Assessing cancer hazards of bitumen emissions a case study for complex petroleum substances

Transcription

1 Critical Reviews in Toxicology ISSN: (Print) (Online) Journal homepage: Assessing cancer hazards of bitumen emissions a case study for complex petroleum substances Anthony J. Kriech, Ceinwen A. Schreiner, Linda V. Osborn & Anthony J. Riley To cite this article: Anthony J. Kriech, Ceinwen A. Schreiner, Linda V. Osborn & Anthony J. Riley (2018) Assessing cancer hazards of bitumen emissions a case study for complex petroleum substances, Critical Reviews in Toxicology, 48:2, , DOI: / To link to this article: The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. Published online: 02 Nov Submit your article to this journal Article views: 919 View Crossmark data Citing articles: 1 View citing articles Full Terms & Conditions of access and use can be found at

2 CRITICAL REVIEWS IN TOXICOLOGY, 2018 VOL. 48, NO. 2, REVIEW ARTICLE Assessing cancer hazards of bitumen emissions a case study for complex petroleum substances Anthony J. Kriech a, Ceinwen A. Schreiner b, Linda V. Osborn a and Anthony J. Riley c a Heritage Research Group, Indianapolis, IN, USA; b C&C Consulting in Toxicology, Meadowbrook, PA, USA; c Independent Consulting, Guilldford, Surrey, UK ABSTRACT When assessing cancer hazard and risk associated with a complex petroleum substance, like bitumen emissions, there are often conflicting results related to human, animal and mechanistic studies. Validation of the complex composition to assure that it matches real-world exposures and control of confounders are pivotal factors in study design to allow the necessary read-across during assessments. Several key studies on bitumen emissions in two-year dermal cancer assays reported variable outcomes ranging from high cancer incidence to no cancer incidence. Here, we synthesize findings from published studies to explain the differences and discuss critical factors in cancer hazard evaluation for complex petroleum substances. Using these critical factors, we reviewed relevant human genetic toxicity, mammalian toxicity and mechanistic studies with bitumen to understand the divergence in results. We assess the most reliable and scientifically supported information on the potential carcinogenic hazards of bitumen emissions and comment on quality and completeness of data. Human hazard data are typically considered highest priority because they eliminate the need for interspecies extrapolation and reduce the range of high -to low-dose extrapolation during the risk assessment process. Finally, two well-conducted comprehensive animal studies are discussed that have well-defined test material, exposure concentration and composition representative of worker exposure, evidence of systemic uptake, no confounding exposures and provide consistency across all elements within both studies. Studies that allow effective read-across from human, animal and mechanistic components, control for confounders and are well-validated analytically against workplace exposures, provide the strongest evidence base for evaluating cancer hazard. ARTICLE HISTORY Received 31 July 2017 Revised 9 October 2017 Accepted 9 October 2017 KEYWORDS Asphalt; fumes; human; mammalian; dermal; inhalation; mechanistic; mutagenicity; risk assessment; genetic toxicity; read-across Table of contents Introduction Complex petroleum substances Critical factors in cancer hazard evaluation for complex petroleum substances Data quality Test material composition Experimental exposure conditions Evidence of exposure and uptake Data consistency Confounding exposures Bitumen human genetic studies DNA adducts and DNA strand breaks Cytogenetic studies Oxidative damage Experimental genetic toxicity (in vitro and in vivo mammalian) Validated animal cancer studies Case study 1 evaluating carcinogenic potential of bitumen emissions following inhalation exposure Case study 2 evaluating the carcinogenic potential of condensed bitumen emissions following dermal exposure Mechanistic studies Human studies Discussion and conclusions Acknowledgements Declaration of interest Supplemental material References Introduction Evaluating the potential health effects caused by simultaneous exposure to many constituents in a complex petroleum substance is of considerable importance, but fraught with challenges, for example, ensuring that experimental CONTACT Linda V. Osborn linda.osborn@hrglab.com Heritage Research Group, Indianapolis, IN, USA ß 2017 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License ( 4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.

