Document 06bE5896gL8Ogww4LGavKxQkx

G Model CBI-6046; No. of Pages 11 ARTICLE IN PRESS Chemico-Biological Interactions xxx (2009) xxxxxx Contents lists available at ScienceDirect Chemico-Biological Interactions journal homepage: www.elsevier.com/locate/chembioint Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis D.C. Glass a,, T.W. Armstrong b,1, E.D. Pearlman b, D.K. Verma c, A.R. Schnatter b, L. Rushton d a Monash Centre for Occupational and Environmental Health (MonCOEH), Department of Epidemiology & Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, The Alfred Hospital, Melbourne, VIC 3004, Australia b ExxonMobil Biomedical Sciences, Annandale, NJ, USA c McMaster University, Hamilton, Ontario, Canada d Imperial College, London, UK article info Available online xxx abstract Background: Three casecontrol studies each nested within a cohort of petroleum workers assessed exposure to benzene in relation to risk of haematopoietic cancers. These studies have each been updated and the cases will be pooled to derive a more powerful study. The benzene exposure of new leukemia cases and controls was estimated in accordance with each respective study's original methods. An essential component of the process of pooling the data was comparison and rationalisation of the exposure estimates to ensure accuracy and consistency of approach. This paper describes this process and presents comparative estimates before and after appropriate revision took place. The original petroleum industry studies, in Canada, the UK and Australia, were conducted at different points in time by different study teams, but the industry used similar technology in similar eras in each of these countries. Methods: A job history for each subject giving job title, dates of starting and leaving the job and location of work, was assembled. For each job or task, the average benzene exposure (Base Estimate (BE) in ppm) was derived from measurements collected at applicable worksites. Estimates of exposure intensity (workplace exposure estimates (WE)) were then calculated for each line of work history by adjusting the BEs for site- and era-specific exposure-related variables such as loading technology and percentage benzene in the product. To ensure that the exposure estimates were comparable among the studies, the WEs were allocated to generic Job Categories, e.g. Tanker Driver (by technology used e.g. bottom loading), Motor Mechanic. The WEs were stratified into eras, reflecting technological changes in the industry. The arithmetic mean (AM), geometric mean (GM) and range of the stratified WEs were calculated, by study, for each generic Job Category. These were then compared. The AMs of the WEs were regarded as substantially similar if they were within 20% in all three studies in one era or for at least two studies in two eras. If the AM of the WE group differed by more than 20%, the data were examined to see whether the difference was justified by differences in local exposure conditions, such as an enclosure versus open work area. Estimates were adjusted in the absence of justification for the difference. Results: Reconciliation of differences resulted in changes to a small number of underlying BEs, particularly the background values, also the BEs attributed to some individuals and changes to the allocation of jobs between Job Categories. Although the studies covered some differing sectors of the industry and different time periods, for 22 Job Categories there was sufficient overlap, particularly in the downstream distribution sector, to make comparisons possible. After adjustment 12 Job Categories were judged to be similar and 10 were judged to be justifiably different. Job-based peak and skin exposure estimates were applied in a uniform way across the studies and a single approach to scoring the certainty of the exposure estimates was identified. Abbreviations: AIP, Australian Institute of Petroleum, Australian study; AM, arithmetic mean; BE, Base Estimate exposure in ppm associated with a specific job or task; EM, Exposure modifier, a factor used to adjust the BE to individual circumstances; FAB, French American British scheme to classify leukemias; GM, geometric mean; IOL, Imperial Oil Limited, Canadian study; IP, Institute of Petroleum, UK study; LH, lympho-haematopoetic; MM, multiple myeloma; NHL, non-Hodgkin lymphoma; WE, workplace exposure estimate, exposure estimate for a subject's job in ppm; WHO, World Health Organisation. Corresponding author. Tel.: +61 3 9903 0554; fax: +61 3 9903 0576. E-mail address: deborah.glass@med.monash.edu.au (D.C. Glass). 1 Present address: TWA8HR Occupational Hygiene Consulting LLC, Branchburg, NJ, USA. 0009-2797/$ see front matter 2009 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.cbi.2009.11.003 Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 2 Keywords: Exposure assessment Pooled analysis Benzene Petroleum industry Leukemia ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx The revised exposure estimates will be used in the pooled analysis to examine the risk of haematopoietic cancers and benzene exposure. This exercise provided an important quality control check on the exposure estimates and identified similarly exposed Job Categories that could be grouped for risk assessment analyses. 2009 Elsevier Ireland Ltd. All rights reserved. 1. Introduction Three casecontrol studies of lympho-haematopoetic (LH) cancers nested in petroleum industry cohorts have previously been carried out. The studies focussed on leukemia, where the exposure to benzene has been quantified for each member of the study [13]. The studies were the IOL study [4] from Canada, the IP study [5] from the UK and the AIP study [6] from Australia. The original studies were conducted at different points in time by different study teams, but the petroleum industry used similar technology in similar eras in each of these countries. Benzene is classified as a Group 1 human carcinogen by the International Agency for Research on Cancer [7], and there is general agreement that benzene can cause leukemia in highly exposed individuals. Exposure to benzene is lower in the petroleum industry compared to that reported in other studies of LH cancers, where the exposure may be to pure benzene rather than benzene mixed with other hydrocarbons [811]. Thus the risk estimates for LH cancers from the petroleum industry studies, particularly for leukemia, are of more relevance to exposures experienced currently, including in the wider community outside the petroleum industry. The three studies were carried out with a view to being able to pool the study subjects to form a single, larger, and more powerful study. A team from the Institute of Occupational Medicine (UK) and Institute for Risk Assessment Sciences, University of Utrecht (Holland) evaluated the consistency and quality of the data in these studies and discussed aspects important for the possible pooling of the studies [12]. The team indicated that although they "could not check in detail whether the metric for exposure assessments was the same across the studies", there was no intrinsic barrier to pooling the data [12]. 