Aetiological factors of the myelodysplastic syndromes (MDS) are largely unknown, with the exception of alkylating agents, ionizing radiation and benzene. Some other risk factors have been suggested by the few epidemiological studies reported (solvents, ammonia, exhaust gases, metals, pesticides, alcohol). We performed a case–control study to assess the relationship between occupational or environmental factors and MDS. Two hundred and four patients with newly diagnosed MDS, and 204 sex- and age-matched controls were included. Medical history, demographic data, lifetime exposure and hobbies were obtained. Qualitative and quantitative exposure to chemical and physical hazards were evaluated with the patients and reviewed by a group of experts in occupational exposure. The median age was 70 years and 62% of the patients were men. In univariate analyses, we found relationships between MDS and smoking habits, gardening, occupations such as health professionals, technical and sale representatives, machine operators, agricultural workers, textile workers, qualitative occupational exposures (exposed/non-exposed) to oil, solvents, ammonia, pesticides, fertilizers, cereal dusts, contact with poultry or livestock and infective risk, and lifetime cumulative exposure to solvents, oil, textile dust and infective risk. The main risk factors of MDS determined by multivariate analyses (conditional logistic regression) were, being an agricultural worker [odds ratio (OR) = 3·66; 95% confidence interval (CI) 1·9–7·0], textile operator (OR = 3·66; 95% CI 1·9–7·9), health professional (OR = 10·0; 95% CI 2·1–48·7), commercial and technical sale representative (OR = 4·45; 95% CI 1·4–14·6), machine operator (OR = 2·69; 95% CI 1·2–6·0), living next to an industrial plant (OR = 2·45; 95% CI 1·5–4·1), smoking (OR = 1·74; 95% CI 1·1–2·7) and lifetime cumulative exposure to oil (OR = 1·1; 95% CI 1·0–1·2). Further studies should be performed to assess specific exposures more precisely and it would be of interest to develop a map of haematological malignancies according to industrial background.
The myelodysplastic syndromes (MDS) are clonal disorders of haematopoietic stem cells characterized by ineffective haematopoiesis leading to blood cytopenias and by a high incidence of progression to acute myeloid leukaemia (AML) (Fenaux, 1996). Their aetiological factors still remain largely unknown and only few epidemiological studies have been conducted specifically in MDS.
The occurrence of MDS after treatment using anti-neoplastic drugs (mainly alkylating agents) is well established (Pedersen et al, 1990; Levine & Bloomfield, 1992). Similarly, reviews of cases of AML secondary to ionizing irradiation (especially in atomic bomb survivors) (Mole, 1990) and benzene exposure (Aksoy, 1985; Brandt, 1992) have shown that they were often preceded by a documented or suspected phase of MDS. A few recent case–control studies have suggested an association with occupational or environmental exposures, including exposure to petrol and diesel vapours, exhaust gases, ammonia, radiations, metals, pesticides, solvents and alcohol consumption (Farrow et al, 1989; Goldberg et al, 1990; West et al, 1995; Ido et al, 1996; Rigolin et al, 1998). We present here the results of a case–control study of environmental or occupational factors in 204 MDS cases diagnosed in the North of France.
Patients and methods
Patients and controls Incidental cases of MDS newly diagnosed between September 1991 and February 1996 in the Department of Haematology of the University Hospital of Lille (France) and classified according to French–American–British (FAB) criteria (Bennett et al, 1982) were included. MDS secondary to anti-neoplastic agents or radiotherapy for another malignant disease, and patients who did not have the ability to answer the questionnaire, were excluded. For each case, we included a control of the general population matched for sex and age (± 3 years). For each MDS patient, five names were randomly selected from the electoral register of a town or village that was itself randomly sampled on the list of towns or villages of similar size in the area (département) of the patient's residence. They all (cases and controls) received a letter inviting them to participate in ‘a study on occupational and environmental exposures and health’. When the agreement for the patient's participation was acquired, we contacted the patient in order to organize an appointment with the interviewer and also the first corresponding control on the list. In the case of absence or refusal (by the first control), the second one was contacted; this process was continued until a control for each patient was found.
