Relation of soft-tissue sarcoma, malignant lymphoma and colon cancer to phenoxy acids, chlorophenols and other agents.

HARDELL L. Relation of soft-tissue sarcoma, malignant lymphoma and colon cancer to phenoxy acids, chlorophenols and other agents. Scand j work environ health 7 (1981) 119-130. An association between exposure to phenoxy acids or chlorophenols and soft tissue sarcoma and malignant lymphoma has previously been reported. An association between exposure to organic solvents and malignant lymphoma has been demonstrated as well. In the present investigation the validity of the assessment of exposure to phenoxy acids and chlorophenols in the previous studies has been further analyzed, partly through a reconsideration of original data and partly through the utilization of another cancer type (colon cancer) for comparison. No observational bias was found which could distort the earlier findings. No significant association was found for these chemicals and colon cancer, whereas exposure to asbestos showed about a twofold increase in risk for colon cancer, an ocurrence in agreement with previously reported findings.

able a meanilngful evaluation of the risks assocLaJred with the exposures studied. In the fimt soft..;tissue sarcoma study (11) and in the lymphoma study (10) the cases consisted of all the individuals with a histopathologically verified diagnosis who had been admi.tted to the Department of Oncology in Umea during specified study pertods. Cases in the second soft-tissue sarcoma ,study (4) were obtained from the Gancer Regi,ster of the Swedish Social Welfar,e BoaI'd, and they were all the patients residing in a particular region of southern Sweden at 1Jhe :time of diagnosis.
Referents were selected through a matching procedu,re, ie, ,they wer,e selected individually to each oase fQr sex (males only included), age, and place of res.idence. In adcliJtion, for deceased patients, deceased referenlts with the same year of death were selected.
n each of the studies, exposure information was obtained. with written questionnaires and was supplemented by telephone interviews. With 5ucfu a method care must be taken to avoild the possibi:l.ity that the mdividuals' a,ttiJtudes and previous ex-periences can introduce bias into the assessment of exposure (1). One possibiliJty could be a tendency, COIlSC'ious or not, for those suffering from a disease to OVeTstate past exposure as compared 10 healthy referents. Eaoh study was designed to minimize such recall bias through the inclusion of a number of questions about various occupational exposures.
Each subject, or the next-of-kin of a deceased subj,e-ot, was mailed a questionnaire whkh consisted of nine pages with a vani:ety of questions concerning previous and present occupation, condit ions in the work environment, smoking habits, etc, or in total 33 different questions. Each question was subdivided into issues about special aspects of work conditions, exposure to different agents, etc, ie, about 130 questions in all. Ten of these subquestions were somehow related to the use of herbicides. This number is to be compared with, for example, 16 questions about the use of organic solvents, 4 about work with plastics, 3 about work with glues, 4 about use of drugs, and several questions about smoking habits. Supplementary interviews were carried out by individuals who did not know whether the subjects were cases or referents. Employers, neighbors, and other individuals who might have special knowledge concerning a subject's past exposure were consulted by the interviewer, if necessary, to verify and monitor the accuracy of the exposure information. Since deceased referents were selected for deceased patients, it is unlikely that there should be differences between the cases and referents because of differences in recall of exposure.
Despite all. these efforts ooncerning the design of the studies, some observational bias mighrtstill exist. The purpose of the present investiga'tion was to analyze further the validity of the assessment of exposure Ito phenoxy acids and chioro-phendLs in the previous studies, partly through a -reconsideration of the origina'l dalta in these studies and partly through the utilization of another cancer type (colon cancer) both as an index series and as a reference for the earlier case series of soft-<tissue sar,cOIIlJaJS and malignant lymphomas. As already mentioned, the interviewers were blinded in aM the studies. It has been suggested (3), however, that some cases might have revealed their 120 diseaAe during the phone call; if so, observational bi-as might have been introduced. For the evaluation of such a hypothetical possibility influencing the results in these studies, the exposure data based on /the questionnaires only were compared to rtlhe exposure data obtained from the combined informatilOn from the phone interviews and the questionnaires.

