Document r6kVMwbz0nZ6aME8BG04Bmnkv

(nternationalJournal of Epidemiology 0 InternationalEpidemiologicalAssociation 1992 Vol. 21, No. 6 Printed in Great Britain Occupational Risk Factors for Acute Leukaemia: A Case-Control Study s RICHARDSON,* R ZIITOUN,** S BASTUJI-GARIN,* V LASSERRE,+C GUIHENNEUC,+ M CADIOU,** F VIGUIEfAND I LAFFONT-FAUST4 Richardson S (INSERM U 170, 16 Avenue Paul Vaillant Couturier, 94807 Villejuif, France), Zittoun R, Bastuji-GarinS, Lasserre V, Guihenneuc C, Cadiou M, Viguie F and Laffont-Faust I. Occupational risk factors for acute leukaemia: A Case-control study. InternationalJournal o f Epidemiology 1992; 21: 1063-1073. A case-control study has been performed for occupational risk factors of acute leukaemia, based on 185 cases more than 30 years old and 513 matched controls. There was a significant excess of polyvalent farming and electronic engineers among professionsof cases, and, in addition of metal workers when consideringthe professionspursued for more than 5 years. The corresponding exposures were analysed through a detailed questionnaire, and assessed by an industrial hygienist after blinding the case-control status. The odds ratios (OR) were computed after adjustment on matching variables and prior chemo- or radiotherapy treatment, and after stratification for the level and total duration of exposure. There was no excess of professionalexposure to ionizing radiation among cases. A significant relationship was observed between acute leukaemia and high or medium exposure to benzene, as well as over 10 years high or medium exposure to exhaust gas. In addition a significant relationship was observed with exposure to pesticides-insecticides and/or weed killers-and to electric and magnetic fields (EMF). The relationship with pesticides was significant when considering high or medium exposure to weed killers and more than 10 years exposure to both subtypes of pesticides. The relationshipwith pesticidesand EMF remained significant when confounding factors were taken into consideration and after adjustment on co-exposure to benzene. The cytological studies showed that acute leukaemiasfollowing exposure to benzene (high or medium) and to EMF were only of myelogenous subtypes, whereas those following exposure to pesticides were divided between lymphoblastic and myeloblastic subtypes. Cytogenetic studies failed to show increasedfrequency of chromosomal abnormalities, as described in acute leukaemias secondary to anti-cancer treatments. Our study adds credence to the hypothesis that pesticides and EMF are leukaemogenic agents, together with benzene. Besides genetic factors, certain exposures such as benzene, ionizing radiation and anti-neoplastic agents are known to cause acute leukaemia, through induction of irreversible DNA damage. Iatrogenic exposure to radiation, which is especially leukaemogenic in utero and in childhood, has been reduced over the past decades, and most `secondary' acute leukaemias occur following administration of alkylating drugs for a first cancer. Meanwhile, occupational exposure to benzene and ionizing radiation still remain potential risk factors for acute leukaemia in Western countries, despite strict limitations and security rules. Acute leukaemias secondary to these exposures share several characteristics in common including predominant * INSERM U 170, 16 Avenue Paul Vaillant Couturier, 94807 Villejuif, France. 8 . Service d'Hematologie, HBtel-Dieu, 1 Placedu Parvis Notre Dame, '75004 Paris, France. Laboratoire de statistiquesmedicales, CNRS URA 13-23.45 rue des 'Saints Peres, 75006 Paris, France. Laboratoire d'histologie, embryologie et cytogenetique, HBtel-Dieu, *75004 Paris, France. Ministere du Travail, 1 place de Fmtenay, 75007 Paris, France. myelogenous cytological subtypes, latency of several years, and a dose-dependent relative risk. Recent epidemiological studies have investigated the role of low-dose exposure to b e n ~ e n e ,i~onizing radiation, background radiation6 and of other occupational and environmental factors: especially pesticides, and non-ionizing radiation resulting from electric and magnetic fields (EMF).8 These studies yielded controversial results, depending on the methods used, the size of the sample, and the geographical area under investigation. Some methodological aspects could partly explain contradictory results: many studies have used the ICD classification without distinguishing between leukaemia subtypes, or have not taken into account possible confounding factors. In addition, the level of exposure can vary from time to time, according to the introduction in work places of new agents with leukaemogenic potential or, conversely, the withdrawal or limitation of such agents. The advantages of case-control studies of acute leukaemia as presented here are several, including the identification of multiple potential leukaemogenic 1063 cE: , 4: I 1064 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY agents, either through consecutive professional exposures, or acting simultaneously, the detailed analysis of their relative risk following evaluation of the levels of exposure and adjustment on confounding factors, as well as correlation between specific agents and certain cytological subtypes. MATERIAL AND METHODS Cases and Controls This study performed from 1984 to 1988 involved 204 cases of acute leukaemia who were over 30 years old and resident in France. Most of them were hospitalized in the clinical department of haematology at HatelDieu, Paris, the others in the department of haematology of HGpital Henri-Mondor in Creteil, near Paris. The age limit was selected to increase the chance of identifying occupational factors. The diagnosis of acute leukaemias, as well as the subtype classification, was made according to the French-American-British (FAB) Cooperative Group rite ria.