herbicides and soft-tissue sarcomas in female rice weeders. A population-based case-referent study.

Phenoxy herbicides and soft-tissue sarcomas in female rice weed ers: A population-based case-referent study. Scand J Work Environ Health 13(1987) 9-17. A population based case-referent study was conducted in an area of northern Italy where rice growing is the predomi nant agricultural activity and phenoxy herbicides have been used since 1950. Manual rice weeding was formerly performed by a seasonal female working population; in the early 1950s these women were con currently exposed to chemical herbicides. Sixty-eight persons representing incident and histologically re vised cases (31 women) and 158 population referents (73 women) were interviewed. The cases were histo logically confirmed independently by two blinded path ologists, and exposure to phenoxy herbicides was assessed by two blinded pesticide researchers. An age-adjusted odds ratio 01'0_91was found for the living men (with suspect exposures; no man diagnosed as a case had been exposed with certainty to phenoxy herbicides). Among the living women the relative risk was 2.7 (90 "70 confidence interval 0.59- 12.37), and it further increased when attention was restricted to women exposed in the whole 1950- 1955 period and to younger age groups.

Reprint reque sts to : Dr P Vineis, Servizio di Ep idemiologia dei Tu mori, lstituto di An ato mia Pa tologica, Via Santena 7, 1-10126 Torino, Italy. have not been sufficiently informative (10,2 1, 24, 26). (See also th e reviews in refe rences I an d 6). T he Internati onal Agency for Resear ch on Cancer, chiefly on the basis of reports by Eriksso n et al (8) an d Hardell & Sandst rom (14), evaluated the evide nce on the carcino genicity of ph eno xy acid herbi cides to human s as limited ( 15).
Most of the afor ementioned authors emphasized the need fo r fu rth er knowledge. Therefore we decided to und ertake an investigation in the provinces of Vercel-Ii, Novara, and Alessandria (northern Italy), where rice grow ing is the predominant agricultural activity. The total population in this area is approximately 1.5 million . Phenoxy herbicides were introduced in large amounts in th e earl y 1950s and have been extensively used up to now. Ac cording to official sta tistics (16,17), more than 4.8 t of 2,4-0 and 21.7 t of MCP A were used in the three provinces in 1959 (data not available for 2,4 ,5-T). Th e corresponding figur es for 1969 were 24.5 t for 2,4-0, 29.4 t for MCP A and 125.1 t fo r 2,4 ,5-T . 2,4,5-T was officially banned in 1970. Before the introductio n of chemical herbicides, a female working popu lation was seasonally engaged in rice weeding. For a period around 1950-1955 rice weeding was still done man ua lly although chemical weeding was being experimented with. It is therefore reasonable to assume that the rice weeders working during this period were exposed to phenoxy herb icides, ma inly throug h skin contac t.

Case identification and ascertainment
New cases of soft-tissue sarcoma with a pro ved or suspected histological diagnosis in 1981-1983 (age of per-son~20 years) were identified through all of the pathology departments of the three provinces (hospitals of Vercelli, Novara, Alessandria, Borgomanero, Biella, Casale Monferrato). The search was extended to six pathology departments in the city of Torino and to the National Cancer Institute in Milan, where residents in the three provinces could be referred for treatment. Overall , 135 potential cases were identified (116 in hospitals located in the three provinces and 19 in Torino or Milan). Visceral sarcomas were excluded.
Histological specimens (with the exception of the cases diagnosed in Torino and Milan and four other cases) were independently reviewed by two pathologists (AD, RP), blindly with respect to the exposure status and to previous diagnosis. The classi fication of the World Health Organization (WHO) was adopted for the soft-tissue sarcomas (7).
Of the 135 identified cases, 18 were excluded because the person's town of residence was unknown or the person was not traced after five attempts (table I) . After the histological review, 26 cases were discarded. For five others the two pathologists agreed on the diagnosis of so ft-tissue sarcoma but did not agree on the histological type. After a check of the clinical records, four cases turned out to be visceral sarcomas and were excluded. For nine histologically confirmed cases, we met with refusal to participate, and for two in which the subject was deceased no relative could be traced . Of the remaining 75 cases, 68 were incident (ie, newly diagnosed in 1981 -1983), and seven turned out to be prevalent (ie, diagnosed prior to 1981) during a check of the clinical records . The seven prevalent cases were not included in the analyses. Of the 68 subjects remaining, 44 were living and 24 were deceased. The age-standardized annual incidence rate (including interviewed and noninterviewed subjects and excluding prevalent and visceral cases) was 2.8/100 000 [95 0,70 confidence limits (CL) 1.3-4.2] for the men and 2.0 (95 % CL 0.8-3.1) for the women (age2 0 years). These rates do not take into account the histological revision in order to be comparable to the rates reported in 1976 -1977 by the Cancer Registry of the Province of Varese (an area close to the study area). The latter rates (per 100 000, age~20 years) were 2.3 (95 % CL 0.4 -4.1) for the men and 1.5 (95 % CL 0.06 -2.9) for the women (23). (All the rates have been age-standardized on the basis of the world population.)

