of the styrene-butadiene rubber industry. Mortality patterns with discussion of the hematopoietic and lymphatic malignancies.

Environmental epidemio logic investigation of the styrene-butadiene rubber industry: Mortality patterns with ciiscussion of the hematopoietic and lymphatic malign;lncies. Scand j work environ .health 8 (1982) 250-259. A retrospective cohort mortality study and an industrial hygiene assessment were undertaken in two styrene-butadiene rubber producing facilities in eastern Texas. Occupational histo!Y records were available from 1943 at plant A and from 1950 at plant B to the study cut-offdate of 31 March 1976. With a two-sided test statistic, no statistically significant excesses in total or cause-specific mortality were observed for the overall worker population of either plant. However, the plant A study group demon strated a nonsignificant statistical excess [standardized mortality ratio (SMR) of 203] for the cause-specific category of leukemia and aleukemia. Additional analyses were per formed on a subgroup consisting of all white males with at least six months of employ ment at plant A between the beginning of 1943 and the end of 1945, a time which coin cided with process and operational changes. An SMR of 278, also not statistically signifi cant, was demonstrated for the leukemia and aleukemia cause-specific category. Due to the relative modest study population sizes, the power of this study to detect statistically significant excesses in leukemias or other malignancies of the hematopoietic and lym phatic tissues is not very large unless one is interested in 'substantial excesses, such as those that would correspond to a fourfold increase in risk.

Styrene-butadiene rubber (SBR) is the most widely used synthetic rubber in the world. The production of this synthetic was not actively pursued in the United States or elsewhere in the world until the supply of natural rubber had been curtailed by war activities in the Pacific during the late 1930s and early 1940s. In 1941, the Rubber Reserve Company of the US Reconstruction Finance Corporation initiated a program to construct plants which would permit the annual production of 40,642 t (metric) of synthetic rubber. Before the end of the war, this annual capacity had been increased to about one million metric tons. As part of this program, the United States government constructed 15 SBR plants, two butyl rubber plants, 16 butadiene production facilities, and five styrene production facilities. After World War II, these facilities were sold to private corporations (13).
In January 1976, two men who had been employed at adjacent SBR facilities in Port Neches, Texas, died of leukemia. On the basis of preliminary observations (10), the National Institute for Occupational Sa- ment records documenting the employment of 2,015 workers were available.from 1950 on. Of this total worker population, 1,094 were white males having at least six months of employment. In this latter group of employees, the average length of employment was 10.78 a, and the average annual turnover rate was 5.10 %. An estimated 485 people were employed at this plant when the study was initiated. Processes and environmental conditions of an industry change with time. To illustrate the changes in this particular industrial process, some of the important process changes which have occurred at plant A since 1943 are briefly described. Originally, this plant produced rubber by a hot temperature, batch polymerization process. The batch polymerization process was then converted to a continuous feed operating system in 1946. In 1949, a cold temperature, continuous polymerization process was developed to go along with the hot process. Extender oils were added to the chemical process in 1951 to insure that the rubber product was not too hard for future processing. A pilot polyethylene operation was constructed at this plant in 1960, but was discontinued in 1965. A general modernization program of the production area was begun in 1967. Undoubtedly, such changes influence environmental conditions within the plant. Similar process evolution has taken place at plant B.
Styrene and butadiene were considered as the target chemicals for this study because they are the two major components of SBR, and benzene exposure was evalu-

