Absence of Synergism between Exposure to Asbestos and Cigarette Smoking in Asbestosis'4 JONATHAN M. SAMET, GARY R. EPLER, EDWARD A. GAENSLER, and BERNARD ROSNER
SUMMARY To assess both independent and synergistic effects of exposure to asbestos and cigarette smoking on the development of asbestosis, survey data from 4 groups of workers exposed to asbestos were analyzed with multivariate statistical models. Survey methods were standardized and included for the 383 subjects a respiratory symptoms questionnaire, occupational history, physical examination, pulmonary function testing, and a chest radiograph. Exposure to asbestos and cigarette smoking were assessed by questionnaire. Synergism between the 2 exposures was not present for previously identified manifestations of asbestosis including bilateral fine crackles, clubbing, dyspnea, radiographic abnormality, decreased forced vital capacity, and decreased single-breath diffusing capacity of the lung for CO. However, additive, independent effects of these 2 exposures were present for each of these parameters.
Introduction Synergistic relationships between occupational exposures a n d cigarette smoking have p r o f o u n d implications for b o t h industry a n d smoking workers. O c c u p a t i o n a l exposure to asbestos markedly increases the risk of l u n g cancer for cigarette smokers, a n d this increased risk is best explained by a multiplicative interaction between t h e 2 exposures (1-4). T h e i m p o r t a n c e (Received in original form November 22, 1978 and in revised form March 28,1979) 1 From the Department of Medicine, School of Medicine, University of New Mexico, Albuquerque, N. M.; the Channing Laboratory, and Departments of Preventive Medicine and Medicine, Harvard Medical School, and the Peter Bent Brigham Hospital Division of the Affiliated Hospitals Center, Inc., and the Thoracic Services, Departments of Medicine and Surgery, Boston University School of Medicine, Boston, Mass. 2 Supported in part by Research Training Grant no. HL-05998, Program Project Grant no. HL-19717, Career Award HL-1173, and contract no. l-HR-53028, all from the National Heart, Lung, and Blood Institute.
of such interactions in t h e pathogenesis of asbestosis is unclear (5-9). T o assess b o t h i n d e p e n d e n t a n d synergistic effects of exposure to asbestos a n d cigarette smoking in the d e v e l o p m e n t of asbestosis, w e applied multivariate statistical models to survey d a t a from 4 groups of workers exposed to asbestos. T h e models e x a m i n e d the relationships between these 2 exposures a n d previously identified clinical, radiographic, a n d physiologic manifestations of asbestosis, including: bilateral fine crackles, clubbing, dypsnea, r a d i o g r a p h i c abnormality, decreased forced vital capacity (FVC), a n d decreased single-breath diffusing capacity of the l u n g for C O ( D L C O ) (5, 10, 11). T h e results do n o t confirm synergism between cigarette smoking a n d exposure to asbestos. Materials and Methods Survey populations. Since 1966, we have conducted 3 Presented in part at the International Conference on Occupational Lung Disease, San Francisco, Calif., February 27-March 2,1979. 4 Requests for reprints should be addressed to Dr. Edward A. Gaensler, Boston University School of Medicine, 80 E. Concord St., Boston, Mass. 02118.
