Document 6R2YRGn7zGwmQwXa0w4LKnLo6

Effects of exposure to vinyl chloride An assessment of the evidence by Sir Richard Doll, FRS1 DOLL R. Effects of exposure to vinyl chloride: An assessment of the evidence. Scand J Work Environ Health 14 (1988) 61--78. This paper reviews the possible effects of vinyl chloride on the mortality of oc cupationally exposed men and the carcinogenic effects that might be observed in the general population as a result of environmental pollution with vinyl chloride. The results of four studies fulfilling the criteria of providing substantial numbers of observations more than 25 years after first exposure and covering a period long enough for more than 10 of the workers to have been expected to die constitute the basis for the assessment of the occupational hazards. Other studies provide only supplementary information. The data permit two conclusions. First, men occupationally exposed to vinyl chloride have experienced a specific hazard of angiosarcoma of the liver. Second, an> other occupational hazards that may have existed have been small. No positive evidence of a hazard of any nonmalignant disease or any type of cancer other than angiosarcoma of the liver has been found except possibly for a small hazard of lung cancer when exposure was heavy. More definite conclusions might be reached if those who have studied exposed employees could present their results in appropriate and comparable ways. A very small risk of angiosarcoma may have occurred as a result of vinyl chloride escaping into the environment around plants handling vinyl chloride in the past, but the evidence indicates that the current risk to the general public (if any) must be negligible. Key terms: angiosarcoma of the liver, cancer, lung cancer, mortality, polyvinyl chloride, review, vinyl chloride monomer. tki ) 608 644 (redactory it number 0.00 air mail address above, paged issues unless :ht Clearance Center ; internal or personal nter's Transactional rage, is paid directly lions that have been re fee code for users For many years the inhalation of large amounts of vinyl chloride has been recognized as potentially haz ardous. Concentrations of the order of 10 000 ppm in the air induce unconsciousness and cardiac arrhyth mia, while prolonged exposure to concentrations an order of magnitude lower have been liable to cause a specific pathological syndrome. This "vinyl chloride illness" has been characterized by four cardinal signs, namely, enlargement of the liver and spleen with a specific histological appearance, patchy infiltration of the skin resembling scleroderma, bony changes in the tips of the fingers described as acroosteoiysis, and peripheral circulatory changes identical with the clas sical picture of Raynaud's disease. These pathologi cal reactions may occur singly or together and may pos sibly be accompanied by other less characteristic ef fects. They can, how ever, be completely avoided if ex posure never exceeds the level of a few- hundred parts per million, ie, the level to which exposures were generally reduced in the mid-1960s. One other serious effeci has, however, been observed that may not be avoidable in the same relatively easyway, namely, the production of angiosarcoma of the liver. It must, indeed, be presumed that some risk of developing the disease will persist from exposure to 1 Imperial Cancer Research Fund, Cancer Epidemiology and Clinical Trials Unit, Radeliffe Infirmary, Oxford, Unued Kingdom. Reprim requests io: Sir R Doll, Imperial Cancer Research Fund, University of Oxford, Gibson Building, The Radeliffe Infirmary, Oxford OX2 6HE, United Kingdom. doses that are even lower than the current industrial levels of 5 ppm or less, as vinyl chloride has been show n to act as a mutagen (23), and it cannot be assumed that a threshold exists below which no carcinogenic risk persists. Moreover, the possibility has to be considered that vinyl chloride may cause some cancers other than angiosarcoma of the liver, partly because laboratory studies have shown that it causes other cancers in animal experiments and partly because the initial studies demonstrating the production of angiosarcoma of the liver in humans were inadequate in size to ex clude a material increase in the risk of cancer in com mon sites, such as the lung and large bowel. Since no threshold dose can be postulated, it also follows that some cancers may have been produced in the general public by the small amounts that have escaped into the general environment. Consideration also needs to be given to the possi bility that exposure to vinyl chloride over a long pe riod may have noxious effects on humans that cannot be seen easily in animal experiments (by, for example, producing chronic respiratory disease), and, as it is a mutagen, there is also a possibility that it may act as a teratogen and cause congenital malformations in offspring. In this review 1 have not examined the possibility that vinyl chloride acts as a teratogen or that it causes mutations in germ cells, as there is too little serious evidence to justify inclusion. Reviews carried out for sections of the industry by Downs et al (unpublished report to the Society of Plastic Industries Inc in 1977) i 61 and by MacMahon (unpublished report to the Chemi cal Manufacturers Association in 1977) concluded that the few reports of positive effects could not be sub stantiated, and no additional evidence was found in a similar later review by Barr (unpublished report to Air Products and Chemicals Inc in 1981), apart from a report that embryos were absorbed and skeletal os sification was produced when pregnant rats were ex posed to doses appreciably lower than those that had been used by other workers without any such effects being observed. The report of absorbed embryos (34) could not be evaluated thoroughly, however, as the ex periment was inadequately described. I have, therefore, examined only the possible effects on the personal health of men occupationally exposed to vinyl chlo ride, other than those related to their reproductive capacity, and the carcinogenic effects that might con ceivably be observed in the general population as a result of the widespread distribution of vinyl chloride as a pollutant. Occupational hazards Many studies of workers exposed to vinyl chloride in the manufacture of vinyl chloride monomer (VCM) and polyvinyl chloride (PVC) have been undertaken since it was first found that vinyl chloride could cause cancer in animals (28, 48) and man (10). These inves tigations have confirmed that exposure causes a haz ard of angiosarcoma of the liver and, in several in stances, have shown excess incidence or mortality rates that were conventionally statistically significant for other diseases. Conventional tests of statistical signifi cance are, however, designed to help answer single questions defined beforehand, and several findings that might be expected to occur by chance alone once in (say) 20 times must be expected to occur if dozens of rates are examined in each of several sets of indepen dent data. Scientists have, therefore, been faced with the problem of deciding whether the excess rates that have been observed in individual studies are due to oc cupational hazards or to the vagaries of chance. This problem can be solved in part with an exami nation of the results of a sumation of data from com parable studies, that is, by a comparison of the sums of the numbers of deaths observed and expected in each study. This procedure does not require the as sumption that the exposures have been the same in each study any more than the same assumption is required for each individual when the results of each study are considered alone. It does require however that each exposed population has been observed over a period when its members were at risk of developing disease (if a genuine hazard existed) and that in each study the reference population from which the expected num bers of deaths were derived was appropriate (that is, at the same risk of developing disease as the exposed population would have been in the absence of expo sure). These requirements do not introduce any new complexity, as both are, of course, also required if cor rect conclusions are to be drawn from the results of the individual studies when they are examined on their own. Sources of information Four studies meet the aforementioned requirements, namely, two large national surveys, one reported by Jones (25) for the United Kingdom (UK) and the other by Environmental Health Associates (14) for the United States (US), and studies of individual plants in Canada, reported by Theriault & Allard (46) and in Italy, reported (as a pan of a national study) by Belli et al (4). All four include observations on men more than 25 years after their first exposure, and the ex pected number of deaths is, in each case, greater than 10 V# of the total number of employees, a value indi cating a long average period at risk. Earlier observa tions on UK and US employees (6, 9, 12, 16, 20, 36, 39,40,45, 50) have been subsumed in the national sur veys and now' serve only as sources of hypotheses and of some detailed information not included in the na tional reports. Studies of German (49), Norwegian (22), Swedish (7), French (41, and unpublished report of Laplanche et al), Japanese (33, 37), and some other Italian (4) workers provide some supplementary infor mation, but, in general, the periods of observation have not been long enough for useful epidemiologic data to be obtained about diseases that are unlikely to occur within 20 years of first exposure, or they report only selected results which are difficult to in terpret, as only excess rates tend to have been selected. Studies of makers of PVC products have not been included, as the workers have had much less exposure to vinyl chloride than those employed in the manufac ture of VCM or PVC and any occupational hazard to which they may have been exposed is more likely to have been produced by PVC dust. US study. The study carried out by Environmental Health Associates (14) on behalf of the US Chemical Manufacturers Association is the largest and most in formative investigation thus far undertaken. It cov ered 10 173 men who had worked in 37 plants owned by 17 companies -- 1 214 men in 11 plants that produced only VCM, 6 848 men in 18 plants that produced only PVC, 935 men in three plants that produced both, and I 176 men in five plants that produced homopolymers and copolymers, with or without VCM or PVC. Twenty-two of the plants were in the southern part of the country, 14 were in the northeastern or north central parts, and one was in the west. Men were included if they had been exposed to vinyl chloride for at least a year before 31 December 1972 and had been employed in 1942 or subsequently (the first year depending on the date the plant began making 62 ASI 00005597  QLintroduce any newrequired if cor- xBRsm the results of are examined on their ttioned requirements, .eys, one reported by >m (UK) and the other >ociates (14) for the t of individual plants ult &. Allard (46) and tational study) by Belli vations on men more xposure, and the exach case, greater than iployees, a value indi risk. Earlier observa- (6, 9, 12, 16, 20, 36, led in the national surces of hypotheses and ot included in the naman (49), Norwegian nd unpublished report 3, 37), and some other supplementary inforeriods of observation meful epidemiologic ^Athat are unlikely ^^^xposure, or they ich are difficult to in1 to have been selected, roducts have not been ad much less exposure iloyed in the manufac>ccupationa! hazard to >osed is more likely to Just. out by Environmental df of the US Chemical he largest and most inar undertaken. It cov ied in 37 plants owned aen in 11 plants that men in 18 plants that n in three plants that en in five plants that I copolymers, with or ,-re in the southern part ; northeastern or north the west. td been exposed to vinyl fore 31 December 1972 42 or subsequently (the ^^lant began making or using vinyl chloride and the earliest date that per sonnel records were deemed to be complete, whichever w'as the later). Individuals who met these criteria were identified from company records by company per sonnel. Racial characteristics were known only for 686 men, 97 Vo of whom were white, and it was presumed, for the purpose of estimating the number of expected deaths, that all 10 173 men were white. Follow-up data were obtained from plant and So cial Security Administration records and (for men who died after 1979) from the National Death Index. Five plants did not collaborate in the extension of Cooper's (9) earlier study, which had been subsumed in the present investigation, and the 955 employees in these plants who were known to be alive on 31 December 1972 were not followed any further. For the rest, follow-up was attempted to death or 31 December 1982, whichever was the earlier. On this basis 92.7 Vo of the men were successfully traced. Those who were untraced were excluded from the last date of contact, which was usually the date when employment ceased. Almost half of the men (46 Vo) were first employed before 1955. A large proportion was, therefore, ob served more than 25 years after first exposure (and in many cases for more than 30 years) when diseases with a long latency period might be expected to be seen. Short-term