As many studies published on MDS or AML (Farrow et al, 1989; Goldberg et al, 1990; Levine & Bloomfield, 1992) focused on exposure to solvents, we thought that sample size should be calculated in order to obtain at least an excess risk of 2 [ods ratio (OR) = 2] for solvents. In a preliminary study, we found a prevalence of exposure to solvents of 25% in the general population of our region. Then, the calculation of the sample size (with the hypothesis of a prevalence of exposure to solvents in 25% of controls and in order to obtain an OR of 2 with a type I error of 5% and a power of 80%) showed a minimal number of 165 patients and 165 controls. We finally included 204 patients and 204 controls.
Data All controls and patients were interviewed at home by a trained interviewer using the same questionnaire. The questionnaire provided information on demographic data, lifetime residence, medical history (past illnesses and drugs), diagnostic X-rays, radiotherapy for non-malignant disease, smoking habits or living near potential hazards (nuclear site, chemical waste, industrial plant). Estimation of the exposure to occupational or environmental factors was based on the approach developed by Siemiatycki (1984) for identifying carcinogens. Occupational history was first assessed by the description of all jobs held for at least 6 months. They were classified according to the International Labour Office (International Standard Classification of Occupation, 1990). A detailed analysis of each job was made with each patient and control in order to estimate as precisely as possible any exposure to a list of about 70 hazards (organic or mineral compounds such as solvents, glues, oils, coal tar, plastics and resins, inks, dyes, formaldehyde, acids, pesticides, fertilizers, wood preservatives, metal fumes and dusts, asbestos, exhaust gases, stone dusts, wood dusts, textile dusts, paper dusts, cereal dusts, soap, detergents, contact with birds or stock, potential exposure to virus or bacteriological agents considered as an infective risk, ionizing and non-ionizing radiations: radar, radiotransmission, high voltage). Frequency and duration of the exposure were assessed for each chemical or hazard (number of h/d; number of d/year; number of years). The number of hours and days was translated into a coefficient of full time equivalent (> 7 h = full d = 1; 5–7 h = 0·75 full d; 3–5 h = 0·5 full d and so on, and > 220 d a year = full year time = 1; 120–220 d a year = 0·75 full year time and so on). Total cumulative exposure was then expressed in full year time equivalent. Lifetime cumulative exposure was estimated by multiplying the coefficient of full day time by the coefficient of full year time and by the number of years. In order to confirm the reality of the exposure and to point out possible exposures not mentioned by the patients or controls during the interview, all questionnaires were reviewed by a group of experts in occupational hazards. This group of experts was composed of a medical occupational inspector, an occupational hygienist, a toxicologist and occupational physicians. They were blind to the status (case or control) when reviewing the questionnaires. Some exposures were then checked again with the cases and controls themselves or with the occupational physicians at the workplace.
Analysis Statistical analysis was performed on epi info and sas software. Means with standard deviation or median and interquartile range (Q1–Q3) are presented for descriptive data. Matched paired tests were used. For qualitative data, Mantel–Haenzel matched odds ratios are presented with 95% confidence intervals. For continuous variables, a paired Student's t-test was used. Multivariate analysis was performed using a conditional logistic regression with ascending stepwise regression. Personal, occupational and exposure data that gave significant results in the univariate analysis were included in the multivariate model. P-values below < 0·05 were considered as statistically significant.
Initial characteristics of the population
Between September 1991 and February 1996, 204 incident cases of MDS and 204 matched controls were included.
The median age of the patients was 70 years (interquartile range = 62–74 years) vs. 70 years (interquartile range = 62–75 years) for the controls (difference: non-significant). Men predominated (61·8% in patients and controls).
