Occupational Exposure to Benzene at the Exxonmobil Refinery In

www.nature.com/jes

Occupational exposure to benzene at the ExxonMobil Reﬁnery in Baytown, TX (1978–2006)

SHANNON H. GAFFNEYa, JULIE M. PANKOb, KEN M. UNICEb, AMANDA M. BURNSb, MARISA L. KREIDERb, RICHARD H. GELATTc, LINDSAY E. BOOHERd AND DENNIS J. PAUSTENBACHa aChemRisk, LLC, 25 Jessie Street, Suite 1800, San Francisco, California, USA bChemRisk, LLC, Pittsburgh, Pennsylvania, USA cExxonMobil Biological Sciences, Annandale, New Jersey, USA dExxonMobil Production Company, Houston, Texas, USA

Although occupational benzene exposure of refinery workers has been studied for decades, no extensive analysis of historical industrial hygiene datahas been performed focusing on airborne concentrations at specific refineries and tasks. This study characterizes benzene exposures at the ExxonMobil Baytown, TX, refinery from 1978 to 2006 to understand the variability in workers’ exposures over time and during different job tasks. Exposures were grouped by operational status, job title, and tasks. More than 9000 industrial hygiene air samples were evaluated; approximately 4000 non-task (43h) and 1000 task-related (o3 h) personal samples were considered. Each sample was assigned to one of 27 job titles, 29 work areas, and 16 task bins (when applicable). Process technicians were sampled most frequently, resulting in the following mean benzene concentrations by area: hydrofiner (n ¼ 245, mean ¼ 1.3 p.p.m.), oil movements (n ¼ 286, mean ¼ 0.23 p.p.m.), reformer (n ¼ 575, mean ¼ 0.10 p.p.m.), tank farm (n ¼ 9, mean ¼ 0.65 p.p.m.), waste treatment (n ¼ 446, mean ¼ 0.13 p.p.m.), and other areas (n ¼ 460, mean ¼ 0.062 p.p.m.). The most frequently sampled task was sample collection (n ¼ 218, mean ¼ 0.40 p.p.m.). Job title and area did not significantly impact task-related exposures. Airborne concentrations were significantly lower after 1990 than before 1990. Results of this task-focused study may be useful when analyzing benzene exposures at other refineries. Journal of Exposure Science and Environmental Epidemiology (2011) 21, 169–185; doi:10.1038/jes.2009.53; published online 28 October 2009

Keywords: benzene, reﬁneries, exposure assessment, industrial hygiene.

Introduction sures during refining operations indicate that full-shift exposures have been typically less than 1 part per million Benzene is a natural, minor constituent of crude oil, where (p.p.m.) since the 1980s (Weaver et al., 1983; Buchet et al., it is present in amounts varying from 0.1% to 3% (Verma 1984; Runion and Scott, 1985; CONCAWE, 1987, 1994; and des Tombe, 1999). During the distillation of crude oil, Nordlinder and Ramnas, 1987; Rappaport et al., 1987; HEI, the lighter weight fractions of the crude are extracted; because 1988; Glass et al., 1994, 2000, 2001; Armstrong et al., 1996; benzene has a low molecular weight and boiling point, it is Verma et al., 2001). Although these studies provide use- removed with the lower boiling fractions. Benzene can also ful information regarding potential occupational exposures be produced in refining processes, such as reforming. to benzene in the petroleum industry, they do not pro- Once produced, pure benzene is isolated in the refinery vide detailed job- or task-specific exposure data or informa- and transported through an enclosed and continuous pro- tion specific to workers at individual facilities. As a result, the cess to the chemical plant, primarily for use as an industry-wide data sets are difficult to use if one wishes to intermediate feedstock in the production of other chemicals. conduct more detailed dose reconstruction studies or to Although the refining process is generally ‘‘closed and estimate the exposures of individual workers on the basis of continuous,’’ occupational exposure to benzene in the oil their job descriptions. This study expands upon the currently refining industry has been evaluated for more than 60 years available literature by analyzing industrial hygiene data (OSHA, 2003). Existing literature reporting benzene expo- collected from one of the world’s largest refineries on the basis of the jobs and tasks being performed. In the United States, ExxonMobil currently owns and 1. Address all correspondence to: Dr. Shannon H. Gaffney, ChemRisk, operates seven petroleum refineries in Baton Rouge, LA; LLC, 25 Jessie Street, Suite 1800, San Francisco, CA 94105, USA. Joliet, IL; Baytown, TX; Beaumont, TX; Chalmette, LA; Tel.: 415-896-2400. Fax: 415-896-2444. Torrance, CA; and Billings, MT. As part of an ongoing E-mail: [email protected] Received 18 June 2009; accepted 31 August 2009; published online 28 historical benzene exposure assessment program, analyses have October 2009 been conducted to understand both non-task (approximately Gaffney et al. Exposures to benzene at refineries full-shift time weighted averages) and task level exposures for were installed in 1988 at process stream sampling points, and refinery employees for these facilities (Panko et al., 2009). these allow workers to sample process streams in a closed This paper presents the quantitative estimates of potential system, thereby reducing the potential for benzene exposure exposure by job category and task for the largest domestic during sampling. A benzene stripper was installed in 1996 to ExxonMobil refinery, located in Baytown, TX. remove benzene from any water stream that contained greater than 0.5% benzene. Such water streams are likely to be those that contact process streams, particularly those Background associated with fuels production. As pure benzene is produced at and exported from the The Baytown refinery was built by Humble Oil in 1919. Baytown refinery and associated chemical plant complex, Humble Oil was eventually purchased by Standard Oil of much of the focus of engineering control projects at the New Jersey, later named Exxon, and thus upon the merger of refinery was placed on lowering benzene exposures at the Exxon and Mobil in 1999, operation of the Baytown refinery loading docks, particularly at the dock where benzene was continued under the ExxonMobil Corporation (Henderson loaded onto barges. and Benjamin, 1996; Segal, 1999). The Baytown refinery is one of the largest in the United States capable of processing over 500,000 barrels of oil per day (b.p.d.). Nearly 95% of Industrial Hygiene Program for Benzene the crude oil processed at the Baytown refinery arrives through ships and barges. Both fuels and lube products are In addition to implementing process modifications designed produced at this refinery. The Baytown refinery produced to reduce benzene exposure in the workplace, ExxonMobil asphalt for product sales until 2002. Fuel products are established benzene programs and practices to ensure that primarily exported through pipeline, whereas specialty employee exposures were less than the generally accepted products, such as lubes, leave through barge. There are no exposure limits and to comply with the Occupational Safety rail or truck loading locations within the refinery. and Health Administration (OSHA) standards. As OSHA Discussion of general refining processes have been detailed changed the benzene exposure limits, similar adjustments elsewhere; however, all refineries are unique (OSHA, 2003; were made to the Baytown refinery’s benzene program to Meyers, 2004). Information specific to the process config- maintain compliance. These programs included the identifi- uration at the Baytown facility is therefore provided in cation of benzene-containing streams and associated tasks Figure 1. In addition, a description of each area in the that potentially resulted in worker exposure to benzene, refinery and its associated functions are provided in Table 1. employee monitoring and hazard communication, medical surveillance of potentially exposed workers, and the im- Historical Engineering and Process Changes that Reduced plementation of benzene exposure controls including hand- Occupational Exposure to Benzene ling practices for the performance of tasks where there may The greatest potential for benzene exposure during refinery be a potential for exposure to benzene, including preparing operations results from direct contact with benzene-contain- equipment for maintenance work, opening and blinding ing process or waste streams. Accordingly, certain jobs or equipment, enclosed space entry, and clean-up of spills or tasks, such as those involving the opening of lines or leaks. These guidelines include gas testing before performing equipment containing benzene, provide the highest likelihood the tasks and establishing a temporary benzene ‘‘regulated of worker exposure to benzene. Occupational health practi- area’’ if test results are greater than 1 p.p.m. benzene. tioners at the Baytown refinery have a history of limiting potential worker exposure to benzene; they have been Overview of Industrial Hygiene Monitoring Strategy evaluating worker exposure to benzene since the 1930s, and Industrial hygiene programs at the ExxonMobil refineries have well-established industrial hygiene guidelines/proce- across the United States are guided by an exposure dures for minimizing potential exposure to benzene. The assessment strategy (EAS) that incorporates both qualitative use of personal protective equipment such as respirators, and quantitative aspects for assessing worker exposure. The protective clothing and gloves, and the containment of EAS is a multistep process that starts with an evaluation of process streams in closed systems has been a focus of the the chemical constituents of the process streams to which industrial hygiene and operations groups since the refinery workers may be exposed, and then assesses the level of was built. In addition, numerous operational changes and potential (independent of the use of respirators) on the basis engineering controls have been implemented to reduce the of the tasks that may bring the workers into contact with potential for worker exposure to benzene and to limit streams or other potentially hazardous materials. Quantita- environmental emissions of benzene. tive measurements are then targeted to assess tasks and The Baytown refinery was one of the first to institute the activities with higher potential for exposure, or to reduce the use of ventilated and enclosed sampling boxes. These boxes uncertainty in exposure assessment results. As part of the

170 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at reﬁneries Gaffney et al.

