts for valid scientific testimony.  The model criteria: 
  •          Is applicable and has been tested.
  •          Has been subjected to peer review and is generally accepted.
  •          The rate of error is known and acceptable.
  •          The existence and maintenance of standards and controls concerning the                     operation.
  •          Is generally accepted in the relevant scientific community.
Joonas goes on to advise that FPMmodels are superior to the above NFPModels (what he calls “imaginary” models) because:

       Mass flows are traceable à Model can be used for environmental, occupational and consumer exposure assessment!!
       There is No unit conversions!!
       Error analysis can be made separately for emission source, emission controls, and dispersion.
       No need for Tier levels;  the Tier level depends on available information.
       Possible ”calibration” is straight forward (e.g. chamber tests)
       In the NF/FF model the NF volume and air mixing are adjustable according to the source (free parameterization).
       Results are easy to interpret
       TRANSPARENT!
       Easy to develop for further needs
       No need to discretize parameters (e.g. room size, ventilation rate,…)
       Accuracy superior when compared to compared to mechanistic or conceptual modes

I took most of the above from a November 29, 2018 presentation that Joonas gave in Denmark.  I will be happy to send the PowerPoint slide deck of that talk to anyone who asks at mjayjock@gmail.com.

4 comments:

  1. Mike, can you give some examples of First Principle Models

    ReplyDelete
    Replies
    1. The simple well mixed box model that tracks the amount of contaminant going in and out of a box of air as a function of generation and ventilation. Please send me an email and I will send you references.

      Delete
  2. Mike, glad to see the blog up and running! Pls send me the presentation.
    Steve

    ReplyDelete
    Replies
    1. Steve, you show up as Unknown. Please send me an email.

      Delete