According to FAB classification, 38·2% of the patients had refractory anaemia (42 men, 36 women), 9·3% had refractory anaemia with ringed sideroblasts (10 men, 9 women), 26·5% had refractory anaemia with excess of blasts (37 men, 17 women), 18·1% had chronic myelomonocytic leukaemia (26 men, 11 women) and 7·9% patients had refractory anaemia with excess of blasts in transformation (11 men, 5 women).
Socio-economic distribution for the last job held, according to the French INSEE classification (Institut National de Statistiques et d'Etudes Economiques) did not differ significantly between patients and controls (data not shown, P = 0·29). The control population was also well representative of the population of the North of France (Nord-Pas de Calais) in terms of socio-economic distribution, as there was no significant difference (P = 0·2) between the observed socio-economic distribution of the controls and the expected distribution according to the socio-economic distribution of people in the North-Pas de Calais, as determined by the French National census of 1990, with respect to the sex and age distribution of the MDS cases.
Demographic data, medical history and environmental exposures
Twenty-five cases and 19 controls were childless (difference not significant: P = 0·4). The mean number of children did not differ between the two groups (2·6 ± 1·9 children in MDS patients and 2·7 ± 1·8 children in controls respectively; P = 0·44).
A history of haematological malignancy (including myelodysplasia) in at least one relative was reported by 7·5% of the patients and 3·5% of the controls (P = 0·08). Past immunizations and previous medical treatments could not be analysed, as many patients had forgotten those data. Sixty-one percent of the MDS patients and 48% of the controls were current or ex-smokers and the difference was significant (P = 0·0003). The cumulative quantity of tobacco smoked over life, estimated in pack-years (number of cigarette packs per d × number of years) was more important in patients (t-test = 2·3; P = 0·023), but there was no clear dose–response relationship, as people who had smoked more than 20 pack-years did not have a higher risk than those who had smoked less than 20 pack-years (Table I). MDS patients did not appear to have been more frequently exposed to diagnostic X-rays (Table II) and there was no relationship between the number of previous X-rays and MDS.
Hobbies and non-occupational exposures were evaluated semiquantitatively (never, sometimes, often) (Table III). More patients than controls used to garden more than once a week (often) (P = 0·0008). For other life-time hobbies such as ‘do it yourself’, car maintenance and hobbies exposing to chemicals (glues and paints), we found no difference between the two groups. MDS patients had not lived more frequently near potential environmental hazards such as a nuclear site, a chemical waste or a high voltage power line. However, the proximity of an industrial plant was more often reported by MDS patients than controls (P = 0·00003).
Analysis by job titles Jobs held for at least 6 months were taken into account for the lifetime occupational exposure. All men and 85·9% of the MDS women had worked compared with only 65·4% of control women. MDS cases had held at least one job significantly more often than controls (odds ratio = 3; 95% confidence interval = 1·3–7·7; P = 0·008). Employment as a plant and machine operator, skilled agricultural worker, for elementary occupations and also as manager (according to the 10 major groups of the International Standard Classification of Occupation, 1990) were more frequent in MDS patients (Table IV). Of note, however, is that major group 1 (legislators, senior officials and managers) includes heterogeneous occupations and people employed in companies with very different types of activities (whole-sale, restaurant, transport, business service, personal care). When numbers allowed, we analysed, in all major groups, specific jobs or groups of jobs that were related. None of the jobs included in the major group 1 emerged as significantly more frequent in MDS patients. In the other major groups, jobs that had been more often held by patients included ‘health professionals’ (medical doctors, dentists, veterinarians, nurses, personal care-workers) (OR = 8; 95% CI = 1·1–355; P = 0·04), ‘technical and commercial sales representatives’ (OR = 3·2; 95% CI = 1–13·7; P = 0·05), ‘machine operators’ (OR = 2·8; 95% CI = 1·3–6·4; P = 0·006), ‘agricultural workers’ (OR = 2·8; 95% CI = 1·4–6·2; P = 0·003) and ‘textile workers’ (OR = 2·8; 95% CI = 1·3–6·46; P = 0·006).