^ During this period, there were 79 other cases of acute leukaemia satisfying the inclusion criteria who could not be interviewed. Of these patients, 34 had died within a week of arrival at the hospital, 43 were in a too serious condi- .tion to be interviewed (aplasia, infection . .) and two were transferred to other hospitals. Hence in total 72% of eligible cases were interviewed. During the same 4-year period, 561 controls, matched for sex, age (k5 years), ethnic group and usual residence, hospitalized in other departments of the same hospitals were interviewed. We have secondarily excluded from the analysis the cases and controls for which the interviewer had recorded poor cooperation (about 5 070 of cases and controls), the cases which were not matched with at least one control, and the controls which corresponded to excluded cases, leaving 185 cases and 513 controls. Among the 185 cases, 147 have three controls, 34 have two controls, and four have one control. The conditions of the controls were partly explained by the relative importance and cooperation of certain specialized departments in the same hospitals (ophthalmological: 30%; cardiovascular and pulmonary: 17%; diabetes and endocrinopathies: 16%; gynaecological: 12%; orthopaedic: 10%; cancer-excluding leukaemia and lymphomas: 7%; miscellaneous: 5 % ; undefined: 3%). Methods Physicians trained in industrial health administered a standardized interview to cases and their appropriate controls. The questionnaire was designed to obtain information concerning: demographic data, information on occupation and health problems of the family, past medical history including any radiotherapy or chemo. therapy treatment, drug use, a complete description of the occupational activities, and some assessment of the environmental and leisure exposure. The occupational history was assessed in detail by a separate questionnaire for each job including, apart from the job title and the branch of activity, recall of chemical and physical exposures with an estimation of exposure frequency and the use of any individual protection. The analysis of occupations was carried out on all the different job titles of cases and controls. The 185 cases recorded 484 occupations and the 5 13 controls recorded 1364 occupations (which corresponds to an average of 2.6 per case and 2.7 per control respectively). For 95 cases (51.4%)and 280 controls (54.6%) there was no recorded industrial or agricultural exposure. For the rest, the number of separate exposures recorded (out of a list of 24) was 447 for the cases and 1114 for the controls (average: 5.0 and 4.8 respectively). Then, blinding the case-control status, the job exposures over the occupational history (24 items) were reassessed by an industrial hygienist in a similar way to that used when building a job exposure matrix.Io Whenever possible, the hygienist coded the exposure into low ( <5% of the working time), medium (5-50Vo) and high (>50% of the working time). A weighted length of exposure (WLE) was defined on the basis of level and total duration of exposure: WLE = 0.025 x duration of low exposure + +0.25 x duration of medium exposure 0.75 X duration of high exposure. The weights were chosen so as to represent approximately the average percentage of exposed working time in each of the three exposure categories. The analysis was performed only on the exposures assessed by the industrial hygienist. A standard case-control analysis was conducted." Results are reported first separately for each exposure and then significant risk factors are analysed jointly. Matching was included in the design of this study SO as to ensure comparability of sociodemographic charmteristics between cases and controls. The matching variables (hospital, sex, age, ethnic group and residence) have a fixed number of categories which is independent of the number of subjects in the study. Hence in the analysis, the matching can be taken into account either by including the matching variables as adjustment variables in an unconditional lOgis*lc regression for large strataJ2 or by performing c a d i tional regressions. After assuring ourselves of the similarity of results between the two approaches, We only report results corresponding to the first approach when each risk factor 1s analysed separately. For the joint analysis of the significant risk factors, results for both unconditional and conditional logistic regression will be given. BMDP progams were used for the unconditional logistic regression and the SAS procedure pH-REG was used for the conditional analysis. RESULTS AS expected there was no significant difference between the cases and controls for the matching variables: age (mean 55.4 k 14.9 for both groups), sex (male/female: 50.2 versus 48.2), and residence (Paris, suburbs or outside). Further sociodemographic variables including ethnic origin of both parents (Europe: 89.7% versus %Yo), age at the end of education (mean: 16.5 +5 versus 16.6), socioeconomic activity, and family situation were also comparable. The professional situation at time of study was similar (working: 50.8% versus 47.6%; out of work: 2.2% versus 1.8%; retired: 36.2% versus 38.1%; no profession: 10.8% versus 12.5%). There was, however, a small but significant difference for place of birth (France: 83.2% versus 88.7%; other European countries: 3.2% versus 5.3%; Africa: 11.9% versus 5.8%; other: 1.6% versus 0.2% for cases versus controls respectively) (P= 0.004). Once 5% poor cooperation was excluded cooperation was evaluated as good in 63% versus 60% and, medium in 37% versus 40% respectively. The analysis of job titles, coded according to the International Labour Office Classification ( 1 968) showed a significant excess of polyvalent farming and electronic engineers among cases as compared to controls. When comparisons were made for professions pursued for more than 5 years, the same significant excesses were maintained, with, in addition, an elevated OR for metal workers among cases (Table 1). The relative risk estimates concerning exposure to suspected professional agents are shown in Table 2. A relatively low number of cases (8/185 = 4.3%) and controls (18/513 = 3.5%) had a previous history of chemotherapy or radiotherapy in a body area including active bone marrow. The OR was computed for comparison of professional exposures between cases and controls after adjustment for prior