2. Background to the petroleum industry and exposure to benzene The petroleum industry can be viewed as having three main sectors: upstream where the oil is extracted from the ground either on land or offshore and undergoes initial processing to separate the gas from the crude oil; refineries where the crude is fractionated and gasoline, diesel, heating oil etc. are produced; distribution terminals where the products are received from the refinery in bulk, stored and distributed to factories, automobile service stations or other dispensing locations, often by road tanker. The three casecontrol studies all included distribution terminals but the IOL study also included marine distribution and the AIP study included upstream and refinery workers. Benzene is present in crude oil and at most stages of petroleum production and distribution. Historically, it has comprised about 3% of gasoline fuels, but is lower today. Exposure typically occurs during maintenance work or during loading or unloading of tanks, ships, rail cars and road tankers, tasks which are common at distribution terminals. In refineries, products are mainly handled in closed systems but exposure can occur from fugitive emissions. In the upstream sector, closed systems are common and the streams handled usually contain low concentrations of benzene so exposures are lowest in this sector. Benzene is also a by-product of combustion of fuels and other materials such as tobacco, wood and coal. It is present in indoor environments from activities such as cooking and heating and it is ubiquitous in urban air at low concentrations [13]. The IOL and IP studies include individuals whose work histories may go back to the beginning of the 20th century. The AIP study is more recent, with most workers being recruited after 1950. Technology has changed over the years, for example, road tankers were commonly loaded by the `top splash' method in the early years i.e. loading took place through hatches on top of the road tankers using loading arms that did not reach the bottom of the tanker, hence causing high exposure of volatile products, including benzene, to the driver who was positioned on top of the tanker. Between 1950 and 1970, these were gradually replaced by top submerged loading road tankers i.e. the loading arm reached the bottom of the tanker. Bottom loading road tankers with an enclosed system were introduced in the 1980s. The three studies used a similar epidemiological design, including for the retrospective benzene exposure estimation and all three studies included distribution workers. An essential component of the process of pooling the data was comparison and rationalisation of the exposure estimates to ensure accuracy and consistency of approach. This paper describes this process and presents comparative estimates before and after appropriate revision took place. 3. Exposure estimation methods A job history for each subject giving job title, dates of starting and leaving the job and location of work, was assembled from company records (IOL), from several company records (IP) or from prospectively collected cohort interviews (AIP). Information was gathered by the respective research teams about the worksites represented in each study and how they might have changed over time with respect to technology, fuel types distributed, engineering controls etc. Data on the nature of the tasks carried out by individuals at the site were also collected. The retrospective exposure assessment for benzene was carried out (blind as to case status) by occupational hygienists using the same basic model developed for the IOL study (the first of the studies) and tailored to the data available in each of the studies. This involved the following processes: Country-specific exposure data measured for specific jobs and tasks, augmented with more generic data from the literature where necessary, were assembled and used to calculate Base Estimates (BEs) which are average exposures of benzene exposure (expressed in ppm) for jobs or tasks. Table 1 presents values attributed to the BEs in the original studies [1416]. Filling of drums, trucks, rail cars and/or tanks with fuel entailed the highest exposures. To take account of changes in technology over time, the type of fuel handled, weather conditions and other factors that might affect exposure levels experienced by individuals, a range of key exposure modifiers (EM) was developed. The EMs varied slightly between the three studies (Table 2). Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx Table 1 Comparison of Base Estimate values between studies [1416]. Job group Airport background Refuelling with Avgas Drum filling, various configurations and ventilation Fitter terminal Tankage inventory Laboratory worker Motor Mechanic Rail car loading Road Tanker Bottom Loading Task Road Tanker Top Loading Task Agency Fuel Delivery Driver Road Tanker Driver Job (bottom load) Road Tanker Driver Job (top submerged) Road Tanker Driving & Unloading Service station attendant Ship dip/gauge Terminal Operator Terminal Operator background Urban background a Task-based BE. b Job-based BE. Base Estimates in ppm benzene AIP 0.08 1.65 1.554.69 0.67 4.20a 0.090.75 0.33 3.77a 0.55 1.76 0.19 0.32 0.16 5.4a 0.14 0.005 IP 0.11 0.48 0.19 0.16 1.27b 0.260.48 0.40 0.13 0.11 1.38b 0.163.45 0.016 0.003 3 IOL 0.04/8.7 0.03b 0.15 0.34 2.6 0.22 0.09 0.14 1.3/0.2 0.12 0.01 0.005 The EMs were assigned a range of values that were then used as multiplying factors for the BEs. For each line of work history an appropriate BE was selected for each job or task. If necessary, BEs were adjusted by multiplying by appropriate values of one or more EM to give a measure of the intensity of benzene exposure (ppm), called a Workplace Estimate (WE). The WEs were multiplied by the number of years for which the job was held and these estimates were summed to give a cumulative exposure estimate in ppm-years for each person. In the AIP study, the WEs were then adjusted by a factor to reflect the length of the reported working week. This adjustment was not possible for the other two studies, and therefore was not used in the pooled study nor for the comparisons described below. In the analyses of the three studies, WE and cumulative exposure estimates were used as a continuous or categorical variable to assess risk of LH cancer, specifically leukemia. Table 2 Exposure modifiers used in the three studies [1416]. Modifier type Workplace Exposure modifiers considered Extent of the system's enclosure, types of handling systems, e.g. bottom loading, vapour recovery, degree of automation Volumes (rate and quantity) of material handled and percentage losses as fugitive emissions rather than spills Task layout, distance to source of exposure Regulations affecting exposure Task Environment Materials Length of time spent on the task Materials handled Cleaning methods and materials used on self, plant and tools Outside jobs Temperature and variability Inside jobs Ventilation type and efficacy Volatility of materials (Reid vapour pressure for gasoline) Benzene content of materials Composition changes, time, season, region Study All IP, AIP IP, AIP AIP All All AIP IOL, IP IOL, AIP IOL, IP All All The pooled study included cases and controls from the three previously published studies together with newly identified cases and controls matched using the same criteria as before. Exposures for the new subjects were estimated by the same methodology and by the same study team members for IOL and AIP studies and by a subgroup of the original study team for the IP study. 