A mortality study of vinyl chloride monomer workers

PG. A mortality study of vinyl chloride monomer workers employed in the United Kingdom in 1940-1974. Scand J Work Environ Health 14 (1988) 153-160. The mortality experience of 5 498 male workers employed for at least one year during 1940-1974 in the vinyl chloride industry of the United Kingdom was followed through to 31 December 1984. There was a significant ex cess of nonsecondary liver tumors with II deaths, of which seven were angiosarcomas. All the angiosar coma deaths occurred in autoclave workers with a median latency of 25 years from date of first exposure. A strong health y worker effect was seen. Other than that for liver cancer, no increased incidence of cancer deaths attributable to vinyl chloride monomer exposure was found. There was no evidence of increased mortality from chronic liver disease. The incidence of death from respiratory disease was low and was not affected by polyvinyl chloride dust exposure.

Reprint requests to: Dr RD Jones , Epidemiology and Medical Statistics U nit, Magdalen Hou se, Sta nley Precinct, Bootle, Merseyside , Great Britain L20 3QZ . this occupation in the United Kingdom und er the Industrial Injury Provisions o f the Social Security Act of 1975 (9). The role of chronic liver disease in the mortality pattern of these workers is not so clear (22).
Pos sible health effects of PVC dust have been the subject of a number of studies (2,15,16,21,26), although some doubt remain s as to whether the recorded changes in lun g function and radiographic app earance are indi cativ e of any increa se in respiratory disease.
Thi s paper describes a mortality an alysis of British VCM workers th at ha s primarily been based on the population rep orted on by Fox & Collier (II) . Ho wever, different criteria redu cing the tot al cohort size, have been used for th e entry of workers into the survey. The expo sure categories have been based upon the more-det ailed occupatio nal information that is now availabl e, and data on an extra IO years of follo w-up have been added .

Study population
A nominal roll wa s compiled from the personnel records of nine chemical plant s manufacturing or polymeriz ing vinyl chloride. A prerequisite for entry into the cohort was employment for at least one year in a job or jobs that invol ved potential expo sur e to VCM fo r at least 25 % of the wo rkweek . Only ind ividuals first employed in the period 1940-1974 were included. Per son s were ent ered int o the study at the time of completion of one year's employment in such an occupation. Identi fication details for each individual were sent to the Office of Population Censuses and Surveys, which flagged the record s of the individual on the Na-

Occupational exposure
A list of 12job titles was drawn up in conjunction with advice from the plants concerned, and each worker's employment history was recorded in accordance with this classification. Profiles of exposure to VCM and PVC dust were constructed for job titles from data supplied by company staff as part of an independent exercise prior to the analysis of any mortality data. Reliable VCM exposure data were available for the more recent years, but for earlier periods estimates of exposure were produced with the use of any existing historical hygiene data and the recollection of employees, interpreted in the light of present knowledge, of the irritant and narcotic effects of VCM and their relationship to exposure levels. VCM exposure profiles were plotted and used for the determination of the exposure groups. The profile produced used much of the data previously presented by Barnes (1). The profile supported a further classification of the 12 job titles into autoclave workers (group A), baggers and driers (group B), craftsmen (group C), and other workers (group D).
Information on PVC dust levels was more sparse and only related to more recent exposures, but mea-154 surements showed the greatest exposures to have occurred among the baggers and driers (group B). Table 1 shows the VCM exposure estimates for the occupational groups. These figures are not dissimilar to those reported by Ott et a1 (23). Because there was some movement between jobs, with a general tendency for workers to move from groups A and B to group D Gob movement among the craftsmen in group C was much more limited) and because this movement tended to be from areas of high exposure to areas of low exposure, mortality by occupational exposure was analyzed from the population categorized as ever group A, ever group B never group A, ever group C never group A or B, and always group D.
Since 1974exposure to VCM in the industry has been much more rigidly controlled. The levels fell almost immediately to less than 50 ppm, and even greater control has been achieved since. The present control limit for VCM is 3 ppm annually with a maximum timeweighted average per work shift of 7 ppm. This level is many orders of magnitude less than the levels that existed for much of the study period, and therefore the present analysis was limited to persons who had worked in the industry prior to 1975. All workers in the study have thus had the opportunity for at least 10 years of follow-up.

