Advanced REACH Tool

1
Advanced REACH Tool

• ART

2
Collaborative project

• HSL Nick Warren, Kevin McNally
• IOM John Cherrie, Martie van Tongeren, Peter
  Ritchie
• BAuA Martin Tischer
• NRCWE Thomas Schneider
• University Utrecht Hans Kromhout
• TNO Jody Schinkel, Wouter Fransman, Hans
  Marquart, Joop van Hemmen, Erik Tielemans

3
Need for higher Tier approach
Tier 1

• First Tier inherently conservative
• Simple, easy-to-use, inexpensive
• Higher Tier approaches needed for subgroups
• As simple as possible, but not simpler
• A generic higher Tier tool increases
  cost-effectiveness and speed of RA
• Case-by-case approach is alternative

Reduced complexity
Generic higher tier
Increased precision
Specific approach
4
Need for consistency

• Different models should approximately give
  similar results for the same ES
• We do not know!
• Applicability domain should be clearly described
• ART may enhance consistency
• Conceptual model for exposure assessment
• Provides good quality data for validation

5
ART approach

• New developments on
• Mechanistic models
• Bayesian statistics
• Databases
• Software
• Validation

6
Making full use of all information

BDA
Uncertainty
Worked example!
Model Data Updated estimate
7
Mechanistic model ART (1)Cherrie et al. Ann
Occup Hyg 1999Tielemans et al. Ann Occup Hyg
2008

• 9 Modifying Factors (MF)
• - Intrinsic emission potential (E)
• - Activity emission potential (H)
• - Local controls (LC)
• Separation (Sep)
• Segregation (Seg)
• - Surface contamination (Su)
• - Dilution (D)
• - Personal behavior (P)
• - RPE

Equations
8
Mechanistic model ART (2)
Jan. 2008 Oct. 2008
June 2009
1. Conceptual model
2. Quantification of MFs
3. Expert review
4. Calibration
5. Variability
9
Bayesian approach similarity algorithm

• ART will explicitly incorporate uncertainty of
  exposure data arising
• through sample size and level of similarity with
  assessment scenario

Assessment Scenario
Measurement series
MF 1 MF 2 MF 3 MF n
MF 1 MF 2 MF 3 MF n
Similarity Algorithm
Uncertainty factor
10
The overall picture
Small, analogous dataset
Own dataset
Large, partially analogous dataset
Mechanistic model estimate
11
Evolving system
Mechanistic model
Larger applicability domain
More detailed insights into MFs
Additional data for calibration
Increased accuracy
Additional data for Bayesian update
12
Where do we stand now?
Stakeholder Meeting September 2009
Jan. 2008 Jan. 2009
Jan. 2010

• Mechanistic and
• Bayesian model

2. Database and software
3. Testing and validation
Steering committee meeting
13
Worked example ART dye manufacture

• Manual scooping and dumping of powdered dyes
• 2hrs of exposure related tasks. No relevant
  exposure during rest of shift
• Fairly hazardous compounds
• Full shift exposure to inhalable dust

14
Dye manufacture user input
Intrinsic emission potential (dustiness) 5
classes 0.1 Firm granules, flakes or
pellets 10 Extremely fine and light
powders

• Localized control
• 12 classes
• No control
• 0.1 Wet suppression
• 0.1 LEV
• 0.01 LEV and partial enclosure
• 0.001 LEV and complete enclosure

15
Dye manufacture mechanistic model
Calibration based on approximately 500
measurements
Final calibration will be based on several
thousands of measurements
16
Dye manufacture exposure measurements

• Completely analogous data
• Most likely added by the user
• 21 measurements, 6 sites
• Median 0.19 mg m-3
• Range 0.06 - 1.29 mg m-3
• GSD 2.8
• First update uses data from just one company
• Second update uses all measurement data

17
Dye manufacture updated predictions
18
Dye manufacture updated predictions
19
Dye manufacture predictions
20
Summary (1)

• Research project is up and running
• ART will be available early 2010
• Beta version might be available earlier
• Approach makes use of modelled estimates and
  measurements
• ART facilitates higher tier exposure assessment
  under REACH
• Provides estimates of whole distribution
• Allows inclusion of any new data that becomes
  available
• Familiar users with ready information should be
  able to complete an assessment within 5 minutes

21
Summary (2)

• Approach facilitates sharing of exposure data
  down and up the supply chain
• Research project provides good quality data for
  validation of different models
• Additional discussions are needed on issues
  related to variability and uncertainty
• Which percentile of distribution should be used
  in RA?
• What level of precision is required?

22
Thank you for your attention

• Questions?