Cardiovascular mortality among munitions workers exposed to nitroglycerin and dinitrotoluene.

W. Cardiovascular mortality among munitions workers exposed to nitroglycerin and dinitrotoluene. Scand J Work Environ Health 1991;18:34-43. A retrospective cohort mortality study with 5529 nitro glycerin, 4989 dinitrotoluene, and 5136 unexposedworkers compared the mortality of the exposedgroups with that of the United States population and that of the unexposed group with life-table analysis and Poisson regression.Mortality from ischemic heart diseasewascloseto that expected,and mortality from cerebrovasculardiseasewas slightlylessthan that expected, for the workers with both nitroglycerin and dinitrotolueneexposureand for those withdinitrotolueneexposureonly. A significantinteractionbetween age and nitroglycerineexposure was detected in the Poisson regression analyses for ischemicheart dis ease, particularly for workers actively exposed to nitroglycerin. The rate ratio for the workers under 45 years of age and activelyexposedto nitroglycerinwas 3.30 (95lJ7o confidence interval 129-8.48). This study did not showa chronic effect of nitroglycerinor dinitrotoluene exposure on cardiovascular disease risk. Potential biases related to the company's medicalscreening program may have limited the ability to detect chronic cardiovascular effects. .

from nitroglycerin exposure. Hogstedt and his coworkers have reported an association between ischemic heart disease (IHD) and cerebrovascular disease in cohort and case-referent studies of Swedish workers exposed to nitroglycerin and ethylene glycol dinitrate (14,15,16). Most of the cardiovascular deaths in the Swedish studies occurred among older workers with a long duration of exposure, months or years after their last exposure, which was interpreted by the authors as being suggestive of a •'chronic" effect. Finally, Craig et ai, in a study of Scottish explosives workers observed an excess of IHD mortality among workers exposed to both nitroglycerin and ethylene glycol dinitrate but not among workers exposed to nitroglycerin alone (17).
Dinitrotoluene (DNT) has also been recently reported to be associated with an increased risk of mortality from IHD in a study from the Chemical Industry Institute of Toxicology (CIIT) (18). The CIIT study was conducted to examine the potential carcinogenicity of DNT, in response to findings of an increased incidence of cancers of the liver, connective tissue, and kidney among DNT exposed rodents (19)(20)(21). The CIIT study failed to detect an excess of cancer for any site.
This retrospective cohort mortality study was conducted to examine the possible relationship of exposure to nitroglycerin and DNT to the risk of cardiovascular disease mortality. The relationship of exposure to DNT and the development of cancer will be examined in a sub sequent report.

Subjects and methods
The study population was identified from current and former white male workers from a United States (US) Arm y munitions facility in Radford, Virginia . Since 1949, this facility has been continually involved in the production of propellants used in munitions and rocket motors.
An extensive medical screening program, including preemployment and periodic med ical examinations, has been employed at this facility. Workers exposed to nitroglycerin were examined every six months , whereas all other employees were examined once a year. These examinations included a resting electrocardiogram (BCG) and blood pressure determinations . Since 1970, these examinations ha ve also included a treadmill ECG for job applicants, for workers employed in nitroglycerin-exposed jobs, and for workers over the age of 40 years. Job applicants with an abnormal ECG or hypertension are not hired. Workers developing hypertension or an abnormal ECG while working at the facility are restricted from employment in operations with potential for nitroglycerin or DNT exposure. The risk of cardiovascular disease has been further limited by the company's policy of not allowing workers to smoke cigarettes in the production areas of the facility .
Personnel records for all workers employed at the facility were microfilmed and demographic information , general dates of employment, and operations workers who had been employed for at least five months at the facility from 1 January 1949 to 21 January 1980 were selected for study.
A comprehensive environmental survey was conducted by industrial hygienists of the National Institute for Occupational Health and Safety (NIOSH) at the study facility in 1981 (22). Personal air-monitoring samples for nitroglycerin were collected and analyzed for 92 individuals from 10 different job locations . The time-weighted averages of the nitroglycerin exposure were found to range from 0.001 to 0.028 ppm. The potential for dermal exposure to nitroglycerin was not evaluated in this survey. We were unable to estimate quantitatively nitroglycerin exposures for our analysis because of a lack of dermal exposure data and possible changes in airborne exposure levels over the period of the study. Instead we developed the followin g qualitative system for evaluating the potential for exposure to nitroglycerin.