3 122 A. J. KRIECH ET AL. exposures are representative of actual exposure in the workplace. We examine experimental studies with bitumen and bitumen emissions as examples of this challenge. Several organizations have reviewed the health aspects of bitumen and bitumen emissions including International Agency for Research On Cancer (IARC), National Institute for Occupational Safety and Health (NIOSH), German Committee for the Determination of Occupational Exposure Limits (MAK Commission), American Conference of Governmental Industrial Hygienists, Inc. (ACGIH), Internal Program on Chemical Safety (IPCS) and Institut National de la Recherche Scientifique or National Institute of Scientific Research (INRS). It is especially challenging to assess the carcinogenic potential of these materials due to the complexity of their chemical composition and the possible presence of small amounts (parts per million) of potentially carcinogenic/genotoxic constituents. Some organizations, notably IARC, have attempted to incorporate mechanistic and genotoxicity data into their hazard evaluation. However, the lack of a standardized approach to evaluating mechanistic data, and its significance for cancer hazard evaluation, has resulted in some criticism and recommendations for a more systematic and transparent approach (Goodman and Lynch 2017). Results of mammalian studies vary broadly from benign (Fuhst et al. 2007; Clark et al. 2011) to carcinogenic (Sivak et al. 1997). With studies showing such vastly different outcomes, it is often difficult to reach sound scientific conclusions on the health effects of a complex substance without an in-depth understanding of the material, how it behaves under different conditions and what factors can influence exposure. Here, we look at how carcinogenicity and genotoxicity data on complex substances could be evaluated and reviewed to better explain the outcomes in the face of apparently conflicting data. We began with published large-scale hazard reviews, considered the individual papers cited within those reviews and expanded our assessment to include new additions to the literature. When assessing hazard, it is important to look for research that includes human, animal and mechanistic studies conducted in well-validated, linked investigations. This allows read-across from one approach to another to better understand the mechanisms of health outcomes. It is important in assessing such studies that proper analytical validation of test material composition against workplace exposure samples is conducted. This ensures that hazard evaluations studies reflect true exposure potential. Studies which lack validation need to be reviewed separately to determine how the composition and biological activity of complex substances can change with end use, such as temperature or environmental conditions. Our review is focused on experimental data and therefore excludes review of the extensive dataset of epidemiological studies; primary outcomes have been published elsewhere (Boffetta et al. 2003; Rhomberg et al. 2015). Boffetta et al. concluded that it was not possible to confirm a causal link between exposure to bitumen emissions and cancer of the lung, head, and neck. Furthermore, no association was identified between exposure to bitumen emissions and any other neoplasm. Rhomberg et al. (2015) have used various methods illustrating that cancer risks to roofers from dermal and inhalation exposure to built-up roofing asphalt (BURA) are within acceptable risk thresholds, concluding that the risk is extremely low. Complex petroleum substances Bitumen is a complex substance, derived from crude oil, that contains tens of thousands of different individual constituents, many of which are likely to present different physicochemical and biological properties (Asphalt Institute, Eurobitume 2015). When bitumen is heated to liquefy the product, it can release low parts per million (<10 ppm) levels of volatile and semivolatile compounds into the atmosphere in the form of aerosols, gases and particulates, to which workers can be exposed. These compounds are initially trapped in the bitumen because of the incomplete distillation of crude oils. Where bitumen is cut-back or fluxed with lower boiling hydrocarbons, additional volatile and semivolatile compounds may be released upon heating. The product, bitumen, is used in many ways due to its engineering properties for building roads, waterproofing roofs and in hydraulic applications such as pond liners. Bitumen is a non-distillable residuum obtained from the distillation of suitable crude oils (Asphalt Institute, Eurobitume 2015). The distillation process normally involves atmospheric distillation followed by either vacuum distillation or steam distillation. Additional processing, such as air oxidation, solvent stripping or blending of petroleum residua of different stiffness characteristics, may be needed to form a material whose physical properties meet the technical requirements for commercial applications. A significant amount of research has been conducted to chemically characterize bitumen emissions under normal application conditions (NIOSH 2000; Kriech et al. 2002). Emissions comprise predominantly straight chain hydrocarbons with lesser amounts of cycloalkanes (approximately 70% aliphatic/30% aromatic). The aromatic portion includes alkyl benzenes, polar and semipolar compounds. Much of the polycyclic aromatic compound (PAC) composition is in the form of alkylated species (primarily alkyl-naphthalenes, alkyl-fluorenes and alkyl-anthracenes) and includes sulfur heterocyclic materials (primarily alkyl benzothiophene and alkyl-dibenzothiophenes). The most prominent of the parent, unalkylated PACs detected in workplace exposures for paving and roofing workers include acenaphthene, anthracene, fluoranthene, fluorene, naphthalene, phenanthrene and pyrene (Kriech et al. 2007; Cavallari et al. 2012a). Normal paving worker exposures contain only trace levels of parent unalkylated PACs, primarily of the 2- and 3-ring variety. Some 4-ring PACs have been reported in fume condensates generated to mimic paving worker exposures (pyrene and triphenylene (Clark et al. 2011) and eight 4 6 ring PACs have been reported in fume condensates generated to mimic roofing worker exposures (Clark et al. 2011), all at trace levels (< ppm) except for pyrene, a noncarcinogenic 4-ring PAC, at 33 ppm. Temperature plays a significant role in determining the amount and composition of bitumen emissions. As the

