Document RpyKYj25Yp9RqvDw2e9aROE2v

Dgvrvev^ / American Journal of Industrial Medicine 4:399-419 (1983) Exposures and Mortality Among Chrysotile Asbestos Workers. Part I: Exposure Estimates John M. Dement, PhD, Robert L. Harris, Jr, PhD, Michael J. Symons, PhD, and Carl M. Shy, md, DrPH A detailed study of plant processes and dust control methods over the period 1930-1975 was conducted in an asbestos textile plant processing chrysotile. Linear statistical models for reconstructing historic dust exposure levels, taking into account textile processes, dust control measures, and job assignments, were developed. Parameters of these statistical models were estimated using 5,952 industrial hygiene sampling measurements covering the period 1930-1975. Por most textile operations, exposure levels were signifi cantly reduced by about 1940, when most engineering dust control measures were in place. Results of the exposure estimates indicated "precontrol" exposure levels to range from 3 to 78 fibers/cc with typical levels well above 10 fibers/cc. After textile operations were provided with dust control measures, estimated exposure levels ranged from 3 to 17 fibers/cc and were usually in the range of 5 to 10 fibers/cc. These exposure estimates were combined with an assessment of mortality among workers at this plant to investi gate exposure-response relationships. Exposure-response results are presented in the companion manuscript in this volume. - Key words: chrysolite, asbestos, textile, exposures, controls INTRODUCTION The term "asbestos" is the generic name for a group of naturally occurring min eral silicates that have commercial value due to their fibrous morphology. There are six fibrous silicates currently classified as asbestos by various governmental and industrial groups. These include the fibrous serpentine, mineral chrysotile, and the fibrous amphibole minerals crocidolite, amosile, anthophyllite.sactinolite, and tremolite. In the United States, more than 95% of all asbestos used is of the chrysotile variety [Clifton, 1975). * University of North Carolina. School of Public Health, Chapel Hill. Address reprint requests to John M. Dement, National Institute of Environmental Health Sciences, P.O. Box 12233, Mail Drop 19-02. Research Triangle Park, NC 27709. Accepted for publication September 17, 1982. 1983 Alan R. Uss, Inc. 400 Dcmenl et al Epidemiologic studies have repeatedly demonstrated associations between expo sure to asbestos and increased mortality due to nonnialignant respiratory diseases (asbestosis), lung cancer, pleural and peritoneal mesothelioma, and in some studies, gas trointestinal cancer; these data have been adequately reviewed elsewhere [1ARC, 1977; Selikoff and Lee, 1978; NIOSH, I980J. In most industries, blends of several fibertypes typically have been used; thus epidemiologic studies of workers exposed to a single fiber type such aschrysotileare rare. Mortality among chrysotile miners and millers has been studied |McDonakl el al, 1980; Rubino et al, 1979; Nicholson et al, 1979]; however, only one small study has been conducted analyzing mortality among factory workers using chrysotile (Weiss, 1970]. To provide further information on mortality risks associated with chrysotile ex posures, a retrospective mortality study was conducted among a cohort of asbestos tex tile workers. Methods for reconstructing historic dust exposures, taking into account jobs and controls, were developed and combined with the mortality data to evaluate ex posure-response relationships for lung cancer and nonnialignant respiratory diseases. This paper describes plant processes and controls and methods used to estimate historic exposure levels. Mortality data and exposure-response analyses are presented in the companion manuscript in this volume. MATERIALS AND METHODS Description of the Facility The plant under study began production of asbestos packing materials for steam engines and pumps in 1896. In 1901 high pressure packings were added and asbestos textiles (yarns and cloths) were first produced in 1909. Production of asbestos textiles remained stable, and in 1937 operations were expanded to include manufacture of rubber rollers and rubber linings for pipes and tanks used in the paper, textile, and chemical industries. In 1955, additional facilities were added for production of roll covers used on flatwork ironers. Textile production operations at this plant were always concentrated in five buildings of one or two stories each. Fortunately, rubber production facilities were al ways physically separated from textile operations with relatively little mobility of em ployees between the two. These factors served to minimize confounding exposures in rubber operations on mortality results. Chrysotile is the only type of asbestos ever processed at this plant as a raw fiber. A small amount of crocidolite yarn was woven into a tape or made into a braided pack ing beginning in the 1950s until approximately 1975. Crocidolite was never carded, spun, or twisted. The total quantity of crocidolite ever processed was extremely small (less than 2000 pounds), and all weaving of crocidolite tapes was done wet on a single loom; thus exposures were low. According to company personnel, an average of ap proximately 6-8 million