Baytown Reﬁnery Process Flow Diagram

Crud Light Ends Range 1%-6% Propanes Light Ends CRUDE LIGHT ENDS Liquid Average 0.02% Butanes Range: 0.0% - 0.5% Reformer Aliphatic Solvents Range: 0.0% - 5.7% LIGHT ENDS REFORMING MOGAS NAPHTHA HYDROFINING ALKYLATION MOGAS Naphtha Range Alkylation - 0% 0.4%-1.6% Varsol Fluids KEROSENE Jet Fuels HYDROFINING Kerosene Hydrofining Diesel Fuels Range: 0.1%-0.7 % Exxsol D Fluids CRUDE OIL PIPESTILLS LGO (PS3, PS7, HYDROFINING DIESEL FUELS (5% Pipeline PS8) Cat Light Ends HAGO 95% Tanker) Catalytic Cracking Range LVGO Range 0%-1.8% Crude Oil HYDROFINING 0.5%-1.0% DIESEL Range CATALYTIC CAT. LIGHT FUELS <0.01%-0.6% LVGO CRACKING ENDS MVGO Mogas Range HVGO 0.25%-1.2% HYDROFINING HYDROFINER MOGAS Distillation Cracking bottoms Range: 0.003%-2.8% FUEL OILS HYDROCRACKING MOGAS Extraction, hydrofining, dewaxing units DEASPHALTING 0% LUBE BASED LUBE PROCESSING STOCKS RESIDUUM COKING COKE Coker Naphtha Range 0.1%-1.6% ASPHALT All benzene ranges listed in weight % (until 2002) LGO - Light gas oil HAGO - Heavy atmospheric gas oil LVGO - Light vacuum gas oil MVGO - Medium vacuum gas oil HVGO - Heavy vacuum gas oil MOGAS - Motor gasoline Figure 1. Process flow diagram for the ExxonMobil refinery in Baytown, TX. This diagram identifies the process streams associated with the major units of the refinery, including benzene content by weight.

benzene program, much of the monitoring has historically program is likely to over-represent workers and work been focused on activities where a worker is working on activities with higher benzene exposures. equipment containing streams with higher percentages of benzene (i.e., reformate) and where the activities involve Personal Protective Equipment open handling of streams (e.g., opening/draining equipment Current and historic ExxonMobil respiratory protection versus operating closed processing equipment), Similarly, the standards were reviewed to identify those tasks and activities industrial hygiene monitoring has historically targeted work- that have historically required respiratory protection. In the ers who are potentially exposed to process streams more past, a few areas of the Baytown reﬁnery required respirators frequently than those who are not. This sampling strategy (parts of the environmental control house, docks, and the oil has resulted in the monitoring of selected jobs and tasks that movements control center) and were designated as permanent potentially have higher levels of benzene exposure than would benzene-regulated areas in accordance with 29 CFR 1910.28. result from a random sampling approach. Furthermore, once These areas have changed over time as engineering controls an area, job, or task is well characterized and determined to have reduced the exposure potential. On the basis of gas be of low exposure potential, or after controls have been testing results, in instances when benzene concentrations are implemented, collecting additional data for that area, job, or greater than 1 p.p.m. (conditions most likely to occur during task becomes infrequent. As a result of this focused or periodic tasks involving work with open process equipment), targeted monitoring strategy (e.g., in contrast to an industrial appropriate respirators are required to provide protection hygiene program that involves collecting random samples, or to the workers. In addition to these instances, certain speciﬁc that requires quarterly or annual sampling of every employ- tasks in the oil movements area require respiratory protec- ee), the benzene air sampling data collected under this tion regardless of gas testing results; such tasks include tank

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 171 Gaffney et al. Exposures to benzene at reﬁneries

Ta bl e 1 . Description of areas within the ExxonMobil reﬁnery in Baytown, TX that are represented in the database.

Area Abbreviations Function

Waste treatment (n ¼ 1159) WT To reduce or prevent environmental contamination from the refining process. Waste treatment is considered separately from the rest of the utilities area. Reformer (n ¼ 823) RF To promote conversion of low octane feed components (naphtha) into high-octane products (reformate). This process results in the generation of aromatics, such as benzene, in the stream. Oil movements (n ¼ 571) OM Area designated for fuel blending, treatment, and storage. Laboratory (n ¼ 565) LAB To analyze samples of feedstocks and end products to determine characteristics. Hydrofiner (n ¼ 466) HF To remove impurities (sulfur, nitrogen, oxygen) in feedstocks or end products through catalytic treatment with hydrogen. Desulfurization (n ¼ 215) DES To remove sulfur containing impurities from feedstocks or end products; similar to hydrofiner. Catalytic Light Ends CLEU To separate various light end hydrocarbons or to remove heavier hydrocarbons from process streams Unit (n ¼ 175) originating from the catalytic cracker. Pipestill (n ¼ 164) PS To separate crude oil into fractions on the basis of the boiling point. Shops (n ¼ 119) SH Areas designated for repair of machinery, including garage repairs and repair of other refinery-related equipment. Light Ends Unit (n ¼ 107) LEU To separate various light-end hydrocarbons into fuel gas, propane, butane, and isobutane or to remove heavier hydrocarbons from process gases. Catalytic cracker (n ¼ 105) CC Uses catalyzed reaction to break larger hydrocarbon molecules into smaller molecules to increase quality and quantity of more desirable products. Garage (n ¼ 83) GA Refinery vehicle maintenance and storage area. MEK Unit (n ¼ 68) MEK Uses a mixture of methyl ethyl ketone and toluene as solvents in extracting wax from the lube oil, making it suitable for low temperature applications. Hydrocracker (n ¼ 55) HDC Uses both a catalyst and hydrogen to crack feedstocks into more desirable products. Tank Farm (n ¼ 45) TF To store starting or end products. Considered separately from the rest of oil movements area due to nature of potential exposure (enclosed space). Coker (n ¼ 40) CK To convert crude oil residues and low-value process oils into lighter stocks or to marketable products (i.e., petroleum coke) in order to maximize end product availability from the refining process. Lube blending and LBS To blend and store oils, greases, and other end products from heavy hydrocarbon process streams. storage (n ¼ 39) Alkylation Plant (n ¼ 30) ALK To add alkyl groups (most commonly isobutane) to the stream using heat, pressure, and/or catalysts to improve octane rating of the products. Utilities (n ¼ 21) UT Area designated for flares and cooling towers, steam production (through boilers) power generation, benzene stripping, and water treatment (both waste and river waters). Asphalt plant (n ¼ 21) AS To recover asphalt from residuum from the atmospheric and vacuum distillation processes. Solvents hydrogenation SHU To produce quality hydrocarbon fluids, which are very low in aromatics, olefins and polar compounds unit (n ¼ 18) through hydrogenation. Dewaxing area (n ¼ 16) DEW To remove wax residues from the oil products through the use of either methylethyl ketone dewaxing or propane dewaxing. Lube extraction LXU Removes and isolates lube oils from process streams. unit (n ¼ 13) Deasphalting unit (n ¼ 10) DEA Remove asphalt and other heavy bottoms from hydrocarbon streams produce more desirable products (i.e., transportation fuels); asphalt is often marketed as an end product itself.

SO2 Plant (n ¼ 10) SO2 To produce elemental sulfur from hydrogen sulfide containing process streams. Administration (n ¼ 4) AD Office staff. Medical (n ¼ 4) MED Occupational medical division. Hydrogen generation HGU To produce high purity hydrogen for use in other refinery operations (i.e., hydrofining). unit (n ¼ 1) Solvent fractionating SFU To produce quality hydrocarbon fluids, which are very low in aromatics, olefins and polar compounds unit (n ¼ 1) through distillation.

gauging, tank water draws, and mechanical work in tanks ﬂushed before opening or in areas where gas testing detects (For tanks containing benzene, steam cracked naphtha, benzene concentrations in air above the exposure limit. In hydroformer feed, and/or reformate.). Temporary areas these cases, respirators are required to be worn and the type requiring benzene respiratory protective equipment have of respirator is dependent on the measured air concentration. historically been used for selected tasks and activities by The potential for dermal exposure to benzene may exist for standard operating and maintenance procedures including anyone working with an open benzene-containing process opening any process equipment that has contained more than stream. A review of all available current and historical 0.1% benzene and could not be completely drained and reﬁnery industrial hygiene documentation was performed to