Living rejferents
A random sample of eligible referents (30 living men and 30 living women aged~20 years from each province) with a distribution, by municipality, representative of the population of each province was drawn from electoral rosters . The overall number of referents was approximately twice the expected number of eligible cases . In fact, a total of 168 individuals from 65 municipalities were identified versus 180 expected. (Four electoral offices did not respond at all, and a few offices provided a lower than required number of individuals, for a total of 12 losses.)

Deceased referents
Thirty-seven of the 135 persons originally enrolled as cases were deceased at the time of the interview. Therefore, for each of these 37 persons, two deceased individuals, to be matched to the deceased subject by age (± 4 years), sex, and year of death, were identified as referents from the same mun icipality as the deceased subject through the demographic offices of the municipalities. There was no selection of the cause of death, with the exception of the exclusion of one suicide (for ethical reasons) .
Seventeen deceased subjects with soft-tissue sarcoma had two matched referents. Only one referent met the matching criteria for six cases . No individual with the corresponding requirements was found for five cases. The municipalities did not respond for eight referents. And, finally, no referent was sought for five persons with soft-tissue sarcoma because they were enrolled in the National Cancer Institute in Milan after the scheduled end of the study. A total of 40 deceased referents were therefore known for interview purposes, and in fact the relatives of 36 were interviewed . a Including ind ividuals not available for interview due to illness, deafness or mental disorders. b Relatives of deceased ind ividuals were not available. C Moved to other provinces; in strict sense not eligible. d The four visceral and seven prevalent sarcomas were identified as such after a check of t he c linical records afte r the cases had been enrolled . (See the te xt .)

Interviews
Each enrolled subject was given a code number unrelated to the nature of their case-referent status. All subsequent procedures until the fina l stage of the analysiswere blind . Each living subject was mailed an invi-tation to be inter viewed, with no mention of the precise purpose of the study. Whenever possible, the invitation was followed by telephone calls in an attempt to schedule an interview date . Otherwise, additional letters inviting the individual to suggest a possible date for the interview were sent with a prepaid envelope inside. For each individual , at least five attempts (either by telephone or by a home visit by the interviewer)were made before the individual was considered untraceable . For deceased persons the next-of-kin was identified, and the same procedure was followed. A simplified postal questionnaire was sent to subjects who, for any reason, refused a direct interview but agreed to provide information by mail. Overall, information on 16cases and 37 referents was obtained in this way. The postal questionnaire was suitable for an exposure assessment identical to that provided by the direct interv iews.
All the interviews were carried out by a trained interviewer (LM). The questionnaires collected (i) demographic information, (ii) a life-long smoking history , (iii) a life-long occupational history (including job titles , addresses of the plants or farms, and types of products or crops), (iv) a history of radiological therapeutic treatment, and (v) a history of residence changes, including a specific question on periods spent in rice-growing areas. In addition the questionnaire included a list of job titles potentially involving exposure to phenoxy herbicides or chlorophenols (including jobs in sawmills, the paper and leather industries, highway and railway maintenance, and flower and market gardening). Two addi tional sheets were filled out for those who had worked in agriculture, one for rice growing and one for other crops. In these sheets information was collected on the involvement in jobs entailing the handling, transportation, and distribution of herbicides; herbicide spraying procedures; other specific jobs involving (or not) herb icide exposure (ie, weeding and water regulation in rice growing); knowledge of exposure to herbicides (and more specifically to phenoxy acids or chlorophenols) or insecticides. This part of the questionnaire was prepared in cooperation with pesticide researchers.
Demographic information (including current occupation and educationallevei) for nonrespondents was collected from demographic offices of the municipalities of residence.