AMERICAN REVIEW OF RESPIRATORY DISEASE, VOLUME 120. 1979
75
76
SAMET, EPLER, GAENSLER, AND ROSNER
pulmonary surveys of employees from several asbestos-related industries. For the present analyses, 4 groups of workers exposed to asbestos were chosen: those from shipyard A, shipyard B, factory A, and factory B. All pipe coverers employed at shipyard A were studied. Asbestos-insulated pipes had been used in this yard for new ship construction since the late 1930s. Participation was obligatory for all pipe coverers employed at shipyard B, and management selected an additional group of paintercleaners who had cleaned the work area of the pipe coverers. Over-all participation was 92 per cent. Asbestos insulation had been in use since 1952. In addition to installing new pipes, these workers were also involved in "ripping out" insulation and refitting old ships. At factory A plant participation was voluntary for all workers, 90 per cent of whom were studied. The employees exposed to asbestos at this plant had manufactured asbestos filters in a special section with high exposure from 1952 to 1956. Factory B is a fireproofing product manufacturing plant where asbestos had been used extensively from the early 1930s to 1975. The survey was voluntary for all employees, including workers exposed directly and indirectly to asbestos as well as administrative personnel. Approximately 85 per cent of the employees participated. For these last 2 groups, only employees with a job description indicating exposure to asbestos were included in this analysis. Nonexposed employees, such as administrative personnel, were excluded. For all 4 groups, we used only data from the initial study. Survey methods. A questionnaire adapted from that of Fletcher and co-workers (12) concerning dyspnea, cough, and sputum production was administered to each worker. In addition, the following occupational information was recorded by trained pulmonary clinicians: present and past jobs; total years employed; years since first employment; total years of exposure to asbestos, with level of exposure; and years since first known exposure. Adventitious lung sounds were coded by quality and quantity of crackles (13). Clubbing was recorded as absent, possible, or definite. Physiologic studies included the FVC and forced expiratory volume in I s (FEVj). A water-sealed spirometer was used for these determinations. The subject was standing, and noseclips were used. The best of 3 FVC values was chosen. The DLCO was determined according to the method of Ogilvie and coworkers (14) with the use of an automated procedure from our laboratory (15). Only tests with an inspired volume of more than 90 per cent of the previously determined FVC were used, and the mean of 2 tests was calculated. Each chest roentgenogram was coded by a panel of pulmonary clinicians, who reached a consensus at a single reading, according to a modified version of the International Labor Office UlCC/Cincinnati (ILO U/C) radiographic classification (16,17).
Data analysis. To determine the independent effects of cigarette smoking and of exposure to asbestos, and to test for possible synergism, linear multiple regression (18) was used for FVC, DLCO, and radiographic profusion; multiple logistic regression (19) was used for the presence of rales, clubbing, and dyspnea. Multiple logistic regression was used for the latter discrete variables because the independent variables are more likely to be related to the logistic transformation of the probability of their occurrence than to the nontransformed probability. The linear multiple regression model used was: y = a + p^ + £ 2 x 2 + £3X^2 + c, where x1 and x 2 are indices of cigarette smoking and asbestos exposure, and y is the FVC, DLCO, or radiographic profusion. Least squares regression was used to estimate a and the p1 and supplied a, a constant term, and the pv the regression coefficients. The latter represent estimates of the change in y per unit change in a particular x± with all other x± held constant. The multiple correlation coefficient squared (R2) is a measure of the ability of the independent variables to explain the variance of the dependent variable. The contribution of each independent variable to R2 can be assessed. In the model described, synergistic or antagonistic interactions between the 2 variables can be tested by entering the multiplicative term x 1 x 2 . If synergism is present, the regression coefficient for this interaction term indicates the magnitude of the effect and is significant. A step-up hierarchical model was used, which entered indices of smoking and exposure to asbestos and then their interaction. Standard programs of the Statistical Package for the Social Sciences were used (20). Before being used in the regression model, the FVC and the DLCO were standardized for age and for height with regression equations obtained from 140 male blue collar workers employed at the same plants but with no exposure to asbestos. Members of this group had no respiratory symptoms and had a normal chest film. The age and height regressions for the FVC and DLCO in the group were as follows: FVC (L) = 0.119* height - 0.023* age - 2.43 DLCO (ml/min/mm) = 0.51* height - 0.11* age -f 1.50 A per cent predicted value was then calculated for each worker by dividing his actual by his predicted value. The values for black subjects were scaled upward by 1.132 (21). The over-all ILO U/C radiographic profusion was made suitable for the regression model by transformation to a normalized z score. The subjects were rank ordered by their radiographic profusion; the ranks were converted into a cumulative frequency distribution, and finally, each rank was assigned a score from a table of areas under a standard normal curve (22). In the linear multiple regression analyses, the dependent variables (y) included the FVC per cent predicted value, the DLCO per cent predicted value,
77
CIGARETTE SMOKING AND EXPOSURE TO ASBESTOS
and the radiographic score. The independent variables (x1) included (1) the log of lifetime pack-years of cigarettes increased by 1, so that for nonsmokers this variable took on the value of zero, and (2) either the total years of exposure to asbestos or the years elapsed since first exposure, as an index of exposure to asbestos. The product of these 2 indices was used to test for a synergistic interaction. The logarithmic transformation of lifetime pack-years of cigarettes was used because of the skewed distribution of the nontransformed variable. The regressions were done separately for each survey group, for the entire population, and for cigarette-smoking groups. When the survey groups were pooled, adjustment for plant-toplant variation in exposure was accomplished by creating interaction terms between the exposure index and indicator variables for the plants. The plant-exposure interaction terms were entered after the main smoking and exposure variables and before the exposure-cigarette interaction. The logistic model used for clubbing, rales, and dyspnea regressed the logit transformation of the probability of an abnormality on the independent variables. As in linear multiple regression analysis, the significance of the effect of each independent variable can be tested. For the logit analyses the survey groups were pooled. The independent variables were entered hierarchically, as in the linear multiple regression model, with the logarithm of lifetime cigarette smoking and years since first exposure preceding their interaction term and the plant-exposure interaction terms.