A listing of all unique operations which occurred more than once in the study computer file was reviewed by NIOSH industrial hygienists (TB and MB) with the assistance of company and union representatives. Each operation was rated concerning the probability of exposure to nitroglycerin and DNT . Using this information we formed the following three groups: (i) workers probably exposed to nitroglycerin, (ii) workers probably exposed to DNT, (iii) workers not exposed to either DNT or nitroglycerin. Workers whose ex-posure to DNT or nitroglycerin was classified as only "possible" were excluded from the analysis.
Detailed work histories were coded for individuals in the groups probably exposed to nitroglycerin or DNT. The histories included dates, production lines when applicable, and operations for all job change s at the plant. Only the first and last date of emplo yment were coded for the unexposed workers.
Each operation which was initially determined to be probably exposed to nitroglycerin was further reviewed by the NIOSH industrial hygienists with the assistance of company and union representatives to assess qualitatively the potential for dermal and respiratory exposure to nitroglycerin. On the basis of these discussions, one of the authors (TB) ranked each nitroglycerin-exposed operation as having either a " high" or " low" potential for nitroglycerin exposure. Although information on the air concentrations of nitroglycerin was available, quantitative estimates of exposure could not be developed since quantitative information on dermal exposure was una vailable.
Using this exposure classification scheme, we identified 5529 workers with probable exposure to nitroglycerin , 4989 workers with probable exposure to DNT, and 5136 workers not exposed to either DNT or nitroglycerin. Excluded from the analyses were 3256 workers only possibly exposed to nitroglycerin or DNT. Altogether 2512 workers were exposed to both nitroglycerin and DNT .
The vital status of the cohorts was ascertained as of 31 December 1982 through the following sources: the Social Security Administration, the Internal Revenue Service, post offices , and motor vehicle departments. Death certificates were obtained from state vital statistics bureaus and coded by a trained nosologist using the International Classification of Diseases revision in effect at the time of death . The results from the vital status ascertainment are presented in table 1. We were able to trace 970/0 of the study population successfully . Loss to follow-up was slightly lower in the nitroglycerin group (1.6%) than in either the DNT (2.8%) or unexposed (4.0%) group. The percentage of missing death certificates was again slightly lower for the nitroglycerin group (l.l 010) than either the DNT (1.5%) or unexposed (2.3%) group. These slight differences in loss to follow-up and failure to obtain death certificates are unlikely to have introduced any serious biases into our study.
A modified life-table program was used to compute the expected numbers of deaths by multiplying mortality rates specific for cause, five-year age groups, and five-year calendar groups from data on white males in the US population by the corresponding personyears distribution of the study population (23). US mortality rates for 1975-1978 were used for 1979-1982 since mortality rates for after 1978 were unavailable . Person-years were enumerated for the exposed workers after they had been employed for at least five months and had worked at least 1 d in an exposed oper- at ion; the unexposed workers simpl y had to have five months of employment. Person-years were counted until either the end of the study (31 December 1982) or until the date of loss to follow-up or until the date of death, whichever occurred first. Standardized mortalit y ratios (SMR) were calculated by dividing the observed deaths by the number expected. Exact statistical tests (two-tailed) and associated 95070 confidence intervals (95% Cl) were estimated On the basis of the Poisson distribution. Direct comparisons were made between the mortalit y of the exposed and unexposed groups for IHD and cerebrovascular disease. A matrix of person-year and deaths was generated for the direct comparisons and the Poisson regression with an SAS (stat istical anal ysis system) program (24). Directly standardized rate ratios (SRR) were computed with the use of the age and calendar-time distribution of the person-years of the entire study population as the weighting factor. A modification of the Mantel-Haenszel procedure was used for computing chi-squares on the bas is of the assumption of a binomial distribution (25). Test-based 95% confidence intervals were computed for the SRR values (26). Testing for trend was accomplished with a modification of the Mantel extension test for incidence density data (27). Tests of linear trend were performed with the use of the natural logarithm of the median duration of the person-years for each duration -of-exposure category.