Assessment of exposure
Two experts with experience in chemical aspects of agriculture (EZ , AC) were asked to assess the exposur e to phenox y acids for each job recorded in the interviews. This assessment was done independently by the two and in a blind manner. The following criteria were used for reading and interpreting the questionnaires: category I = not exposed, category 2 = exposure could not be ruled out, and category 3 = certainlyexposed . Attempts to distinguish between skin contact and other paths of exposure and between low and high levels of exposure proved not to be feasible. Rice weeders were considered to be exposed to phenoxy herbicides when they worked after 1950 and did not work exclusively in a small rice allotment of their own. Category 2 was used particularly for people engaged after 1950in corn, wheat and pasture growing. As a first step, a separate exposure assessment was performed by each of the two experts . Subsequently, a joint assessment was made , in which they reached a common evaluation for each recorded job . The latter assessment was used in the analyses.
Herbicide spraying in the area has been chiefly done by tractor.

Coding of occupational histories
In addition to the aforementioned exposure assessment, all occupational histories were coded according to the classification for job titles of the International Labour Organization (4) and the United Nations classification for economic activities of plants or farms (22). All coding work was done by a single person (GC) in a blind manner.

Analyses
The analyses were performed with an HP-61 calculator using programs for epidemiology (25). Ageadjusted relative risks were computed according to Mantel-Haenszel's estimator (19); 90 % confidence intervals were obtained with Miettinen's approximate procedure (20) (the lower limit of each interval corresponding to a 95 % one -tailed confidence limit). In addition to age (in four groups), therapeutic X rays and smoking were included as potential confounders. The X-ray exposure represents a known risk factor for soft-tissue sarcoma (3), but there is no evidence of association with herbicide exposure. Smoking is not known to be a risk factor for this kind of tumor, but it was conservatively included as a potential confounder. Table 2 reports the educational levels represented by the persons with soft-tissue sarcoma broken down into respondents and nonrespondents. The presented educationallevels are those of the persons scheduled for interview (ie, the next-of-kin of deceased subjects). Responding men showed an unbalanced distribution for the cases and referents in the different educational categories « 6 years: 26 cases and 37 referents,~6 years: IO cases and 43 referents; unknown: I case and 5 referents); the distribution of the women was more balanced « 6 years: 17 cases and 42 referents;~6 years: 12 cases and 30 referents; unknown: 2 cases and I referent). Table 2. Educational levels (years of schooling) for the responding and nonresponding living subj ects or the next -of-kin of deceased sub jects. Nonresponding individuals include refusals, untraced individuals, individuals whose relatives were not identified , and individuals who had changed residence.

Living subjects
Next-of-kin of deceased subjects  Table 3. Distribution of the living sub jects by age group and exposure level (1 = not exposed , 2 = exp osu re could not be ruled out , 3 = exposed) . Exposure refers to any calendar period. The number of subjects for whom information came from postal questionnai res is presen ted in parenth ese s.  Table 4. Distribution of the living female subjects by age group and exposure level (1 = not exposed, 2 = exposure could not be ruled out , 3 = exposed). Only women exposed cont inu ousl y during 1950-1955 have been included. Table 3 reports the absolute numbers of living subject s with soft-tissue sarcoma and their referents by expo-Among the men 26 (30.5010) of 85 referents and 15 (40.5 %) of 37 subjects with soft-tissue sarcoma had worked in agriculture during at least some period of their worklife . The women with such occupations were 21 (28.5 %) of 73 referents and 14 (45.5 0J0 ) of 31 women with soft-tissue sarcoma. Two male referents and no men with soft-tissue sarcoma were assigned a job title code of rice grower. The respective numbers for the women were three and three (crude odds ratio 2.5). This code was the only job title code mentioning rice growing, and it was assigned when rice growing was the only activity performed in a given work period . Table 5 reports the age and exposure distribution for the deceased subjects. There were 24 next-of-kin of persons with soft-tissue sarcoma interviewed and 36 nextof-kin of referents. Three women and no men with soft-tissue sarcoma belonged to expo sur e category 3, and one woman and one man with soft-tissue sar coma were classified into category 2. The odds ratio (ageadjusted in two age groups) for the women in exposure categories 2 and 3 was 1.05 (90 % CL 0.21-5.1). When matched pairs or triplets were considered, there sure levels and age groups fo r the two sexes. No men with soft-tissue sa rco ma were definitely exposed, and expo sure could not be ruled out for onl y two, the ageadjusted odds ratio being 0.91 (90 % CL 0.21-3.91) .