4. Comparison and revision of the exposure estimates from the three studies An exposure assessment team that included three of the occupational hygienists that had worked on the exposure assessment for the original studies and on the updates to the studies and an external independent experienced hygienist contributed to this process outlined as follows. 4.1. Step 1: Review of the original methodology One member of the team (DG) visited the other two centres to discuss the exposure assessment process for each of the studies and assemble all necessary documentation. This was made available to the other team members after local input and correction. The original exposure estimations had taken place 1015 years previously; however, the original study documentation was identified and available for reference. The original exposure assessment of intensity and cumulative exposure had been carried out following the process described above. However, each study had also separately assigned measures of intermittent (peak) and skin exposure and these differed between studies, as did the measures of uncertainty assigned to the original exposure estimates. The team met face to face early in the comparison process, and for a week towards the end of the comparison process and had regular teleconferences over a 6-month period. 4.2. Step 2: Consolidation of the job categories Much of the technology and processes in the oil industries in the three countries were similar and thus the jobs and tasks were also similar. However, a wide range of job titles were used, many of which changed over the period of the three studies. The team therefore agreed on a list of 45 major Job Categories, e.g. Engineer, Motor Mechanic, Terminal Fitter, Tank Farm Operator, Tanker Driver; Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 4 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx which reflected the range of jobs within the three casecontrol studies (Table 3). The job title for each line of work history for each individual study subject was then allocated to one of the 45 Job Categories. 4.3. Step 3: Comparison of Workplace Estimates (WEs) For each study, the WEs (the benzene exposure intensity associated with each job) were grouped by generic Job Category and by Table 3 All Job Categories grouped by Analysis Group and indicating whether skin and or peak (>3 ppm at least weekly) exposure was allocated. Analysis Group White collar work Job category name Administrative Manager Engineer Terminal Supervisor (hands off) Description Clerical/admin (hands off), draughtsmen, graphic artists, admin supervisor, sales reps, car chauffeur, dispatchers Manager (hands off) Engineer (non-marine), Aircraft pilot/engineers Terminal Supervisor Craft work Motor Mechanic Terminal Fitter Laboratory Pump Mechanic Refinery Fitter Instrument Technician Motor Mechanic (vehicle maintenance) Terminal Fitter (Craftsman) Laboratory Workers Pump Mechanics, gas station pump repair and proving Refinery Fitter Instrument Technician--Refinery Product handling white collar Terminal Supervisor (hands on) Manager (hands on) Tankage Inventory Airfield Supervisor Terminal Supervisor Manager (hands on) Gauger--mostly office work but also checking tankage levels inc clerks (hands on) Airfield supervisor Product handling blue collar Terminal Operator Terminal Operator Tank Farm Operator Road Tanker Loader Rail Car Loader Drum Filler Jetty Operator Service Station Attendant Labourer (hands on) Marine Deck Marine Pumpman Air Attendant Air Attendant Avgas Tanker Driver Tanker Driver (no loading) Tanker Driver Bottom Load Tanker Driver Top Submerged Tanker Driver Top Splash Driver's Mate Terminal Operator--unspecified Terminal Operator--general (depot hand); dipper/checker (checking after filling) Tank farm operator (pumpman) Road/Tanker Loader (any refined) Rail Car Loader (any refined) Drum or Barrel Filler (any refined) Jetty operator Service station workers Labourer (hands on) Marine Deck--misc marine jobs, loading, unloading as main source of exposure Marine Pumpman, some significant exposure to bilge and product in bilge Air Station Attendants--uncertain about Avgas exposure Air Station Attendants Avgas Tanker Driver (unspecified) Tanker Driver No loading but mogas (exposure from unloading) Tanker Driver White Oil Bottom Load Tanker Driver White Oil Top Load Tanker Driver Top Splash Load Driver's mate (lorry boy, youth/adult mate) Unexposed blue collar work Refinery work (exposed) Tanker Driver (no mogas/avgas) Unexposed Terminal Operator Manual Worker Security Refinery Craftsman Refinery Operator (non-benzene) Other Upstream Air Attendant Avtur Bitumen Recovery Burner Repair Marine Non-Deck Refinery Supervisor (hands on) Refinery Operator (benzene) Refinery Operator-Wharf Refinery Operator-Tank Farm Tanker Driver (no mogas) Black Oil, Bitumen, Grease, Packaged drums, LPG, truck drivers) Other truck/equipment op Greasemaker, Terminal Operator, drum filler etc. at Lubes Site or Black Oil Terminal Operators Manual Worker, Labourer (hands off), Storekeeper, chef, cleaner, cook (facility background Security, Gatekeeper and Firemen Other Craftsmen, generally unexposed, e.g. electrician, painter, carpenter, rigger Refinery Operator--non-benzene stream, e.g. poly, alky, PDU, RCU, SRU, etc. Other Upstream Workers not exposed to product, e.g. radio room, helipad, instrument technician, electrician, first aider, crane operator Air Station Attendants Avtur only (no avgas) Bitumen Recovery (similar to heavy oil well work) Burner repair--home oil burner service and repair Marine Non-Deck--stewards, cooks, marine engineers & firemen, officers with background exposure Refinery Supervisor Refinery Operator--benzene stream, e.g. tank farm operators, wharf and jetty, CCU, CDU, reformers, DAP, mogas blending Refinery Operator-Wharf and Jetty Refinery Operator-Tank Farm Upstream operations Upstream Supervisor Upstream Fitter Upstream Onshore Upstream Offshore Upstream Supervisor Upstream Fitter Upstream Onshore Operator Upstream Offshore Operator a Indicates most usual designation but allocation may change for certain individuals. Skin No No No No High High Lowa High Med Med Lowa Lowa Lowa Low High High High High High High High High High Med High Med Med High High High High High Low No No No No No No No No Low Low No No Med High High No Low Low Low Peak No No No No Yes pre 1980 Yes Noa Yes Yesa No Yesa Noa No No Yes Yes Yes Yes Yes Yes Yes No Yesa Yes Yes No No Yes No No Yes Yes No No No No No No No No No No No No No No Yes Yes No No No No Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx Table 4 Workplace Estimates (ppm) by era, job category and study: original and revised estimates data for 19451979 only. Original estimates Revised estimates 19451959 19601979 19451959 IOL Administration AM 0.011 GM 0.010 n 131 Min 0.010 Max 0.020 IP 0.014 0.013 67 0.003 0.016 AIP 0.040 0.007 33 0.001 0.350 IOL 0.010 0.010 192 0.010 0.020 IP 0.015 0.013 87 0.003 0.016 AIP 0.015 0.005 150 0.001 0.168 IOL 0.006 0.005 63 0.005 0.016 IP 0.012 0.010 133 0.003 0.016 Manager (hands off) AM 0.005 GM 0.005 n 239 Min 0.005 Max 0.005 0.014 0.013 6 0.013 0.016 0.005 0.005 945 0.005 0.005 0.014 0.013 18 0.013 0.016 0.030 0.010 31 0.001 0.168 0.007 0.006 8 0.005 0.016 0.013 0.010 27 0.003 