INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Insulation wools constitute the major group of man-made vitreous fibers (MMVF). A large proportion of the population is either regularly or occasionally in contact with these fibers. Especially exposed are workers in the MMVF production industry and insulation workers, owing to longer exposure periods and hence higher accumulated fiber exposures. Because of some similarities between MMVF and asbestos fibers, there has been concern about the potential risks that may be associated with MMVF exposure. The potential for causing lung cancer/mesothelioma has been studied in experimental models and in large international mortality surveys and was not dealt with in our study. The risk of nonmalignant respiratory disease associated with MMVF exposure has been evaluated in a series of epidemiological surveys, primarily with MMVF production workers as exposed group. Most studies were of cross-sectional nature and the methodological problems have been considerable: choice of reference population; confounders, especially smoking and former asbestos exposure; healthy worker effect; etc. All contribute to the difficulties in establishing an association or lack of association between respiratory disease and MMVF exposure.

In 1988 a review by the International Agency for Research on Cancer (IARC) concluded that there is no proof of an association between occupational exposure to MMVF and nonmalignant respiratory disease (1). A review by De Vuyst and colleagues from 1995 also concluded that there is no firm evidence that exposure to glass-, rock-, and slag wool is associated with lung fibrosis, pleural disease, or nonspecific respiratory diseases in humans (2).

The above conclusions are based on a number of epidemiological studies with large groups of MMVF production workers. Weill and associates studied 1,028 male workers from seven plants, with median employment length of 18 yr and found no respiratory symptoms or adverse lung function related to exposure (3). In a follow-up, Hughes and coworkers found prevalences of respiratory symptoms and functional values consistent with a healthy population (4). In an Australian study of 671 rockwool and glasswool factory workers from eight plants, Woolcock and Mellis found no evidence of occupational asthma, pulmonary fibrosis, lung cancer, or occupational pleural disease (5). The lung function was assessed as being normal in the study population. In contrast, Clausen and coworkers found significantly lower values of FEV₁ in 340 insulation workers compared with 166 bus drivers (6). The observed difference was independent of smoking habits and self-assessed former asbestos exposure. Kilburn and coworkers found reduced expiratory flows in 284 appliance manufacturing workers using fiberglass for refrigerator insulation, compared with predicted regional values (7).

The present study was undertaken to examine the respiratory health of the Danish stonewool (rockwool) factory workers, which has not previously been studied in a controlled design. The aim was to evaluate self-assessed respiratory health and objective lung function parameters in this workforce, compared with similar data on a reference group of blue collar workers with no exposure to MMVF.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

The workforce of a Danish stonewool plant in Hedehusene, in production since 1937, was examined to find male workers with five or more years of exposure in the period 1955 to 1987, using the factory archives and the files of the Danish Cancer Register. Workers older than 70 yr of age or who had moved too far away for convenient examination at the Factory Health Center were excluded. Furthermore, workers had to be mainly employed in factory production. Present as well as former employees were examined to reduce potential selection bias. The basis for our comparison group was a random sample from the general population which in 1991 was offered a general health examination at the Center of Preventive Medicine, Copenhagen University Hospital. From this sample consisting of 2,163 men and women we picked a subsample of male subjects who matched the exposed group on age and social class (mainly skilled and unskilled workers). However, the control subjects were not individually matched to the cases. The control subjects, as well as the exposed subjects, lived in the suburbs west of Copenhagen. Subjects with occupational exposure to MMVF were excluded from the comparison group.

Questionnaire

All in the exposed group and subsequently in the comparison group were mailed an introductory letter and a self-administration questionnaire, which they were asked to complete and return. The questionnaire focused on detailed smoking history and a health history with emphasis on respiratory problems. The members of the comparison group were asked for a brief occupational history whereas the exposed workers were asked for a similar occupational history and additional information regarding their job history within the stonewool industry. After returning the questionnaire the responders were given an appointment for performing the lung function tests. The nonresponders were mailed a new letter and questionnaire andafter a while if still not respondinga third and final letter. All letters stressed that participation was voluntary. On entry the participants signed an informed consent. The study was approved by the local ethical committee.