Results Data were available for 409 survey subjects, of whom 383 underwent complete physiologic testing. The mean age and smoking habits of the
4 survey groups were similar, as was the prevalence of radiographic abnormality (table 1). T h e 2 shipyard groups had sustained nearly identical exposure to asbestos. Exposure in factory A was of shorter duration than that in factory B, but the mean number of years elapsed since first exposure in factory A exceeded that in factory B (table 1). The multiple regression analyses were first performed separately for each plant. Both total years of exposure to asbestos and years since first exposure to asbestos were evaluated as exposure indices. Because the multiple R 2 tended to be higher with the latter, and because the results with regard to synergism were not affected by the choice of exposure index, only the results obtained with years since first exposure are reported. A significant effect of exposure to asbestos on FVC was found in both shipyards and factory B, whereas a significant effect of cigarette smoking was present only in shipyard B (table 2). The interaction between cigarette smoking and years elapsed since first exposure to asbestos was not significant in any of the survey populations. T h e regression coefficients for exposure to asbestos were identical in the 2 shipyard groups (table 2). Cigarette smoking was associated with a decreased DLCO in each of the survey groups (table 3). An additional effect of exposure to asbestos was found only in factory A and factory B. In factory B, where exposure to asbestos had been sustained longer than in factory A, the contribution of exposure to asbestos to R 2 was nearly twice that of cigarette consumption.
TABLE 1 CHARACTERISTICS OF STUDY SUBJECTS Shipyard A No. of subjects* Mean age, yr Cigarette smokers, % Present Past Never Over-all I LO U/C Profusion .t, % 0/0-0/1 1/0-1/2 > 2/1 Mean exposure to asbestos, yr Mean no. years since first exposure to asbestos
Shipyard B
Factory A
Factory B
89
94
42.6
45.4
44.9
41.5
60
56 27
40 42
52 30
16
18
18
70 25
76 12
12.2
98
30 10 59
128
11.6
11.0
66 21 13 3.4
17.0
14.6
15.1
31
10
5
12 7.6
Complete radiographic and physiologic data were not available for 5 subjects from Shipyard A, 5 from Shipyard B, 9 from Factory A, and 7 from Factory B. tin each population, 95 to 100 per cent of the opacities were of the s, t, u variety.
78
SAMET, EPLER, GAENSLER, AND ROSNER
TABLE 2 R E G R E S S I O N * OF FORCED VITAL CAPACITY, EXPRESSED AS A PERCENTAGE OF THE PREDICTED V A L U E , ON LIFETIME CIGARETTE CONSUMPTION, Y E A R S SINCE FIRST EXPOSURE TO A S B E S T O S , A N D THEIR INTERACTION Cigarettes
Coefficient
Interaction
Asbestos Regression
Regression AR2
Shipyard A
NSt
_
Shipyard B
-4.92§
0.07
Factory A
NS
Factory B
NS
-
Coefficient
Regression AR2
Coefficient
-0.39t
0.11
-0.35§
0.04
NS
-
NS
0.38
NS
NS -0.82$
NS
Each row represents the results of a separate stepwise regression. For each variable that attained significance at P < 0 . 0 5 , the regression coefficient and the change in R due t o that variable are given. The regression coefficients and P values are for that step at which all variables were significant. ' N o t significant. tp < 0.01 §p < 0 . 0 5 .