Reconsideration of earlier data and the aims of the present study
In the previous studies (4, 10) rtlhe subjects were dtvided into the occupational categories agriculture/forestry and other. Risk ratios were then calculated within these occupations for both exposed and nonexposed individuals. For the nonexposed cases the risk ratios were 1.1 and 0.9, respectively, in these studies, values which would be inconsistent with the occurrence of an observational bias in the assessment of exposure (1). Since it has been suggested that occupati:ons were reported secondariJ.y to the reoall of eX'posure (3) however, the material has now also been further analyzed by subdividing the individuals occupied in agricul-ture/forestry into two groups, those who had been occupied in the same job throughout tile whole period silnce 1950 (approximately when phenoxy acids were introduced in Sweden) and those who had been occupied for shorter periods and not continuously since 1950. This subdivision rules out the possibility that occupation was reported secondarily to Ifhe recall of exposure as far as those occupied only in agriculture/forestry are concerned.
The cases in the present new study con-siJsted of male patients with colon cancer, a disease not previously suspected to be associated with ex;posure to phenoxy acids or chlorophenols.
The debate about phenoxy acids and their presumptive ris:ks was intense in Sweden during the period when this new study \VIaS conducted. A!ccordingly, any systematic hi-as :iI!l the observation of eX'posure that might have been present in !lfue first three studies should also 'be present in this study.
Additional subjects with regard to colon cancer Cases. The cases consisted of all male patients with colon caIlJCer aged 25-85 a who were residenlts in Jthe region of the Department of OIl!cology, Umea, and who had been repoIited to the Swed~sh Cancer Register in [1978][1979]. They all had the h'i:stopathdlogical diagnosis of adenocarcinoma.
Referents. All the referents in the two earHer studiles in the Ullllea region (10,11) w,ere used as the reference entity in the colon cancer study. Consequently bOith cases aIlId referents wer,e derived from the same population source. The information previously obtained from the questionnaires and interviews of these referents were thereby 'Used and analyzed. Stratjji'cation was made for age and place of residence (rural or urban domicile). Table 1. Exposure frequencies (%) to phenoxy acids on the basis of the questionnaires in three casereferent studies: I = soft-tissue sarcoma study I (11), II = soft-tissue sarcoma study II (4), and III = malignant lymphoma study (10). Data are shown for all exposure and with exposure to phenoxy acids within a latency period of 5 a or less than 1 d excluded. Assessment of exposure. The pI'ocedure already described was used in the assessment of exposure for rthe colon cancer cases. The cases were mailed a question-naiI'e iden'ti.caJ. d:o the one previously used wti.lth a variety of questions about different wor'kpla,ces, chemIcal exposures, use of drugs, smoking habits, etc. The answers were tllen, ilf necessary, completed over the phone by an interviewer who was not aware of whether the interviewee was a colon cancer patient or another SUlbject since the questionna,ires were mixed with identical questionnaires used in a simultaneous case-referent study of nasopharyngeal carcinoma (resuLts to be published).
As in rtlhe previous studies the following criteria weI"e used in the assessment of exposme. Subjects with a total exposure of less than 1 d to phenoxy acids were considered nonexposed in the analysis. Exposure to chlorophenols or organic solvents for less than one month over the years or ,less :than one week continuously was considered low-grade; higher dura-t~ons were termed higlh-grade. Exposure to phenoxy acids, chlorophenols, or organic solvents withiJn 5 a prior to the diagnosis was not included, ie, exposure 1973 or later in ,tMs study, in the analysis o,f cases and referentlS jllilt as in the earlier studies. For other agents analyzed no such exclusion was made.

Statistical methods
The statiJstLcal analysis of the data was based on rt!he Man!bel-Haenszel procedures for 'the ca'lculartion of p-values and for the estiJrnation of overall rate ratio (16). The pdnciples for determining standardized ra,te ratios have been outlined by Miettinen (17,18), just as the method for calculating the confidence interval of the raTt:e ratio (19).

