Validation and Recalibration of the Asbestos Removal Exposure Assessment Tool (AREAT).

OBJECTIVES: The Asbestos Removal Exposure Assessment Tool (AREAT) was previously developed to estimate exposure to respirable asbestos fibres during abatement processes. The current study describes the validation and recalibration of the AREAT model with external data. During model validation, the AREAT model was expanded to be able to estimate asbestos exposure from an additional source category: 'unspecified asbestos remnants'. METHODS: The validation dataset (n = 281) was derived from exposure measurement studies where for each exposure measurement the AREAT model parameters were coded and estimates were calculated. Pearson correlation coefficients (r) and intra class correlation coefficients (icc) were calculated as an indication of the agreement between the AREAT estimates and measured concentrations. In addition, the bias and the proportion of measurements with higher concentrations than model estimates were calculated. To expand and investigate model performance on exposure from 'unspecified asbestos remnants', a separate dataset was created with measurements collected during working with unspecified asbestos remnants, and similar validation comparisons were performed. Lastly, linear regression techniques were used to investigate possible improvements in model parameters. The model was recalibrated on a combined dataset consisting of the validation dataset and the original calibration dataset to increase model robustness. RESULTS: The validation comparisons showed good relative agreement (r) between AREAT estimates and measurements (r = 0.73) and a moderate absolute agreement (icc = 0.53). The overall relative bias was 108%, indicating an overall overestimation of exposure, and 4% of the estimated concentrations were higher than the actual measured concentrations. For the data subset concerning unspecified asbestos remnants, a moderate correlation between model estimates and measurement outcomes was found (r = 0.63). However, based on the low number of data in this subset, and moderate r, it was decided that cleaning of unspecified asbestos remnants is out of scope until more data are available. The results of this validation study suggested that two input parameters (product type friable material, efficacy of control measure foam) underestimated exposure. The effects of these parameters were updated to improve model performance. Compared to the original model, the recalibrated model resulted in slightly higher explained variance (62% compared to 56%) and lower uncertainty (15 compared to 17.3). CONCLUSION: The original AREAT model provided reliable asbestos exposure estimates with a sufficient level of conservatism taking into account the 90-percentile estimates. The model was further improved via the addition of a new feature and recalibration to predict asbestos exposure during the clean-up of unspecified asbestos remnants.

The Dermal Advanced REACH Tool (dART): A Bayesian Model for Dermal Exposure Assessment.

The dermal Advanced REACH Tool (dART) is a tier 2 exposure model for estimating dermal exposure to the hands (mg min-1) for non-volatile liquid and solid-in-liquid products. The dART builds upon the existing ART framework and describes three mass transport processes (deposition (Dhands), direct emission and direct contact (Ehands), and contact transfer (Thands)) that may each contribute to dermal exposure. The mechanistic model that underpins the dART and calibration of the mechanistic model, such that the dimensionless score that results from encoding contextual information about a task into the determinants of the dART can be converted into a prediction of exposure (mg min-1), have been described in previous work. This paper completes the methodological framework of the dART model through placing the mechanistic model within a wider statistical modelling framework. A mixed-effects model, within a Bayesian framework, is presented for modelling the rate of dermal exposure per minute of activity. The central estimate of exposure for a particular task is provided by a calibrated mechanistic model (and thus based upon contextual information about a task). The model also describes between- and within-worker sources of variability in dermal exposure, with prior distributions for variance components based upon the literature. Estimates of exposure based upon informative prior distributions may be updated using measurement data associated with the task. The dART model is demonstrated using three worked examples, where estimates are initially obtained based upon the prior distributions alone, and then refined through accommodating measurement data on the tasks.

The Development and Calibration of a Mechanistic Asbestos Removal Exposure Assessment Tool (AREAT).

