INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Refractory ceramic fibers (RCF) are man-made vitreous fibers (MMVF) that are utilized for industrial high-temperature insulation applications such as kilns and furnaces. They are noncrystalline or amorphous and are manufactured from melted kaolin clay or a combination of Al₂O₃ and SiO₂ and fiberized by a wheel centrifuge or steam-jet fiberization process. The final products include bulk fiber, blankets, boards, paper, and textiles. Overall production represents 1 to 2% of total MMVF production within the United States (1). Approximately 32,000 workers are involved with the manufacture, secondary use, and end-use applications of RCF (2).

Animal inhalation studies of RCF in hamsters and rats at a maximum tolerated dose of 220 fibers/cc reported pulmonary and pleural fibrosis, mesothelioma, and lung cancer. Multidose inhalational studies at 36, 91, and 162 fibers/cc demonstrated cellular response at the lowest dose, one mesothelioma at 91 fibers/cc, and minimal fibrosis at 91 and 162 fibers/cc (3).

A cross-sectional study of workers employed with European manufacturing of RCF demonstrated an association between irritant symptoms and RCF exposure as well as decrements in FEV₁ related to cumulative exposure in current and ex-smokers (7). Two chest radiographic studies of workers involved with manufacturing RCF in five United States plants demonstrated an association between risk of pleural plaques and time since first RCF production job, years of RCF production employment, and cumulative fiber-mo/cc exposure. No increased prevalence of interstitial fibrosis was noted (8, 9). As part of the United States morbidity study of RCF manufacturing workers, spirometry testing was offered to each current worker at five manufacturing facilities on an approximately yearly basis. This report represents the longitudinal results of this testing across the five plant sites from 1987 through 1994 for those current male workers with five or more pulmonary function results.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The study was designed to evaluate any effects of RCF employment on longitudinal pulmonary function in workers involved with RCF manufacturing operations. Spirometry tests were offered at yearly intervals between June 1987 and June 1994 to workers at five RCF manufacturing locations. Subjects eligible for analysis were chosen on the basis of being employed at an RCF manufacturing location for at least 1 mo preceding pulmonary testing, and having participated in five to seven pulmonary function testing sessions by June 30, 1994. The provision of participating in at least five pulmonary function testing sessions required that RCF workers be hired on or before June 30, 1990 in order to be included in the analysis.

The initial sample consisted of 963 workers who were actively employed in RCF manufacturing operations for at least 1 mo preceding the first test. The 963 actively employed workers included 256, 153, 285, 93, and 176 individuals from Georgia, Indiana, New York, Oklahoma, and Tennessee, respectively. Numbers of men and women were 754 (78%) and 209 (22%) respectively. Of the 754 men, 569 (75%) were hired prior to June 30, 1990. Of the 569 men, 361 (63%) were included in the longitudinal analysis by participating in at least five pulmonary function testing sessions. The percentages of male participants from each site were 18, 21, 29, 10, and 22 from Georgia, Indiana, New York, Oklahoma, and Tennessee, respectively.

There were 85 women hired by June 30, 1990, who provided five to seven pulmonary tests. Lifetime cumulative RCF exposure calculated to each test was available for 43 women from two plant sites. These numbers were considered too small for the same statistical modeling approach as that applied to men due to small numbers in categories of analysis variables, which would likely result in unstable parameter estimates.