4 CRITICAL REVIEWS IN TOXICOLOGY 123 temperature of the bitumen increases, so does the concentration of emissions released as a logarithmic function (Asphalt Institute, Eurobitume 2015). Also, with increased temperature, the proportion of higher molecular weight compounds increases, which can lead to the increased presence of 4 6 ring PACs. Although minor data variations occur, the emissions generated from air rectified (partially oxidized) and straight-run vacuum distilled bitumen, appear similar in composition and physical properties (Asphalt Institute, Eurobitume 2015) when heated to the same temperature. The chemical composition of bitumen emissions is dependent on a wide variety of other variables such as crude oil source, processing, handling, oxidation (Trumbore et al. 2015), exposure levels and distance from the emission source. In 1981, Thayer et al. conducted a two-year mouse dermal cancer assay on condensates of built up roofing asphalt (BURA) emissions, derived from oxidized bitumen. Tumors were observed with fume condensates generated at 232 C (450 F) and 316 C (601 F) in a laboratory. The higher the temperature of preparation of the bitumen emission condensates, the greater tumorigenic activity observed. Kriech et al. (2007) showed that these emissions were not the same as worker exposures and that the high temperature and continuous agitation contributed to significantly higher levels of 4 6 ring PACs than seen in worker exposures. A follow-up mouse study was conducted in 1989 by Sivak et al. on five fractions of this same fume condensate with increasing polarity A, B, C, D and E. The carcinogenic activity was limited to Fractions B and C, which represented only 10.2% of the total fume condensate. These studies provided valuable insight into the complex composition and helped identify potential causative components of the potential carcinogenicity of bitumen emissions. For road paving, conventional hot-mix asphalt is a mixture of bitumen and mineral aggregate (stone) materials that is typically produced and applied at temperatures in the range of C. In Europe, the term is synonymous with asphalt, whereas the petroleum portion (the binder) is referred to as bitumen. Warm-mix asphalts have also been developed to reduce the energy required to build roads and to reduce worker exposures. Warm-mix asphalts are applied at temperatures typically C lower than conventional rolled or dense-graded asphalt. Use of diesel oil as a releasing/cleaning agent is widespread in the hot mix asphalt paving industry and can contribute significantly to worker hydrocarbon exposures, including PACs, which may also be present in bitumen emissions (Cavallari et al. 2012a, 2012b; Osborn et al. 2013). Mastic asphalt is an asphalt mixture in which the volume of filler (normally finely crushed rock) and bitumen binder exceeds the volume of remaining voids producing a stiffer material requiring higher mixture temperatures of >200 C. Mastic asphalt is only used in Europe (mainly in Germany and France) and contains aggregates <2 mm in size. The mastic industry represents only 1.1% of the worldwide asphalt market (Asphalt Institute, Eurobitume 2015; IMAA 2013). Mastic asphalt pavement systems include a leveling course called pulver asphalt placed over a concrete base. Once the pulver layer is laid, the crew pours a semi-fluid, high temperature (250 C) mastic layer. This is troweled down and then precoated stone chips are spread before rolling. Coal tar contamination from the subbase (pulver layer) has been shown to be a significant confounder in evaluations of the hazard from mastic asphalt use (Raulf-Heimsoth et al. 2008; Blackburn et al. 1999). Roofing products containing bitumen include bitumen shingles (applied at ambient temperatures), rolled roofing (often torch applied), Built-Up Roofing Asphalt (BURA), which is hot mopped at temperatures of C ( F) and has kettles that hold bitumen that can be heated up to 288 C (550 F), polymer-modified membranes, saturated felt underlayment and roofing adhesives. These are the highest temperature roofing application uses for bitumen. Hotapplied bitumen is a small and shrinking part of the roofing market. Exposures for the other bitumen roofing products are far lower (torch jobs) to negligible (shingles). Studies of hot applied bitumen applications are of little meaning for most of the roofing industry. Manual removal of old roofing materials ( roofing tear-off ) containing coal tar is one of the primary confounders involved in assessing workplace roofing worker exposures and is associated with a sixfold increase in relation to asphalt emission exposure alone (McClean et al. 2007a). Critical factors in cancer hazard evaluation for complex petroleum substances Petroleum-derived substances, such as bitumen and bitumen emissions, can contain many thousands of individual hydrocarbon species, many of which are likely to present different physico-chemical, toxicological and environmental properties. The properties and exposure profile of the complex petroleum substance will depend upon the nature and amount of the chemical species present and the operational/environmental conditions. Release of individual constituents from the complex substance and hence potential for occupational and environmental exposure will be determined by the overall properties of the material and conditions of use. This complexity of composition and properties presents a significant challenge when designing toxicological studies or evaluating existing data, to determine the carcinogenic hazards of complex petroleum substances. Review of the experimental methodology used in the studies is therefore critically important to ensure that experimental results, to the greatest extent possible, are scientifically robust and that conclusions reached are based on actual hazard properties, rather than hazard predictions based solely on the presence of individual hazardous constituents. Consideration of the following factors provides a structured approach to the design of toxicological studies and the evaluation of both individual studies and the overall data set that contributes to the final hazard conclusion. Data quality An analytical approach for assessing the quality of data from all experimental and occupational/environmental health effect studies is required to ensure that conclusions reached