Deceased subjects
Among the alive women , four subjects with softtissue sarcoma and five of their referents were certainly exposed, and one and two, respectivel y, belonged to exposure category 2. This distribution corresponds to an odd s ratio of 2.42 (90 % CL 0.56 -10.34) when exposure categories 2 and 3 are combined and an odds ratio of 2.70 (90 0J0 CL 0.59-12.37) for cate gor y 3 alone. The odd s ratios were age-adjusted in the four age groups. Table 4 shows the distribution of the living female subjects who were engaged in relevant jobs in the whole 1950-1955 period. Three with soft-tissue sarcoma and one referent met these criteria, all from exposure categor y 3. Unexposed were 16 women with soft-tissue sarcoma and 53 female referents, as shown in table 3.
a Cate gori es 2 (exp osure could not be ruled out) and 3 (exposed) have been co mbined und er th is head ing .
were only two: On e was a tri plet in which the su bject with so ft-tissue sarcoma and the two referents had been definitely expo sed to phenoxy herbicides (catego ry 3), a nd the seco nd was a triplet in which one of the two referents (but not the sarcoma subject) belonged to category 3. Therefo re a matched an al ysis of decea sed female subjects was uninformative. When the to wns of reside nce of the deceased referents were compared with tho se of the living referent s, important differences appeared . Six of the 23 deceased male referen ts lived in four towns of rice-growing areas versus 3 of 62 living referents. Among the women, 6 o f 13 deceased referen ts lived in two to wns with the same cha racteristics, versus 2 o f 60 living referent s. Thi s compari son stro ngly suggests that overm atch ing of deceased subje cts was introduced into the study by th e inclusion of residence as a matching fact or.

Conf ounding
Th erapeutic X rays and smo king were co nsidere d as potential confounders . Overall , onl y two men and one woman with so ft-tissue sarcoma, and on e male and no female refer ent s had been expos ed to th erapeuti c radiation befor e 1981. Confounding by thi s var iable, therefore, can be ruled ou t.
As for smoking, 24 (65 070) of 37 men and 3 (11 % ) of 28 women with soft-tissue sa rcoma and 54 (65 It /o) of 83 male refe rents and 2! (40 It /o) of 52 female referent s were smo kers or exsmokers. (For two male re ferents information on smoking was not availabl e.) Th ese find ings ind icat e a clear deficit of smok ers amo ng the women with soft-tiss ue sarco ma. Such a pattern can be int erpreted as a con seque nce of the greater proportion of agri cultural workers among the person s with so ft -tiss ue sarcoma . Table 6 repo rts the distribution o f the cases by histological type according to the WHO classification. Table  7 gives the histological type and site of the tum ors for th e exposed cases. Among th e exposed women with so ft -tiss ue sa rco ma (cat egories 2 a nd 3 of expos ure), there were two cho ndros arco mas , two Kap osi sarcomas, two mal ign ant fibrous histio cytomas, one cho r-  doma , and one leiom yosarcoma. (One case was uncert ain with respect to histolo gical type .) No gross differences were noti ced bet ween the expo sed and une xpo sed women with respect to the sites of the tumors. Their locat ions were as follows: head and neck (0 expos ed/3 unexposed) , retroperitoneal (I exposedl2 une xposed), upper limb s (I exposed/4 unexposed), lower limb s (3 expos ed/7 unexposed), thorax (1 expo sed/3 unexpos ed ), abdomen (I exposed/ ! un expo sed) , pelvis (2 expo sed /! unexpos ed) , site not specified (0 exposed/ l unexposed) . Table 7. Sex, age, exposure category, job tille, years of exposure, and histology and site of tumor (if a case) for the exposed subjects with soft-tissue sarcoma (cases) and the exposed referents.  Living referent  Male  30  2  Wheat, corn and  1967-1984  pasture grower  Living referent  Male  47  2  Corn and pasture  1953-1957  grower  1970-1982  Living  Male  63  2  Wheat and corn  1959-1984  referent  grower b  Deceased referent  Male  77  3 Farm laborer 1950-1968 in rice growing a 2 = exposure could not be ruled out; 3 = exposed. b Also engaged in railway maintenance . Table 7 reports the ages at the time of the interview, vital status, job titles, years of start and cessation of exposure, and the level of exposure (as assessed by experts) for the exposed subjects, in addition to the histological types and tumor sites of those with soft-tissue sarcoma.

Exposure of persons with prevalent cases of soft-tissue sarcoma and those with visceral sarcomas
Of the five women interviewed who were considered to have prevalent cases of soft-tissue sarcoma, four were unexposed and one was definitely exposed, and they belonged to the 55-to 64-year age group. Two persons with visceral sarcoma were interviewed; they were both unexposed.