Exposure to asbestos fibres is linked to numerous adverse health effects and the use of asbestos is currently banned in many countries. Still, asbestos applications are present in numerous residential and professional/industrial buildings or installations which need to be removed. Exposure measurements give good insight in exposure levels on the basis of which the required control regime is determined to ensure that workers are protected against adverse health effects. However, it is a costly and time-consuming process to measure all situations as working conditions and materials may vary greatly. Therefore, the mechanistic model 'Asbestos Removal Exposure Assessment Tool (AREAT)' was developed to estimate exposure to respirable asbestos fibres released during asbestos abatement processes where measurements are not available. In such instances tailored control regimes can be implemented based on modelled exposure levels. The mechanistic model was developed using scientific literature, an in-house asbestos abatement dataset, and knowledge with regard to previously developed models. Several exposure determinants such as the substance emission potential, activity emission potential, control measures, and dilution in air were identified and specific modifiers were developed for each category. Through an algorithm, AREAT calculates a dimensionless score based on the model inputs. The model was calibrated using a statistical model on an extensive measurement dataset containing a broad variety of exposure scenarios. This statistical model enabled the translation of dimensionless AREAT scores to actual estimated fibre concentrations in fibres m-3. In total, 370 personal inhalation exposure measurements from 71 different studies were used for calibration of AREAT. Of these measurements, in 191 cases (52%) with microscopic analysis (all asbestos fibre analyses were conducted with scanning electron microscopy/energy dispersive X-ray analysis in accordance with ISO 14966) no fibres were detected and the limit of detection values(LODs) were given. To assess the influence of the large number of measurements with exposures below LOD values on the performance of the model, calibrations were performed on the total dataset and the selection of data excluding measurements below LOD. The AREAT model correlated well with the datasets, with a Pearson correlation of 0.73 and 0.8 and Spearman rank correlation of 0.56 and 0.8. The model was fitted to estimate a typical exposure value [i.e. geometric mean (GM) exposures], but it is recommended to use a more conservative worst case higher percentile (for example the 90th percentile; which adds a factor of 17.3 based on the model uncertainty on the GM estimate), to account for variability in the measurements and uncertainty in model estimates. This work has shown the development and calibration of a mechanistic model, capable of estimating asbestos fibre exposures during asbestos abatement processes. The AREAT model will be implemented as a lower tier exposure model in a risk assessment tool used within the Netherlands to plan abatement processes and to develop control strategies.

Calibration of the Dermal Advanced REACH Tool (dART) Mechanistic Model.

The dermal Advanced REACH Tool (dART) is a Tier 2 exposure modelling tool currently in development for estimating dermal exposure to the hands (mg min-1) for non-volatile liquid and solids-in-liquid products. The dART builds upon the existing ART framework and describes three mass transport processes [deposition (Dhands), direct emission and direct contact (Ehands), and contact transfer (Thands)] that may each contribute to dermal exposure. The mechanistic model that underpins the dART and its applicability domain has already been described in previous work. This paper describes the process of calibrating the mechanistic model such that the dimensionless score that results from encoding contextual information about a task into the determinants of the dART can be converted into a prediction of exposure (mg min-1). Furthermore, as a consequence of calibration, the uncertainty in a dART prediction may be quantified via a confidence interval. Thirty-six experimental studies were identified that satisfied the conditions of: (i) high-quality contextual information that was sufficient to confidently code the dART mechanistic model determinants; (ii) reliable exposure measurement data sets were available. From these studies, 40 exposure scenarios were subsequently developed. A non-linear log-normal mixed-effect model was fitted to the data set of Dhands,   Ehands, and    Thands scores and corresponding measurement data. The dART model was shown to be consistent with activities covering a broad range of tasks [spray applications, activities involving open liquid surfaces (e.g. dipping, mixing), handling of contaminated objects, spreading of liquid products, and transfer of products (e.g. pouring of liquid)]. Exposures resulting from a particular task were each dominated by one or two of the identified mass transport processes. As a consequence of calibration, an estimate of the uncertainty associated with a mechanistic model estimate is available. A 90% multiplicative interval is approximately a factor of six. This represents poorer overall precision than the (inhalation) ART model for dusts and vapours, although better than the ART model for mists. Considering the complexity of the conceptual model compared with the ART, the wide variety of exposure scenarios considered with differing dominant routes, and the particular challenges that result from the consideration of measurement data both above and beneath a protective glove, the precision of the calibrated dART mechanistic model is reasonable for well-documented exposure scenarios coded by experts. However, as the inputs to the model are based upon user judgement, in practical use, the reliability of predictions will be dependent upon both the competence of users and the quality of contextual information available on an exposure scenario.