5 124 A. J. KRIECH ET AL. are solely based on sound and reliable scientific evidence. The approach should include assessment of data Relevance, Reliability and Adequacy for each study. Under Registration, Evaluation, Authorization and Restriction of Chemicals (REACH), the regulatory program for registration of chemicals in Europe a systematic approach (Klimisch et al. 1997) is used to evaluate the quality of toxicological and ecotoxicological data for hazard and risk assessment. Klimisch is a widely used and accepted method for assessing the quality and adequacy of individual studies. It is recognized that human data do not readily fit the Klimisch scheme, so a modified approach has been proposed (Money et al. 2013). Both approaches result in individual studies being assigned to one of four categories (Table 1). Only studies that meet the criteria for Category 1 (Reliable without restrictions) or 2 (Reliable with restrictions) should be considered adequate for the purposes of reliable hazard and risk assessment (Klimisch et al. 1997). Test material composition The complex composition of the bulk petroleum substance will differ significantly from the emissions to which workers and others are likely to be exposed upon heating bitumen (Asphalt Institute, Eurobitume 2015). The individual hydrocarbon constituents present in the substance each have specific and different physical chemical properties, which determine its behavior and availability for exposure under a given set of conditions. It is important therefore to understand the nature and the amounts of individual constituents released under realistic, operational and environmental conditions. Analytical characterization of the test material used in experimental studies is critical to ensure that it is representative of the material workers and others may be exposed to under normal working conditions. Experimental exposure conditions An understanding of how complex petroleum substances are used/handled, and how operational conditions/environmental conditions can affect release of, and exposure to constituents, is a key consideration. This will help determine the most appropriate route of exposure and the nature of the test material. It is also important to validate the composition of the test material used against collected occupational field samples, to ensure the test material reflects human exposure under realistic conditions. Evidence of exposure and uptake When evaluating data from experimental animal studies or human occupational/environmental health studies, it is important to ascertain, either directly or indirectly, whether exposure has resulted in systemic uptake and exposure of the target cells. This can be confirmed with toxicokinetic data, if available, or indirectly by methods such as the presence of urinary metabolites or the identification of DNA adducts such as in white blood cells of workers or animals, and in skin, nasal and lung tissue samples in animal studies. Data consistency For some petroleum substances, an extensive set of health effects data are available, particularly for genotoxicity and carcinogenicity endpoints. Different studies can provide conflicting results, so careful evaluation is needed to determine whether factors such as use of unrepresentative test materials, different operational/environmental conditions, confounding exposures or experimental design bias can help explain conflicting or contradictory results. Confounding exposures In laboratory studies, it is possible to control the composition of exposure and how much the organism is exposed to. Unfortunately, in health effect studies of complex petroleum substances in exposed populations, possible confounding exposures to occupational or environmental hydrocarbons from other sources occur frequently and have the potential to influence study outcomes. Strategies to eliminate or control for possible confounding exposures are critically important in both study design and data analysis. With these critical factors guiding the path forward, we discuss human toxicology studies, genetic toxicity, validated animal cancer studies, mechanisms and other human studies relevant to bitumen emissions. Bitumen human genetic studies Human studies investigating the genotoxic potential of exposure to bitumen emissions have been reviewed and assigned modified Klimish scores. Criteria to identify studies suitable for hazard assessment included characterization of bitumen/asphalt samples, exposure levels and duration of exposure, adequate group size of exposed subjects and unexposed control populations, background data on subjects (e.g. work experience, health, smoking, alcohol consumption), workplace monitoring data and urinary metabolites and confounding variables (e.g. diesel oil, diesel exhaust, coal tar pitch). In general, all studies included were representative of workplace environment. Sample size ranged from 20 to 60 subjects with concurrent controls, except for the extensive and much larger German Human Bitumen Study ( ), which included more than 300 subjects. Details about the bitumen samples used were often limited to general descriptive terms without compositional characterization, but the studies typically included air sampling or personal exposure monitoring of total or individual PACs and urinary 1-OH pyrene content, or other biomarkers (e.g. hydroxylated phenanthrenes) used to quantify and characterize worker exposure. Studies are briefly summarized in text tables, with assigned modified Klimisch scores and discussed in greater detail in Supplemental materials. DNA adducts and DNA strand breaks Studies involving roofers exposed to bitumen vapor and aerosols were generally positive for DNA adducts/strand breaks in peripheral white blood cells (WBC) (Herbert et al.

