Skip directly to: content | left navigation | search
  1. ATSDR ›
  2. NER ›
  3. TCE ›
  4. Trichloroethylene (TCE): Mortality Data Analyses

National Exposure Registry, Trichloroethylene (TCE): Mortality Data Analyses

    SECTION 6

    MORTALITY DATA ANALYSES

    In accordance with the Policies and Procedures Manual (1), attempts made in preparing the TCE Subregistry to identify and locate at the time of baseline interview all eligible persons who were living or deceased. The initial identification of potential off-site registrants was accomplished by following leads furnished by the on-site and former occupants of the residences with contaminated water supplies. For most decedents, identification leads were furnished by spouses or relatives. Given the information supplied, death certificates were then located. Only those deceased individuals for whom death certificates were obtained were included in the TCE Subregistry data file and the subsequent analyses.

    The estimated start date of exposure varied across sites (see Section 7, Figure 7-1) from 1970 through 1980; for most sites and for the majority of registrants, exposures were thought to have begun prior to 1978. The start date is a conservative estimate based on available information. Eligibility of deceased and living registrants was based on their residency at an exposure address during the time frame calculated for the particular site where they had resided.

    Of the 239 death certificates obtained for the TCE Subregistry, 229 (96%) were for registrants who died in state(1); 140 (59%) were male and 99 (41%) were female. All decedents were classified as white on the death certificates. The underlying cause of death for each decedent was classified according to ICD-9 codes (19) using procedures described in the National Exposure Registry procedures (1).

    The possibility existed that a smaller percentage of eligible persons, particularly decedents, would be identified if they had moved from the area compared with those who remained in the immediate vicinity. Table 6-1 provides a comparison of the percentage of registrants (both living and deceased) located out of state versus in state at the time of the baseline interview. Approximately the same percentage (9%) of decedents and living registrants were located out of state.

    Table 6-2 also presents information on the in-state versus out-of-state status of all registrants, but in a different way. Table 6-2 indicates the percentage of deceased registrants who were located out of state at the time of the baseline interview. Of those who were located out of state at the time of the baseline interview (n = 378), 6% were decedents. Of those who were located in state at the time of the baseline interview (n = 3,903), 6% were decedents. Thus, it appears as if the percentage of deceased registrants was about the same both on site and off site. However, it is unknown whether this was indeed the case or if differential rates were confounded by the researcher's ability to identify the potential registrants on site and off site.

    Table 6-1.--Comparison of living and deceased by location and sex.

    Status

    		 All Males Females
    Living 4,042 1,942 (48%)* 2,100 (52%)*
    In state 3,686 (91%)† 1,789 (49%)* 1,897 (51%)*
    Out of state 356 (9%)† 153 (43%)* 203 (57%)*
    Deceased 239 140 (59%)* 99 (41%)*
    In state 217 (91%)† 125 (58%)* 92 (42%)*
    Out of state 22 (9%)† 15 (68%)* 7 (32%)*

    *Row percent.

    †Column percent.

    Table 6-2.Comparison of location for deceased and living.

    Location All Males Females
    Out of State 378 168 (44%)* 210 (56%)*
    Deceased 22 (6%)† 15 (68%)* 7 (32%)*
    Living 356 (94%)† 153 (43%)* 203 (57%)*
    In State 3,903 1,914 (49%)* 1,989 (51%)*
    Deceased 217 (6%)† 125 (58%)* 92 (42%)*
    Living 3,686 (94%)† 1,789 (49%)* 1,897 (51%)*

    *Row percent.

    †Column percent.

    Based on the information available, it was not possible to determine the total number of persons who were eligible for participation at the time of the baseline interview, but who were not identified. For this reason, it is not possible to determine the comparability of the identification and location rates according to in- and out-of-state status. Only if the identification and location rates were the same for all registrantsliving and deceasedwould the calculated mortality rate for the TCE Subregistry population be accurate. The comparisons of the overall population mortality rate presented in this section were based on this assumption: the location rates were the same for deceased and living registrants. However, such a comparison of the TCE Subregistry mortality rate to the national rate should be interpreted with caution; although the previous calculations suggest this assumption is valid, it cannot be fully investigated. This section also presents the calculation of the percentage of the total number of deceased registrants who died from a given cause by age groups and provides comparisons with national ratios.