Process description and environmental concentrations
SBR is a copolymer of styrene and butadiene made through an emulsion polymerization process. A typical emulsion system contains water, monomers, initiators, and an emulsifier (soap) (11). A typical recipe, as described by one of the plant managers, is shown in table 1. The two plants included in this study began operations in 1943. Plant A has been operated continuously by BF Goodrich, which purchased the plant from the United States government in 1955. Personnel employment records and occupational histories were available for all personnel since 1943. According to best estimates, a total of3,494 persons have worked at this plant. Of this total population, 1,662 were white males who had at least six months of employment. In this latter group of employees, the average length of employment was 9.48 a and the average annual turnover rate was 7.00 %. An estimated 700 people were employed at this plant when the study was initiated.
Plant B was operated by Firestone from 1943 to 1947 and then shut down. Attempts to gain Firestone authorization for access to social security data which would identify early employees have failed. The plant was reactivated in 1950 by the US Rubber Company and operated until 1955. At that time, the plant was purchased by a company jointly owned by US Rubber and Texaco. This jointly owned company operates the facility today. Personnel employ-fety and Health decided to conduct detailed environmental and epidemiologic studies at these facilities to test the hypothesis that employment in the SBR production industry was associated, specifically, with an increased risk of leukemia and, more generally, with an increased risk of other malignancies of hematopoietic and lymphatic tissue. In the course of this investigation, mortality information on malignancies of hematopoietic and lymphatic tissue, as well as on other causes of death, was collected. Information on other causes of death was used to explore the possible existence of other previously unrecognized unusual mortality patterns and is presented along with the information on leukemia.
ated due to its known leukemogenic activity. Measurements were taken in all areas of both SBR production facilities with an approved method (3).
The results of the environmental samples are presented in table 2. All of the samples were well within existing standards of the Occupational Safety and Health Administration and threshold limit values recommended by the American Conference of Governmental Industrial Hygienists (l00 ppm of styrene, 1,000 ppm of butadiene and 10 ppm of benzene for an 8-h time-weighted average). No historical monitoring data were available.

Methods
Vital status determinations Were made through 31 March 1976 for all cohort members with the help of several follow-up sources, including the Social Security Administration, the US Post Office, the Texas Bureau of Motor Vehicles, and a commercial case location service. Death certificates were obtained for any study members known to be deceased. The underlying causes of death taken from the certificate were coded by one nosologist in accordance with the revision of the International Lists of Disease and Causes of Death in effect at the time of death and then converted into the 7th revision code for the purposes of comparison.
A modified life-table (4,5) technique was used to obtain the person-years at risk of dying by 5-a age groups, by 5-a calendar time periods, by duration of work experience (exposure), and by time from initial employment (latency) in the SBR production industry. Only white males with at least six months of nonmanagement and nonadministrative employment were included in these analyses because the other sex-ethnic subgroups were not large enough for meaningful analyses and a shorter employment period would be difficult to correlate with chronic health conditions. Person-years at risk of death began accumulating at six months after initial employment. The age, race, sex, calendar time and cause-specific mortality rates of the US population were then applied to the appropriate stratum of person-years at risk to generate the expected number of cause-specific deaths in the study population. This expected number of deaths was then compared to the observed number of deaths in the study population so that any statistically significant differences and the standardized mortality ratio (SMR) could be determined. The magnitude of the differences was evaluated by a two-sided test statistic based on the Poisson distribution (with p < 0.05).

