Document LKOZvM779EGLMDXbBpZJRrjkX

152 Occupational and Environmental Medicine 1997;54: 152-166 A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum marketing and distribution workers in the United Kingdom Lesley Rushton, Helena Romaniuk Department of Public Health Medicine and Epidemiology, University of Nottingham, Queen's Medical Centre, Notingham NG7 2UH L Rushton H Romaniuk Correspondence to: Dr Lesley Rushton, Department of Public Health Medicine and Epidemiology, University of Nottingham, Queen's Medical Centre, Notingham NG7 2UH. Accepted 11 September 1996 Abstract Objectives-To investigate the risk of leukaemia in workers in the petroleum distribution industry who were exposed to low levels ofbenzene. Methods-From the cohort of distribution workers, 91 cases were identified as having leukaemia on either a death certificate or on cancer registration. These cases were compared with controls (four per case) randomly selected from the cohort, who were from the same company as the respective case, matched for age, and alive and under follow up at the time of case occurrence. Work histories were col- lected for the cases and controls, together with information about the terminals at which they had worked, fuel compositions, and occupational hygiene measurements of benzene. These data were used to derive quantitative estimates of personal exposure to benzene. Odds ratios (OR) were calculated conditional on the matching, to identify those variables in the study which were associated with risk of leukaemia. Examination of the potential effects of confounding and other variables was carried out with conditional logistic regression. Analyses were carried out for all leukaemia and separately for acute lymphoblastic, chronic lymphocytic, acute myeloid and monocytic, and chronic myeloid leukaemias. Results-There was no significant increase in the overall risk of all leukaemias with higher cumulative exposure to benzene or with intensity of exposure, but risk was consistently doubled in subjects employed in the industry for > 10 years. Acute lymphoblastic leukaemia tended to occur in workers employed after 1950, who started work after the age of 30, worked for a short duration, and experienced low cumulative exposure with few peaks. The ORs did not increase with increasing cumulative exposure. The risk of chronic lymphocytic leukaemia seemed to be related most closely to duration of employment and the highest risk occurred in white collar workers with long service. These workers had only background levels of benzene exposure. There was no evidence of an association of risk with any exposure variables, and no evidence of an increasing risk with increasing cumulative exposure, mean intensity, or maximum intensity of exposure. The patterns of risk for acute myeloid and monocytic leukaemia were different from those of the lymphoid subgroups, in which duration of employment was the variable most closely related to risk. Risk was increased to an OR of 2-8 (95% confidence interval (95% CI) 0-8 to 9.4) for a cumulative exposure between 4'5 and 45 ppm-years compared with < 0 45 ppmyears. For mean intensity between 0-2 and 0-4 ppm an OR of 2-8 (95% CI 0'9 to 8.5) was found compared with < 0-02 ppm. Risk did not increase with cumulative exposure, maximum intensity, or mean intensity of exposure when treated as continuous variables. Cases of acute myeloid and monocytic leukaemia were more often classified as having peaked exposures than controls, and when variables characterising peaks, particularly daily and weekly peaks, were included in the analysis these tended to dominate the other exposure variables. However, because of the small numbers it is not possible to distinguish the relative influence of peaked and unpeaked exposures on risk of acute myeloid and monocytic leukaemia. There was no evidence of an increased risk of chronic myeloid leukaemia with increases in cumulative exposure, maximum intensity, mean intensity, and duration of employment, either as continuous or categorical variables. Analyses exploring the sensitivity of the results to the source and quality of the work histories showed similar patterns in general. However, no increases in ORs for categories of cumulative exposure were found for acute myeloid and monocytic leukaemia in the data set which included work histories obtained from personnel records still in existence, although numbers were reduced. Analyses excluding the last five and 10 years of exposure showed a tendency for ORs to reduce for chronic lymphocytic leukaemia and chronic myeloid leukaemia, and to increase for acute myeloid and monocytic leukaemia. Limitations of the study include uncertainties and gaps in the information collected, and small numbers in subcategories of exposure which can lead to wide CIs around the risk estimates and poor fit of the mathematical models. Conclusions-There is no evidence in this study of an association between exposure to benzene and lymphoid leukaemia, either acute or chronic. There is some suggestion of a relation between exposure A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum workers in the United Kingdom 153 to benzene and myeloid leukaemia, in particular for acute myeloid and monocytic leukaemia. Peaked exposures seemed to be experienced for this disease. However, in view of the limitations of the study, doubt remains as to whether the risk of acute myeloid and monocytic leukaemia is increased by cumulative exposures of < 45 ppm-years. Further work is recommended to review the work histories and redefine their quality, to explore the discrepancies between results for categorical and continuous variables, and to develop ranges around the exposure estimates to enable further sensitivity analyses to be carried out. (Occup Environ Med 1997;54:152-166) Keywords: leukaemia; benzene; petroleum distribution In this paper we present the results of a nested case-control study, in which cases and controls were identified from a cohort of petroleum distribution workers, to investigate the risk of leukaemia associated with exposure to benzene. Quantitative benzene exposure was estimated for all cases and controls. The influence of possible confounding variables was explored and sensitivity analyses were carried out to assess the effects of possible misclassification of exposure. There have been many studies reporting the carcinogenicity and haematotoxicity of benzene at high exposures, both in animal and epidemiological studies.' 2 Benzene has been associated with leukopenia, thrombocytopenia, and aplastic anaemia3-5; the exposures responsible being generally well in excess of 25 ppm. The association of leukaemia with high exposures to benzene has also been reported, with acute myeloid leukaemia being most commonly described.3 8 Evidence from human studies linking benzene to leukaemias other than acute myeloid is contadictory and inconsistent, with, for example, chronic lymphocytic leukaemia comprising only a small percentage (of the order of 2%) of the leukaemias found in these studies.691' Smokers are thought to be at an increased risk of leukaemia, in particular myeloid.'2 Cigarette smoke contains measurable quantities of benzene. Wallace'3 found smoking to be the single most important source of benzene exposure in a general population, with the average smoker (32 cigarettes a day) taking in about 1 8 mg benzene a day. Recent interest has focused on the derivation of quantitative estimates of the doseresponse relation between benzene exposure and leukaemia, with an emphasis on evaluating the risk at low exposures. Two studies examined the risk in chemical workers,'_'7 and several others have carried out risk assessments of workers exposed to benzene in the rubber hydrochloride (Pliofilm) manufacturing industry. 18-21 Most critiques of the risk assessment by Rinsky et al'8 have reassessed these projections, for example with different assumptions, exposure estimates, or mathematical models, and have not used new data.22 24 All the current risk assessments have been based on very small numbers of leukaemia deaths-for example, 14 in the Pliofilm study-and have used extrapolation to assess risk at low concentrations of ben- zene-for example, < 5 parts per million (ppm). These studies were carried out on chemical workers for whom benzene was the main solvent used. None of these risk assessments were based on employees in the petroleum industry, where daily exposures are generally well below 5 ppm, and for whom potential exposure is from a mixture of volatile organic compounds. Four recent cohort studies of oil distribution workers have reported mortalities. These are the Institute of Petroleum study in the United Kingdom,25 26 the American Petroleum Institute study27 in the United States, a Canadian study,28 and the Australian Health Watch study.29 Bisby29 has also reported can- cer incidence in the Australian study. A finding common to all these studies is some excess of leukaemia and other cancers of the haematopoietic system. In the United Kingdom study, in which 23 300 men from four oil companies were followed up for nearly 40 years, the standardised mortality ratio (SMR) for the total population of the distribution centres for all leukaemias was 108 (95% confidence interval (95% CI) 83 to 140) and for acute myeloid leukaemia was 121 (95% CI 78 to 179). For those whose last job title was driver the corresponding figures were 125 (95% CI 83 to 181) and 155 (95% CI 82 to 265). The American study found an SMR of 89 (95% CI 59 to 129) for all leukaemias and an SMR of 151 for acute myeloid leukaemia (95% CI 80 to 257). The Canadian study found an SMR of 101 for all leukaemias (95% CI 40 to 208), and an SMR of 335 (95% CI 108 to 781) for all leukaemias in drivers. The Australian study found an SMR of 220 for all leukaemias (95% CI 70 to 500), and a standardised incidence ratio (deaths plus incidence) of 360 (95% CI 170 to 660). The American, Canadian, and Australian studies have carried out nested case-control analyses of leukaemia and the other lymphatic cancers.29 31 The American case-control study estimated exposure to total hydrocarbons only. The Canadian and Australian studies estimated exposure to both benzene and total hydrocarbons, but the Canadian study produced a quantitative estimate and the Australian simply ranked exposures. The American study found no increased risk or exposure-effect relation between exposure to total hydrocarbons and leukaemia. The Australian study found an increasing risk of lymphohaematopoietic cancer with increasing exposure to benzene. Schnatter et al, in the Canadian study,30 found no dose response relation between all leukaemias and benzene intensities, which were primarily in the 0 1 and 1 0 ppm range. The strongest risk factors were smoking, a family history of cancer, and duration of employment. 