Discussion
Testing the hypothesis of an association between exposure to phenoxy acids in agricultural practice and soft-tissue sarcomas required a study to be carried out (i) on a population basis and (ii) in an area where phenoxy acids have been widely used. The present study meets both these requirements.
The most important exposures took place in the early 1950s, due to the concurrent presence of manual weeding and chemical spraying. Indeed, an increased risk of soft-tissue sarcoma was shown among the female rice weeders who had been employed in the early 1950s. Exposure among the men was negligible.
Although the present study is limited by a low power, on qualitative grounds our findings are consistent with both those obtained by Hardell & Sandstrom (14) and Eriksson et al (8) in similar populationbased studies and with those of reports on small series of cases among workers exposed to chlorophenols and/or their derivatives in the chemical industry (9,18).
Studies failing to suggest an association either (i) selected referents from cancer patient files (2) or adopted an approach involving proportional mortality ratios, which implied some loss in sensitivity (10,21), or (ii) used a crude categorization of exposures, in terms of occupation at the time of diagnosis (2) or at the time of death (21). A misclassification of exposure may lead to an underestimation of the relative risk, if not a systematic bias (28).
Intensity and duration of exposure should also be kept in mind when different studies are compared, although no clear picture of the levels of exposure in different jobs and different countries is available.
The statistical power of the present study is low since we had planned a sample size suitable for increases in risk of the same order of magnitude of those reported by Swedish authors (8,14). This outcome is reflected by the large confidence intervals around the relative risks. The one-tailed lower confidence limit was, in most instances, lower than one. Indeed, the target population (female rice weeders affected by soft-tissue sarcomas) was inherently limited in size.
Major biases were prevented. The similarity of the incidence rates of soft-tissue sarcoma in the study area and in the near province of Varese suggests that case ascertainment was likely to be complete. The analysis was restricted to histologically revised incident cases. Population referents were selected from electoral rosters. Respondence was reasonable. (Nonrespondents were 9 of 80 for the confirmed incident cases and 26 of 208 referents.) Selection bias is therefore a remote explanation of the findings. As table 8 suggests, almost all of the exposed living females were in fact in the lowest educational level (relative risk in the category < 6 = 3.86, 90 070 CL 0.87 -17.15). The interviewer was thoroughly trained, all questionnaires were coded blindly with respect to caseness, and exposures were blindly assessed by expert pesticide researchers.
However, the relative risks may have been underestimated because of the following three types of problems:  2. Elderly women (~75 years) were less likely to be engaged in highly-exposing jobs in the 1950s, since they were already 45 years of age or older in that period. Their inclusion in the study may have introduced an underestimation of the risks.
3. The procedure of matching deceased referents to the deceased persons with soft-tissue sarcoma with respect to town of residence may have artifactually raised the proportion of exposed referents. The proportion of definitely exposed women was similar among the deceased (3 of 10) and living (4 of 21) women with soft-tissue sarcoma but quite different among the deceased (5 of 13) and living (5 of 60) referents. The choice of a group of living referents, extracted from the general population, for comparison with the deceased subjects inducted as cases is considered to be the best option in case-referent studies; the only flaw is the quality of the interview, obtained from the nextof-kin of deceased subjects (5,11). If we compare all of the female subjects with soft-tissue sarcoma (living + deceased) with the living referents only, the ageadjusted odds ratio is 2.71 (90070 CL 0.51-14.17). We intend, in any case, to select another reference group of deceased individuals, unmatched to the subjects with soft-tissue sarcoma in respect to town of residence.
In conclusion, our study showed an odds ratio of 2.7 for living women who were exposed to phenoxy herbicides in any period of their life. When we restricted our attention to living women less than 75 years of age and exposed in the 1950-1955 period, an ageadjusted odds ratio of 15.5 (90 % CL 1.3 -180.3) was determined (table 4). When exposure to phenoxy herbicides among living females who had a regular job in agriculture (ie, excluding unexposed nonagricultural workers) was analyzed, the age-adjusted odds ratio (Mantel-Haenszel) for exposure category 3 was 3.0.
The number of definitely exposed subjects with softtissue sarcoma and referents was negligible among the men. There is no specialized job such as pesticide ap- 16 plicator in rice growing in Piedmont, and annual herbicide spraying is performed, among numerous other jobs, by unspecialized rice growers or farm laborers.
No dose-response relationship could be considered because we were not able to distinguish between high and low levels of exposure in the exposure assessment (with the exception of the a priori hypothesis concerning the 1950-1955 period). Most of the exposed subjects (both those with soft-tissue sarcoma and the referents) had been employed in rice weeding for more than 10 years (table 7).