Lung Function Tests

Lung function tests were performed by one trained technician, using a MedGraphics system PF/Dx (Medical Graphics, St. Paul/Minneapolis, MN). The device was calibrated daily and all values were recorded at BTPS. Each subject completed a dynamic spirometry with at least 3 acceptable and 2 reproducible maneuvers in accordance with American Thoracic Society (ATS) recommendations (8). The highest FVC and FEV₁ were recorded. After dynamic spirometry the transfer factor (TL_CO) and the transfer coefficient (diffusion per unit of alveolar volume [TL/VA]) were measured, using the single-breath carbon monoxide (CO) technique in accordance with ATS recommendations (9). The average value from at least 2 acceptable maneuvers was recorded. TL_CO was adjusted for level of hemoglobin, measured with a Reflotron (Boehringer Mannheim, Mannheim, Germany) at the same time as the other tests were performed. A few refused to deliver a blood sample, and were then included with unadjusted TL_CO. Lung function testing was performed with the same device and by the same technician for the two groups. After completing the examination of the exposed group at the Factory Health Center, all equipment was transferred to the Center of Preventive Medicine, where the examination of the comparison group took place. This sequence did not allow for blinding of the technician to the occupational status of the subjects.

Sample of Nonparticipants

From the 130 workers of the exposed group who had refused to participate in the study or who had not responded on three letters, a random sample of 32 was picked. They were mailed another letter in which we offered to visit them in their own home at a suitable time. Afterwards they were contacted by phone and asked for their consent. Workers who by this contact refused participation were not contacted further. The others were visited by a physician (E.F.H.) and a lung function test was performed on a portable Micro Spirometer (Micro Medical, Rochester, UK) measuring FEV₁ and FVC three times, after which the effort with the highest FEV₁ was chosen. They furthermore completed an interviewer-administered questionnaire regarding smoking history and a brief respiratory health history.

Calculations and Statistical Analysis

Data were coded and entered in a Microsoft Excel worksheet, version 5.0. Using published standard reference equations for ventilatory flows (10) and TL_CO (11), lung function in percentage of predicted was calculated for each individual. Data were subsequently analyzed using the SPSS statistical program for Windows, version 6.1.

Demographic distributions were described with mean, median, and range. Continuous dependent variables were described with mean and 95% confidence interval (CI) and tested for significance with a two-sided t test. Dichotomous dependent variables were described with their frequency and tested for significant differences between the two groups with two-sided ² test. Yates' correction for continuity was applied for all 2 × 2 tables analyzed.

The proportion of subjects with airflow obstruction was compared between the two groups. Obstruction was defined by a reduced FEV₁/ FVC ratio (below 88% of predicted), according to the European Respiratory Society (ERS) consensus statement on chronic obstructive pulmonary disease (12). The obstruction was graded as mild (FEV₁  70% of predicted), moderate (FEV₁ between 50 and 69% of predicted) or severe (FEV₁ at or below 50% of predicted) also according to ERS criteria (12).

Regression analyses were used to test for relationship between response variables and possible explanatory variables. Age, height, pack-years, and exposure duration were entered as continuous variables in the regression analyses, whereas current smoking, exposure group, and self-assessed asbestos exposure were entered as dummy variables. Multiple stepwise linear regression was used for the continuous response variables (FEV₁, FVC, FEV₁/FVC, TL_CO, and TL/VA) whereas multiple logistic regression was used for the dichotomous response variables (cough, phlegm, and dyspnea).