As with the FVC, synergism between cigarette consumption and exposure to asbestos was not present. Exposure to asbestos was the principal determinant of the radiographic score, but small additive effects of cigarette smoking, without synergism, were noted (table 4). In these analyses the contribution to R 2 of the interaction term was minimal (table 5). With the exception of the FVC in factory A, comparison of the regression coefficients for the interaction term with their standard errors revealed that none approached significance. In addition, the direction of effect indicated by the sign of the regression coefficient was not consistently that anticipated for synergism. When the 4 survey groups were pooled, only
additive effects on FVC, DLCO> a n d radiographic score of cigarette smoking and exposure to asbestos were found (results not shown). T o confirm further the absence of synergism, the regression analyses described were done separately within cigarette-smoking groups. T h e regression coefficients on exposure to asbestos for FVC and DLCO tended to be highest in those who had never smoked; they were the same for the radiographic score in those who had never smoked and in past smokers (table 6). Members of the 3 cigarette-smoking categories did not differ in mean years of exposure to asbestos, mean years elapsed since first exposure to asbestos, or type of current job. The results of the logit model for the pres-
TABLE 3 R E G R E S S I O N * OF DIFFUSING CAPACITY OF THE LUNG FOR CO, E X P R E S S E D AS A PERCENTAGE OF THE PREDICTED V A L U E , ON LIFETIME CIGARETTE CONSUMPTION, Y E A R S SINCE FIRST EXPOSURE TO A S B E S T O S , A N D THEIR INTERACTION Cigarettes Regression Coefficient Shipyard A Shipyard B Factory A Factory B
-14.0t -11.6t -10.6§ -11.3t
*For explanation, see table 2 . t p < 0.01. + N o t significaM §p
SUMMARY To assess both independent and synergistic effects of exposure to asbestos and cigarette smoking on the development of asbestosis, survey data from 4 groups of workers exposed to asbestos were analyzed with multivariate statistical models. Survey methods were standardized and included for the 383 subjects a respiratory symptoms questionnaire, occupational history, physical examination, pulmonary function testing, and a chest radiograph. Exposure to asbestos and cigarette smoking were assessed by questionnaire. Synergism between the 2 exposures was not present for previously identified manifestations of asbestosis including bilateral fine crackles, clubbing, dyspnea, radiographic abnormality, decreased forced vital capacity, and decreased single-breath diffusing capacity of the lung for CO. However, additive, independent effects of these 2 exposures were present for each of these parameters.
Introduction Synergistic relationships between occupational exposures a n d cigarette smoking have p r o f o u n d implications for b o t h industry a n d smoking workers. O c c u p a t i o n a l exposure to asbestos markedly increases the risk of l u n g cancer for cigarette smokers, a n d this increased risk is best explained by a multiplicative interaction between t h e 2 exposures (1-4). T h e i m p o r t a n c e (Received in original form November 22, 1978 and in revised form March 28,1979) 1 From the Department of Medicine, School of Medicine, University of New Mexico, Albuquerque, N. M.; the Channing Laboratory, and Departments of Preventive Medicine and Medicine, Harvard Medical School, and the Peter Bent Brigham Hospital Division of the Affiliated Hospitals Center, Inc., and the Thoracic Services, Departments of Medicine and Surgery, Boston University School of Medicine, Boston, Mass. 2 Supported in part by Research Training Grant no. HL-05998, Program Project Grant no. HL-19717, Career Award HL-1173, and contract no. l-HR-53028, all from the National Heart, Lung, and Blood Institute.
