WORKING FOR A HEALTHY FUTURE

INSTITUTE OF OCCUPATIONAL MEDICINE . Edinburgh . UK www.iom-world.org

The Advanced REACH Tool

John Cherrie

Summary…

• Background• The concept of ART• The mechanistic model tool• Dispersion, an example of the work to underpin

the model

• Bayesian updating• Model output

Background…

• ART is a generic higher tier tool• Included in the REACH Guidance• Applicable for exposure to dust, vapour, mist, and

fumes• Co-development project funded by:• Shell, GSK, Eurometaux, Cefic• HSE, Dutch government, Afsset• BOHS

• Delivered by:• TNO, IOM, HSL, BAuA, UU and NRCWE

The conceptual basis…

• Based on the premise that both models and measurement data can be informative about exposure

• Set within a Bayesian framework

Combining data reduces uncertaintyU

ncer

tain

ty

+ BDA

Model Data Updated estimate

The ART approach…

Model

Bayesian process to combine data

and model output

Exposure estimates for risk assessment

Similarity module to select data for risk

assessment

Exposure database, with contextual

information

Modelling the exposure distribution

Scenario

Median exposure

Exposure variability

Mechanistic model

Information from meta-analysis

Model specification…

jklikijiijkli wcY )()()()()( )(ln

- Between-company variability

- Between-worker variability

- Within-worker variability

Underpinning the mechanistic model…

• The early work by Cherrie, Schneider et al• Developed a model for use in epidemiological studies• Multiplicative deterministic model• Near-field/Far-field• Relied on the assessor to choose the parameters• Validated in a number of different situations

• Dispersion parameters derived from a 2-box model

Cherrie and Schneider. (1999) Validation of a new method for structured subjective assessment of past concentrations. The Annals of Occupational Hygiene; 43 (4): 235-245.

Cherrie. (1999) The effect of room size and general ventilation on the relationship between near and far-field concentrations. Applied occupational and environmental hygiene; 14 (8): 539-46.

An example of the use of this model…

0.001

0.01

0.1

1.0

10.0

100.0

1000.0

0.0 0.0 0.1 1.0 10.0 100.0

Measured exposures (ppm)

Esti

mate

d e

xp

osu

res (

pp

m)

101

Validation of the model…

The ART mechanistic model…

The mechanistic model…

Source

LCIR

Source barrier

Far-field

Personal barrier Surfaces

Near-field

RPE

2nd

Personal behaviour

Inhalation boundry

Nine modifying factors…

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

Three basic equations…

RPECCC ffnft )(

nfnfnfnfnfnfnf DSuPLCHEC )(

ffffffffffffffff SepDSuSegLCHEC )(

Substance emission potential…

Category Relative weight

Firm granule 0.01

Granule 0.03

Coarse dust 0.1

Fine dust 0.3

Extremely fine dust

1.0

Substance emission potential…

iiimixi pp ,

Partial vapour pressure of substance i in mixture

Vapour pressure of substance i

Mole fraction

Activity coefficient

UNIFAC

Activity emission potential…

Powders

Liquids

Solid objects

Hot / molten metal

Local controls…

• No localized controls• Suppression techniques• Wetting at the point of release• Knockdown suppression

• Containment (non-extracted)• Local ventilation systems• Receiving hoods• Capturing hoods

• Enclosing hoods

Efficacy = 0.0001

Efficacy = 0.5

Dispersion…

Q

FF Q

NF eT

Indoor

Outdoor

Spray booth

Calibration with measurements…

• Using multiple sources• Also including measurements in the low range• Only good quality data: N~2,500• Quantification in units (mg/m3)• Provides model uncertainty• Separate predictions for:• Dust• Vapour• Mist• Fumes

Bayesian updating…

• Bayesian techniques allow for different types of information to be integrated in a theoretically rigorous way via probability theory

• Bayes approach takes account of the relative strength of the different information sources

• Bayesian module is integrated in ART to update model estimate with user’s own data

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

-4 -3 -2 -1 0 1 2 3

Data

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

-4 -3 -2 -1 0 1 2 3

Data Prior

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

0.7

0.75

0.8

-4 -3 -2 -1 0 1 2 3

Data Prior Posterior

ART output…

• Explicit treatment of variability and uncertainty• Percentile of distribution• Uncertainty of estimate

Variability

Uncertainty

90th , 75th , 50th percentile

Inter-quartile, 80 %, 90 %, 95 % CI

Bayesian update

Key scientists involved in ART…

• Erik Tielemans (TNO)• John Cherrie (IOM) • Wouter Fransman (TNO)• Hans Marquart (TNO) • Jody Schinkel (TNO) • Thomas Schneider (NRCWE)• Martin Tischer (BaUA• Martie van Tongeren (IOM)• Nick Warren (HSL)

Acknowledgements…

• S. Bailey (GSK, UK)• D. Brouwer (TNO, The Netherlands)• M. Byrne (University of Galway, Ireland)• D. Dahmann (Bochum University, Germany)• J. Dobbie (BP, UK)• D. Gazzie (Industrial Health Control, UK)• A. Gillies (Gillies Associates Ltd, UK)• E. Joseph (GSK, UK)• I. Kellie (OHS Scotland Ltd, UK)• D. Mark (HSL, UK)• P. McDonnell (University of Galway, Ireland)• M. Newell (Bayer Crop Sciences, UK)• M. Nicas (University of California, USA)• D. O’Malley (Genesis Environmental Ltd, UK)• M. Piney (HSL, UK)• A. Robertson (IOM, UK)• J. Saunders (HSL, UK)• M. Smith (Bayer Crop Sciences, UK)• H. Tinnerberg (Lund University, Sweden)• H. Westberg (Örebro University Hospital, Sweden)• A. Wolley (Wolley Associates, UK)• J. Wren (GSK, UK)• D. Iddon (Lilly, UK)• D. Noy (Dow chemicals, NL)• J. Urbanus (Shell, NL)• R. Battersby (EBRC consulting, Germany)