154 Rushton, Romaniuk Our paper presents the result of a nested correct. There was no evidence that this error case-control study from the United Kingdom occurred in other companies. cohort. Six key areas were identified for which data had to be collected to provide quantitative exposure estimates, and to provide data on Method other variables which might potentially influ- The cases were men from the United Kingdom ence exposure. These related to: (a) work his- oil distribution cohort who (a) died before 1 tories, (b) job descriptions, (c) terminal January 1993 with a mention of leukaemia histories, (d) fuel compositions, (e) occupa- (either underlying or contributory cause) on tional hygiene measurements, and (f) possible the death certificate or, (b) had a cancer regis- confounding variables-such as smoking tration of leukaemia (international classifica- habits. tion of diseases revision nine (ICD-9) codes When the data were collected for the origi- 204-208). Altogether 91 cases were identified, nal distribution cohort only the last job title 88 from death certificates with or without can- and location for those who had left and the cer registration and three from cancer registra- current job title and location for those still tion alone. Of the 88 identified from death working on 31 December 1975 were recorded. certificates, 75 had leukaemia as an underlying For the purpose of estimating exposure over cause of death and 13 had leukaemia as a con- the working life of each case and control a tributory cause of death. Cancer registrations work history was required-that is, dates of were also received for 40 of the 88 deaths (can- starting and leaving each job, and the site. This cer registrations were not available before was carried out blind to the subject's case or 1971). In 11 of these there were differences control status. The main source of information between the diagnoses on the death certificates on work history in this study was company per- and cancer registration, in most cases the cancer sonnel records. However, their availability var- registration was less specific-for example, ied according to company record retention "other leukaemia" rather than "acute policies. For those study members for whom a leukaemia". The more specific diagnosis was personnel record could not be found or was adopted for analysis. Permission to contact incomplete, the information was obtained from histopathology departments to ask for histolog- a variety of sources including pension records, ical confirmation of the diagnoses was received medical records, and interviews with retired or from the Office Population Censuses and long service staff. Surveys (OPCS). Up to the present time con- When a work history was incomplete, it was firmation of the diagnosis has been received for necessary to make certain assumptions or 38 of the cases, with no histology record avail- assign date values-for example, the mid- able for a further 31. Histology reports have month or mid-year was used if days or months been received for nine of the 11 cases for were unknown, and a date midway between whom differences were found between the known dates was taken if there were gaps in the diagnosis on the death certificate and on the dates for a change in either a job or location. cancer registration, all of them confirming the When a work history was largely unavailable a more specific diagnosis chosen from the death typical work history was used. When assump- certificate for analysis. None of the histology tions were made these were flagged on the data reports for other cases resulted in a change of base. As a measure of the quality of the work diagnosis for the study. history, each study member was assigned a Four controls per case were selected as "job confidence" code of 1, 2, or 3, where code being both informative and efficient for the 1 indicated a complete work history, or one for power of the study.32 The controls were ran- which last job only was known, but the dura- domly selected from all men from the same oil tion of employment was < 10 years, code 2 company with a year of birth within three years indicated a partially complete work history for either side of the date of birth of the case. So which an assumption had been made, and that the controls should be at risk of the disease code 3 indicated a poor work history, and was at the time of case occurrence, the controls assigned when only the last job title was were selected from people alive and under fol- known, this was a supervisory or managerial low up at the time of case occurrence (date of post and the duration of employment was over death for the cases identified from death cer- 10 years. tificates, and date of diagnosis for those identi- None of the cases and only eight (2%) of the fied from cancer registration only). Two controls had job confidence score 3, the poor controls were erroneously selected who were category. The partially complete classification not under follow up on the relevant date. was assigned to 21 % of the work histories over- These have been excluded from the study. all (26% cases, 19% controls). Also, during preparation for an external quality Altogether 315 terminals were identified audit of the study, it was discovered that eight of from the work histories, over 95% of which the controls for the four cases from one com- were closed by the time of data collection. A pany had been incorrectly matched for age. detailed proforma was developed for collecting These controls have also been excluded from information on the terminals over the complete the analysis. This has resulted in one case, a period of the terminal's existence. Data col- patient with acute myeloid leukaemia, also lected included, for broad periods, approxi- being excluded from the analyses. After the mate numbers of staff, the range of products discovery of this error the algorithm for select- stored, sources and methods of supply, storage ing controls was rechecked and found to be capability, distribution practices for products A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum workers in the United Kingdom 155 handled at the terminal, information on working practices for different tasks carried out around the terminal, and use of protective equipment by all staff. All terminals still open were visited to obtain these data. Retired and long service employees were interviewed about closed sites. Other sources of information included site plans, booklets, photographs, company maga- zines, and other material available from com- pany libraries, and property and engineering departments. A major source of archive mater- ial on all aspects of the oil industry was the British Petroleum (BP) Archive held at Warwick University. Data on other variables which might influ- ence risk of leukaemia-such as smoking habits and jobs held before work in the oil industrywere also sought. Information on smoking was occasionally available from medical records and information on previous jobs was available from some personnel records. To gain infor- mation on the terminals retired or long service employees were interviewed and they were also asked if they could recall the smoking habits of relevant study members. However, the resulting data on smoking was sparse. Data collection for fuel compositions and occupational hygiene measurements, and the method of developing quantitative exposure estimates relating to the work histories of each study member is described in detail in a separate paper,33 and is therefore only briefly given here. The method used for this study was an extension of that developed for the Canadian case-control study.30 A representative set of benzene measurements (referred to as base estimates), which were also well characterised-that is, had values for factors considered to affect exposure-was assembled for different tasks in the distribution centres. The base estimate exposure measurements (ppm) were time weighted to eight hours (8-hour TWA). These base estimates were then adjusted to take account of changes in factors which were considered to affect exposure to give a workplace estimate. The six adjustment factors used were job activity (to adjust for known differences in tasks from those covered in the base estimate), number of loads handled per day, loading technology (in particular to adjust for differences between top splash and top submerged loading), percentage of benzene in the fuel, product mix (to adjust for handling products other than gasolines-such as diesel-during the same work), and air tem- perature. For some terminals the data on individual employees were insufficient to indicate whether drivers were assigned to road tankers carrying black oils-such as, heating oils-or white oils-such as motor gasolines, (black oils do not contain benzene). Two workplace estimates were derived when this arose, and a probability based on the product mix at the terminal was assigned. The workplace estimates were derived for each line of a work history, the base estimate being multiplied by the six modifying factors. The adjusted estimate for each job and site for each line was multiplied by the time spent for that job, and these were then summed over each study member's complete work history to give a cumulative exposure (ppm-years). Exposures were further classified qualita- tively into 12 categories according to whether they were likely to have occurred in intermittent peaks, defined by frequency (daily, weekly, or monthly), intensity (1-3 ppm, > 3 ppm), and duration (1-15 minutes, 15-60 minutes). The potential for skin exposure