Mortality analysis
The use of lCD-categorized data is not ideal for studying rare tumors such as angiosarcoma, which would normally be expected to make only a tiny contribution to any particular ICD category. Even when all primary liver tumors are considered together, they account for less than 0.25 070 of deaths. Because the primary nature of the tumor is often not specifically mentioned on the death certificate, many of these deaths are not coded as primary liver cancer. In order to include these deaths, we found it necessary to include liver cancer deaths that were "not specified whether primary or secondary." By the aggregation of deaths coded to ICD categories 155.0 and 197.8 (eighth revision) and categories 155.0 and 155.2 (ninth revision), cases of liver cancer not specifically certified as metastatic were considered. Aggregated deaths from these causes have been referred to as "nonsecondary liver cancer." Noncirrhotic fibrosis of the liver is not specifically mentioned in any ICD revision. It would presumably be coded to liver disease (lCD 570-573, eighth and ninth revisions). Within lCD categories 570-573 there are differences of classification between the eighth and ninth revisions, but it is very likely that noncirrhotic portal fibrosis would have been coded to 573 in the eighth revision and 571.9 or 573 in the ninth revision. The five certificates coded to an underlying cause between 570 and,573 were extracted and scrutinized. The underlying caus~of death in all of these had been coded to 571. An expected figure was therefore calculated for ICD 571, whic h accounts , in the genera l population , fo r the vast majority of deaths from liver d isease by either ICD revision.
T he mortali ty analysis ha s been carried out with the OCMAP com puter program (18). The sta ndardized mortal ity ratio s have been calcula ted in th e normal manner, and the 95 010 con fiden ce inter vals for these ra tio s were calculated under the assumption that the ob ser ved number o f death s follo wed a Poisson distribution. The resu lts are sta ted as significa nt when the sta ndardized mortality rat io (SMR) 100 lies outside the 95 % co nfid ence interval (95 % C I).

Analysis of cancer by occupation
Ca ncer mortalit y by o ccup at ion is shown in ta ble 3 fo r mali gnancies which have been reported in excess in at least one populati on of VC M wor kers. Th ere was an ob vious excess of no nseco nda ry liver ca nce rs among autoclave workers, but no statistically significant excesses of ot her can cers .
From tab le 1 it is clear that th e grea test expos ure to VCM was experien ced by the autoclave wor kers,  a The workers we re cla ssi fied by the j ob held wi th the high est expos ure to vinyl ch lor ide monomer, ie, ever A, ever B, never A, etc, as des c ribed in the tex t. • P <0.5. Table 4. Observed (0) and expected (E) mortality for nonsecondary liver cancers by latency and cumulative exposure among autoclave workers (ie, workers exposed only as autoclave operators).

Time since
Length of exposure first starting 51 year 2-4 years 5-9 years <:10years craftsmen had had substantial exposure, but baggers, driers and "other" workers had had much lower exposures. The group of men who had ever worked on the autoclaves (highest exposuregroup A) had a statistically significant excess of liver tumors coded as "tumors of the liver and intrahepatic ducts not specified as secondary tumors" (07, E 0.38, SMR 1 842, 95 070 CI 741-3 795). The same group did not show any statistically significant excess for any other malignant disease. Table 4 shows an analysis of observed and expected deaths from nonsecondary liver cancers in autoclave workers by length of exposure and latency. Four of the seven decedents were men with more than 10 years of cumulative exposure at a time more than 20 years after their first exposure to vinyl chloride. All seven deaths were due to angiosarcoma. The latency period between time of first exposure to VCM and death ranged from 8 to 33 years with a median latency of 25 years. The analysis of other malignant diseases for which two or more deaths occurred revealed an excess of urinogenital cancer among workers with a low level of exposure (group D) (0 23, E 14.7, SMR 156, 95 % CI 99-235), largely due to excess bladder and prostate cancer (019, E 11.7, SMR 163, 95 % CI 98-254). The excess of bladder cancer that was seen in the population overall was confined to three associated factories at which a variety of chemicals was made (0 10 Vinyl chloride exposure and nonmalignant disease Nonmalignant liver disease (lCD 570-573) was responsible for five deaths, all of which had been coded to ICD 571 by the Office of Population Censuses and Surveys. For this disease category there was no excess of mortality in the population as a whole (0 5, E 4.9) although the autoclave workers did contribute two cases (0 2, E 1.1), one of which had been certified as hepatic fibrosis and accepted by the coroner as a case of industrial disease. The remaining four deaths were due to cirrhosis of the liver. Death certificates for the 11 cases of nonsecondary liver cancer (which included the seven angiosarcomas found in the autoclave workers) gave no mention of hepatic fibrosis, although two (one hepatoma and one angiosarcoma) contained reference to cirrhosis.