Results
The determination of vital status for the workers in plants A and B are shown in tables 3 and 4, respectively. At plant A, 1,356 persons were identified as alive, 252 persons were found to be deceased (with death certificates received for 246 individuals), and for 54 individuals (3.25 %) vital status was undetermined. In the data analyses, the latter 54 were all considered alive, and they thus contributed to the person-years at risk and made the causespecific death results conservative. A total of 34,187 person-years at risk of dying had accumulated for the entire plant A study group. At plant B 980 persons were identified as alive, 80 persons were found to be deceased (with death certificates received for 78), and vital status was undetermined for 34 individuals (3.11 %). As with their counterparts at plant A, this last group was considered to be alive. A total of 19,742 Table 3. Vital status of white males with at least six months of employment at plant A between the beginning of January 1943 and the end of March 1976. a person-years at risk of dying had accumulated for the entire plant B study group. Table 5 shows the total mortality pattern for plant A, and it indicates deficits for the categories total mortality, all other cancer, and disease of the respiratory system. Elevated SMRs, though not statistically sig- nificant, were observed for the categories pertaining to malignant neoplasms of the lymphatic and hematopoietic tissues. The deficits in overall and some cause-specific mortality may partially be explained by the widely documented healthy worker effect (6,7). This effect has been explained by the selection criteria used by most physically demanding industries; these criteria require that new employees have a certain level of good health before being hired.
When the records of the individuals identified with leukemia in plant A were F~viewed (presented in table 6), it was ob-served that most of these employees had started work before the end of December 1945. This date corresponds to the time when the batch process was converted to a continuous feed operation and the war time production conditions were discontinued. Because of these process changes, it was decided to evaluate the mortality experience of white male employees with at least six months of employment in plant A between the beginning of January 1943 and the end of December 1945. This subgroup of the plant A population consisted of 600 workers. For this group of employees, the average length of employment was 11.89 a and the average annual turnover rate was 6.84 %. Table 7 presents the vital status: 365 were identified as alive, 201 were found to be deceased, and for 34 (5.67 %) the vital status was unknown. The experience of this subgroup resulted in the accumulation of 17,086 person-years at risk of death. Table 8 contains the observed and expected numbers of deaths and the SMRs for the cause-specific death categories for this subgroup. Deficits of mortality were observed for the categories total mortality, all other cancers, and accidents. The same difficulties with classification prior to 1950, as was previously discussed, affect the same cause-specific categories in table 8. Excesses in cause-specifc mortality were observed in the overall category malignant neoplasms of the lymphatic and hemato-poietic tissues and its subcategory leukemia and aleukemia with associated SMRs of 212 and 278, respectively. Table 9 shows the total mortality pattern for plant B, and it indicates deficits for  Subsequently, the mortality rates for 1968 to 1975 have been incorporated, and the computer program has undergone minor revisions which have resulted in slightly different expected numbers of deaths. b ICD = International Classification of Diseases.
total. mortality, overall malignant neoplasms, malignant neoplasms of the digestive organs and peritoneum, and all other causes. These deficits may also be explained in part by the healthy worker effect. None of the elevated SMRs were statistically significant or appeared to represent a trend in mortality.