Spirometric testing for FVC and FEV₁ was done on a yearly basis for current workers using technicians from the University of Cincinnati with extensive spirometric testing experience and completion of a spirometry testing training course approved by the National Institute of Occupational Safety and Health (NIOSH). The testing was performed on Ohio-Med 822 dry rolling seal 8-L spirometers coupled to Spirotech 300 microprocessors (Graseby Andersen, Smyrna, GA). These instruments meet all American Thoracic Society (ATS) instrument specifications for spirometric testing. The same equipment was used throughout the testing period. Calibration testing on the spirometers and microprocessors was done before and after each half-day testing. The spirometers were leak-tested before and after each half-day and the tubing was separately leak-tested prior to each site visit. Spirometric testing met or exceeded ATS (1987) criteria with the maximum values for FVC and FEV₁ chosen from acceptable tracings (10). Testing continued until the highest and second highest FVC and FEV₁ values were within 5% of each other. If the highest valid FVC or FEV₁ occurred on the last trial, another trial was performed to ensure maximal effort. Spirometry testing was performed in the standing position unless a medical condition precluded this, with nose clips and loosening of tight-fitting clothing. Weight and height were measured in stocking feet to the nearest pound and quarter inch. Workers were asked to refrain from smoking and eating 1 h prior to testing. Results were rejected if there was technical difficulty unlikely to be related to any underlying lung condition. Tests were randomly reviewed by an investigator (R.T.M.) for quality assurance and were electronically entered into the database. Spirometric values were expressed as a percentage of a predicted value standardized for age, sex, race or ethnicity (Caucasian, Mexican-American, Native American), and height (11). The predicted values for African-Americans were obtained by reducing the Caucasian values by 15% (14). Each worker was provided with a copy of his or her spirometric results.

Occupational history interviews were administered at the time of each test session; these provided job titles and dates of all RCF jobs held. An independent review by two investigators (C.H.R. and J.E.L.) enabled each job title to be classified as "production" or "nonproduction" according to whether or not at least 10% of the job was performed in a production area. Duration of time spent in RCF production employment up to the date of each test session (RCF production years) was calculated. Measurements of RCF fiber levels (fibers/cc) were obtained from industrial hygiene sampling during the years 1987 through 1994. Details of the sampling protocol are reported elsewhere (15). Exposure measurements in each work location were averaged for specified time periods, and the mean value was used to estimate RCF exposure for a group of jobs considered to represent similar exposures. Exposure levels were assigned to each individual from these job and time-specific estimates according to the job history reported by the worker. Cumulative RCF exposure (fiber-mo/cc) from the first test session (approximately 1987) to each subsequent test session was calculated by multiplying job-specific exposure level by number of months on the job and totaling across all jobs held. In two plant locations, pre-1987 fiber levels were available from a previous industry-wide study (16, 17). For workers employed in these locations, pre-1987 exposure data were combined with current exposure data to calculate cumulative RCF exposure from first RCF job to the date of each pulmonary function test session. For the other three sites, information concerning pre-1987 job-specific exposure levels was not available to calculate working lifetime cumulative RCF exposure (fiber-mo/cc).

Statistical Analysis

Regression models. Cross-sectional and longitudinal analyses were used to examine the association of spirometry values with RCF exposure. First, a cross-sectional analysis was carried out using data from the initial pulmonary function test of men who participated in at least one pulmonary function test session and were hired by June 30, 1990. Longitudinal analyses were of two types, repeated measures and a two-stage random effects analysis (18, 19). The longitudinal study was designed to estimate the effect of continued exposure to RCF on lung function change during follow-up testing (1987 to 1994), adjusted for previous RCF exposure and covariates related to lung function outcome. In order to allow enough time to detect possible longitudinal changes in lung function related to RCF exposure, longitudinal results were based on data from men who provided five to seven measurements. The cross-sectional and longitudinal analyses focused on two outcome variables, FVC and FEV₁, which were height-adjusted by multiplying FVC and FEV₁ values by the square of the average height of all male subjects and then dividing by the square of each individual's average height measured over all tests. The units of pulmonary function outcomes in the cross-sectional and repeated measures analyses were milliliters (ml); units for the random effects analysis were ml/yr. Least squares regression provided results of the cross-sectional analysis. The method of generalized estimating equations (GEE) (20, 21) was used for the repeated measures analysis, using a model similar to that of the cross-sectional analysis. The GEE method was chosen because it is suitable for analyzing longitudinal data in which observations are missing or unequally spaced. The hypothesis that data were missing completely at random was investigated by qualitative examination of mean values of percent predicted FVC at initial test versus number of tests provided using data from men who provided one or more pulmonary function tests. Subjects were divided into groups according to the number of tests provided. The rank orders of mean percent predicted FVC values at initial test and number of tests provided were sufficiently different to conclude that the frequency of follow-up testing was not related to initial test value. Preliminary analyses showed the correlation between all pairs of observations on the same individual to be approximately equal; hence, GEE results were obtained assuming a uniform correlation structure. The original GEE macro written by M. R. Karim (Johns Hopkins University, Baltimore, MD) was used. The second type of longitudinal analysis consisted of two stages. First, individual slopes of height-adjusted FVC and FEV₁ measurements on age were calculated by simple linear regression of FVC and FEV₁ on age, recorded to the nearest tenth of a year. This yielded a linear age regression coefficient (slope) for each worker. Second, each individual's slope was regressed on covariates. Results were obtained using the SAS procedure PROC MIXED assuming a random effects model, in which maximum likelihood estimates of slope parameters were obtained allowing slope variances to differ as a result of differences in number and timing of test sessions among individuals. The two-stage random effects model fits a line to all the data regressed on age, and simultaneously fits a line to each individual's deviation from the population line (22). In the analyses hypothesis testing was performed at the p = 0.05 significance level, unless stated otherwise.

