Professional Documents
Culture Documents
5
O International Epldemlotoglcal Association 1995 Printed in Great Britain
Gulguet M (INSERM U263, Centre de Blolnformatique, Universite Paris 7, 2 place Jussieu, 75005 Paris, France),
Aplastic anaemia (AA) is a rare but severe haem- However, no toxic effect has been shown in epi-
atological disease, often of unknown origin. In a recent demiological studies. 1415
French study by Mary et al.' the annual incidence rate The high number of cases of unknown origin, the
was 1.5 per million inhabitants, the fatality rate was diminishing effects of known toxic substances due to
34% one year after diagnosis and 74% of new cases modification of exposure levels or increase in protec-
were declared idiopathic by physicians in charge of the tion as well as debatable aetiologies justify the reassess-
patients. Medical history, therapeutic drugs and envir- ment of the risk of AA associated with occupational
onmental factors have all been implicated.2'3 Some exposures. Such studies were performed simultaneously
effects concerning the relationship between occupa- in the US 16 and in France. Our case-control study,
tional exposure and the occurrence of AA have been based on the national register of aplastic AA initiated in
well documented, e.g. benzene4"3 or radiation expos- 1984,' was conducted in 1985-1988. This study has
ure,6 but these are now disappearing as a result of either been previously analysed for drug associated risks.17
replacement of the toxic substance,7-8 increased protec- The results presented here are that part of our study
tion and control of the exposure level.9 Conversely, assessing toxic exposures.
some toxic effects are still being debated due to the lack
of consistency of published results. The relationship
between AA and pesticides remains controversial, with MATERIAL AND METHODS
many case reports suggesting their possible role.10"13 Selection of cases and controls has already been des-
cribed in detail.17 A diagnosis of AA required at least
two depressed blood cell lineages: haemoglobin
• INSERM U263, Centre de Bioinformalique, University Pari» 7,
2 place Jiusieu, 75005 Paris, France.
< 1 0 g / 1 0 0 m l and reticulocytes < 5 0 x l 0 9 / l , poly-
Reprint requests to: M Guiguet, INSERM U263, CHU St-Antoine, morphonuclears <1.5xlO 9 /l, platelets < 1 0 0 x l 0 9 / l ,
27 me Chaligny, 75012 Paris, France. and a bone marrow biopsy specimen with decreased
993
994 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
cellularity compatible with the disease and showing no whether or not the occupation had been done for at least
significant fibrosis or neoplastic infiltration. Patients one year. Each subject could have been assigned to
coming from abroad for treatment and those with con- more than one job. The toxic exposures were invest-
stitutional disease (e.g. Fanconi anaemia) were ex- igated across all jobs and coded by the highest exposure
cluded. Moreover, the study was restricted to cases level recorded. The first analysis compared the exposed
aged 18-70. During the study period, 222 cases of an subjects, whatever their level of exposure, to the non-
appropriate age were recorded in the register and, of exposed subjects. Subsequent analysis was restricted
these, 98 were included in the case-control study con- to subjects exposed at a Marge' level versus those who
cerned with occupational history. The other cases were were not exposed.
not interviewed as either no investigator was available Statistical analysis was conducted according to the
at that time in the centre or the investigator, already matching design with a variable number of controls."
involved in another case, did not have time to conduct
C 1
N+/N-' OR" 9 5 * CI N+/N- ORb 95% CI*
Specific occupations
Automechanics and service station employees
(843, 451^»9O) 1/5 0.4 (0.06-3.4) 1/1 1.0 (0.06-16.0)
Construction workers (950-959) 8/3 4.8 (1.2-18.7) 3/1 3.0 (0.3-28.8)
Painters (930-939) 1/4 0.5 (0 06-4.5) 0/0 NE
• N+ and N- are the number of discordant matched sets in which either the aplastic anaemia (AA) patient was exposed and at least one control was
not, or the AA patient was unexposed and at least one control was.
b
Odds ratio.
C
95% confidence interval.
d
Not evaluable.
between the cases and the neighbour controls for None of the other exposures investigated was
history of occupational exposure to pesticides, but a associated with AA.
borderline non-significant difference appeared when
comparing cases to hospitalized controls. There was no
evidence of an increased risk for those exposed to any DISCUSSION
of the investigated categories of solvent. In a case-control study, selection of controls is crucial,
Cases were at least twice as likely as both groups of and in the present study, two groups of controls were
controls to have been exposed to glues, a relationship investigated. The first group was composed of patients
still present after restriction to the 'large' level of ex- admitted to the same hospital as the cases. Among these
posure. When both groups of controls were combined, hospitalized controls no specific disease was over-
the association was always significant with a global OR represented,17 and this diversity prevented any potential
equal to 2.7 (95% C I : 1.5-5.0) and 2.8 (95% C I : 1.3- bias due to hospitalization for specific diseases related
6.2) for all and 'large' exposure, respectively. The same to some occupational exposures. However, a potential
trend towards an increased risk after exposure to paints referral bias could not be excluded as the hospitals have
was observed with both categories of controls, with OR specific specialities for different pathologies, implying
of 1.8 (95% C I : 1.0-3.4) and 1.8 (95% C I : 0.8-4.2) that cases and hospitalized controls may not reflect a
for all and 'large' exposure, respectively. When using population from the same area. This was confirmed
the conditional regression model for any or 'large' when looking at the number of inhabitants by their
exposure to paints and glues, the association with AA place of residence. The AA patients were more likely
remained only for glues with an OR of 2.8 (95% C I : to live in rural places (<2000 inhabitants) than the
1.3-5.8) and 3.2 (95% CI : 1.0-10.2) for any and hospitalized controls (33% versus 17%), while no dif-
'large' exposure, respectively. Amongst those exposed, ference was observed with the neighbours (37%).