All tests of significance were two-sided and p values less than 0.05 were considered significant. At this significance level we estimated a power of 95% to detect a difference of 5% between the two groups in the age- and height-adjusted lung function values, provided a standard deviation of approximately 15%.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Participation and Population Description

The characteristics of the exposed and the comparison group are shown in Table 1. A total of 377 male production workers exposed to MMVF matched the inclusion criteria and were mailed an introductory letter and a questionnaire. A total of 247 (66%) returned the questionnaire after up to two follow-up letters; 242 (64%) showed up for examination; 235 (97%) of the examined subjects performed acceptable spirometry, whereas 221 (91%) had acceptable measurement of TL_CO. The comparison group consisted of 381 male workers who received an introductory letter and a questionnaire; 246 (65%) completed the questionnaire and were examined; 243 of the examined (99%) performed acceptable spirometry and 213 (88%) had acceptable measurement of TL_CO. The study population was the 235 exposed subjects and 243 nonexposed subjects who completed the questionnaire and who performed acceptable spirometry. Subsequent analyses were done on this population. The mean employment length in the exposed group was 16.6 yr (range, 5 to 46 yr). In this calculation, all employment was included, even if outside the period 1955 through 1987 which defined the selection for study.

Among the nonresponders a random sample of 32 exposed workers were contacted. Two (6%) were dead in the meantime, 4 (13%) were alive but lost to follow-up, 8 (25%) refused participation, and 18 (56%) accepted to participate. The eight who continually refused participation gave reasons like "lack of interest" and "distrust of this kind of investigation." None argued illness as a reason. Of 18 accepting participation, 17 performed an acceptable lung function test. The characteristics of these nonresponders are shown in Table 1.

Asbestos Exposure

In the exposed group 62 subjects reported former occupational asbestos exposure, whereas 96 reported no asbestos exposure and 77 did not know or provided no information regarding this exposure. In the comparison group there were 51 subjects with self-assessed asbestos exposure, whereas 137 had no such exposure and 55 did not know or provided no information. The number of asbestos-exposed subjects was not significantly different between the two groups. The 113 subjects with self-assessed asbestos exposure were not different from the other subjects on any of the demographic variables (age, height, smoking) or on any of the lung function parameters and they were included in subsequent analyses.

Lung Function

The means and 95% confidence intervals of the means of FEV₁, FVC, FEV₁/FVC, TL_CO, and TL/VA in percentage of predicted are shown in Figure 1. There were no significant differences between the exposed group and the comparison group except for FEV₁/FVC ratio which was significantly lower in the exposed group (98.6%) than in the comparison group (101.4%) (p < 0.005).

View larger version (47K): [in this window] [in a new window]   Figure 1.   Mean lung function parameters in percentage of predicted values in the exposed group (n = 235) and the comparison group (n = 243). TLCO and TL/VA were measured in 221 exposed subjects and 213 comparison subjects. FEV1/FVC was significantly lower in the exposed subjects than in the comparison subjects (p < 0.005).

The number of subjects with a significantly reduced FEV₁/ FVC ratioand thus with probable obstructive lung disease is shown in Table 2. A total of 47 subjects had a FEV₁/FVC ratio below 88% of predicted34 (14.5%) in the exposed group and 13 (5.3%) in the comparison group. Subgroup analyses in respect to the severity of airflow limitation showed that in the exposed group 17 had mild obstruction, 13 had moderate obstruction, and four had severe obstruction according to ERS criteria. In the comparison group the numbers were 8, 2, and 3. Testing the differences with a two-sided ²-test showed that the number of obstructive patients was significantly higher in the exposed group (p = 0.001). The relative risk (RR) ratio of airflow obstruction in the exposed group was 2.7 with a 95% CI of 1.5 to 5.0. Omitting subjects with self-assessed asbestos exposure from the analysis resulted in 26 (15.0%) with airflow limitation in the exposed group versus 12 (6.3%) in the comparison group, which was still statistically significant (p = 0.01). If subjects with self-reported asthma furthermore were excluded, the number with airflow limitation was 22 (13.5%) in the exposed group versus 11 (6.4%) in the comparison group (p = 0.04).