workers had been excluded from the cohort, and most of the men had continued in em ployment for many years, two-thirds being employed for 10 years or more and the average duration of em ployment being 16 years. Fifteen hundred and thirty-six men were found to have died. In 1 439 cases, the cause of death was ob tained from the death certificate, but no cause was ob tained for the other 97 persons (6.3 Vo). The numbers of deaths expected from each of 38 causes or groups of causes were obtained by multi plying the person-years at risk by the disease-specific national rates for white males, for the corresponding age groups and five-year periods of the study. This important study is open to three minor criti cisms, which are unlikely to have had any material effects on the results. First, the lists of employees were compiled by comparn personnel from company re cords without any independent check. Second, the as sumption that all the employees were white will have caused the expected deaths to have been very slightly underestimated, as the few' black employees are likely to have had higher mortality rates and there is no reason for supposing that the small sample from which the proportion of black employees was estimated was necessarily representative. Third, an element of uncer tainty was introduced by the failure to trace as many as 7.3 Vo of the employees. Two other criticisms are more important. First, the expected numbers of deaths were calculated on the as sumption that the men would have experienced the same mortality rates as the white male population of the whole country at the corresponding dates. The use of national rates is common practice in studies of in dustrial populations and tends to result in an over estimation of the expected numbers of deaths so that the employees appear to be unusually healthy. This "healthy worker effect" is well known and has been taken into account in my discussion of the results. A more serious objection to the use of national rates is the way mortality varies from one part of the country to another, due to differences in the prevalence of en vironmental and social factors unrelated to the occu pation of interest. It is, therefore, generally preferable to use state (if not county) rates in place of national rates. With 37 plants, however, it might be thought that their geographic distribution would be sufficiently wide to make the use of national rates appropriate. Unfortunately 22 of the plants were located in the south, and a check would have been desirable to see whether their location could have caused any material distortion of the results. Second, causes were not obtained for 97 of the 1 536 deaths. This deficiency was allowed for in the calcu lation of the overall mortality by the inclusion of deaths due to unknow n causes. It was not allowed for, how ever, in the calculation of the disease-specific mortality rates and will have caused the standardized mortality ratios to be underestimated by an average of 6.3 %. For the present purpose, therefore, the numbers of deaths attributed to specific diseases have each been multiplied by 1.0674 [100/(1--97/1536)] and rounded off to the nearest integer. UK study. The study reported by Jones (25 and un published) on behalf of the British Health and Safety Executive covered 5 498 men who were employed for at least one year in jobs that involved potential expo sure to VCM for at least 25 Vo of the work week and who were first employed in the period 1940--1974. De tails of the men were compiled from the personnel records of nine chemical plants manufacturing or poly merizing vinyl chloride, and the vital status of the men was determined at the end of 1984 from the records of the National Health Service Central Register. Five thousand four hundred and ninety-eight men were traced (98.9 Vo). Seven hundred and eighty deaths were identified, and copies of the death certificates [coded to the eighth revision of the International Classifica tion of Diseases (ICD) if they occurred before 1979 and to the ninth revision if they occurred later] were sent to the investigators. Several specific points about the study need to be noted. First, national mortality rates for England and Wales were calculated for five-year age groups over quinquennial periods for 66 causes of death, and these rates were used in the estimation of the numbers of deaths that might have been expected in the cohort by multiplying them by the corresponding numbers of person-years under observation. Some difficulty which could have been related to the causes of death being 63 ! coded according to the eighth and ninth revisions of the 1CD was, however, experienced in obtaining suit able rates for all causes of death, and rates for a rela tively late period had to be used for estimating the numbers of deaths from many diseases that might have been expected to occur in earlier periods. For two categories the earliest available rates were 1960--1964, for one they were 1965--1969, for 24 they were 1970-- 1974, and for one they were 1975--1979. Second, an attempt was made to classify men ac cording to whether they had high, intermediate, or low exposure to VCM or PVC dust, and each man's em ployment history was recorded according to 12 job titles with advice from the plants concerned. The men were then grouped according to w hether exposure to VCM was likely to have been high (group A), expo sure to PVC dust was likely to have been high with exposure to VCM low (group B), or exposure to VCM and PVC dust was intermediate and intermittent (group C). All other men, who would generally have had low exposure to both VCM and PVC dust, were classed as group D. Within all the groups, exposure to VCM was likely to have been higher if it had begun before 1956. The study makes an important contribution to the knowledge concerning the long-term effects of vinyl chloride. The use of national rates to calculate the ex pected numbers of deaths may be justified on the grounds that the men were employed in nine plants, which were presumably distributed about the country, but no details of their location are given. In general, mortality rates tend to be higher in the parts of Britain where heavy industry is located than in other parts of the country so that the expected numbers of deaths are more likely to be biased downwards than upwards; but whether this is so or not needs to be shown. The use of recent rates to calculate expected num bers of deaths from many specific causes of death was presumably necessary if the diseases were to be studied individually and will have done no harm if the inci dence and fatality of the diseases in question remained stable. It would have been desirable, however, for the diseases to have been specified so that the reader would know which were liable to be distorted. The system used to classify the men into four expo sure groups is the sort of system that is commonly used if precise measures of exposure are not available. It creates some difficulties in the statistical analysis if men move from one job to another and are classed (as in this instance) as having had high exposure if they have ever had a particular type of employment (eg, ever been employed as an autoclave worker). No evidence is pro vided to show that the person-years at risk before a man entered the category have been subtracted and added to another exposure group before the numbers of expected deaths were calculated. Movement from one job to another was said to have tended to be out of groups A and B into D, but even so it must be presumed that the expected numbers of deaths in the first two exposure categories are likely to have been overestimated. Canadian study. The Canadian study (46) was limited to employees of a single plant in Shaw inigan, Quebec. The plant, which was situated in an industrial com plex, w'as opened in 1943. VCM and PVC were both made until the late 1960s, when the production of VCM ceased, while the production of PVC continued. An attempt was made to trace all the production work ers whose names appeared on the unions' lists or the payrolls of the companies in the whole industrial com plex, including the vinyl chloride plant, between 1 January 1948 and 31 December 1972, and contact was made with the worker or his next-of-kin in 1 611 out of 1 659 instances (97.1 o). Detailed occupational and smoking histories were obtained by questionnaire, and 156 men who had been employed by the companies for less than five years were excluded. The remaining men were categorized as (i) exposed to VCM if they had worked on the production of VCM or PVC for at least five years (451 men), (ii) unexposed to VCM if they had worked similarly for less than six months (870 men), and (iii) other men (134 in total). The last group was excluded from the study. Follow-up was closed on 31 December 1977. Copies of the death cer tificates were obtained, and the causes for all who had died (59 exposed and 233 unexposed) w ere coded ac cording to the eighth revision of the 1CD. Informa tion was also sought for histological or cytological con firmation of all the diagnoses for all the exposed men who had died of cancer. The results were examined in two ways. First, the mortalities of the exposed and unexposed men were compared after standardization for five-year periods of the study and five-year age groups. Second, the mor tality of the exposed men was compared with that ex pected if the men had had the sex- and age-specific mortality rates recorded in Quebec for the year 1971. In both comparisons the causes of death used were those specified on the death certificate, and the addi tional pathological information was used later only for interpretation of the results. Most of the exposed men were exposed for more than 10 years (75 to), the average length of exposure was approximately 17 years, and 44 to were observed more than 25 years after first exposure. Although small, the study makes a useful contribu tion to the overall results. The histological review of the cancer cases is particularly helpful. It showed that all eight cancers diagnosed as liver cancer (including two specified as hepatoma and one specified as angio sarcoma) were angiosarcomas of the liver, as well as one that had been diagnosed as angiosarcoma of the peritoneum. Two other cases of angiosarcoma of the liver were found to have been certified as hepatic cir rhosis. It is also helpful to have a comparison between 64 00005599 AS I * mbeg\rs of deaths in the wkely to have been study (46) was limited Shawinigan, Quebec, in an industrial com1 and PVC were both en the production of on of PVC continued. . the production workhe unions' lists or the whole industrial comide plant, between I 1972, and contact was xt-of-kin in 1 611 out tiled occupational and by questionnaire, and ed by the companies luded. The remaining rosed to VCM if they of VCM or PVC for i) unexposed to VCM r less than six months 134 in total). The last itudy. Follow-up was opies of the death cercauses for all who had nosed) were coded ac4^khe 1CD. Informa- or cytological Con or all the exposed men t two ways. First, the unexposed men were i for five-year periods >ups. Second, the morompared with that exsex- and age-specific bee for the year 1971. _*s of death used were tificate, and the addiwas used later only for ere exposed for more ige length of exposure d 44 "o were observed exposure. tkes a useful contribuhistological review of oelpful. It showed that iver cancer (including one specified as angioof the liver, as well as s angiosarcoma of the f angiosarcoma of the ertified as hepatic cir^afpmparison between the exposed and "unexposed" employees of the same companies as it shows that the low mortality observed for all nonmalignant diseases could be attributed to a healthy worker effect and was not due to bias in the recording of exposure (relative risk for all nonmalig nant causes compared to that of the "unexposed" men 0.95). One aspect of the study has to be criticized however, ie, the use of provincial rates for one year (1971) to calculate expected mortality spread over a 30-year period (1948 to 1977 inclusive). Deaths will have tended to bunch up towards the end of the period of obser vation so that the rates for this particular year may have been fairly representative, but it must have caused some distortion of the expected numbers of deaths, the size (and even the direction) of which is impossible to estimate. For most disease groups the distortion is un likely to have been large. Italian study. A study of all men employed in the pro duction of vinyl chloride and PVC in nine Italian plants was begun in 1983. All men were included who w'ere employed for at least six months at any time from the start up of the plant to the end of 1981. The study is still incomplete, but results are now available for men in three plants (4). Two plants (in Ferrara and Rosignana) began operation in 1953. Four hundred and thirty-seven men were employed in one plant and 181 in the other. All but three (from the Ferrara plant) were followed to the end of 1984. Expected deaths were esti mated by multiplying the person-years at risk by the corresponding national mortality rates for each fiveyear age group and each five-year period of the study. The total expected deaths in each case amounted tc more than 10 7o of the employees in the two plants (12.4 and 12.8 %). Clinical information was sought about the cause of death of all the 55 employees of the Ferrara plant who had died. Revised diagnoses, which were not used for comparison with the expected deaths, revealed four deaths from cancers of the liver in place of one. The Ravenna plant did not begin operation until 1959. Six hundred and thirty-eight men were employed. All but four were traced to the end of 1983, and 17 were found to have died. No man could have been fol lowed for more than 24 years, and only 25.1 deaths (3.9 <7o of the work force) were expected. The data for this plant have not, therefore, been used in the prin cipal analyses. It may be noted, however, that one death was attributed to liver cancer when 0.1 was ex pected. Other sources. The Norwegian study (22) provided ob servations on 454 men who had been employed in a plant in Telemark where VCM had been manufactured from 1950 to 1971 and PVC from 1950 to the end of the study period. Every man was included whose name was recorded in the company's personnel register and health department records who had ever been em ployed from the start of production to the end of 1969 and had worked for at least one year. The men were followed from 1953 to 1979 inclusive. Deaths and cases of cancer were identified from the records of the Cen tral Bureau of Statistics and the national cancer regis try. No reference was made to any men being lost to follow-up, but it can be assumed that the number (if not zero) was small, as all citizens have an identity number which is used by both employers and central agencies. Fifty men were found to have died against 59.34 expected if the sex-, age-, and quinquenniumspecific national mortality rates had operated. Twentyone men were found to have developed 23 cancers against 20.16 cancers expected from the comparable national incidence rates, the use of which was justi fied by the finding that the incidence in the county in which the plant Was situated was between 90 and 95 % of the rate of the country as a whole. One man who had been employed in PVC production developed an giosarcoma of the liver. The observed and expected numbers of cases were given for cancers of the lung, colon, and thyroid, for melanomas, and for all can cers, but no expected numbers were given for other types of cancer. It is evident that several other types of cancer must have been in deficit, as there were eight cases in all against 14.93 expected, and it is difficult to know what weight to give the excesses observed for the reported types of cancer, as they seem likely to have been reported specifically because the numbers were in excess of those expected. The authors noted that one further case of melanoma had occurred after the closure of the study and that one "incipient case" was also known to them. The German study (49) included the following three groups: (i) 7 021 men w ho had been exposed to VC in the course of their employment in any of the 11 plants in which VC and PVC had been produced in the Federal Republic of Germany, (ii) 4 820 men who had been employed in seven chemical plants without having had any exposure to vinyl chloride, and (iii) 4 007 men employed in two other plants where PVC was processed. Employees were included only if they were of German or Austrian nationality, and they were regarded as exposed to vinyl chloride if they were production workers or other skilled workers or laborers assigned regularly to the plants, but not if they were employed in them only occasionally. All the men were included from the time of opening of the plants to the end of 1974, and they were followed to the end of 1974. Many of the men were therefore observed for only a few years after first employment, and only 14, 36, and 19 %, respectively, of the three groups were first em ployed before 1954 and were therefore capable of con tributing person-years at risk more than 20 years after first employment, when an occupational hazard of can cer could be expected to be observed. 65 I Of the exposed group 93.2 V# were successfully fol lowed, and causes of death were discovered for 92.8 Vo of the 414 men discovered to have died. The propor tions for the other two groups were respectively 89.8 and 88.7 Vo for the unexposed and 92.1 and 86.9 % for the PVC process workers. The failure to obtain causes of death for all the men who had died was allowed for in the subsequent analysis by the weight ing of the numbers attributed to each cause by a sys tem which took account of the age group and calen dar period in which death with an unknown cause oc curred. The expected numbers of deaths from each cause was calculated by multiplying the person-years at risk by the sex-, age-, and cause-specific mortality rates for the Federal Republic of Germany. National data before 1968 used an idiosyncratic classification system, and the 1968 rates had to be used to multiply all the person-years at risk up to the end of 1968. For subsequent years (1969 to 1974) the person-years at risk were multiplied by the corresponding rates for the same calendar year. Epidemiologic studies are more difficult to carry out in the Federal Republic of Germany than in North America, the United Kingdom, or Scandinavia because the medical cause of death is not recorded publicly, and there is no central system which can be used for checking whether an individual is alive or dead. In these circumstances, the German authors have made valiant efforts to obtain reliable data, and the propor tions of men in the exposed groups who were not suc cessfully followed (6.8 170) and the proportions of deaths for which the cause was not obtained (7.2 Vo) were similar to those in the study of the Environmental Health Associates (14). Two defects, however, make the data less useful. First, no national mortality rates were available before 1968, and the use of the 1968 rates to estimate the num bers of deaths in and before 1968 will have overesti mated the numbers attributable to diseases that were becoming more prevalent or were being diagnosed more often and underestimated those due to diseases that were becoming less prevalent. Second, and more importantly, a large proportion of the men had been first employed less than 10 years before the follow-up ended. Therefore the useful observations on the few men who had been exposed long enough to have had much chance of developing an occupational disease with a long latency period must have been swamped by a mass of other observations that had little to con tribute. The expected deaths amounted to only 6.2 Vo of the exposed men, and there is, therefore, little to be gained, and something to be lost, by including the German data in the overview. It may be noted, how ever, that 12 deaths were attributed to cancer of the liver among the workers exposed to vinyl chloride against 0.9 expected and that smaller excesses were also observed among the unexposed chemical workers (4 observed against 1.1 expected) and the PVC process workers (3 deaths against 0.8 expected). Two Swedish plants have produced VCM and PVC, one since 1945 and the other since 1971, and employees of the first plant bave been studied by Byren et al (7). All persons who had ever been employed when expo sure to VCM coaid occur were listed from the per sonnel files of tl*factory. Twenty-one were excluded because they were foreigners who left the country after a short period ofemployment. The remaining 750 were followed to October 1974. Expected numbers of deaths were estimated by multiplying the person-years at risk by the corresponding age-specific mortality rates for the whole country, and the expected numbers of can cer cases from 1958 to 1971 inclusive (during which period all cancercases had been registered nationally) were estimated by multiplying by the national agespecific cancer kcidence rates. In both instances, the rates used were those recorded in 1969. Fifty-eight deaths were found, but no figure was given for the ex pected number. Detailed figures were given only for the numbers of deaths and cases observed and expected for cancer of the lung and for cancers of the liver and pancreas combined and for the numbers of deaths from brain cancer and three categories of cardiovas cular disease. T*o men known to have angiosarcoma of the liver were certified as having died of liver can cer or pancreatic cancer, and a third man died of angiosarcoma of the liver 17 months after the close of the follow-up. Two French sudies provide the results of a long term follow-up of men employed in one plant (41) and of a short-term follow-up of men employed in 12 plants (Laplanche et al. unpublished). The first provided ob servations on 1 311 men exposed to vinyl chloride in the production of VCM and PVC and in selected an cillary operations from the opening of the Tavaux plant in 1953 to the ad of 1976 (41). Six other employees were excluded from the study because of lack of oc cupational histories and 160 men because their vital status at the end of the study period was undetermined. Twenty-five men were found to have died against 48.75 expected from contemporaneous sex- and age-specific national mortality rates (3.7 Vo of the men at risk). One death was attributed to angiosarcoma of the liver, in a man who had been exposed for more than 15 years. The reported data are so incomplete and cover such a relatively short period from the opening of the plant that they add nothing of epidemiologic value to the results of the other studies, apart from the addition of a further case of angiosarcoma. The second study provided observations on 1 100 men aged 40 to 55 years who, in 1980, were exposed or had been exposed to vinyl chloride in 12 plants, which constituted "most of the French VCM polym erisation plants" (Laplanche et al, unpublished). Many of the men were, or had been, employed at Tavaux and were presumably survivors of the cohort studied by Pierre et al (41). The men were followed for five years, and their morbidity and mortality were com pared with those observed for 1 100 men of the same 66 ASI 00005601 # VCM and PVC, e 1971, and employees died by Byren et al (7). employed when expoe listed from the pernty-one were excluded 0 left the country after he remaining 750 were ted numbers of deaths te person-years at risk fie mortality rates for rcted numbers of canclusive (during which registered nationally) by the national ageIn both instances, the * in 1969. Fifty-eight 1 was given for the exs were given only for observed and expected mcers of the liver and e numbers of deaths egories of cardiovaso have angiosarcoma ing died of liver cana third man died of tonths after the close ^Aresults of a long- one plant (41) and employed in 12 plants The first provided obd to vinyl chloride in C and in selected anig of the Tavaux plant Six other employees ecause of lack of ocm because their vital od was undetermined, ave died against 48.75 sex- and age-specific "the men at risk). One coma of the liver, in more than 15 years, plete and cover such . opening of the plant niologic value to the rt from the addition ma. bservations on 1 100 r 1980, were exposed nloride in 12 plants, French VCM polym, unpublished). Many employed at Tavaux if the cohort studied rre followed for five tality were cornmen of the same ages ( 2 years) who were employed in the same plants but who had never been exposed to VCM. The men in both groups were interviewed personally, and iniormation was obtained about their smoking and drinking habits, which were found to be similar in the two groups. The men in the exposed group had been first exposed for an average of about 14 years and had first been employed in the plant about 18 years previ ously. Morbidity and mortality data were recorded an nually by the plant physician, who successfully traced 98 Vo of the exposed men and 96 % of the referents. One of the exposed men, but none of those unexposed, developed an angiosarcoma of the liver. Data were not gi\en separately for different periods after first em ployment, and it is impossible to assess the significance of the finding that six of the exposed men developed lung cancer against two of the referents, which may well reflect a chance occurrence of unusually few cases in the reference group, as the proportion of all lung cancers in that group (2 out of 15) was unusually low. One exposed man developed a cancer of the lymphohematopoietic system against none of the referents, but none of the men in either group were known to have developed melanomas or cancer of the brain or thy roid. A Japanese study has reported the mortality ex perience of 4 524 men employed for at least one year before 1965 in 25 Japanese plants which began pro ducing VCM or PVC between 1949 and 1964 inclu sive (37). The men were followed to 31 October 1975. Twenty-eight percent of the men were observed more than 20 years after first employment, but none was ob served more than 26 years. Only 0.6 Vo of the men were untraced, and copies of the death certificates were ob tained for all the 209 men who had died (4.6 Vo of the initial cohort). Individuals were classified according to the job in which they had been longest employed at the termination of their follow-up, and data were given separately for the 2 546 men classed as employed in PVC production and 1 978 others (including 900 classed as VCM production workers). If this