pounds of chrysotile were processed annually at this point. Chrysotile was received from Quebec, British Columbia, and Rhodesia. Plant Processes and Controls Processes for production of asbestos textiles al this plant were typical of this in dustry and remained remarkably unchanged for the period 1940-1975. Historically, the plant studied has been progressive in the application of modern dust control measures in asbestos textile production. This plant was the subject of an Exposures Among Chrysolilc Workers 401 extensive study of dust control measures and occupational exposures by the U.S. Public Health Service in January and February, 1937 [Page and Bloomfield, 1937], Page and Bloomfield stated regarding this plant: `These results [of the study] should not be interpreted as representing the maximum possible efficiency in the control of asbestos dust, but it is believed that they are representative of the best practice in this country at this lime." Subsequent to the Page and Bloomfield study, additional local exhaust ventila tion was added to twisting frames and cop (shuttle) winding machines. At the request of the company, an industrial hygiene study was conducted in November, 1937 to evalu ate the effectiveness of these added controls [Brown, I937J. Plant processes remained fairly constant over the period 1940-1975, with most engineering controls being in place by 1940. For the current study, an attempt was made to document changes in textile pro cesses and engineering control measures between 1930 and the study ending dale of 1975. This was a painstaking process involving detailed studies of Public Health Ser vice, insurance carrier, state and plant engineering records. These data were supple mented through discussions with long-time plant engineering personnel employed at the plant for the greater part of the study period. A summary of important control measures is given in Table I. Detailed process and control descriptions may be found elsewhere [Dement, 1980]. Most engineering controls in this plant were in place by 1940 and their effectiveness documented by the U.S. Public Health Service Study [Page and Bloom field. 1937], Photographs of textile operations demonstrating engineering controls in place before 1940 are shown in Figures 1 -- 10. In addition to local exhaust ventilation shown in these figures, some spinning frames were supplied with wetting rolls to dampen rovings prior to twisting. In addition, wet weaving w'as begun in approxi-* malely 1937. TABLE I. Siimniun of Engineering Controls by Operation in Textile (1930-1975) Operation Dates Summary of engineering controls Tiber picparnlinn Crushing Opening Blending Picking and combing Carding Spinning Twisting Winding Poster Universal Cop Weaving Inspection Bi aiding Waste recovery 1930-1965 1930-1975 1930-1945 I945-I9M 1964-1975 1930-1975 1935-1975 1930-1975 1938-1975 No local exhaust ventilation. Phased out in approximately 1965 Openers and combers enclosed and exhausted, screens exhausted Blending done in exhausted blending booths (canopy hood) Wending done on elevated station without ventilation Automatic blending with total enclosure and exhaust Machines enclosed and exhausted, screens exhausted raiding machines enclosed and exhausted at 11 points No local exhaust ventilation, some wet spinning since 1966. al though modest Bottom of twisting frame enclosed and exhausted 1937-1975 1930-1975 1937-1975 1937-1975 1937-1975 1930-1975 1930-1975 Individual conical hood around each spindle with exhaust No ventilation Exhaust on each spindle Combination of local exhaust and wet weaving, light weaving done wet since 1930 Exhaust at cleaning brushes Exhaust on some braiders, improved by better enclosure in 1958 Exhaust and ventilation on willowcr and roving reopener 402 Dement et al l ip. 2. I'ickcr machines showing vcmilalion a( charging hopper in 1937. Exposures Among Chrysotile Workers 403 Tig. 3. Picker machines in 1937 showing ventilation at discharge apron. Fig. 4. A 1937 photograph showing carding machine ventilation. I 'C. ft. Riiit! spinning I mine wiilioui local oxliniivi ventilmion in 1937. Exposures Among Chrysolite Workers 405 Pig. 7. Twisting frame in 1937 with partially installed downdraft ventilation system. 406 Dement et al Fi*;. 8 Spooking machine showing downdraft ventilation system in 1937. Exposures Among Clirysotile Workers 407 Pig. 9. A 1937 photograph of doth loom with local exhaust ventilation. 408 Dement et al Fig. 10. Cloth inspection table in 1937 showing ventilation at brushes. Industrial Hygiene Data All known industrial hygiene data were collected from many sources including the company insurance carrier, the State Board of Health, the U.S. Public Health Ser vice, and the Company sampling program. In total, 5,952 environmental samples were available covering the period 1930-1975. Prior to 1965 all samples were oT the impinger type; from 1965 until 1971 membrane filter samples were also collected. In 1971, the impinger method was discontinued and from that time on the membrane filter method was used exclusively. For purposes of evaluating worker exposure levels, it was necessary to estimate a conversion between the impinger and membrane filter data. Fortunately, there were two good sources of data for this conversion