The Poisson regression was used to evaluate confounding and the possible interactions between nitroglycerin, DNT, and other covariates (ie, age and calendar year). The models were fitted with the use of the GUM (general linear interactive modeling) statistical package (28,29). Parameters for exposure, age, and year were included in all the models. Terms representing the interactions between the exposure variables and the other covariates were evaluated and were only included in the final model if they were determined to be statistically significant (P < 0.05) according to the likelihood ratio test statistic.
The same categories were used to represent exposure, age at risk (five-year), and calendar year at risk 36 (lO-year) for the direct standardization and regression analyses. Analyses were performed with the use of the following exposure categories: ever versu s never exposed (nitroglycerin and DNT), duration of expo sure (nitroglycerin and DNT), high versus low nitroglycerin exposure , and time since last exposure to nitrogl ycerin. For the analysis with time since last exposure to nitroglycerin, the workers were categorized as being "actively" exposed to nitroglycerin if it had been less than one month since their last date of nitroglycerin exposure. Potential confounding by employment status (active versus retired) was also controlled for in the model for time since last exposure to nitroglycerin. The workers were categorized as being "actively" employed if it had been less than one month since their last date of emplo yment at the study facilit y.

Results
The results from the life-table analysis for all causes and for the major noncardiovascular death categories are presented in table 2. All cause mortality was close to that expected for all three groups. All three groups experienced an excess in mortality from mental and per sonality disorders, which was primarily due to alcoholism, and an excess in mortality due to violence.
The results for cardiovascular and cere brovascular causes of death are presented in table 3. Mortality from cerebrovascular disease was observed to be less than expected for the nitroglycerin (SMR 0.90, SRR 0.87) and DNT (SMR 0.95, SRR 0.89) groups. IHD mortality was close to that expected in the DNT group (SMR 0.98, SRR 0.99) and slightly elevated in the nitroglycerin group (SMR 1.07, SRR 1.07). In the unexposed group a nonsignificant excess was observed for mortality from hypertension with heart disease (SMR 1.59) and for mortality from hypertension without heart disease (SMR 2.06) . Hypertension without heart disease (SMR 1.17) and other myocardial degeneration (SMR 1.41) were slightly elevated in the DNT group.
The results from the life-table anal ysis stra tified by age at risk are presented in table 4. A large excess of IHD was evident among the nitroglycerin-exposed work ers in the < 35-year age group (SRR 5.46, 95070 CI 0.83-36.02). An excess in mortality from cerebrovascular disease was also observed among the DNTexposed workers in the 55-to 59-year age group (SRR 4.46,95% CI 1.11-17.84). A significant (P=0.03) interaction between nitroglycerin exposure and age was detected in the Poisson regression model for IHD , which was primarily due to a significan t effect among workers who were exposed to nitrogl ycerin and under 35 years of age (SRR 10.31, 95% CI 1.33-79.86). There was no indication of a significant interact ion or of confounding between DNT and nitro glycerin exposure in this, or in any, of the other regression models evaluated.
The results of the life-tabl e analysis stratified by calendar year at risk are presented in table 5. No consistent trends with calendar year were evident for mor-tality from IHD or cerebrovascular disease for any of the exposure groups. In the Poisson regression models, calendar year did not appear to be a significant predictor for mortality from either IHD or cerebro vascular disease, and there were no stati stically significant intera ctions between calendar year and any of the exposure variabl es.
The results for mortality from IHD and cerebrovascular disease by duration of exposure are presented in table 6. An inverse relationship between duration of exposure and mortality from IHD or cerebrovascular disease was evident for the nitroglycerin group and to some extent for the DNT group . For IHD mortality the inverse trend was statistically significant (P = 0.003) for DNT exposure and nearly significant (P = 0.08) for nitroglycerin exposure.
The nitroglycerin group analysis was further str atified by qualitative high and low exposure categories (table 7). Mortality from cerebrovascular disease (SRR high 0.77, SRR low 0.93) and from IHD (SRR high   a SRR standardized for calendar year with the care of the unexposed group as the reference group and the person-years dlstrlbution of the unexposed group to weight the straturn-specltlc rates . b SRR could not be est imated because the reference group rate was zero for these categories. * P <0.05, ** P <0.01 (two-tailed). Table 5. Standardized rate ratios' (SRR) for ischemic heart disease and cerebrovascular disease by calendar year at risk.