172 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at refineries Gaffney et al. determine whether specific benzene-related jobs or tasks personal protection practices throughout the refinery. Final- actually offered a significant opportunity for dermal exposure ly, refinery process engineers, industrial hygienists, and to benzene, and whether PPE was recommended for these operations personnel were consulted regarding information tasks. On the basis of this review, it appears that refinery not found in written records. workers were always required to wear protective clothing and gloves to prevent dermal exposure. The industrial hygiene Data Analysis records support this assessment as records after about 1970 Air samples collected during routine and turnaround indicate that when a potential dermal exposure to benzene operations were organized by area (Table 1), job title existed, workers wore gloves. In addition, the electronic (Table 2), and task description (Table 3). Air samples were industrial hygiene records indicate that more than 99% of the also classified by sample duration (i.e., o180 min and tasks and jobs that have been monitored for benzene Z180 min) and type of sample (i.e., personal and area). exposure did not involve dermal contact. Therefore, absorp- Samples less than 180 min were typically collected to capture tion of benzene due to dermal contact was judged to be periods in which specific task activities were underway and de minimis, and no attempts to quantify exposure were were considered task samples representative of peak or task- performed. specific exposures. Samples with durations of 180 min or greater were considered non-task samples, as they most likely characterized more than one task performed by a worker as Methods part of routine job duties and were not targeted in monitoring programs as task-specific. Samples were also classified as Data Collection non-task samples when the sample collection time could not Historical industrial hygiene data used in this analysis were be determined. gathered from multiple sources. A total of 9650 (total records Each personal air sample was assigned to one of 27 job in database) post-1978 samples were gathered from three titles and one of 29 work areas. In addition, the numerous electronic ExxonMobil databases: Personal Computer In- tasks conducted by workers at the refinery were consolidated dustrial Hygiene System (PCIHS, 1978–1998), Medgate into 16 task bins on the basis of the nature of the task. All (1999–2003), and EAS (2005-present). Information con- samples were classified by a refinery operational status of tained in these databases pertinent to our analysis included routine or turnaround. The protocol for assigning job title, the following: sample date, sample duration, sample type work area, and task assignments was reviewed by Exxon- (personal/area), analytical result, sample media, job title, Mobil industrial hygienists to ensure that samples were department, area and task associated with the sample. The properly characterized. Traditionally, worker exposure accuracy of the information contained in all three databases groups have been defined by ExxonMobil as job title-area- was verified through an independent and random review of task combinations. These worker exposure groups were the corresponding paper records associated with 25% of the evaluated to determine whether any of them could be air sample results. An error was considered critical if it combined to increase statistical power for a quantitative involved the sample result (concentration, laboratory result, assessment of exposures using an analysis of variance units, qualifier). All other errors, including typographical (ANOVA) model. The ANOVA was used to determine errors, were considered non-critical, given that they would whether the average benzene concentrations differed sig- not lead to miscalculation of the airborne concentration. nificantly by operational status, by work area for a given job Overall, based on the fact that there were fewer than 0.8% title (non-task samples), and by job title or work area for a critical errors in the database, it was concluded that the given task bin (task samples). database accurately reflected the original documentation. For Censored data (i.e., those samples less than the limit of any errors (typically transcription errors) found during the detection (LOD)) were included in the statistical analysis verification process, appropriate changes were made to the using the regression on order statistics (ROS) method database. (Helsel, 2005; U.S. EPA, 2007), which is equivalent to the In addition to the electronic databases, information robust log probit regression method presented in the IH regarding the refinery processes, job and task descriptions, literature (Hewett, 2007; Ignacio and Bullock, 2006). The benzene process changes, industrial hygiene surveys, and general approach of the ROS method includes fitting a linear exposure assessment initiatives relevant to the study were regression model of the detected values of the data set to the gathered from numerous sources. Industrial hygiene surveys quantiles of the assumed distribution (values from the y axis conducted over time throughout the refinery were reviewed of a probability plot) and replacing the values for samples less for both additional exposure data and details regarding jobs than the LOD with the values extrapolated from the linear and tasks. Furthermore, both historical and current indus- regression. As this data set has multiple limits of detection trial hygiene programs specific to benzene and respiratory associated with it, the robust ROS method developed by protection were reviewed to gain an understanding of Helsel and Cohn (1988) was used. This method was used

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 173 Gaffney et al. Exposures to benzene at reﬁneries

Ta bl e 2 . Description of job titles within the ExxonMobil reﬁnery in Baytown, TX that are represented in this database. The areas associated with each job title (as represented in the database) are presented as well.

Job title Description of duties Area(s)

ExxonMobil Process technician Responsible for process operations such as draining ALK; CC; CK; DES; HF; LAB; LEU; OM; PS; RF; SH; employees (n ¼ 2707) process equipment or lines prior to work and taking TF; UNK; UT; WT; LXU; DEW; Var; HDC; LBS; AS; (n ¼ 4830)a samples of process streams. CLEU;DEA;MEK;SO2;SFU;SHU Machinist (n ¼ 863) Equipment maintenance/repair on refinery units. ALK; CC; CK; DES; HF; LAB; LEU; OM; PS; RF; SH; TF; UNK; UT; WT; LXU; DEW; Var; HDC; HGU; LBS; GA; CLEU; DEA; MEK; SHU Laboratory technician Responsible for analyzing samples of process streams LAB; LBS (n ¼ 478) and performing other laboratory work. Pipefitter/welder Piping maintenance, including blinding and breaking ALK; CC; DES; HF; LEU; OM; PS; RF; SH; TF; UNK; (n ¼ 274) of lines. UT; WT; LXU; Var; HDC; LBS; CLEU; SHU Process supervisor Responsible for supervision of process technicians. CC; DES; HF; LEU; OM; PS; RF; WT; HDC; CLEU (n ¼ 191) Mobile equipment Responsible for operating motorized equipment, AD; HF; OM; RF; UNK; WT; GA operator (n ¼ 87) such as cranes, backhoes, or vacuum trucks Fuel truck driver/ Operates fuel truck throughout refinery. GA refueler (n ¼ 51) Electrician (n ¼ 33) Responsible for electrical work at the refinery. CC; LAB; OM; PS; RF; SH; UNK; CLEU; DEA Instrument technician Maintenance of instrumentation associated with units, CC; HF; LAB; LEU; OM; PS; RF; SH; UNK; CLEU; (n ¼ 33) including chromatographs and other equipment. SHU Garage mechanic Maintenance work on mobile equipment at the AD; SH; UNK; HDC; LBS; GA (n ¼ 30) refinery. Industrial hygienist Occupational health and safety including chemical CC; OM; RF; WT; MED (n ¼ 15) exposure monitoring. Maintenance supervisor Supervision of maintenance employees throughout HF; RF; WT; Var (n ¼ 7) refinery. Truck driver (n ¼ 5) Operates vehicles throughout refinery. GA Inspector (n ¼ 3)b Unit inspections and meter proving. OM; RF; UT Engineer (n ¼ 1) Responsible for supervision and design of refinery CK processes and mechanical functions Security officer (n ¼ 1)b Security operations throughout refinery. RF Contract Contractor F laborer Performance of unskilled maintenance tasks at the HF; OM; TF; WT; HDC employees (n ¼ 106) refinery, including sludge removal, filter replacement, (n ¼ 409) equipment flushing, and bundle cleaning. Contractor F Laboratory Analyze samples of process streams and perform LAB technician (n ¼ 61) other laboratory work. Contractor F pipefitter Piping maintenance, including blinding and CC; HF; RF; CLEU (n ¼ 57) breaking of lines. Contractor F catalyst Load catalyst into units or remove spent catalyst HF; RF (n ¼ 52) from units. Contractor F machinist Equipment maintenance/repair on refinery units. ALK; CC; WT; CLEU (n ¼ 45) Contractor F mobile Responsible for operating motorized equipment, OM; WT; LBS equipment operator such as cranes, backhoes, or vacuum trucks. (n ¼ 28) Contractor F gauging/ Gauge tanks or measure product temperature. OM; AS inspection (n ¼ 17)C Contractor F building Building/carpentry operations including general SH; WT trades (n ¼ 15) construction and scaffold building. Contractor F Maintenance of grounds. WT groundskeeper (n ¼ 12) Contractor F process Process operations such as draining process WT operations (n ¼ 11) equipment or lines before work and taking samples of process streams. Contractor F field Supervision of contract maintenance employees TF; WT supervisor (n ¼ 5) throughout refinery.

AD, administration; ALK, alkylation plant; AS, asphalt plant; CC, catalytic cracker; CK, coker; CLEU, catalytic light ends unit; DEA, deasphalting unit; DES, desulfurization; DEW, dewaxing area; GA, garage; HDC, hydrocracker; HF, hydroﬁner; HGU, hydrogen generation unit; LAB, laboratory; LBS, lube blending and storage; LEU, light ends unit; LXU, lube extraction unit; MED, medical; MEK, MEK unit; OM, oil movements; PS, pipestill; RF,

reformer; SFU, solvent fractionating unit; SH, shops; SHU, solvents hydrogenation unit; SO2,SO2 plant; TF, tank farm; UNK, unknown; UT, utilities; Var, various locations; WT, waste treatment. aIncludes 51 samples where job title is undefined due to insufficient data (42 non-task; 8 task; 1 time unknown). bThese job titles are only associated with task-specific samples.

174 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at reﬁneries Gaffney et al.

Ta bl e 3 . Description of task bins. The task bins were created to group similar tasks with similar exposure potential together, and were reviewed by industrial hygienists from ExxonMobil. Both the job titles and areas associated with each task (as represented in the database) are presented.