estimate specific for the plant under study. These included 120 paired impinger and membrane filler samples collected in 1965 by the U.S. Public Health Service and 986 concurrent samples by these two methods col lected in plant operations during 1968-1971. Linear correlation analyses were used to evaluate impinger-membrane filter conversionsTtaking into account plant operations by means of indicator variables. For the paired sample data, a square root transformation of X and Y variables was used to stabilize variances of impinger and membrane filter particle counts which are approxi mately described by a Poisson distribution (Leidel et al, 1979; Kleinbaum and Kupper, 1978]. For concurrent sample data, a log conversion of the dependent variable was used to take into account the approximate log-normal distribution of environmental sample Exposures Among Chrysolite Workers 409 data collected over a period of time [Breslin el al, 1967; Sherwood, 1971; Gale, 1965; Coenen, 1966]. Possible differences in conversions with calendar time were analyzed using the concurrent samples. More details concerning statistical models employed are given elsewhere [Dement, 1980J. Using the paired sample data, no statistically significant differences (a = 0.05) were observed by plant operation nor any significant differences in conversion factor with increasing impingcr concentration. For the paired sample data, a mean conversion of 2.9 fibers >5 /an/cc was estimated to be equivalent to one MPPCF. The concurrent sample data allowed an assessment of changes in conversion esti mates with time (1968-1971) and differences by plant operation. No significant trends with calendar time were noted nor were there any significant differences among plant operations except fiber preparation. For all operations except preparation, a mean con version of 2.5 fibers > 5 pin/cc per MPPCF was obtained. A higher conversion of 7.8 fibers > 5 ^m/cc per MPPCF was observed for preparation. Comparisons of conversions obtained using the two independent data sets indi cate excellent agreement except for fiber preparation where a higher conversion was ob tained. For the present study, impinger sample data were converted to fiber concentra tions using a conversion of 3 fibers longer than 5 pm per MPPCF for all operations except preparation where a factor of 8 was used. These factors are near the upper limits of the estimated 95% confidence intervals and are likely conservatively high conversion values (ie, may overestimate true fiber exposure levels). Exposure Estimates For evaluating exposure-response in the current study, it was necessary to have estimates of each worker's cumulative dust exposure throughout his working lifetime. Such estimates require detailed knowledge of all jobs held by the worker, the dates these were held, and estimates of average exposure levels for each job. Stated mathe matically, a worker's cumulative dust exposure (Ey) at any time (T) since initial em ployment may be expressed as follows: ' N Er = E Qtj i= 1 where C; = dust concentration for job i during calendar time period worker was employed in that job, l; = time spent in job i, and N = number of jobs held until time T. ` For this study, occupational histories for each worker were obtained from plant personnel records, and dust concentrations for each job were estimated using an expo sure model whose parameters were estimated using historical exposure data. Occupational histories. Detailed occupational histories for personnel at this plant were first recorded in approximately 1930. Occupational histories included both an indication of plant department and job held by the worker and beginning and end ing dates for each job. These records also contain dates when workers were absent and dates of termination and rehire. 410 Dement ct al In order to calculate a worker's dust exposure, a method of relating work history information to estimates of exposure was needed. For this purpose, a four-digit job coding scheme was developed through discussions with plant personnel and by review ing company personnel records. Fortunately, processes in this plant remained reason ably constant with the result being that job titles remained stable over the study period. All work histories were coded according to a four-digit coding scheme, and estimates of exposure levels were made specific to the job coding scheme. Exposure classification. Since exposure data are not available for each worker in the present cohort, methods of estimating exposure by plant job category were required. For estimating historic exposures, Woilowitz ct al (1970] suggested breaking a plant into "exposure zones'' with each zone ranked according to exposure. Corn and Esmen [1979] have proposed an extension of the exposure zone concept for purposes of designing industrial hygiene sampling programs. Corn and Esmen described an expo sure zone as a grouping of workers based on similarity of job and exposure characteris tics. An exposure zone may be a definable area of the plant or, alternatively, persons from different areas but sharing the requirement of similarity of job and exposure characteristics may form an exposure zone. By using exposure zones, available en vironmental data may be used more efficiently in estimating