A novel strategy for retrospective exposure assessment in the Norwegian silicon carbide industry

Our plain-English summary of A novel strategy for retrospective exposure assessment in the Norwegian silicon carbide industry, by S. Føreland, M.D. Bugge, B. Bakke, et al (Journal of Occupational and Environmental Hygiene, 2012).

The study

• Authors: S. Føreland, M.D. Bugge, B. Bakke, E. Bye, W. Eduard
• Journal: Journal of Occupational and Environmental Hygiene
• Published: 2012;9(4):230-41
• Cited by: 9 articles, as counted by NIH iCite
• Read it: PubMed · Publisher

Abstract

The objective was to construct a retrospective job-exposure matrix (JEM) for the Norwegian silicon carbide industry. More than 3300 historical total dust measurements were available (1967-2005); however, there were few measurements of other agents. Total dust measurements were therefore used as the basis for the JEM, and a novel method was developed to estimate exposure to other agents. Multiple linear regression models were developed to describe historical exposure to total dust. Exposure estimates were extrapolated backward to periods without exposure data by adjustments for process and work-hour related changes. An exposure assessment study was performed where total dust was sampled in parallel with fibers or respirable dust. The respirable dust was analyzed for the content of quartz, cristobalite, and silicon carbide. Mixed-effect models were developed to estimate the exposure to these agents from total dust exposure, plant, and job group. Exposure to asbestos and polycyclic aromatic hydrocarbons was assigned qualitatively. Multiple linear regression models of total dust described historical exposure best in the furnace department (R(2) (adj) = 0.49-0.74). Models in the other departments explained less variance (R(2) (adj) = 0.12-0.32). Exposure determinants and total dust explained a substantial proportion of the between- (70-100%) and within-worker (8.0-54%) variance in the mixed-effect models. The relative bias between available historical measurements and the estimated exposure to dust components varied between -39% (fiber) and 40% (quartz). However, corrections were not considered necessary due to limitations in the historical data. The component-specific metrices were sufficiently different from each other (r(Pearson) < 0.7), with the exception of total and respirable dust (r(Pearson) = 0.84) and total dust and cristobalite (r(Pearson) = 0.72), and will enable component-specific epidemiologic analyses in the future. Improved and less correlated estimates of exposure levels for the different components in the dust were obtained with the updated exposure assessment. Due to limitations in the measurement data, assumptions had to be made, especially in the period before 1967. [Supplementary materials are available for this article. Go to the publisher's online edition of Journal of Occupational and Environmental Hygiene for the following free supplemental resource: a file containing tables outlining multiple linear regression models for prediction of total dust exposure in the processing departments of Norwegian SiC producing plants, evaluation of the predictive abilities of the reduced total dust models, and mixed models for pedicting exposure to fibers and respirable quartz, cristobalite, non-fibrous silicon carbide and respirable dust from total dust exposure.].