Analysis of exposures We subsequently analysed in detail occupational and environmental exposure to a list of about 70 chemical or physical hazards. A comparison between patients and controls was performed when at least four people had been exposed. The most frequent exposures are presented in Table V.
Table V. Risk estimates of occupational exposure: qualitative exposure and lifetime cumulative exposure.
|Glue adhesives||11/4||2·8 (0·8–11·8)||0·2*|
|Aromatic polycyclic hydrocarbons||17/10||1·8 (0·7–4·6)||0·93*|
|Exhaust gases||33/33||1·0 (0·5–1·9)||0·27|
|Plastic fumes and dusts||7/2||3·5 (0·7–34·5)||0·09*|
|Hydrogen peroxid||13/12||1·1 (0·4–3·1)||0·49*|
|Metal dusts||23/13||2·3 (0·9–6·0)||0·3*|
|Mineral dusts||34/40||0·8 (0·5–1·4)||0·07|
|Wood dusts||8/8||1·0 (0·3–3·3)||0·71|
|Cereal dust||31/15||2·6 (1·2–6)||0·58*|
|Infective risk||59/26||2·9 (1·7–5·4)||0·05*|
|Cotton and flax dusts||34/12||3·4 (1·6–8·2)||0·013*|
Qualitative exposure (being or not being exposed to the hazard) was analysed first, and exposure to several hazards appeared significantly more often in MDS patients, including solvents, oil, ammonia, pesticides, fertilizers, cereal dusts, contact with poultry or livestock, infective risk, textile dust, metal fumes or metal dust, as detailed in Table V. These exposures were representative of the jobs that appeared previously in excess of risk, particularly agricultural jobs, industrial jobs (solvents and metals) and textile jobs. ‘Infective risk’ could be encountered with activities involving contact with poultry or livestock, with sick people or animals, and jobs like cleaners or sweepers, or garbage collectors. Other hazard exposures were not significantly more frequent in MDS cases.
Cumulative time exposure over life was assessed for more than 70 hazards. Evaluation of the difference between cases and controls for lifetime cumulative exposure to the main hazards is presented in Table V. A significant dose–response relationship was found for the exposure to solvents (OR = 1·1; 95% CI = 1·01–1·22), oil (OR = 1·13; 95% CI = 1·02–1·24), textile dust (OR = 1·05; 95% CI = 1·01–1·09) and infective risk (OR = 1·04; 95% CI = 1–1·07). The relationship reported here (OR calculated by conditional logistic regression for one variable) means that, for instance, for solvents there was a 10% excess of risk of MDS by year of full-time exposure to solvents, for oil there was a 13% excess of risk of MDS by year of full-time exposure to oil, and so on.
Finally, a conditional logistic regression was performed, using ascending stepwise regression owing to the great number of parameters. The following parameters were included in the multivariate analysis: personal data (smoking habits, number of diagnostic X-rays, gardening, living next to an industrial plant, number of years close to an industrial plant, radiotherapy, familial history of haematological malignancy, working or not, being or not being exposed to one of the 70 hazards), professional data (International Labour Office major groups and specific occupations including agricultural workers, miners, restaurant workers, sales representatives, health professionals, machine operators, printing-machine operators, textile operators, food-machine operators, labourers) and exposure data (quantitative exposure to textile dust, petrol, solvents, oil, polycyclic aromatic hydrocarbons, exhaust gases, formaldehyde, ink, metal, pesticides, fertilizers, cereal dusts, poultry, livestock, infective risk, paper dust, ionizing radiation). The main risk factors of MDS that emerged from this multivariate study are presented in Table VI: they included agricultural workers, textile operators, health professionals, machine operators, and commercial and technical sale representatives as jobs, smoking as a personal habit, living next to an industrial plant as an environmental factor, and quantitative oil exposure as an occupational exposure.
