Historical Respirable Quartz Exposures of Industrial Sand Workers: 1946-1996

Background

Besides a clear relationship to silicosis, crystalline silica - quartz - has been associated with lung cancer, nonmalignant renal disease, and auto-immune disease. To study diseases associated with crystalline silica further, NIOSH conducted a cohort mortality study of workers from 18 silica sand plants, which had quarry , crushing , and bagging operations to produce industrial sand. Twelve of these plants also had grinding mills to produce fine silica powder. The historical crystalline silica exposures of workers at these plants were estimated to facilitate exposure-response analyses in the epidemiologic study.

Methods

NIOSH obtained personal respirable dust measurement records from Mine Safety and Health Administration (MSHA) compliance inspections at all 18 plants and from the archives of seven plants which had collected samples. These samples had been analyzed for quartz content by x-ray diffraction. Although no personal samples were available before 1974, impinger dust measurements were reported for 19 silica sand plants in 1946; these data were converted and used to estimate exposures prior to 1974. Statistical modeling of the samples was used to estimate quartz exposure concentrations for workers in plant-job-year categories from the 1930s when mortality follow-up of the cohort began until 1988 when follow-up stopped.

Results

Between 1974 and 1996, there were 4,269 respirable dust samples collected at these 18 plants. The geometric mean quartz concentration was 25.9 µm/m3 (GSD = 10.9) with a range from less than 1 to 11,700 µm/m3. Samples below 1 µm/m3 were given a value of 0.5 µm/m3. Over one-third of the samples (37%) exceeded the MSHA permissible exposure limit value for quartz (PEL = 10 µm/m3 / ( % quartz + 2 )) and half (51%) of the samples exceeded the NIOSH recommended exposure limit (REL = 50 µm/m3). The samples were collected from workers performing 143 jobs within the 18 plants, but too few samples were collected from many of the jobs to make accurate estimates. Therefore, samples were combined into 10 categories of jobs performing similar tasks or located within the same plant area.

Conclusions

The quartz concentrations varied significantly by plant, job, and year. Quartz concentrations decreased over time, with measurements collected in the 1970s significantly greater than those collected later. The modeled exposure estimates improve upon duration of employment as an estimate of cumulative exposure and reduce exposure misclassification due to variation in quartz levels between plants, jobs, and over time.

METHODS

Historical respirable quartz measurements were sought from a variety of sources to estimate quartz exposures across all jobs within the 18 plants in the industrial sand cohort. NIOSH obtained the greatest number of personal, respirable dust measurement records for these plants from the Mine Safety and Health Administration (MSHA). Industrial hygiene samples had been collected in all 18 plants by MSHA inspectors beginning in 1974 to determine compliance with MSHA exposure standards [NIOSH, 1996]. The data for personal exposure samples included the sampling date, contaminant code, airborne concentration, occupation, permissible exposure limit, percent silica, silica concentration, standard industrial classification, and the mine at which the sample was obtained. This data system is currently maintained by MSHA Technical Support in Denver, Colorado.

Personal respirable quartz measurements were also available from the archives of seven of the 18 plants. These plants had collected respirable quartz samples in the same manner as MSHA to document workers' silica exposures and guide their dust control intervention efforts. These measurements were coded into a computer spreadsheet using the same variable fields as the MSHA data in order to combine the data if they were found to comparably represent workers' exposures.

These samples were collected by having workers wear portable battery-powered pumps drawing air at 1.7 liters per minute (L/min) through cyclone pre-separators followed by dust collection filters. The cyclone pre-separator selectively allowed "respirable" dust-mass median aerodynamic diameter of 3.5 mm-to pass on to the filter [Lippman, 1970; Hearl, 1997]. The filters were weighed to determine the mass of dust collected and those with a mass of at least 0.1 mg were analyzed for quartz content by x-ray diffraction [Watts and Parker, 1995]. The lowest quartz concentrations that could have been measured directly on the samples with this method would have been approximately 20-30 µg per sample or 25 µg/m3 over an 8-hour work day [NIOSH, 1994].

If the dust mass on the filters was less than 0.1 mg, the reported airborne quartz concentrations were derived by multiplying the dust mass on the filters by the percentage of quartz believed to be in the dust, rather than analyzing the quartz content directly on the filters. The percentage of quartz was estimated by analyzing the quartz content of bulk dust or high-volume air samples. Using this technique, quartz concentrations lower than 25 µm/m3 were calculated for samples with low dust mass. Quartz concentrations as low as 1 µm/m3 were reported with sample concentrations lower than 1 µm/m3 typically reported as zero. In this study, all sample concentrations lower than 1 µm/m3 were given the value 0.5 µm/m3 , the least biased estimator of concentrations between zero and one [Hornung and Reed, 1990]. Samples greater than 15 µm/m3 respirable quartz dust were deleted because it was believed these samples were outliers not representative of sand plant environments and were possibly transcription errors [Watts and Parker, 1995]. Only three samples were greater than 15 µm/m3 .