chemoor radiotherapy and on the matching variables. The comparisonswere made for all levels of exposures and, separately, for high or medium level exposure (HM). Professional exposure to ionizing radiation was not found to be linked to acute leukaemia in our study. ExPosure to EMF is associated with an elevated OR. This OR is strongly increased and becomes statistically significant when exposure to EMF other than arc welding is considered separately. No association is found with exposure to solvents in general but OR for benzene (HM) and oxygenated solvents are significantly increased. Moderately increased OR not reaching statistical significance, are found for HM exposure to fuel and exhaust gas. All the OR measuring the association with pesticide exposure are elevated, with a high and significant risk attached in particular to HM exposure to weed killers. No subject (case or control) was coded for exposure to ethylene oxide. Exposure to other agents which are not suspected in relation to leukaemia was also investigated as a way of checking whether our study could reproduce negative results. We did find OR close to 1.O for various agents such as disinfectants, oils and grease, metallic dust, and other dusts (results not shown). There were fewer cases than controls with a history of exposure to asbestos dusts (2.7% versus 6.49'0, OR = 0.36, 95% CI : 0.14-0.96). Other differences were noted between cases and controls when the comparison was made for weighted length of exposure which was significantly longer for benzene and exhaust gas (Table 3). In addition the adjusted OR showed a significant link with acute leukaemia in case of exposure for more than 10 years to exhaust gas, weed killers, and insecticides. As expected, professional exposures were more frequently observed for both groups in men than in women (22.6% versus 3.6%, P < 0.001). Double or multiple exposures to benzene HM, pesticides and/or EMF were more frequently observed in cases than in controls (3.78% versus 0.58%, OR = 7.20, 95% 1066 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY TABLE 2 Relative risk estimates concerning exposure to suspected agents Cases Controls ORa 95% CI No. % No. 7 0 Electric and magnetic fields - all types of exposure Arc welding Exposures other than arc welding (a) + (b)+ f (a) (b) 14 7.6 8 4.2 4 2.2 7 3.8 3 1.6 23 4.5 1.7 (0.9-3.5) 18 3.5 1.2 (0.5-3.0) 9 1.7 1.3 (0.4-4.2) 5 1.o 3.9 (l.2-12.5)* 3 0.6 2.9 (0.6-14.4) ....................................................... Ionizing radiation 4 2.2 17 3.3 0.7 (0.2-2.O) ............................................................................................................................................................................ Solvents - all types Benzene Other hydrocarbon solvents 71 38.4 188 36.7 1.1 (0.7-1.5) (a) 22 11.9 47 9.02 1.3 (0.8-2.3) (b) 15 8.1 16 3.1 2.8 (1.3-5.9)** (a) 28 15.1 71 15.0 1.o (0.6-1.6) (b) I 1 6.0 34 6.6 0.9 (0.4-1.8) Halogenated solvents (a) 44 23.8 115 22.4 1.1 (0.7-1.6) (b) 23 12.4 62 12.1 1.o (0.6-1.7) Oxygenated (a) 42 22.7 82 16.0 1.5 (1.O-2.4)' solvents (b) 15 8.1 3 1 6.0 1.5 (0.7-2.7) ........................................................................................................................................................................................... Fuel (a) 25 13.5 50 9.8 1.4 (0.8-2.4) (b) 14 7.6 26 5.1 1.5 (0.7-3.0) Exhaust gas (a) 11 6.0 35 6.8 0.8 (0.4-1.7) (b) 8 4.3 13 2.5 1.7 (0.7-4.3) ........................................................................................................................................................................................... Pesticides - all types 22 11.9 40 1.8 1.6 (0.9-2.8) Weed killers (a) 16 8.7 30 5.9 I .5 (0.8-2.8) (b) 7 3.8 6 1.2 3.5 ( I . 1-10.8)* Insecticides (a) 22 11.9 37 7.2 1.7 ( I .O-3. I ) (b) 8 4.3 11 2.1 2.1 (0.8-5.4) a Adjusted OR on the matched variables and prior history of radiotherapy or chemotherapy. +(a) All exposure. f + (b) High or medium exposure. OR significant at the 5% level. ** OR significant at the 1'70 level. three agents the OR is still doubled: (OR = 1.95; 95% CI : 1.20-3.16). Among the 31 cases and controls who had a history of exposure to benzene HM there was a significantly higher proportion of people exposed to EMF (9.7% exposed versus 1.35 non-exposed, P = 0.0005) and similarly to pesticides (22.6% versus 8.2'70, P = 0.006), and to oxygenated solvents (38.7% versus 16.8'70, P = 0.002). After adjustment for benzene HM, the OR for exposure to oxygenated solvents is reduced (OR = 1.4, 95% CI : 0.9-2.2); after further adjustments for exposure to EMF and weed killers HM its effects become weak and non-significant. The other three clearly significant exposures (benzene HM, EMF, weed killers HM) were then analysed simultaneously by logistic regression. Results from both un- OCCUPATIONAL RISK FACTORS FOR LEUKAEMIA TABLE 3 Cornparson of duration of exposure I o suspected agents between cases and controls 1067 Cases Controls ORa 95% CI P valueb Electric and magnetic fields Arc welding' WLE (m S s) Exposures other than arc welding' WLE (m f s) ...................................................... 3.0 f 3.7 1.9 S 2.9 2.9 It 3.1 8.1 i 12.9 .................. ................................. - ........................ - Solvents Benzene WLE (m f s) Years of HM exposure 4.7 f 5.1 3.3 f 6.2 - - 0 170 491 1 - < 10 7 5 4.3 (1.3-14.3)** > 10 8 11 2.1 (0.8-5.4) Other hydrocarbon solvents WLE (m f s) Years of HM exposure 0 < 10 > 10 3.8 f 6.8 174 2.6 f 4.1 479 1 - 2 15 0.4 (0.1-1.6) 9 19 1.3 (0.5-2.9) Halogenated solvents WLE (m f s) Years of HM exposure 6.0 f 8.4 3.6 f 5.6 0 162 45 1 1 - ,<10 3 26 0.3 (0.1-1.1) > 10 20 36 1.5 (0.8-2.7) Oxygenated solvents WLE (m f s) Years of HM exposure .................................... 0 <10 > 10 .................. 2.3 f 2.8 170 4 11 2.5 f 3.8 482 12 19 ...................... 1 I .O I .7 ... - (0.3-3.1) (0.8-3.6) Fuel WLE (m f s) Years of HM exposure 6.3 f 7.5 4.4 f6.4 0 171 487 1 - < 10 3 8 1.O (0.3-3.8) > 10 11 18 1.8 (0.3-3.9) Exhaust gas WLE (m =t6 ) Years of HM exposure .................................... 