Task bin Abbre- Description Job titles(s) Area(s) viations

Sample collection- SP Collecting process sample from line/unit. Contractor F gauging/inspection; CK; DES; HF; LEU; OM; PS; RF; process (SP), process TF; UNK; UT; WT; LXU; Var; HDC; n ¼ 218 technician; security officer LBS; AS; CLEU; DEA; SFU; SHU Sample analysis- SL Tasks involved in laboratory analysis of Contractor F laboratory technician; LAB laboratory (SL), samples including washing of glassware; laboratory technician n ¼ 147 pouring of samples; disposing of samples; and other laboratory work. Liquid transfer (L), L Tasks involved in transfer of liquid product Contractor F laborer; contractor F AD; HF; OM; PS; RF; SH; UNK; n ¼ 137 or waste, including water draws; product mobile equipment operator; fuel truck WT; HDC; LBS; GA; AS loading driver/refueler; garage mechanic; or unloading; sludge removal; and vacuum machinist; mobile equipment operator; truck operations. process technician Blinding and B Installing blinds, breaking lines or opening Contractor F machinist; contractor F CC; DES; HF; LEU; OM; PS; RF; breaking (B), equipment. (including coke drums) and pipefitter; machinist; pipefitter/welder; SH; UNK; UT; LBS; CLEU; SHU n ¼ 97 installing or removing equipment. process technician Gauging (G), G Gauging tanks or barges to determine Contractor F gauging/inspection; OM; PS; TF; WT; LBS; AS n ¼ 65 liquid levels process technician Equipment cleaning ECR Maintaining or repairing units/equipment; Contractor F catalyst; inspector; CC; CK; DES; HF; OM; PS; RF; and repair (ECR), calibrating equipment or meters; changing machinist; pipefitter/welder; process SH; UNK; UT; MEK n ¼ 58 or cleaning filters and screens; and cleaning technician equipment. Equipment EP Draining, bleeding or blowing down lines Contractor F laborer; contractor F CC; DES; HF; LEU; OM; PS; RF; preparation (EP), or equipment; preparing equipment for pipefitter; machinist; pipefitter/welder; SH; WT; LXU; Var; HDC; MEK n ¼ 47 subsequent work. process technician Waste treatment WTM Maintaining or repairing units/equipment; Machinist; process technician; WT plant maintenance calibrating equipment or meters; changing contractor F laborer; contractor F (WTM), n ¼ 46 or cleaning filters and screens; and cleaning machinist; contractor F process equipment at the waste treatment plant. operations; pipefitter/welder Inspection/ INSP Unit inspections or bystander observation. Contractor F catalyst; contractor- OM; PS; RF; WT; Var; CLEU observation (INSP), gauging/inspection; contractor F n ¼ 21 machinist; contractor F process operations; machinist; process technician Separator skimmer SK Skimming the wastewater separator. Contractor F laborer; process OM; WT (SK), n ¼ 16 technician Painting (P), P Spray painting or silk screening of signs Electrician; instrument technician; SH n ¼ 10 and other objects. pipefitter/welder Repair Leak (RL), RL Repairing leaks in lines or equipment Contractor F laborer; machinist; HF; OM; PS; WT; LBS n ¼ 8 process technician Catalyst work CA Loading or removing catalyst. Contractor F catalyst RF (CA), n ¼ 5 Electrical ECRE Electrical maintenance and repair work. Electrician RF; SH maintenance and repair (ECRE), n ¼ 4 Tank farm blinding TFB Installs blinds at the motor gas and/or Pipefitter/welder TF and breaking (TFB), the LCN tank. n ¼ 4 Material transfer MT Tasks involved in the transfer of soil that Mobile equipment operator WT (MT), n ¼ 1 may have been contaminated with hydrocarbon.

AD, administration; ALK, alkylation plant; AS, asphalt plant; CC, catalytic cracker; CK, coker; CLEU, catalytic light ends unit; DEA deasphalting unit; DES, desulfurization; DEW, dewaxing area; GA, garage; HDC, hydrocracker; HF, hydroﬁner; HGU, hydrogen generation unit; LAB, laboratory; LBS ,lube blending and storage; LEU, light ends unit; LXU, lube extraction unit; MED, medical; MEK, MEK unit; OM, oil movements; PS, pipestill;

RF, reformer; SFU, solvent fractionating unit; SH, shops; SHU ,solvents hydrogenation unit; SO2,SO2 plant; TF, tank farm; UNK, unknown; UT, utilities; Var, various locations; WT, waste treatment; WTM, waste treatment plant maintenance.

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 175 Gaffney et al. Exposures to benzene at reﬁneries

F instead of the typical substitution method psubstitutingffiffiffi the blanks, instantaneous or grab samples, samples taken during LOD by the LOD divided by 2 or the 2Fbecause it emergency response or fire training, and any samples rejected produces fairly robust estimates of the mean and SD even by the ExxonMobil industrial hygienists for being invalid or with modest departures from the lognormal distribution and not representative of personal exposure were eliminated from even if 50–70% of the data are below the LOD (Huybrechts the data set (Figure 2). Of the remaining 9310 data points, et al., 2002; Lubin et al., 2004; Baccarelli et al., 2005; Ignacio 7402 were taken in the refinery and 1908 were collected at the and Bullock, 2006). docks or loading areas. Docks and loading area samples are The benzene concentration data for the task and non-task reported separately. Of the 7402 refinery samples, 1988 were data sets were tested for distribution fit using the Kolmogor- area samples, 5239 were personal samples, 18 were source- ov–Smirnov goodness-of-fit test for a normal, lognormal, specific samples, and 157 lacked sufficient data to classify and gamma distribution, and none of these distributions sample type. Source-specific or unclassifiable samples may (Po0.05) were found at a 95% confidence level. However, not be characteristic of typical exposures of refinery employ- the data were found to be approximately lognormal on the ees, and therefore not considered further in this analysis. basis of probability plots. Therefore, the robust ROS method The refinery data set used for the analysis contained 5239 for a lognormal distribution was used for these two data sets. personal samples, of which 4160 were considered non-task As these data sets were approximately lognormal, the samples (Z180 min) and 1079 were characterized as task natural log-transformed sample results including non-detect samples (o180 min) (Figure 2). Of the non-task samples, values estimated by the ROS model were used in the 3897 were taken during routine operations and 263 during ANOVA analyses. The Tukey multiple comparison test was turnarounds. Of the task samples, 1027 were taken during used to identify any significant pairwise differences between routine operations, 52 during turnaround operations. area for a given job title (non-task samples) and between job Air sampling for benzene at Baytown was conducted by title or area for a given task bin (task samples) at a 95% ExxonMobil industrial hygienists according to the standard confidence level (Neter et al., 1990). Job categories and task operating procedures involving the use of either charcoal bins characterizing potential differences by area (non-task sorbent tubes or passive organic vapor badges. Samples were and task samples) or job title and area (task samples only) analyzed by a laboratory accredited by the American were established on the basis of the results of the ANOVA Industrial Hygiene Association according to NIOSH or and post hoc pairwise analysis. other methods consistent with the internal standard operating Standard industrial hygiene descriptive summary statistics procedures (National Institute of Occupational Safety and including geometric SD, geometric mean and arithmetic Health, 2003). In general, over the past 20 years, data mean were calculated for the resulting job categories and task indicate that the majority of benzene concentrations were bins. However, the primary purpose of this study was to below the LOD. A detailed description of detection present comprehensive job category and task bin benzene frequencies and the limits of detection for each data set can concentration data for use in historical exposure reconstruc- befoundintheTable4. tion. Therefore, the results and discussion emphasize the arithmetic mean, which is considered the best metric for Non-Task Concentration Estimates in the Refinery Data Set worker dose (Ignacio and Bullock, 2006). The results of the non-task benzene concentrations by job To determine whether there was a trend over time by any category are presented in Table 5. Not surprisingly, the job category, pairwise comparisons were carried out to ANOVA and pairwise comparisons indicated that air identify statistical differences between samples collected from concentrations for five job titles (Process Technician, 1978 to 1989 and those collected from 1990 to 2006. The Machinist, Pipefitter/Welder, Instrument Technician, and task data set does not contain refinery personal samples from Contractor-Catalyst) were influenced by the area of the plant 2004 to 2006. The year 1990 was used as the cutoff point, as in which the employee was working. As such, separate job regulations intended to reduce employee exposure to benzene categories were created for each job title/area combination and to reduce benzene emissions to the ambient environment that had benzene concentrations that were significantly from petroleum refineries were mostly implemented by 1990 different than that particular job title with all areas combined. (OSHA, 1987; U.S. EPA, 1989). Concentrations of airborne benzene for non-task samples were analyzed for each job category during routine and turnaround operations. Of those job categories that had a Results sufficient sample size (nZ10), the highest mean benzene air concentration (3.4 p.p.m.) was associated with the pipefitter/ In total, 9650 samples were initially assembled to perform welder working during routine operations in the tank farm this analysis: 8845 from the ExxonMobil data bases and 805 (n ¼ 11). Of the 11 samples associated with this job category, from archived hard copy files. Before analysis, duplicate three had results 41 p.p.m.; these samples were collected in 1984 entries, samples lacking quantification for benzene, field during blinding and gas-freeing of the light cat naphtha tank.

176 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at reﬁneries Gaffney et al.

Excluded samples* Total Records n= 162 (1.7%) in Database *Includes samples coded or commented n=9,650 by IH as void, not representative,or rejected.

Miscellaneous Samples Other Records Chemical Plant (n=36; 0.4%) Duplicate entry (n=4; 0.0%) Refinery and Instantaneous or grab (n=31; 0.3%) Sample blank or test sample (n=15; 0.2%) Docks Samples FEDS (n=51; 0.5%) Sample not benzene (n=14; 0.1%) n=9,310 (96%) Incomplete concentration data (n=27; 0.3%)

Docks and Loading Unknown/Source/ Unknown/Source/ Refinery Rack Other Other n=7,402 (77%) n=1,908 (19%) n=175 (1.8%) n=79 (0.8%)

Area n=238 (2.5%) Area n=1,988 Personal Personal (20.6%) (<3 Hr TWA) (3-12 Hr TWA) n=572(5.9%) n=1,019 (10.6%) Personal Personal (<3 Hr TWA) (3-12 Hr TWA) n=1,079 (11.2%) n=4,160 (43.1%)

Figure 2. Exposure Monitoring Data Categorization (1978–2006): Breakdown of the data included in the database. For this analysis, only those personal samples associated with the reﬁnery were considered.