to benzene was estimated, for each job in the work history, as none, low, medium, or high. As well as cumulative exposure (ppm-years; both categorised into discrete ranges and as a continuous variable), mean intensity (mean 8- hour TWA, (ppm); cumulative exposure divided by duration of employment), maxi- mum intensity (maximum 8-hour TWA, (ppm); highest intensity for any job in the work history), and years of employment have been analysed. Cumulative exposure was analysed in (a) quintiles, primarily to check assumptions of linearity, and (b) in four categories which rep- resented a working lifetime of 45 years (< 0 45, 0-45-4-49, 4-5-4A99, > 45 ppm-years). To aid comparison of results between different sub- groups the cumulative exposure quintiles for all leukaemias (< 0-26, 0-26-0-59, 0-60-1-64, 1-65-4-78, > 4-79) have been used for all analyses, assuming the most probable exposure to white oil or black oil when this uncertainty arose. The categorisations for maximum and mean intensities of exposure and duration of employment were chosen by examining the distribution of these variables for the whole study sample before separating them into cases and controls. The end date for the cases was taken as the date of diagnosis when this was available or date of death otherwise. Exposures for controls were summed to the end date of the corre- sponding case. Lagged exposures of five and 10 years were also analysed excluding exposures received five and 10 years immediately before the end date. Potential confounding or effect modifying variables considered were smoking, employment status at end date, socioeconomic status based on the job title of longest duration, age, and date started work, ever had a previous job, and ever had a previous job as a driver. The highest potential skin contact in the work history was used to characterise dermal exposure. Two variables which characterised peak exposure were derived-namely, the number of years exposed to the peak-and ever experiencing the peak for more than one year. This was compared with those never experiencing any of the 12 peaks and who also had the lowest working lifetime cumulative exposure category (< 0-45 ppm-years). Although these derived variables attempt to characterise the nature of peaked exposure, the results for these variables proved difficult to interpret and should be treated circumspectly. The data were put into the Paradox database. Data manipulation and exploratory analysis was carried out with the statistical package SAS. The epidemiological package EGRET was used for case-control analyses. 156 Rushton, Romaniuk Table 1 Numbers of leukaemias by subtype Lymphoid: Acute Chronic Other Total Myeloid: Acute Chronic Other Total Monocytic: Acute Chronic Other Total Other: Acute Other Total Total leukaemia 7 31 1 39 31 11 2 44 1 1 1 3 3 2 5 91 The main comparison variable in all analyses was case-control status. Odds ratios (ORs) (the odds, or risk, of disease in higher v lowest exposure groups) were calculated, conditional on the matching, to identify those variables in the study which were associated with risk for leukaemia. Investigation of the potential effects of other variables and adjustment of the ORs to take into account confounding variables has been carried out with conditional logistic regression. The ORs have been estimated for exposure variables which were treated as categorical variables and, if the categorical analysis suggested that this was appropriate, as continuous variables to aid comparison with other case-control studies." The method of condi- tional logistic regression assumes an exponen- tial relation between the disease risk and other continuous variables, such as exposure. However, a linear model is often used for the assessment of cancer risk with epidemiological data, and was the preferred model mentioned by the United States Occupational Safety and Table 2 Characteristics of the study sample Vanrablc Cases Il (?/O) Age at 1st exposure (y): < 25 25 -34 > 35 Mean age of 1st exposure 36 (40) 33 (37) 21 (23) 2837 Date of hire: Before 1930 1930-39 1940-49 1950-59 After 1960 22 (24) 20 (22) 15 (17) 18 (20) 15 (17) Years of employment: < 10 10-19 20-29 30 39 > 40 Mean duration of employment 14 (16) 24 (27) 22 (24) 21 (23) 9 (10) 23 4 Highest potential skin exposure: None Low Medium High 28 (31) 8 (9) 35 (39) 19 (21) Cumulative exposure (ppm-y): <1 1-4 5-9 > 10 Mean cumulative exposure 44 (49) 27 (30) 11 (12) 8 (9) 56 Mean intensity (ppm): <0 02 002-0 19 0 2-0-39 > 0-4 Mean of mean intensity (ppm) 32 (36) 33 (37) 19 (21) 6 (7) (0 20 Maximum intensity (ppm): < 0-02 002-0 19 0 2-0 39 >0 4 Mean maximum intensity (ppm) Longest job: Professional or managerial Clerical Skilled Unskilled Smoking: Non-smoker Smoker Not known 31 (35) 30 (33) 9 (10) 20 (22) 0-39 4 (4) 12 (13) 63 (69) 11 (12) 5 (5) 6 (7) 79 (88) Controls o ('/o) 137 (39) 125 (35) 92 (26) 28 3 98 (28) 71 (20) 68 (19) 66 (19) 51 (14) 95 (27) 76 (21) 84 (24) 65 (18) 34 (10) 20 8 97 (27) 25 (7) 154 (44) 78 (22) 183 (52) 105 (30) 40 (11) 26 (7) 4.4 114 (32) 163 (46) 45 (13) 32 (9) 0 22 104 (29) 131 (37) 43 (12) 76 (22) 0 41 8 (2) 38 (110) 274 (76) 41 (12) 18 (5) 24 (7) 312 (88) Total I (/O) 173 (39) 158 (36) 113 (25) 28 4 120 (27) 91 (20) 83 (19) 84 (19) 66 (15) 109 (25) 100 (22) 106 (24) 86 (19) 43 (10) 21 4 125 (28) 33 (7) 189 (43) 97 (22) 227 (51) 132 (29) 51(12) 34 (8) 46 146 (33) 196 (44) 64 (14) 38 (9) 0 21 135 (30) 161 (36) 52 (12) 96 (22) 0 40 12 (3) 50 (11) 330 (74) 52 (12) 23 (5) 30 (7) 391 (88) Health Administration (OSHA) in their ruling on occupational exposure to benzene.'-4 The first approach of categorizing continuous variables facilitated the estimation of ORs for each level of exposure without constraint to any specific pattern. ` The effect of important potential biases, such as potential misclassification of exposures where product mix was uncertain (exposure to black oils which do not contain benzene, or to white oils which do), and sensitivity to quality and source of the work histories was examined in separate analyses, to assess the effect of the uncertainties on the results. Separate analyses were carried out for all leukaemias, and for four leukaemia groups: (a) acute lymphoblastic leukaemia, (b) chronic lymphocytic leukaemia, (c) acute myeloid and acute monocytic leukaemia, and (d) chronic myeloid leukaemia. Numbers of cases were too small to carry out separate analyses for other leukaemia subtypes. Previous studies have generally been too small to enable such separate analyses to be carried out, and hence, have analysed all leukaemias together. However, a recent paper by Wong3` has shown the importance of taking specificity of this disease into consideration in causation analysis. Others have also recognised the distinct aetiology, pathogenesis, treatment and prevention of these four different groups of leukaemia. Results The following sections present many ORs, many with wide confidence intervals (CIs) due to small numbers. Relatively few hypotheses were postulated for this study, although the main focus is the evaluation of the possible relation between acute myeloid leukaemia and cumulative exposure to benzene. In general, therefore, the ORs have not been used in formal testing of hypotheses, in particular as this can lead to problems of multiplicity (multiple testing) and hence difficulties in interpretation. Emphasis has been placed on exploration of patterns and magnitude of risk, with sensitivity analyses where appropriate. Table 1 gives the number of leukaemias by leukaemia subtype. Five leukaemias were "other leukaemias", three being acute and two not specified as to type or whether acute or chronic. The 13 leukaemias identified as a contributory cause of death were eight chronic lymphocytic leukaemias, three chronic myeloid leukaemias, one other myeloid leukaemia and one other lymphoid leukaemia. The underlying causes of death for these 13 leukaemias were heart disease (six), stroke (two), pneumonia (two), cancer of the pancreas (one), chronic obstructive airway disease (one), and hyperplasia of the prostate (one). All the acute leukaemias identified from death certificates were the underlying cause on the certificate. The three cases identified only from cancer registration data were one case each of acute myeloid leukaemia, chronic lymphocytic leukaemia, and unspecified leukaemia. The remaining results presented in A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum workers in the United Kingdom 157 this paper exclude the 10 controls whirch were incorrectly selected, and one corresi?onding matched case. Table 2 summarises some of the chatracteristics of the study sample. The distribution of age at first exposiure was similar for cases and controls, range ('12-6 to 57) as was date of hire from 1909 t(c 1970. Mean duration of employment was slightly higher for cases than controls, with a larger percentage of controls working for < 1 0 years. Cases and controls for whom the lon~gest job was unskilled had a much shorter meaIn duration of employment (15 5 years for casees, 11-7 years for controls) than the overall stuwdy sample. The distribution of the highest potential skin contact was similar in cases and controls, with slightly higher proportions of conitrols in the medium and high contact groups. The distribution of cumulative eiexposure was similar for cases and controls. Cunnulative Table 3 Odds ratios and (95% CIs) for selected variables for all leukaemia Variable Cases n Cumulative exposure (continuous) Cumulative exposure quintiles (ppm-y): < 0-26 0-26-0-59 0 60-1-64 1-65-4-78 > 4.79 Cumulative exposure working lifetime (ppm-y): < 0-45 0*45-4*49 4.5-44.9 > 45 Duration of employment (continuous) Years of employment: < 10 10-19 20-29 30-39 > 40 Maximum intensity (continuous) Maximum intensity (ppm): < 0-02 002-0 19 0-2-0-39 > 04 Mean intensity (continuous) Mean intensity (ppm): < 0-02 002-0 19 0-2-0-39 > 04 Highest potential skin exposure: None Low Medium High Date of hire: Before 1950 After 1950 57 33 Employment status at study end date: Not employed 68 Employed 22 Socioeconomic status: Blue collar White collar 74 16 Smoking: Non-smoker Smoker 5 6 Controls n OR (95% CI) 1-004 (0 99 to 1-02) 76 (1) 66 2-58 (1-12 to 5-92) 68 2-28 (0-95 to 5-43) 76 1-31 (054 to 3-19) 68 2-13 (0 90 to 5 03) 109 (1) 172 1-42 (0 77 to 2-61) 69 2-48 (0 73 to 3-00) 4 1-35 (0-14 to 12-8) 1-03 (1-00 to 1-05) 95 (1) 76 2-64 (1-20 to 5 82) 84 2-40 (1-03 to 5 62) 65 3-34 (1-33 to 8-41) 34 2-75 (0-87 to 8 65) 0 99 (0-82 to 1-20) 104 (1) 131 0 77 (0-44 to 1-35) 43 0-66 (0-28 to 1-53) 76 0-88 (0-46 to 1-71) 0-97 (0-67 to 1-40) 114 (1) 163 0-71 (0-41 to 1-24) 45 1-44 (0 74 to 2 79) 32 0-65 (0-25 to 1-72) 97 (1) 24 1- 16 (0-47 to 2-91) 154 0-77 (0 44 to 1-34) 78 0-83 (0-43 to 1-62) 237 (1) 117 1-28 (0-69 to 2-36) 315 (1) 39 6-29 (2-51 to 15-77) 308 (1) 46 1-46 (0-78 to 2-73) 18 (1) 24 1-78 (0-34 to 9-21) Goodness offit (P value) 0-55 012 - - 006666 0-02 0-08 - - 0 91 00-7722 0 86 0 15 - 0-74 - 0-43 <0001 0-25 0 50 exposures ranged from close to 0 to > 200 ppm-years, although 81% were < 5 ppmyears. The upper tail of the distribution was distorted by 15 subjects who were known to have worked previously for a company which marketed benzene enriched products, and hence had greater exposures to benzene. Only one of these was a case. There was a larger proportion of cases than controls with a mean intensity of exposure between 0-2 and 0-4 ppm. The distribution of maximum intensity of exposure was similar for cases and controls. Just under a third of the study sample had never experienced a peak at all, with proportionately more cases than controls in this group. More cases experienced daily or weekly peaks than controls. To provide a measure of socioeconomic status, study members were classified according to the job that they held longest within their work history-that is, as professional or managerial, clerical or office workers, skilled manual (driver, operator, or supervisor), unskilled. There were proportionately more cases than controls in the professional and clerical groups. The distribution of smokers and non-smokers was the same for cases and controls but information on smoking was unobtainable for nearly 90% of the study sample from medical or personnel records, or from interview. Of the cases 24% were in employment at the study end date (diagnosis or death) compared with only 11% of the controls. For all leukaemias together cumulative exposure tended to be highly correlated with both mean intensity of exposure (r = 0-84) and maximum intensity of exposure (r = 0-82), but not with duration of employment (r = 0-15). Similar correlations were found for leukaemia subtypes. ALL LEUKAEMIAS The results for leukaemia subtypes show rather different patterns which are hidden in the results for all leukaemias. However, to put the subgroup analyses in context and to pro- vide comparisons with previous studies the results for all leukaemias taken together are presented. Table 3 gives ORs and 95% CIs for some of the main variables of interest for all leukaemia. The OR for cumulative exposure as a continuous variable was just > 1. This result, together with the result for cumulative exposure categorised into discrete ranges, suggests that there is little evidence of an increased risk of all leukaemias with increased cumulative exposure. Fitting duration of employment as a continuous variable gave the best fitting model. The result for duration of employment as a categorical variable shows the risk being consistently > 2 for all duration groups over 10 years. A test for trend was significant (P < 0-05) and approached the 5% signifi- cance for linearity. Results of sensitivity analyses of uncertain- ties in product mix and quality and source of work histories showed similar patterns, although there were slight variations in the 158 Rushton, Romaniuk magnitudes of the ORs. The patterns of risk did not change substantially when analyses were carried out excluding any exposures in the last five and 10 years before the date of diagnosis. ACUTE LYMPHOBLASTIC LEUKAEMIA There were seven cases of acute lymphoblastic leukaemia, including the only case who was previously employed by a company marketing benzene enriched products. This case had a much greater exposure to benzene than the other cases of acute lymphoblastic leukaemia. The mean cumulative exposure for all seven cases was 29-6 ppm-years compared with 2-0 Table 4 Odds ratios (95% CIs) for chronic lymphocytic leukaemia Variable Cases n Cumulative exposure (continuous) Cumulative exposure (quintiles) (ppm-y): < 0-26 0-26-0-59 0-60-1-64 1-65-4-78 > 4.79 12 9 2 7 Cumulative exposure working lifetime (ppm-y): < 0-45 0 45-4 49 4.5-44.9 > 45 8 16 7 0 Duration of employment (continuous) Years of employment: < 20 10 20-29 10 > 30 11 Maximum intensity (continuous) Maximum intensity (ppm): < 0-02 0-02-0-19 0-2-0-39 04 15 8 2 6 Mean intensity (continuous) Mean intensity (ppm): < 0-02 0-02-0-19 0-2-0-39 > 04 16 8 5 2 Highest potential skin exposure: None Low Medium High 12 3 7 9 Date of hire: Before 1950 After 1950 23 8 Employment status at study end date: Not employed 25 Employed 6 Socioeconomic status: Blue collar White collar 21 10 Age started work: < 25 25-34 ) 35 12 9 10 Previous job driver Years of work: < 20 > 30 White collar Years of work: < 20 > 30 White collar < 20 y work white collar > 30 y work white collar Controls n OR (95% CI) 0-98 (0-91 to 1-05) 23 (1) 27 10-57 (1-27 to 87 92) 22 11 -21 (1-20 to 104-4) 28 1-33 (0-11 to 16-48) 24 5-64 (0-64 to 49-61) 41 (1) 57 1-07 (0 40 to 2 86) 26 1-22 (0-38 to 3-89) 20 1-06 (1 01 to 1 11) 53 (1) 35 1-81 (0 59 to 5 52) 36 2-11 (063 to 703) 0 61 (0 21 to 1 78) 33 (1) 49 0-31 (0-11 to 0-88) 19 0-16 (0-02 to 0 96) 23 0-51 (0-17 to 1 56) 0-51 (0 09 to 3 00) 40 (1) 60 0-25 (0-08 to 0 78) 16 0-65 (0-18 to 2 27) 8 0-64 (0 13 to 3 23) 31 (1) 7 -11 (0-24to 5-17) 60 0-29 (010 to 085) 26 0-82 (0-28 to 2-41) 91 (1) 33 0 94 (0 31 to 2-87) 118 (1) 6 17 7 (2-04 to 153-3) 107 (1) 17 3-66 (1-27 to 10-51) 43 (1) 40 0-79 (0-29 to 2-18) 41 0-87 (0 33 to 2-31) 15 0 48 (0-10 to 2 30) - 168 (054 to 520) - 1-58 (0-42 to 5-87) - 3-46 (1-16 to 10-35) - 2-16 (0-64 to 728) _ 0054 (009 to 337) - 1-51 (0-14 to 16-74) - 0-29 (0-01 to 6 62) - 11-36 (0-48 to 266-4) ppm-years for controls (reducing to 0-2 and 1-3, respectively, if the case with high exposure to benzene and matched controls are excluded). Five of the seven cases started work after 1950. The OR for cumulative exposure examined as a continuous variable was 1 02 when all seven cases were included but fell to 0-32 when the case with very high exposure was excluded. The ORs for the exposure variables in discrete ranges decreased with increasing cumulative exposure (OR = 0-27, 95% CI 0-02 to 2-9 for > 0 45 compared with < 0-45 ppm-years) and maximum intensity (OR = 049, 95% CI 0 05 to 4 9 for > 0-02 compared with < 0-02 ppm). Four cases had worked for < 10 years, with a mean duration of 15-3 years for cases compared with a mean of 19-4 years for controls. The cases tended to start work at an older age (mean 36-6 years) compared with controls (mean 28-3 years). Five of the seven cases had no or low skin contact. Only one of the cases was in employment at the date of diagnosis or death. All seven were blue collar workers. Job histories were complete or almost complete for all cases and controls. Duration of employment < 10 years (OR = 6-9, 95% CI 0 7 to 72.9) compared with > 10 years, and date of hire after 1950 (OR = 6X8, 95% CI 0-7 to 64-8) compared with date of hire before 1950 were associated with increased risk. As there were only one or two cases who had experienced each of the peak exposures the ORs for the different types of peaks and duration of time in these peaks were either not calculable or unstable due to small numbers. CHRONIC LYMPHOCYTIC LEUKAEMIA There were 31 cases of chronic lymphocytic leukaemia. The cases had a lower mean cumulative exposure (2-6 ppm-years) than the controls (3-6 ppm-years). However, in contrast to acute lymphoblastic leukaemia, the cases tended to have a longer duration of employment (mean 26-7 years) compared with controls (mean 21-7 years). There was no difference in the age of starting work. None of the cases had worked for < 10 years, compared with 21% of the controls. Nearly a third of the cases were white collar workers, compared with only 14% of the controls. This is reflected in the type of exposures experienced, with fewer cases (58%) than controls (75%) ever experiencing peaks, and more cases (48%) than controls (31%) with either no or low skin contact. Two of the controls had poor work histories, and 10 of the cases (32%) and 21 of the controls (17%) had partially complete work histories. Table 4 gives the ORs and 95% CIs for selected variables in conditional logistic models. The OR for cumulative exposure as a continuous variable was < 1. Analysis of cumulative exposure categorized into discrete ranges again showed no clear trend of increased risk with increasing exposure. The high ORs in the lowest categories of quintiles were due mainly to the lowest quintile, the base category, against which the other categories were compared, containing only one A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum workers in the United Kingdom 159 case. Combination of the first two quintiles would reduce the ORs for the remaining three to 1-57, 0-27, and 1d12, respectively. Associations with mean and maximum intensity, both as continuous and categorical variables, were negative. There is thus no evidence for an increasing risk with increased exposure. The ORs for experiencing any of the peaks (not shown) were < 1. Starting work at the age of 25 or above compared with starting work before 25, date of hire 1950 or later compared with date of hire before 1950, and having a previous job as a driver compared with not having a previous job as a driver had low ORs showing no increased risk of chronic lymphocytic leukaemia for these variables. The OR for potential skin exposure was highest in the low exposure category. The ORs were greater than unity for three variables, duration of employment as a continuous and categorical variable, socioeconomic status (white collar compared with blue collar), and employment status at the end date (employed compared with not employed). The influence of other variables on duration of employment was explored in models, as well as socioeconomic