Polyvinyl chloride dust exposure and disease
Baggers and driers (group B) were considered the high exposure group for PVC dust, craftsmen having less exposure and groups A and D the least. Table 5 shows the mortality from respiratory disease for the men who had ever been baggers or driers, regardless of previous or subsequent VCM or PVC exposure, and craftsmen who had never been autoclave workers or baggers or driers. The table shows that there was a nonsignificant deficit of malignant disease and a significant deficit of nonmalignant respiratory disease among the baggers and driers. The mortality for craftsmen showed no evidence of any significant excess of respiratory disease.

Discussion
This study reports the mortality experience of workers engaged in the manufacture and polymerization of vinyl chloride for the major chemical companies in the United Kingdom from 1940 on. In order to restrict the population to persons with definite occupational exposure and thereby prevent dilution of the cohort with substantial numbers of workers who had very small exposures, we have included only persons who have worked for more than one year in jobs judged to involve some exposure to VCM for at least 25 % of the workshift. These selection criteria reduced the size of the cohort originally reported on by Fox & Collier (11) from 7 409 to 5 498. Restriction of recruitment to the period for which the previous report was written ensured a minimum possible follow-up period of 10 years. In addition to any considerations of latency, work conditions would have been very different in the po st-1974 period with much lower exposures to YCM .
Ob ject ive hygiene data, alon g with est imate s based on ind ividu als' personal memory a nd judgement, showed that, while levels of expo su re had fallen during the time period studied, sub stantial redu ction s had been particularly associated with the mid to late 1950s and then again in the earl y 1970s. The hygiene data available indi cated that autoclave workers had the highe st expo sure to YCM , baggers and dr iers the highest PVC dust exposure , and craftsmen intermediate exposur e for both the monomer and the pol ymer .
In the ab senc e of detailed information on ind ivid ual expo sure, the cor relatio n of expos ur e and effect could have been attempted in a number of ways. After du e con sideration we decided that the analysis should be by occupation in accord anc e with the hygiene data. It is accepted that grouping the population in this way is a relatively crude method of assessing exposure and correlating this exposure with effect, and the accuracy of such a method depends upon the relevance of the groupi ngs chosen and the pre cision of historical occup ati on al data.
Another possible weakness of the study is that, while comparable mo rtality rates were availabl e for the period in which over 80 % of the death s occurred , expected death s fo r the period before the eighth ICD revision were calculated with the use of mortality data that were not dir ectly comparable. Thi s period would have accounted fo r o ver 25 % of the expected mortality. The dif ferences bet ween the eighth and ninth revisions dicta te the way in which the observed deaths have been categorized. An y bias introduced by att empts to define comparable data coded under diff erent ICD revisions would operat e differently for different diseases accord ing to the cha nges made in th e classi fication.

Overall mo rtality
Th e overa ll mortality of the subj ects studied (table 2) shows a stro ng healthy worker effe ct , more obvious for respiratory than circulatory diseases. The fact that the effect is more evident for some occupational groups than for oth ers may be a reflection of different physical demand s of work activity in different job s.