Reconsideration of earlier data
T.he arualysis of the data in the questionnaires prior to the interviews conceI'ning spraying operations with phenoxy acids is shown in table 1. The risk ram'os were ca1lrculateid without any consideration of laltency period or exposure time but also with 'exposure to phenoxy acids within the latency period of 5 a or less than 1 d excluded. In the analysis of exposure no consider,ation was given to matters dis- Table 3. Number of cases and referents exposed to phenoxy acids divided by occupation in the malignant lymphoma study (10). closed during 'the intervriews, such as some individuals answering that phenoxy acids had been used in their neighborhood without their own active participation or some individuals not mentioning what preparation had been used but referring to spraying in their own garden with nicotine, etc. This analysis was merely a validation of the questionnaire with respect to exposure to phenoxy acids. About a fourfold increase in risk was demonstrated in all three studies, a finding indicating that no major change in the results was introduced by the interviewer.
For the further elucidaJtion of a possilble observaltional bia,s, aU the subjects em-played in agricu1turelforestry in the second soft-tissue sarcoma study and I\lhe malignant lymphoma stUidy were divided into two groUJPs, '1O'1lg-(ie, since the 1950s) and short-term farm/forestry work, and the risk ratios for exposed and nonexposed individuals employed in agriculture/ forestry were calcu1arted wiJth nonexposed subjects in other occupations as the reference group (tables 2 & 3). For the nonexposed i!ndiviJduals occupied in agriculture/forestiry during the whole period since 1950, a risk ratio of 1.2 and 0.8 was obtained for the two studies, respectively, and for those employed for only some period a risk ratio of 1.0 was Table 4. Exposure frequencies (Ofo) to different agents among the cases and referents of the total sample and after exclusion of those exposed to phenoxy acids and chloropheno/s, respectively. calculalted fOT bOIth studies. If the exposure oi cases was 'COnsiderably exagge.ralted and the exposure of the referenrts highly underestimated, a value far below 1.0 would have 'been obta:ined in the ca1culation of ris'k ratios of nonexposed individuals in ag.n,culture/forestry as compared to other occupat~ons. The obtained risk ratios of aboUit 1.0 indiJcal1;e t'hal1; no sUlbstan;tJi.a!1 dis.tortil()n in the observart::itoo:l of exposure to phenoxy acids could be pres-ent in the studies. Furthermore, the rather simiJar risk ratios determined for nonexposed subjects among individuals wilth shoIit-and long-term occupa.tion in agricul!1Jure/forestry in the two studies indlcated that there could hardly be any secondary reca]l about occupation iby individuals wi1lh a shor:t time of employment [as suggested (3)] which should have disqualified the technique (1) for evaluating observaltional bias in illhe studoies. Ohloro- Table 5. Number of colon cancer cases and referents exposed to phenoxy acids and chlorophenol.
Exposed Place of residence Nonexposed Total Phenoxy acids Chlorophenols  a One control exposed both to phenoxy acids and chlorophenols is included among the subjects exposed to phenoxy acids. Table 6. Number of cases and referents exposed to asbestos. phenoxy aoids. The use of motor saws was less common among the colon cancer patienrts than among the referenrts. Exposure to asbestos was analyzed in more detail.
Asbestos. The answers regaI"ding exposure to a'Sbestos were not supplemented over the phone for more details about work procedures, exposure ti;mes, latencyperi'Ods, etc, since detailed information about such exposure w.as not ,the primary aim of the s1ludies. The one case and three referents who -did not answer the question were excluded from rtlhe analysis. AccoI"ding to table 6 the risk ratio was cal-culated to be phenol exposure is more common in other occupation's, such as woodwork, including saw mills. WiJ1fu1Jhe same technique the riosk raJtios were ailso calculated for nonexposed subjects empl,oyed. in agriculture/ for'es1Jry/woodwork With<lUt any distortiolll of 'the resutts (:tables 2 & 3).