7.0 f 6.5 2.3 f 3.1 0 177 500 1 - 4 10 1 6 0.4 (0.1-3.7) >10 7 7 2.8 (1.0-8.3). ....................................................................................................................... Pesticides Weed killers WLE (m f s) Years of HM exposure 6.6 f 12.3 1.5 f 3.8 0 178 507 1 - < 10 1 3 1.o (0.1-9.7) > 10 6 3 6.0 (1.5-25.0)** Insecticides WLE (m f s) Years of HM exposure 5.0 f 10.7 1.5 f 3.5 0 177 502 1 - < 10 1 6 0.5 (0.1-4.1) > 10 7 5 4.0 (1.2-13.2)** NS NS 0.007 NS NS NS NS ....................... NS 0.01 ..... NS NS :Adjusted OR on the matching variables and prior history of radiotherapy or chemotherapy Non-parametric Mann-Whitney test These categories had been already speciiied on the questionnaire. WLE is the weighted length of exposure defined in Material and Methods OR significant at the 5% level ** OR significant at the 1% level 1068 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY matched and matched analyses are presented in Table 4. The three risk factors remain strongly associated with acute leukaemia. There is an overall similarity of results between both types of analyses, the matched analysis giving slightly higher relative risk estimates. The OR associated with EMF exposure is the largest but is estimated with much imprecision. TABLE 4A. Logistic regression analysing simultaneously three exposures adjusted f o r matching variables and prior history of chemotherapy or radiotherapy EXPOSURE OR CI (95Vo) Benzene (HM) Electric and magnetic fields Weed killers (HM) 2.49 3.19 3.06 1.16-5.34 0.95-10.67 0.97-9.68 TABLE 4B. Conditional logistic regression (matched analysis adjustment for prior history of chemotherapy or radiotherapy) EXPOSURE Benzene (HM) Electric and magnetic fields Weed killers (HM) OR 2.74 3.99 3.45 CI (95%) 1.27-5.94 1.06-14.69 1.05-11.34 Finally we present comparisons on morphology and cytogenetics within the case group, contrasting in particular professionally exposed and non-exposed cases. For this purpose we have chosen a global definition of 'exposed' as any patient for whom an exposure to one or more of the following agents-EMF, solvents (all types), fuel, exhaust gas and pesticidesis recorded. There were 26 cases of transformation of prior myeloproliferative or myelodysplastic syndromes. Fifteen cases corresponded to transformation of a previous chronic myelocytic leukaemia into acute leukaemia, six (7.3%) among the exposed and nine (8.7%) among the non-exposed cases. There was only one case exposed among the four cases of transformation of previous non-leukaemic myeloproliferative syndromes. On the other hand, six (7.3%) cases of transformation of previous myelodysplastic syndromes were exposed versus only one case (1%) non-exposed (P= 0.025). It is interesting to note that exposure to solvents other than benzene was found in the occupational history of all six cases of transformation of myelodysplastic syndromes. With our small number of cases of transformation, it is not possible to determine whether exposure promotes transformation or rather helps to create the primitive syndromes. To shed some light on this matter, we excluded the 26 cases of transformation from our analysis and ascertained again by joint logistic regression (conditional) the part played by benzene HM, EMF and weed killers HM. In comparison with Table 4, all the CI are increased, as expected since 26 cases and 68 controls are now excluded from the analysis. The OR for benzene HM is nearly identical to that reported in Table 4~ (OR = 2.34, 95% CI : 1.02-5.37), those for EMF (OR = 3.58, 95% CI : 0.77-16.7) and weed killers HM (OR = 3.47, 95% CI : 0.95-21.1) are a little higher but the effect of EMF is now poorly estimated and significant only at the 10% level. The distribution of the cytological types of acute leukaemia according to the FAB classification was in agreement with the one expected in adult patients, with 31 (16.5%) acute lymphoblastic leukaemia (ALL) and 154 (83.2%) acute myelogenous leukaemia (AML). All the 15 cases for whom an exposure to benzene HM and the seven cases for whom an exposure to EMF had been identified were of the AML type, whereas the 22 cases with a history of exposure to pesticides were split between ALL and AML. The relative risk was higher for ALL and pesticide exposure than it was for AML. The same pattern is also found for weed killers, insecticides and exposure to fuel (Table 5). The distribution of cytological types of acute leukaemia was also investigated for other agents but no significant results emerged. The distribution of the cytological subtypes according to the FAB classification was not different from the one expected in this age group, taking into account the number of cases resulting from transformation of previous myeloproliferative or myelodysplastic disorders. The small numbers in each subtype precluded from finding significant differences between exposed and non-exposed cases. However the exposed cases have fewer L1, and more L2, M3 and M6 subtypes: (Ll: 1.2% versus 3%; L2: 12.4% versus 9%; M3: 7.8% versus 5.2%; M6: 3.7% versus 1%). Cytogenetics were available for 102 cases. There were no more global abnormal karyotypes in patients with a history of professional exposures than in those without such exposure (54% versus 50%). A trend was observed for more deletions of chromosomes 5 or 7, Or trisomy 8 in the exposed cases (11% versus 5 % ) , but the difference was not statistically significant. Philadelphia (Ph) chromosome was identified in 1' subjects, eight (17.4%) among the exposed and nine (16.1To) among the non-exposed. Twelve of the C@eS with a Ph chromosome corresponded to transformation of previous chronic myelocytic leukaemia (Six exposed and six non-exposed); the other cases Corresponded to primary AML or ALL with Ph chromosome. Overall, we found no notable difference in cytogenetics between the exposed and the non-exPosed cases in our sample. OCCUPATIONAL RISK FACTORS FOR LEUKAEMIA -TABLE 5 Cyrologrcal types of acute leukaemia according to the type of exposure Acute lymphoblastic leukaemia (NO. = 31) / Electric and magnetic fields 0% (0) Benzene (HM) 0% (0) Fuel pesticides Weed kiilers Weed killers (HM) Insecticides 22.58% (7) OR = 2.46 (0.93-6.55) 19.35% (6) OR = 2.82 (1.01-7.92) 19.35% (6) OR = 3.61 (1.27-10.20) 6.45% (2) OR = 4.78 (0.81-28.0) 19.35% (6) OR = 3.29 (1.16-9.36) Acute myelogenous leukaemia (NO. = 154) 4.55% (7) OR = 4.83 (1.48-15.80)* 9.74% (IS) OR = 3.61 (1.69-7.7)- 11.69% (18) OR = 1.23 (0.67-2.25)' 10.39% (16) OR = 1.38 (0.73-2.62)' 6.49% (IO) OR = 1.10 (0.52-2.36). 