of such interactions in t h e pathogenesis of asbestosis is unclear (5-9). T o assess b o t h i n d e p e n d e n t a n d synergistic effects of exposure to asbestos a n d cigarette smoking in the d e v e l o p m e n t of asbestosis, w e applied multivariate statistical models to survey d a t a from 4 groups of workers exposed to asbestos. T h e models e x a m i n e d the relationships between these 2 exposures a n d previously identified clinical, radiographic, a n d physiologic manifestations of asbestosis, including: bilateral fine crackles, clubbing, dypsnea, r a d i o g r a p h i c abnormality, decreased forced vital capacity (FVC), a n d decreased single-breath diffusing capacity of the l u n g for C O ( D L C O ) (5, 10, 11). T h e results do n o t confirm synergism between cigarette smoking a n d exposure to asbestos. Materials and Methods Survey populations. Since 1966, we have conducted 3 Presented in part at the International Conference on Occupational Lung Disease, San Francisco, Calif., February 27-March 2,1979. 4 Requests for reprints should be addressed to Dr. Edward A. Gaensler, Boston University School of Medicine, 80 E. Concord St., Boston, Mass. 02118.
AMERICAN REVIEW OF RESPIRATORY DISEASE, VOLUME 120. 1979
75
76
SAMET, EPLER, GAENSLER, AND ROSNER
pulmonary surveys of employees from several asbestos-related industries. For the present analyses, 4 groups of workers exposed to asbestos were chosen: those from shipyard A, shipyard B, factory A, and factory B. All pipe coverers employed at shipyard A were studied. Asbestos-insulated pipes had been used in this yard for new ship construction since the late 1930s. Participation was obligatory for all pipe coverers employed at shipyard B, and management selected an additional group of paintercleaners who had cleaned the work area of the pipe coverers. Over-all participation was 92 per cent. Asbestos insulation had been in use since 1952. In addition to installing new pipes, these workers were also involved in "ripping out" insulation and refitting old ships. At factory A plant participation was voluntary for all workers, 90 per cent of whom were studied. The employees exposed to asbestos at this plant had manufactured asbestos filters in a special section with high exposure from 1952 to 1956. Factory B is a fireproofing product manufacturing plant where asbestos had been used extensively from the early 1930s to 1975. The survey was voluntary for all employees, including workers exposed directly and indirectly to asbestos as well as administrative personnel. Approximately 85 per cent of the employees participated. For these last 2 groups, only employees with a job description indicating exposure to asbestos were included in this analysis. Nonexposed employees, such as administrative personnel, were excluded. For all 4 groups, we used only data from the initial study. Survey methods. A questionnaire adapted from that of Fletcher and co-workers (12) concerning dyspnea, cough, and sputum production was administered to each worker. In addition, the following occupational information was recorded by trained pulmonary clinicians: present and past jobs; total years employed; years since first employment; total years of exposure to asbestos, with level of exposure; and years since first known exposure. Adventitious lung sounds were coded by quality and quantity of crackles (13). Clubbing was recorded as absent, possible, or definite. Physiologic studies included the FVC and forced expiratory volume in I s (FEVj). A water-sealed spirometer was used for these determinations. The subject was standing, and noseclips were used. The best of 3 FVC values was chosen. The DLCO was determined according to the method of Ogilvie and coworkers (14) with the use of an automated procedure from our laboratory (15). Only tests with an inspired volume of more than 90 per cent of the previously determined FVC were used, and the mean of 2 tests was calculated. Each chest roentgenogram was coded by a panel of pulmonary clinicians, who reached a consensus at a single reading, according to a modified version of the International Labor Office UlCC/Cincinnati (ILO U/C) radiographic classification (16,17).