(DNT = dinitrotoluene)
Year 1949-1959 1960-1964 1965-1969 1970-1974 1975-1979 1980-1982 Group Nitroglycerin-exposed DNT-exposed • SRR standardized for age with the use of the unexposed group as the reference group and the person-years distribution of the unexposed group to weight the stratum-specific rates.  • SRR standardized for age and calendar year with the use of the unexposed group as the reference group and weighting of the stratum-specitic rates by the co mbined person -years distribution of the exposed (hi.gh or lo~) and une xpos ed groups. . . . b High and low exposure mutually exclusive. Individuals who worked onboth hIgh and low exposure jobs were only oncluded onthe high exposu re category. b Active =less than one month since last exposure to nitroglycerin,1 month-<5 years = between 1 month and less than 5 years since last exposure to nitroglycerin, 2:5 years = 5 or more years since last exposure to nitroglycerin.
0.96, SRR low 1.14) was higher overall in the low-exposure group than in the high-exposure group. A highly significant (P<O.OOOOI) interaction between high exposure and age was detected in the regression model; it was primarily due to an excess of high exposure to nitroglycerin among workers under the age of 35 years (SRR 19.38, 95% CI 2.47-152.07). Finally, an analysis was performed of IHD mortality by time since last exposure to nitroglycerin (table 8). A small and statistically nonsignificant excess of IHD was observed for workers in the actively (less than one month since last exposed) exposed group (SRR 1.17, 950/0 CI 0. 23-5.92) and in the group whose exposure occurred more than five years ago (SRR 1.22, 95% CI 0.82-1.71). Poisson regression and life-table analyses of the relationship between duration of exposure and IHD mortality were performed for the workers with more than five years since last exposure to assess further the potential for a chronic effect. The rate ratio for IHD mortality was observed to decrease with increasing duration of exposure among the workers with more than five years since last exposure in both analyses (results not shown). A highly significant (P = 0.001) interaction between age and active nitroglycerin exposure was detected in the Poisson regres-

AGE
sion model for IHD mortality. The results from this model by age are presented in figure 1. As with the previous analyses for overall nitroglycerin exposure and high nitroglycerin exposure, this interaction was primarily due to a large excess among workers actively exposed to nitroglycerin in the < 35-year age group (SRR 21.71,95% CI 5.34-88.23). Smaller and non-\ significant excesses of IHD mortality were also observed in the regression model for IHD among workers actively exposed to nitroglycerin in the 35-to a Time since last exposure to nitroglycerin, that is, the number of days between the last date employed in an exposed job in the personnel record and the date of death . It is likely that the last date exposed in the personnel record is not accurate to the day. b Time between the onset of symptoms and death as reported on the death certificate. c Findings from the company's preemployment medical examination.
According to the Poisson regression, the rate ratio for IHD mortality among workers <45 years of age who were actively exposed to nitroglycerin was over three times that of unexposed active workers (SRR 3.30, 95% CI 1.29-8.48). A summary of key characteristics of the cases who died from IHD before reaching the age of 45 years while actively exposed to nitroglycerin is presented in table 9. According to their death certificates these casesgenerallydied shortly after the onset of symptoms, and during the weekend or at the beginning of the week (Monday or Tuesday). According to the personnel records, the cases appeared to have a relatively long duration of exposure to nitroglycerin (mean = 7.7 years), and, with the exception of two cases, they died on the same day that they were last exposed to nitroglycerin . All but two of the cases had had normal preemployment medical examinations. One WaS however hypertensive (systolic blood pressure 170, diastolic blood pressure 118), and the other had a family history of IHD and had previously suffered a myocardial infarction. All but one of the cases had occurred prior to the 1970s.

Discussion
Mortality from IHD was close to that expected, and mortality from cerebrovascular disease was less than that expected among the nitroglycerin-and DNT-exposed workers when either the US population or the unexposed group is used as the reference group. The fact that the cardiovascular mortality was close to the expected level for these groups is unusual, since car-40 diovascular mortality has generally been observed to be less than that expected among workers in comparisons with the US population (30,31). This finding may be explained by the fact that our study facility was located in a county in which the cardiovascular mortality rate is substantially higher than that of the nation as a whole.