study is continued for another 10 years, it should provide useful additional information, but the present data include too few observations on men more than 20 years after first exposure to be of any material use. They confirm the evidence of a hazard of liver cancer with six deaths among the PVC production workers against 2.54 expected from national rates, w'hile only one such death was observed among the other workers against 1.82 expected. One of the six deaths from liver cancer among the PVC production workers was certified as due to angiosarcoma of the liver, and at least one of the other liver cancer deaths was due to the same cause. Lung cancer deaths were not in excess (2 observed in PVC workers against 2.33 expected). No data were given for cancers of the lym phatic and hematopoietic systems, for cancer of the brain or thyroid, or for melanomas. The mortality reported by Masuda (33) for 305 Japanese vinyl chlo ride workers has presumably been subsumed in Na kamura's (37) later and larger study. Hazards of cancer The results of the four most useful studies are listed individually in tables 1 and 2. The overall results for all causes, liver cancer, and three broad groups of con ditions are shown in table 3, and those for 10 types, or classes, of cancer are presented in table 4. Data have not been reported for each type of cancer in each study, and the sources of the data are, therefore, specified separately for each type. Additional information ob tained from four other less informative studies (4, 7, 22, 49) is given in table 5 for seven types or classes of cancer. Table 3 shows that, apart from cancer of the li\er, the overall mortality is what would be anticipated for an industry without any major hazard of accident or disease. In particular the standardized mortality ratio (SMR) of 84 for diseases other than cancer is typical of the ratios that are commonly observed for groups of employed men. A low SMR of this order reflects the "healthy worker effect," which results from the selection process that inevitably excludes some of the least healthy members of the population from indus trial employment. This effect does not, however, nor mally affect the mortality from cancer beyond that ob served in the first few years after the start of employ ment, and an SMR of 102 for cancers other than can cer of the liver is compatible both with the absence of hazard and with SMR values of 84 for other diseases and 77 for accidents, poisonings, and violence. Angiosarcoma. Death certificates are an unreliable source of information about the histology of cancers that cause death, but there is no reason to suppose that the excess mortality attributed to liver cancer (or, in the US series, liver and gallbladder cancer) is not entire ly accounted for by the known hazard of angiosar coma. Fifteen of the 37 deaths2 attributed to cancers of the liver and gallbladder in the US series are know n to have been due to angiosarcoma of the liver (14). In the UK series, seven of the 11 deaths attributed to liver cancer, not specified as secondary, were known to be angiosarcomas, and they all occurred in auto clave workers against 0.38 expected liver cancers of all types (P< 10-5) (25). In the Canadian series, histo logical review showed that seven of the eight so-called liver cancers were angiosarcomas (one had been de scribed as an angiosarcoma on the death certificate, two as hepatomas, and five as unspecified liver can cers). One so-called liver cancer death was found to have been due to cancer of the sigmoid colon, while one death attributed to angiosarcoma of the perito- Increased to 39 in table I to take account of the 97 extra deaths from an unknown cause. 67 ASI 00005602 Table 1. Observed and expected numbers of deaths from different cancers reported in the (our principal studies (4,14, 25. 46). (O = observed number of deaths, E = expected number of deaths) Type or class of cancer United States 0* E United Kingdom OE Canada OE Italy OE Buccal cavity and pharynx Esophagus Stomach Large intestine Rectum Liver Liver and gallbladder Pancreas Other digestive Larynx Lung Other respiratory Bone Skin (nonmelanoma Melanoma Prostate Testis Bladder Kidney Other and unspecified urinary Brain Eye and central nervous system Thyroid Lympho- and reticulosarcoma Hodgkin's disease Leukemia Multiple myeloma 1 Other lymphatic } Other All cancers 13 11.55 7 8.07 11 16.01 21 28.79 39 5.77 17 18.40 118 115.87 5 6.38 2 1.81 6 7.36 16 15.20 5 8.46 12 9.06 25 12.76 12 7.98 3 5.45 14 13.94 46 40.50 383 341.73 4 3.58 6 14.34 26 23.91 9 13.94 11 10.49 11 1.94 7 9.88 4 2.21 81 92.12 0 8 6 !1 12 0.64 0.14 526 5.78 2 1 74 12 9.59 2 1.38 14 8.00 3 4.10 3 4.29 4 6.18 2 0.43 4 2 35 3 2.50 7 5.16 2 2.35 18 8.12 235 228 60 ] I1[ 1 J 11! v 1i j .i J 2 20 1*33 1.67 0.95 16.37 1 0.8 1 0.6 3 3.0 0 1.2 1 0.6 0 07 1 09 12 6,1 0 02 1 0,7 1 0.7 0 04 0 07 9 45 30 21,1 * Observed deaths multiplied by 1.0674 and rounded off to the nearest integer to allow for deaths without discovered cause. Table 2. Numbers of deaths from nonmalignant and all causes reported in the four principal studies (4,14,25,46). (O - observed number of deaths, E = expected number of deaths) Cause of death United States O* E United Kingdom OE Canada OE Benign and other unspecified tumors Cerebrovascular disease Ischemic heart disease Other circulatory disease Bronchitis b Pneumonia Other respiratory disease Cirrhosis of the liver Other digestive disease Disease of the genitourinary system Other diseases Suicide Accidents and other violence All nonmalignant causes All causes 4 5.08 75 91 93 521 597.73 157 123,55 44 22.83 16 31-94 15 32.84 37 56.06 27 3952 11 20.41 67 115.68 49 52.78 130 173.19 1 153 1 363.54 1 536 1 705.27 276 105 36 } 40 5 43 } 40 545 780 288 141 644, i) 5\ I 76 50 665 894 25 31.67 | 1 6 3.21 } 3.85 2 5.40 10.58 39 54.76 59 71.07 See footnote to table 1. * Emphysema in data from the United States c Includes two cases certified as cirrhosis of the liver which proved to be angiosarcoma of the liver. 6 Includes one case with cause unknown. Italy OE 0 0.5 19 27 4 3 5.0 4 52 3 27 2 7.0 1 0.9 4 7,7 36 564 66 77.5 neum was found to have been due to angiosarcoma of the liver (46). In the Italian study, further evidence revealed that three further deaths should have been at tributed to cancer of the liver (for a total of four), but only one of the four was described as an angiosarcoma (4). 68 ASI 00005603 1 Italy OE 1 0.8 1 0.6 3 3.0 0 1.2 1 06 6 0,7 1 0.9 12 6.1 0 0.2 1 07 1 07 0 04 0 07 9 4,5 30 21 1 h^^discovered cause. 14.25,46). (0 = observed Italy 0E 0 } 19 13 4 3 2 1 4 36 66 05 27 4 5.0 5.2 2.7 70 0.9 7.7 56.4 77,5 a total of four), but an angiosarcoma i Table 3. Mortality from cancer of the liver and other causes among vinyl chloride workers in 49 plants in the four principal studies combined (4,14, 25, 46). (0 = observed number of deaths, E = expected number of deaths, SMR = standardized mortality ratio) Cause of death 0E Cancer of the liver* Cancer of other sites Other diseases Accidents, poisonings, and violence All causes 59 609 1 547 226 2 441 8.45 599.35 1 844.89 295.15 2 747.84 * Including cancers of the gallbladder in the series from the United States. SMR 698 102 84 77 89 Table 4. Mortality from various cancers among vinyl chloride workers in 49 plants in the four principal studies combined (0 = observed number of deaths, = exposed number of deaths, SMR = standardized mortality ratio) Type or class of cancer 0E SMR Source o' information* Mouth and pharynx Digestive system (other than liver) Respiratory system Lung Genitourinary system Melanoma Brain Thyroid Lymphatic and hematopoietic system Other All other than of the liver 18 16.57 125 154.59 223 229.36 211 214 09 70 62.81 2 1.94 29 19.54 2 0.43 57 50.87 83 63.24 609 599,35 109 1, 2, 3. 4 81 1,2, 3.4 97 1, 2, 3. 4 99 1, 2, 4 111 1, 2, 3. 4 2, 4 148 1, 2, 3 2 112 1, 2, 3. 4 131 1. 2, 3, 4 102 1, 2, 3. 4 * 1 = United States study <14), 2 = United Kingdom study (25). 3 = Canadian study (46), and 4 = Italian study (4) Table 5. Mortality* from various cancers among vinyl chloride workers. Supplementary evidence (4, 7, 22. 49). (O-observed number of deaths, E = expected number of deaths, SMR = standardized mortality ratio) Type or class of cancer Federal Republic of Germany (11 plants) OE Norway (1 plant) OE Digestive system (excluding the liver) Lung Melanoma Brain Thyroid Lymphatic and hemato poietic system Other (excluding the liver) All excluding the liver 35.0 31.8 23 5 24.6 2.1 1.3 16.5 77 107 24.3 87.8 89.7 3 1.44b 5 2.84 4 0.79 2 0-16 8 14.93 22 20.16 * Incidence and cases in the Norwegian study c Cancer of the intestine only r- Cancer of the lung and brain only. Sweden (1 plant) OE 3 1.78 2 0.33 5= 2 11* Italy (1 plant) 0E 1 1.2 0 1.5 0 0.1 0 0.7 4 20 5 57 Four countries combined (14 plants) 0 E SMR 39 0 31.5 4.0 4.1 2.0 16.5 22.7 119.8 34.44 30.72 0.89 1.63 0 16 8.4 41.43 117.67 113 103 196 55 102 Further evidence that the excess mortality from liver cancer (or liver and gallbladder cancer in the US series) can be attributed principally if not wholly to the known hazard of angiosarcoma is obtained in a comparison of the excess deaths with the numbers of deaths from angiosarcomas recorded in the Register of Liver Angio sarcoma Cases (maintained on behalf of the Associa tion of Plastics Manufacturers in Europe by the Im perial Chemical Industry PLC) before the end of the follow-up period (Bennett, unpublished). Fifty-one ex cess liver cancers are recorded in the combined data, and 49 angiosarcoma are recorded in the Register for 69 ASI 00005604 Table 6. Mortality from lung cancer In the series from the United States (US) (14) and the United Kingdom (UK) (25) by characteristics relevant to an occupational hazard. (0 = ob served number ot deaths, E = expected number ot deaths, SMR = standardized mortality ratio) Data characteristic* Category 1 Ob E SMR Category 2 Ob E SMR Observed 20 years or more after first employment (1), others (2) Employed 10 years or more in the US (1), others in the US (2) Employed belore 1956 in the UK (1), others in the UK (2) Ever employed as autoclave worker in the UK (1), others in the UK (2) 114 113.96 100 55 52.45 105 52 51.39 101 16 17.08 94 85 93.83 91 63 63.44 99 29 40.50 72 65 74 82 87 * The numbers in parentheses designate the category. b See footnote to table 1 for observed deaths in the US. the relevant periods for the three countries and the two Italian plants3 that are covered by the survey. None of the 120 cases yet recorded in the Register were in men who were first exposed after 1969, and none of the 45 men affected in North America were first exposed after 1964, All may, therefore, have been exposed to concentrations of several hundred parts per million, and many may have been exposed to concen trations appreciably higher (unpublished report by Barr to Air Products and Chemicals Inc in 1981). Lung cancer. The idea that exposure to vinyl chloride might cause cancer of the lung was suggested by Monson et al in 1974 (36), when they noted 13 cases against an expected number of 7.9 in a study of proportional mortality. The combined data shown in table 4 do not provide any support for the hypothesis, either for res piratory cancer as a whole (SMR 97) or for the speci fied data for lung cancer in the US, the UK, and Italy (SMR 99). There are, however, consistently higher risks in the subgroups of men in the US and UK series, in which occupational hazards would be more likely to i Tweniy-nine were registered as occurring in the United Slates against an excess of 33; only 20, however, were iden tifiable in both series. Inquiry has, as yet, failed to reveal information about the histology of the remaining 13 in the cohort study and the origin of the nine extra deaths in the register. Nine deaths were registered as occurring in the United Kingdom against an excess of nine, but one of the registered cases was certified as due to a benign hemangioma and not related to the liver (code 227 in the eighth revision of the International Classification of Diseases). Ten cases were registered as occurring in Canada against an excess of eight; two were recorded as being in men who had been employed for five years, and it is possible that the actual duration had been slightly less than five years with con sequent exclusion from the Canadian cohort. One case was registered as occurring in one of the two Italian plants against an excess of less than one. be seen than in other groups. This circumstance is illustrated by table 6, which shows that the SMR values are slightly higher for men observed 20 years or more after first exposure than for men observed earlier, for men employed before 1956 in the UK than for men first employed after 1955 (when exposure levels are believed to have been lower), for men employed for longer than for shorter periods in the US, and for autoclave workers in the UK (among whom the angio sarcoma cases have mostly occurred) than for other workers. The differences are all small or very small. They are, however, all in the same direction, and the probability that the rates should all be higher in the groups in which an occupational hazard is more like ly to be seen in each of the four pairs of groups is I in 16. Additional information from other sources is given in table 5. A total of 30 deaths (or cases) was observed, and this figure