The airborne benzene concentration distributions for those and tank farms during routine operationsdrivesthestatistical job categories with sufficient data (nZ10) are presented in difference in benzene concentration between the operation Figure 3. The most frequently sampled job categories statuses (Table 5). This result was consistent with operational included those with the greatest potential for contact with practices that require draining/purging of liquid streams benzene-containing process streams, such as process techni- before the commencement of turnaround operations. There- cian, machinist, pipefitter-welder, and laboratory technician. fore, during much of the turnaround there is little material remaining in the equipment to generate significant exposure Influence of Operating Status to benzene. The ANOVA comparing benzene concentrations during routine and turnaround operations indicated that there Benzene Concentrations over Time was a statistical difference (Po0.05) between operational The distributions of personal airborne benzene concentra- statuses when all job categories were considered. Within job tions by job category with sufficient sample size (n410) categories, differences between operational status were not for the two time periods (1978–1989 and 1990–2006) are meaningful or were attributable to specific targeted activities, shown in Figure 4. Statistical differences (Po0.05) in mean such as catalyst skimming at the hydrofiner or hydraulic benzene concentration by time period were observed for excavation at the waste treatment plant. For most job cate- process technician (hydrofiner (1.5 and 0.010 p.p.m.), waste gories, potential differences in exposure during turnaround treatment (0.14 and 0.012 p.p.m.), and all other areas (0.070 could not be discerned because too few samples were and 0.010 p.p.m.)), machinist (waste treatment (0.37 and available for the job category or job categories in specific 0.019 p.p.m.) and all other areas (0.10 and 0.012 p.p.m.)), areas. One exception was the Process Technician category laboratory technician (0.12 and 0.13 p.p.m.), and mobile (n450, for routine and turnaround groups), where the mean equipment operator (0.23 and o0.11 p.p.m.) during routine benzene concentration for all areas combined is lower during operations. For all of these job categories, mean benzene air turnaround operations (n ¼ 102, mean ¼ 0.067 p.p.m.) than concentrations were statistically significantly lower for the routine operations (n ¼ 2021, mean ¼ 0.27 p.p.m.). Data time period from 1990 to 2006 than for 1978–1989 with the collected in areas such as the hydrofiner, oil movements, exception of the laboratory technician (Figure 4). In this

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 177 Gaffney et al. Exposures to benzene at reﬁneries

Ta bl e 4 . Detection frequency and average limit of detection for non-task and task samples in the reﬁnery, docks, and loading rack data sets.

Data set n Detection frequency Benzene concentration (p.p.m.) (%) Geometric mean Arithmetic mean Minimum Maximuma (Geometric SD)

Task 1079 36 0.15 (3.0) 0.32 0.0010 9.0 Non-task 4160 41 0.050 (2.6) 0.077 0.0010 1.5 aNine task sample detection limits exceeded 2.5 p.p.m.; five non-task sample detection limits exceed 0.5 p.p.m. case, the simple arithmetic mean for the laboratory technician statuses when all task bins were considered. However, for increased over time. Although the simple arithmetic mean is most task bins, potential differences in exposure during an appropriate and unbiased predictor of the true mean, for turnaround could not be discerned because too few samples subsets of data with large sample sizes and large geometric were available for the task bin. Nonetheless, task level SDs, such as this one, it is less applicable to the ANOVA samples were analyzed separately on the basis of operations comparison than other estimates (Ignacio and Bullock, status (Figure 5). 2006). Reanalysis of the arithmetic mean for this group on the basis of the lognormal minimum variance unbiased Benzene Concentrations over Time estimator revealed a decrease over time for the laboratory The distributions of personal airborne benzene concentra- technician from 0.13 to 0.02 p.p.m. tions by task bins with statistical differences (Po0.05) in mean benzene concentration for the two time periods (1978– Task Exposure Concentration in the Refinery Data Set 1989 and 1990–2003) are shown in Figure 6. For those task The average benzene air concentrations and related summary bins with sufficient sample size (nZ10), a statistical difference statistics for task samples are summarized in Table 6. The in mean benzene concentration by time period was observed ANOVA and pairwise comparisons indicated that the area in for sample collection (0.87 and 0.079 p.p.m.) and sample which these tasks were performed did not influence the analysis (4.6 and 0.10 p.p.m.). Although the repair leak task benzene concentrations associated with the tasks. In addition, bin only had a sample size of eight, it too was statistically the job title of the individual performing the task appears to significant by time period (36 and 0.037 p.p.m.). For these have had no influence on the benzene concentrations task bins, benzene air concentrations were statistically associated with the tasks. significantly lower for the time period from 1990 to 2003 Of the task bins with sufficient sample size (nZ10), the than for 1978–1989. laboratory sample analysis bin had the highest mean benzene air concentration (1.9 p.p.m.). To perform this task, the technician must work directly with small quantities of process Discussion stream samples, and work is usually performed under controlled conditions. However, this result is driven by nine This study presents the results of 25 years of personal air sample results above 10 p.p.m. that were collected in 1988 monitoring for benzene at the Baytown, TX ExxonMobil during the filtration of reagent grade benzene for saybolt refinery. Air concentrations of benzene collected while color test while wearing powered air-purifying respirators. performing specific jobs and tasks were found to be consistent The distributions of benzene air concentrations for those with the results reported in other studies of benzene exposure task bins with sufficient data (nZ10) are presented in in the petroleum industry (Weaver et al., 1983; Buchet et al., Figure 5. The tasks with a potential for contact with a 1984; Runion and Scott, 1985; CONCAWE, 1987, 1994; process stream were the most frequently sampled, and Nordlinder and Ramnas, 1987; Rappaport et al., 1987; HEI, include the following: sample collection (n ¼ 218), sample 1988; Verma et al., 2001). When comparing the results of this analysis (n ¼ 147), liquid transfer (n ¼ 137), blinding and study to the industry-wide data sets, the average exposures breaking (n ¼ 97), gauging (n ¼ 65), equipment cleaning and at the Baytown refinery decrease within the range of repair (n ¼ 58), equipment preparation (n ¼ 47), and waste those reported in the literature, despite the fact that treatment plant maintenance (n ¼ 46). the Baytown data set was targeted to over-represent benzene handling activities (Verma et al., 2001). The average Influence of Operating Status benzene air concentration at the Baytown refinery for all The ANOVA that compared benzene concentrations during non-task samples is 0.23 p.p.m. (during routine operations), routine and turnaround operations indicated that there compared with the arithmetic means reported in the was a statistical difference (Po0.05) between operational literature, which range from 0.05 to 1.62 p.p.m., with an

178 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at reﬁneries Gaffney et al.

Ta bl e 5 . Summary statistics for the non-task data set for (a) ExxonMobil employees, (b) contractors, including job title, operational status, and area.

Job title Status Area(s) n Detection Geometric mean, geometric Minimum Maximum frequency SD, arithmetic mean, detected detected (%) 75th, and 95th percentile benzene concentration (p.p.m.)

(a) Process technician Hydroﬁnera 245 49 ND, ND, 1.3, 0.20, 1.0 0.015 91 Oil movementsb 286 24 ND, ND, 0.23, ND, 0.21 0.010 49 Routine Reformerc 575 46 ND, ND, 0.10, 0.20, 0.24 0.010 2.0 Tank farm d 9 100 0.35, 2.6, 0.65, ND, ND 0.15 3.6 Waste Treatment e 446 52 0.045, 5.6, 0.13, 0.20, 0.40 0.013 1.6 All other areasf 460 35 ND, ND, 0.062, 0.055, 0.20 0.010 6.0 Hydrocracker 9 100 0.11, 1.8, 0.13 ND, ND 0.060 0.42 Turnaround Catalytic cracker 9 22 ND, ND, 0.020, ND, ND 0.025 0.094 All other areas 84 50 0.022, 5.1, 0.065, 0.072, 0.26 0.0070 0.44 Shopsg 68 7 ND, ND, 0.13, ND, 0.15 0.021 5.6 Routine Waste treatmenth 228 72 0.10, 6.6, 0.35, 0.40, 0.93 0.026 5.8 Machinist All other areasi 421 27 ND, ND, 0.089, 0.042, 0.34 0.0090 7.7 Turnaround Waste treatment 3 100 0.40, 1.1, 0.40, ND, ND 0.34 0.44 All other areas 15 20 ND, ND, 0.061, ND, 0.21 0.020 0.76 Laboratory technician Routine All areasj 352 37 ND, ND, 0.12, 0.040, 0.34 0.0035 12 Routine Tank farmk 11 73 0.33, 14, 3.4, 1.8, 15 0.12 16 Pipeﬁtter/welder All other areasl 182 21 ND, ND, 0.055, ND, 0.23 0.0090 2.0 Turnaround All areas 13 62 0.018, 3.5, 0.034, 0.058, 0.081 0.020 0.12 Process supervisor Routine All areasm 179 30 ND, ND, 0.050, 0.039, 0.23 0.010 1.2 Turnaround All areas 11 27 ND, ND, 0.038, 0.045, 0.12 0.057 0.13 Mobile equipment operator Routine All areasn 76 55 0.046, 7.3, 0.20, 0.22, 0.66 0.019 3.2 Turnaround All areaso 6 100 0.24, 4.6, 0.48, ND, ND 0.017 1.6 Fuel truck driver/refueler Routine All areasp 33 76 0.16, 6.5, 0.41 0.53, 1.3 0.10 1.9 Instrument technician Routine Shopsq 1 100 39, ND, 39, ND, ND 39 39 All other areas 29 34 ND, ND, 0.016, 0.020, 0.039 0.010 0.040 Electrician Routine All areas 27 48 ND, ND, 0.016, 0.020, 0.037 0.010 0.035 Turnaround All areas 1 0 ND, ND, o0.02, ND, ND ND ND Garage mechanic Routine All areas 17 53 0.026, 5.6, 0.073, 0.080, 0.26 0.021 0.35 Industrial hygienist Routine All areas 8 100 0.14, 3.1, 0.24, ND, ND 0.040 0.83 Turnaround All areas 2 50 0.0069 15, 0.024, ND, ND 0.047 0.047 Maintenance supervisor Routine All areas 3 0 ND, ND, o0.06 ND, ND ND ND Turnaround All areasr 4 75 0.17, 8.4, 0.44, ND, ND 0.17 1.0 Truck driver Routine All areas 5 80 0.17, 3.3, 0.27, ND, ND 0.18 0.69 Engineer Turnaround All areas 1 0 ND, ND, o0.04, ND, ND ND ND