Colon cancer study
The cases of the colon can~e-r sturdy consisted of 157 men. '.Dhree !(l.9 0J0) of them did nOit answer the questionnaire. Out of the 541 referents, 206 were derived frO'.ll1 the soft-tissue sarooma study and 335 from the malignant lymphoma study; thus all the referents from :the two studiles were used. Sixty-five (41 0J0) cases and 245 (45 0J0) referents were deceased. Exposure to various agents is presented in table 4. Of the cases and refe-rents 11.0 and lOA %, respeotJively, had been exposed to phenoxy acids or chlorophenols. Table 5 shows the risk ratios for exposure to phenoxy acids and chlorophenols. Stratification was made for age, vital status, and place of residence (urban vs rural domicile).
Other exposures. Exposure :to various other ag,ents is listed in table 4. Details about exposure to organic solvents, smoking habits, and use of antihypertensive dl'Ugs were supplemented over the phone if necessary. Use of antihypertensive drugs was somewhat more common amOlllg the colon canc,er patiellits !than among the I"srerent:s. No significant diff.erence between cases and I"eferents was shown for organic solvents, dic'hloro-diphenyl-tri-chIoro-ethane (DDT), glass fiber, and smoking. EX'posure to mercury seed dressi'ngs covaried. with exposure to

Utilizing colon cancer as another reference entity
'Dhe pI'evious studi'es of cases with softtissue sarcoma (11) or malignant lymphoma (10) wifu their residence in the region of '1Jhe Department or Oncology in Umea were evaluated using r1Jhe colon oancer patients from the same population source as another reference entity. .oNH rtilie individua·ls were thereby strattfied ac-cording to poss1ble confounders such as plac.e of residence (urban or rural domidle) and age. Tables 7 and 8 show aboUit a fivefold increase in rthe risk (p < 0.001) for soft-tissue sarcoma and maligIllalllt lymphoma for cases exposed ,to plhenoxy acids or chlomphenols, i1e, ahoU't the saJme risks as in the pub'ltshed studiJes when matched referents from the population were used.

Discussion
One pQssilble reason for misleading results in case-referent studies coruld be that cases Table 7. Distribution of soft tissue sarcoma cases in northern Sweden (11) and colon cancer referent patients according to exposure history, place of residence, and age group.  have a tendency to pay more aJbt~mtion to questions about .previous chemical exposure 'tnan their referents or that the interviewer shows greater interest in the cases than 'the referents. The results obtained from a recons1deraJtion of earlier studi'ffi show that 'there was no substantial difference in the exposure frequenClies as obtained 'through questionnaires only or from questionnaires with a supplementary telephone interview. The value of the supplementary interview mainly seems to be that 00 removing misconceptions about questions and obtaiIDng more detailed information on certain aspects. Thus rohe data on expoisuTe was improved, but bias was hardly introduced, at least if one can judge from rthe present evaluation. A bias in reportirngexposure to phenoxy acids halS heen 'suggested by Cole (3) as the explamattion of rtlhe results of these studies. lit has also been suggested that people, especially ca'ses, might not always directly recall jobs in agriculture/forestry, but first remember exposure :to phenoxy acids and then, conditionally on that recall, also Table 8. Distribution of malignant lymphoma patients (10) and colon cancer referent patients according to exposure history, place of residence, and age group. Five patients with malignant lymphoma exposed both to phenoxy acids and chlorophenols included among individuals exposed to phenoxy acids.  (3); ie, there seems 'to be no longer any reason to believe in any observ.altional bias in fuese studies and Ithere is obviously not any rationale for such a 'belief. Risk ratiors wer,e also calculatted with the colon cancer pa'u-ents as referents. About the same increase was found in the risk for soft~ti'S:SIUe sarcoma and malignant ly:mphoma for cases exposed to phenoxy adds or 'chlOTop'henols as pr,eviously report,ed. This r,esuilt again indic-alted ,fuaJt no sUbsrtJantial observational bias could exist in 1lhe S'hIdies, since individuals sutffering from colon cancer should be expected to reca!lil exposure simitlarly as o1Jhers suffering fmm cancer and the deoba,te about possihle hea~tJh hazalds from phenoxy adds was perhaps even more vigorous at the ,time of the colon cancer study than it waLs during the period of the earlier studies. This siltu,ation is noteworthy since i,t has been suggested (3) that the great interest in the adverse effects of phenoxy adds might have influenced the recall of exposure of the case individuals in comparioon to the rererents.