increases to 31.5 when allowance is made for the number of deaths due to unidentified causes in the German study (SMR becoming 103). In the German study the SMR was higher for the men who had been exposed for 10 years or more than for those who had been exposed for shorter periods (111 against 79), and, in the Norwegian study, four of the five cases observed occurred in men whose occupations were regarded as involving high exposure against 1.82 of the 2.84 expected. Both the German and the Swed ish studies derived the expected numbers of deaths from national mortality rates for a single year towards the end of the study period. The expected numbers of deaths are likely, therefore, to have been overestimated and the SMR values correspondingly underestimated as the mortality from lung cancer had been rising throughout the period of observation. Brain cancer. The idea that vinyl chloride might cause brain cancer was also suggested by Monson et al (36) when they reported five cases against 1.2 expected. The combined data that are shown in table 4 provide some support for this hypothesis. The cases of Monson et at (36) were, however, observed in US workers and must be presumed to be included in the total reported by Environmental Health Associates (14); therefore they will have contributed a substantial proportion of the total in table 4. As a test of the hypothesis the data of Monson et al ought, therefore, to be subiracted from those in the table. Their investigation was not a cohort study, and their expected deaths do not cor respond exactly to those in table 4. If, however, the observed and the expected cases are both subtracted from the totals, twenty-four observed deaths remain against approximately 18.3 expected, a difference which might easily occur by chance (P one-tailed = 0.1).4 1 If the study of Waxweiler et al (50) is regarded as the origin of the hypothesis, 26 deaths are left against 18.94 expected (P one-tailed = 0.07), 70 ASI 00005605  -..Jhis circumstance is the SMR values 20 years or more en observed earlier, for the UK than for men ien exposure levels are for men employed for ds in the US, and for mong whom the angiocurred) than for other .11 smalt or very small, ame direction, and the ild all be higher in the al hazard is more like-ir pairs of groups is 1 other sources is given ir cases) was observed, .5 when allowance is is due to unidentified vIR becoming 103). In as higher for the men ears or more than for r shorter periods (111 ian study, four of the ten whose occupations exposure against 1.82 jerman and the Swedd numbers of deaths jAtingle year towards ^^Bected numbers of i^been overestimated Jingly underestimated ncer had been rising `vation. I chloride might cause by Monson ei al (36) linst 1.2 expected. The i table 4 provide some e cases of Monson et d in US workers and J in the total reported >ciates (14); therefore stantial proportion of te hypothesis the data >re, to be subtracted nvestigation was not ed deaths do not core 4. If, however, the > are both subtracted served deaths remain peered, a difference ance (P one-tailed = is regarded as the origin i against 18.94 expected Additional information from two other sources is given in table 5. The small excess reported provides little further evidence of an occupational hazard, as one of the two deaths observed in the Swedish study ccurred in a young man who had been employed for less than a year when the diagnosis was made, while the excess death rate for brain cancer observed in the German study was less than that observed among chemical workers not exposed to vinyl chloride (2.9 deaths after allowance for deaths from unknown causes against 1.6 expected) and among workers in the PVC fabrication industry (5.9 deaths after allowance for deaths from unknown causes against 1.1 expected). 'ancers of lymphatic and hematopoietic tissues. The ;Jea that vinyl chloride might cause cancer of the lym phatic and hematopoietic tissues -- more specifically the lymphatic tissue -- was suggested by Tabershaw & Gaffey (45) and by Waxweiler et al (50) in two cohort studies, when they found, respectively, five deaths from lymphomas in the most heavily exposed workers against 2.54 expected and four deaths from cancers of the lymphatic and hematopoietic tissues against 2.5 ex pected. These small excesses might have been ignored if the laboratory findings had not been interpreted as suggesting that lymphomas were produced experimen tally in animals exposed to vinyl chloride by inhala tion (29). The idea that similar exposure might also cause lymphomas in humans, therefore, merits seri ous consideration. The data from the four principal studies that are summarized in table 4 provide little support for the hypothesis when all cancers of the lymphatic and hematopoietic tissues are considered to gether (57 deaths against 50.87 expected, SMR 112) and very little more is obtained from the separate data for cancers of the lymphatic system (Tabershaw & Gaffey's definition of ICD list numbers, eighth revision, 200--203 and 205 being used) that are shown in table 1 (35 deaths against 29.40 expected). The position is, moreover, hardly altered if the data in Tabershaw & Gaffey's initial report are subtracted (29 deaths against 23.36 expected, SMR 124), Little additional information is provided by the results of the German study (49). (See table 5.) This study obtained an SMR of 214 for exposed workers (based on 15 observed deaths, increased to 16.5 when allowance is made for the number of deaths from un known causes) against SMR values of 77 and 34 for an unexposed group of chemical workers and a group of PVC fabricators. It showed that the excess of the exposed workers was present only for men who had been exposed for more than one year and that this ex cess was most marked for men who had been exposed for five years or more (10.7 deaths after allowance for the number of deaths from unknown causes against 4.0 expected, SMR 268, P one-tailed <0.01). Melanoma. An excess of melanoma was reported for Norwegian workers by Heldaas et al (22), who raised the possibility that vinyl chloride might have produced the disease. Four cases were observed when 0.79 were expected, and three of the four were in men whose oc cupations involved the highest exposures (against 0.51 expected). At the time of the writing of their report, one further case had been detected with onset three years after the closure of the study. Subsequent studies in other countries have, so far, reported only two deaths against 2.0 expected. (See tables 1 and 5.) Thyroid cancer. An excess of thyroid cancer was also reported in the Norwegian study (22), in which two cases were observed against 0.16 expected. The inves tigators were not aware of any other studies indicating an excess of this type of cancer, and they drew no con clusion from their observation. Two of the three major studies that have been reported since the Norwegian observation was made gave no data for thyroid can cer; the third reported two deaths against 0.43 ex pected. (See table 1.) One death from thyroid cancer, it may be noted, was reported in the US by Monson et al (36). Cancers of the digestive tract. Suggestions that vinyl chloride might cause cancers of the digestive tract in general have sometimes been made, but they have not taken adequate account of the contribution of cancers of the liver to the total number of cancers of the digestive system, particularly when it is borne in mind that some liver cancers are likely to be misdiagnosed as cancers of other organs. The combined data from the four principal studies shown in table 4 weigh heavily against the idea that any such effect has been produced. Other cancers. One of the remaining types, or classes, of cancer listed in table 4 shows a statistically signifi cant excess, namely, the heterogeneous group of "other cancers" (83 observed deaths against 65.24 ex pected, P two-sided <0.05). This excess is only mar ginally significant and may be a chance observation. The most likely explanation is, however, that a few angiosarcomas of the liver were not recognized and were diagnosed as secondary liver cancer or carci nomatosis, site unknown, the number of deaths in this category therefore being increased. Hazards of nonmalignant disease No previous study has suggested that any nonmalig nant cause of death other than cirrhosis of the liver w'ould be likely to be increased as a result of exposure to vinyl chloride, and cirrhosis of the liver is presumed to be increased only because of the liver changes that were observed as part of the "vinyl chloride illness" (24, 31, 33). Two other possibilities have, however, been raised, namely, the production of nonmalignant respiratory disease, because of the changes in lung function and radiographic appearances that have been 71 ! Table 7. Mortality 'rom selected nonmalignant causes and all causes in the tour principal studies combined. (0 = observed number of deaths, E = expected number of deaths, SMR = standardized mortality ratio) Type of disease 0 E SMR Source of information* Bronchitis, emphysema11 80 66.83 120 1. 2 Other respiratory disease 71 125.78 56 1. 2 All respiratory disease Ischemic heart disease Other circulatory diseasec All Circulatory disease6 Cirrhosis of the liver Other disease 160 797 252 1 103 46 238 200.82 885.73 264.55 1 209.35 66.26 36807 All nonmalignant disease All external causes 1 547 1 844.50 226 29515 All nonmalignant causes All causes 1 773 2139 65 2 441 2 747.85 80 1. 2. 3, 4 90 1.2 95 1,2 91 1. 2. 3, 4 69 1, 2, 4 65 1. 2, 3, 4 84 1. 2. 3, 4 77 1, 2. 3, 4 85 1, 2, 3, 4 69 1, 2. 3, 4 * 1 = United Stales study (14), 2 = United Kingdom study (25), 3 = Canadian study (46), and 4 = Italian study (4). b Bronchitis in the United Kingdom study, emphysema in the United States study. c Includes cerebrovascular disease in the United Kingdom and Italian studies. Table 8. Mortality from chronic obstructive lung disease* in the series from the United States (US) (14) and the United King dom (UK) (25) by characteristics relevant to an occupational hazard. (O = observed number of deaths, E = expected number of deaths. SMR = standardized mortality ratio) characteristic11 Category i O E SMR Observed 20 years or more after first employment in the US (1). others in the US (2) Employed 10 years or more in the US (1), others in the US (2) Employed before 1956 in the UK (1), others In the UK (2) Ever employed as an autoclave worker in the UK (1), others in the UK (2ic 30 16 26 3 15.8 190 10.9 147 30.17 86 6.55 46 Category 2 O E SMR 11 7.0 157 25 12.0 208 10 13.60 74 33 37.22 89 Described as emphysema in the US study and as bronchitis in the United Kingdom study. b The numbers in parentheses designate the category. c Men ever employed as a bagger or drier, occupations which would have caused the greatest occupational exposure to polyvinyl chloride dust, experienced one death from bron chitis against 4.98 expected. recorded for men exposed to PVC dust (2, 26,27,44), and acute cardiac death, from analogy with the effect of other halogenated hydrocarbons (25) and the ob servation of an increased mortality from myocardial infarction in the few years following the cessation of exposure in the Swedish PVC processing industry (35). Relevant figures for the numbers of deaths from these and other nonmalignant causes that are obtainable from the four principal studies were given in table 2, and they have been summarized in table 7. Cirrhosis of the liver. Three of the four principal studies gave separate figures for cirrhosis of the liver, none of which showed an increased mortality (table 2); in combination they gave an SMR of 69 based on 46 deaths. The fourth study, which did not give sepa rate data for cirrhosis of the liver, reported four deaths from all diseases of the digestive system combined against 3.85 expected and noted that the four included two that were certified as due to cirrhosis of the liver, but actually due to angiosarcoma (46). In the two sup plementary studies in which data were given for this disease, the SMR was 82 in one, based on 15.1 deaths after allowance for the number with unknown causes (49), and 133 in the other, based on seven deaths (37). Nonmalignant respiratory disease. The data for non malignant respiratory disease are confusing in that the total SMR from the combined data for the four prin cipal studies is 80 and is the sort of figure that is com monly found in healthy industrial populations, yet the US study recorded a substantially increased mortality from emphysema (41 deaths and an SMR of 180 be fore any allowance was made for deaths from un known causes). No such excess was found in the UK, where 36 deaths from bronchitis gave an SMR of 82. International comparisons of chronic nonmalignant respiratory disease are complicated by the usage of dif ferent terms to describe what it is now agreed is best called chronic obstructive lung (or pulmonary) disease, but which in the past tended to be called emphysema in the US and chronic bronchitis in the United King dom. It must, therefore, be presumed that the two categories of "emphysema" and "bronchitis" used re spectively in the two large national studies were meant to describe the same thing. One must assume, there fore, that the experiences in the two countries were very different, despite the fact that both related to