(b) Contractor F laborer Routine All areass 78 64 0.063, 7.7, 0.24, 0.28, 1.0 0.030 2.1 Turnaround All areast 5 60 0.076, 12, 0.70, ND, ND 0.055 3.2 Reformer 35 63 0.012, 3.7, 0.023, 0.036, 0.067 0.0070 0.088 Contractor F catalyst Turnaround Hydrofineru 9 78 1.4, 22, 19, NDv ND 1.1 103 Contractor F pipefitter Turnaround All areas 41 20 ND, ND, 0.036, ND, 0.071 0.0080 0.88 Contractor F machinist Routine All areas 24 8 ND, ND, 0.015, ND, 0.048 0.050 0.055 Turnaround All areas 11 9 ND, ND, 0.012, ND, 0.022 0.022 0.022 Contractor F laboratory Routine All areasv 34 47 ND, ND, 0.56, 0.29, 2.3 0.042 6.1 technician Contractor F mobile Routine All areas 23 43 ND, ND, 0.073, 0.053, 0.31 0.023 0.45 equipment operator Contractor F building trades Routine All areas 14 7 ND, ND, 0.0044, ND, 0.012 0.012 0.012 Contractor F groundskeeper Routine All areas 12 0 ND, ND, o0.02, ND, ND ND ND Contractor F process Routine All areasw 8 100 0.38, 2.0, 0.48, ND, ND 0.15 1.3 operations Contractor F field Routine All areasx 4 75 0.086, 10.2, 0.41, ND, ND 0.020 1.4 supervisor

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 179 Gaffney et al. Exposures to benzene at reﬁneries

Ta bl e 5 . Continued

Job title Status Area(s) n Detection Geometric mean, geometric Minimum Maximum frequency SD, arithmetic mean, detected detected (%) 75th, and 95th percentile benzene concentration (p.p.m.)

All job titles Routine All areas 3897 40 ND, ND, 0.23, 0.11, 0.44 0.0035 91 Turnaround All areas 263 46 ND, ND, 0.73, 0.060, 0.51 0.0070 103 All job titlesy All Samples All areas 4160 41 ND, ND, 0.26, 0.10, 0.44 0.0035 103

ND: geometric mean not calculated because 50% of the samples were below the limit of detection. Percentiles not calculated because the proportion of results below the detection limit exceeds the selected percentile or the sample size was less than 10. aFourteen sample results above 1 p.p.m. collected between 1978 and 1980 at the naphtha fractionation unit (NFU) post; task was undefined or indicated routine work. bThree sample results above 1 p.p.m. collected in 1994, task was routine work including gauging and sample collection. cTwo sample results above 1 p.p.m. were collected in 1980 and 1981; no additional information was available. dMaximum sample result collected in 1978; no additional information was available. eFour sample results above 1 p.p.m. collected between 1979 and 1982 at the waste treatment retention basin or separator posts. fThree samples above 1 p.p.m. collected at a MEK unit post (n ¼ 1; 1981) and light ends unit posts (n ¼ 2; 1978). gMaximum sample result collected in 1978 during boiler house maintenance. hTen sample results above 1 p.p.m. collected between 1978 and 1983 during cleaning and maintenance of the waste treatment plant preseparator. iTwo sample results above 1 p.p.m. collected in 1984 during pump slab repair work in the oil movements area (liquid was steam cracked naphtha). jMaximum sample result collected in 2005; no additional information was available. kThree samples greater than 1 p.p.m. collected in 1984 during blinding and gas-freeing of light cat naphtha (LCN) tank. lMaximum sample result collected in 1984 during work in multiple areas including blinding at the fuel blend manifold and the LCN tank. mMaximum sample result collected in 1981 at the hydroskimmer; no additional information was available. nTwo sample results above 1 p.p.m. were collected in 1979 and 1985, during refueling and operation of a hydraulic excavator at the waste treatment preseparator, respectively. oMaximum sample result collected in 1984 during operation of a hydraulic excavator at the waste treatment preseparator. pThree samples above 1 p.p.m. where collected in 1978 and 1980; no additional information was available. qMaximum sample result collected in 1983; task and unit were undefined but technician stated that a supplied air respirator was used during periods of high hydrocarbon exposure. rMaximum sample result collected in 1993 during supervision of catalyst skimming operation at the hydrofiner. sFive sample results above 1 p.p.m. collected between 1988 and 1996 during oil movements separator skimmer maintenance (n ¼ 3) or waste treatment unit maintenance (n ¼ 2). tMaximum sample result collected in 1993 during scaffold erection at the hydrofining unit. uSeven sample results above 1 p.p.m. were collected in 1993 during catalyst skimming at the hydrofining unit. vSix sample results above 1 p.p.m. were collected in 2005 during sampling activities in the drum retain area. wMaximum sample result collected in 1989 during slop oil tank filtration at waste treatment. xMaximum sample result collected in 1988 during supervision at the tank farm; no additional information was available. yIncludes 43 samples where job title could not be assigned.

average of 0.22 p.p.m. (Runion and Scott, 1985; Verma Analysis by refinery area indicates that with the et al., 2001). exceptionoffivejobcategories,theareainwhichan Fifty-nine percent of the non-task samples showed benzene individual worked did not influence the average benzene air air concentrations below the LOD. Given the large number concentration. These findings are similar to data from other of samples in the data set and the targeted nature of the refinery studies, and demonstrate that the nature of the sampling program (sampling is focused on jobs where direct work being performed is a major determinant of overall contact with benzene-containing streams is most likely), the exposure to benzene (Verma et al., 2001). However, because low detection frequency is indicative of generally overall low there are often very few samples associated with a job concentrations of benzene in the air. Specifically, nearly 98% category and area, the absence of any statistical difference of the non-task sample results were less than 1 p.p.m., a may be a function of a small sample size. For the Baytown finding consistent with an industry-wide study indicating that refinery job categories that displayed area dependence 95% of benzene results from air samples taken at several US during routine operations, the sample sizes were large refineries were below 1 p.p.m. (Runion and Scott, 1985). (n ¼ 30–2021), and the results were consistent with expecta- Furthermore, 65 of the samples classified as non-task in these tions; for example, in areas where the benzene content of the data set had unknown sampling times, and were assumed to process stream was low, occupational exposure was low, and represent samples collected for over 180 min. This is a vice versa. conservative assumption, because any error in the analysis The analysis of air concentration trends over time associated with this assumption would overestimate indicated that for most job categories, exposures were slightly long-term exposure. higher in the 1978–1989 time period than in 1990–2003, and

180 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at reﬁneries Gaffney et al.

100 95th Percentile-Routine Status 95th Percentile-Turnaround Status Arithmetic Mean 10

0.1

Benzene Concentration (ppm) 0.01

0.001

Electrician Garage Mech Machinist-shops Laboratory TechPipefitter/Welder ProcessProcess Supervisor SupervisorFuel Truck Diver Contractor - Laborer Contractor -Contractor-Machinist Pipefitter Process Tech-Reformer Mobile Equip Operator Contractor - Machinist Process Tech-Hydrofiner Machinist-AllMachinist-All other otherareas areas Machinist-Waste Teatment Process Tech-Oil MovementsProcessProcess Tech-All Tech-All other areasother areas Pipefitter/welder-Tank Farm Contractor -Contractor Laboratory - BuildingTech Trades Process Tech-Waste Treatment Pipefitter/Welder-All other areas Instrument Tech-All Contractorother areas - Catalyst - Reformer Contractor - Mobile Equip Operator Figure 3. Distribution of non-task benzene air concentrations by job category. This ﬁgure demonstrates the overall distribution of air concentrations of benzene for those job categories with greater than 10 samples.

10 95th Percentile (1990-2006) 95th Percentile (1978-1989) Arithmetic Mean

Laboratory Mobile Equipment Process Technician Machinist Technician Operator 1

† †† 0.1

Benzene Concentration (ppm) 0.01

0.001 Hydrofiner* Waste Treatment*All other areas* Waste Treatment* All other areas* All areas* All areas* Figure 4. Distribution of non-task benzene air concentrations over time by job category. This ﬁgure demonstrates the overall distribution of air concentrations of benzene for each window of time (1978–1989 and 1990–2006). Distributions are presented by job category, and are limited to those job categories with greater than 10 samples in each time window. *Denotes statistical signiﬁcance by time period (Po0.05). wCalculation of the MVUE of arithmetic mean revealed a decrease over time for the laboratory technician from 0.13 p.p.m to 0.02 p.p.m. ww All results for samples collected between 1990–2006 (n ¼ 12) were below detection limit; mean detection limit plotted.

were statistically signiﬁcantly higher for three of the job the changes occurred could not be discerned from the categories (Figure 4). These results are consistent with relatively small data sets available. expectations, given the general time period in which various The benzene air concentrations associated with the 16 task process changes were implemented. However, decreases in bins were largely non-detectable. Only two of ten task bins airborne concentrations of benzene in the speciﬁc areas where (nZ10) had a detection frequency greater than 50%.

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 181 Gaffney et al. Exposures to benzene at reﬁneries

Ta bl e 6 . Summary statistics for the task data set, including job title and area.