Data analysis. To determine the independent effects of cigarette smoking and of exposure to asbestos, and to test for possible synergism, linear multiple regression (18) was used for FVC, DLCO, and radiographic profusion; multiple logistic regression (19) was used for the presence of rales, clubbing, and dyspnea. Multiple logistic regression was used for the latter discrete variables because the independent variables are more likely to be related to the logistic transformation of the probability of their occurrence than to the nontransformed probability. The linear multiple regression model used was: y = a + p^ + £ 2 x 2 + £3X^2 + c, where x1 and x 2 are indices of cigarette smoking and asbestos exposure, and y is the FVC, DLCO, or radiographic profusion. Least squares regression was used to estimate a and the p1 and supplied a, a constant term, and the pv the regression coefficients. The latter represent estimates of the change in y per unit change in a particular x± with all other x± held constant. The multiple correlation coefficient squared (R2) is a measure of the ability of the independent variables to explain the variance of the dependent variable. The contribution of each independent variable to R2 can be assessed. In the model described, synergistic or antagonistic interactions between the 2 variables can be tested by entering the multiplicative term x 1 x 2 . If synergism is present, the regression coefficient for this interaction term indicates the magnitude of the effect and is significant. A step-up hierarchical model was used, which entered indices of smoking and exposure to asbestos and then their interaction. Standard programs of the Statistical Package for the Social Sciences were used (20). Before being used in the regression model, the FVC and the DLCO were standardized for age and for height with regression equations obtained from 140 male blue collar workers employed at the same plants but with no exposure to asbestos. Members of this group had no respiratory symptoms and had a normal chest film. The age and height regressions for the FVC and DLCO in the group were as follows: FVC (L) = 0.119* height - 0.023* age - 2.43 DLCO (ml/min/mm) = 0.51* height - 0.11* age -f 1.50 A per cent predicted value was then calculated for each worker by dividing his actual by his predicted value. The values for black subjects were scaled upward by 1.132 (21). The over-all ILO U/C radiographic profusion was made suitable for the regression model by transformation to a normalized z score. The subjects were rank ordered by their radiographic profusion; the ranks were converted into a cumulative frequency distribution, and finally, each rank was assigned a score from a table of areas under a standard normal curve (22). In the linear multiple regression analyses, the dependent variables (y) included the FVC per cent predicted value, the DLCO per cent predicted value,
77
CIGARETTE SMOKING AND EXPOSURE TO ASBESTOS
and the radiographic score. The independent variables (x1) included (1) the log of lifetime pack-years of cigarettes increased by 1, so that for nonsmokers this variable took on the value of zero, and (2) either the total years of exposure to asbestos or the years elapsed since first exposure, as an index of exposure to asbestos. The product of these 2 indices was used to test for a synergistic interaction. The logarithmic transformation of lifetime pack-years of cigarettes was used because of the skewed distribution of the nontransformed variable. The regressions were done separately for each survey group, for the entire population, and for cigarette-smoking groups. When the survey groups were pooled, adjustment for plant-toplant variation in exposure was accomplished by creating interaction terms between the exposure index and indicator variables for the plants. The plant-exposure interaction terms were entered after the main smoking and exposure variables and before the exposure-cigarette interaction. The logistic model used for clubbing, rales, and dyspnea regressed the logit transformation of the probability of an abnormality on the independent variables. As in linear multiple regression analysis, the significance of the effect of each independent variable can be tested. For the logit analyses the survey groups were pooled. The independent variables were entered hierarchically, as in the linear multiple regression model, with the logarithm of lifetime cigarette smoking and years since first exposure preceding their interaction term and the plant-exposure interaction terms.
Results Data were available for 409 survey subjects, of whom 383 underwent complete physiologic testing. The mean age and smoking habits of the
4 survey groups were similar, as was the prevalence of radiographic abnormality (table 1). T h e 2 shipyard groups had sustained nearly identical exposure to asbestos. Exposure in factory A was of shorter duration than that in factory B, but the mean number of years elapsed since first exposure in factory A exceeded that in factory B (table 1). The multiple regression analyses were first performed separately for each plant. Both total years of exposure to asbestos and years since first exposure to asbestos were evaluated as exposure indices. Because the multiple R 2 tended to be higher with the latter, and because the results with regard to synergism were not affected by the choice of exposure index, only the results obtained with years since first exposure are reported. A significant effect of exposure to asbestos on FVC was found in both shipyards and factory B, whereas a significant effect of cigarette smoking was present only in shipyard B (table 2). The interaction between cigarette smoking and years elapsed since first exposure to asbestos was not significant in any of the survey populations. T h e regression coefficients for exposure to asbestos were identical in the 2 shipyard groups (table 2). Cigarette smoking was associated with a decreased DLCO in each of the survey groups (table 3). An additional effect of exposure to asbestos was found only in factory A and factory B. In factory B, where exposure to asbestos had been sustained longer than in factory A, the contribution of exposure to asbestos to R 2 was nearly twice that of cigarette consumption.