Mortality from IHD or cerebrovascular disease also did not appear to differ appreciably between our highand low-exposure groups. Although these categories were based upon the best judgement of company, union, and NIOSH personnel, the potential for a negative bias due to misclassification errors must be recognized. An excess of IHD ·mortality was observed among the young workers highly exposed to nitroglycerin.
Significant interactions between age and nitroglycerin exposure were detected for IHD mortality in several of the Poisson regression models. The age interaction was particularly pronounced for the workers who were under 35 years of age and actively exposed to nitroglycerin. A nonsignificant excess of IHD mortality was also apparent for the workers actively exposed to nitroglycerin in the 35-to 39-year age group and the 40-to 44-year age group. IHD mortality was, however, generallyless than expected among the workers who were activelyexposed to nitroglycerinand were 45 years of age or older.
Several features of the eight IHD deaths of workers who were under the age 45 years and actively exposed to nitroglycerin appear to be consistent with previous case reports of sudden death among workers following withdrawal from exposure to nitroglycerin and ethylene glycol dinitrate (2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). The workers in most previously reported cases were also under the age of 50 years and died suddenly without prior cardiovascular symptoms. Our cases would be considered sudden deaths since the time between the on-set of symptoms and death ranged from 0 to 12 h. In six of the eight cases there was no evidence of preexisting cardiovascular disease according to the workers' preemployment medical examination.
The previously reported cases generally occurred 24 to 48 h after withdrawal from exposure to nitroglycerin . Six of the eight deaths in our study occurred on the same day that the workers in question were last employed in an operation with potential for nitroglycerin exposure. However, according to the company personnel · manager, it is likely that the personnel records used in our study were not accurate to the day and that the personnel department in some cases may have.used the date of death as the last date employed even if the person died during a break from work. Thus it is impossible to conclude that the deaths observed in this study among young workers actively exposed to nitroglycerin are due to withdrawal from nitroglycerin, although this would appear to be the most likely explanation for our findings.
Seven of the eight IHD deaths which occurred among the workers who were under the age of 45 years and actively exposed to nitroglycerin happened prior to the 1970s. In the regression analysis for IHD mortality the interaction between year at risk and time since last exposure to nitroglycerin was not statistically significant (P = 0.13). According to a model which included this interaction, the rate ratio for the workers actively exposed to nitroglycerin was 1.94 for 1949 to 1960, 1.07 for 1960 to 1969, and 0.95 from 1970 on. The declining IHD risk over time was likely to be due to the medical program that has been initiated since the early 1970s.
Our findings can be discussed in relation to the earlier reports of a chronic cardiovascular effect among Swedish explosive workers (14)(15)(16). In these studies an excess of cardio vascular and cerebrovascular disease was primarily observed among older workers (over 50 years) with a long duration of exposure, and most of the exposed cases occurred months or years following the cessation of exposure. This result was interpreted by the authors as being suggestive of a chronic cardiovascular effect. In contrast, we observed an inverse relationship between IHD mortality and duration of exposure to nitroglycerin and no excess in mortality from IHD and cerebrovascular disease among older workers. A slight excessin IHD mortality was observed among the workers five years after the cessation of nitroglycerin exposure ; however, IHD mortality was also inversely related to duration of nitroglycerin exposure in this group. In one of the Swedish studies, the excess of card iovascular mortality was limited to workers after a 20-year latency period (16). IHD mortality did not appear to be elevated among nitroglycerin -exposed workers after a 20-year latency period in our stud y (SMR.0.97).
Unlike the Swedish studies, the workers in our study were not exposed to ethylene glycol dinitrate. Ethylene glycol dinitrate is an even more potent vasodilator than nitroglycerin, and it may be potentially more toxic than nitroglycerin when one considers its higher volatility and its higher potential for skin absorption (32). Thus the differences between our findings and the Swedish studies might be attributed to the influence of ethylene glycol dinitrate. This interpretation is supported by a study of Scottish explosive-manufacturing workers which examined the mortality experience of workers exposed to both nitroglycerin and ethylene glycol dinitrate and workers exposed to nitroglycerin alone (17). A statistically significant excessin IHD mortality (SMR 1.50) was only observed among young workers (age 15-49 years) exposed to nitroglycerin and ethylene glycol dinitrate, and not among workers exposed to nitroglycerin alone .