worker exposures, since all workers need not be monitored. While dividing a plant into zones of .similar dust exposure characteristics results in more homogeneous exposure classifications, there still may exist differences in jobs or tasks for selected workers within these zones which may have an effect on exposure. Esmen |I979] suggests that jobs may be further grouped into "uniform task cate gories." Workers within these task categories experience more homogeneous exposures and further reduce the variability of exposures for a given exposure z.one. Both the exposure zone concept and the uniform task concept were utilized to de velop exposure models for the present study. Plant textile production operations were assigned to one of nine exposure zones (Table II) according to similarity of processes and exposures. For textile operations at this plant, these zones correspond to physically defined areas of the plant. The uniform task concept suggested by Esmen was modified for the present study. All jobs within each exposure zone where asbestos exposure occurred were as signed to one of four Uniform .lob Categories (UJC) based upon tasks associated with the particular job. These categories are as follows: Category A: General area personnel. All personnel generally within a given plant area but not specifically associated with a particular machine or operation. These jobs include service personnel, fixers, helpers, oilers, elevator operators, doffers, clerks, supervisory personnel, etc. Category B: Machine operation. Persons who normally operate textile produc tion machinery. These jobs would include operators of opener machines, screens, card spinning frames, looms, winders, etc. Each exposure zone may contain several subcalegories of machine operators. Category C: Clean-up. Persons involved with cleaning production areas such as lloor sweeping, cleaning of machines, etc. Category D: Rawfilter handling. Persons handling or transporting raw unspun fiber or fiber waste. These jobs would include slockrollers, fly handlers, card grinders, etc. Exposures Among Clirysotile Workers TABLE II. Plant Operations and Exposure Zone Assignments Operation Exposure zone number Preparation/waste recovery Carding Ring and gang spinning Mule spinning Spooling (Foster winding) Twisting Universal winding ' Heavy weaving Draper weaving t 2 3 4 5 6 7 8 9 411 Within each exposure zone, estimates of exposure for the different UJCs require different considerations. Workers in category A move throughout a given exposure zone and are not usually associated with a particular process or piece of equipment. The best estimate of average exposure at any given time for these workers would be ob tained by pooling all dust samples within a given exposure zone to obtain an average ex posure for that zone. In a few instances, workers such as service personnel may spend part of their working day in more than one exposure zone. An estimated average expo sure would then be obtained as follows: 'all zones. Cifi where X a = average exposure for U.IC A, C; = average dust concentration for exposure zone i, and fj = usual fraction of working day spent in exposure zone i. The above estimation procedure is particularly applicable to the present study, as the textile operations being considered consist of many dust sources nearly uniformly distributed throughout a given exposure zone. Workers in categories B, C, and D require different exposure estimates. These workers also experience the average exposure for a given zone; however, workers in these categories also experience an increment in exposure attributable to machine operation or performance of especially dusty manual tasks such as cleaning machines and floors or handling of raw unspun fiber or fiber waste. For the category machine operator, there may be several subcategories. For example, cop winders and weavers would be subcategories of Uniform Job Category B for exposure zone 8. Exposure esti mates for workers in categories B-D may be expressed as follows: Xk = Eall; [Qfi] + Ea,, zones, i f ki where X = average exposure for UJC k, Ci = average dust concentration for exposure zone i, fi = usual fraction of working day spent in exposure zone i, and In = increment in average concentration in exposure zone i due to Uniform Job Category k. 412 Dement et at AH workers in categories B-D remain within one exposure zone during a working day; therefore, their exposure may be estimated by the following reduced expression: Xk = Q + Ik;. Statistical model for exposure zones. Using the exposure zones and Uniform Job Categories described above, log-linear models were developed to allow estimation of worker exposures in each exposure zone over time. Numerous investigators have demonstrated that environmental data collected over time are best described by a log-normal frequency distribution (Breslin et al, 1967; Sherwood, 1966, 1971; Jones and Brief, 1971; Hounam, 1965; Juda and Budzinski, 1967], Since regression techniques require normality of the dependent variable, the ex posure models for each zone were based on the log of the dust concentration values. For each exposure zone, the log concentration values may be described by the following model for any given point in time: Yj = Ek&kZjk where Yj = mean of log