Dermal Advanced REACH Tool (dART)-Development of a Dermal Exposure Model for Low-Volatile Liquids.

This article describes the development of a mechanistic model for underpinning the dermal Advanced REACH Tool (dART), an extension of the existing ART model and its software platform. It was developed for hand exposure to low volatile liquids (vapour pressure ≤ 10 Pa at 20°C) including solids-in-liquid products. The model is based on an existing conceptual dermal source-receptor model that has been integrated into the ART framework. A structured taxonomy of workplace activities referred to as activity classes are adopted from ART. Three key processes involved in mass transport associated with dermal exposure are applied, i.e. deposition, direct emission and contact, and transfer. For deposition, the model adopts all the relevant modifying factors (MFs) applied in ART. In terms of direct emission and contact (e.g. splashes) and transfer (e.g. hand-surface contacts), the model defines independent principal MFs, i.e. substance-related factors, activity-related factors, localized- and dispersion control and exposed surface area of the hands. To address event-based exposures as much as possible, the model includes crucial events during an activity (e.g. hand immersions) and translates objective information on tools and equipment (manual or automated) to probable events (e.g. splashes) and worker behaviours (e.g. surface contacts). Based on an extensive review of peer-reviewed literature and unpublished field studies, multipliers were assigned to each determinant and provide an approximated (dimensionless) numerical value. In the absence of (sufficient) evidence, multipliers were assigned to determinants based on assumptions made during discussions by experts in the consortium. A worked example is presented to illustrate the calculation of hand exposure for a specific scenario. The dART model is not yet implemented in the ART software platform, and a robust validation of the model is necessary to determine its predictive ability. With advancing knowledge on dermal exposure and its determinants, this model will require periodic updates and refinements, in addition to further expansion of the applicability domain of the model.

Comparing the Advanced REACH Tool's (ART) Estimates With Switzerland's Occupational Exposure Data.

The Advanced REACH Tool (ART) is the most sophisticated tool used for evaluating exposure levels under the European Union's Registration, Evaluation, Authorisation and restriction of CHemicals (REACH) regulations. ART provides estimates at different percentiles of exposure and within different confidence intervals (CIs). However, its performance has only been tested on a limited number of exposure data. The present study compares ART's estimates with exposure measurements collected over many years in Switzerland. Measurements from 584 cases of exposure to vapours, mists, powders, and abrasive dusts (wood/stone and metal) were extracted from a Swiss database. The corresponding exposures at the 50th and 90th percentiles were calculated in ART. To characterize the model's performance, the 90% CI of the estimates was considered. ART's performance at the 50th percentile was only found to be insufficiently conservative with regard to exposure to wood/stone dusts, whereas the 90th percentile showed sufficient conservatism for all the types of exposure processed. However, a trend was observed with the residuals, where ART overestimated lower exposures and underestimated higher ones. The median was more precise, however, and the majority (≥60%) of real-world measurements were within a factor of 10 from ART's estimates. We provide recommendations based on the results and suggest further, more comprehensive, investigations.

Validation of the dermal exposure model in ECETOC TRA.