0.016 Terminal Supervisor (hands on) AM 0.404 0.328 GM 0.335 0.039 n 22 11 Min 0.088 0.016 Max 0.892 3.261 0.421 0.297 3 0.168 0.928 0.305 0.214 20 0.043 0.892 0.033 0.025 21 0.016 0.138 0.156 0.107 50 0.001 0.383 0.204 0.179 5 0.075 0.310 0.136 0.016 42 0.003 3.250 Motor Mechanic AM 0.051 GM 0.040 n 13 Min 0.010 Max 0.099 0.088 0.088 4 0.088 0.088 0.826 0.754 3 0.396 1.041 0.070 0.051 6 0.020 0.119 0.088 0.088 15 0.088 0.088 0.720 0.647 19 0.356 1.438 0.030 0.028 3 0.015 0.046 0.013 0.011 12 0.003 0.016 Terminal Fitter AM GM n Min Max 0.364 0.205 8 0.016 0.480 1.107 1.104 7 0.965 1.206 0.418 0.305 15 0.016 0.480 0.764 0.579 28 0.005 1.206 0.476 0.435 13 0.165 0.756 0.455 0.126 33 0.003 4.619 Terminal Operator--general (depot hand) AM 0.814 3.185 0.499 GM 0.541 0.391 0.301 n 43 33 10 Min 0.036 0.016 0.168 Max 1.785 0.016 1.872 1.134 0.893 108 0.058 2.096 0.191 0.149 59 0.018 0.425 0.361 0.252 28 0.140 1.572 0.839 0.751 15 0.357 1.525 1.617 0.191 90 0.004 32.47 Drum Filler AM GM n Min Max 1.310 1.310 1 1.526 1.526 1 1.526 1.526 1.678 1.451 15 0.291 3.780 1.000 1 1.220 0.825 24 0.153 4.908 1.305 1.305 2 1.305 1.305 1.321 0.743 6 0.086 2.952 Tanker Driver (no mogas/avgas) AM 0.016 GM 0.016 n 29 Min 0.016 Max 0.016 0.005 0.005 3 0.005 0.005 0.016 0.016 45 0.016 0.016 0.016 0.006 15 0.005 0.168 0.008 0.008 11 0.008 0.008 0.013 0.007 110 0.003 0.058 Tanker Driver Top Submerged Load AM 0.218 GM 0.123 n 19 Min 0.036 Max 2.159 0.353 0.330 14 0.228 0.777 0.156 0.118 206 0.022 2.053 0.364 0.206 51 0.174 1.147 0.402 0.397 4 0.304 0.457 0.341 0.138 59 0.016 3.148 Tanker Driver Top Splash Load AM 0.675 GM 0.198 n 166 Min 0.025 Max 10.26 0.695 0.690 2 0.619 0.770 0.528 0.151 30 0.036 6.451 0.781 0.674 18 0.250 1.488 0.647 0.200 409 0.013 10.26 Unexposed Terminal Operator AM 0.016 GM 0.016 n5 Min 0.016 Max 0.016 0.147 0.118 12 0.005 0.168 0.016 0.016 21 0.016 0.016 0.115 0.082 44 0.005 0.168 0.016 0.016 2 0.016 0.016 0.009 0.006 39 0.003 0.016 AIP 0.012 0.007 31 0.001 0.084 0.315 0.069 3 0.019 0.907 0.503 0.387 4 0.264 1.220 1.107 1.104 7 0.965 1.206 0.959 0.419 4 0.019 1.872 1.50 1.23 17 0.16 3.71 0.005 0.005 3 0.005 0.005 0.329 0.313 14 0.228 0.753 0.695 0.690 2 0.619 0.770 0.018 0.018 19 0.005 0.019 19601979 IOL IP 0.006 0.006 96 0.005 0.016 0.012 0.010 178 0.003 0.016 0.005 0.005 16 0.005 0.016 0.012 0.009 40 0.003 0.016 0.341 0.182 11 0.043 2.550 0.082 0.023 22 0.004 0.976 0.023 0.023 2 0.015 0.024 0.014 0.012 36 0.003 0.016 0.724 0.678 9 0.180 0.825 0.342 0.146 20 0.003 0.480 0.702 0.359 35 0.014 1.947 0.202 0.118 187 0.004 1.611 0.035 0.035 1 0.035 0.035 0.306 0.306 2 0.306 0.306 0.014 0.010 8 0.008 0.070 0.014 0.008 125 0.003 0.088 0.289 0.230 33 0.082 0.660 0.179 0.121 426 0.009 2.053 1.016 0.912 9 0.250 1.488 0.952 0.303 62 0.023 7.391 0.016 0.016 9 0.016 0.016 0.005 0.004 95 0.003 0.016 5 AIP 0.012 0.008 151 0.001 0.084 0.014 0.010 33 0.005 0.084 0.067 0.034 49 0.005 0.350 0.349 0.318 19 0.175 0.892 0.821 0.766 26 0.362 1.206 0.336 0.168 6 0.019 0.651 1.32 0.92 21 0.11 4.87 0.007 0.006 25 0.005 0.038 0.389 0.345 56 0.174 1.980 0.014 0.012 51 0.005 0.019 Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 6 Table 4 (Continued ) Original estimates 19451959 IOL Manual Worker AM GM n Min Max IP 0.016 0.016 3 0.016 0.016 Security AM GM n Min Max 0.016 0.016 2 0.016 0.016 Engineer AM GM n Min Max 0.010 0.006 5 0.003 0.016 Air Avgas AM GM n Min Max 0.046 0.043 18 0.026 0.075 1.200 0.996 4 0.400 2.356 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx Revised estimates 19601979 19451959 AIP IOL IP AIP IOL IP AIP 0.199 0.042 3 0.001 0.428 0.016 0.016 21 0.016 0.016 0.145 0.132 5 0.079 0.254 0.032 0.025 9 0.016 0.070 0.012 0.010 85 0.003 0.016 0.010 0.004 2 0.001 0.019 0.126 0.119 2 0.084 0.168 0.144 0.020 13 0.001 0.604 0.070 0.070 2 0.070 0.070 0.012 0.009 21 0.003 0.016 0.052 0.040 2 0.019 0.084 0.035 0.017 4 0.005 0.084 0.008 0.006 5 0.003 0.016 0.041 0.018 50 0.001 0.260 0.007 0.006 13 0.005 0.016 0.006 0.004 14 0.003 0.016 0.045 0.020 4 0.005 0.084 0.054 0.049 14 0.019 0.107 0.546 0.272 15 0.072 2.002 0.066 0.066 2 0.066 0.066 0.076 0.074 36 0.057 0.110 0.795 0.733 3 0.400 1.045 0.223 0.223 7 0.223 0.242 19601979 IOL IP 0.018 0.016 8 0.005 0.070 0.012 0.010 50 0.003 0.016 0.016 0.016 1 0.016 0.016 0.009 0.006 7 0.003 0.016 0.005 0.005 20 0.005 0.016 0.008 0.005 14 0.003 0.016 0.057 0.051 29 0.008 0.110 0.399 0.117 14 0.016 1.891 AIP 0.025 0.014 10 0.005 0.084 0.017 0.006 8 0.001 0.072 0.039 0.018 47 0.001 0.133 0.046 0.043 18 0.026 0.075 era (pre 1945, 19451959, 19601979 and 1980 onwards). The eras were chosen to reflect technological changes that had taken place in the industry. The arithmetic mean (AM) and geometric mean (GM) of the WE groups were estimated and the number, maximum and minimum values identified. These data were then assembled into a spread sheet and compared between studies. The AMs of the WEs were regarded as substantially similar if they were within 20%, in all three studies in one era or for at least two studies in two eras. If the AM of the WE group differed by more than 20% the data were examined to see if the reason for the difference could be ascertained. In many cases, differences in technology, fuel handled, job processes or other factors affecting exposure for the specific job group/era led to appropriate assignment of different BEs and/or EMs between the studies so that the dissimilarity was thought to be correct. This was documented. In other cases, differences between the WEs appeared to be a difference in the estimation process used by the three studies. The data were then investigated to see whether and how the WE should be adjusted. Changes to the exposure estimates were agreed by the group and implemented by the individual study exposure assessor. 4.4. Step 4: Development of common definitions of indices of intermittent (peak) and skin exposure Two measures of peak exposure were used, firstly whether a subject was likely to have been exposed to more than 3 ppm of benzene for between 15 min and 60 min at least weekly (Yes or No) and secondly if they were likely to have been exposed to a product containing more than 20% benzene (Yes or No). The skin exposure metric of No Exposure, Low, Medium or High was allocated on the basis of whether or not there was a probability of at least weekly skin exposure. This metric did not have dimensions that included extent of skin area, product concentration or period of contact. 