    STATISICAL ANALYSES

    STATISTICAL ANALYSES

    Variables

    Table 6-3 provides a summary of information by age on the underlying cause of death for each of the 239 deceased registrants. The variables--age and underlying cause of death--were used in the related statistical analyses. Control variables other than age were considered for the analyses, but the cell sizes were too small and results unreliable when other variables (such as death in or out of state) were included in the statistical models. The variable sex, however, was considered and included; the analyses were run using only males and only females.

    Statistical Methods

    Life table analysis (26,27) was used to evaluate mortality rates for the persons who were eligible for the TCE Subregistry at the time of the baseline interview. Standardized mortality ratios (SMRs) and proportional mortality ratios (PMRs) were computed using published methods (27). SMRs were calculated by dividing the observed number of deaths by the expected number, after stratification to adjust for the confounding effects of age and year of death. Two-sided 95% confidence intervals were computed for each cause-specific SMR, with use of the Byar approximation for eight deaths or more and Fisher's exact method for fewer than eight deaths (28). The PMRs were basically a comparison of death proportions, for example the proportion of deaths due to cancer divided by all deaths in the TCE Subregistry file versus the same proportion in the general population. Two-sided 95% confidence intervals were calculated as described previously. The calculations were standardized for age, time interval, and sex. The appropriate U.S. population rates (white only) were used as the reference group for comparison.

    RESULTS

    Table 6-3 provides a summary of the underlying causes of death for all 239 decedents by age group, and includes cancer deaths by ICD-9 codes. An examination of the TCE Subregistry deaths by year of death (Table 6-4) indicates that 5 (2%) of the 235 occurred from 1970 through 1974, 20 (8%) from 1975 through 1979, 79 (34%) from 1980 through 1984, 123 (52%) from 1985 through 1989, and 8 (3%) from January through August 1990. This distribution

    Table 6-3.Summary of death certificate data.

    Cause of Death

    		 Age Groups (years)
    24

    N (%)*

    		 25 - 34

    N (%)

    		 35 - 44

    N (%)

    		 45 - 54

    N (%)

    		 55 - 64

    N (%)

    		 65 - 74

    N (%)

    		 75

    N (%)

    		 Total
    All causes 13 (5)† 7 (3)† 6 (2)† 19 (8)† 46 (19)† 56 (23)† 92 (38)† 239
    Malignant neoplasms

    (ICD 140 - 208)

    		 0 (0) 1 (14) 2 (33) 3 (16) 15 (33) 18 (32) 16 (17) 55 (23)
    Digestive 0 0 0 1 3 5 4 13 (5)
    Respiratory 0 0 1 1 4 6 4 16 (7)
    Breast 0 0 0 1 2 3 0 6 (2)
    Genital, F 0 0 0 0 0 1 0 1 (1)
    Genital, M 0 0 0 0 1 0 1 2 (1)
    Urinary 0 0 0 0 1 0 0 1 (<1)
    Others 0 1 1 0 2 2 2 8 (3)
    Lymphatic 0 0 0 0 2 1 5 8 (3)
    Cardiovascular disease

    (ICD 390-459)

    		 1 (8) 0 0 10 (53) 21 (46) 25 (45) 47 (51) 104 (44)
    Chronic respiratory disease

    (ICD 460-496)

    		 1 (8) 0 0 2 (11) 2 (4) 5 (9) 12 (13) 22 (9)
    Accidents and adverse effects

    (ICD E800-E949)

    Ill-defined

    (ICD 780-799)

    		 10 (77) 6 (86) 4 (67) 3 (16) 3 (6) 2 (4) 2 (2) 30 (12)
    All others

    (ICD 250, 280-389, 520-

    739)

    		 1 (8) 0 0 1 (5) 5 (11) 6 (11) 15 (16) 28 (12)

    *Number of deaths by cause within age group, % - percentage of deaths by cause within age group.

    †Percentage of total deaths by age group.

    was quite different from the expected numbers based on national data; the national rates were consistent across time periods.

    The PMR analyses did not yield significant results at the p  0.01 level. There were significant increases at the p 0.05 level for some age groups by sex for varying time periods. Those results are not reported; in all cases the significance was based on two or fewer cases, making interpretation tenuous.