DNT exposure did not appear to be associated with an increased risk of either IHD or cerebrovascular disease in our study . This finding is in contrast with the results of a report by Levine and his co-workers (18) from the CUT of a significant excessof IHD mortality (SMR 1.43) among workers exposed to DNT at both US Army armament facilities studied. Our study site was one of the facilities included in the CUT investigation; however, its report included workers from a somewhat different time period (1940)(1941)(1942)(1943)(1944)(1945)(1946)(1947)(1948)(1949)(1950)(1951)(1952)(1953)(1954)(1955)(1956)(1957)(1958)(1959) than ours . When we ran our life-table analysis on workers employed during the period which overlapped with the CUT study (1949)(1950)(1951)(1952)(1953)(1954)(1955)(1956)(1957)(1958)(1959), we failed to observe an excess of IHD mortality among the DNT exposed workers. An alternative explanation for these discrepancies is that the definition of exposure to DNT differed between the two studies. The stud y population in the CUT study was restricted to worker s who had worked in jobs with substantial DNT exposure. A similar restriction was not made when we defined the DNT group for our investigation. There was also an indication in the CUT study that the risk of IHD mortality was highest among workers 15 years after their first exposure (latency). IHD mortality was only slightly elevated (SMR 1.06, 95070 CI 0.93-1.21) among the DNT-exposed workers with more than 15 years of latency in our study.
At least two important potential sources of bias need to be considered when the results of our investigation are interpreted. Foremost is the potential for bias introduced by the medical screening program. Mortality from cardiovascular disease is usually substantially less in working populations than in the US population due to the healthy worker effect (29,30). This effect should be particularly strong in our study, since workers were required to pass a preemployment medical examination that included cardiovascular tests. Because of this potential source of bias, we used an internal unexposed group, in addition to the US population, as a reference group. Unfortunately this internal comparison group may also have been susceptible to selection biases related to the medical screening program.
The workers who were employed in areas of the facility with potential for exposure to nitroglycerin were required to pass a more rigorous medical examination than the preemployment examination, and the individuals with an abnormal ECG or abnormal blood pressure could be restricted from work in operations with potential exposure to either nitroglycerin or DNT. A potential bias could have been introduced since the individuals at higher risk of cardiovascular diseasemay have been preferentially assigned to unexposed areas of the facility. The results from our study do suggest that such a bias may have occurred since an excess of hypertensive heart diseasewas observed among the unexposed workers, whereas mortality from this cause was observed to be less than that expected among the nitroglycerin and DNT groups. The observed inverse relationship between cardiovascular mortality and duration of exposure to either nitroglycerin or DNT may also be a reflection of the bias introduced by the medical screening program since workers with a long duration of exposure would have been required to pass numerous medical examinations. A second potential source of bias is related to our inability to control for other important cardiovascular risk factors. Tobacco smoking, a strong risk factor for cardiovascular disease, is of particular concern. Workers employed in the production areas of this facility are not allowed to smoke due to the explosion hazard. This policy may have introduced a negative bias in the comparisons between the US population and the various groups examined in this study. The influence of this policy on internal comparisons made in this study is less clear. However, the fact that respiratory cancer mortality did not appear to vary appreciably between the nitroglycerin (SMR 0.83), DNT (SMR 0.86), and unexposed (SMR 0.86) groups provides indirect evidence that the smoking habits of these groups were similar.
In summary, our study demonstrated a significant excess of IHD mortality among the workers who were actively exposed to nitroglycerin and under the age of 45 years (SRR 3.30, 95070 CI 1. 29-8.48). These findings are generally consistent with those of earlier case reports of sudden death due to IHD after withdrawal from nitroglycerin exposure. Our study failed to detect a chronic cardiovascular effect, as previously reported among Swedish explosives workers. This difference may be due to bias introduced by the medical screening program or to the absence of the ethylene glycol dinitrate exposure which had occurred in the Swedish studies. Our study also failed to detect a previously reported association between DNT exposure and IHD mortality. Although the previous report included our study facility, differences in the time period studied and the definition of DNT-exposed operations could possibly explain our discrepant findings.