concentration values for exposure zone i, 0ik = multiple regression parameter for UJC k in exposure zone i, and Zjk = independent variable used to identify UJC k in exposure zone i. The variables Zjk are nominal variables used to identify each UJC and may take the values 0 and 1. Since the estimate of exposure for UJC A is the overall average for a* given exposure zone and estimates for other UJCs are represented as increments to this overall average, the Zjk are defined as follows: ZjA = 1 if exposure zone i: 0 otherwise, ZjB = 1 if UJC = B: 0 otherwise, Zjc = 1 if UJC = C: 0 otherwise, and Zjo = 1 if UJC = D: 0 otherwise. This model estimates exposures for UJCs B-D as the overall area average plus an increment or "effect''due to the particular UJC within the exposure zone. The estimates for UJC A are the exposure zone averages "corrected" for the effects of particularly dusty jobs within the exposure zones (UJCs B-D). The above model represents exposures at a given point in time; however, expo sures over time are affected by both changes in the manufacturing processes and engi neering control methods. To account for these changes, additional nominal variables were introduced into the model for each exposure zone. These variables then assumed the value of 0 and 1 depending on whether the process of engineering control change was present or absent at a given point in time. These variables are expressed in the model as follows: Yj -- L/ /JjkZjk + ofjjZjj kj Exposures Among Chrysolite Workers 413 where Yi, and Zj^ are as previously defined, ajj = multiple regression parameter for control j in exposure zone i, and Z-,j = independent variable used to represent dates of control changes for exposure zone i. Even after job categories and changes in engineering controls are accounted for in the regression models, there still may exist trends in dust concentration for an expo sure zone as a function of calendar time. These may be the result of such things as a gradual increase in production volume or perhaps residual interaction not accounted for by the models. To account for such trends, it is desirable to enter calendar time into the model. However, it is undesirable to enter calendar time into the model as continu ous variable (eg, exact year samples were collected), since this would tend to extrapolate any trends into calendar time periods where few samples may have been taken. To avoid this problem, calendar time was broken into discrete intervals depending on the particular exposure zone and assigned a nominal value. The entire model may then be expressed as follows: Yj = 22 fiik k + 1, orij Zjj + 22 6j( Zjt jt where all other variables are as previously defined, 6j, = multiple regression parameter for lime interval t in exposure zone i, and Zji = independent variable used to represent calendar time periods. Estimation of model parameters. Parameters of the previously described regres sion models for each txlile exposure zone were estimated by multiple linear regression using historic industrial hygiene sampling data. These data were obtained from a num ber of sources, as previously described. For estimating model parameters, each impinger or membrane filter sample was assigned to both an exposure zone and a Uniform Job Category. For the membrane filter samples, which were collected on individual workers, this assignment was done using the worker's department and job code. Using descriptions of sample location from industrial hygiene reports and plant engineering layout diagrams, each impinger sample was first assigned to one of the rou tine sampling points established for the company sampling program in about 1956. Using this sampling network, all impinger samples were then assigned to an appropri ate exposure zone Uniform Job Category. Uniform Job Categories were assigned ac cording to location of the particular sampling point within a given exposure zone. For UJC B in preparation (zone 1) and heavy weaving (zone 8), several subcategories of "machine operator" were required. These categories are as follows: Preparation (zone I). Fly machine operator, shaker screen operator, vertical machine operator, crusher operator, waste machine operator. Heavy weaving (zone 8). Beamers and creelers, winders, weavers, inspectors. 414 Dement et al The data concerning process modifications or changes in engineering controls were used to assign nominal values for these variables in the exposure models. A sum mary of significant process or control modifications considered in the models is shown in Table III. The parameters for each exposure zone model were estimated using typical least squares filling procedures. For each exposure zone, the variables for UJC and engi neering controls were first entered into the model along with a continuous variable (t) representing the calendar year the samples were collected. This latter variable was used to determine if any important residual trends of dust concentration with calendar time existed after all other main effect variables were ill the models. The significance of this time variable was tested using a partial F statistic. If the time variable was not signifi cant (p > 0.05), the model was refitted without this variable. Only carding and ring spinning had residual lime trends in the