The ECETOC TRA model (presently version 3.1) is often used to estimate worker inhalation and dermal exposure in regulatory risk assessment. The dermal model in ECETOC TRA has not yet been validated by comparison with independent measured exposure levels. This was the goal of the present study. Measured exposure levels and relevant contextual information were gathered via literature search, websites of relevant occupational health institutes and direct requests for data to industry. Exposure data were clustered in so-called exposure cases, which are sets of data from one data source that are expected to have the same values for input parameters in the ECETOC TRA dermal exposure model. For each exposure case, the 75th percentile of measured values was calculated, because the model intends to estimate these values. The input values for the parameters in ECETOC TRA were assigned by an expert elicitation and consensus building process, based on descriptions of relevant contextual information.From more than 35 data sources, 106 useful exposure cases were derived, that were used for direct comparison with the model estimates. The exposure cases covered a large part of the ECETOC TRA dermal exposure model. The model explained 37% of the variance in the 75th percentiles of measured values. In around 80% of the exposure cases, the model estimate was higher than the 75th percentile of measured values. In the remaining exposure cases, the model estimate may not be sufficiently conservative.The model was shown to have a clear bias towards (severe) overestimation of dermal exposure at low measured exposure values, while all cases of apparent underestimation by the ECETOC TRA dermal exposure model occurred at high measured exposure values. This can be partly explained by a built-in bias in the effect of concentration of substance in product used, duration of exposure and the use of protective gloves in the model. The effect of protective gloves was calculated to be on average a factor of 34 in this data set, while factors of five to ten were used in the model estimations. There was also an effect of the sampling method in the measured data on the exposure levels. Exposure cases where sampling was done via an interception method, such as gloves, on average showed a factor of six higher 75th percentiles of measured values than exposure cases where sampling was done via a removal method, such as hand washing. This may partly be responsible for the apparent underestimation of dermal exposure by the model at high exposure values. However, there also appeared to be a relation between expected exposure level (as indicated by the model estimate) and the choice of sampling method.In this study, solid substances used in liquid products were treated as liquids with negligible volatility. The results indicate that the ECETOC TRA dermal exposure model performs equally well for these substances as for liquids. There were suggestions of a difference in performance of the model between solids and liquids.For several parts of the ECETOC TRA dermal model, no or hardly any measured dermal exposure data were available. Therefore, gathering of more dermal exposure levels is recommended, specifically for situations not yet sufficiently covered in the present data set.

Advanced REACH Tool: a Bayesian model for occupational exposure assessment.

This paper describes a Bayesian model for the assessment of inhalation exposures in an occupational setting; the methodology underpins a freely available web-based application for exposure assessment, the Advanced REACH Tool (ART). The ART is a higher tier exposure tool that combines disparate sources of information within a Bayesian statistical framework. The information is obtained from expert knowledge expressed in a calibrated mechanistic model of exposure assessment, data on inter- and intra-individual variability in exposures from the literature, and context-specific exposure measurements. The ART provides central estimates and credible intervals for different percentiles of the exposure distribution, for full-shift and long-term average exposures. The ART can produce exposure estimates in the absence of measurements, but the precision of the estimates improves as more data become available. The methodology presented in this paper is able to utilize partially analogous data, a novel approach designed to make efficient use of a sparsely populated measurement database although some additional research is still required before practical implementation. The methodology is demonstrated using two worked examples: an exposure to copper pyrithione in the spraying of antifouling paints and an exposure to ethyl acetate in shoe repair.

Reliability of the Advanced REACH Tool (ART).