3.25% (5) OR = 3.15 (0.92-10.8); 10.39% (16) OR = 1.51 (0.79-2.89)' 1069 The OR are calculated after adjustment on matching variables and prior history of radiotherapy or chemotherapy. All the agents included in Table 2 have been analysed but only those with increased OR either for ALL or AML are reported in this Table. * 95% c1. DISCUSSION Our study shows a significantly higher frequency of past history of occupational exposure to EMF, benzene and pesticides in patients hospitalized for acute leukaemia than in control patients hospitalized for various other reasons. As in all hospital-based case-control studies, selection bias cannot be totally excluded. The first to be addressed is a referral bias, since recruitment of neither cases nor controls is exhaustive for their respective diseases. Cases and controls were matched for place of residence in order to control this bias at the design stage. Furthermore, similarities in terms of demographic and socioeconomic characteristics were checked. In particular the proportions of shopfloor workers were nearly identical (23.9% for the cases versus 21.7% for the controls). The second potential bias arises from the disease for which the controls were hospitalized, which could be related to some occupational exposure. The diversity of control diseases make it unlikely that any particular exposure is overrepresented among controls. Besides, no exposure difference between cases and controls was found for various agents such as disinfectants, oil and grease, metallic dust and other dusts which have not been associated with leukaemia in the literature. This strongly supports the validity of our results. Concerning recall bias we note that both cases and controls were patients in a hospital and had similar motivation for answering. Patients who did not cooperate with the interviewer were excluded from the analysis, and for the others, degree of cooperation was similar. This study was designed to assess lifetime work exposure and to control for other types of work exposure with a complete work history questionnaire coded at a second stage by an industrial hygienist. This method is close to the approach described by Gerin et ai.13for assessing industrial exposure which has been shown to work well. l4 Nevertheless a questionnaire only provides a surrogate measure of exposure with possible misclassifications which reduce the power of the study to detect a difference in the exposure of cases and controls.15 However, the use of blind coding of each individual work history guarded against any differential misclassification. Breaking down the analysis by 'intensity'represented here by the frequency of exposure per unit of time-and for 'length'-total duration of exposure-has led to interesting observations: benzene was associated with leukaemia in case of high or medium intensity of exposure or longer weighted length of exposure. On the other hand a long duration of exposure of more than 10years, was more frequently observed in cases than in controls for pesticides in- 1070 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY cluding both weed killers and insecticides, and for exhaust gas. The professions corresponding to these occupational exposures were mainly farming-includingfarm owners, tenants and labourers-electronic and metal workers. The corresponding activities were agriculture, equipment manufacture, iron and steel industries, and the chemical industry. Exposure to EMF was only significantly associated with acute leukaemia when electricarc welding was excluded; the corresponding exposed occupations were engineers or technicians involved in electrical engineering or electronics, furnace workers, work near an electronic bath and jewellery polisher. One X-ray technician was included in the EMF exposure as well as in ionizing radiation exposure. Several case-control studies for occupational exposure of leukaemia patients have been reported re~ently.1~6.-20 They differ according to the definition of cases, either restricted to AML,5*1o8r extended to acute leukaemia + chronic myelocytic leukaemia. The cases in other studies included related diseases such as myelodysplastic syndromes,21-which are generally considered as preleukaemic stateslymphomas and myelomas. l9 The sizes of samples, as well as the methods of recruitment also differ from one series to the other. The present study is among the largest in terms of the total number of cases and controls; one case-control study, from Lindquist et al. was devoted specifically to acute leukaemia, with a questionnaire addressed to 125 cases and 125 controls originating from a population register.20However, this study was selectively focused on exposure to organic solvents, and concluded that there was a significant risk of developing acute leukaemia in professions like painters, and other exposed occupations, suggestingan aetiological role for benzene. Two other case-control studies yielded similar results: Flodin et al. found, from questionnaires mailed to the next-of-kin of 42 cases who died from AML and 244 controls, a sixfold increase in the leukaemic rate ratio related to solvent exposure. The additional role of background radiation was also suggested.'* Farrow et af. investigating 63 cases of myelodysplastic syndromes, found a significantly higher frequency of occupational exposure to petrol diesel fumes or liquids, and of hobbies involving chemical exposure, again suggesting an effect of benzene.21 Another case-control study" explored specifically the link between leukaemia (all types) and farming or health-related occupations and failed to show a significantly higher relative risk. Our study, based on a high number of cases and controls and on specificity of disease, shows a signifi- cant relationship between acute leukaemia and exposure to pesticides and EMF in addition to benzene. This relationship was confirmed by multivariate analysis. Benzene is a well-known leukaemogenic and car- cinogenic agent. It induces leukaemias and other haemopoietic neoplasms in several models in mice.22 Several cohort studies have shown a significantly increased risk of death from leukaemia following exposure to benzene. 