Liver cancer
Th ere was a n overall sta tistically significa nt excess of morta lity fro m non secondary liver can cer (0 II , E 1.94, SMR 567, 95 % CI 283-1 015), lar gely due to seven cases of angiosarcoma, six of which occurred at the same plan t and all of which occurre d in autoclave workers. No oth er nonsecondary liver tumors occurred among th e autocl ave wor kers, but these seven cases were suffici ent to pro duce a huge excess for th is disease cat egory in this group (0 7, E 0.38, SMR 1842, 95 % C I 74 1-3 795). The result s vivid ly demonstr ate the assoc iatio n between YCM expos ur e and an in-2 creased risk of pr imary liver tumor, even though thi s oncogenic effect is probabl y limited to the induction of angios arcoma. This was not the case a t the time of Fox & Coll ier 's ana lysis when onl y two an giosarcoma deaths had occurred . Ten years of addit ional follo wup have produced sufficient deaths from angiosar coma for non secondary liver tumors to be significa ntly in excess for th e whole population ; yet it must be realized that thi s excess is the result of seven angiosar comas , all of whic h occurred in autoclave wor kers, and the remaining workers contribute little to the nonsecondar y liver tumor death s (0 4, E 1.58, SMR 253, 95 % CI 69-648). Thi s non signifi cant excess cannot be taken as evidence of an association between YCM and nonangiosarcoma primary liver tumors . If VCM doe s playa part in the genesis of other primary liver tumors, its effect is clearl y far weaker than for angiosar coma, and the pow er of this study is insufficient to permit further comment.
Mortality studies are dependent upon the accur acy of death certi fication . Subsequent classification of the information on the certificate is subject to coding as carried out by a nosologist within the con straints imposed by the leo. Com parative mortality dat a are not therefor e a co mpletely accurate measure of disease incidence, even for fatal cond itions. The smaller the true incidence of the disease, the more easily an y measure of it will be a ffected by allocation of ind ividual cases to a part icular category . Because of co-existing recordings of an giosarcom a (3, 10), it was po ssible to check the result s of thi s mortality analysis against the figure for cases of VCM -relat ed an giosarcoma of the liver in the United Kingdom. T wo cases of an giosar coma were ident ified in our coho rt, for which the underlying cau se of death had not been classified to an y code fo r liver ca ncer. One described as hemangioendothelioma of t he liver was coded to ICD (eighth revision) 227 (benign hemang ioma/lymphangioma). In the case of the o ther man, histological evidence of angiosarcoma was found, but no mention was made of it on the death certificate, even though a verd ict of industrial disea se was given. Both men had been autoclave work ers.

Lung cancer
There was a no nsignifi cant deficit of lun g cancer in the co hort overa ll. If YC M doe s have a car cino genic effect on the lung, then, given the usual latency fo r occup ational lung cancer, any excesses should be seen in the lon ger follo w-up groups. Table 6 shows the results for different period s of follow-up . No clear eviden ce of any increased lun g cancer risk emerges with lon ger period s of follow-u p. Autoclave workers followed for 20 years or more show a non statistically significant excess of lun g cancer (0 10, E 6.5, SMR 154, 95 % CI 74-283), but a non stati stically significant excess in such sma ll numbers mu st be viewed with caution. Thu s the results o f the study do not demonstrate any association between VCM exposure and lung cancer deaths. For any study of this kind its power (ie, ability to demonstrate excess disease risks where they exist) will be far greater for common causes of death such as lung cancer than for uncommon tumors such as primary liver cancer. A negative result could therefore be taken as evidence that, if VCM had an y etiologic role in the development of lung cancer, it must be small in comparison with that seen for angiosarcoma of the liver. It should be remembered, however, that lung cancer is a common tumor becau se of the high prevalence of smoking during the time of this study period. Without smoking, lung cancer would be a far more uncommon disease. If vinyl chloride exerted a carcinogenic effect on the respiratory system that was not synergystic with the carcinogenic effect of smoking, then such an effect would be masked by the commonality of smokinginduced tumors. Thu s the power of the stud y to detect excess lung cancer is not the only consideration to be borne in mind when VCM is being assessed as a possible respiratory carcinogen.

Brain cancer
On the whole, the population showed no excess of brain cancer, and, because of the small numbers associated with the individual exposure categories, the results do not warrant further discussion. The exposure and latency details for the four cases included in the analysis are given in table 7 along with two cases that have occurred in the post-1974 starters. This study does not offer any anecdotal or statistical evidence for an association between brain cancer and VCM exposure.