exposure data. For all analyses, impinger con centrations were converted to fiber concentrations using the conversions previously de rived. Each exposure model used the log of the dust concentration data. A significance level of 0.10 was used for inclusion of any variable in the models. Nearly all estimated regression parameters we:e significant at the 0.05 level, however. The lower level of significance of 0.10 was chosen for considerations of validity of the models in accounting for jobs and controls. Mean exposures and confidence intervals. The regression models incorporate the assumption of log-normality of dust concentration values. Estimated model pa rameters for the statistical models can be used to predict the mean for the log of dust concentration values and, by exponentiating these values, geometric mean dust expo sures are estimated. While the regression models work well using the log normality assumption, others have shown that the arithmetic mean may be the best estimator of the time aver aged concentration to which workers are exposed [Sherwood, 1966; Coenen, 1966, 1970; Leidel and Busch, 1975). This is especially true when short "grab" samples of ap proximately equal sampling duration are used to estimate worker exposures. Such is the case for the present study, since both impingers and membrane filter samples were of short duration and can be considered "grab" samples for statistical considerations. Arithmetic mean exposures and confidence interval estimates must, therefore, be ap proximated from the regression analysis estimates of parameters and variances. Two TAIILK III. Summary of Process and Control Variables in Exposure Zone Models Operation (zone) Control or process modifications Preparation (1) Carding (2) Ring spinning (?) Mule spinning (4) Foster winding (5) Twisting (6) Universal winding O) 1 leavy weaving (R) Light weaving (9) Mixing beds phased out Automatic blending lines added Carding ventilation completed Some wet spinning No major changes Ventilation of winders completed Twister ventilation completed No major changes Loom, cop winding and inspection ventilation added No major changes Date 1945 1964 1935 1966 1937 1938 1937 -- Exposures Among Chrysolite Workers 415 situations exist with regard to worker job categories. The first involves those workers who spend ail their time in one plant exposure zone, whereas the second concerns workers, although few in number, who may spend time in more than one zone. The former situation will be considered first. The regression models predict the mean values for the log of dust concentration as follows: Y = EfrZj (1) where Y = mean of log concentration, /3j = estimated regression coefficients, and Z, = independent variables for controls, UJC, time, etc. An estimate of the arithmetic mean dust concentration may be obtained from the following expression: X = exp (Y + 1/2 <j2) (2) where X = estimated arithmetic mean concentration, Y = mean of log concentration values, and a1 = pooled estimate of variance of Y from the regression model. For workers who may spend part of their working day in more than one exposure zone, their average exposures may be estimated as follows: X = WjXi where W; = fraction of working day in exposure i as determined from job descriptions plant interviews, and Xj = arithmetic average dust concentration for specified UJC in exposure zone i. Using the above derived expressions and estimated regression parameters, mean exposure values by calendar time and 95% confidence intervals were estimated for each job. An expression for estimating the variance of the means exposure values by use of a Taylor series approximation was developed and is described elsewhere [Dement, 1980]. RESULTS AND DISCUSSION Estimates or mean exposure levels along with 95% confidence intervals are given in Tables 1V-XII for the various textile production operations. For most operations, exposures were significantly reduced by about 1940 when most engineering controls were in place. Significant reductions were accomplished in most textile operations with "precontrol" exposures ranging from about 3 to 78 fibers/cc and were typically well above 10 fibers/cc. After textile operations were provided with controls, exposures ranged from approximately 3 to 17 fibers/cc and were usually in the range of 5 to 10 fibers/cc. , 416 Dement et al TABLE IV. Summary of Exposure Estimates for Fiber Preparation and Waste Recovery Operation Uniform job category Estimated mean fiber exposures (fibers > 5 /tm/cc) and 95Vo confidence intervals 1930-1944 1945-1964 1965-1975 A: General area personnel B: Machine operators Fly machine operator Crusher operator Waste machine operator C: Clean-up personnel D: Raw fiber handling 26.2 (12.2-40.2) 78.0 (38.1-117.8) 78.0(38.1-117.8) 45.9 (21.0-70.8) 54.4 (10.7-98.1) 35.0 (17.0-53.0) 8.1 (6.0-10.1) 23.9 (17.7-30.2) 23.9(17.7-30.2) 14.1 (9.5-18.7) 16.7 (5.4-28.0) 10.8 (7.3-14.2) 5.8 (4.4-7.2) 17.2(13.1-21.3) - 10.1 (7.1-13.2) 12.0 (4.4-19.6) 7.3 (5.7-9.0) TABLE V. Summary of Exposure Estimates for Carding Operation Uniform job category A: General area personnel B: Card operators C: Clean-up personnel D: Raw fiber handling Estimated mean fiber exposures (fibers > 5 /ori/cc) and 95V confidence intervals 1930-1935 1936-1945 1946-1965 