cohorts that had very similar experiences of angiosarcoma of the liver and so, presumably, fairly similar exposures to vinyl chloride. Separate figures are shown in table 8, where avail able, for the mortality observed among men with dif ferent durations and intensities of exposure. Unlike the data for cancer of the lung that were show n in table 6, they provide no consistent evidence of a greater risk in the groups in which an occupational hazard would 72 0000567 dust (2, 26, 27,44), logy with the effect rbons (25) and the obtality from myocardial lowing the cessation of Processing industry (35). -rs of deaths from these >es that are obtainable < were given in table 2, ed in table 7. of the four principal >r cirrhosis of the liver, eased mortality (table n SMR of 69 based on hich did not give sepav, reported four deaths -tive system combined J that the four included 0 cirrhosis of the liver, ta (46). In the two supna were given for this based on 15.1 deaths ' with unknown causes d on seven deaths (37). The data for non^Bmfusing in that the ^Hta for the four prin- t of figure that is comial populations, yet the lly increased mortality id an SMR of 180 be- for deaths from unwas found in the UK, is gave an SMR of 82. chronic nonmalignant ted by the usage of dift is now agreed is best or pulmonary) disease, be called emphysema is in the United Kingresumed that the two 1 "bronchitis" used re nal studies were meant e must assume, therewo countries were very poth related to cohorts es of angiosarcoma of tirly similar exposures n table 8, where avail1 among men with diff exposure. Unlike the t were shown in table dence of a greater risk nal hazard would 1 be expected to be concentrated. The authors of the En vironmental Health Associates report (14) were unable to give any explanation for the increased mortality iom emphysema, and they point out that it could hardly be due to excess cigarette smoking, as there was no overall excess for cancer of the lung. It is striking, however, that the excess is more than compensated for by deficiencies in the other categories of nonmalignant respiratory disease (pneumonia 15 deaths,5 SMR 47.0; other respiratory disease 14 deaths, SMR 42,6), and the question arises whether the emphysema excess could be a classificatory artifact. Environmental Health Associates (14) list all the 41 deaths which show that they were coded under ICD number 527 which, in the out-of-date seventh revision that was used for the coding of all deaths in the study, w-as the code for "other respiratory disease not otherwise classified" and included emphysema. Under that revision, however, emphysema that was associated with bronchitis should be classified with bronchitis under ICD numbers 500 to 502, and the possibility may be considered that some of the emphysema deaths should have been classified in some category of respiratory disease other than ICD number 527. If this were the situation, it could account for both the excess mortality from emphysema and the grossly deficient mortality from other nonmalignant respiratory diseases. No excess mortality from "bronchitis, emphysema, and asthma" was observed in the German study (SMR 44 with 6.3 deaths observed after allowance for the number of deaths from an unknown cause) (49). Cardiovascular disease. Data for ischemic (or arterio sclerotic) heart disease (which may be presumed to in clude the vast majority of all deaths certified as due to acute cardiac disease) were given only by the two big national studies, and they provide no evidence of an increased mortality. The SMR values of 90 for this group of diseases and of 91 for all cardiovascular dis ease recorded in the four principal studies are typical of the SMR values of healthy industrial populations, and there is no suggestion of any occupational hazard in the subsidiary analyses provided by the two national studies. In particular, there is no evidence of an in creased mortality within one month of leaving em ployment in the UK study either for all workers (52 deaths, SMR 61) or for the most heavily exposed auto clave workers (9 deaths, SMR 42). A slight increase in ischemic heart disease mortal ity was recorded for the exposed workers in the Ger man study (49), but it was less than that recorded for* * The deaths attributed to different groups of respiratory dis eases and the corresponding SMR values that are cited in this section for the US study are as given by the Environ mental Health Associates (14) and have not been adjusted to account for the number of deaths from an unknown cause. To take account of these deaths, the observed deaths and SMR values can both be multiplied by 1.0674. the unexposed chemical workers and the PVC fabri cators (SMR values of 127,131, and 158 based on 97.2, 126.7, and 109.7 deaths, respectively, after allowance for the number of deaths from unknown causes). Discussion The information that has now been obtained about the long-term health of men occupationally exposed to vinyl chloride is massive and compares favorably with that available for any other occupational group. Two facts are outstanding. First, the men have experienced a specific hazard of a type of cancer that is normally extremely rare, namely, angiosarcoma of the liver. The rarity of this disease under other conditions made the detection of the hazard easy; but the long latency period before the disease appears after first exposure (almost always more than 10 years and usually more than 15 years) meant that a large number of men had been exposed before the hazard was detected and that it will still be many years before the extent of the pro tection provided by the reduction in exposure in the 1960s and that of the further reduction that followed the recognition of the hazard in 1974 are known. There is, unfortunately, no effective treatment for the dis ease, and the number of cases is reflected in the num ber of deaths. Some 50 deaths have occurred among the 16 740 men who were followed in the four prin cipal studies that have been reviewed in this report, so that approximately 1 in 335 men have been affected, 2 % of the deaths having been due to this one cause. Eventually many more men must be expected to de velop the disease. One estimate (38) suggests that the total may be increased 10 times, but a more realistic estimate is two to three times (19). The second outstanding observation is that the mor tality of the exposed men, other than that due to angiosarcoma of the liver, is typical of the normally healthy industrial worker -- that is not to say that no other hazard exists, but that the effect of any other hazard is small. The massive data that are now available provide no reason for thinking that any hazard other than one of cancer has been overlooked. It is, however, still dif ficult to decide w hether vinyl chloride produces a risk of developing cancer other than angiosarcoma of the liver which might be small compared to the risks produced by nonoccupational causes, but yet abso lutely almost as large as the risk of developing the nor mally very rare angiosarcoma. One of the many hazards suggested can be dismissed, as there is no evidence to support it, namely, that of vinyl chloride as a cause of any cancer of the diges tive tract other than angiosarcoma of the liver. Two hazards (of melanoma and cancer of the thyroid) have been suggested only very recently, and few of the avail able studies have provided information about them. There is no good theoretical reason or laboratory evi dence to suggest that either should be produced by 73 vinyl chloride, and, in light of present evidence, the simplest explanation is that the reported excesses are the chance effects that must be expected when many different types of cancer are studied in several different populations. So far as melanoma is concerned, it has to be remembered that the disease has become much more common in recent years in Scandinavia (where the excess was reported) due, it is believed, to the popularity of sunbathing and the increased oppor tunities for Scandinavians to travel to the warmer parts of Southern Europe and North Africa. The extent to which this change may have affected the observation in Norway needs to be examined. Two other hazards (of lymphoma and brain cancer) were suggested by the early results of some of the US studies. That vinyl chloride might produce a hazard of lymphoma was initially supported by the prelimi nary results of animal studies, but the complete results of the many investigations that have been undertaken (see reference 29) do not suggest that lymphoma or any other cancer of the hematopoietic system is liable to be produced. There is, however, some evidence that brain tumors can be produced in rats (29). The hypoth eses that lymphomas and brain cancers might be produced by vinyl chloride have been supported by the observation that both these types of cancer have caused death more often than might be expected from national mortality rates, but the excesses observed in the com bined data from the four principal studies in this re view are small and not statistically significant, and the hypotheses remain unproved. The small excess of brain cancer is particularly difficult to evaluate, as mortal ity rates from this disease have changed rapidly over time as methods of diagnosis have improved and the suspicion of an occupational hazard (which was raised in 1975) could have influenced the findings. What ex cess has occurred has been limited to the US and Ger many, and the German findings carry little weight, as the excess was found in each of the three occupational groups studied, irrespective of the chemicals to which they were exposed. The supplementary data from the German study showing an increased mortality from lymphatic and hematopoietic cancers are more impres sive, particularly as the excess was the most marked for men who had been employed for at least five years. In these circumstances, judgment must still be sus pended until the data for each study are analyzed for each specific type of cancer, by intensity and duration of exposure, and by time since exposure began. There remains the suggestion that vinyl chloride might cause lung cancer. At first sight, this possibil ity is ruled out by the SMR of 97 for the combined data for respiratory cancer for the four principal studies. Lung cancer is, however, normally so com mon (accounting for about 8 Vo of the expected deaths) that an increase in mortality that was half as impor tant (numerically) as the increase in mortality from angiosarcoma of the liver might easily be overlooked (95 Vo confidence limits of the SMR 85--112). The in cidence of the disease varies moreover within a coun try, and there must be doubts as to whether the na tional experience provides a suitable reference for men employed in plants that are not evenly distributed about the country. In these circumstances one cannot exclude an occupational hazard unless it can be shown that the mortality of the exposed men is independent of the factors that might be expected to influence it if some of it were occupational in origin, namely, the intensity and duration of exposure and the time since exposure began. It is not possible to examine these relationships in detail, as the reports do not provide all the necessary information. Such information as they do provide, which was summarized in table 6, supports the idea that exposure to vinyl chloride involves a small hazard of lung cancer. Taken in conjunction with the knowledge that lung tumors have been produced in sev eral species of animals exposed to vinyl chloride by in halation (29), it would seem that a small hazard of lung cancer probably did occur. The evidence is not, how ever, strong enough to conclude that it definitely did. If it did, the hazard was evident only for men who had been employed for many years at a time when expo sures of several hundred parts per million or more were common, and any persisting risk can be only minute and incapable of detection. The questions that have been left unanswered by this discussion might well be answered definitely if (i) all the exposed men could be followed to (say) the end of 1984, (ii) the investigators could present their data in comparable ways, taking account of duration of employment and time since first employment and presenting data separately for men first employed be fore (say) 1965 and between 1965 and 1974, and (iii) estimates could be made of the effect of correcting the results for each group of employees for the locality in which they lived and worked. Hazards to the general population As vinyl chloride has been proved to cause cancer in man and is a mutagen in laboratory experiments, it must be presumed that even the minute doses that escaped into the general environment from production plants or (in the early days of manufacture) from PVC materials will have caused some risk of cancer to the general public. These risks must, however, have been very small, as air concentrations of vinyl chloride, even w'ithin a kilometer of plants handling vinyl chloride, used to be (in or around 1975) of the order of 10 to 40ppb(l, 3, 15), and this level is about one-ten thou sandth of the concentration that has caused an occu pational hazard. It is obvious, therefore, that it would be impossible to detect the risk of any cancer that might be produced