Table VI. Occupational and environmental risk factors of MDS: multivariate analysis (conditional logistic regression with ascending stepwise regression).
|Living next to an industrial plant||2·45||(1·5–4·1)||0·007|
|Commercial and technical sales representatives||4·45||(1·4–14·6)||0·013|
Relatively few case–control studies of epidemiological factors of MDS have been performed, especially with a large number of cases, and their results are summarized in Table VII. Our study was based on the protocol established by the Cardiff group (West et al, 1995). We used the same structured questionnaire, and the exposure evaluation was based on the Siemiatycki method (Siemiatycki, 1984). In case–control studies, the choice of controls is of major importance. Contrary to West et al (1995), we did not choose controls among patients referred to other departments of our Hospital because of differences between the French and British health care systems. There is in France, in most specialities except for haematology, a large private sector that treats people with an overall higher socio-economic status, whereas public Hospitals treat a higher proportion of people of overall lower socio-economic status, with a potentially higher incidence of exposure to industrial hazards. Therefore, in order to avoid a recruitment bias (if controls had been chosen in another department of our public hospital), our controls were randomly sorted on electoral lists and matched for sex and age with MDS patients. We checked that this control population was well representative of the population of the North of France in terms of socio-economic distribution. Therefore, we compared the observed distribution of the controls (according to the French National socio-economic classification) with the expected distribution (with respect to the sex and age distribution of the MDS cases) using the socio-economic distribution of people in the North of France, given by the French National census of 1990. Observed and expected distributions were comparable (data not shown). Although the interviewer should have been blind as to the patient/control status, we cannot discount the fact that some of them may have revealed their status to the interviewer. However, in order to minimize exposure errors or omission, all the questionnaires were reviewed for exposure assessment and blindly coded by the expert group. Recall bias is always of concern in case–control studies. To obtain the same motivation of controls and patients, the letter of information asked for the participation to a study on health and occupational and environmental exposures. At the time of the appointment with the interviewer, they were advised that it was a research study on ‘Leukaemia’, and controls as well as cases were told of the importance of a detailed description of past exposures.
Table VII. Main previously published case–control studies of occupational and environmental factors in MDS.
|Farrow et al (1989)||Petrol and diesel (vapours)||35/63||23/63||P < 0·01|
|Petrol and diesel (liquid)||29/63||14/63||P < 0·01|
|Ammonia||10/63||1/63||P < 0·05|
|Goldberg et al (1990)||Pesticides||17/24||6/21||P = 0·002|
|Solvents||5/24||9/21||P = 0·009|
|Asbestos, fibres, welding fumes|| || ||NS|
|West et al (1995)||High occupational exposure (> 2500 h, medium intensity)|| || || |
|Radiation||27/400||12/400||OR = 2·25 (1·1–4·7)|
|Metals||66/400||37/400||OR = 1·78 (1·2–2·7)|
|Halogenated organics||26/400||12/400||OR = 2·17 (1·1–4·5)|
|Petroleum products||66/400||53/400||OR = 1·25 (0·8–1·8)|
|Plants, animals|| || ||NS|
|Non occupational exposure|| || || |
|Dental X-rays||123/400||83/400||OR = 2·85 (1·3–2·8)|
|Other X-rays, hobbies, gardening|| || ||NS|
|Smoking (ever smoked)||286/400||275/400||OR = 1·16 (0·8–1·6)|
|Ido et al (1996)||Smoking||70/110||62/116||NS|
|Alcohol||56/116||42/116||OR = 2·15 (1·1–4·2)|
|Solvents, lead, animals|| || ||NS|
|Rigolin et al (1998)||Solvents||25/178||4/178||OR = 7·1 (2·4–20·9)|
|Pesticides||48/178||27/178||OR = 2·1 (1·3–3·6)|
|Nagata et al (1999)||Hair dye use (ever used)||34/111||146/830||OR = 1·99(1·2–3·4)|
|Solvents occupational exposure||12/111||39/830||OR = 1·99(0·97–4·1)|
|Former smokers||26/111||149/830||OR = 0·79(0·4–1·6)|
|Current smokers||37/111||257/830||OR = 0·94(0·5–1·8)|
|Former alcohol drinkers||7/111||27/830||OR = 1·01(0·4–2·7)|
|Current alcohol drinkers||48/111||384/830||OR = 0·82(0·5–1·4)|
Our study does not confirm the findings of West et al (1995) of a higher risk of MDS in childless people. A higher incidence of history of blood malignancy in at least one relative was found in MDS patients compared with controls. This could suggest a genetic susceptibility to develop haematological malignancies. However, one cannot exclude that MDS patients could have paid more attention to blood diseases in family relatives than controls, explaining an underevaluation of familial blood malignancies in controls.