1966-1975 10.8 (7.1-14.5) 13.3 (9.5-18.1) 18.1 (9.2-27.1) 22.8 (13.2-32.0) 5.3 (3.6-6.9) 6.5 (4.3-8.7) 8.8 (4.6-13.0) 11.0 (8.3-13.7) 2.4 (2.3-2.5) 29(2.6-3.3) 4.0 (2.6-5.4) 5.0 (3.8-6.1) 4.3 (4.0-4.6) 5.3 (4.8-5.8) 7.2 (4.7-9.7) 9.0(7.0-11.0) TABLE VI. Summary of Exposure Estimates for Ring Spinning Operation Estimated mean fiber exposures (fibers > 5 (tm/cc) and 95Vo confidence intervals________________ Uniform job category1930-19651966-19701971-1975 A: General area personnel B: Spinner operators C: Ctcan-up personnel D: Raw fiber handling 8.2 (7.6-8.8) 6.6 (5.6-7.6) NP* IMP 8.6 (7.7-9.5) 6.9 (6.0-7.9) 6.2 (5.3-7.1) 5.0 (4.5-5.4) *NP: No personnel within job category in this operation. TABLE VII. Summary of Exposure Estimates for Mule Spinning Operation Uni form job category A: General area personnel B: Spinner operators C: Clean-up personnel D: Raw fiber handling Estimated mean fiber exposures (fiber > 5 /im/cc) and 95V confidence intervals 1930-1975 6.3 (5.8-6.9) 4.8 (4.2-5.0) 6.7 (4.3-9.1) NP*- aNP: No personnel with job category in this operation. TABLE VIII. Summary of Exposure Estimates for Foster Winding Operation Uniform job category Estimated mean fiber exposures (fibers > 5 pm/cc) and 95V confidence intervals 1930-1937 1938-1975 A: General area personnel B: Winder operator C: Clean-up personnel D: Raw fiber handling 10.4 (2.1-19.3) 13.6 (2.4-24.8) 20.9(1.1-40.7) NP* 4.2 (3.5-5.1) 5.5 (2.8-8.1) 8.4 (3.4-12.5) aNP: No personnel with job category in this operation. Exposures Among Chrysotile Workers 417 TADLE IX. Summary of Exposure Estimates for Twisting Operation Uniform job category Estimated mean fiber exposures (fibers > 5 ^m/cc) and 95Vo confidence intervals 1930-1938 1939-1975 A: General area personnel B: Twister operators C: Clean-up personnel D: Raw fiber handling 24.6(14.0-35.2) 36.0(19.2-52.7) 31.9(13.1-50.7) 31.9(13.1-50.7) 5.4 (4.8-6.1) 7.9 (6.7-9.2) 7.0 (4.3-9.8) 7.0 (4.3-9.8) TABLE X. Summary of Exposure Estimates for Universal Winding Operation Uniform job category Estimated mean fiber exposures (fibers > 5 jtm/cc) and 95*7* confidence intervals 1930-1975 A: General area personnel B: Winder operators C: Clean-up personnel D: Raw fiber handling 4.1 (3:4-4.8) 4.1 (3.4-4.8) 8.4 (4.0-12.7) NP *NP: No personnel within job category in this operation. TABLE XI. Summary of Exposure Estimates for Heavy Weaving Operations Uniform job category General atea personnel Machine operators Bcaniers and creelers Winders Weavers inspectors Clean-up personnel Raw fiber handling Estimated mean fiber exposures (fibers > 5 /un/cc) and 95% confidence intervals _______ 1930-1936 1937-1975 9.2 (5.6-12.8) 2.6 (1.8-3.1) 14.3 (8.1-20.5) 9.7 (5.3-14.1) 14.3 (8.1-20.5) 5.3 (2.1-8.5) 30.6 (22.4-36.8) 30.6 (22.4-38.8) 3.8 (3.5-4.1) 2.6 (2.2-3.0) 3.8 (3.5-4.1) 1.4 (1.1-1.8) 8.2 (4.3-12.0) 8.2 (4.3-12.0) TABLE XII. Summary of Exposure Estimates for Light Weaving Operation Uniform job category A: General area personnel B: Weavers, winders and calendar operators C: Clean-up personnel D: Raw fiber handling Estimated mean fiber exposures (fibers > 5 /un/cc) and 95% confidence intervals 1930-1975 2.7 (2.1-3.2) 2.7 (3.1-3.2) 6.9 (4.2-9.5) NP *NP: No personnel within job category in this operation. Exposures in a large cross-section of the U.S. asbestos textile industry have been studied by the U.S. Public Health Service [Lynch and Ayer, 1966]. These studies were conducted using the membrane filter method. A comparison between the Lynch and Ayer exposure measurements and those estimated for the plant under study is given in Table XIII. In the Lynch and Ayer study, those plants with highest exposures likely ap proximate dust controls in the textile industry prior to implementation of dust controls. Table XIII shows these levels to be approximated by the precontrol exposure estimates 418 Dement et al TABLE XIII. Comparison of Exposure Estimates with Data for Other Textile Mills Operation Estimated mean exposure (fibers > 5 pm/cc) Lynch and Ayer (1966) Present study" Best plant Worst plant Precontrols With controls Preparation Carding Spinning1' Twisting Windingc Weaving 2.6 2.0 1.8 l.l 1.3 1.5 17.0 15.2 18.9 22.4 17.5 17.8 26.2-78.0 10.8-22.1 4.8- 8.2 24.6-36.0 4.1-20.9 5.3-30.6 5.8-17.2 4.3- 9.0 4.8- 6.7 5.4- 7.9 4.1- 8.4 1.4- 8.2 "For the present study, data shown are for the highest and lowest exposures by job within a particular operation. ^Includes ring and mule spinning as Lynch and Ayer did not present separate data. `Includes Foster winding and universal winding as Lynch and Ayer did not present separate data. studied here. Likewise, data from the best controlled plants studied by Lynch and Ayer are lower than postcontrol estimates in this plant. These comparisons suggest that the exposure estimates for the plant being studied are well within reasonable limits and consistent with findings of others relative to textile operations. The exposure estimates shown in Tables 1V-X11 were used in conjunction with mortality data to estimate expo sure-response relationships. It is realized that some exposure misclassification is unavoidable in any retrospec tive epidemiological study. However, the exposure models derived and used for this study appear to give reasonable exposure estimates without significant upward or downward