4.5. Step 5: Allocation of job certainty score A certainty score of Low, Medium or High was allocated to the exposure estimate for each line of work history. This was driven firstly by the certainty with which the Job Title was known and then by the certainty in the knowledge of the site characteristics, products handled, the technology used and the robustness of the study's Base Estimate for that job. 4.6. Step 6: Debate and agree on changes to original exposure estimates The team met face to face for a week-long meeting to debate and agree on differences and similarities in the individual studies, and to agree on the changes necessary to pool the studies. Detailed minutes were kept of the teleconferences and the meeting, to record decisions and the reasoning behind them. 5. Results There were originally 45 Job Categories identified, four further Job Categories were created in discussion. A `Tankage Inventory' Job Category was created for clerical workers who dipped tanks for inventory or customs purposes. Road and Rail Car Loaders were separated into two categories. `Supervisor' and `Manager' categories were each split into `hands on' and `hands off' to make the jobs within the categories more uniform and comparable. For 27 of the Job Categories it was not possible to meaningfully compare between studies because there was insufficient overlap between the studies. This was defined as fewer than five jobs in more than one study in at least one era, e.g. lorry boy or driver's mate was an IP study-only job, the IOL study had three marine distribution Job Categories, with a substantial number of jobs, the AIP study had six refinery Job Categories and five upstream Job Categories which were not adequately covered in the other studies. Table 4 presents selected WEs for 19501979 before and after changes were made. The changes were either as a result of reallocation of jobs between categories or as a result of changes to the values attributed to the BE or a change in the BE attributed (mainly allocation of a population rather than a site background). Some of the initial differences in the WEs were explained by the differences in Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx Table 5 Workplace Estimates of benzene (ppm) by Job Category, era and study which were similar between studies after adjustment. pre 1945 19451959 19601979 IOL IP AIP IOL IP AIP IOL IP AIP Administration AM 0.006 GM 0.005 n 24 Min 0.005 Max 0.016 0.010 0.008 94 0.003 0.016 0.008 0.007 5 0.005 0.019 0.006 0.005 63 0.005 0.016 0.012 0.010 133 0.003 0.016 0.012 0.007 31 0.001 0.084 0.006 0.006 96 0.005 0.016 0.012 0.010 178 0.003 0.016 0.012 0.008 151 0.001 0.084 Mechanic AM GM n Min Max 0.550 0.173 1 0.028 1.071 0.012 0.009 3 0.003 0.016 0.030 0.028 3 0.015 0.046 0.013 0.011 12 0.003 0.016 0.503 0.387 4 0.264 1.220 0.023 0.023 2 0.015 0.024 0.014 0.012 36 0.003 0.016 0.349 0.318 19 0.175 0.892 Terminal Fitter AM 0.358 GM 0.358 n1 Min 0.358 Max 0.358 0.692 0.299 23 0.003 6.781 0.476 0.435 13 0.165 0.756 0.455 0.126 33 0.003 4.619 1.107 1.104 7 0.965 1.206 0.724 0.678 9 0.180 0.825 0.342 0.146 20 0.003 0.480 0.821 0.766 26 0.362 1.206 Drum Filler AM GM n Min Max 0.74 0.74 1 0.74 0.74 1.858 1.119 8 0.269 4.374 1.305 1.305 2 1.305 1.305 1.321 0.743 6 0.086 2.952 1.50 1.23 17 0.16 3.71 0.035 0.035 1 0.035 0.035 0.306 0.306 2 0.306 0.306 1.32 0.92 21 0.11 4.87 Tanker Driver Bottom Load AM GM n Min Max 0.191 0.191 1 0.191 0.191 0.198 0.196 3 0.166 0.215 Tanker Driver Top Submerged Load AM 0.059 GM 0.055 n2 Min 0.037 Max 0.081 0.402 0.397 4 0.304 0.457 0.341 0.138 59 0.016 3.148 0.329 0.313 14 0.228 0.753 0.289 0.230 33 0.082 0.660 0.179 0.121 426 0.009 2.053 0.389 0.345 56 0.174 1.980 Tanker Driver Top Splash Load AM 0.391 0.614 GM 0.381 0.131 n 4 130 Min 0.296 0.020 Max 0.581 14.40 0.781 0.674 18 0.250 1.488 0.647 0.200 409 0.013 10.26 0.695 0.690 2 0.619 0.770 1.016 0.912 9 0.250 1.488 0.952 0.303 62 0.023 7.391 Unexposed Terminal Operator AM 0.009 GM 0.007 n 19 Min 0.003 Max 0.016 0.016 0.016 2 0.016 0.016 0.009 0.006 39 0.003 0.016 0.018 0.018 19 0.005 0.019 0.016 0.016 9 0.016 0.016 0.005 0.004 95 0.003 0.016 0.014 0.012 51 0.005 0.019 Engineer AM GM n Min Max 0.005 0.005 5 0.005 0.005 0.008 0.006 12 0.003 0.016 0.007 0.006 13 0.005 0.016 0.006 0.004 14 0.003 0.016 0.045 0.020 4 0.005 0.084 0.005 0.005 20 0.005 0.016 0.008 0.005 14 0.003 0.016 0.039 0.018 47 0.001 0.133 Refinery Fitter AM GM n Min Max 0.400 0.400 2 0.400 0.400 0.425 0.320 21 0.001 1.206 0.360 0.348 5 0.200 0.400 0.395 0.364 37 0.082 1.206 Refinery Operator (non-benzene stream) AM GM n Min Max 0.070 0.070 2 0.070 0.070 0.075 0.061 9 0.005 0.084 0.070 0.070 6 0.070 0.070 0.070 0.062 54 0.005 0.096 7 1980 on IOL IP AIP 0.006 0.005 57 0.005 0.016 0.014 0.012 13 0.003 0.016 0.015 0.009 116 0.001 0.070 0.031 0.030 1 0.024 0.047 0.295 0.282 12 0.220 0.527 0.286 0.247 1 0.110 0.400 0.480 0.480 1 0.480 0.480 0.553 0.530 10 0.234 0.670 0.70 0.60 14 0.16 1.73 0.123 0.107 9 0.059 0.363 0.140 0.140 1 0.140 0.140 0.195 0.192 41 0.108 0.255 0.190 0.154 10 0.074 0.521 0.123 0.100 28 0.016 0.321 0.308 0.295 38 0.127 0.456 0.016 0.016 3 0.016 0.016 0.006 0.006 10 0.005 0.016 0.340 0.325 3 0.200 0.400 0.020 0.014 3 0.010 0.070 0.003 0.003 8 0.003 0.003 0.016 0.016 1 0.016 0.016 0.013 0.011 29 0.005 0.017 0.033 0.012 39 0.001 0.191 0.335 0.308 36 0.070 1.020 0.066 0.058 68 0.005 0.140 Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 8 Table 5 (Continued) pre 1945 IOL Air Avgas AM GM n Min Max 0.110 0.110 1 0.110 0.110 IP 0.110 0.110 2 0.110 0.110 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx 19451959 AIP IOL IP 19601979 AIP IOL IP AIP 0.066 0.066 2 0.066 0.066 0.076 0.074 36 0.057 0.110 0.795 0.733 3 0.400 1.045 0.223 0.223 7 0.223 0.242 0.057 0.051 29 0.008 0.110 0.399 0.117 14 0.016 1.891 1980 on IOL IP 0.223 0.223 1 0.223 0.223 0.089 0.089 1 0.089 0.089 AIP 0.206 0.054 6 0.015 0.836 background BE values that had been assigned. Changes were made as follows: AIP: the Terminal Background BE (originally 0.14 ppm) and the Airport Background BE (originally 0.08 ppm) were revised down to 0.016 ppm (use of local data and imputed data gave inappropriately high BEs). IP study: where black oil only was handled a population background level of 0.003 ppm was assigned IOL study: the Refinery Background was increased to 0.07 ppm in line with the AIP study after examination of data from CONCAWE [17]. AIP study: office jobs at airports and terminals were allocated the Terminal Background BE. AIP study: the BE for Motor Mechanics was reduced from 0.33 ppm to 0.22 ppm. The higher value had been based on data from the literature which included gasoline mechanics [18]. The above changes have resulted in a general reduction of many of the mean WEs (Table 4). After the rationalisation and adjustment, there were 12 Job Categories where the WEs were substantially similar, Table 5; these included Administration, Tanker Drivers, Drum Fillers and Fitters. When using a less rigorous definition of similar, i.e. any 2 studies with WEs within 20% in any one era, there were 20 of the comparable 25 sets had similar mean WEs. There were 10 Job Categories where the differences in the WEs were deemed to be a result of `real' and justifiable differences in exposure for the subjects in the three countries (Table 6). Some of these differences were a result of different background exposures being allocated because of the different industry sectors included in the jobs, e.g. security which included refinery personnel in the AIP study but mainly terminal personnel in the other studies. There were differences in products handled between the studies, many of the IP sites had little benzene-containing product after 1960 so most Manual Workers and Terminal Operators at these sites had lower exposure allocated than in the other studies. In addition, there were a small number of sites in the UK which, until the late 1960s handled high benzene products. Thus there was a wide range of exposures for the same Job Category in the UK. Some IP top loading tanker drivers only carried gasoline in an estimated 1 in 20 trips so exposure was lower for these jobs. WEs were similar for some eras and not at others, e.g. drum filling jobs were identified into the 