As many jobs within the 18 plants were represented by too few samples, accurate exposure estimates could not be made for these jobs over time. Therefore, samples were combined into 10 broad categories of jobs performing similar tasks or located within the same area of the plant (Table I). These categories were similar to job categories used by other researchers [Hatch et al., 1947; Severns, 1979]. Statistical modeling was used to stratify the samples into plant and time categories and to estimate the quartz concentrations for workers in each plant-job-year category.

No personal respirable quartz measurements were available from MSHA or the companies before 1974. However, researchers with the Industrial Hygiene Foundation of America, Inc. had conducted a cross-sectional dust exposure assessment study of 19 silica sand plants in 1946 [Hatch et al., 1947]. The dust samples in this study were obtained using midget impingers with n-propyl alcohol as the collecting fluid. Both breathing zone and general area samples were collected. Where workers were stationed at one position, samples were collected in the breathing zone within a few inches of their noses. In many cases it was necessary to follow workers around a department to obtain representative samples. Where no obvious source of dust was present, the researchers collected general air samples to estimate dust exposures of groups of workers.

The dust particles in alcohol suspension were placed in a Sedgwick Rafter Cell and allowed to settle for 20 min. All particles at or near the bottom of the cell, smaller than 5 mm, were then microscopically counted in five widely spaced locations within the cell. Since the toxic action of crystalline silica particles had been shown to increase progressively with a decrease in size, these small particles were considered to be the most important for evaluating respiratory hazard. The results of these counts were reported as million dust particles per cubic foot of air (mppcf).

To estimate the quartz concentration of the dust particles smaller than 5 µm, airborne dust was collected with the I-H-F sampler which consists of a high volume air pump drawing air through two filters one foot square at 30 cubic feet per minute (849 L/min); the filters were composed of felted layers of salicylic acid crystals [Hatch et al., 1947]. The filters were placed in alcohol to dissolve the acid and the dust was filtered and washed. Samples of settled dust from less accessible areas (rafters) were also collected for size separation and quartz analysis. The dust from either the I-H-F samplers or settled dust was agitated in alcohol to provide a uniform suspension and allowed to settle for 40 min and then the supernatant was withdrawn by suction. This procedure was repeated 8-10 times, reportedly separating the dust into two portions-one less than 5 mm and the other greater. The amount of quartz in the portion of dust less than 5 mm was determined by x-ray diffraction analysis.

Unfortunately, the identity of the 19 plants in the Hatch study were not available; therefore, the impinger measurements could not be directly compared to the plant-specific quartz samples collected since 1974. However, these impinger measurements could provide estimates of the airborne quartz exposures for specific jobs in the silica sand industry in the late 1940s across all plants. The impinger dust measurements in mppcf had to be converted to mass per volume (µm/m3) to compare them to the respirable quartz measurements collected with a cyclone pre-separator. Based on comparisons of side-by-side impinger and respirable dust cyclone measurements reported by other researchers, the impinger dust count measurements in mppcf were converted to respirable dust mass in µm/m3 by multiplying them by 0.1 [Ayer et al., 1973; Rice et al., 1984; Sheehy and McJilton, 1987]. These converted values were then multiplied by 1000 and by the average percentage of quartz found in the historical dust mass samples to provide an estimate of the respirable quartz concentration in µm/m3 .

Exposure Standards

Exposure standards for quartz have been developed to prevent respiratory disease; the historical exposure measurements were compared to these criteria. MSHA incorporated the 1973 American Conference of Governmental Industrial Hygienists (ACGIH) threshold limit value (TLV) as the permissible exposure limit (PEL) for respirable mass of mineral dust containing quartz [CFR, 1997]. This standard employs the following formula:

PEL = 10 mg/ m3 / (% quartz + 2)

where the "% quartz" is the percent-by-weight of quartz in the personal respirable dust sample. The PEL is measured as respirable dust mass, in contrast to the NIOSH recommended exposure limit (REL) of 50 µm/m3 , which is measured as a time-weighted respirable quartz level for up to 10 h day during a 40-hour work week [NIOSH, 1992]. The ACGIH currently recommends a TLV for respirable quartz of 100 µm/m3 [ACGIH, 1999].