bias. Since exposures for both diseased and nondiseased individuals are cal culated in the same manner by use of work histories, misclassification on exposure would be nondifferential in nature (ie, effecting both diseased and nondiseased indi viduals). Shy et al [1978] have shown that nondifferential misclassification on exposure biases toward the null effect. In the current study of various cumulative exposure levels, the effect would be to make strata more homogeneous with respect to exposure, thus reducing any exposure-response gradients. The linear model approach for exposure estimation as was used in this study has several advantages. First, and perhaps most important, the models allowed for simul taneously adjusting exposure estimates for changes in engineering controls, job cate gories, and process changes. Second, this approach serves to minimize variances in the exposure estimates while accounting for the log-normal distribution of environmental sample data. The linear model approach may be useful in other retrospective studies where historic exposure estimates are desired. REFERENCES Bresiin AJ,Ong L, Glauberman H, George AC, LeClare P (1967): The accuracy of dusl exposure estimates obtained from conventional air sampling. Am Ind Hyg Assoc .1 28:55-61. Brown RM (Nov. 1937): Report of dusl study in one department of the General Asbestos and Rubber Company plant at North Charleston, S.C. South Carolina State Board of Health. Clifton RA (1975): Asbestos. "Mineral Facts and Problems." Washington, DC: US Department of Interior. Coenen W (1966): The confidence limits for the mean values of dust concentration. Staub 26:36-45. Coencn W (1970): Measurement of the concentration of heallh>impairing especially silicogenic dusts at work places of surface industries. Staub 31:16-23. Exposures Among Clirysotile Workers 419 Corn M. Esmen NA (1979): Workplace exposure rones for classification of employee exposures to physical and chemical agents. Am Jd Hyg Assoc J 40:47-57. Dement JM (1980): "Estimation of Dose and Evaluation of Dose-Response in a Retrospective Cohort Mor tality Study of Clirysotile Asbestos Textile Workers." North Carolina: Doctoral Dissertation, Uni versity of North Carolina at Chapel Hill. Esmen NA (1979): Retrospective industrial hygiene surveys. Am Ind Hyg Assoc J 40:58-65. Gale MM (1965): The log-normal distribution and some examples of its application in the field of radiation protection. In "Atomic Energy Research Establishment Report AERE-R 4736." England: Her Majesty's Stationery Office. Hounarn RF (1971): An application of the log-normal distribution to some air sampling results and inter pretation of air sampling data. In "Atomic Energy Research, Establishment Report AERE-M 1469." England: Her Majesty's Stationery Office. IARC (International Agency for Research on Cancer) (1977): "IARC Monographs on the Evaluation of Carcinogenic Risk of Chemicals to Man: Asbestos." I von, France: WHO. Jones AR, Brief RS (1971): Evaluating benzene exposures. Am Ind Hyg Assoc J 32:610-613. Juda J, Budz.inski K (1967): Determining the tolerance range for the mean value of dust concentration. Staub 27:12-16. Kleinbatim DG, Kuppcr LL (1978): "Applied Regression Analysis and Other Multivariable Methods," 1st Ed. Belmont, California: Duxbury Press. Lcidef NA, Bayer SG, Zimrwaldc RD (1979): "USPHS/N/OSH Membrane Filter Method for Evaluating Airborne Asbestos Fibers." DHEW Publication No. (NIOSH) 79-127. Lcidel NA, Busch KA (1975): "Statistical Methods for the Determination of Noncompliance with Occupa tional Health Standards.*' DHEW Publication No. (NIOSH) 75-159. Lynch JR, Ayer HE (1966): Measurement of dust exposures in the asbestos textile industry. Am Ind Hyg Assoc J 27:431-437. McDonald JC, Liddell FDK, Gibbs GW, F.yssen GE, McDonald AD (1980): Dust exposure and mortality in clirysotile mining, 1910-75. Br J Ind Med 37:11-27. Nicholson WJ. Sclikoff IJ, Seklmau H, l.ilis R, Formby P (1979): Long-term mortality experience of clirysotile miners and millers in Theflord Mines, Quebec. Ann NY Acad Sci 330:11-31. NIOSH (National Institute for Occupational Safety and Health) (1980): "Workplace Exposure to Asbes tos: Review and Recommendations." DHEW Publication No. (NIOSH) 81-103. Page RT, Bloomfield .1.1 (1937): A study of dust control methods in an asbestos fabricating plant. Public Health Rep 52, No. 48:1-16. , Rubino OF, Piolatlo G, Newhousc ML, Scansctti G, Arcsin GA. Murray A (1979): Mortality of chrysolilc asbestos workers at the Balangeio Mine, Northern Italy. Br J Ind Med 36:187-194. Selikoff IJ, Lee DH (1978): "Asbestos and Disease." New York: Academic Press. Sherwood RJ (1966): On the interpretation of air sampling for radioactive particles. Am Ind Hyg Assoc J 27:98-109. Sherwood RJ (1971): Monitoring of benzene exposure by air sampling. Am Ind Hyg Assoc J 32:840-846. Shy CM, Klcinbaum DG, Morgenstern H (1978): The effects of niisclassification on exposure status in epi demiological studies of air pollution health effects. Bull NY Acad Med 54. No. 11:1155-1165. Weiss W (1970): Mortality of a cohort exposed to clirysotile asbestos. J Occup Med 19:737-740. Woilowilx HJ, Schake G, Woitowuz. R (1970): Ranking estimation of dust exposure and industrial-medi cal epidemiology. Slaub-Reinhalt Lufl 30:15-19.