In order to examine the relation between MMVF exposure, smoking, and airflow obstruction, the exposed and comparison subjects were stratified as never-smokers (n = 109), smokers with a history of 20 or less pack-years (n = 147), smokers with more than 20 but not more than 40 pack-years (n = 129), or smokers with a history of more than 40 pack-years (n = 80). Thirteen subjects could not be classified according to the number of pack-years. The risk of airflow obstruction in relation to MMVF exposure was only significantly increased in the group with a smoking history of more than 40 pack-years where 16 of 45 (35.6%) in the exposed group had airflow obstruction versus 3 of 35 (8.6%) in the nonexposed group (p = 0.01). The smoking history in those two subgroups was similar with a mean of 54.1 pack-years in the exposed group versus 55.9 pack-years in the comparison group. The association between smoking history, group category, and degree of airflow obstruction is shown in Figure 2.

View larger version (29K): [in this window] [in a new window]   Figure 2.   Proportion with airflow obstruction in different smoking strata in the exposed group (n = 232) and the comparison group (n = 233). In the stratum with a smoking history of more than 40 pack-years, a significantly higher proportion of the exposed subjects had airflow obstruction, relative to the proportion with airflow obstruction among the comparison subjects (p = 0.01).

The number of subjects with a substantially reduced TL_CO or TL/VA was not significantly different in the two groups. In the exposed group 22 of 221 (10.0%) had a TL_CO below 80% of predicted versus 27 of 213 (12.7%) in the comparison group. TL/VA was below 80% of predicted in 33 (14.9%) from the exposed group versus 24 (11.3%) from the comparison group. Of those subjects with acceptable spirometry, 14 exposed (6%) and 30 nonexposed (12%) had no valid measurement of TL_CO. The different rate of performing this test was due to technical difficulties with the gas analyzer on two occasions when subjects from the comparison group were examined.

Multiple Regression Analysis

The associations between lung function parameters and continuous independent variables (age, height, pack-years and exposure duration) are shown in Table 3. FEV₁, FVC, FEV₁/ FVC, TL_CO, and TL/VA were all associated with age and height, with most values being comparable to the applied reference equations (10, 11). FEV₁, FEV₁/FVC, TL_CO, and TL/VA were all negatively associated with smoking history in terms of pack-years. TL_CO and TL/VA were furthermore associated with smoking status, as current smokers had lower values than never- and ex-smokers after controlling for number of pack-years. None of the lung function parameters were negatively associated with MMVF exposure except for FEV₁/FVC which was 3.6% lower in the exposed group after controlling for age, height, and smoking. However, FEV₁/FVC was not negatively associated with exposure duration in terms of employment length. An interaction term between smoking and exposure duration was tried in all models. The interaction term was defined as the product of a dichotomous smoking variable (more or less than 20 pack-years) and a dichotomous exposure variable (exposed or nonexposed). After controlling for age, height, pack-years, current smoking, exposure duration, and exposure group, this interaction term was negatively associated with FEV₁ (p = 0.004), FEV₁/FVC (p = 0.002), and TL_CO (p = 0.02).

The regression models were tried separately on the comparison group and the exposed group, to examine for differences in regression equations, but this had only very minor influence on the found regression coefficients. The self-assessed asbestos exposure, entered as a dichotomous variable, had no significant association with any of the lung function parameters.

The correlation between the explanatory variables age, employment length, and number of pack-years was examined by visual inspection of scatterplots to exclude nonlinear correlations and by calculation of Pearson correlation coefficient (r). Age was positively correlated to pack-years (r = 0.36, p < 0.01) and to employment length (r = 0.23, p < 0.01). Employment length was not correlated to number of pack-years (r = 0.07, p = 0.27).

Self-reported Symptoms and Disease

The prevalences of self-reported symptoms and disease are shown in Figure 3. There was no significant difference between the exposed and the comparison group in the frequency of reported asthma and chronic bronchitis nor in the reported symptoms (cough, phlegm, and dyspnea). The prevalence of reported emphysema was 3.8% in the exposed group versus 0.9% in the comparison group, resulting in a RR of 4.5. However, the number of diseased patients was very small (9 versus 2) and the 95% confidence interval of the RR was 1.0 to 20.6. 