by vinyl chloride other than a risk of angiosarcoma of the liver, as it has proved so diffi cult to detect any other risk among men who were ex posed occupationally. The position with regard to an- 74 ASI 00005609 I >ver within a coun}o whether the na- e reference for men tot evenly distributed umstances one cannot unless it can be shown ;d men is independent *pected to influence it in origin, namely, the ure and the time since ible to examine these eports do not provide tch information as they ted in table 6, supports iloride involves a small i conjunction with the . been produced in sevo vinyl chloride by ina small hazard of lung evidence is not, howr that it definitely did. only for men who had at a time when expor million or more were >k can be only minute left unanswered by this red definitely if (i) all owed to (say) the end ^present their data Ini of duration of irst employment and nen first employed be*65 and 1974, and (iii) -ffeet of correcting the yees for the locality in ulation ved to cause cancer in ratory experiments, it he minute doses that ment from production anufacture) from PVC e risk of cancer to the i, however, have been of vinyl chloride, even mdling vinyl chloride, i of the order of 10 to is about one-ten thoutt has caused an occuherefore, that it would f any cancer that might other than a risk of t has proved so diffiong men who were exu^^with regard to an giosarcoma of the liver is different. This disease is nor mally so rare that, in the absence of specific exposure to one of the known causes (vinyl chloride, thorium dioxide, and arsenic in pesticides and medicines), the annual incidence is on the order of 1--2 10~7 (5, 8).6 In these circumstances the discovery of even one case in a man living close to a factory in which vinyl chlo ride was used in the days before exposure was tightly controlled may be regarded as presumptive evidence of the effect of environmental pollution. Several surveys have been undertaken to determine whether any such cases have occurred. Saric et al (43) and Blinder & Pershagen (13) sought for cases in the vicinity of plants handling vinyl chloride in Yugoslavia and Sweden and found none. In Holland Dalderup et al (11) found eight confirmed cases not attributable to thorotrast or arsenic and could trace "no contact with vinyl chloride," but they made no specific men tion of the patient's place of residence. Baxter et al (3) found 14 confirmed cases in Great Britain over a 12-year period, one of which was in a man who had lived half a kilometer from a PVC manufacturing plant, and, in New York State over an 18-year period, Brady et al (5) found 19 cases that could not be at tributed to any known cause, five of which W'ere in peo ple living within a mile of plants manufacturing or using vinyl chloride. The overall incidence rates in these last two studies were not unduly high, but the occur rence of as many as six cases among people living so close to manufacturing plants is surprising. Brady and his colleagues, moreover, compared their series of pa tients with matched referents and found that none of the referents lived equally close to a plant. Two of these six neighborhood cases (one in England and one in New York State) cannot be attributed to environmental pollution with vinyl chloride, as the men who devel oped the disease had lived near the plants for six and eight years, respectively, before developing the disease, and this period is too short to allow for the necessary latency. The other four cases, however, all occurred after 15 or more years of local residence, and the dis covery of these cases strongly suggests that pollution of the environment around plants manufacturing VCM or PVC may have caused a minute hazard to the gen eral public. Current concentrations around plants handling vinyl chloride are certainly much lower than those reported previously by the US Environmental Protection Agency. Recent British measurements made within a few hundred meters of the VCM areas have given aver age values below the daily limit of detection (5 ppb) for three of five plants, the readings at the two others being 20 ppb (100 m outside the boundary fence) and 88 ppb Oust inside it) (47), although substantially higher values were recorded on two occasions asso ciated with putting one plant into operation and with 11 The figure of 1.4 lO-8 cited by Heath et al (21) seems to have been a misprint for 1.4- I0~7. an accident at the other. According to any reasonable criterion the hazard to the general public (if there is any at all) must be negligible (42). No other hazard to the general population, other than a hazard of cancer, can reasonably be postulated. Summary This paper reviews (i) the possible effects of vinyl chloride on the personal health of men exposed by virtue of their occupation (other than the early effects of the very high concentrations to which men were ex posed when the industry was first developed -- uncon sciousness, cardiac arrhythmia, and the characteristic "vinyl chloride illness") and (ii) the carcinogenic effects that might conceivably be observed in the gen eral population as a result of the widespread distribu tion of vinyl chloride as a pollutant. The possibility that vinyl chloride might act as a teratogen or might cause mutations in the germ cells has not been exam ined, as the little evidence that has been adduced relating to such possible effects has been reviewed else where and the conclusion was reached that no such ef fects have been demonstrated. Many groups of workers exposed to vinyl chloride in the manufacture of VCM or PVC have been studied since the carcinogenic potential of vinyl chloride was first recognized. Some results have shown that occu pational exposure can cause angiosarcoma of the liver, and others have suggested that it may cause several other types of cancer as well. The actual situation can be determined only in an examination of all the evi dence, especially the combined results of those studies that include a substantial proportion of observations on men more than 25 years after their first exposure and cover a long enough period for more than 10 7o of the employees to have been expected to die. The results of four studies can be usefully combined for this purpose. They are two national studies, one from the US and the other from the UK, and two studies of employees in one plant in Canada and two plants in Italy. The results of other studies from the Federal Republic of Germany, Norway, Sweden, Italy, France, and Japan can be used only to provide sup plementary information. The many earlier reports of exposed workers in the US and the UK concern men covered more completely in the two recent national studies, and their results serve only as sources of hy potheses. Minor criticisms can be made of three of the four most useful studies. They do not seriously affect the value of the results, except that allowance has to be made for the failure to determine the cause of 6.3 Vo of the deaths recorded in the US study. Three of the studies use national rates to estimate the numbers of deaths that might have been expected to occur in the absence of any special occupational hazard, and the fourth (Canadian) uses rates for the province in which I the plant was situated. It must, therefore, be kept in mind that the rates used may not have been wholly ap propriate for the localities in which the plants were situated. This circumstance is potentially important for the US study, which covered workers in 37 plants, 22 of w'hich were situated in the southern part of the country. The other less informative studies are, for the most part, open to more serious criticism, and the \alue of each set of results needs to be assessed separately in relation to each disease. The combined results of the four principal studies show that the SMR values, reflecting the ratios between the numbers of deaths observed and those expected in the absence of an occupational hazard multiplied by 100, have been (i) 77 for accidents and other violence, (ii) 84 for diseases other than cancer, and (iii) 102 for cancers other than cancer of the liver. All these results are what might be anticipated for an industry devoid of any specific occupational hazard. The low ratio for diseases other than cancer reflects the "healthy worker effect," which results from the selection process that inevitably excludes some of the less healthy members of the population from industrial employment and is compatible with a higher ratio for cancer, as the mor tality from cancer is not normally subject to such an effect, apart from the first two or three years imme diately following the start of employment. The mortality from cancer of the liver was nearly seven times that expected. Most of the 51 excess deaths were known to be due to angiosarcoma, even though this diagnosis was not recorded on the death certifi cate. The excess corresponds closely with the 49 deaths due to angiosarcoma reported to the International Register of Angiosarcoma Cases as occurring in em ployees of the plants concerned during the periods under observation. All the men who developed the dis ease were likely to have been exposed to concentrations of vinyl chloride of several hundred parts per million or more. Three other types of cancer which have been sug gested to occur as a result of exposure to vinyl chlo ride are cancers of the lung, brain, and lymphatic and hematopoietic systems. The combined data for the mortality from respiratory cancer fail, at first sight, to support the hypothesis regarding lung cancer (SMR 97). Higher ratios for lung cancer have, however, been observed consistently in the subgroups in which the effect of an occupational hazard would be most like ly to be seen (that is, men employed for more than 10 years, exposed to higher than average concentrations, or observed more than 20 years after first exposure). In two of the supplementary studies it was also noted that the mortality from lung cancer was specifically increased among the most heavily exposed workers. The combined data show small excesses in the mor tality from cancers of the brain and of the lymphatic and hematopoietic systems. The excesses are, however, not statistically significant, and there is nothing to sug gest that they are occupational in origin. An excep tion is the observation of an increased mortality from cancers of the lymphatic and hematopoietic systems in the supplementary study from the Federal Republic of Germany. Two types of cancer were reported to be in excess in the Norwegian study, namely, thyroid cancer and melanoma. The significance of this finding is difficult to assess because very little information about these cancers has been provided by other studies. Suggestions that vinyl chloride might cause cancers of the digestive tract have failed to account for the contribution of angiosarcoma of the liver. When this disease is excluded, the mortality from digestive tract cancer decreases to below the average (SMR 82 for the four principal studies). A small excess mortality from the heterogeneous group of "other cancers" in the combined results of the four principal studies was statistically marginally significant (83 deaths against 65.25 expected, P<0.05). Some of the excess was likely to have been due to the misclassification of angiosarcomas as secondary can cers of the liver or carcinomatosis, site unknown. The following three nonmalignant causes of death have required special examination: cirrhosis of the liver because of damage to the liver in "vinyl chloride ill ness, " myocardial infarction (from analogy with the effect of other halogenated hydrocarbons and because of some observations from Swedish PVC fabricators), and nonmalignant respiratory disease because of changes in lung function and the radiographic appear ance of the lungs observed in men exposed to PVC dust. Far from being raised, the mortality from cir rhosis of the liver was less than expected in the three principal studies and in one of the two supplementary studies which gave separate figures for the disease (SMR values of 69, based on 46 deaths, and 82, based on 15 deaths), while in the other supplementary study the increase was trivial. Data for myocardial infarction have not been re ported separately. But myocardial infarction accounts for most of the deaths attributed to ischemic heart dis ease, and there is no evidence that either ischemic heart disease or cardiovascular disease as a whole was un duly common (SMR values of 90 and 92, respective ly) or related to occupational exposure. The data for the third category of nonmalignant dis ease (nonmalignant respiratory disease) are confusing, because the two large national studies give conflicting results. The combined data for the four principal studies show the low mortality that is commonly found in healthy industrial populations (SMR 80). This fig ure hides, however, an increased mortality from chronic obstructive lung disease (SMR 120), which in cludes emphysema and is due to a grossly increased mortality attributed to emphysema in the US study (SMR 193). The corresponding mortality in the British study, which was preferentially described as due to bronchitis, was less than expected (SMR 82), as was the mortality from pneumonia (SMR 50) and other res- 76 ASI 0000561 an increased mortality from * ^)[dHi h1 ematopoietic