6 CRITICAL REVIEWS IN TOXICOLOGY 125 Table 1. The four categories of review (adapted from Money et al. 2013). Category 1 Reliable without restriction Toxicology/ecotoxicology data This includes studies or data from the literature or reports that were carried out or generated according to generally valid and/or internationally accepted testing guidelines (preferably performed according to GLP) or in which the test parameters documented are based on a specific (national) testing guideline (preferably performed according to GLP) or in which all parameters described are closely related/comparable to a guideline method Human Data (chronic, nonspecific outcomes) Applies to human data (HD) where the study design was appropriate to the research question under consideration and was executed with high attention to the details of study protocols and procedures. Specifically: (1) subjects were selected to represent the appropriate exposure distributions of persons or person time at risk, with emphasis on measuring and reporting response parameters, and adequate recruitment and follow-up procedures in place to maximize participation and reduce loss to follow up with cohort studies; (2) exposure assessment was made independent of outcome, with as little measurement error as possible (relative to between subject variation), using wellestablished methods appropriate to the study design; quantitative, validated, individual-level data will be essential for exposure response assessment; (3) outcome data were collected independent of exposure status and ascertained with equal probability for both cases and noncases (or controls in a case - control study), using standardized data collection procedures (e.g. registries); (4) the possibility for serious biases have been reduced by design, controlled through statistical adjustment and/or quantified through sensitivity analyzes; (5) the sample size/exposure range was sufficient to study the question under investigation so that effect estimates are not constrained by high imprecision, and (6) these data were analyzed comprehensively, using appropriate statistical techniques to address the research questions and model assumptions, if applicable, were met. In addition, the methodology and results were comprehensively and transparently reported according to relevant guidelines e.g. the STROBE guidelines for observational data (von Elm et al. 2007) Human data (acute or specific outcomes) Similar principles to those applying for chronic or nonspecific outcomes also apply to this type of study, although the focus of the review should lie more with how well exposure has been characterized and the disease/effect outcome is defined, (e.g. no restrictions: precise exposures measured on an individual basis, outcome verified via objective criteria, etc.) and less emphasis on whether there are minor confounders and/or biases. However, in the case of very rare outcomes, reports of cases among clearly exposed population may be interpretable without restrictions although such reports would be categorized as not reliable in the case of chronic, nonspecific outcomes Category 2 Reliable with restriction Toxicology/ecotoxicology data This includes studies or data from the literature, reports (mostly not performed according to GLP), in which the test parameters documented do not totally comply with the specific testing guideline but are sufficient to accept these data or in which investigations are described which cannot be subsumed under a testing guideline, but which are nevertheless well documented and scientifically acceptable. Similar principles to those applying for chronic or nonspecific outcomes also apply to this type of study, although the focus of the review should lie more with how well exposure has been characterized and the disease/effect outcome is defined, (e.g. no restrictions: precise exposures measured on an individual basis, outcome verified via objective criteria, etc.) and less emphasis on whether there are minor confounders and/or biases. However, in the case of very rare outcomes, reports of cases among clearly exposed population may be interpretable without restrictions although such reports would be categorized as not reliable in the case of chronic, nonspecific outcomes Human Data (chronic, nonspecific outcomes) Applies to HD which possess most of the elements of a Type 1 quality study, but where their overall quality has been compromised due to minor, but obvious, methodological limitations. Examples of such limitations include well-designed and well-conducted studies, but where there are limited measurement data available to validate estimated individual-level exposure data, some residual confounding is possible or there is some imprecision because of a small sample size or low exposure range. Alternatively, the study design may not be optimal, forexample, a cross-sectional design that does not allow inferences to be made about the time order of events. Another example would be a study in which subject selection procedures and/or postentry loss to follow up introduces the possibility of selection bias. Human data (acute or specific outcomes) Likely to apply to HD which possess most of the elements of a Type 1 study, but where some restrictions on interpretation are appropriate, for example, individual-level exposure data are lacking; effects derive from self-reported outcomes, etc. As in the case of type-1 studies, some studies with serious methodological limitations may provide reliable information for an acute or specific outcome. It is also possible for the same study with specific and nonspecific outcomes to have a different Klimisch scores for these outcomes. Category 3 not reliable Toxicology/ecotoxicology data This includes studies or data from the literature/reports in which there are interferences between the measuring system and the test substance or in which organisms/test systems were used which are not relevant in relation to the exposure (e.g. nonphysiologic pathways of application) or which were carried out or generated according to a method which is not acceptable, the documentation of which is not sufficient for an assessment and which is not convincing for an expert judgment Human data These studies fail to meet one or more of the most basic standards necessary to interpret epidemiologic research, such as appropriate study design for the research question of interest or adequate selection of study subjects. Nevertheless, many such studies will be well-conducted investigations and be useful for the purpose intended, for example, hypothesis generation, but their design renders them uninformative for hazard identification or risk assessment. Examples of such studies include ecological studies and linkage studies using census job titles as a surrogate for exposure. Studies classified as not reliable will often have serious methodological flaws that make the results uninterpretable regarding a causal association, particularly with respect to chronic, nonspecific outcomes Category 4 not assignable This includes studies or data from the literature, which do not give sufficient experimental details and which are only listed in short abstracts or secondary literature (books, reviews, etc.) 1990a, 1990b; Fuchs et al. 1996; Toraason et al. 2001), whereas DNA adducts observed in paving workers and asphalt painters demonstrated merely suggestive evidence of possible genotoxic effects (Fuchs et al. 1996) (Table 2). Sample sizes were small, but exposures were monitored and worker backgrounds provided. Toraason et al. (2001) demonstrated that when roofers were exposed to coal tar pitch dust as well as bitumen emissions in roof tear-off operations,