6,23*25These leukaemias are mainly of acute myelogenous type, as in our study. The period of latency can be several years long: in the study from Yin et af. the cumulative mortality was in proportion to duration of exposure up to 20 years.24Standardized mortality ratios have demonstrated a marked, progressive increase with increasing cumulative exposure to benzene, leading to recommendation of lowering exposure to less than 1 p.p.m. per year. Our study, however, like that performed by Rushton and Alderson on death certificates of male oil refinery employees,25 concludes that an increased relative risk only occurs in cases of high or medium exposure or longer duration of exposure. Besides the difficulties related to assessment of exposure, this contradiction could be explained by some experimental observations according to which intermittent dosing leads to greater toxic response than chronic low-dose exposures.26 In addition, many workers are exposed to benzene from exhaust gas, leading to persistent exposure even in countries with `no benzene indu~tries'~'O. ur study shows a higher risk of acute leukaemia in cases of prolonged exposure (>10 years) to exhaust gas. This should reinforce the need for limiting exposure to benzene, whatever its sources. Several studies have shown an increased risk of leukaemia in agricultural areas among farming-related professions. 28-33 Other cancer diseases were also related to these occupations, such as malignant lymphomas, 31,33,34 multiple myeloma30 and various solid tumours including lip, prostate, stomach carcinomas30.3593a6nd soft tissue sarcomas. The methods used in these studies were mainly based on death certificates or cancer registers, and the comparisons were made with other types of cancer or the general population, allowing evaluation of the proportionate (Cancer) mortality ratio or the relative risk. Methodological differences or biases could explain why some of these studies have shown an increased relative risk of leukaemia following exposure to pesticides,28,32 while others failed to show a significant relationship. 17334'36 As emphasized by Linos et al. lack of association between leukaemia and farming in some studies, codd be a result of the size of samples, over-matching and the , OCCUPATIONAL RISK FACTORS FOR LEUKAEMIA 1071 possibility that farmers were more likely to be chosen as control subjects than other occupational groups. Another possible bias could be due to changes over time: artificial fertilizers and pesticides have been extensively used since the 1960s, and the relative risk for several tumours has been shown to have increased over time in agricultural workers.36 The responsibility of pesticides could be questioned since people working in agriculture are exposed to several other agents including chemicals (such as benzene) and, hypothetically, bovine leukaemia virus.31332 In our present series, a past history of exposure to pesticides alone was observed in 1 1 cases, to pesticides and benzene (with or without exposure to exhaust gas) in 10 cases, and to pesticides, benzene and EMF in one case. All the 16 cases exposed to weed killers also described contact with insecticides and only six out of 22 cases exposed to pesticides described contact with insecticides alone. Hence it is difficult to separate the effects of weed killers from those of insecticides. The specific role of pesticides is shown, following detailed job description and multivariate analysis, by the fact that OR for weed killers remains significant following adjustment for exposure to benzene. Another argument can be drawn from the cytological differences between acute leukaemias following the two types of exposures: those following exposure to benzene were mainly of AML subtypes, whereas those following exposure to pesticides were of ALL as well as AML subtypes. Furthermore it should be emphasized that pesticides, as other leukaemogenic agents such as benzene and ionizing radiation, can induce either bone marrow injury and aplastic anaemia following high-dose exposure,)'or leukaemia as a consequence of prolonged exposure. The role of occupational or environmental exposure to EMF has been much debated during the recent Years. The hypothesis was first raised by Wertheimer and Leeper,38 who reported an association in Colorado between residential electrical wiring configurations and childhood cancer, including leukaemia. Biological mechanisms which could be involved are largely hypothetical, since EMF, unlike benzene or ionizing radiation, are not mutagenic:39 weak EMF, acting through a calcium-modulated mechanism, can alter membrane responses to surface-stimulating molecules.40Our results suggest a relationship between to EMF and acute leukaemia, as shown in Our Preliminary analysis.41 Some caution is warranted, however, before drawing firm conclusions: the OR was statistically significant only when arc welding was excluded and there was no association with level or length of exposure. Several epidemiological studies have tried to correlate leukaemia with occupational ex- posure to A meta-analysis of studies published by 1987 has been performed by Savitz et al.: out of ten studies, five showed an increased relative risk, and the overall relative risk computed in this meta-analysis was 1.2 (95% CI : l.l-l.3).6 The risk was highest for AML compared to acute leukaemias or leukaemias in general, an observation close to that of the present study. The largest risk elevation was for aluminium workers in one but an excess risk was noted across most studies for electronics tech- nicians, electrical engineers, telegraph, radio or radar operators, and for power and telephone linesmen. Other studies have been published since this meta- analysis, which mostly indicated an increased risk of leukaemia following occupational exposure to EMF. Only Tornquist et al. did not find an increased risk of leukaemia or cancer in general in Swedish power linesmen or power station operators,49while Stern et al. found an increased risk, especially for lymphoid leukaemias, among electricians, and for myeloid leukaemias among welder^.