status and employment status at the end date on variables characterizing exposure. Inclusion of employment status did not substantially alter the ORs of any of the other variables but always remained significant itself. There seemed to be an interaction between duration of employment and socioeconomic status, with the highest risk being for white collar workers with over 30 years duration of employment. The results were not substantially altered by sensitivity analyses according to product mix or quality and source of the work history. Similar patterns were found for latency analyses excluding the last five and 10 years of exposure, although there was a tendency for Table 5 Odds ratios (95% CIs) for acute myeloid and monocytic leukaemia Variable Cases Controls nn Cumulative exposure (continuous) Cumulative exposure (quintiles) (ppm-y): < 0-26 0-26-0-59 0-60-1-64 1-65-4-78 > 4-79 - 6 5 6 6 8 - 35 17 23 24 22 Cumulative exposure working lifetime (ppm-y): < 0 45 7 045-449 15 4.5-44.9 9 45 0 46 51 23 1 Duration of employment (continuous) Years of employment: < 10 10-19 20-29 30-39 > 40 7 10 8 5 48 29 23 16 5 Maximum intensity (continuous) Maximum intensity (ppm): < 0-02 0-02-0-19 0-2-0-39 0-4 6 32 13 47 5 14 7 28 Mean intensity (continuous) Mean intensity ((ppm): < 0-02 0-02-0-19 0-2-0-39 0-4 6 34 14 56 10 18 I 13 Highest potential skin contact: None Low Medium High 6 30 3 12 16 51 6 28 Date of hire: Before 1950 After 1950 15 63 16 58 Ever employed as previous driver 4 Employment status at study end date 11 26 Socioeconomic status: Blue collar White collar 29 108 2 13 Age started work (continuous) Age started work: < 25 25-34 ) 35 14 49 12 50 5 22 *Assumes exposure to white oil products which contain benzene. tAssumes exposure to black oil products which do not contain benzene. OR (95% CI) 1-00 (0-96 to 1-04) (1) 1-88 (0-49 to 7-16) 1-68 (0-46 to 611) 1-60 (0-44 to 5-79) 2-38 (0-65 to 8-73) (1) 2-17 (0-77 to 6-09) 2-82 (0-82 to 9-38) 0 1-03 (0-99 to 1-07) (1) 2-65 (0-88 to 7-93) 2-76 (0-79 to 9-58) 2-86 (0-66 to 12-29) 1-86 (0-12 to 29-52) 0-66 (0-28 to 1-52) (1) 1-45 (0-51 to 4-10) 1-69 (0-46 to 6-17) 1-34 (0-37 to 4-86) 0-68 (0-17 to 2-82) (1) 1-34 (0-48 to 3-74) 2-76 (0-90 to 8-48) 0-43 (0-05 to 4-05) (1) 1-19 (0-26 to 5-42) 1-54 (0-54 to 4-39) 1-11 (0-32to3-87) (1) (0-46 to 3-17) 1-21- 1-71 (0-49 to 6-03) 2-97 (0-95 to 9-27) (1) (0-12 to 2-53) 0-56 1-01 (0-96 to 1-05) (1) 0-84 (0-36 to 1-93) 0-81 (0-25 to 2-62) Goodness offit (P value) 0-95 0-75 0-18 0-19 0-38 OR White oil* 1-00 (1) 2-57 1-60 2-16 3-21 (1) 1-95 2-78 0 Black oilt 1-00 (1) 1-67 1-47 1-78 2-16 (1) 1-61 2-28 0 0-22 0-68 0-69 0-86 (1) (1) 1-35 1-62 1-70 2-27 1-53 1-56 0-57 0-77 0-72 0-13 (1) (1) 1-20 1-31 2-89 2-84 0-43 0-46 0-84 Does not (1) converge 1-70 1-56 1-35 0-69 0-41 0-06 0-42 0-75 0 90 160 Rushton, Romaniuk the ORs for all exposure categories to reduce as the lagged exposure increased from zero to 10 years. For example, the ORs for cumulative exposure 0 45-4-5 compared with < 0-45 ppm-years reduced from 1 07 for no lagged exposure to 0-87 for a lag of five years and to 0-79 for a lag of 10 years. The corresponding ORs for > 4-5 ppm-years compared with < 0 45 were 1-22, 1-06, and 0-88. ACUTE MYELOID AND MONOCYTIC LEUKAEMIA Acute myeloid and monocytic leukaemia were considered together because of their similar aetiology. There were 31 cases of acute myeloid and monocytic leukaemia included in the analysis, all but one being myeloid. (One further case was excluded as all four controls were erroneously matched for age). In all but one case of acute myeloid and monocytic leukaemia this was given as the underlying cause of death on the death certificate, the remaining one being identified from cancer registration only. Most of the cases (94%) were blue collar workers, compared with 89% of the controls. The cases and controls had similar mean cumulative exposures (cases 3-7 ppm-years, controls 3-8 ppm-years), with the range of exposure for the controls being much greater (up to 103-8 ppm-years) than for the cases (up to 22-3 ppm-years). The cases had a slightly longer mean duration of exposure (19.1 years) than controls (16.6 years), with no difference in the age of starting work. More controls (40%) had worked for < 10 years than cases (23%). Cases had slightly more exposure to peaks than controls (75% cases, 66% controls), and slightly more medium or high skin contact (cases 71%, controls 65%). There were no cases or controls in this data set with poor work histories, but five cases (16%) and 16 controls (13%) had partially complete work histories. Table 5 shows the ORs of acute myeloid and monocytic leukaemia for exposure variables, potential confounding variables, and other variables. Unlike the lymphatic subgroups duration of employment was not the variable most closely related to risk. The ORs for cumulative exposure categorised into discrete ranges showed a tendency to increase as exposure increased, and a similar pattern was found as mean intensity of exposure as a categorical variable increased. A test for trend for cumulative exposure gave a probability value of 0 09, and a test for a linear trend was not significant. There were no cases with a cumulative exposure above 45 ppm-years, the fourth category of the working lifetime categorisation. To explore the risk between 4-5 and 45 ppmyears, ORs were calculated for different cut off Table 6 Odds ratios (95% CIs) for conditional logistic regression models for acute myeloid and monocytic leukaemia Variable OR (95% CI) Goodness offit (P value) OR White oil Black oil Model 1: Cumulative exposure: 045-449 > 4-5 Employed at study end date 1-82 (0-63 to 5 29) 2-48 (0 73 to 8 47) 2 61 (0-81 to 8-41) 0 12 1 70 1-26 2 37 1-98 2 63 2-86 Model 2: Mean intensity: 0-02-0-19 0-2-0-39 > 04 Employed at study end date 123 (043 to 348) 2-66 (0-84 to 8 47) 0-42 (0 04 to 3 94) 2-92 (0 92 to 9-29) 0-06 1-14 1-17 2 64 2-76 0-41 0-42 2 64 3 05 Model 3: Cumulative exposure: 045-449 > 45 Date of hire After 1950 2-82 (0-89 to 8 89) 3-71 (0-99 to 13-94) 2-08 (0-66 to 6-58) 0-19 2-54 2-02 3-71 2-81 2 07 1-83 Model 4: Mean intensity: 0-02-0-19 0-2-0-39 > 04 Date of hire After 1950 1-45 (0-51 to 411) 3 26 (1-00 to 10-57) 047 (005 to 438) 1 67 (0-58 to 4-83) 0-16 1-31 1-42 3-32 3-35 047 0-51 1 65 1-64 Model 5: Years at peak exposure (weekly, > 3 ppm, 15-60 minutes) 1-05 (1-00 to 1 10) 0 04 1-06 1-04 Model 6: Cumulative exposure: 0-45-4-49 > 4.5 Years at peak exposure (weekly, > 3 ppm, 15-60 minutes) 1 57 (0 49 to 5 09) 1-31 (0-25 to 698) 104 (0-98to -11) 0-19 094 1-19 0-63 1-04 1 08 1-04 Model 7: Ever experienced peak (weekly, > 3 ppm, 15-60 minutes) 186 (063 to 548) 0-24 2 19 2-13 Model 8: Cumulative exposure: 045-449 > 4-5 Ever experienced peak exposure (weekly, > 3 ppm, 15-60 minutes) 0 34 (0 04 to 2 59) 0-38 (0-04 to 3 28) 4-72 (0-57 to 39 03) 0 49 0 35 0-18 0 41 0-26 5-42 9 41 A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum workers in the United Kingdom 161 Table 7 Acute myeloid and monocytic leukaemia Excluding job confidence score 3 Variable Cases Controls n n OR Cumnulative exposure (ppm-y): < 0 45 7 46 (1) 0 45-4-49 15 51 2 17 >45 9 24 2 82 Mean intensity (Ppm-y): < 0-02 6 34 (1) 0-02-0-19 14 56 1-34 0-2-0-39 10 18 2-76 > 0-4 1 12 046 Excluding job confidence scores 2 and 3 Cases Controls n n OR 7 44 (1) 13 40 2 13 6 21 1 62 5 29 (1) 13 49 1-30 7 15 2 14 1 12 0-39 Excluding work histories from majority interview Cases Controls n n OR 7 42 (1) 14 45 1-94 7 21 2 18 6 30 (1) 13 50 1-23 8 16 2 16 1 12 0-42 Including work histories from personnel records only Cases Controls n n OR 6 31 (1) 12 26 1 42 2 15 0 33 5 19 (1) 11 33 0 85 4 13 0 73 0 70 Table 8 Odds ratios (95% CI) for lagged analyses for acute myeloid and monocytic leukaemia Excluding exposures before date of diagnosis of: 0 years 5 years Cumulative exposure (ppm-y) OR (95% CI) Goodness offit OR (95% CI) Goodness offit <045 0 45-4-49 4 5-44-9 > 45 (1) 0-18 (1) 0-16 217 (077 to 609) - 210 (070 to 625) - 2 82 (0-82 to 9-38) 3 06 (0-92 to 10 18) 0- 10 years OR (95% CI) (1) 232 (083 to 645) 3-67 (1-03 to 13 07) - Goodness offit 0 10 - points equivalent to intensities of 0-2, 0-3, and 0 4 ppm for 45 years-that is, cumulative exposure (a) between 4-5 and 9 ppm-years (OR = 2-54, 95% CI 0-64 to 10-10) and > 9 ppm-years (OR = 2-93, 95% CI 0-63 to 13-59), (b) between 4-5 and 13-5 ppm-years (OR = 2-65, 95% CI 0 75 to 9 40) and > 13-5 ppm-years (OR = 2-88, 95% CI 0 37 to 22 32), and (c) between 4-5 and 18 ppm-years (OR = 2 90, 95% CI 0-85 to 9-90) and > 18 years (OR = 1-64, 95% CI 0 15 to 18-19). The mean cumulative exposure for the con- trols was influenced by several high estimates. These also affected the OR for cumulative exposure as a continuous variable which was 1-0, but was 1-09 when the four controls with cumulative exposures > 23 ppm-years were excluded. The ORs for mean intensity and maximum intensity of exposure as continuous variables were < 1. There was little evidence for an increased risk of acute myeloid and monocytic leukaemia associated with date of hire, socioeconomic status, and age at starting work. The OR for being employed at the end date compared with not being employed was increased, although it was not nearly as high as that for chronic lymphocytic leukaemia (table 4). Table 6 gives the results of fitting different variables with cumulative exposure and mean intensity of exposure categorised into discrete ranges, and an example of the effect of including a variable defining a peaked exposure (weekly, > 3 ppm, 15-60 minutes). Inclusion of employment status at the end date with cumulative exposure or mean intensity slightly reduced the ORs for the exposure categories. Inclusion of date of hire after 1950 compared with date of hire before 1950 increased the ORs for cumulative exposure and mean intensity. No interaction effect was found for this variable. Inclusion of variables relating to peaks tended to decrease the ORs for the categories of cumulative expo- sure and give high OR values for the peak vari- able. Although the general patterns remain the same, the values of the ORs for exposure vari- ables did vary if different exposure estimates were used for those work history lines where uncertainty