7 126 A. J. KRIECH ET AL. Table 2. Summary of human mechanistic studies DNA adducts and strand break studies. References Test material Results Klimisch score and comments Herbert et al. (1990a,b) Tear-off of coal tar pitch roof, application of asphalt roof Roofers (12), Control (12) Positive DNA adducts in roofers postshift correlated with skin levels of total PAH and B(a)P but not personal Klimisch 4. Coal tar confounds relevance to bitumen. All subjects exposed to both coal tar pitch and Fuchs et al. (1996) Toraason et al. (2001) Marczynski et al. (2006) Marczynski et al. (2010) Marczynski et al. (2011) Raulf-Heimsoth et al. (2011a) Raulf-Heimsoth et al. (2011b) Cavallo et al McClean et al. (2007a,b) Kendzia et al. (2012) Roofing fumes (350 C); paving fumes (150 C); bitumen paint. Roofers (7), pavers (18), painters (9), Control (34) Steep asphalt. Some Coal tar pitch exposure Roofers (26), Control (15) Mastic asphalt for paving indoor garages. Pavers (66), Control (45) Mastic asphalt for paving indoor garages. Pavers (42), no control cited Mastic asphalt for paving indoor garages. Pavers (320), Control (118) Mastic asphalt for paving indoor garages. Pavers (320), Control (118) Rolled asphalt followed by mastic asphalt in tunnel pavers (6) Concrete asphalt fumes, 8 h shifts Pavers (19), Controls (22) Paving asphalt some high levels recycled asphalt product Pavers (49), Controls (36) over 4 seasons. Road construction workers nonbitumen exposed (118) air samples Positive strand breaks for roofers, negative overall for pavers and painters. Low levels of DNA adducts in pavers and 2/9 painters Positive for strand breaks. No oxidative damage. Coal tar þ asphalt exposed workers had higher PAH exposure and slightly higher strand breaks Positive increased strand breaks before and after shift compared to control. OxodGuo adducts did not correlate with strand breaks, or exposure Negative for increased strand breaks in induced sputum [IS] leukocytes postshift. DNA damage overall high but did not change during shift. Correlation with IL-8 measure of inflammation Positive: Higher 8-oxodGuo adducts and DNA strand break in pre- and postshift blood samples compared to concurrent controls but within range of nonexposed healthy individuals. Indicators of lower airway irritation and inflammatory effects in induced sputum samples seen preshift in exposed workers did not increase postshift. DNA strand breaks higher pre-shift, decreased during shift. Increased 8-oxodGuo independent of asphalt application. Positive for increased DNA strand breaks, some oxidative damage in WBC. Negative for SCE Equivocal: Increases in adduct levels during work week. Levels similar in spring, summer and fall, higher in winter. Positive: Nonsmokers pre-dna stand breaks preshift; more DNA adducts and apoptosis postshift. Conc. of IL-8 in lower airways enhanced in smokers. asphalt. Klimisch 2. Well-performed study albeit small sample size for roofers (7) and painters (9). Pavers and painters DNA effects comparable to controls. Klimisch 2. Demonstrated coal tar tearoff exposure a factor in increasing DNA damage in roofers. Well performed study, good monitoring Klimisch 2. Cross shift study with large sample size. Part of German Human Bitumen Study from Klimisch 2. Follow-up to large German study Klimisch 1. Large cross-sectional, crossshift study. Authors considered increases in genotoxic effect not be attributed to bitumen exposure due to lack of association with exposure levels Klimisch 2. Large cross-sectional, crossshift study No significant effect on upper airway by NALF. Lung function factors in normal range with slight decrease in FEV1 & FEC during shift Klimisch 2. Small longitudinal study compared results with larger asphalt exposed population to identify influence of tunnel setting. Klimisch 2. Well-performed study, background data. Did not address specific asphalt product Klimisch 3. Complex results, multiple confounders and the presence of diesel oil used in cleaning tools Klimisch 2. Cross-sectional cross-shift data set for comparison asphalt exposed workers in German Human Bitumen study breathing zone measurements of total particulates, benzenesoluble and total PAC content were higher and DNA damage was statistically significantly greater (p <.05). Exposure to bitumen alone was associated with a small but significant increase in DNA stand breaks. Thus, studies evaluating DNA damage among roofing workers must correct for the potential presence of coal tar dust (confounding exposure) in evaluating positive results. Bitumen emission-induced DNA adducts have been characterized as various benzo[a]pyrene diol epoxide adducts (Wey et al. 1992). Herbert et al. (1990a, 1990b) identified one major adduct and several fainter ones from peripheral white blood cells of roofing workers. When mixed with BPDE-1-DNA, this principal unnamed DNA adduct did not comigrate with the major N 2 -guanine-adduct of B[a]P diol epoxide [BPDE-I guanine N 2 adduct], suggesting that B[a]P in bitumen emissions is not the principal source of adducts. Most of the studies have addressed exposure of paving workers. Many conventional paving worker studies show the incidence of DNA strand breaks in bitumen-exposed workers was higher than unexposed controls and sometimes increased with work week exposure. McClean et al. (2007a, 2007b) designed a longitudinal study evaluating DNA adduct formation in hot mix asphalt paving workers (49) and nonpavers (36) over 12 months during the spring, summer and autumn working seasons and nonworking season (winter). They reported that DNA adducts increased during the work week, consistent with absorbed dose measured by urinary 1- hydroxypyrene (1-OHP) and the magnitude of response varied with paving tasks. However, the adduct levels were no higher among paving workers than among nonpaving workers, suggesting that either the presumed unexposed controls were inadvertently exposed to asphalt dust or the apparent correlation between bitumen emissions exposure and DNA