Statistical Analysis

Statistical Analysis System (PC-SAS) computer software was used for all statistical analyses [SAS, 1996]. Because the distributions of the quartz measurements were right skewed, the measurements were transformed using the natural logarithm. Therefore, geometric means (GM) and geometric standard deviations (GSD) of the measurements were calculated for each plant, job, and year category. Analysis of variance and Student's t-test were used to test differences in the mean quartz measurements across various plant, job, and year categories. Analysis of variance using general linear models was used to evaluate the effect of plant, job, and year on quartz concentrations, and to estimate the adjusted quartz exposure levels by the plant, job, and year categories.
RESULTS

A total of 4,269 respirable quartz samples were available for estimating historical exposures in the 18 silica sand plants between 1974 and 1996; 2,975 samples collected by MSHA inspectors and 1,294 samples collected by seven companies. Approximately one-fourth (23.7%) of the samples were recorded to be 1 µm/m3 or less and 1,585 samples (37%) were less than 25 µm/m3. The MSHA and company-collected data were combined into a single dataset, since no significant differences were found by analysis of variance (P = 0.54) between the quartz measurements collected by MSHA and those collected by the seven companies across plants, jobs, and years. Also, no trend was observed indicating that MSHA measurements were usually either higher or lower than company-collected measurements. The overall geometric mean quartz concentration was 25.9 µm/m3 (GSD = 10.9) and the highest concentration was 11,700 µm/m3 . Over one-third (37%) of all samples exceeded the MSHA-PEL and one-half (51%) of the samples exceeded the NIOSH-REL.

Summary statistics of the personal respirable quartz measurements collected between 1974 and 1996 are presented by plant in Table II. The plants are presented in order of increasing adjusted geometric mean dust concentration; the geometric means were adjusted by job category and year. The number of samples collected, geometric means, and maximum quartz concentrations varied dramatically across the plants. These data indicate that silica sand workers have historically encountered overexposure to quartz. Even among those plants with relatively low geometric mean quartz concentrations, a portion of the samples exceeded the PEL; in only one plant (4A)-represented by 38 samples-did none of the samples exceed the PEL.

Summary statistics of the personal respirable quartz measurements collected between 1974 and 1996 are presented by job category in Table III. Geometric means and standard deviations of the quartz concentrations were adjusted by plant and year. Screening and bagging jobs tended to have the greatest quartz exposures; approximately one-half of the samples from these two job categories exceeded the PEL. Milling, bagging, and loading jobs encountered dusts with the greatest percentage of quartz. Quarry and administration jobs had the lowest quartz exposures.

The geometric mean respirable quartz measurements by year, adjusted for plant and job category, are presented in Figure 1. This figure shows that respirable quartz levels have fallen dramatically since 1974. The average quartz exposure in 1974-based on only 44 samples-was 100 µm/m3 , dropping steadily until 1985 when the average annual quartz exposure was 16 µm/m3 . The annual geometric mean quartz exposure has remained below 20 µm/m3 since 1982 and has remained below 10 µm/m3 since 1990. Mining regulations were strengthened in 1977 with the passage of the Federal Mine Safety and Health Act. These measurements indicate that quartz exposures were already falling before 1977 in the silica sand industry, and continued to decrease after 1977. The data also indicate that quartz exposures dropped further after 1981 and 1991. They have been fairly stable since 1992, averaging about 5 µm/m3 . The quartz exposure levels appeared to increase somewhat in 1988 after MSHA changed the analytical standard for quartz analysis (Minusil V) to more accurately analyze for quartz and conform to international methodology [Watts and Parker, 1995].

Summary statistics of the personal respirable quartz measurements collected between 1974 and 1996 are presented by year category in Table IV. These data show that silica sand workers have historically had increased risk of overexposures to quartz. Over 40% of the personal respirable samples were not in compliance with the quartz standard between 1974 and 1984 and 57% exceeded the NIOSH-REL during these years. However, the proportion of samples exceeding respirable quartz criteria has been steadily declining since the early 1970s. These measurements may be compared to a NIOSH study of crushed stone workers conducted between 1979 and 1982 at limestone, granite, and traprock mines in which only 14% of the respirable quartz samples exceeded the MSHA-PEL and 25% exceeded the NIOSH-REL [Kullman et al., 1995].

In creating the historical exposure matrix for epidemiologic analysis, too few samples were available to estimate quartz exposures for each plant-job category (n = 180) for every year between 1974 and 1988. Therefore, the 18 plants were reduced to four categories (Plants 1-6; Plants 7-9; Plants 10-14; and Plants 15-18) and the years grouped into three categories (1974-1979; 1980-1984; and 1985-1988) for a total of 120 plant-job-year categories.

Combining the 18 plants into four categories was accomplished by comparing the least-squares geometric mean quartz concentrations (adjusted by job and year) of the plants and grouping those plants with very similar means. The least significant difference test was used to guide grouping of the plants. Divisions between the plant groups attempted to maximize the similarity of the geometric mean quartz concentrations of plants within a group, while maximizing the difference between groups. The geometric mean quartz concentrations for plants within a category were more closely related than plant concentrations outside their category. The geometric mean quartz concentrations of the four plant categories were significantly different from each other. The letter codes in Table II indicate which plants were grouped together.