OBJECTIVES: The aim of this study was to assess the reliability of the Advanced REACH Tool (ART) by (i) studying interassessor agreement of the resulting exposure estimates generated by the ART mechanistic model, (ii) studying interassessor agreement per model parameters of the ART mechanistic model, (iii) investigating assessor characteristics resulting in reliable estimates, and (iv) estimating the effect of training on assessor agreement. METHODS: Prior to the 1-day workshop, participants had to assess four scenarios with the ART. During two 1-day workshops, 54 participants received 3-h training in applying the mechanistic model and the technical aspects of the web tool. Afterward, the participants assessed another four scenarios. The assessments of the participants were compared with gold standard estimates compiled by the workshop instructors. Intraclass correlation coefficients (ICCs) were calculated and per model parameter and the percentage agreement and Cohen kappa statistics were estimated. RESULTS: The ICCs showed good agreement before and almost perfect agreement after training. However, substantial variability was observed between individual assessors' estimates for an individual scenario. After training, only 42% of the assessments lay within a factor of three of the gold standard estimate. The reliability appeared to be influenced by several factors: (i) information provided by text and video hampered the assessors gaining additional information required to make the assessments, (ii) for some parameters, the guidance documentation implemented in the tool may have been insufficient, and (iii) in some cases, the assessors were not able to implement the information explicitly provided. CONCLUSIONS: The ART is an expert tool and extensive training is recommended prior to use. Improvements of the guidance documentation, consensus procedures, and improving the training methods could improve the reliability of ART. Nevertheless, considerable variability can be expected between assessors using ART to estimate exposure levels for a given scenario.

The Advanced REACH Tool (ART): incorporation of an exposure measurement database.

This article describes the structure, functionalities, and content of the Advanced REACH Tool (ART) exposure database (version 1.5). The incorporation of the exposure database into ART allows users who do not have their own measurement data for their exposure scenario, to update the exposure estimates produced by the mechanistic model using analogous measurement series selected from the ART exposure measurement database. Depending on user input for substance category and activity (sub)classes, the system selects exposure measurement series from the exposure database. The comprehensive scenario descriptions and summary statistics assist the user in deciding if the measurement series are indeed fully analogous. After selecting one or more analogous data sets, the data are used by the Bayesian module of the ART system to update the mechanistically modeled exposure estimates. The 1944 exposure measurements currently stored in the ART exposure measurement database cover 9 exposure situations for handling solid objects (n = 65), 42 situations for handling powders, granules, or pelletized material (n = 488), 5 situations for handling low-volatility liquids (n = 88), 35 situations for handling volatile liquids (n = 870), and 26 situations for handling liquids in which powders are dissolved or dispersed (resulting in exposure to mist) (n = 433). These 117 measurement series form a good basis for supporting user exposure estimates. However, by increasing the diversity of exposure situations and the number of measurement series in the database, the usefulness of the ART system will be further improved. Suggestions to stimulate the process of sharing exposure measurement data both to increase the available data in the ART and for other purposes are made.

Use of the MEGA exposure database for the validation of the Stoffenmanager model.

OBJECTIVES: This paper explores the usefulness of the exposure database MEGA for model validation and evaluates the capability of two Stoffenmanager model equations (i.e. handling of powders/granules and machining) to estimate workers exposure to inhalable dust. METHODS: For the task groups, 'handling of powders and granules' (handling) and 'machining of wood and stone' (machining) measurements were selected from MEGA and grouped in scenarios depending on task, product, and control measures. The predictive capability of the model was tested by calculating the relative bias of the single measurements and the correlation between geometric means (GMs) for scenarios. The conservatism of the model was evaluated by checking if the percentage of measurement values above the 90th percentile estimate was ≤10%. RESULTS: From 22 596 personal measurements on inhalable dust within MEGA, 390 could be selected for handling and 1133 for machining. The relative bias for the task groups was -25 and 68%, respectively, the percentage of measurements with a higher result than the estimated 90th percentile 11 and 7%. Correlations on a scenario level were good for both model equations as well for the GM (handling: r(s) = 0.90, n = 15 scenarios; machining: r(s) = 0.84, n = 22 scenarios) as for the 90th percentile (handling: r(s) = 0.79; machining: r(s) = 0.76). CONCLUSIONS: The MEGA database could be used for model validation, although the presented analyses have learned that improvements in the database are necessary for modelling purposes in the future. For a substantial amount of data, contextual information on exposure determinants in addition to basic core information is stored in this database. The relative low bias, the good correlation, and the level of conservatism of the tested model show that the Stoffenmanager can be regarded as a useful Tier 1 model for the Registration, Evaluation, Authorisation and Restriction of Chemicals legislation.