8 CRITICAL REVIEWS IN TOXICOLOGY 127 adducts is not a definitive measure for health hazard identification. Adding to this question was the observation that for both groups DNA adduct levels were higher in non-working winter months indicating the potential importance of seasonal variations and other environmental factors in contributing to DNA damage. The complexity of these results and multiple confounding factors brings into question the reliability of this study for predicting genotoxic effects. Concern for high exposure, at higher application temperatures of bitumen in the small population of mastic workers in Germany resulted in the performance of the German Human Bitumen Study evaluating mastic asphalt workers (320 workers, 118 unexposed controls) employed in paving indoor garage spaces from 24 to 180 months. This extensive study produced six DNA adduct and strand break studies (Marcyznski et al. 2006, 2010, 2011; Raulf-Heimsoth et al. 2011a, 2011b; Kendzia et al. 2012), one cytogenetic study (Welge et al. 2011) and two biomarker studies (Pesch et al. 2011; Lotz et al. 2016). DNA strand breaks were higher in bitumen-exposed workers before and after work-shift compared to unexposed controls but strand break levels did not increase significantly during the work week. The higher levels of strand breaks detected preshift suggest DNA cross-links from the previous week s exposure were repaired over the weekend leaving more strand breaks pre-shift (Marcyznski et al. 2006, 2010). Interestingly, some unexposed controls showed a similar pattern of strand breaks albeit at lower levels. In a subsequent publication evaluating all subjects in the study, Marcyznski et al. (2011) reported that elevated levels of DNA strand breaks and 8-oxo-dG adducts in bitumen exposed mastic asphalt workers compared to reference construction workers, were considered within the range of nonexposed healthy individuals, leading the authors to conclude that there was no positive association between DNA damage and the magnitude of bitumen emissions or between urinary metabolites and DNA damage in the blood. Conventional paving asphalt is handled at lower temperatures [180 C] than mastic asphalt [250 C]. To evaluate differences in response to exposure to vapors and aerosols, Raulf-Heimsoth et al. (2011a) examined a small subgroup of 6 workers employed at a tunnel construction site over 2 weeks, one week working with conventional paving asphalt, the second week applying mastic asphalt. Subjects were sampled mid-week. Although mastic asphalt application led to higher exposure to PAC and bitumen containing vapor and aerosol than conventional paving asphalt workers, no differences were seen in urinary metabolite excretion, lung function impairment or genotoxic markers between examination days. In a subset of 42 mastic asphalt workers (no controls) from the German Human Bitumen Study, induced sputum (IS) was collected pre- and postshift and evaluated for DNA strand breaks in sputum leukocytes compared with DNA strand breaks in blood lymphocytes; interleukin-8 (IL-8) levels, a marker of irritation, were also measured (Marcyznski et al. 2010). DNA damage in IS leukocytes was high overall but did not increase during workshift. The values correlated with total cell and neutrophil count pre- and postshift but did not correlate with DNA strand breaks in blood lymphocytes, which were higher before work shift than after work shift. DNA stand breaks in IS cells did correlate with IL-8 levels. Raulf- Heimsoth et al. (2011b) followed this study by evaluating the irritating effects of bitumen-containing vapor and aerosol exposure for all 320 mastic asphalt workers and 118 reference controls using spirometry, nasal lavage fluid (NALF) and induced sputum. IS concentrations of interleukin 8, total protein and matrix metalloproteinase-9, all biomarkers of inflammatory effects in lower airways were statistically significantly higher (p <.05) in exposed workers before work-shift and did not increase during work-shift. No significant effects were seen between groups for upper airways by NALF and lung function factors were within normal range for both pre- and post-shift groups. Results suggest a chronic irritation effect in lower airways, which does not seem to be exacerbated by additional workplace exposure. In 2012, Kendzia et al. expanded the German Human Bitumen study by evaluating pre- and post-shift levels of inflammatory biomarkers and DNA-damage in non-bitumen exposed constructions workers (59 smokers, 59 nonsmokers). Values for DNA strand breaks and 8-oxodGuo adduct from peripheral white blood cells were like those reported for controls in the Marczynski et al. (2011) publication evaluating DNA damage in bitumen exposed workers. The pattern of response for nonsmokers indicated more DNA strand breaks preshift than postshift while more DNA adducts and greater percentage of apoptotic cells were seen postshift. Cross-shift differences were also seen in concentrations of preinflammatory IL-8 in lower airways with higher levels in the morning before work-shift. Concentration of IL-8 in smokers was enhanced both pre- and postshift compared to nonsmokers. This study indicated the potential influence of circadian rhythm on biomarkers. When considered with Marczynski et al. (2011), a similar shift-work effect from general occupational exposure was observed, marginally enhanced by bitumen exposure but within the acceptable range of healthy individuals. Pesch et al. (2011) and Lotz et al. (2016) reported that PAC metabolites in urine of mastic asphalt workers could not be considered specific biomarkers of bitumen exposure in the workplace due to the weak correlation with airborne bitumen exposure levels. Smoking however showed strong impact on PAC metabolite concentrations. Pesch et al. (2011) found only a weak association between airborne concentrations of bitumen during a working shift and post-shift concentrations of urinary 1-OHP and hydroxy-phenanthrene metabolites and no association with hydroxy-naphthalene metabolites in 317 mastic asphalt workers compared to 117 unexposed reference controls. Lotz et al. (2016) showed similar results investigating diol epoxide pathways of phenanthrene (PHE) and phenolic pathways of PHE and pyrene in urine of 91 mastic asphalt workers compared to 42 unexposed reference controls, before and after work-shifts. Urinary diol epoxides of PHE and pyrene increased during and after shift and were highest in smokers. Metabolites of diol epoxides of PHE were excreted in higher concentrations than phenolic metabolites. Overall, correlations of these PAC metabolites with bitumen exposure were weak or negligible and likely result from low concentrations of PACs (0.05%) in airborne bitumen emissions.