View larger version (44K): [in this window] [in a new window]   Figure 3.   Prevalences of respiratory symptoms and self-reported respiratory disease among the exposed subjects and the comparison subjects. None of the RR ratios was significantly elevated for the exposed subjects.

In the logistic regression analyses, cough was found to be significantly associated with the number of pack-years and was most prevalent in current smokers, whereas there was no significant difference between ex- and never-smokers. Phlegm was significantly associated with age and current smoking again there was no difference between ex- and never-smokers. Dyspnea increased with age and number of pack-years. There was no significant association between respiratory symptoms and MMVF exposure.

Sample of Nonresponders

The nonresponder sample performed spirometry on a portable device which could not be directly calibrated, which makes exact comparisons of lung function between this sample and the two major groups unjustified. The sample mean FEV₁ in percentage of predicted was 89.3 with a 95% CI of 82.1 to 96.6. FVC in percentage of predicted was 92.4 with 95% confidence limits of 85.8 to 98.9. Three subjects (17.6%) had an obstructive flow pattern with a FEV₁/FVC ratio below 88% of predictedone having mild obstruction and two having moderate obstruction according to ERS criteria (12). The smoking history of these three subjects was 56, 58, and 35 pack-years. One person reported a diagnosis of asthma, whereas none reported chronic bronchitis or emphysema. Two persons reported frequently having cough and phlegm whereas none reported dyspnea.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study has shown that workers on a Danish stonewool plant with a considerable duration of exposure to MMVF have self-assessed respiratory health comparable to an unexposed workforce. The lung function in general is normal and almost similar in the exposed and the unexposed group, after controlling for differences in age and height. However, heavy smoking subjects exposed to MMVF seem to have an increased risk of airflow obstruction compared with subjects with a similar smoking history but without MMVF exposure. Despite this observation the regression models fail to correlate exposure length to a decrease in lung function or to an increase in the prevalence of respiratory symptoms. In fact, exposure duration tends to be associated with improvement in lung function, indicating a possible "healthy worker effect."

It could be argued that employment length is a poor surrogate marker of exposure. However, all employees had jobs in proximity to either the manufacturing process or in the handling/secondary processing of the stonewool and furthermore had a substantial employment length (median: 15 yr, range: 5 to 46 yr). Most workers had different positions within the factory during their employment, and this job rotation tended to reduce the variation in individual exposure levels. An experimental model for calculating the level of respirable fibers in the past MMVF production processes (13) indicates an average airborne respirable fiber concentration of 0.1 to 0.2 fiber/ ml in the examined period extending back to 1955, and the majority of the exposed subjects probably had average exposure levels within this range.

The prevalences of self-reported respiratory disease were low in both groups and thus our study power on this issue was modest, reflected in broad CI for the RR ratios (Figure 3). Respiratory symptoms were more frequently reported, but with no significant differences between the groups. The finding that respiratory symptoms were not correlated to exposure but essentially were associated with age and smoking parameters is consistent with other studies on this subject (3, 4, 14).

The regression models fairly accurate determined the influence of age and height on lung function in accordance with previously published reference values (10, 11) and detected the adverse effect of smoking. None of the lung function indices were significantly associated with exposure length. We examined for linear correlation of the explanatory variables and found no correlation between smoking amount and exposure length. However, age was significantly correlated to smoking as well as exposure length. Using standardized age- and height- adjusted lung function values (10, 11) in the regression models did not significantly modify the contribution of smoking or exposure to lung function. We are thus confident that a possible adverse effect of exposure was not camouflaged by the strong confounders, smoking and age.

We used TL_CO as a single measure to evaluate the prevalence of pneumoconiosis, as this marker in an epidemiological context has been found very effective in discriminating pneumoconiotic subjects from normal subjects (15). We found normal values for TL_CO as well as TL/VA. Furthermore the values in the exposed group and the comparison group were similar, as were the number of subjects with a TL_CO or TL/VA below 80% of predicted. This finding is reassuring and consistent with other studies which have found no radiographic evidence of excess risk of lung fibrosis (4, 5). The independent association of smoking status to TL_CO and TL/VA is probably an artefact, resulting from carbon monoxide back pressure in current smokers. However, the association with smoking amount is also found in ex-smokers and probably reflects emphysematous changes related to smoking.