Advanced REACH Tool (ART): overview of version 1.0 and research needs.

This paper provides an outline of the Advanced REACH Tool (ART) version 1.0 and a discussion of how it could be further developed. ART is a higher tier exposure assessment tool that combines mechanistically modelled inhalation exposure predictions with available exposure data using a Bayesian approach. ART assesses exposure for scenarios across different plants and sites. Estimates are provided for different percentiles of the exposure distribution and confidence intervals around the estimate. It also produces exposure estimates in the absence of data, but uncertainty of the estimates will decrease when results of exposure measurements are included. The tool has been calibrated using a broad range of exposure data and provides estimates for exposure to vapours, mists, and dusts. ART has a robust and stable conceptual basis but will be refined in the future and should therefore be considered an evolving system. High-priority areas for future research are identified in this paper and include the integration of partially analogous measurement series, inclusion of company and site-specific assessments, user decision strategies linked to ART predictions, evaluation of validity and reliability of ART, exploring the possibilities for incorporating the dermal route and integration of ART predictions with tools for modelling internal dose. ART is initially developed in the scope of REACH but is equally useful for exposure assessment in other areas.

Advanced REACH Tool: development and application of the substance emission potential modifying factor.

The Advanced REACH Tool (ART) is an exposure assessment tool that combines mechanistically modelled inhalation exposure estimates with available exposure data using a Bayesian approach. The mechanistic model is based on nine independent principal modifying factors (MF). One of these MF is the substance emission potential, which addresses the intrinsic substance properties as determinants of the emission from a source. This paper describes the current knowledge and evidence on intrinsic characteristics of solids and liquids that determine the potential for their release into workplace air. The principal factor determining the release of aerosols from handling or processing powdered, granular, or pelletized materials is the dustiness of the material, as well as the weight fraction of the substance of interest in the powder and the moisture content. The partial vapour pressure is the main intrinsic factor determining the substance emission potential for emission of vapours. For generation of mist, the substance emission potential is determined by the viscosity of the liquid as well as the weight fraction of the substance of interest in the liquid. Within ART release of vapours is considered for substances with a partial vapour pressure at the process temperature of 10 Pa or more, while mist formation is considered for substances with a vapour pressure ≤ 10 Pa. Relative multipliers are assigned for most of the intrinsic factors, with the exception of the weight fraction and the vapour pressure, which is applied as a continuous variable in the estimation of the substance emission potential. Currently, estimation of substance emission potential is not available for fumes, fibres, and gases. The substance emission potential takes account of the latest thinking on emissions of dusts, mists, and vapours and in our view provides a good balance between theory and pragmatism. Expanding the knowledge base on substance emission potential will improve the predictive power of occupational exposure models and thereby the accuracy and precision of the exposure estimates.

Advanced Reach Tool (ART): development of the mechanistic model.

This paper describes the development of the mechanistic model within a collaborative project, referred to as the Advanced REACH Tool (ART) project, to develop a tool to model inhalation exposure for workers sharing similar operational conditions across different industries and locations in Europe. The ART mechanistic model is based on a conceptual framework that adopts a source receptor approach, which describes the transport of a contaminant from the source to the receptor and defines seven independent principal modifying factors: substance emission potential, activity emission potential, localized controls, segregation, personal enclosure, surface contamination, and dispersion. ART currently differentiates between three different exposure types: vapours, mists, and dust (fumes, fibres, and gases are presently excluded). Various sources were used to assign numerical values to the multipliers to each modifying factor. The evidence used to underpin this assessment procedure was based on chemical and physical laws. In addition, empirical data obtained from literature were used. Where this was not possible, expert elicitation was applied for the assessment procedure. Multipliers for all modifying factors were peer reviewed by leading experts from industry, research institutes, and public authorities across the globe. In addition, several workshops with experts were organized to discuss the proposed exposure multipliers. The mechanistic model is a central part of the ART tool and with advancing knowledge on exposure, determinants will require updates and refinements on a continuous basis, such as the effect of worker behaviour on personal exposure, 'best practice' values that describe the maximum achievable effectiveness of control measures, the intrinsic emission potential of various solid objects (e.g. metal, glass, plastics, etc.), and extending the applicability domain to certain types of exposures (e.g. gas, fume, and fibre exposure).