Lymphatic cancer
Because of the close association between non-Hodgkins lymphoma and chronic lymphatic leukemia, it is pertinent to consider the diseases which span the three different ICD codings 200,202, and 204 together. Two deaths coded to lCD 200 for autoclave operators were numerically in excess of the expected figure (0 2, E 0.47, SMR 426 , 95 070 CI 52-1 537), but this value did not reach statistical significance. There was also a nonstatistically significant excess of lymphatic leukemia among the workers with low exposure (group D) (0 3, E 1.1 , SMR 273, 95 % CI 56-797). The problems of interpreting such small numbers are considerable. However, the evidence of this study alone cannot be considered suggestive of a causative role for VCM in the production of lymphatic cancer.

Other cancers
Although an association for VCM exposure and cancer of the colon has previously been commented upon (12,24), this study provides no evidence of an excess of large bowel cancer in persons exposed to vinyl chloride.
In the population described in this study, only two deaths from malignant melanoma occurred. Both involved men from the low-exposure group (02, E 0.94, SMR 213, 95 % CI 26-769). There was therefore, nothing in these results to support an association between VCM and the development of melanoma of the skin.
Two deaths from cancer of the thyroid occurred. One autoclave operator who worked in this job for one year died of thyroid cancer II years after his first exposure to vinyl chloride. The other case involved a man who was a bagger for 14 years and died some 23 years after first exposure. Although not as rare as angiosarcoma, cancer of the thyroid is still a very uncommon tumor in men (approximately 120 deaths per year and about double that number of male cancer registrations Frontal astrocytoma of the brain in England and Wales). While it is not pos sible to interpret the occurrence of these two cases as being due to VCM, they are described in detail because of a similar nonsignificant excess having been described in another study (12). The significant excess of urinogenital cancers (largely bladder and prostate) limited to three factories was an unexpected finding. The facts that autoclave operators did not show such an excess, that it was hea vily contributed to by low-exposure workers, and that it was confined to three particular factories make an association with VCM unlikely. Occupational exposure to other chemicals on such multiprocess sites must be considered. However, in a study in which so many comparisons have been drawn, the possibility of this being a chance finding must remain. A case-referent study of these cases is to be carried out in order to investigate other possible occupational explanations.

Nonmalignant disease
In addition to the well-known aSSOCIatIOn between vinyl chloride and angiosarcoma of the liver, noncirrhotic periportal fibrosis and portal hypertension is a prescribed occupational disease in the United Kingdom under the Industrial Injuries Provisions of the Social Security Act 1975 (6). An analysis by underlying cause of death showed no evidence of an excess of nonmalignant liver di sease , although one certificate included mention of hepatic fibrosis which was adjudged by the coroner to be an industrial disease.
Noncirrhotic periportal fibrosis in VCM wor kers has, of course, been described (25), and some of the se workers ha ve subsequently developed angio sarcoma of the liver. The certificates for one case of hepatoma and one of angiosarcoma mentioned cirrhosis, but this disease was not considered to be the underl ying cause of death. Ca ses of periportal fibrosis among VCM workers are probably underreported in mortality studies because other pathological conditions, including subsequent development of angiosarcoma of the liver, are likely to be quoted as the underlying cause of death .
PVC dust exposure has been shown to produce radiographic chan ges and a reduction of the ratio of forced expiratory vo lume in 1 s to forced vital capacity, but the clinical significance of these changes is unclear. The workers deemed to be the mo st exposed to PVC du st in th is study (table 5) had a deficit of mortality from respiratory di sease; thus there is no evidence in this study to suggest a relationship between PVC expo sure and incre ased mortality from respiratory disease. Respiratory di sea se is, of course, considerably affected by smo king habits . Little inference can therefore be drawn from such small numbers with smo king data unavailable.
To conclude, the overall mortality pattern of VCM workers in the United Kingdom shows a typical health y worker effect associated with such cohort populations. These workers have a significant excess of primary liver tumor illustrating a link between the VCM exposure of autoclave workers and the development of angiosarcoma. There is little evidence of any relationship between other causes of mortality and VCM expo sure .