Exposure modeling. From the work history, each job was identified as production or nonproduction and the duration of each type of employment was calculated. Categories at initial test session were formed for the cross-sectional and longitudinal analyses. These were nonproduction employment, production employment of 7 yr or less, and production employment greater than 7 yr. Categories were determined by the mean number of years of RCF production employment at initial test. The repeated measures analysis was performed using one metric characterizing exposure up to initial test and a second metric to characterize exposure from initial to follow-up test. Three approaches were used, as follows: (1) categorical duration RCF production employment up to initial test, and continuously measured duration RCF production employment from initial to follow-up test (all plant sites); (2) categorical duration RCF production employment up to initial test, and calculated cumulative RCF exposure (fiber-mo/cc) from initial to follow-up test (all plant sites); (3) categorical cumulative RCF exposure up to initial test, and calculated cumulative RCF exposure from initial to follow-up test (two plant sites).

Each of the above analyses provided separate estimates of any exposure effects on lung function to the initial test and those between initial and follow-up test. Categories of cumulative RCF exposure at initial test were (1) up to 15 fiber-mo/cc, (2) greater than 15 fiber-mo/ cc but no more than 60 fiber-mo/cc, and (3) greater than 60 fiber-mo/ cc. These fiber ranges yielded significant differences between each pair of mean cumulative exposure values and at the same time provided sufficient numbers of subjects in each category. The two-stage random effects analysis used categorical duration RCF production employment to initial test and calculated cumulative RCF exposure from first to last test. Separate estimates of the effect of RCF exposure to initial test and RCF exposure from initial to last test were provided.

Race adjustment. It has been shown that average lung function values of African-Americans are lower than those of Caucasians of the same age, sex, and height (26). Several strategies were tried to adjust for the effect of race in the regression analyses, resulting in height-adjusted values of African-Americans being increased by 15%. The criterion for choice was based on accurately predicting observed values of African-Americans in a repeated measures analysis of height-adjusted FVC values which included all male RCF workers. Mean predicted values for each of several categories of production employment and age were obtained for African-Americans. FVC means of African-Americans, after individual values were increased by 15%, were least discrepant by this strategy.

Age modeling. Cross-sectional (between-individual) and longitudinal (within-individual) effects of age were modeled simultaneously in the repeated measures analysis. Age was modeled as a restricted cubic spline function with deflection points at ages 25, 35, and 55 (27). Plots of FVC versus age for nonproduction nonsmokers were evaluated visually and shown to approximate a spline function defined by these intervals. Slope of the tangent to the spline curve at age 40 was obtained to estimate the average effect of age on pulmonary function at approximately the mean age of the study population (37.3 yr). A similar spline function was used to model age cross-sectionally in the analysis of initial test data.