existed whether exposure occurred through handling white oils or whether black oils were handled. The last two columns of tables 5 and 6 show these ORs, black oil indi- cates an assumption that these products (which do not contain benzene) were handled, and white oil indicates an assumption that these products (which do contain benzene) were handled. Table 7 gives the results for cumulative exposure and mean intensity of exposure for four sets of data evaluating the sensitivity of the results to the quality of the work histories. The patterns are similar for the two data sets defined by excluding work histories with job confidence score 3 (poor) and work histories mainly taken from interviews. No increase in ORs was found for the data set which excludes work histories with job confidence scores 2 and 3 (poor and partially complete) and for the data set which only includes work histories from personnel records. However, the numbers of cases and controls on which these results were based were much reduced, especially at higher levels of exposure and in the data set which only includes work histories from personnel records. When analyses were carried out for acute myeloid and monocytic leukaemia excluding the past five and 10 years of exposure, the general patterns remained (table 8). Although the ORs were not substantially altered, there was a tendency for the ORs for the categories of cumulative exposure and mean intensity to increase, and for the fit of the models to improve, as the lag increased from zero to 10 years. 162 Table 9 Odds ratios (95% CIs) for chronic myeloid leukaemia Variable Cumulative exposure (continuous) Cumulative exposure (quintiles) (ppm-y): < 0 60 0 60-1-64 1 65-4-78 > 4.79 Cumulative exposure (working lifetime) (ppm-y): < 0-45* 0-45-4 49 4.5-44.9 > 45 Duration of employment (continuous) Years of employment: < 20 20-39 > 40 Maximum intensity (continuous) Maximum intensity (ppm): < 0-02 002-0 19 0-2-0 39 -> 04 Mean intensity (continuous) Mean intensity (ppm): < 0 02 002-0 19 0 2-0-39 > 04 Date of hire: Before 1950 After 1950 Socioeconomic status: Blue collar White collar Age started work (continuous) Age started work: < 25 25-34 > 35 Cumulative exposure: 0-60-1 64 1 65-4 78 > 4.79 Age started work: 25-34 > 35 Maximum intensity: 0-02-0 2 > 02 Age started work: 25 34 > 35 Mean intensity: 002-0-19 > 0-2 Age started work: 25-34 > 35 Cumulative exposure: 0 60-1 65 1 65-4-78 > 4.79 Years at peak exposure (monthly, >3ppm, 1 15 minutes) *Categories analysed are < 45 > 4.5. Cases II 3 3 3 2 0 9 0 4 5 1 3 4 2 1) 3 5 1 9 2 10 I 3 6 25 Controls 1i 16 12 6 10 10: 24: 10 0 14 21 9 14 18 4 8 15 20 4 5 34 10 39 5 20 13 11 OR (95% CI) 1-004 (0 94 to 1-07) (1) 139 (024 to 791) 2 45 (0-41 to 14-65) 1 05 (0-15 to 7 36) (1) 076 (014 to 406) 1-04 (0-96 to 1 12) (1) 071 (011 to 4-51) 0 63 (0-06 to 6-85) 0 81 (0-30 to 2 18) (1) 1.02 (0 20 to 5 21) 2 26 (0-29 to 17-57) 116 (015 to 887) 1004 (016 to 628) (1) 1 26 (0-27 to 5-82) 2 30 (030 to 17 39) 1-06 (0 09 to 12 47) (1) 050 (003 to 761) (1) 080 (009 to 685) 1-01 (0 95 to 1 08) (1) 2-88 (0-63 to 13-28) 1-23 (0-18 to 8 42) 3 29 (0-40 to 27 08) 4-83 (0 54 to 43 21) 0-74 (0 10 to 5 62) 6-45 (0-85 to 49-00) 2-22 (0-22 to 22-48) 1 50 (0-24 to 9-21) 1 30 (0 22 to 7-72) 315 (053 to 1883) 1-29 (0-16 to 10-46) 1-75 (0 33 to 0.28) 156 (021 to 11-33) 3 25 (0-58 to 18-35) 1 30 (0-14 to 11-85) 1-69 (0 25 to 11-37) 2-98 (0-42 to 21 03) 1-49 (0 17 to 12-83) 1-42 (0 85 to 2-37) Rushton, Romaniuk Goodness offit (P value) 0-89 0-78 0-74 0-28 0-92 0 60 0-88 0-997 0 88 - 0-61 0-84 0-75 0 33 OR White oil 1 003 (1) 0 98 2 94 1 12 (1) 0 76 0 82 (1) 0 99 2-68 1 03 0 98 (1) 130 2 26 0 90 Black oil 1-005 (1) 1-63 2 64 1 14 (1) 0 76 0-82 (1) 1 02 2 26 1-16 1 02 (1) 1 26 2 30 1-06 0 38 2 33 6-32 -- 0082 6 87 7.32 0 82 - 6641 10-48 - 2219 2 29 0-66 1950 -- 130 3315 --- 129 1 50 1 30 3 15 1-29 0 61 1 84 - 141 - 3.42 1 38 1-75 1-56 3.25 1 30 0 57 1 22 - 3383 - 1*55 - 139 1 84 3 11 1-55 1 41 CHRONIC MYELOID LEUKAEMIA There were 11 cases of chronic myeloid leukaemia. (One case of acute myeloid leukaemia also had chronic myeloid leukaemia on the death certificate but was classified as acute myeloid leukaemia for this study). Nine of the 11 cases were in the skilled manual socioeconomic group. Four of the controls had poor work histories with seven of the cases (64%) and 12 of the controls (27%) having almost complete work histories. Nine cases had a cumulative exposure between 0 45 and 4-5 ppm-years, (mean 5-4 ppm-years for cases, 4-9 ppm-years for controls), with no cases having a cumulative exposure of < 0-45 ppmyears. Five cases had an mean intensity between 0-02 and 0-2 ppm. Seven cases had more than 30 years duration of employment (mean for cases 29-8, controls 26 5). Table 9 shows these patterns. The first two quintiles of cumulative exposure and the categories < 0-45 and 0 45-4-49 of working lifetime cumulative exposure have been combined as there were no cases in the first A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum workers in the United Kingdom 163 quintile (< 0-45). Increasing duration of some assumptions to complete the work histo- employment analysed as a continuous variable ries. Analysis by smoking and previous jobs was associated with increased risk. The ORs was restricted due to incomplete information. for cumulative exposure and mean intensity as The limitations of the methods for esti- continuous variables were just > 1. There was mating exposure are discussed in detail in a no evidence of an increasing risk with increas- separate paper.33 Incomplete or missing infor- ing categories of cumulative exposure, dura- mation which had to be assumed-for exam- tion of exposure, maximum intensity, or mean ple, hygiene data for base estimates, data on intensity of exposure, with the only increased closed terminals and on product source-con- ORs being in categories 0-2-0-4 ppm for max- tributed to uncertainties in the exposure esti- imum and mean intensity of exposure, based mates. However, a validation procedure with on only two cases. independent hygiene data showed that the The ORs for experiencing peaks were not methods of estimation produced reasonable calculable as all the cases experienced some estimates when compared with known sam- kind of peak-that is, there were no cases in pling results. the "never experienced a peak and cumulative A restriction of this study, which was per- exposure in the lowest working lifetime cate- haps not anticipated before exposure estima- gory" group. Unlike the results for acute tion, is that the ranges of both intensity and myeloid and monocytic leukaemia, duration of cumulative exposure are narrow. Half the employment > 20 years compared with < 20 study sample had < 1 ppm-years of cumulative years, date of hire 1950 or later compared with exposure, with over 80% < 5 ppm-years. The date of hire before 1950, and white collar precision of this study to estimate risks > 10 workers compared with blue collar workers ppm-years is therefore reduced. showed reduced ORs and inclusion of these In general cumulative exposure was highly variables with other variables in conditional correlated with both mean and maximum logistic regression models did not improve the intensity of exposure but not with duration of fit of any of the variables. There is thus no evi- exposure. The lack of correlation with dura- dence of an increased risk of chronic myeloid tion of exposure was partly due to about a leukaemia with these variables. third of the study sample having background Starting work at the age of > 25, compared or population levels of exposure to benzene for with < 25, acted as a negative confounding their entire working life. variable by increasing the ORs for the lower An attempt was made to characterise the cumulative exposure categories and to a lesser peaked nature of the exposure experienced in extent for the lower mean and maximum oil distribution centres with a qualitative clas- intensity categories. No significant interactions sification combining frequency, intensity, and were found with any of the potential con- duration of exposure. The classifications have founding variables, including age started work, been assigned primarily from the job titles in or other variables. the work histories, but take into account the Similar to the results for acute myeloid type of terminal. They are limited in their use leukaemia, some of the results were sensitive for describing in any detail the quantitative to choice of exposure estimate where there was nature of peak exposure and should therefore uncertainty as to product mix. The last two be interpreted with caution. columns of table 9 show the ORs for exposure The results for all leukaemias taken together to white and black oil. However, the general are obscured by the differing patterns of risk patterns already described were not altered for found in the leukaemia subtypes. However, these estimates. The patterns were also similar the OR for cumulative exposure when treated for sensitivity analyses by quality and source of as a continuous variable was almost exactly the work history. same as that found by Schnatter et al in the There was no change in the results when analysis of the Canadian data.30 The influence exposures were excluded for the five years of chronic lymphocytic leukaemia on the before the date of diagnosis. There was a slight results for total leukaemia is reflected in the reduction of the OR for the cumulative expo- dominance of duration of employment as the sures categories assuming a lagged exposure of variable that best related to risk of leukaemia, 10 years (OR = 0 40, 95% CI 0-05 to 3-5 for and the lack of such a relation with either > 4-5 compared with < 4-5 ppm-years). cumulative exposure or intensity. There was also a suggestive relation with duration of employment in the Canadian study. Discussion There were only seven cases of acute lym- Although this is the largest case-control study phoblastic leukaemia. This tended to occur in to be carried out to investige the risk of workers employed after 1950 for a short dura- leukaemia associated with exposure to ben- tion who experienced low cumulative exposure zene, there are still only 91 cases, and