the length of exposure to glues did not differentiate The second group of controls, taken from the general
between the case or control subjects, with an equal population, were personally named by the cases
median duration of 6 years. For paint exposure, the from among his/her neighbours. This geographical
median durations were also comparable between the overmatching could never be totally excluded, espe-
cases (10 years) and the controls (8 years). cially for occupational exposures. Clearly, neither of
996 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 2 Toxic substances evaluated and exposure" among the control subjects
n (%)
Pesticides
all types 24 (13) 17 (24)
insecticides 24 (13) 17 (24)
11 (6) 10 (14)
weedkillers 18 (10) 14 (19)
10
* Exposure was defined as (+) any exposure during the investigated period, or (++) a large level of exposure.
these two control groups were perfect. Nevertheless, subject during an interview, and exposures were co-
careful analysis of differences in exposure levels be- ded at a second stage by a toxicologist. This approach
tween the two groups of controls and in the association has often been used when no job-exposure matrix
with AA according to control group, emphasizes the was available. Classification of exposure into the 'low'
importance of having these two groups. The ideal con- or 'large' category was performed by a single toxic-
trol group lies somewhere in between, according to ologist using information on the job, the description
each factor studied. of the tasks done, and environmental conditions and
In this study we considered occupational exposures protection reported during the interview. Because of
which occurred during jobs done for at least one year. the relatively high proportion of controls with low
Many respondents reported short-term occupations, exposure, it is likely that the category, 'low' exposure,
13% for the cases, 15% and 13% for the hospitalized was rather wide. In any case, as the classification was
and neighbour controls, respectively. In addition, occu- performed blindly by the same toxicologist, no bias
pational exposures resulting from unofficial jobs were could be introduced in the comparison between cases
probably not reported to the interviewers. If these jobs and controls.
resulted in intense, although short-term exposures, a No relationship was found between any major group
more detailed interview would have been required to of occupations and AA. This supports the lack of sys-
detect them. tematic selection bias. Only one major occupational
Occupational exposures were assessed by a two- group, farming, was studied for its hypothetical rela-
stage process. Work history was obtained from each tionship with AA. In accordance with the difference
OCCUPATIONAL EXPOSURES AND APLASTIC ANAEMIA 997
Pesticides
all types 18/18 1.6 (0.8-3.0) 4/10 0.4 (0.1-1.3)
insecticides 18/18 1.6 (0.8-3.0) 4/10 0.4 (0.1-1.3)
11/8 2.3 (0.9-5.8) 1/5 0.2 (0.02-1.7)
weedkillers 15/15 1.7 (0.8-3.6) 4/9 0.4 (0.1-1.4)
9/8 2.0 (0.8-5.3) 2/6 0.3 (0.1-1.6)
1
Exposure was defined as (+) any exposure during the investigated penod, or (++) a large level of exposure; the reference group being people who
were not exposed.
b
N+ and N - are the number of discordant matched sets in which either the aplastic anaemia (AA) patient was exposed and at least one control was not,
or the AA patient was unexposed and at least one control was.
c
Odds ratio.
d
95% confidence level.
'Not evaluable.
in residential location, the percentage of agricultural unable to find any increasing risk among people who
workers was different between the two groups of reported farming, landscaping, gardening or pest
controls, with 6% and 17% among the hospitalized exterminating.14 A positive association between AA
controls and the neighbours, respectively. The national and exposures to pesticides has been postulated as a
percentage of agricultural workers was 8% in 1985.2I result of case reports.10"13 When exposure to pesticides
Therefore, a comparison of agricultural occupational was directly investigated, no risk was observed in the
exposure was undertaken between the AA cases and the comparison with neighbours but the diminishing
hospitalized controls and no excess of cases was numbers of pairs could indicate a possible over-
observed. A previous descriptive study of 251 French matching between cases and neighbours. Exposure to
AA cases suggested excess frequency in the rural pesticides was associated with a moderate, borderline,
population. 22 However, subsequent epidemiological non-significant risk when cases were compared to the
studies did not support this association. A cohort hospitalized patients. Whatever the categories, the OR,
study has indicated an excess of AA mortality for even if not reaching statistical significance, were
farmers, but the excess was not consistently observed increased after restriction of the exposure to the Marge'
after I960. 13 A case-control study of fatal A A was level. In the study presented here, only occupations
998 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
done after the 1970s were investigated, and, if there is suspicion to these mixtures which will need to be better
still a risk, it could currently be low. investigated in future studies.
In our study, no significant risk was detected for In conclusion, this large-scale case-control study
agents with recognized bone marrow toxicity. No asso- confirmed the vanishing role of previously demon-
ciation was found between AA and occupational radia- strated toxic substances related to the occurrence of
tion exposure. Another recent case-control study found AA. Nevertheless, findings showed a borderline non-
OR equal to 1.0 for occupational exposure to radia- significant association with pesticides and an associa-
tion. 16 The mortality trends in successive cohorts of tion with glues and/or paints was noted. The latter
radiologists indicated that the excess risk of AA dra- clearly requires further study to investigate the various
matically decreased for those entering this speciality individual compounds contained in these products.
after 1940.9 In the same way, the relationship which
was reported between airborne concentration and ben-