We found an excess of smokers and ex-smokers in MDS patients. Similar results were found in our preliminary report (Nisse et al, 1995) (OR = 1·83; P = 0·03). Pasqualetti et al (1997) found a clear relationship between smoking and MDS (OR = 2·33; P < 0·03). West et al (1995) also found an elevated but not significantly increased risk of MDS owing to smoking (OR = 2·03) in their study. As in our study, West et al (1995) did not observe smoking as a single dose–response for the risk of MDS. They found that most excess of risk was among people who smoked 1–4 cigarettes/d, whereas, in our study, most of the excess of risk occurred in people who smoked 1–19 pack-years cumulatively over life. Thus, smoking appears to be a risk factor for MDS. Tobacco smoke contains carcinogenic substances, including benzene and radioactive compounds, both of which have been shown to increase the risk of AML. Indeed, smoking may also increase the risk of AML. Brownson et al (1993)) showed a positive association between smoking and AML (relative risk = 1·4; confidence interval = 1·2–1·6) in a meta-analysis of case–control studies and suggested that 17% of AML cases could be attributed to tobacco smoking.
The leukaemogenic role of diagnostic X-rays remains controversial. We did not find any excess of exposure to X-rays in our study, while West et al (1995) showed that patients had received significantly more dental X-rays than controls. We also found no significantly higher incidence of previous radiotherapy in MDS cases, but this could have been owing to the fact that MDS occurring after radio- or chemotherapy for a prior neoplasm were excluded from the study. We cannot draw any conclusion regarding the environmental exposure to high-power voltage lines and nuclear plants, as the number of exposed people was too small. West et al (1995) suggested trends of association between non-ionizing radiation and MDS for high level exposure (OR = 1·88, 95% IC = 0·75–4·82 for high voltage, OR = 3, 95% IC = 0·75–13·94 for radio transmission). An excess of leukaemia and particularly of AML has also been reported in people professionally exposed to low frequency electromagnetic fields (Theriault et al, 1994; Kheifets et al, 1997). Ciccone et al (1993) also noted a high but not significant exposure to electromagnetic fields in their case–control study. West et al (1995) also suggested a dose–response relationship between MDS and radiation (both ionizing and non-ionizing) (OR = 2·05, 95% IC = 1·15–3·68 for medium exposure, OR = 2·25, 95% IC = 1·1–4·7 for high exposure).
Living near an industrial site appeared more frequently in MDS patients in our study. This is an original finding and it would be of interest to develop this association by studying the characteristics of production and emission of those industrial plants.
Regarding hobbies, in contrast to the results of West et al (1995), we found an association between MDS and gardening (at least once a week). This is in agreement with the relationship between MDS and agricultural work.