In plain English

The authors built a job-exposure matrix, estimating past exposure by job and period, for three Norwegian silicon carbide plants. It drew on 3345 historical total dust measurements but few of other agents. A 2002–2003 study sampled total dust alongside fibers or respirable dust. Models then estimated fibers, silicon carbide and respirable dust from total dust, plant and job group. Quartz and cristobalite were modelled only for furnace jobs; elsewhere, fixed percentages were applied to respirable dust estimates. No historical asbestos exposure measurements were available. Jobs were classed as exposed or unexposed by task. Maintenance personnel, control room and charger operators were classed as exposed when asbestos was used. Against historical measurements, relative bias was –39% for modelled fibers and 40% for quartz. The earlier matrix's estimates averaged 18 times higher for silicon carbide and 36 times higher for crystalline silica. The authors attributed most differences to the earlier matrix mostly using total dust rather than respirable dust. The authors concluded that component estimates were considerably less correlated than in the earlier matrix. They noted large uncertainties, especially before 1967, and called the asbestos classification crude.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Retrospective exposure assessment: construction of a job-exposure matrix using statistical models of historical measurements plus a comparative parallel-sampling study
• Who was studied: Workplace air measurements from three Norwegian silicon carbide plants (Plants A, B and C). These were 3345 total dust, 224 crystalline silica, 89 fiber and 16 PAH personal measurements from 1967 to 2005, plus parallel measurements taken in 2002–2003. The matrix was intended for an epidemiologic study of the workers.
• Exposure: Total dust, respirable dust, airborne fibers, respirable quartz, cristobalite and silicon carbide, all estimated quantitatively. Asbestos and polycyclic aromatic hydrocarbons (PAH) were assigned qualitatively. Asbestos tasks included cutting, removing and replacing asbestos plates in the furnaces, plugging with asbestos between furnace elements, and maintenance on machinery, pipes and insulation.
• Main results:
  • No historical asbestos exposure measurements were available. Maintenance personnel, control room operators and charger operators were classified as exposed to asbestos in the time periods when asbestos was used.
  • Quartz and cristobalite were modelled with mixed-effect models only for furnace department jobs. In the processing and maintenance departments, 60–93% of measurements were below the limit of detection. For those departments, fixed percentages were applied to respirable dust estimates instead.
  • Total dust models explained the most variance in the furnace departments, with adjusted R-squared from 0.49 in Plant A to 0.74 in Plant C. They explained the least in the maintenance departments, from 0.12 in Plant A to 0.30 in Plant C.
  • According to the models, total dust exposure declined over time.
  • In split-sample testing of the total dust models, the mean relative bias was 0.76%, ranging from –12% to 12%.
  • In the mixed models for dust components, exposure determinants explained 8.0–54% of the within-worker variance and 70–100% of the between-worker variance.
  • Compared with historical measurements, the relative bias was –39% for the fiber model, 40% for the quartz model and 1% for the cristobalite model. These comparisons were based on 125 to 158 historical measurements.
  • Compared with the earlier matrix by Romundstad et al., total dust estimates in this study were on average 1.5 times higher. The earlier matrix's estimates were on average 18 times higher for silicon carbide, 36 times higher for crystalline silica and 3.2 times higher for fibers.
  • The authors attributed most of the differences from the earlier matrix to method. The earlier matrix mostly based its dust-component estimates on total dust, whereas this study analyzed the respirable dust fraction.
  • Correlations between exposure estimates for the different dust components were moderate to low. The exception was total dust and respirable dust (Pearson r = 0.84). Total dust and cristobalite had a Pearson r of 0.72.
• The authors' conclusion: The authors concluded that large uncertainties remained in the estimates, especially before 1967 and for job groups without measurements. They also concluded that the new matrix's estimates for different components were considerably less correlated than in the previous matrix. They expected this to help future epidemiologic analyses identify possible causal agents.
• Limitations the authors state:
  • Asbestos was classified only as exposed or unexposed, with no measurements. The authors called this a crude description and said misclassification was possible. Maintenance personnel may also have had sporadic asbestos exposure after its use was prohibited.
  • Worker identification was available for less than 40% of the total dust samples. Repeated measurements on the same worker therefore could not be accounted for, so some bias is possible.
  • Other factors affecting dust composition may not have been accounted for, particularly in the earliest periods.
  • Historical crystalline silica had been measured in different size fractions. This made comparison with the modelled estimates difficult and uncertain.
  • Large uncertainties remained, especially before 1967 and for job groups without measurements.

Related studies

• Toxicity in vitro of some silicon carbides and silicon nitrides: whiskers and powders — I. Svensson, E. Artursson, et al. (1997)
• Histopathological changes in rat lung following intratracheal instillation of silicon carbide whiskers and potassium octatitanate whiskers — A. Ogami, Y. Morimoto, et al. (2007)
• Expression of Clara cell secretory protein in the lungs of rats exposed to silicon carbide whisker in vivo — Y. Morimoto, L. Ding, et al. (2003)
• Carcinogenicity of fluoro-edenite, silicon carbide fibres and whiskers, and carbon nanotubes — Y. Grosse, D. Loomis, et al. (2014)
• Pulmonary response, in vivo, to silicon carbide whiskers — G.L. Vaughan, S.A. Trently, et al. (1993)
• In vitro activity of silicon carbide whiskers in comparison to other industrial fibers using four cell culture systems — N.F. Johnson, M.D. Hoover, et al. (1992)

Browse more research in Journal of Occupational and Environmental Hygiene, or return to medical research.