fewer with few peaks. The ORs did not increase with when the leukaemia subtypes are considered. increasing exposure. There was no evidence The study may be subject to errors inherent to that the cases of acute lymphocytic leukaemia case-control studies, such as bias in selection left employment after a short time due to any of controls and recall bias.32 About 20% of the adverse working conditions. There have been work histories were incomplete (sometimes very few published reports of an association of due to reliance on interviews for obtaining the acute lymphoblastic leukaemia with exposure information). Although many of these were to benzene and the results from this study sup- minor omissions, it was necessary to make port this. 164 Rushton, Romaniuk The risk of chronic lymphocytic leukaemia with long service. Similar patterns of peak seemed to be related most closely to duration exposure were found in chronic myeloid of employment with the highest risk occurring leukaemia as in acute myeloid and monocytic in white collar workers with long employment. leukaemia. It is known that a proportion of These workers had background or population chronic myeloid leukaemia cases do progress levels of exposure to benzene. Many cases to acute myeloid leukaemia,'8 but this occurred when in employment. There was no occurred in only one case in this study evidence for an association of risk with any (included in acute myeloid and monocytic exposure variables, and no evidence of an leukaemia with haematological evidence pro- increasing risk with increasing cumulative vided by the hospital). Chronic myeloid exposure, mean intensity, or maximum inten- leukaemia is also related to the Philadelphia sity. The higher ORs in the lowest categories chromosome anomaly.38 of quintiles were an artifact of the definition of We analysed the sensitivity of the results to the categories, chosen to be consistent with the source of the work histories. In general the the analyses for the other subtypes. The ORs patterns were similar although the magnitude for any type of peak exposure were low or not of the ORs and width of CIs varied. A potential calculable showing that any exposure that was source of bias might have been introduced experienced did not occur in the form of through information being more readily avail- peaks. As previously stated, there has been little able for surviving controls, particularly from evidence in previous studies of a relation interviews. However, results were similar between exposure to benzene and chronic when these data were excluded. In data sets lymphocytic leukaemia and our results seem to which included only work histories taken substantiate this. mainly from personnel records which still exist Acute myeloid leukaemia has been most fre- the patterns were similar for three of the quently reported as being associated with leukaemia subtypes but different for acute exposure to benzene.35 8 In contrast with myeloid and monocytic leukaemia. However, chronic lymphocytic leukaemia 94% of cases the numbers of cases and controls were greatly of acute myeloid and monocytic leukaemia in reduced, adding to the uncertainty of the esti- the current study occurred in blue collar work- mates and the predominance of random varia- ers. Risk of acute myeloid and monocytic tion. The strict criteria of inclusion of the data leukaemia was highest in men with cumulative sets from existing personnel records meant exposure of 4-5-45 ppm-years compared with that some fairly reliable work histories were cumulative exposure < 0-45 ppm-years. Only excluded. Examples of exclusions are; short one man (a control) had cumulative exposures service workers (< 5 years) where the person- > 45 ppm-years. There was no association nel record does not exist any more, although with cumulative exposure when analysed as a the original cohort data were abstracted from continuous variable. The excess of cases with personnel records; histories from regular med- cumulative exposures of 4-5-45 ppm-years ical records which were consistent for job title occurred particularly in men with daily or and location; and complete work histories weekly peaks of exposure. Risk also increased which had most of the history from medical to > 2 for an mean intensity of 0-2-0-4 ppm records but some from personnel records. It compared with < 0-02 ppm. No clear trend was not possible to carry out analyses examin- was found in analyses of risk according to ing the sensitivity of the estimates to the quality maximum intensity of exposure. The recent of the terminal information. reanalysis of the Pliofilm cohort showed dou- There were occasional inconsistencies in ble the risk of all leukaemias (nine out of 14 results for some exposures when analysed as were acute myeloid leukaemias) at 55 ppm- continuous variables or categorised into dis- years cumulative exposure.2' crete ranges. This suggests that the assump- It was not possible to assess the effect of tions inherent in the analysis as a continuous smoking on the risk of acute myeloid and variable may not be justified. As Rothman35 monocytic leukaemia in this study due to lack has pointed out, categorisation facilitates the of data. Siegel'2 reviewed 12 studies which estimation of the ORs for different levels of investigated the risk of myeloid leukaemia in exposure without the relational constraints smokers, and found 10 of them to be consis- imposed for continuous variables. tent with an increased risk of about 1-5. In several of the subtype analyses, date of Smoking is thus not a potent confounder and hire had a significant effect on the risk, with would be unlikely to affect the shape of a dose- the risks generally being higher for those who response relation. However, it could be an started work after 1950. It should be noted, additional source of benzene which, in this however, that this nested case-control study is study, could not be taken into account, and derived from a cohort with criteria for entry could also be an effect modifier. including at least one continuous year's work There were 11 cases of chronic myeloid between 1950 and 1975. Thus, men who both leukaemia. There was no evidence of an joined and left before 1950 were not included increased risk as cumulative exposure, maxi- in the cohort. The lack of inclusion of any mum intensity, mean intensity, and duration leukaemia cases occurring before 1950 means of employment increased, either as continuous that the risk before 1950 cannot be adequately or categorical variables. There were no cases evaluated. with an exposure < 0 45 ppm-years, with nine The other confounding variable which of the 11 being exposed between 0-45 and 4-5 showed consistently raised ORs was employ- ppm-years. Most were blue collar workers ment status at each subject's study end date, A case-control study to investigate the risk of leukaemia associated with exposure to benzene in petroleum workers in the United Kingdom 165 with ORs being high for those in employment. Employment status at the study end date is highly correlated with duration of employment. A larger proportion of the controls had shorter service, and so were less likely to be employed at the case's date of diagnosis or death. This variable was included primarily to assess the existence of a healthy worker effect, the process by which those remaining employed tend to be healthier than those leaving employment. Under the assumption that there was no effect of exposure on risk of disease, study members who were employed for longer will on average have a higher cumulative exposure than those with shorter employment histories, but due to the healthy worker effect, risk may seem lower in the higher cumulative exposure group.39 Exposure may thus seem to be protective. If the ORs for exposure variables go up when this variable is included this would be evidence of such an effect. In the present study there was little evidence of a healthy worker effect occurring for any of the leukaemia subtypes. In fact, there was a tendency for the exposure ORs to go down slightly or remain the same when employment status was included in the models. Conclusions This study on assessment of the risk of leukaemia with exposure to low concentrations of benzene emphasises the importance of recognising that the separate leukaemia subtypes may have different aetiologies. This has been substantiated by the varying patterns of risk shown in these results, particularly between the lymphocytic and myelogenous groups. The limitations of the study discussed earlier should be taken into account when interpreting these results, including the wide CIs of some of the estimates, the poor fit of some of the models, and inconsistencies between the results obtained when the variables characterising exposure are continuous or categorical. There is no evidence in this study of an association between exposure to benzene and lymphocytic leukaemia, either acute or chronic. There is some suggestion of a relation between exposure to benzene and myeloid leukaemia, in particular acute myeloid and monocytic leukaemia. The nature of the exposure experienced for this disease also seemed to be peaked. However, in view of the limitations, doubt remains as to whether the risk of acute myeloid and monocytic leukaemia is increased by cumulative exposures of < 45 ppm-years. This study estimated ORs for exposed versus less exposed oil distribution workers in the United Kingdom. It was limited in the range of exposures experienced by the study sample. A tendency for risk of acute myeloid and monocytic leukaemia to be associated with peaked exposure has been found in the present study but further work is needed to confirm this relation epidemiologically, including quantification of peaks, and to explore possible toxicological mechanisms that could explain it. The completeness of the work histories varied and information from personnel records had sometimes to be supplemented with information from other sources, such as medical and pension records and interviews. Some attempt has been made to carry out analyses exploring the robustness of the results to uncertainties in the quality and source of this information. Some inconsistencies have arisen