Leukemia cases
A number of leukemia cases were identified in former plant A or plant B employees who did not qualify for consideration in the analyses for any of several reasons, including employment of less than six months, employee nonwhite, the death having occurred after the end of the study follow-up, and leukemia listing as secondary cause of death.  (1), also from the University of North Carolina's Occupational Health Studies Group, reported the mortality experience of a population from another rubber manufacturing plant observed between 1964 and 1973. The population stl,ldied was defined as "any persons who, as of January 1, 1964, were 40 or more years of age and an active or living retired hourly worker from the plant [p 387]." The SMR for neoplasms of the lymphatic and hematopoietic tissues for white male cohort members in the 65-to 84-year-old age group and the 40-to 85-year-old age group demonstrated a statistically significant excess in the monocytic leukemia and the other neoplasms of the lymphatic and hematopoietic tissues subcategories.
In a second paper, Andjelkovich et al (2) reported SMRs for persons employed in specific work areas at the same rubber manufacturing plant studied earlier; they used the entire cohort's mortality experience as a reference for comparison. A statistically significant excess of deaths for malignant neoplasms of lymphatic and hematopoietic tissues was reported for persons whose most representative work area was general services, which would not necessarily involve contact with SBR production.
In an earlier paper by Monson et al (9), mortality experience of all members of the same local union who worked for one company in Akron, Ohio, was presented. The study included 13,571 men who had worked during or after 1935 for at least 5 a.
Excesses of deaths due to leukemia were observed for those men working in the tire and the processing divisions. The results of the Monson et al study reinforce the possibility of an association between exposure to SBR production and the development of lymphatic and hematopoietic malignancies.
Both plants in the current study experienced statistically significant deficits in the total mortality category. Excesses of mortality in plant A, though not statistically significant, were observed in the categories of overall malignant neoplasms of the hematopoietic and lymphatic tissues and its subcategories lymphosarcoma and reticulosarcoma, Hodgkin's disease, and leukemia and aleukemia, the largest number of d~aths occurring in the last mentioned. The mortality experience of white males with six months of employment in plant A between the beginning of January 1943 and the end of December 1945, a period prior to a major process change, revealed increases in the mortality excesses of the same three categories. The p-values associated with the excesses in the overall category malignant neoplasms of the lymphatic and hematopoietic tissues and the subcategory leukemia and aleukemia, though greater than the arbitrarily chosen significance level of 0.05, were less than 0.1. Such excesses, though derived from relatively small numbers of observed deaths, are approaching statistical significance. These findings are consistent with those of McMichael et al (8) and the unpublished report by Spirtas et al. In addition, the independent investigations of rubber manufacturing facilities by Andjelkovich et al (1, 2) and Monson et al (9) reinforce the possibility of an association between working with SBR and the development of lymphatic and hematopoietic malignancies.
Analyses using detailed work history assignments of the study group members could not be conducted at the time of this report. Such analyses are planned and will help clarify whether the observed excess in malignancies of the hematopoietic and lymphatic tissues might be based on a risk experienced by specific subgroups of the overall cohort. Length of employment and time since initial employment were considered but did not demonstrate any discernible pattern in these three popula-tions. In an attempt to avoid the over-s~plification of these issues, we have deferred their discussion until the detailed work history assignment can be incorporated into the analyses. Of particular interest would be the association between mortality experience and specific work assignments of employees who had worked at least six months before the end of December 1945. The available information indicates that the excesses in mortality of neoplasms of the lymphatic and hematopoietic tissues were greater for this group, though not statistically significant.
In closing we would like to make two observations about the results of this study. One concerns the ability of an investigation to detect a significant difference between observed and expected cause-specific mortality which is the statistical power of the study. The statistical power is derived from the expected rate of the cause-specific mortality and the person-years at risk. In this study, three modest study cohorts (which had accumulated 34,187; 19,742 and 17,086 personyears at risk) were considered, and the cause-specific mortality of interest was that of leukemia (which has an annual mortality rate for white males of approximately 8.81/100,000 in the United States). These data result in low power, ie, limited ability to detect significant differences, unless one is interested in large relative risk associated with exposure. In the table of appendix 1, the power percentages indicate that it is unlikely that a doubling of the expected occurrence of leukemia would be identified as significant, but likely that four times the expected occurrence would be identified.
The second observation has to do with the use of a two-sided test statistic, which is conservative in its ability to detect significant differences, if there is no reason to believe the environment would be protective against cause-specific mortality (acknowledging the operation of the employment selection bias known as the healthy worker effect). Historically, it was the custom at the National Institute for Occupational Safety and Health to use the conservative two-sided test statistic. Prior to and at the time this report was prepared and originally presented, two of the authors (TJM and RAL) were debating the relative merits of one-sided versus two-258 sided test statistics. Although we subsequently agreed that the one-sided test statistic was more appropriate for tests of a specific hypothesis about a potential excess risk of cause-specific mortality, we have left the two-sided test statistic in this paper since the results had already been presented in that way. The reader may want to consider these two observations in evaluating the results, especially in those related to malignancies of the hematopoietic and lymphatic tissues. If a one-sided test statistic had been used, the mortality for the leukemia and aleukemia subcategory would have been statistically significant at the 0.05 level for the subgroup of plant A.

Conclusions
No statistically significant excess in causespecific mortality was detected in any of the three study groups (overall plant A, plant B, or restricted plant A). Excess mortality, though not statistically significant for specific categories of neoplasms of the lymphatic and hematopoietic tissues, was observed for both overall plant A and the restricted plant A study groups. These findings continue to suggest, as have four previous reports, that the production or manufacturing of SBR may be associated with an excess of lymphatic and hematopoietic neoplasms.