Job title Status Area(s) n Detection Benzene concentration (p.p.m.) frequency (%) Geometric Geometric Arithmetic mean, 75th, Minimum Maximum mean SD and 95th percentile detected detected

Process technician Hydrofinera 245 49 ND ND 1.3, 0.20, 1.0 0.015 91 Oil Movementsb 286 24 ND ND 0.23, ND, 0.21 0.010 49 Routine Reformerc 575 46 ND ND 0.10, 0.20, 0.24 0.010 2.0 Tank farm d 9 100 0.35 2.6 0.65, ND, ND 0.15 3.6 Waste treatmente 446 52 0.045 5.6 0.13, 0.20, 0.40 0.013 1.6 All other areasf 460 35 ND ND 0.062, 0.055, 0.20 0.010 6.0 Hydrocracker 9 100 0.11 1.8 0.13, ND, ND 0.060 0.42 Turnaround Catalytic cracker 9 22 ND ND 0.020, ND, ND 0.025 0.094 All other areas 84 50 0.022 5.1 0.065, 0.072, 0.26 0.0070 0.44 Shopsg 68 7 ND ND 0.13, ND, 0.15 0.021 5.6 Routine Waste Treatmenth 228 72 0.10 6.6 0.35, 0.40, 0.93 0.026 5.8 Machinist All other areasi 421 27 ND ND 0.089, 0.042, 0.34 0.0090 7.7 Turnaround Waste treatment 3 100 0.40 1.1 0.40, ND, ND 0.34 0.44 All other areas 15 20 ND ND 0.061, ND, 0.21 0.020 0.76 Laboratory technician Routine All areasj 352 37 ND ND 0.12, 0.040, 0.34 0.0035 12 Routine Tank Farmk 11 73 0.33 14 3.4, 1.8, 15 0.12 16 Pipefitter/welder All other areasl 182 21 ND ND 0.055, ND, 0.23 0.0090 2.0 Turnaround All areas 13 62 0.018 3.5 0.034, 0.058, 0.081 0.020 0.12 Process supervisor Routine All areasm 179 30 ND ND 0.050, 0.039, 0.23 0.010 1.2 Turnaround All areas 11 27 ND ND 0.038, 0.045, 0.12 0.057 0.13 Mobile equipment operator Routine All areasn 76 55 0.046 7.3 0.20, 0.22, 0.66 0.019 3.2 Turnaround All areaso 6 100 0.24 4.6 0.48, ND, ND 0.017 1.6 Fuel truck driver/refueler Routine All areasp 33 76 0.16 6.5 0.41, 0.53, 1.3 0.10 1.9 Instrument technician Routine Shopsq 1 100 39 ND 39, ND, ND 39 39 All other areas 29 34 ND ND 0.016, 0.020, 0.039 0.010 0.040 Electrician Routine All areas 27 48 ND ND 0.016, 0.020, 0.037 0.010 0.035 Turnaround All areas 1 0 ND ND o0.02, ND, ND ND ND Garage mechanic Routine All areas 17 53 0.026 5.6 0.073, 0.080, 0.26 0.021 0.35 Industrial hygienist Routine All areas 8 100 0.14 3.1 0.24, ND, ND 0.040 0.83 Turnaround All areas 2 50 0.0069 15 0.024, ND, ND 0.047 0.047 Maintenance supervisor Routine All areas 3 0 ND ND o0.06, ND, ND ND ND Turnaround All areasr 4 75 0.17 8.4 0.44, ND, ND 0.17 1.0 Truck driver Routine All areas 5 80 0.17 3.3 0.27, ND, ND 0.18 0.69 Engineer Turnaround All areas 1 0 ND ND o0.04, ND, ND ND ND aFourteen sample results above 1 p.p.m. collected between 1978 and 1980 at the Naphtha Fractionation Unit (NFU) post; task was undefined or indicated routine work; bThree sample results above 1 p.p.m. collected in 1994, task was routine work including gauging and sample collection; cTwo sample results above 1 p.p.m. were collected in 1980 and 1981; no additional information was available; dMaximum sample result collected in 1978; no additional information was available; eFour sample results above 1 p.p.m. collected between 1979 and 1982 at the waste treatment retention basin or separator posts; fThree samples above 1 p.p.m. collected at a MEK unit post (n ¼ 1; 1981) and light ends unit posts (n ¼ 2; 1978); gMaximum sample result collected in 1978 during boiler house maintenance; hTen sample results above 1 p.p.m. collected between 1978 and 1983 during cleaning and maintenance of the waste treatment plant preseparator; iTwo sample results above 1 p.p.m. collected in 1984 during pump slab repair work in the oil movements area (liquid was steam cracked naphtha); jMaximum sample result collected in 2005; no additional information was available; kThree samples greater than 1 p.p.m. collected in 1984 during blinding and gas-freeing of light cat naphtha (LCN) tank; lMaximum sample result collected in 1984 during work in multiple areas including blinding at the fuel blend manifold and the LCN tank; mMaximum sample result collected in 1981 at the hydroskimmer; no additional information was available; nTwo sample results above 1 p.p.m. were collected in 1979 and 1985, during refueling and operation of a hydraulic excavator at the waste treatment preseparator, respectively; oMaximum sample result collected in 1984 during operation of a hydraulic excavator at the waste treatment preseparator; pThree samples above 1 p.p.m. where collected in 1978 and 1980; no additional information was available; qMaximum sample result collected in 1983; task and unit were undefined but technician stated that a supplied air respirator was used during periods of high hydrocarbon exposure; rMaximum sample result collected in 1993 during supervision of catalyst skimming operation at the hydrofiner;

182 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at reﬁneries Gaffney et al.

100 95th Percentile-Routine Status 95th Percentile-Turnaround Status Arithmetic Mean 10

0.1

Benzene Concentration (ppm) 0.01

0.001

Gauging

Liquid Transfer Sample Analysis Sample Collection Separator Skimmer Blinding andBlinding Breaking and Breaking EquipmentEquipment Preparation PreparationInspection/Observation

Waste Treatment Maintenance Equipment Cleaning and Repair Figure 5. Distribution of task benzene air concentrations by task bin. This ﬁgure demonstrates the overall distribution of air concentrations of benzene for those task bins with greater than 10 samples.

100 Each task bin was created to characterize subsets of all 95th Percentile (1990-2003) tasks performed at the refinery, with the intention of 10 95th Percentile (1978-1989) Arithmetic Mean including in each task bin activities sharing a similar exposure n=3 (1978-1989) potential. As each job category is defined by the types of 1 activities that an individual may perform, each task bin has only a few job titles associated with it. Nevertheless, there 0.1 was no difference between job titles within each task bin, 0.01 n=5 indicating that the tasks were the primary determinant of (1990-2003) exposure. The analysis of the task bin by area further Benzene Concentration (ppm) 0.001 supports this concept, as it also indicated no statistical Sample Collection* Sample Analysis* Repair Leak* difference by area. Figure 6. Distribution of benzene air concentrations over time by task. Task data can also be used to gain a better understanding This figure demonstrates the overall distribution of air concentrations of exposure by job title or area. In this data set, the ANOVA of benzene for each window of time (1978–1989 and 1990–2003). Distributions are presented by task bin, and are limited to those bins and pairwise comparisons for non-task concentration with greater than 10 samples in each time window. estimates indicated that operational status and area of the *Denotes statistical significance by time period (Po0.05). plant was a determinant of exposure. By examining these results, it can be seen that in some instances, while the Overall, the results from the analysis of the task-level data concentrations are low, the variability within a given job indicated that tasks that may require interaction with title, area, and operational status combination is appreciable. process streams, such as sample collection, sample analysis, In these instances, it is not possible to ascertain which liquid transfer, repairing leaks or blinding, and breaking in activities might be associated with the underlying cause of tank farms generally have higher air concentrations than variability without evaluating the task data. For example, the those where exposure is minimal (Table 6). These results air concentrations associated with the process technician in reinforce the historic focus of the benzene exposure control the waste treatment area under routine conditions was program described earlier. Of the 1079 task level samples, characterized by a geometric mean of 0.045 p.p.m., geo- only 39 (3.6%) had results greater than 5 p.p.m. In metric SD of 5.6, and an arithmetic mean of 0.13 p.p.m. By comparison, in a study of European refineries, 14% of the reviewing the tasks performed by process technicians and benzene air concentrations associated with the tasks con- associated job titles and areas, it can be seen that waste ducted by refinery operators exceeded 5 p.p.m. (CON- treatment plant maintenance task, with a geometric mean of CAWE, 1994). 0.25 p.p.m., geometric SD of 7.2 p.p.m., and arithmetic