Weight and pack-years modeling. Weight and pack-years were assumed to be linearly related to FVC and FEV₁. This assumption was checked by obtaining scatter plots of each outcome versus each variable, using data from the initial test session, then fitting a smoothing function to the plotted values (28). Globally, fitted functions were approximately linear.

In the longitudinal analyses we modeled the contributions of cross-sectional and longitudinal effects of weight separately. That is, we tried to separate the effect of weight measured by differences among subgroups of male RCF workers at a fixed point in time (cross-sectional effect) from the effect of within-individual weight change (longitudinal effect). In the repeated measures analysis, for example, the value of weight at each test was decomposed into the sum of weight at initial test and weight change from initial to subsequent test, allowing for separate regression coefficients.

Similarly, the contributions of cross-sectional and longitudinal effects of smoking were separated in the longitudinal analyses. In the repeated measures analysis, for example, cumulative pack-years at initial test was modeled in order to estimate the effect of smoking across individuals, and a dichotomous variable (smokingyes,no) was used to describe the effect of individual smoking patterns on lung function outcome.

Analysis of pulmonary function. The variables included in the cross-sectional analysis of initial pulmonary function test results were age (years), RCF production employment (categorically defined), pack-years, current smoking compared with never smoking, previous smoking compared with never smoking, weight, and plant (manufacturing site). A categorical plant variable was included in the model to adjust for possible geographic, socioeconomic, and demographic differences as well as exposure or process differences that might alter pulmonary function.

Longitudinal repeated measures variables included age, initial duration RCF production categories (none, 7 yr or less, greater than 7 yr), follow-up years in RCF production measured from initial test to the date of each subsequent test (five or more), pack-years to initial pulmonary function test, current smoking compared with never smoking at the time of each test, weight at initial pulmonary function test, weight gain or loss from initial test to time of each subsequent test, and categorical plant site. Additional repeated measures analyses were carried out in which exposure metrics changed and other variables remained the same. The methods followed those of Ware (29).

Variables included in the two-stage random effects analysis were mean age from first to last test, initial duration RCF production categories (none, 7 yr or less, greater than 7 yr), accumulated RCF exposure from first to last test, continuous smoking over study period compared with never smoking, previous smoking before initial test compared with never smoking, starter or quitter during study period compared with never smoking, pack-years during study period, weight at initial test, and slope of weight change during study period obtained by a least squares regression of weight on time. All were multiplied by the difference of age at each test minus age at first test to obtain slope estimates.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Table 1 shows descriptive statistics for the 361 of 569 male workers who provided five to seven tests (analysis group), and 191 of 208 men employed in RCF during the testing period who provided one to four measurements (nonanalysis group). Fifteen workers who provided one pulmonary function test and worked less than 1 yr are not included in Table 1. Of the 193 workers in the nonanalysis group, 107 left employment because of retirement, layoff, another job, disability or death, and 86 did not participate because of medical exclusion criteria (elevated blood pressure, n = 20, acute illness, n = 5), vacation, or refusal. Two men provided no measurements. On average, the nonanalysis group was older, smoked more, weighed more, and had lower height-adjusted and percent predicted lung function values at initial test. Mean numbers of years of RCF production employment were 7.0 and 7.3 in the analysis and nonanalysis groups, respectively.

Table 2 shows descriptive statistics of the analysis group by categories of duration RCF production employment at initial test. There were 52 men who never held a production job, 124 men who were employed in RCF production 7 yr or less at the time of initial test, and 185 men who were employed in RCF production longer than 7 yr at the time of initial test. Nonproduction workers had a mean age equal to 37.7 yr, were primarily Caucasian (98.1%), and about half had never smoked (48.1%). The group with 7 yr or less RCF production employment at initial test had an average age of 33.3 yr, were primarily Caucasian (77.4%), and 42.7% were current cigarette smokers. The group with more than 7 yr RCF production employment at initial test were slightly older and had a mean age of 39.8 yr, were primarily Caucasian (83.2%), and 35.1% were current cigarette smokers. Mean pack-years of cigarette smoking of the three groups were 11.8, 9.1, and 14.0 in the same order. Mean weight of all groups was about the same. The distribution of categories of RCF production employment was similar across plants except for Plant 2, which had a larger percentage of nonproduction employees (35.6%) included in the analysis.