systems Wpm the Federal Republic ere reported to be in excess namely, thyroid cancer and ice of this finding is difficult tie information about these d by other studies. .hloride might cause cancers e failed to account for the oma of the liver. When this ortality from digestive tract the average (SMR 82 for the tv from the heterogeneous in the combined results of was statistically marginally st 65.25 expected, P<0.05). elv to have been due to the -arcomas as secondary canomatosis, site unknown, imalignant causes of death mation: cirrhosis of the liver liver in "vinyl chloride illion (from analogy with the I hydrocarbons and because : Swedish PVC fabricators), atory disease because of nd the radiographic appear- men exposed to PVC ^^Phe mortality from cir- tnan expected in the three e of the two supplementary ate figures for the disease in 46 deaths, and 82, based other supplementary study farction have not been recardial infarction accounts buted to ischemic heart disce that either ischemic heart disease as a whole was uns of 90 and 92, respectivesnal exposure, egory of nonmalignant dis tory disease) are confusing, nal studies give conflicting .ta for the four principal lity that is commonly found ations (SMR 80). This figincreased mortality from sease (SMR 120), which in due to a grossly increased iphysema in the US study iing mortality in the British itially described as due to xpected (SMR 82), as was lia (SMR 50) and other res piratory diseases (SMR 46) in the US study. There is no consistent evidence that the mortality from em physema or bronchitis was specifically occupational, and it seems possible that the reported excess in the ' 5 study was an artifact due to nosological difficul- v with the use of the seventh revision of the ICD. Review of the massive data now available on the long-term health of men occupationally exposed to vinyl chloride leads to two dear conclusions. First the men have experienced a specific hazard of the normally extremely rare angiosarcoma of the liver. Approxi mately 1 in 335 of the men exposed in the 49 plants studied died of the disease, and approximately 2 % of the observed deaths were attributed to it. In the course of time the numbers of cases of angiosarcoma must expected to increase two to three times. Second, .,e mortality from all other causes has been typical of that of normally healthy industrial workers. If any hazard has existed, its effect has been small. The data provide no reason to think that any haz ard other than one of cancer has been overlooked. It is, however, still difficult to decide whether vinyl chlo ride produces small risks of cancer, compared to those due to nonoccupational causes, at sites other than the liver, and, if so, whether, in total, these risks might cause almost as many deaths as angiosarcoma of the liver. There is too little evidence either to confirm or refute the suggestion that vinyl chloride might cause mela noma or cancers of the thyroid, brain, and lymphatic and hematopoietic systems. None of the small excesses that have been recorded point specifically to an occu pational hazard, apart from that attributable to can cers of the lymphatic and hematopoietic systems in the German study reviewed, and most are likely to be the son of chance effect that is certain to be observed when many types of cancer are examined in many different studies. The lack of any increased mortality from lung can cer in the combined results of the four principal studies reviewed does not exclude the possibility that there may have been a small occupational hazard of developing the disease, as geographic variations in the incidence of the disease throw doubt on the validity of using national rates for estimating the expected numbers of deaths. The greater mortality in groups of w orkers w ho would be more likely to show- an occupational hazard than other groups suggests that a small hazard may have existed. The evidence is, however, weak, and the existence of a hazard has not been proved. Clearer answers to some of the questions that have been posed in this review might be obtained if the various groups of investigators could present their results in more appropriate and comparable ways. As vinyl chloride is a mutagen in laboratory experi ments and a proved human carcinogen, the minute doses that have escaped into the general environment as pollutants must be presumed to have caused com parably minute risks to the general public. No such risk could possibly be detected, other than one of angiosarcoma of the liver which is normally an ex tremely rare disease. Several surveys have sought evi dence of the existence of such an effect, and sugges tive evidence that such an effect may have occurred at a time when environmental pollution was much greater than it is now has been found in one. References 1. Air Products and Chemicals, Inc. Comments on the pro posed standard for vinyl chloride. Letter to DR Goodwin, Environmental Protection Agency', Washing ton, DC 23 September 1976, (Cited in an unpublished report by Barr to Air Products and Chemicals, Inc, in 1981). 2. Baser ME, Tockman MS, Kennedy TP. Pulmonary function and respiratory symptoms in polyvinyl chlo ride fabrication workers. Am Rev Respir Dis 131 (1985) 203--208. 3. Baxter PJ, Anthony PP, MacSween NM, Scheuer PJ. Angiosarcoma of the liver in Great Britain, 1963--73. Br Med J 2 (1977) 919--921. 4. Belli S, Bertazzi PA, Comba P, Foa V, Maltoni C, Masina A, Pirastu R, Regianni A, Vigotti MA. Indagine sulla mortalita dei produttori di PVC in Italia: Disegno dello studio e primi risultoti. Cancer Lett (in press). 5. Brady J, Liberatore F, Harper P, Greenwald P, Bur nett W, Davies TN, Bishop M, Poian A, Vianna N. Angiosarcoma of the liver: An epidemiologic survey. J Natl Cancer Inst 59 (1977) 1383--1385. 6. Buffler PA, Wood S, Eifler C, Suarez L, Kiliane DJ. Mortality experience of workers in a vinyl chloride monomer production plant. J Occup Med 21 (1979) 195-203. 7. Byren D, Engholm G, Englund A, Westerholm P. Mor tality and cancer morbidity in a group of Swedish VCM and PVC production workers. Environ Health Perspect 17 (1976) 167--170. 8. Byren D, Holmberg B. Tw o possible cases of angiosar coma of the liver in a group of Swedish \myl chloride workers. Ann NY Acad Sci 246 (1975) 249--250. 9. Cooper WC. Epidemiologic study of sinyl chloride workers: Mortality through December 31, 1972. Environ Health Perspect, 41 (1981) 101 -- 106. 10. Creech JL, Johnson MN. Angiosarcoma of liver in the manufacture of polyvinyl chloride. J Occup Med 16 (1974) 150--151. 11. Dalderup LM, Freni SC, Bras G, Bronckhursi FB. An giosarcoma of the liver. Lancet I (1976) 246. 12. Duck BW, Carter JT, Coombes EJ. Mortality study of workers in a polwinylchloride production plant. Lancet 2 (1975) 1197--1199. 13. Elinder CG, Pershagen G. Pilot study concerning the mortality in Njurunda Community. Swedish Nature Conservancy Board, 1978. (Cited in an unpublished report by Barr to Air Products and Chemicals, Inc, in 1981). 14. Environmental Health Associates. An update of an epi demiological study of vinyl chloride workers 1942--82: Final report to the Chemical Manufacturers Association. Environmental Health Associates, Oakland, CA 1986. 15. Environmental Protection Agency. Standard support document and environmental impact statement: Emis sion standard for vinyl chloride. Environmental Protec tion Agency, Washington, DC 1975. (EPA 450/2-75009). 16. Equitable Environmental Health. Epidemiological study of vinyl chloride workers: Final report to Manufacturing Chemists Association. Rockville, MD 1978. r 17. Falk H, Tetles NC, Ishak KG, Thomas LB, Popper H. Epidemiology of thorotrast-induced hepatic angiosar comas. Environ Res IS (1979) 65--73. 18. Fiechtner J, Reyes C, Rcntmerster K, et al. Epidemi ologic notes and reports: Angiosarcoma of the liver -- Wisconsin. Morb Mortal Wkly Rep 25 (1976) 57-58. 19. Forman D, Bennett B, Stafford J, Doll R. Exposure to vinyl chloride and angiosarcoma of the liver: A report of the register of cases. Br J Ind Med 42 (1985) 750-- 753. 20. Fox AJ, Collier PF. Mortality experience of workers ex posed to vinyl chloride monomer in the manufacture of polyvinyl chloride in Great Britain. Br J Ind Med 34 (1977) 1--10. 21. Heath GW, Falk H, Creech JL. Characteristics of cases of angiosarcoma of the liver among vinyl chloride work ers in the United States. Ann NY Acad Sci 246 (1975) 231--236. 22. Heldaas SS, Langard SL, Andersen A. Incidence of can cer among vinyl chloride and polyvinyl chloride work ers. Br J Ind Med 41 (1984) 25--40. 23. International Agency for Research on Cancer. Some monomers, plastics and synthetic elastomers and acro lein. Lyon 1979, pp 377--438. (1ARC monographs on the evaluation of the carcinogenic risk of chemicals to humans, volume 19). 24. Jones DP, Smith PM. Progression of vinyl chloride in duced hepatic fibrosis to angiosarcoma of the liver. Br J Ind Med 39 (1982) 306--307, 25. Jones RD. A mortality study of vinyl chloride monomer workers employed in the United Kingdom in 1940--1984. Scand J Work Environ Health (in press). 26. Lilis R, Anderson H, Miller A. Selikoff I. Pulmonary changes among vinyl chloride polymerisation workers. Chest 2 (1976): suppl, 299--305. 27. Lloyd MH, Gauld S, Copland L, Souiar CA. Epidemi ological study of lung function of workers at a factory manufacturing polyvinyl chloride. Br J Ind Med 41 (1985) 328-333. 28. Maltoni C, Ciliberti A, Gianni L. Chieco P. Vinyl chlo ride carcinogenesis: Current results and perspectives. Med Lav 65 (1974) 421--444. 29. Maltoni C, Lefemine G. Carcinogenicity bioassays of vinyl chloride: Current results, Ann NY Acad Sci 246 (1975) 195-218. 30. Maltoni C, Lefemine G, Ciliberti A, Cotti G, Carretti D. Experimental research on vinyl chloride carcino genesis. In: Maltoni C, Mehlman MA, ed. Archives of research on industrial carcinogenesis. Volume 2. Prin ceton Scientific Publishers, Princelon, NJ 1984. 31. Marsteller HJ, Delbach WK, M uller R, Gedigk P. Un usual splenomegalic liver disease as evidenced by peri toneoscopy and guided liver biopsy among polyvinyl chloride production workers. Ann NY Acad Sci 246 (1975) 95-134. 32. Marsteller HJ, Delbach WK Muller R, Juhe S, Lange CE, Rohner HG, Veltman G. Chronisch-toxische Lebersehaden bei Arbeitern in der PVC-Produktion. Dtsch Med Wochenschr 98 (1973) 2311--2314. 33. Masuda Y. Long term mortality study of vinyl chloride and polyvinyl chloride workers in a Japanese plant. Arch Ind Hyg Toxicol 30 (1979): suppl, 403--409. 34. Mirkova E, Mihailova A, Nosko M. Embriotakrichno i leratogenno destuige no vinilkhloride. Khig Zdraveopaz 21 (1978) 440. (Cited in an unpublished report by Barr to Air Products and Chemicals, Inc, in 1981). 35. Molan 1, Molan G, Holmberg B, Elofsson S, Holmund L, Moosing R, Westerholm P. Mortality and cancer rates among workers in the Swedish PVC processing indus try. Environ Health Perspect 41 (1981) 145--151. 36. Monson RR, Peters JM, Johnson MN. Proportional mortality among vinyl-chloride workers. Lancet 2 (1974) 397-398. 37. Nakamura K. A mortality study of vinyl chloride work ers in Japan. Sangyo Ika Diaguka Zasshi 5 (1983): suppl, 49--57. 38. Nicholson WH, Henneberger PK, Tarr D. Trends in can cer mortality among workers in the synthetic polymers industry. In: Industrial hazards of plastics and synthetic elastimers. Liss, New York, NY 1984, pp 65--78, 39. Nicholson WJ, Hammond EC, Seidman H, Selikoff IJ. Mortality experience of a cohort of sinyl chloridepolyvinyl chloride workers. Ann NY Acad Sci 246 (1975) 225-230. 40. Ott MJ, Langner RR, Holder PB. Vinyl chloride expo sure in a controlled industrial environment. Arch En viron Health 30 (1975) 333--339. 41. Pierre C, Tassignon JP, Pernin H, Spelkens J. Etude de la mortalite chez des travailleurs exposes au chlorure de vinyle. Arch Mai Prof Med Trav Secur Soc 40 (1979) 1131-1145. 42. Royal Society Study Group. Risk assessment: Report by a study group. Royal Society, London 1983. 43. Saric M, Kulcar Z, Zorica M, Gelic J. Malignant tumors of the liver and lungs in an area with a PVC industry Environ Health Perspect 17 (1976) 189--192. 44. Soutar CA. Epidemiological study of respiratory diseases in workers exposed to polyvinvl chloride dust. Thorax 35 (1980) 644--652. 45. Tabershaw IR, Gaffey WR. Mortality study of workers in the manufacture of vinyl chloride and its polymers. J Occup Med 16 (1974) 509- 518. 46. Theriault G, Allard P. Cancer mortality of a group of Canadian workers exposed to vinylchloride monomer. J Occup Med 23 (1981) 671--676. 47. Turner CA, Payne AP, Bushby BR. Determination of ambient levels of vinyl chloride monomer (YCM) around manufacturers in the UK: Part 7. Warren Spring Labo ratory, Department of Trade and Industry. Stexenacc (United Kingdom) 1984. 48. Viola PL, Bigotti A, Caputo A. Ontogenic response of rat skin, lungs, and bones to vinvl chloride. Cancer Res 31 (1971) 516-519. 49. von Greiser E, Reinl W, Weber H. Vmyl-chlorid ex position und mortalitai deutscher chemiearbeiter im vergleich zur mortalitai nichtexponiener chemiearbei ter und PVC-verarbeiter. Zentralbl Arbeitsmed Arbeitssch Prophyl Ergonomic 32 (1982) 44--62. 50. Waxweiler RJ, Stringer W. Wagoner JK. Jones J, Falk H, Carter C. Neoplastic risk among xsorkers exposed to vinyl chloride. Ann NY Acad Sci 271 (19'6) 40--48 Received for publication: 16 February 1988 78 AS I 00005613