In this study we controlled for the most important confounder, that is, smoking, but other confounders have to be considered. During production the exposed workers may have been exposed to asbestos, formaldehyde, polycyclic aromatic hydrocarbons, arsenic, and fumes from metal slags and curing ovens. Such exposure would have been dependent on employment period and job type, as these substances were present primarily in earlier phases of production (16). The sources of asbestos exposure were primarily personal protective equipment and thermal insulation. However, a small number of workers have been exposed to loose asbestos fibers from experimental products for a short duration (16). We found it important to examine if such coexposure to asbestos did influence the results. The proportion reporting some degree of occupational exposure to asbestos was not significantly different between the two groups, and although quantitative exposure data could not be obtained, most questionnaire data indicated only a brief or infrequent exposure to asbestos. This was indirectly confirmed by subgroup analysis on the asbestos- exposed group, which revealed no decrease in spirometric indices nor any lung diffusing capacity impairment in this group. Based on the assumption that the asbestos exposure was minimal and of minor importance in respect to nonmalignant respiratory disease, the exposed group was included in subsequent analyses.

To reduce the "healthy worker effect" we included subjects who had left the industry, provided they had 5 yr of employment in the stipulated period. However, this bias is not completely eliminated, demonstrated by the trend toward increase in most lung parameters with increasing employment length. Preemployment selection of healthy or unhealthy individuals into the industry constitutes another possible bias. Positive preemployment selection is not to be expected, as no particular health risk is assumed to be associated with the employment. Negative preemployment selection is demonstrated in this population, especially for workers with shorter employment length (17). This bias can be expected to be minimal by excluding subjects with less than 5 yr of employment.

A considerable number of subjects in the a priori selected groups chose not to participate in the study, and it was essential to acquire some knowledge about this group. The results of this nonresponder survey were very reassuring, as none argued respiratory illness as reason for not participating. The lung function testing showed expiratory flow values that were consistent with the values in the major groups. Apart from somewhat heavier smoking history among the nonresponders, no differences were found and it seems reasonable to assume that a higher participation rate would not have changed the overall conclusions.

In summary, the present study provides no consistent evidence of any harmful effect of long-term occupational exposure to MMVF. Of the spirometric indices, adjusted for age and height, only the FEV₁/FVC ratio was moderately lower in the exposed group than in the comparison group. Smoking history accounts for some of this difference and the residual variation is not correlated with the duration of exposure. TL_CO and TL/VA were normal in the exposed group and without adverse association to exposure. Thus, the present study indicates that pneumoconiosis has not developed in this workforce. The prevalence of respiratory symptoms and disease is not significantly higher in the exposed group and the symptoms are essentially associated with smoking parameters and not with exposure duration. The most prominent finding of the study is that a larger proportion of heavy smokers among the exposed workers had spirometric signs of airflow obstruction relative to a similar smoking group of workers in the comparison group. This finding may be the result of confounding from other exposure to lung hazards, but alternatively, an additive or synergistic action between smoking and fiber exposure on airflow obstruction can be speculated. It is well known that smoking reduces mucociliary clearance and depresses the phagocytic properties of the macrophages (18). Both mechanisms tend to favor a possible adverse effect of fibers on the small airways, owing to an impaired clearance function. Observations supporting a hypothesis of interaction between smoking and dust or fiber exposure have previously been published in association with foundry manganese dust (19), dust from coal and gold mines (20), and recently in a study on ceramic fibers (21), although it has not been reported in association with stonewool exposure.

A longitudinal study with a more detailed individual fiber-exposure assessment is necessary in order to further investigate whether MMVF exposure constitutes a hazard for developing obstructive lung disease, possibly by potentiating the adverse effect of smoking on lung function.