Table 2 also shows average yearly changes of height-adjusted FVC and FEV₁ and average cumulative RCF exposure from initial to last test. Average changes in FVC and FEV₁ per year were approximately equal across employment categories. Mean cumulative RCF exposure values increased with longer duration of RCF production employment at initial test.

Table 3 shows results of cross-sectional analyses relating height-adjusted FVC and FEV₁ lung function outcomes at initial test to categorical duration of RCF production employment. Results are based on data from 552 men tested one to seven times, including 361 in the analysis group and 191 in the nonanalysis group (Table 1). Initial test FVC and FEV₁ values were 219.4 ml (p < 0.01) and 205.2 ml (p < 0.01) lower among workers with more than 7 yr RCF production employment compared with nonproduction workers, and 65.6 ml and 80.6 ml lower among workers with 7 yr or less RCF production employment compared with the same group. Cigarette smoking, measured by pack-years, was significantly associated (p = 0.02) with cross-sectional FEV₁ results at initial test. After adjusting for the decrement in lung function due to pack-years of smoking, current smokers had an additional decrement in FVC of 97.4 ml, and an additional decrement in FEV₁ of 146.2 ml compared with never-smokers. Previous smokers recovered 29.0 ml FVC and 70.8 ml FEV₁ after pack-years adjustment.

Tables 4 and 5 show GEE results which describe the effect of initial categories of RCF exposure, primarily a cross-sectional (between-individual) effect, and the effect of continuing exposure to RCF, primarily a longitudinal (within-individual) effect. Table 4 is based on 2,211 observations from 361 male workers. Table 4a shows that initial RCF production employment of 7 yr or less was related to a FVC value 66.3 ml lower compared with nonproduction employment. A statistically significant decrease in FVC was related to initial RCF production employment longer than 7 yr (regression coefficient = 171.0, p = 0.04). Follow-up RCF production years was associated with an increase in FVC equal to 5.3 ml per production year. Results of Table 4b are similar to those of Table 4a. Initial RCF production employment of 7 yr or less was related to an FVC value 55.2 ml lower compared with the nonproduction employment group. A statistically significant decrease in FVC was related to initial RCF production employment longer than 7 yr (regression coefficient = 168.3, p = 0.04). Follow-up cumulative RCF exposure was associated with an increase in FVC of 0.7 ml per fiber-mo/cc. Corresponding results for FEV₁ in Table 4, parts a and b were similar but smaller in magnitude than FVC results. No FEV₁ results for RCF exposure were significant at a 5% level or less. Similar results were obtained when the analyses presented in Table 4 were repeated, excluding the plant contributing most of the differences among plant locations.

Table 5 is based on 1,116 observations of 181 male workers from two plants where historical exposure measurements were available. The category of cumulative RCF exposure measured at initial test, which includes values greater than 15 to 60 fiber-mo/cc, was related to a decrease in FVC of 36.2 ml compared with the baseline category which includes cumulative RCF exposures between zero and 15 fiber-mo/cc. A decrease in FVC equal to 156.0 ml was related to initial cumulative RCF exposure values greater than 60 fiber-mo /cc compared with the same baseline category. Follow-up cumulative RCF exposure (fiber-mo/cc) was associated with an increase in FVC of 0.8 ml per fiber-mo/cc. FEV₁ regression coefficients, measuring the effect of the same cumulative exposure categories compared with baseline, were 100.2 and 104.7 for 15 to 60 fiber-mo/cc and greater than 60 fiber-mo/cc, respectively. Follow-up cumulative RCF exposure was associated with an increase in FEV₁ of 0.2 ml per fiber-mo/cc. Results of the two-stage random effects analysis were consistent with the repeated measures analysis with respect to a nonsignificant association between FVC and FEV₁ change with additional cumulative RCF exposure during the study period.