The colon cancer study was not performed during quite ,the same period as the other studies. With the same reference entity this difference might sHghtly bias the estimate of risk ratios in the colon cancer study although not with respect to deceased sulbjects. All the illlvestigations wer,e performed during 1978-1980.
The etiology of colon cancer is not well known. Environmental factors are believed to be important. A positive correlation has been reported between the incidence of colon cancer and general nutritional pattern, fat and protein consumption, arteriosclerotic heart disease, and economic development (7). Support for the assumption that environmental factors are important is given by migrant studies. Populations immigrating from a low incidence area to a high inciJdence area assume the cancer ,risk of the ;latter (6,25). Regarding diet, African populaitions wil11h a high daily intake of fiber and a low inta.ke of refined carbohydrates, as compared to the Western diet, have a lower :incidence of colon cancer than in Western 'countries (26). BHe acids a,nd their degradation products secreted in response to ingested fat are believed to promote eancer (2,13). Furth,ermore diet determines, in the intestine, the composition of the microbial flora which might produce carcinogenic or cocarcinogenic compounds from food or gastrointestinal secretions (5,12). Chronic long-standing inflammatory disease such as ulcerative colitis (20), granulomatous colitis (15), and adenomatous polyps in the colon (24) may all sometimes be precursors of colon carcinoma.
The main purpose of this study was to evaluate a possible relationship between colon cancer and exposure to phenoxy acids or ohlorophenols and to check validioty aspects of earlier studies. No signi:fikant association between this cancer form and exposure to phenoxy acids or chlorophendls was shown. On the oiher hand Oliver et al (21) reported, among other diseases, an overall excess of deaths fr,om gastrointestinal cancer among patients treated wilth the phenoxy acid derivate clofibrate as 'compared to the two l"ef,erent groups. The inquiries in this colon cancer inv,estigation also included a question about the use of doHbra'te. Its use was reported by two cases and It.hree refer,ents, aJJ 01f whom, !however, had been exposed within 5 a prior to the diagnosis.
Among asbeStos workers an ex<Je'Ss of lung cancer, pleural and per~t.oneal mesothelioma, andoanoer of the stomach, colon and rectum has been reported (22,23). For colon and r,ectum cancer the reported risk ratiJos were 2.1 and 2.9, respect;ively, in two cohorts (23). These results are in accordance witih 1Jhe f~ndi:ngs of the present study. In the pr,evious studies O'f soft-tissue sarcoma and malignant lymphoma no differenoe in exposure to asbestos wa:s seen between cases and referents in spiJte of an intense deba'te and information about asbestlos as a cause of cancer.
Regarding exposure to Vlarious other agents in :table 4, no djlfference was seen between colon cancer cases and referents. Colon cancer is less common among physically adive individuals (14), which is the situation for lumlberjacks using motor saws. This can explain the fact tihat motorsaw use was less common among the colon cancer cases ilian among the referents. The use of antihypertensive drugs was somewhat more common among :the colon canoer patients than among l!he referents.
In suunmary these analyses showed that 1Jhe previously reported assoda'tions between exposure to phenoxy acids or chlorophoo'Ols and soft. . . .ti,ssue sarcoma and malignant~ymphoma cannot to any essential degree be explaim.ed by observabona'l bi,as in the studies. Regardring colon cancer, no significant association was found for these dhemicals, even though phenoxy acids and chlorophenols might be of some marginal importance for the eUology of colon cancer. Exposure 00 asbestos was eX'Cessive among the colon cancer patients as compared to the referents, a finding in agreement with previously reported results.