^ A recent study from Garland et ai. has shown also an increased standard- ized incidence ratio among electrician's mates in US navy personneLS0Finally an increased risk of AML was observed by Flodin et al. among Swedish electrical workers including technicians, welders, and computer and telephone mechanics. Our study supports the belief that EMF is an occupational factor for acute leukaemia: the OR indicate a strong association bet- ween acute leukaemia and EMF, once possible con- founding factors have been taken into account, with adjustment for prior chemo- or radiotherapy treat- ment. Thus, despite the fact that, as in most other reports, exposure to EMF is only extrapolated from professional classification or job description rather than direct measurements of the fields, the hypothesis of EMF being leukaemogenic has reached consistency. This study did not disclose any other occupational risk factor, except for oxygenated solvents. There was no relationship with such professions as health care, abattoir work, painting, or exposure to styrene or ethylene oxide as shown in other studies.2 We also failed to show an increased risk of leukaemia following occupational or iatrogenic exposure to ionizing radia- tion. A clear relationship had been observed between moderate to high-dose ionizing radiation and oc- currence of AML. However, recent observations argued against considering limited fields or low-dose ionizing radiation as a major leukaemogenic factor: leukaemias-mostly AML-secondary to the treat- ment of a first malignancy, are mainly due to alky- lating agents whereas radiotherapy induces solid 1072 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY tumours rather than leukaemias. 51 In their recent study Flodin et al. have pointed out a significantly increased risk of AML after X-ray treatment, but not after X-ray examination and radiological work. In a case-control study performed by Stern et al. in a naval nuclear shipyard for all leukaemias combined, an OR above 1.0 was seen only for those workers who had accumulated at least 1.0 rem of exposure, but this, however, was not statistically significant. A recent review has shown the difficulty, for most recent studies, in confirming the classical link between occupational exposure to ionizing radiation and acute leukaemia, with, at a maximum, a rise in cases of exposure over long periods of time.52The fact that no significant difference was observed between cases and controls in our study for exposure to ionizing radiation could mean that the risk of exposure is currently low in France. This should not lead to the conclusion that ionizing radiations are not leukaemogenic, but casecontrol studies are somewhat limited in their ability to detect such associations since few workers accumulate lifetime moderate or high-dose radiation. Finally our study does not provide evidence of a specific cytological subtype and cytogenetic pattern in AML following occupational exposure. This is contrary to other authors who have observed different distribution of cytological subtypes and more abnormal karyotypes, especially of group C, with more deletions of chromosomes 5 and 7 or trisomy 8, in AML patients with a history of occupational exposure. s3*s4 However, we observed significantly more transformation of previous myelodysplastic syndromes in exposed than in non-exposed cases. The failure to observe different cytogenetic patterns in the exposed cases in the present study could be due, therefore, to the relative infrequency of these specific chromosomal abnormalities. ACKNOWLEDGEMENTS The authors wish to thank Professor H Rochant and Dr H Jouault, from Henri-Mondor Hospital in Creteil, for their collaboration in this study, S Cenee from INSERM U 170 for her expert statistical work, and C Vey for typing the manuscript. Supported by Grant no 858034 of the Institut National de la Sante et de la Recherche Medicale (INSERM). REFERENCES I Brandt L. Environmental factors and leukaemia. Med Oncol & Tumor Pharmacother 1985; 2: 7-10. Sandler D P. Epidemiology of acute myelogenous leukaemia. Seminars Oncoll987; 14: 359-64. Rosner F, Griinwald H W . Chemicals and leukaemia. I,: Henderson E S, Lister T A (eds). Leukemia 5th Philadelphia: W B Saunders, 1990, pp 271-87. Rinsky R A, Smith A B, Hornung R et al. Benzene and leukaemia, An epidemiologic risk assessment. N Engl J Med 1987; 316 1044-50. Stern F B,Waxweiler R A, Beaumont J J et 01. A case-control study of leukaemia at a naval nuclear shipyard. A m J Epidemior 1986; 123: 980-92. Flodin U, Fredriksson M, Axelson 0, Persson B, Hardell L. Background radiation, electrical work, and some other exposures associated with acute myeloid leukaemia in a case- ' referent study. Arch Environ Health 1986; 41: 77-84. Lynge E. A follow-up study of cancer incidence among workers in manufacture of phenoxy herbicides in Denmark. Br J Cancer 1985; 52: 259-70. Savitz D A, Calle E E. Leukemia and occupational exposure to electromagnetic fields: review of epidemiologic surveys. J Occup Med 1987; 2 9 47-51. Bennett J M, Catovsky D, Daniel M T et a/. Proposals for the classification of the acute leukaemias of the French-AmericanBritish (FAB) Cooperative Group. Br J Haematol 1976; 33: 451-58. "Coggon D, Pannett B, Acheson E D. Use of the job exposure matrix in an occupational analysis of lung and bladder cancers on the basis of death certificates. J Natl Cancer Inst 1984; 7 2 61-65. ' I Breslow N E, Day N E. Statistical Methods in Cancer Research. The Analysis of Case-Control Studies. Lyon: IARC Scientific Publications, 1980. l 2 Prentice R L. Use of logistic model in retrospective studies. Biometrics 1976; 32: 599-606. l3 GCrin M, Siemiatycki J , Kemper H, Begin D. Obtaining occupational exposure histories in epidemiological case-control studies. J Occup Med 1985; 2 169-77. l4 Goldberg M S, Siemiatycki J, Gerin M. Inter-rate agreement in assessing occupational exposure in a case-control study. Br J Industr Med 1986; 43: 667-76. "Kleinbaum D G, Kupper L L, Morganstern H. Epidemiologic Research. Belmont CA: Lifetime learning, 1982. 