    The results for the statistical analyses which are presented in Tables 6-4 and 6-5 were restricted to registrants who were 15 years of age or older at the time of death because of the sparsity of data for individuals below this age. Other cases were omitted from the analyses in Tables 6-4 and 6-5 because there were incomplete data on the variables of interest. These two sources of missing data left 235 (out of 239) deceased registrants and 3,463 (out of 3,565) living registrants for the analyses.

    Table 6-4.Summary of standardized mortality ratios (SMRs).

    Years Observed (N) Expected (N) SMR*
    1970-1974 5 7 0.73
    1975-1979 20 47 0.42 †
    1980-1984 79 89 0.89
    1985-1989 123 69 1.84§
    1970-1990 235 209 1.12

    *All values were standardized for age, time, and sex.

    †Significantly different at p 0.05.

    §Significantly different at p 0.01.

    The number of deaths from all causes that occurred from the beginning date of exposure through data collection was 235. The expected number of deaths was 209, based on national rates, for the years 1970 through 1990. The calculated SMR (1.12, 95% CI= .98, 1.27) was not statistically significant. The calculated SMRs (Table 6-4) for the time periods consistently increase from 0.42 for 1975 through 1979 to 1.84 for 1985 through 1989. It was of interest that in the period 1975 through 1979 only two of the age groups (20 through 24 years and 55 through 59 years) had SMRs >1.0; in the 1985 through 1989 period, all but one age group (80 through 84 years) had SMRs >1.0. The SMRs that were <1.0 (some of which were significant) for the earlier time periods might reflect decreased success in identifying deaths that were further removed from the time of data collection, as opposed to a truly reduced rate. The number of deaths for the more recent time periods might reflect more closely the true number of deaths within the time periods.

    Table 6-5 displays the significant results for the comparison of mortality ratios by cause of death and age group. As is indicated, two significant differences (two-sided Poisson - p.01) were found for the years 1975 through 1979; they indicate fewer deaths for all causes and heart- related causes in the TCE population. Significant increases were found for the time period 1985 through 1989 for the age groups 25 through 29 years (all deaths); 60 through 64 years (all deaths);

    Table 6-5.Significant results of life table analysis (analysis restricted to registrants 15 years of age or older at time of death).

    Age Group Cause of Death Observed Expected SMR*
    1975 - 1979
    All ages Diseases of the heart 7 18 0.38
    All ages All causes 20 47 0.42
    1985 - 1989
    75 - 79 Malignant neoplasm of

    respiratory system

    		 4 <1 6.62
    75 - 79 Other diseases of

    circulatory system

    		 5 1 4.64
    All ages Diseases of the heart 39 25 1.59
    All ages Accidents 11 2 3.91
    25 - 29 All causes 4 1 6.28
    60 - 64 All causes 15 6 2.67
    75 - 79 All causes 24 9 2.54
    All ages All causes 123 67 1.84

    *SMR (standardized mortality ratio) significant, two-sided Poisson - p 0.01.

    75 through 79 years (malignant neoplasm of respiratory system, other disease of the circulatory system, and all deaths); and the total population (disease of the heart, accidents, and all diseases).

    To summarize, the analyses do not reveal a significant excess number of deaths in the time period 1970 through 1985. There appeared to be a deficiency in identification of deaths occurring in the 1975 through 1979 time period based on the low rate for this period. The number of deaths, however, steadily increased the closer the time period was to the time of data collection. For the time period beginning in 1985, there appeared to be a significant increase in the reporting of deaths for the 75 years and older age group for malignant neoplasms of the respiratory system, heart disease, and other circulatory diseases, as well as for all causes. There were also significant increases in all causes of death for the 25 through 29 and 60 through 64 years of age groups and for all age groups combined.

    1 In general, in-state cases include registrants who resided at sites located in the states of Illinois, Indiana, or Michigan. In-state decedents died in the state where exposure occurred (that is Illinois, Indiana, or Michigan). In-state registrants who were living at the time of the baseline interview resided somewhere in the statepossibly in the site areaat the time of the baseline interview. Out-of-state cases either resided or died in a state other than the state where the exposure occurred.

top