Classification of occupational activities for assessment of inhalation exposure.

There is a large variety of activities in workplaces that can lead to emission of substances. Coding systems based on determinants of emission have so far not been developed. In this paper, a system of Activity Classes and Activity Subclasses is proposed for categorizing activities involving chemical use. Activity Classes share their so-called 'emission generation mechanisms' and physical state of the product handled and the underlying determinants of emission. A number of (industrial) stakeholders actively participated in testing and fine-tuning the system. With the help of these stakeholders, it was found to be relatively easy to allocate a large number of activities to the Activity Classes and Activity Subclasses. The system facilitates a more structured classification of activities in exposure databases, a structured analysis of the analogy of exposure activities, and a transparent quantification of the activity emission potential in (new) exposure assessment models. The first use of the system is in the Advanced REACH Tool.

Validation of the inhalable dust algorithm of the Advanced REACH Tool using a dataset from the pharmaceutical industry.

As it is often difficult to obtain sufficient numbers of measurements to adequately characterise exposure levels, occupational exposure models may be useful tools in the exposure assessment process. This study aims to refine and validate the inhalable dust algorithm of the Advanced REACH Tool (ART) to predict airborne exposure of workers in the pharmaceutical industry. The ART was refined to reflect pharmaceutical situations. Largely task based workplace exposure data (n = 192) were collated from a multinational pharmaceutical company with exposure levels ranging from 5 × 10(-5) to 12 mg m(-3). Bias, relative bias and uncertainty around geometric mean exposure estimates were calculated for 16 exposure scenarios. For 12 of the 16 scenarios the ART geometric mean exposure estimates were lower than measured exposure levels with on average, a one-third underestimation of exposure (relative bias -32%). For 75% of the scenarios the exposure estimates were, within the 90% uncertainty factor of 4.4, as reported for the original calibration study, which may indicate more uncertainty in the ART estimates in this industry. While the uncertainty was higher than expected this is likely due to the limited number of measurements per scenario, which were largely derived from single premises.

Advanced REACH Tool (ART): calibration of the mechanistic model.

The mechanistic model of the Advanced Reach Tool (ART) provides a relative ranking of exposure levels from different scenarios. The objectives of the calibration described in this paper are threefold: to study whether the mechanistic model scores are accurately ranked in relation to exposure measurements; to enable the mechanistic model to estimate actual exposure levels rather than relative scores; and to provide a method of quantifying model uncertainty. Stringent data quality guidelines were applied to the collated data. Linear mixed effects models were used to evaluate the association between relative ART model scores and measurements. A random scenario and company component of variance were introduced to reflect the model uncertainty. Stratified analyses were conducted for different forms of exposure (abrasive dust, dust, vapours and mists). In total more than 2000 good quality measurements were available for the calibration of the mechanistic model. The calibration showed that after calibration the mechanistic model of ART was able to estimate geometric mean (GM) exposure levels with 90% confidence for a given scenario to lie within a factor between two and six of the measured GM depending upon the form of exposure.

Stoffenmanager exposure model: company-specific exposure assessments using a Bayesian methodology.