16Timonen T T T, Ilvonen M. Contact with hospital, drugs, and chemicals as aetiological factors in leukaemia. Lancet 1978; 2: 350-52. " Linos A, Kyles R A, O'Fallon W M. Kurland L T. A case-control study of occupational exposures and leukaemia. Int J Epidemiol1980; 9: 131-35. " Flodin U, Anderson L. Anjou C-G. A case-referent study on acute myeloid leukaemia, background radiation and exposure to solvents and other agents. Scand J Work Environ Health 1981; 7 : 169-78. l9 Brownson R C, Reif J S. A cancer registry-based study Of OccuPational risk for lymphoma, multiple myeloma and leukaemia, Int J Epidemiol 1988; 17: 27-32. 2o Lindquist R, Nilsson B, Eklund G, Gahrton G. Increased risk of developing acute leukaemia after employment as a Painter, Cancer 1987; 60:1378-84. 21 Farrow A, Jacobs A, West RR. Myelodysplasia, chemical exPosure and other environmental factors. Leukemia 1989; 3: 33-35. 22 Cronkite E 0. Chemical leukemogenesis: benzene as a mode', Seminars Hemat 1987; 24: 2-1 I . 23 Infante P F, Rinsky R A, Wagoner J K, Young R J. Leukaemia in benzene workers, Lancet 1977; 2: 76-78. 24 Yin S N, Li G L. Tain D F et al. Leukaemia in benzene workers: a retrospective cohort study. Br J IndustrMed 1987;44: OCCUPATIONAL RISK FACTORS FOR LEUKAEMIA 1073 25 Rushton 1,Alderson M R. A case-control study to investigate the association between exposure to benzene and deaths from leukaemia in oil refinery workers. Br J Cancer 1981; 43: 77-84. 26 Parodi S, Lutz W K, Colacci A et 01. Results of animal studies suggest a non-linear dose-response relationship for benzene effects. Environ Health Perspect 1989; 82: 171-76. 27 Nordlinder R, RamnaF 0. Exposure to benzene at different work ** places in Sweden. A n n Occup Hyg 1987; 31: 345-55. Blair A, Thomas T L. Leukemia among Nebraska farmers: a death certificate study. Am JEpidemiol 1979; 110: 264-73. 29 Burmeister L. Cancer mortality in Iowa farmers, 1971-78. J Nafl Cancer Inst 1981; 66: 461-64. 30 Burmeister L, Everett G D, Van Lier S F, Isacson P. Selected cancer mortality and farm practices in Iowa. A m J Epidemiol 1983; 118: 72-77. 31 Pearce N E, Sheppard R A, Howard J K, Fraser J , Lilley B M. Leukemia among New Zealand agricultural workers. A cancer ' registry-based study. A m 3 Epidemiol 1986; 124:4 2 - 0 9 , 32 Pearce N, Smith A H, Reif J S. Increased risks of soft tissue sarcoma, malignant lymphoma, and acute myeloid leukaemia 3 in abattoir workers. A m J Industr Med 1988; 14: 63-72. 33 Alavanja M C R, Rush G A, Stewart P, Blair A. Proportionate mortality study of workers in the grain industry. JNaflCancer Inst 1987; 18: 247-52. 1 34 Balarajan R. Malignant lymphomas in agricultural and forestry workers in England and Wales. Public Health 1988; 102: 585-92. 35 Buesching D P , Wollsstadt L. Cancer mortality among farmers. J Natl Cancer Inst 1984; 12: 503-04. 36 Wiklund K, Holm L E. Trends in cancer risks among Swedish agricultural workers. J Natl Cancer Ins1 1986; 77:657-64. 37 Najean Y, Pecking A, Le Mercier N, Zittoun R, Chaussade R. Recherche etiologique de l'aplasie mtdullaire. Analyse du sexe, de I'ige, de la profession et de I'habitat dans 251 cas d'aplasie medullaire. Nouv Presse Mhd 1978; 7: 2027-30. ,' Wertheimer N, Leeper E. Electrical wiring configuration and childhood cancer. Am JEpidemiol1979; 109: 273-84. 39 Easterly C E. Cancer link to magnetic field exposure. Am J Epidemiol 1981; 114: 169-84. Adley W R. Cell membranes: the electromagnetic environment and cancer promotion. Neurochem Res 1988; I : 671-77. 4 ' Bastuji-Garin S, Richardson S, Zittoun R. Acute leukaemia in workers exposed to electromagnetic fields. Eur J Cancer 1990; 26: 1119-20. 42 Milham S Jr. Mortality from leukaemia in workers exposed to electrical and magnetic fields, letter. NEngl JMed 1982; 307: 249. 43 Wright W E, Peters J M, Mack T M. Leukaemia in workers exposed to electrical and magnetic fields, letter. Lancet 1982; 2: 1160-61. McDowall M E. Leukaemia mortality in electrical workers in England and Wales, letter. Lancet 1983; 1: 246. Coleman M, Bell J , Skeet R. Leukemia incidence in electrical workers, letter. Lancef 1983; 1: 982-83. 46 Milham S Jr. Silent keys: leukaemia mortality in amateur radio operators, letter. Lancer 1985; 1: 812. 47 Pearce N E, Sheppard R A, Howard J K et al. Leukaemia in electrical workers in New Zealand, letter. Lancet 1985; 1: 811-12. 48 Callee E E, Savitz D A. Leukemia in occupational groups with presumed exposure to electrical and magnetic fields, letter. N Engl J Med 1985; 313: 1476-77. 49 Tornquist S, Norell S, Ahlbom A, Knave B. Cancer in the electric power industry. Br 3 Indusrr Med 1986; 43: 212-13. 50 Garland F C, Shaw E, Gorham E D, et ai. Incidence of leukaemia in occupation with potential electromagnetic field exposure in United States navy personnel. Am J Epidemiol 1990, 132: 293-303. Henry-Amar M, Aeppli D M, Anderson J et al. Workshop statistical report. IX. Study of second cancer risk. In: Somers R, Henry-Amar M, Meerwaldt J H, Carde P (eds). Trearment Strategy in Hodgkin's Dkease. Colloque INSERM n o 1%. London, Paris: Les Editions INSERM/John Libbey Eurotext, '* 1990; pp 355-80. Cartwright R A, Staines A. Acute leukaemias. In: A F Fleming (ed.) Epidemiology of Haematolagical Disease. Baillibre's Clinical Haematology, 1992; 5: 2-26. 53 Mitelman F, Brandt I, Nilsson P G. Relation among occupational exposure to potential mutagenic/carcinogenic agents, clinical findings, and bone marrow chromosomes in acute nonlymphocytic leukaemia. Blood 1978; 52: 1229-37. 54 Narod S A , Dub6 I D. Occupational history and involvement of chromosomes 5 and 7 in acute non-lymphocytic leukaemia. Concer Genet Cyrogenet 1989; 38: 261-69. (Revised version received June 1992)