Differences in the pack-years regression coefficients between the repeated measures analyses (Tables 4 and 5) and cross-sectional analysis (Table 3) were investigated by excluding categorical smoking from the cross-sectional analysis. Regression coefficients and p values for pack-years were 2.9 (p = 0.09) for FVC and 6.2 (p < 0.001) for FEV₁. Results were closer in magnitude to regression coefficients of pack-years at initial test in the repeated measures analyses.

Differences between the slopes of tangents to spline age curves in the cross-sectional (Table 3) and longitudinal analyses (Tables 4 and 5) may be partially explained by the fact that the longitudinal analysis models the cross-sectional (at initial test) and longitudinal age effects simultaneously. Previously, differences between cross-sectional and longitudinal age effects have been documented (30). When the cross-sectional analysis was repeated using the analysis cohort of 361 men, differences between cross-sectional and longitudinal age effects persisted.

An interesting observation was the 13.2 to 17.5 ml decrease (p < 0.001) in both FVC and FEV₁ for each kilogram in weight gain in the repeated measures analysis for all three exposure classifications (Tables 4 and 5).

Within the nonanalysis group, the possibility of a low pulmonary function value at initial test being related to worker availability for future testing was investigated (Table 6). The percents of individuals having percent predicted FVC and/or FEV₁ values below 80% at the initial test were obtained for the 107 workers who left RCF employment, and the 84 of 86 workers who were unable to participate. Thirteen of 83 (16%) and 27 of 83 (33%) production workers who left RCF employment had a percent predicted value lower than 80% for FVC and/or FEV₁, respectively. Numbers of workers unable to participate who had a percent predicted value lower than 80% for FVC and/or FEV₁ were 8 of 76 (11%) and 12 of 76 (16%), respectively. For the 361 men with at least five tests, the percent of production workers having percent predicted FVC and/or FEV₁ values below 80% at the initial test was lower at 19 of 309 (6%) and 34 of 309 (11%), respectively. The percent of workers in production with initial percent predicted FVC and/ or FEV₁ values below 80% was significantly greater in workers who left RCF employment (p < 0.01) compared with the analysis group, based on a chi-square statistic. Among current and previous smokers, the same group also had a significantly greater proportion of workers with percent predicted FVC and/or FEV₁ values below 80% (p < 0.01) in comparison to the analysis group. Overall, 22 of 23 workers (95%) with FVC values below 80% of predicted and 38 of 42 (90%) workers with FEV₁ values below 80% of predicted were current or previous smokers in the nonanalysis group.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

MMVF consist of glass fiber (glasswool, special-purpose glass fiber, and continuous glass filament), mineral wool (rock and slag wool), and refractory ceramic fibers. MMVF are noncrystalline or vitreous in comparison with asbestos, which has a crystalline structure. The fiber diameters of MMVF are dependent on the production process and their end use application and can fall well within the respirable range. The chemical composition of fibers differs according to the raw material used in the manufacturing process and the percent of stabilizers such as Al₂O₃, TiO₂, and ZnO and mobilizers such as MgO, CaO, Na₂O, and K₂O. Increased percentage of stabilizers such as Al₂O₃ gives MMVF increased durability and heat resistance properties. RCF are noncrystalline durable fibers that fall within the respirable range of less than 3.5 µm diameter and 200 µm in length (1, 31).