9 128 A. J. KRIECH ET AL. Table 3. Summary of human mechanistic studies cytogenetic studies. References Test material/subjects Results Klimisch score and comments Burgaz et al. (1998) [Turkey] Paving asphalt steam-refined Pavers (21), plant (7) Control (28) Murray and Edwards, (2005) Paving asphalt. Road pavers (12), Control (18) J arvholm et al. (1999) Karaman and Pirim, (2009) Major, Jakab and Tompa, (2001) Tompa et al. (2007) Sellapa et al [South India] Welge et al. (2011) Celik et al. (2013) [Mersin Turkey] Paving asphalt from one source. Road pavers (28); Controls (30) Concrete asphalt. Pavers (26), Controls (24) Paving Asphalt tar-free [Hungary] Pavers (8), drivers (14), Concurrent (6) & historical (101 þ 87) controls Paving asphalt tar-free follow up to Major et al. (2001) above Road paving asphalt Pavers (36), Controls (37) Hot application mastic asphalt Mastic pavers (225), Controls (69) Paving asphalt Road pavers (40), Controls (40) Positive: SCE; % micronucleated [MN] cells higher in asphalt workers than controls Positive: Significant micronucleated cell frequency in urethral cells and lymphocytes Negative: No significant increase in SCE or micronucleated cells in lymphocytes. Positive: SCE and % MN correlated with increasing years of exposure Positive for chromosome aberrations (CA) and SCE in 1996 with decreasing results to control levels in 1999 CA levels at control levels in 1999 began to increase in with use of crude oil again for cleaning and inadequate ventilation in finisher cabs Positive: MN and DNA damage in WBC. Smokers and heavy drinkers had higher MN and DNA damage. Increased with years of employment. Negative: MN frequency comparable exposed and control preshift and slightly elevated for both postshift; regression modeling suggested slight weak group difference postshift. Positive: Increase in MN in buccal mucosal cells, % karyorrhectic cells higher and repair index for cytogenetic damage lower. BMCyt assay Klimisch 3. Use of diesel oil to clean tools in Turkey confounds results. Klimisch 3. Study designed to compare events in urethane workers (12) with asphalt workers; design weak, no exposure data, potential use of diesel oil. Klimisch 2. Adequate study. Asphalt single sourced in Sweden. No coal tar present. Klimisch 2. SCE and MN higher than controls pre- and post- shift but comparable to each other Klimisch 1. Study demonstrated from improved working conditions, better ventilation replacement of crude oil for cleaning equipment Klimisch 1. Mutagenic activity may be related more to organic solvents and diesel exhaust than asphalt fume exposure. Immunomodulation as shifts in lymphocyte subpop n and activation of T & B cells Klimisch 3: Diesel oil likely present based on typical work practices, confounder of effects, no urinary exposure measurements. Klimisch 2. Large cross-sectional, crossshift study from German Human Bitumen Study Klimisch 3: absence of specific detail on asphalt composition, air monitoring for PAC exposure or urinary measurements [e.g. 1-OHP] Table 2 shows that most studies meet criteria for use for hazard assessment with only McClean et al. (2007a, 2007b) considered unacceptable due to the complexity of results, multiple confounders and Herbert et al. (1990a, 1990b), unassignable due to exposure to coal tar pitch exposure of all subjects. Although most studies show primarily positive results for induction of DNA damage, the extensive German Human Worker study and even the McClean et al. (2007a) study with its limitations suggest that these results fall within the range of values seen in populations of normal, healthy adults and may not be predictive of potential health effects from exposure to bitumen. Cytogenetic studies Cytogenetic studies evaluated the occurrence, or increases in sister chromatid exchange (SCE), micronuclei formation (MN) or chromosome aberrations (CA) primarily in paving workers (Table 3). Although studies of the large population from the German Human Bitumen study showed DNA strand breaks and DNA adduct formation with exposure to bitumen vapor and aerosol, a micronucleus study indicative of actual cytogenetic damage produced comparable results for mastic asphalt workers and reference controls pre- and postshift (Welge et al. 2011). Regression modeling did show a weak and nonstatistically significant group difference in the postshift mode but age was the strongest predictor of increased micronuclei formation in both control and mastic asphalt workers. Similarly, Cavallo et al. (2006) evaluating 18 pavers exposed to bitumen emissions compared to 22 unexposed controls reported statistically significant increased DNA strand breaks and increased oxidative damage (p <.05), but no increase in sister chromatid exchanges. J arvholm et al. (1999) in comparing 28 road paving workers to 30 unexposed controls, all nonsmokers, reported no statistically significant increase in SCE or MN (p >.05). Urinary 1-OHP levels were statistically significantly higher in bitumen workers than in unexposed controls (p <.01) but levels did not increase during shift. The bitumen used came from a single manufacturer, who provides all the bitumen in Sweden. Statistically significantly positive SCE and MN results (p <.05; p <.001, respectively) were reported by Burgaz et al. (1998) in peripheral blood samples from paving workers exposed for a workplace duration of 9.8 ± 3 years. Smoking habits did not have a statistically significant effect among exposed and control subjects (p >.05). Murray and Edwards (2005) also found statistically significantly increased SCE and MN (p <.01) in peripheral blood lymphocytes and exfoliated urethral cells in a comparative study of paving and urethane workers with controls (all subjects sampled once); smoking did not affect results within exposure groups (p >.05). Karaman and Pirim (2009) determined that Turkish road paving workers exposed to bitumen at 170 C had similar