TABLE 1 CHARACTERISTICS OF STUDY SUBJECTS Shipyard A No. of subjects* Mean age, yr Cigarette smokers, % Present Past Never Over-all I LO U/C Profusion .t, % 0/0-0/1 1/0-1/2 > 2/1 Mean exposure to asbestos, yr Mean no. years since first exposure to asbestos
Shipyard B
Factory A
Factory B
89
94
42.6
45.4
44.9
41.5
60
56 27
40 42
52 30
16
18
18
70 25
76 12
12.2
98
30 10 59
128
11.6
11.0
66 21 13 3.4
17.0
14.6
15.1
31
10
5
12 7.6
Complete radiographic and physiologic data were not available for 5 subjects from Shipyard A, 5 from Shipyard B, 9 from Factory A, and 7 from Factory B. tin each population, 95 to 100 per cent of the opacities were of the s, t, u variety.
78
SAMET, EPLER, GAENSLER, AND ROSNER
TABLE 2 R E G R E S S I O N * OF FORCED VITAL CAPACITY, EXPRESSED AS A PERCENTAGE OF THE PREDICTED V A L U E , ON LIFETIME CIGARETTE CONSUMPTION, Y E A R S SINCE FIRST EXPOSURE TO A S B E S T O S , A N D THEIR INTERACTION Cigarettes
Coefficient
Interaction
Asbestos Regression
Regression AR2
Shipyard A
NSt
_
Shipyard B
-4.92§
0.07
Factory A
NS
Factory B
NS
-
Coefficient
Regression AR2
Coefficient
-0.39t
0.11
-0.35§
0.04
NS
-
NS
0.38
NS
NS -0.82$
NS
Each row represents the results of a separate stepwise regression. For each variable that attained significance at P < 0 . 0 5 , the regression coefficient and the change in R due t o that variable are given. The regression coefficients and P values are for that step at which all variables were significant. ' N o t significant. tp < 0.01 §p < 0 . 0 5 .
As with the FVC, synergism between cigarette consumption and exposure to asbestos was not present. Exposure to asbestos was the principal determinant of the radiographic score, but small additive effects of cigarette smoking, without synergism, were noted (table 4). In these analyses the contribution to R 2 of the interaction term was minimal (table 5). With the exception of the FVC in factory A, comparison of the regression coefficients for the interaction term with their standard errors revealed that none approached significance. In addition, the direction of effect indicated by the sign of the regression coefficient was not consistently that anticipated for synergism. When the 4 survey groups were pooled, only
additive effects on FVC, DLCO> a n d radiographic score of cigarette smoking and exposure to asbestos were found (results not shown). T o confirm further the absence of synergism, the regression analyses described were done separately within cigarette-smoking groups. T h e regression coefficients on exposure to asbestos for FVC and DLCO tended to be highest in those who had never smoked; they were the same for the radiographic score in those who had never smoked and in past smokers (table 6). Members of the 3 cigarette-smoking categories did not differ in mean years of exposure to asbestos, mean years elapsed since first exposure to asbestos, or type of current job. The results of the logit model for the pres-
TABLE 3 R E G R E S S I O N * OF DIFFUSING CAPACITY OF THE LUNG FOR CO, E X P R E S S E D AS A PERCENTAGE OF THE PREDICTED V A L U E , ON LIFETIME CIGARETTE CONSUMPTION, Y E A R S SINCE FIRST EXPOSURE TO A S B E S T O S , A N D THEIR INTERACTION Cigarettes Regression Coefficient Shipyard A Shipyard B Factory A Factory B
-14.0t -11.6t -10.6§ -11.3t
*For explanation, see table 2 . t p < 0.01. + N o t significaM §p