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 183 Gaffney et al. Exposures to benzene at refineries mean of 0.85 p.p.m., could be an appreciable contributor to determining potential benzene exposure; rather, most of the variability. As process technicians only rarely conduct this literature has focused only on one variable, such as refinery task, this potentially important determinant of exposure area or job. Furthermore, within the review of the data, an would not have been obvious by review of the non-task exhaustive independent review of the data was conducted to samples alone. ensure data quality and accuracy with respect to original documentation. Of the existing published literature, no other Targeted Monitoring Strategy and Use of the Data studies have ensured this level of data quality review on such According to the ExxonMobil monitoring strategy and a large body of data. standard industrial hygiene practice, air sampling for benzene The analysis by task bin is unique to this study; other is focused on jobs where there is a potential for exposure refinery studies have focused primarily on long-term based on the nature of the work or the work area. This exposure estimates. Although using task-based exposure sampling approach is evident in the number of samples taken levels in epidemiology has its limitations (Cherrie, 1996: for each job, area, and task. Those areas with process streams Verma et al 2001), which summarized data from several that contained the highest percentage of benzene were the studies evaluating benzene exposure to employees in most frequently sampled; similarly, the jobs and tasks with petroleum industries, explicitly recommend moving toward the highest likelihood of exposure to process streams task-based exposure assessments as opposed to long-term containing benzene were most frequently sampled (see time weighted average estimates for the petroleum industry Tables 1–3). (Verma et al., 2001). This study addresses the recommenda- On the basis of the targeted sampling approach used tion by Verma et al. (2001) by providing a detailed task-level by the industrial hygienists at this refinery, the non-task analysis of benzene air concentrations at the Baytown averages calculated for each job category may not represent refinery. In addition, because analysis of the task level data long-term (over the span of months or years) averages for suggests that exposure to benzene is driven by the task or by all job titles. For those job categories where the individuals the work being performed (as opposed to other factors such have regular contact with process streams (e.g., process as area, job, or operational status); the results from this study technicians, pipefitter-welders, and so on) and are therefore support that observation. more likely to be targeted for sampling, these daily averages Although this study improves upon the currently available maybemorerepresentativeoflong-term averages. However, literature, it is not without limitations. Although this facility for those job categories where individuals do not have regular has been actively involved in assessing worker exposures to contact with process streams (e.g., instrument technicians, benzene since the 1950s, only limited documentation of the mobile equipment operators, and so on), the benzene air results before 1978 were available. As a result, quantification concentrations are more likely to represent the unique of workers’ exposures before 1978 will require some activities being performed on the sampling day, rather than extrapolation from the current data set (Proctor et al., those activities being performed on an average day, and 2004). Similarly, industrial hygiene benzene data exist for therefore the daily averages, as predicted by the data 1978–2006; however, although personal samples were presented here, would not be representative of long-term collected from 2004 to 2006, task-related samples were averages. unable to be identified as such in the data base, necessitating As of the targeted nature of the monitoring strategy, care extrapolation from 2004 to the present. In addition, samples must be taken when using the data presented in this analysis were not collected every year for each job category. to estimate employee exposures to benzene at this refinery. As Interpolation will therefore be necessary to estimate for some job categories, the non-task daily averages may not exposures for specific job categories during years when data accurately represent long-term averages, it is necessary to were not collected. Finally, respiratory protection factors or reconstruct these exposures on a case-by-case basis. This adjustments have not been applied to the measured air reconstruction of daily exposures may require using task- concentrations, although, on the basis of the available based data, time-activity patterns, and exposure frequencies. documentation, respiratory protection is part of the standard Furthermore, by reconstructing an individual’s exposure on a work practice for some tasks, and should be taken into case-by-case basis, considerations can be made, if appro- account when using the data for exposure reconstruction. priate, for respirator use.

Strengths and Limitations Conclusions This study provides analysis of a robust data set of benzene air samples, and considers several factors: operational status The data presented here indicate that the benzene air of the refinery; area of the refinery, job title, and task concentrations at the ExxonMobil refinery in Baytown, performed. Of the studies available in the existing literature, TX, are generally low for both non-task and task-based none have attempted to consider all of these factors in sampling events. When considering these data for exposure

184 Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) Exposures to benzene at reﬁneries Gaffney et al.

reconstruction, it is especially important to bear in mind the Glass D.C., Spurgeon A., Calvert I.A., Clark J.L., and Harrington J.M. nature of the samples analyzed in this data set, which were Retrospective assessment of solvent exposure in paint manufacturing. Occup Environ Med 1994: 51(9): 617–625. primarily collected using a targeted strategy. As such, these HEI. Gasoline Vapor Exposure and Human Cancer: Evaluation of Existing Scientific data are likely to represent the upper end of occupational Information and Recommendations for Future Research. Cambridge, MA: exposure distributions at the refinery, rather than air Health Effects Institute (HEI), 1998. Helsel D. More than obvious: better methods for interpreting nondetect data. concentrations representative of typical exposures. This study Environ Sci Technol 2005: 39: 419A–423A. demonstrates the necessity of analyzing refinery benzene Helsel D.R., and Cohn T. Estimation of descriptive statistics for multiply censored exposures on the task level, and is useful for understanding water quality data. Water Resour Res 1988: 24: 1997–2004. HendersonW.,andBenjaminS.Standard Oil: The First 125 Years. Motorbooks which activities drive refinery worker exposures to benzene. International, New York, 1996. pp 12. Hewett P. User Guide for IH Data Analyst. Available from: http://www.oesh.com/ x%20Software/IHDA.php. 2007. Huybrechts T., Thas O., Dewulf J., and Van Langenhove H. How to estimate Conflict of interest moments and quantiles of environmental data sets with non-detected observations? A case study on volatile organic compounds in marine water This work was funded by ExxonMobil, a firm that has been samples. JChrom2002: 975: 123–133. involved in the study of benzene for several decades. At least Ignacio J.S., and Bullock W.H., (eds.) A Strategy for Assessing and Managing Occupational Exposures. AIHA Press, Fairfax, VA, 2006. pp 415–421. two of the authors have served or are likely to serve as expert Lubin J.H., Colt J.S., Camann D., Davis S., Cerhan J.R., Severson R.K., witnesses for ExxonMobil on matters relating to industrial Bernstein L., and Hartge P. Epidemiologic evaluation of measurement data hygiene, exposure assessment, risk assessment, or toxicolo- in the presence of detection limits. Environ Health Perspect 2004: 112(17): 1691–1696. gical issues related to benzene. Meyers R.A., (ed.). Handbook of Petrolem Refining Processes. McGraw-Hill, New York, 2004. National Institute of Occupational Safety and Health. NIOSH manual of analytical methods: aromatic Hydrocarbons, Method 1501. 4th edn. 2003. Acknowledgements Neter J., Wasserman W., and Kutner M.H. Applied Linear Statistical Models,3rd edn. Irwin, Burr Ridge, IL, 1990. We thank Paul Scott of ChemRisk for his assistance with the Nordlinder R., and Ramnas O. Exposure to benzene at different work places in Sweden. Ann Occup Hyg 1987: 31(3): 345–355. statistical analysis and Matt Le and Kathleen Navarro for their OSHA 29 CFR 1910.1028: Benzene 1987. assistance in assembling and organizing the electronic data. OSHA. OSHA technical manual: petroleum refining processes. Vol Section IV, Chapter 2. Occupational Safety and Health Administration (OSHA) 2003. Panko J.M., Gaffney S.H., Burns A.M., Unice K.M., Kreider M.L., Booher L.E., Gelatt R.H., Marshall J.R., and Paustenbach D.J. Occupational References exposure to benzene at the ExxonMobil refinery at Baton Rouge, Louisiana (1977–2005). J Occup Environ Hyg 2009: 6(9): 517–529. Armstrong T.W., Pearlman E.D., Schnatter A.R., Bowes S.M., Murray N., and Proctor D.M., Panko J.P., Liebig E.W., and Paustenbach D.J. Estimating Nicolich M.J. Retrospective benzene and total hydrocarbon exposure historical occupational exposure to airborne hexavalent chromium in a chromate assessment for a petroleum marketing and distribution worker epidemiology production plant: 1940–1972. JOccupEnvironHyg2004: 1(11): 752–767. study. Am Ind Hyg Assoc J 1996: 57: 333–343. Rappaport S.M., Selvin S., and Waters M.A. Exposures to hydrocarbon Baccarelli A., Pfeiffer R., Consonni D., Pesatori A.C., Bonzini M., Patterson components of gasoline in the petroleum industry. Appl Ind Hyg 1987: 2(4): D.G., Bertazzi P.A., and Landi M.T. Handling of dioxin measurement data in 148–154. the presence of non-detectable values: overview of available methods and their Runion H.E., and Scott L.M. Benzene exposure in the United States 1978–1983: application in Seveso chloracne study. Chemosphere 2005: 60: 898–906. an overview. Am J Ind Med 1985: 7: 385–393. Buchet J.P., Van Eyken J., and Lauwerys R. Evalation de l0exposition au benzene Segal D. Exxon-Mobil Merger Wins FTC Approval. Washington Post. Washing- des travailleurs preposes aux qais de chargement de l’essence dans une ton, D.C, 1999. raffinerie. Cahiers de Medecine de Travail Cahiers Voor Arberdsgeneeskunde US EPA. 40 CFR 61 Subpart FF: National Emission Standards for Benzene in 1984: 21: 25–27. Waste Operations, 1989. Cherrie J.W. Are task-based exposure levels a valuable index of exposure for US EPA. Statistical Software ProUCL 4.0 for Environmental Applications for Data epidemiology? Ann Occup Hyg 1996: 40(6): 715–722. Sets with and without Nondetect Observations. Technical Support Center for CONCAWE. A Survey of Exposures to Gasoline Vapour.OilCompaniesEuropean Monitoring and Site Characterization, Las Vegas, NV, 2007. Available from: Organization for Environment, Health and Safety (CONCAWE), The Haguc, http://www.epa.gov/nerlesd1/tsc/tsc.htm. The Netherlands, 1987. Verma D.K., and des Tombe K. Measurement of benzene in the workplace and its CONCAWE. Review of European oil Industry Benzene Exposure Data (1986– evolution process, Part I: overview, history, and past methods. Am Ind Hyg 1992). Oil Companies European Organization for Environment, Health and Assoc J 1999: 60(1): 38–47. Safety (CONCAWE), Brussels, 1994. Verma D.K., Johnson D.M., Shaw M.L., and des Tombe K. Benzene and total Glass D.C., Adams G.G., Manuell R.W., and Bisby J.A. Retrospective exposure hydrocarbons exposures in the downstream petroleum industries. Am Ind Hyg assessment for benzene in the Australian petroleum industry. Ann Occup Hyg Assoc J 2001: 62(2): 176–194. 2000: 44(4): 301–320. Weaver N.K., Gibson R.L., and Smith C.W. Occupational exposure to benzene Glass D.C., Gray C.N., Adams G.G., Manuell R.W., and Bisby J.A. Validation in the petroleum and petrochemical industries. In: Mehlman M.A. (Ed.). of exposure estimation for benzene in the Australian petroleum industry. Advances in Modern Environmental Toxicology, Vol IV. Princeton Scientific Toxicol Ind Health 2001: 17(4): 113–127. Pub, 1983.

Journal of Exposure Science and Environmental Epidemiology (2011) 21(2) 185