The same approach was used to combine the years into three time categories. The yearly least-squares geometric mean quartz concentrations (adjusted by plant and job) were compared and years with very similar means were grouped. The least significant difference test was also used to guide grouping of the years.

The adjusted geometric mean quartz concentrations for the plants were usually lower than the unadjusted mean because, in general, more measurements were collected during the early years when quartz exposures were higher and greater weighting was given these high-early samples when unadjusted mean exposures were calculated. The adjusted means were used for decision making, because these means were less influenced by dramatically differing numbers of samples across the plant-job-year categories.

A statistical model was used to predict the quartz exposure estimates from 1974 to 1988 for the 120 plant- job-year categories (four plant, ten job, and three year categories). The independent variable terms in the model were categorical (plant, job, and year) and were highly significant (P < 0.001). The square of the regression coefficient for this model was r 2 = 0.21. All two-way interaction terms between the independent variables were also statistically significant; however, they were not included in the model to predict quartz exposure estimates because some plant-job-year cells contained few or no samples and including the interaction terms often led to either unreasonably high or low exposure estimates in these cells.

Since no personal respirable dust samples had been collected before 1974, quartz exposures before this year were estimated using impinger dust measurements which had been collected by researchers in 1946 [Hatch et al., 1947]. Table V presents a summary of the impinger samples collected during this study grouped into eight of the 10 job categories of 1974-1996 data. No impinger measurements were available for the "other" and "administrative" job categories. Impinger samples collected across all job categories (overall) in 1946 were used to estimate exposures for the "other" job category. The measurements across all job categories were selected for the "other" job category because-as shown in Table III-the geometric mean of the "other" job category was similar to the overall geometric mean of the 1974-1996 measurement data. The "administrative" job category was given the value of approximately one-fifth the overall median of the impinger measurements across all job categories in 1946 because-as shown in Table III-the measurements collected between 1974 and 1996 indicated that administrative jobs were considerably lower than other silica sand jobs.

As the identities of the plants where the impinger samples were taken were not known, these samples could not be used to directly estimate quartz exposures for the 18 plants in the cohort study. Instead, an indirect plant adjustment was made to estimate quartz exposures for the years 1946 through 1974 using the following steps.

First, for the years 1974-1979, the ratios of the geometric means of the plant-job-specific respirable quartz measurements to the geometric means of the 10 job-specific respirable quartz measurements were calculated for each of the four plant categories. This ratio was multiplied by the median quartz level of the 1946 job-specific measurements (in µm/m3) to estimate the predicted plant effect on job exposures. This adjustment assumes that the differences in concentrations across the four plant categories remained relatively constant between 1946 and 1974. The medians of the 1946 job-specific estimates were chosen because they most represent the geometric means of these measurements, which is consistent with using the geometric means of the later measurements.

Second, to adjust for exposure changes between 1946 and 1974, exposure estimates were incrementally changed each year in a linear fashion. Exposures may have actually fluctuated up and down during these years. On the other hand, the 1946 study commissioned by the silica sand industry probably indicated concern about silicosis risk among employees, possibly leading to process changes or environmental controls to reduce workers' exposures. Therefore, exposures more likely changed in a stepwise fashion across the various plants as controls were implemented. However, the specific points in time when dust control measures may have been instituted at the various plants remains unknown. Therefore, changing the quartz levels at a linear rate was assumed to be the least biased estimator of quartz exposure changes over time. However, the yearly rate of change varied by job category depending of the absolute difference between the 1946 median quartz exposure estimate and the 1974-1977 geometric mean. Figure 2 provides a graph of the predicted quartz exposure estimates of bagging workers in the four plant categories for the years 1946-1988.

Since some cohort members were employed before 1946 and no dust exposure measurements were available before this date, exposures for years pre-1946 were assumed to be the same as exposures in 1946. No information was available indicating industry changes which may have resulted in exposure increases or decreases before 1946.

Interestingly, the quartz concentrations were found to change little between 1946 and the 1970s for quarry, wet process, drying, and milling workers. This was not surprising for quarry and wet process jobs, because they had the lowest exposures in 1946. Therefore, there was less incentive to reduce exposures for workers in these jobs. But it is unknown why drying and milling job exposures, which were some of the higher quartz exposures in 1946, were not shown to decrease over 30 years. Possibly quartz exposures did decrease over time for these jobs, but the decrease was just not reflected by historical sampling. Or, perhaps dust control was not implemented for these jobs at these particular plants to the extent that it was for bagging, loading, and screening jobs.