1980s in the AIP study but not after the 1970s in the IP study. Drum filling is one of the more highly exposed jobs that a terminal operator may carry out. Differences in technology were thought to explain some other differences, e.g. the extent of enclosure of filling sheds which differed between countries. Few individuals in the IOL study and none in the IP study loaded rail cars with petrol, a relatively highly exposed job at Australian terminals. The allocation of peak and skin exposure by Job Category is shown in Table 3. The independent allocations made in the three studies were very similar. After discussion, a few changes were made from one level to the next, e.g. AIP study reduced Motor Mechanic from high to medium, the IP study changed Road Tanker Driver, Bottom Load from high to medium. Motor Mechanic was allocated a peak exposure in the AIP and IOL study and revised to this allocation in the IP study. Allocation of a Job Certainty score was similar in the three studies. The most important element to the score was the certainty of the job title. In the IP study, there were a number of individuals where only the last job was known and so the exposure estimate for the majority of the career was scored as uncertain. In addition, most jobs carried out before 1940 were scored as uncertain. The IOL study was thought to have understated the certainty of the estimates of some jobs, compared to the other studies. These were re-evaluated. 6. Discussion The comparison of the WEs was carried out to ensure that it was appropriate to pool the cases and controls from the three studies. We did not want to reduce the quality of the estimation work by ignoring the careful subject by subject benzene exposure estimations [19]. However, large differences in the exposure estimates for similar jobs between the studies that could not be explained by local real differences in exposure conditions, would have suggested that potentially the data should not be pooled. There has been little published research looking at the comparability of quantified retrospective exposure estimates used in pooled occupational epidemiology studies although there have been more studies using categorical variables [20,21]. A notable exception is a paper that looked at crystalline silica exposure in ten cohorts in support of a pooled analysis [22]. In this study, the data available for each cohort were compared to create exposure estimates in comparable units (mg/m3) of respirable crystalline silica. This was necessary because the measurements had been collected in different ways for the different cohorts, e.g. as mass concentration or particle counts, and the dust had variable silica content between the studies. The adapted exposure estimates were then partially validated by the successful prediction of silicosis in the different constituent cohorts. In our study there was less variability in the exposure estimation methods used as all the exposure estimates were in ppm benzene. In addition, our study made comparisons at the level of the exposure estimates for subjects in the nested casecontrol study rather than at the level of the cohort. In the initial discussions about pooling the data, it became clear that there were inter-study differences in the way that jobs had been allocated to the agreed Job Categories. Most of the changes that took place following the inter-study comparison were reallocations of jobs between categories. The use of AM, GM and range values allowed investigation of the distribution of the data in each cell. The comparison process could then identify outliers between the study sets, e.g. the AIP Motor Mechanics exposure, and within the data sets such some Terminal Operators in the IP study who handled benzene enriched fuels. The outliers could be examined, changes made where necessary and real differences in probable exposure identified. Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx Table 6 Workplace Estimates of benzene (ppm) by Job Category, era and Study which were dissimilar between studies even after adjustment. pre 1945 19451959 19601979 1980 on IOL IP AIP IOL IP AIP IOL IP AIP IOL Manager (hands on) AM 0.005 GM 0.005 n1 Min 0.005 Max 0.005 0.011 0.008 13 0.003 0.016 0.007 0.006 8 0.005 0.016 0.013 0.010 27 0.003 0.016 0.005 0.005 16 0.005 0.016 0.012 0.009 40 0.003 0.016 0.014 0.010 33 0.005 0.084 0.005 0.005 12 0.005 0.016 Terminal supervisor (hands off) AM 0.008 GM 0.006 n8 Min 0.003 Max 0.016 0.016 0.016 3 0.016 0.016 0.006 0.004 27 0.003 0.016 0.015 0.015 5 0.005 0.016 0.005 0.004 55 0.003 0.016 0.012 0.011 3 0.005 0.016 Terminal Supervisor (hands on) AM 0.249 GM 0.045 n 10 Min 0.005 Max 1.470 0.204 0.179 5 0.075 0.310 0.136 0.016 42 0.003 3.250 0.315 0.069 3 0.019 0.907 0.341 0.182 11 0.043 2.550 0.082 0.023 22 0.004 0.976 0.067 0.034 49 0.005 0.350 0.066 0.066 1 0.066 0.066 Terminal Operator--general (depot hand) AM 0.433 0.990 GM 0.431 0.189 n4 31 min 0.389 0.013 max 0.500 7.340 0.839 0.751 15 0.357 1.525 1.617 0.191 90 0.004 32.469 0.959 0.419 4 0.019 1.872 0.702 0.359 35 0.014 1.947 0.202 0.118 187 0.004 1.611 0.336 0.168 6 0.019 0.651 0.130 0.088 10 0.014 0.360 Tank farm operator AM GM n min max 0.654 0.295 8 0.039 1.434 0.522 0.095 15 0.003 1.361 0.455 0.444 4 0.309 0.551 0.025 0.014 2 0.009 0.090 0.299 0.141 5 0.016 0.656 0.448 0.394 19 0.083 1.196 Tanker Driver No loading AM 0.018 GM 0.016 n 31 Min 0.003 Max 0.067 0.010 0.007 10 0.003 0.016 0.160 0.160 25 0.160 0.160 0.018 0.017 5 0.016 0.024 0.160 0.160 12 0.160 0.160 Tanker Driver (no mogas/avgas) AM 0.008 0.006 GM 0.008 0.004 n3 35 Min 0.008 0.003 Max 0.008 0.044 0.008 0.008 11 0.008 0.008 0.013 0.007 110 0.003 0.058 0.005 0.005 3 0.005 0.005 0.014 0.010 8 0.008 0.070 0.014 0.008 125 0.003 0.088 0.007 0.006 25 0.005 0.038 Manual Worker AM 0.016 GM 0.016 n3 Min 0.016 Max 0.016 0.012 0.010 62 0.003 0.016 0.032 0.025 9 0.016 0.070 0.012 0.010 85 0.003 0.016 0.010 0.004 2 0.001 0.019 0.018 0.016 8 0.005 0.070 0.012 0.010 50 0.003 0.016 0.025 0.014 10 0.005 0.084 0.023 0.018 4 0.016 0.141 Security AM GM n Min Max 0.010 0.007 2 0.003 0.016 0.070 0.070 2 0.070 0.070 0.012 0.009 21 0.003 0.016 0.052 0.040 2 0.019 0.084 0.016 0.016 1 0.016 0.016 0.009 0.006 7 0.003 0.016 0.017 0.006 8 0.001 0.072 0.016 0.016 1 0.016 0.016 Air Attendant Avtur AM GM n Min Max 0.014 0.013 15 0.003 0.016 0.016 0.016 11 0.016 0.016 IP 0.016 0.016 2 0.016 0.016 0.007 0.005 3 0.003 0.016 0.083 0.074 11 0.034 0.163 0.014 0.009 4 0.003 0.030 0.003 0.003 1 0.003 0.003 0.008 0.006 5 0.003 0.016 9 AIP 0.014 0.010 81 0.001 0.070 0.039 0.025 36 0.005 0.203 0.415 0.119 3 0.005 0.844 0.293 0.196 16 0.016 0.872 0.005 0.005 13 0.005 0.005 0.038 0.019 2 0.005 0.070 0.024 0.011 15 0.001 0.071 0.016 0.016 6 0.016 0.016 There are of course inter-day and inter-individual differences in exposure in work places which were not taken into account in this study. The use of the BEs and EMs can take account of inter-workplace or era differences but not inter-individual differences in exposure which can be large [23]. Measured exposures for a specific job typically follow a lognormal distribution [24,25]. It is possible that some workers are consistently more highly exposed than others, e.g. because of individual worker practice, and these individuals may be at increased risk of LH cancer. This is a limitation of most studies because daily personal exposure measurement data are usually not available for each study individual. Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 10 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx The WEs were grouped by job type and by era to allow interstudy comparisons. This comparison could not be done at the level of the BE because different EMs were used in the different studies. The inter-study comparison could not take place at the level of the cumulative exposure estimates because these were dependent on the length of an individual's service. The WEs were allocated to the era in which the majority of service of the relevant job took place. It was considered that each WE should appear once in the comparison process rather than for each year that it was applied to. The latter approach would have weighted the comparison towards jobs held over a number of years. There was a balance struck between tight job/era definitions and number of observations. If job categories were small and eras short, there would have been few WEs in each cell. The eras and 45 of 49 Job Categories were agreed before the comparison process took place. The inter-study comparison process was an additional and powerful quality control measure. Even though there was a strong similarity in the individual study's exposure estimation process, and in the range of exposure estimates from the original studies [26], it was important to check that the data could be pooled. The investigators identified outliers in their own data sets and were able to check that these were justified, or make changes when appropriate. In addition, for the exposure estimate certainty score, by sharing exposure estimates and comparing scores across the studies we could identify whether the Low, Medium and High scores had been interpreted in the same way. The allocation of the peak and skin metrics were examined across the studies and some adjustments made to improve consistency. The changes were most usually from one category to the next and reflected slightly different interpretations of the metric and were usually easily resolved between the investigators. The initial Job Categories had not, however, been interpreted in the same way and this gave rise to new Job Categories and changes to the individuals assigned to existing categories. Jobs were not grouped on the basis of job title, because this may not be a good guide to exposure [27]. Job titles vary between companies and by industry sector. For example, in the offshore petroleum industry the term Mechanic is used for a plant maintenance worker, at terminals and refineries it means a vehicle maintenance worker. Further, it has been shown that job titles vary over time [28]. It was important to have occupational hygiene expertise on the industry processes, typical tasks performed in jobs present for the different industry sectors, changes in tasks over time and limitations of precision in retrospective exposure estimation. It was useful to have occupational hygienists from more than one study who had independently assessed exposure. It has been shown that individuals are inevitably selective about the information that they use to make judgements [29]. Having more than one expert exposure assessor meant that more information was shared and opinions were challenged, a process that likely resulted in a more accurate exposure assessment. It was valuable for later reference to have both the discussion and outcomes documented. It proved important to understand the estimation process that had taken place within each of the studies. The exposure estimation methodology was well documented in each study but the reasoning behind individual exposure estimates could be complex. The IP study used a series of codes to identify the reasons for decisions on the BE and EMs included for individual exposure estimates and this improved the transparency of expert exposure estimation [30]. However, individual decisions often depended on site data which was complex to identify and time consuming to check. The early face to face contact was important in establishing the trust between the investigators necessary for open sharing of data and ideas. It was also necessary to ensure a uniform approach to common understanding of the task to be undertaken [31]. The week-long face to face meeting allowed an in depth exploration of the inter-study similarities and differences in the exposure estimates and resolution of the sources of differences. The use of the population background BE in the IP study for jobs at terminals where no benzene was found and for tanker drivers carrying non-benzene products and the reduction in the Terminal and Airport Background BEs in the AIP study will affect a large number of jobs and hence subjects. This has important implications for the doseresponse analyses as the background comparison group will be larger and contain a number of workers that had previously been assessed as having higher exposure. The inclusion of a larger number of workers in the comparison group as was done in previous analyses of the AIP study was found to have had a large effect on odds ratios [26,32]. The Miller et al. review recommended making some standardisations across the exposure calculations, including standardisation of background exposures [12]. This standardisation was described here. It took considerable detailed review of the data and collaboration to ensure compatibility between the exposure estimates from the three studies. The studies were all nested casecontrol studies and used the same approach to the retrospective exposure assessment and there had been communication between the investigators when the original studies were carried out. It was important that a principle exposure assessor from each of the three of the original studies was involved in the exposure resolution so that differences could be resolved or agreed upon. This has wider implications for the general pooling of subjects from casecontrol and other studies which may have used differing exposure assessment methods. 7. Conclusions Retrospective exposure estimates, even those based on the same model, can result in different estimates for similar jobs. In some cases, there were genuine differences resulting from different technology or products being used. In other cases the apparent differences were a result of lack of definition in the original Job Category so that `apples' and `pears' were being compared. When these differences had been resolved the groups were both more homogenous and more comparable with respect to benzene exposure. Some differences in exposure estimates were not justified and alterations to the Base Estimates used or the value attributed to a Base Estimate were made. Many of the changes that were made, reduced exposure by adjusting background values or changing BEs and Job Categories for less highly exposed jobs such as managers or terminal workers with only background exposure, e.g. black oil Terminal Operators. There was more agreement for the more highly exposed jobs such as tanker drivers and Terminal Operators. This exercise provided an important quality control check on the exposure estimates and improved the homogeneity of the exposure groups which will be used in the pooled analysis. It was important to have expertise on the industry