A study of 70 workers with a mean of 20 yr working in the fiberglass manufacturing industry demonstrated no change in FVC or FEV₁ in comparison to a matched control group. The average number of respirable fibers was 0.9 fibers/cm³ (32). A more recent cross-sectional morbidity study of seven glass fiber and mineral wool plants did not demonstrate any abnormalities related to pulmonary symptoms or spirometry results and MMVF manufacturing. The limited changes on chest radiographs at profusion level 1/0 or 1/1, found in 23 of 1,435 workers, were not statistically different from a non-MMVF comparison group and were not associated with spirometric changes (33, 34). A European cross-sectional study of 628 current workers at seven RCF manufacturing plants demonstrated statistically significant relationships between decreases in FEV₁, FEF_25-75%, and cumulative exposure (f-ml/yr) in current smokers, and FEV₁ and cumulative exposure in ex-smokers. The mean (range) cumulative respirable fibers exposure was 3.84 (0-22.94) f-ml/yr based on current mean exposures within the plants which varied from 0.2 to 0.88 f/ml for primary production workers to 0.49 to 1.36 f/ml for secondary production workers (7).

Within this study the results at the initial pulmonary function test session, offered in 1987, reflect the effects of previous exposures, including exposure within the RCF manufacturing facility. Results of later test sessions reflect residual effects from exposure prior to 1987 and subsequent ongoing exposures. All plants were in production prior to 1980 and as early as 1953. The cross-sectional analysis of results from initial pulmonary function tests indicated a greater decrease in FVC and FEV₁ for production workers employed more than 7 yr than production workers with up to 7 yr employment in comparison with the baseline group of nonproduction workers. Similar results were seen in the longitudinal repeated measures analysis. A greater effect was seen in workers with longer than 7 yr in production and with the highest categorical cumulative RCF exposure prior to initial test.

These findings, however, did not persist with the longitudinal analyses of follow-up production years and follow-up cumulative RCF exposure (fiber-mo/cc) from initial pulmonary function test. Both follow-up production years and follow-up cumulative exposure were associated with essentially no change in FVC and FEV₁ (Tables 4 and 5). These data indicated that more recent exposures from the late 1980s until 1994 had no deleterious impact on the longitudinal trend of FVC and FEV₁. The results corresponded to the historical higher exposure levels in the 1950s (estimated maximum 10 fibers/cc in carding) in comparison to more recent exposures that ranged from approximately 1 fiber/cc to below the limit of detection (16).

Selection bias can occur in longitudinal studies with repeated measures since loss to follow-up or nonparticipation can be associated with decreased pulmonary function values (35, 36). In the current study there was a significant difference with respect to percent of FVC and FEV₁ values below 80% of predicted values between production workers who left RCF employment and the production participants. The selection and comparison of cross-sectional and longitudinal data from those individuals who have completed longitudinal data may help to control for discrepancies between cross-sectional and longitudinal estimates of decline in pulmonary function (35, 36). This method to control for selection bias may be appropriate for nonoccupational population studies but needs further evaluation in occupational based populations in view of the "healthy worker effect" (37). In this population the workers who left RCF employment or who were unable to participate were more likely to be "less healthy," as reflected by the increased number with percent predicted FVC and/or FEV₁ below 80%. Cigarette smoking was a likely contributing factor, as 92% of these workers were previous or current smokers with a 28.2 mean pack-year smoking history. There also could be an interactive effect between smoking and RCF exposure. These potential biases will be further evaluated with completion of five pulmonary function test sessions on workers who left RCF employment as part of this ongoing pulmonary morbidity study of RCF manufacturing workers.

This study demonstrated that employment in an RCF production job and accumulated RCF exposure were not associated with a decrement in longitudinal FVC or FEV₁ values from 1987 through 1994 in a current RCF manufacturing workforce. The effects on FVC and, to a lesser extent, FEV₁ that were demonstrated in the cross-sectional analysis of initial pulmonary function test results in 1987, may have resulted from earlier exposure levels to RCF which were higher than current values. The decrease in RCF exposure levels over the last 10 yr through engineering and work practice changes has reduced any detectable continued effect of RCF exposure on FVC and FEV₁.