The web-based tool "Stoffenmanager" was initially developed to assist small- and medium-sized enterprises in the Netherlands to make qualitative risk assessments and to provide advice on control at the workplace. The tool uses a mechanistic model to arrive at a "Stoffenmanager score" for exposure. In a recent study it was shown that variability in exposure measurements given a certain Stoffenmanager score is still substantial. This article discusses an extension to the tool that uses a Bayesian methodology for quantitative workplace/scenario-specific exposure assessment. This methodology allows for real exposure data observed in the company of interest to be combined with the prior estimate (based on the Stoffenmanager model). The output of the tool is a company-specific assessment of exposure levels for a scenario for which data is available. The Bayesian approach provides a transparent way of synthesizing different types of information and is especially preferred in situations where available data is sparse, as is often the case in small- and medium sized-enterprises. Real-world examples as well as simulation studies were used to assess how different parameters such as sample size, difference between prior and data, uncertainty in prior, and variance in the data affect the eventual posterior distribution of a Bayesian exposure assessment.

Cross-validation and refinement of the Stoffenmanager as a first tier exposure assessment tool for REACH.

OBJECTIVES: For regulatory risk assessment under REACH a tiered approach is proposed in which the first tier models should provide a conservative exposure estimate that can discriminate between scenarios which are of concern and those which are not. The Stoffenmanager is mentioned as a first tier approach in the REACH guidance. In an attempt to investigate the validity of the Stoffenmanager algorithms, a cross-validation study was performed. METHODS: Exposure estimates using the Stoffenmanager algorithms were compared with exposure measurement results (n=254). Correlations between observed and predicted exposures, bias and precision were calculated. Stratified analyses were performed for the scenarios "handling of powders and granules" (n=82), "handling solids resulting in comminuting" (n=60), "handling of low-volatile liquids" (n=40) and "handling of volatile liquids" (n=72). RESULTS: The relative bias of the four algorithms ranged between -9% and -77% with a precision of approximately 1.7. The 90th percentile estimate of one out of four algorithms was not conservative enough. Based on these statistics and analyses of residual plots the underlying algorithm was adapted. Subsequently, the calibration and the cross-validation dataset were merged into one dataset (n=952) used for calibrating the adapted Stoffenmanager algorithms. This new calibration resulted in new exposure algorithms for the four scenarios. CONCLUSIONS: The Stoffenmanager is capable of discriminating among exposure levels mainly between scenarios in different companies. The 90th percentile estimates of the Stoffenmanager are verified to be sufficiently conservative. Therefore, the Stoffenmanager could be a useful tier 1 exposure assessment tool for REACH.

Development and evaluation of an exposure control efficacy library (ECEL).

OBJECTIVES: This paper describes the development and evaluation of an evidence database on the effectiveness of risk management measures (RMMs) to control inhalation exposure. This database is referred to as Exposure Control Efficacy Library (ECEL). METHODS: A comprehensive review of scientific journals in the occupational hygiene field was undertaken. Efficacy values for RMMs in conjunction with contextual information on study design, sampling strategy and measurement type (among other parameters) were stored in an MS Access database. In total, 433 efficacy values for six RMM groups (i.e. enclosure, local exhaust ventilation, specialized ventilation, general ventilation, suppression techniques and separation of the worker) were collected from 90 peer-reviewed publications. These RMM categories were subdivided into more specific categories. RESULTS: Estimated average efficacy values ranged from 87% for specialized ventilation to 43% for general ventilation. Substantial variation in efficacy values was observed within RMM categories based on differences in selected covariables within each study (i.e. study design, sampling strategy, measurement type and others). More contrast in efficacy values was observed when evaluating more detailed subcategories. CONCLUSIONS: It is envisaged that ECEL will contribute to exposure modelling, but should be supplemented with expert opinion, preferably in a formal expert elicitation procedure. The work presented here should be considered as a first attempt to collate and analyse RMM efficacy values and inclusion of additional (unpublished) exposure data is highly warranted.