processes, the jobs carried out in the different industry sectors and how that have changed over time, about the difficulties and limitations of precision in retrospective exposure estimation and also an understanding of the estimation process that had taken place within each of the original studies. Conflict of interest D.C. Glass: Funding from the Australian Institute of Petroleum for the ongoing maintenance and updating of the Health Watch Cohort. D.K. Verma, L. Rushton: None. T.W. Armstrong: For part of the study period, Dr. Armstrong was an employee of ExxonMobil Biomedical Sciences Inc., a subsidiary of ExxonMobil Corporation. The funding via CONCAWE and an Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003 G Model CBI-6046; No. of Pages 11 ARTICLE IN PRESS D.C. Glass et al. / Chemico-Biological Interactions xxx (2009) xxxxxx 11 oversight by an independent Science Advisory Panel mitigated any influence of employment. A.R. Schnatter, E.D. Pearlman: Employment with ExxonMobil Biomedical Sciences, Inc. which may give appearance of conflict of interest. Fundings D.C. Glass, T.W. Armstrong, E.D. Pearlman, D.K. Verma, L. Rushton: CONCAWE. A.R. Schnatter: CONCAWE. No sponsor involvement in design, collection, analysis, interpretation of data, writing of manuscript nor decision to submit manuscript. Acknowledgements This research was made possible under contract with CONCAWE and funded by CONCAWE, API, CPPI and CEFIC APA. References [1] T.W. Armstrong, E.D. Pearlman, et al., Retrospective benzene and total hydrocarbon exposure assessment for a petroleum marketing and distribution worker epidemiology study, Am. Ind. Hyg. Assoc. J. 57 (1996) 333343. [2] S.J. Lewis, G.M. Bell, et al., Retrospective estimation of exposure to benzene in a leukaemia casecontrol study of petroleum marketing and distribution workers in the United Kingdom, Occup. Environ. Med. 54 (1997) 167175. [3] D.C. Glass, G.G. Adams, et al., Retrospective exposure assessment for benzene in the Australian Petroleum Industry, Ann. Occup. Hyg. 44 (4) (2000) 301320. [4] R.A. Schnatter, T.W. Armstrong, et al., Lymphohaemotopoietic malignancies and quantitative estimates of exposure to benzene in Canadian petroleum distribution workers, Occup. Environ. Med. 53 (1996) 773781. [5] L. Rushton, H. Romaniuk, A casecontrol study to investigate the risk of leukaemia associated with exposure to benzene in petroleum marketing and distribution workers in the United Kingdom, Occup. Environ. Med. 54 (1997) 152166. [6] D.C. Glass, C.N. Gray, et al., Leukemia risk associated with low level benzene exposure, Epidemiology 15 (5) (2003) 569577. [7] IARC, IARC Monograph on the Evaluation of the Carcinogenic Risk of Chemicals: Some Industrial Chemicals and Dyestuffs, IARC WHO, Lyon, 1982. [8] K.S. Crump, Risk of benzene-induced leukaemia: a sensitivity analysis of the Pliofilm cohort with additional follow-up and new exposure estimates, J. Toxicol. Environ. Health 42 (1994) 219242. [9] R.B. Hayes, S.-N. Yin, et al., Benzene and the dose-related incidence of hematologic neoplasms in China, J. Natl. Cancer Inst. 89 (14) (1997) 10651071. [10] M.B. Paxton, V.M. Chinchilli, et al., Leukemia risk associated with benzene exposure in the Pliofilm cohort. II. Risk estimates, Risk Anal. 14 (2) (1994) 155161. [11] R.A. Schnatter, M.J. Nicolich, et al., Determination of leukemogenic benzene exposure concentrations: refined analyses of the Pliofilm cohort, Risk Anal. 16 (6) (1996) 833840. [12] B.G. Miller, W. Fransman, et al., A review of the data quality and comparability of casecontrol studies of low-level exposure to benzene in the petroleum industry, Int. Arch. Occup. Environ. Health (2009), online Wednesday September 16th 2009. [13] L.A. Wallace, Environmental exposure to benzene: an update, Environ. Health Perspect. 104 (6) (1996) 11291136. [14] T.W. Armstrong, E.D. Pearlman, et al., Retrospective benzene and total hydrocarbon exposure assessment for a petroleum marketing and distribution worker epidemiology study, Occupational Health Division Exxon Biomedical Sciences Ltd East Millstone NJ (1994). [15] G.M. Bell, N. Cordingley, et al., Retrospective Estimation of Exposure to Benzene in a Leukaemia CaseControl Study of Petroleum Marketing and Distribution Workers in the United Kingdom, Institute of Petroleum, London, August 1995. [16] D. Glass, C. Gray, et al., Lympho-haematopoietic cancer and exposure to benzene in the Australian Petroleum Industry, Melbourne: Report to AIP, 2001 May 2001. [17] CONCAWE, Exposure to Atmospheric Benzene Vapour Associated with Motor Gasoline, European Oil Companies' Organization for Environmental and Health Protection, Brussels, 1981, February 1981. Report No.: 2/81. [18] D.C. Glass, C.N. Gray, et al., Validation of exposure estimation for benzene in the Australian Petroleum Industry, Toxicol. Ind. Health. 17 (4) (2001) 113 127. [19] J.M. Elwood, Childhood leukemia and residential magnetic fields: are pooled analyses more valid than the original studies? [see comment], Bioelectromagnetics 27 (February (2)) (2006) 112118. [20] P.C. Valery, G. Williams, et al., Parental occupation and Ewing's sarcoma: pooled and meta-analysis, Int. J. Cancer. 115 (July (5)) (2005) 799806. [21] B. Schlehofer, I. Hettinger, et al., Occupational risk factors for low grade and high grade glioma: results from an international case control study of adult brain tumours [see comment], Int. J. Cancer. 113 (January (1)) (2005) 116 125. [22] A. Mannetje, K. Steenland, et al., Development of quantitative exposure data for a pooled exposureresponse analysis of 10 silica cohorts, Am. J. Ind. Med. 42 (2) (2002) 7386. [23] S.M. Rappaport, H. Kromhout, et al., Variation of exposure between workers in homogenous exposure groups, Am. Ind. Hyg. Assoc. J. 54 (11) (1993) 654662. [24] N.A. Esmen, Y.Y. Hamad, Log-normality of environmental sampling data, J. Environ. Sci. Health A12 (1 & 2) (1977) 2941. [25] P. Oldham, The nature of the variability of dust concentrations at the coal face, Br. J. Ind. Med. 10 (1953) 227234. [26] D. Glass, C. Gray, et al., Health watch exposure estimates--do they underestimate exposure? Chem. Biol. Interact. 153154 (2005) 2332. [27] P.A. Stewart, J.N. Zey, et al., Exposure assessment for a study of workers exposed to acrylonitrile. III. Evaluation of exposure assessment methods, Appl. Occup. Environ. Hyg. 11 (11) (1996) 13121321. [28] P.A. Stewart, A. Blair, et al., Estimating historical exposure to formaldehyde in a retrospective mortality study, Appl. Ind. Hyg. 1 (1) (1986) 3441. [29] D. Kahneman, P. Slovic, et al. (Eds.), Judgement Under Uncertainty: Heuristics and Biases, Cambridge University Press, 1982. [30] D. Kriebel, The dosimetric model in occupational and environmental hygiene, Occup. Hyg. 1 (1994) 5568. [31] N.C. Hawkins, J.D. Graham, Expert scientific judgement and cancer risk assessment: a pilot study of pharmacokinetic data, Risk Anal. 8 (4) (1988) 615625. [32] D.C. Glass, C.N. Gray, et al., The health watch casecontrol study of leukaemia and benzene--the story so far, Ann. NY Acad. Sci. (2005). Please cite this article in press as: D.C. Glass, et al., Ensuring comparability of benzene exposure estimates across three nested casecontrol studies in the petroleum industry in support of a pooled epidemiological analysis, Chem. Biol. Interact. (2009), doi:10.1016/j.cbi.2009.11.003