The odds ratio for job titles classified according to the International Standard Classification of Occupation (1990) showed an excess of risk in occupations usually associated with chemical exposure such as plant and machine operators, agricultural workers, elementary occupations, and also, unexpectedly, managers. As these large groups include sometimes very different jobs, further analysis was made with more specific jobs when the number of patients and controls was sufficient. We found a higher incidence of MDS in health professionals, textile workers, agricultural workers, machine operators and sales representatives. Health professionals may have been more often exposed to radiation, drugs, and infective or biological risks (contact with patients). West et al (1995) also reported an excess of risk in ‘professionals and related in education, welfare and health’ (OR = 1·9, 95% IC = 1·1–3·4). An excess risk of leukaemia and particularly of myeloid leukaemia has been previously reported in health occupations, for example, physicians or nurses handling anti-neoplastic drugs (Skov et al, 1992) and pathologists being exposed to formaldehyde (McLaughlin, 1994), but aetiological factors were not clearly identified. Regarding the increased risk of MDS in textile workers, the hypothesis of the exposure to pesticides may be suggested, as fibres (cotton, flax) are often treated, especially those that have been imported. The relationship between MDS and technical and commercial sales representatives is unclear. Sales representatives probably have a more intensive car use than the rest of the population and may be more exposed to exhaust gases that contain hydrocarbons and benzene oxidation products, known for their mutagenic or carcinogenic potential. Automobile-related activities such as driving a car or refuelling have been associated with significant increased levels of benzene exposure (Hoffman et al, 2000). The relationship between car use and blood disorders has previously been reported (Robinson, 1982; Wolff, 1992). Sales representatives also have more contact with the public and could thus be more exposed to viruses, for example.
Lymphoma and leukaemia were reported to be more frequent in agricultural workers (Franceschi et al, 1991; Blair et al, 1992; Pukkala & Notkola, 1997; Baris et al, 1998). Risk factors potentially associated with these diseases include microorganisms or oncogenic viruses and agricultural chemicals (fertilizers, herbicides, insecticides). We found an excess of MDS patients in agricultural workers, contrary to West et al (1995) and Rigolin et al (1998). However, in their studies, Goldberg et al (1990) and Rigolin et al (1998) reported an increased relative risk for pesticide exposure. In our study, all the exposures encountered in agricultural work (pesticides, cereal dusts, fertilizers, contact with livestock or poultry) appeared significantly related to MDS in univariate analysis. However, we could not identify the specific agricultural factors associated with MDS by multivariate analysis.
Solvents are the class of chemicals probably most often reported as having a link with leukaemia and MDS. The result of benzene exposure in leukaemia and particularly myeloid leukaemia is well documented (Aksoy, 1985; Savitz & Andrews, 1997). It is more difficult to demonstrate the responsibility of other specific solvents, as many different solvents are often used by the same person in the workplace. Therefore, solvents are usually analysed globally. We found an excess risk for the exposure to solvents and the relationship was still significant when studying quantitative exposure (excess of risk of 10% by full year of exposure). Results of Goldberg et al (1990) are consistent with this finding (Table VII). West et al (1995) showed a significant association between MDS and the exposure to halogenated hydrocarbons with an OR = 2·17, 95% CI = 1–4·5 for high exposures (i.e. more than 2500 h). Finally, the recent study of Rigolin et al (1998) confirmed this relationship between MDS and solvents for exposures > 2400 hours (OR = 2·12, 95% CI = 1·26–3·59).
In conclusion, this case–control study suggests an association between MDS and several occupations including agricultural workers, textile operators, health professionals, machine operators, and commercial and technical sales representatives. The risk factors that could be incriminated in each of those occupations could not be identified precisely. This may be because several concomitant or successive factors can play a role in leukaemogenesis, or because the evaluation of exposure was not precise enough to identify the relationship. Further studies should be developed to assess specific exposures more precisely. Further analysis should also be performed taking into account genetic polymorphisms and modalities of exposures (cumulative exposure, time latency, high and short exposure). Finally, it would also be of interest to develop a map of haematological malignancies according to the industrial background.
We thank all the haematologists who contributed to the inclusion of MDS cases: Dr B. Dupriez, Dr M. Wetterwald, Dr C. Rose, Dr M. Simon, Dr J. P. Pollet, Dr D. Sautière, Dr M. T. Caulier and also Dr R. West (University of Cardiff) who gave us the questionnaire. This work was supported by the Centre de Recherche en Santé-Travail-Ergonomie, the Ministère du travail, the Ligue contre le Cancer (Comités du Nord et du Val d'oise) and the Caisse Nationale de l'Assurance Maladie.