from these analyses. Further work is recommended to review the work histories and redefine their quality. Further exploration of the choice of cut offs for categories of variables, perhaps with smoothing techniques, would also be useful to investigate the apparent discrepancies between results for some variables analysed both continuously and categorically into discrete ranges. The development of ranges around the exposure estimates would enable other sensitivity analyses to be carried out. Although this study grouped acute myeloid and acute monocytic leukaemias for analysis, others, including the recent benzene risk characterisation carried out for the European Commission,40 considered all acute non-lymphocytic leukaemias as an entity, although acute myeloid leukaemia is the predominant cell type. There were three cases of other acute leukaemia in this study (one aleukaemic leukaemia, one erythroleukaemia, one acute leukaemia not specified but with pathological evidence of thrombocytopenia and possible aplastic anaemia) which could thus be included in a new grouping. However, these additions are unlikely to alter the results substantially. The study was funded by the Institute of Petroleum, under the general management of a Steering Group chaired by Dr Chris Roythorne. The provision of death certificates and cancer registration notifications for the cases by the Office of Population Censuses and Surveys is gratefully acknowledged. The study has benefited from the advice and support from an independent Scientific Advisory Board. The third member of the research team at the University of Nottingham was Ms Sarah Grace, who was responsible for all the clerical and data handling aspects of the study. We also thank other members of the Department of Public Health Medicine and Epidemiology who contributed to the study. Many people have been involved in the four oil companies and in other organisations such as the Transport and General Workers Union and British Petrolium Archives, in the many aspects of this study, including data collection, development of exposure estimates, data input and checking, and statistical and computing advice. We thank them all for their contributions to the study. Finally we thank colleagues at Exxon Biomedical Sciences for their considerable support, advice, and practical help, including Mr Tom Armstrong, Ms Gail Jorgensen, Dr Mark Nicholich, and Dr Rob Schnatter. 1 McMichael AJ. Carcinogenicity of benzene, toluene and xylene: epidemiological and experimental evidence. In: Fishbein L, O'Neill IK, eds. Environmental carcinogens: selected methods of analysis and exposure measurements. Lyon: IARC, 1988:3-18. 2 Aksoy M. Benzene carcinogenicity. Boca Raton: CRC Press, 1988. 3 Vigliani EC, Saita G. Benzene and leukemia. N Engl _J Med 1964;271:872-6. 4 Aksoy M, Dincol K, Erdem S, Akgun T, Dincol G. Details of blood changes in 32 patients with pancytopenia associated with long-term exposure to benzene. Br Jf Ind Med 1972;29:56-64. 5 Vigliani EC, Forni A. Benzene and leukemia. Environ Res 1976;11:122-7. 6 Aksoy M, Erdem S, Dincol G. Types of leukemia in chronic benzene poisoning. A study in 34 patients. Acta Haematologica 1976;55:65-72. 7 Brandt L, Nilsson PG, Mitelman F. Occupational exposure to petroleum products in men with acute non-lymphocytic leukaemia. BMJ 1978;i:553. 8 Vigliani EC. Leukemia associated with benzene exposure. 166 Rushton, Romaniuk Ann NYAcad Sci 1976;271:143-51. 9 Infante PF, Schwartz E, Cahill R. Benzene in petrol: a con- tinuing hazard. Lancet 1990;336:814-5. 10 McMichael AJ, Spirtas R, Kupper LL, Gamble JF. Solvent exposure and leukemia among rubber workers: an epidemiologic study. Jf Occup Med 1975;17:234-9. 11 Checkoway H, Wilcosky T, Wolf P, Tyroler H. An evaluation of the associations of leukemia and rubber industry solvent exposures. Am Ind Med 1984;5:239-49. 12 Siegel M. Smoking and leukaemia: evaluation of a causal hypothesis. Am I Epidemiol 1993;138:1-9. 13 Wallace LA. The exposure of the benzene population to benzene. Cell Biol Toxicol 1989;5:297-314. 14 Ott MG, Townsend JC, Fishbeck WA, Langner RA. Mortality among individuals occupationally exposed to benzene. Arch Environ Health 1978;33:3-10. 15 Bond GG, McClaren EA, Baldwin CL, Cook RR. An update of mortality among chemical workers exposed to benzene. BrJInd Med 1986;43:685-91. 16 Wong 0. An industry wide mortality study of chemical workers occupationally exposed to benzene. I. General results. BrJ Ind Med 1987;44:365-81. 17 Wong 0. An industry wide mortality study of chemical workers occupationally exposed to benzene. II Dose response analyses. BrJ Ind Med 1987;44:382-95. 18 Rinsky RA, Smith AB, Hornung R, et al. Benzene and leukaemia-an epidemiologic risk assessment. N Engl J7 Med 1987;316: 1044-50. 19 White MC, Infante PF, Chu KC. A quantitative estimate of leukaemia mortality associated with occupational exposure to benzene. Risk Anal 1982;2: 195-204. 20 Austin H, Delzell E, Cole P. Benzene and leukaemia-a review of the literature and a risk assessment. Am J Epidemiol 1988;127:419-39. 21 Paxton MB, Chinchilli VM, Brett SM, Rodericks JV. Leukaemia risk associated with benzene exposure in the Pliofilm cohort, I: mortality update and exposure distribution. Risk Anal 1994;14:147-54. 22 Byrd DM, Barfield ET. Uncertainty in the estimation of benzene risks: application of an uncertainty taxonomy to risk assessments based on an epidemiology study of rubber hydrochloride workers. Environ Health Perspect 1989;82: 283-7. 23 Brett SM, Rodricks JV, Chinchilli VM. Review and update of leukaemia risk potentially associated with occupational exposure to benzene. Environ Health Perspect 1989;82: 267-81. 24 Lamm SH, Walters AS, Wilson R, Byrd DM, Grunwald H. Consistencies and inconsistencies underlying the quantitative assessment of leukaemia risk from benzene exposure. Environ Health Perspect 1989;82:289-97. 25 Rushton L. Further follow up of mortality in a United Kingdom oil distribution centre cohort. Br J Ind Med 1993;50:561-9. 26 Rushton L. A 39 year follow up of the United Kingdom oil refinery and distribution centre studies: results for kidney cancer and leukaemia. Environ Health Perspect 1993; 101(suppl 6):77-84. 27 Wong 0, Harris F, Smith TJ. Health effects of gasoline exposure. II Mortality patterns of distribution workers in the United States. Environ Health Perspect 1993;101 (suppl 6):63-76. 28 Schnatter AR, Katz AM, Nicolich MJ, Theriault G. A ret- rospective mortality study among Canadian petroleum marketing and distribution workers. Environ Health Perspect 1993;101(suppl 6):85-99. 29 Bisby JA. Health watch, the Australian Institute of petroleum health surveillance program, 9th report. Melbourne: Department of Public Health and Community Medicine, University of Melbourne, 1993. 30 Schnatter AR, Armstrong TW, Nicolich MJ, Thompson FS, Katz AM, Heubner WW, Pearlman ED. Lymphohaematopoietic malignancies and quantitative estimates of exposure to benzene in Canadian petroleum distribution workers. Occup Environ Med 1996;53: 773-81. 31 American Petroleum Institute. A nested case-control study of kidney cancer, leukaemia and multiple myeloma in a cohort of land-based terminal workers exposed to gasoline in the petroleum industry. Washington: Health and Environmental Services, 1994. 32 Breslow NE, Day NE. Statistical methods in cancer research. Vol IL the analysis of case-control studies. Lyon: International Agency for Research on Cancer, 1980. (IARC Scientific Publications No 32.) 33 Lewis SJ, Bell GM, Cordingley N, Pearlman ED, Rushton L. Retrospective estimation of exposure to benzene in a leukaemia case-control study ofpetroleum marketing and distribution workers in the United Kingdom. Occup Environ Med 1997;54:167-75. 34 Occupational Safety and Health Administration. Final rule on occupational exposure to benzene. Cincinnati: OSHA, 1987. (Federal Register 52.) 35 Rothman KJ. Modern epidemiology. Boston: Little, Brown, 1986. 36 Wong 0. Risk of acute myeloid leukemia and multiple myeloma in workers exposed to benzene. Occup Environ Med 1995;52:380-4. 37 Wintrobe MM, Lee GR, Boggs DR, et al. Clinical haematol- ogy. Philadelphia: Lea and Febiger, 1981. 38 Linet MS. The leukaemias-epidemiologic aspects. New York: Oxford University Press, 1985. 39 Arrighi HM, Hertz-Picciotto I. The evolving concept of the healthy worker survivor effect. Epidemiology 1994;5: 189-96. 40 Exxon Biomedical Sciences. Benzene risk characterisation report prepared for EUROPIA, CONCA WE, and CEFIC. New Jersey: Exxon Biomedical Sciences, 1996. Vancouver style All manuscripts submitted to Occup Environ Med should conform to the uniform requirements for manuscripts submitted to biomedical journals (known as the Vancouver style.) Occup Environ Med, together with many other international biomedical journals, has agreed to accept articles prepared in accordance with the Vancouver style. The style (described in full in the BMJY, 24 February 1979, p 532) is intended to standardize requirements for authors. References should be numbered consec- utively in the order in which they are first mentioned in the text by Arabic numerals above the line on each occasion the reference is cited (Manson' confirmed other reports2-5 . . .). In future references to papers submitted to Occup Environ Med should include: the names of all authors if there are seven or less or, if there are more, the first six followed by et al; the title of journal articles or book chapters; the titles of journals abbreviated according to the style of Index Medicus; and the first and final page numbers of the article or chapter. Titles not in Index Medicus should be given in full. Examples of common forms of references are: 1 International Steering Committee of Medical Editors, Uniform requirements for manuscripts submitted to biomedical journals. BMJ 1979;1:532-5. 2 Soter NA, Wasserman SI, Austen KF. Cold urticaria: release into the circulation of histamine and eosinophil chemotactic factor of anaphylaxis during cold challenge. N Engly Med 1976;294:687-90. 3 Weinstein L, Swartz MN. Pathogenic properties of invading micro-organisms. In: Sodeman WA Jr, Sodeman WA, eds. Pathologic physiology, mechanisms ofdisease. Philadelphia: W B Saunders, 1974:457-72.