Variability in Quartz Exposure in the Construction Industry: Implications for Assessing Exposure-Response Relations

ABSTRACT

The aims of this study were to determine implications of inter- and intraindividual variation in exposure to respirable (quartz) dust and of heterogeneity in dust characteristics for epidemiologic research in construction workers. Full-shift personal measurements (n = 67) from 34 construction workers were collected. The between-worker and day-to-day variances of quartz and respirable dust exposure were estimated using mixed models. Heterogeneity in dust characteristics was evaluated by electron microscopic analysis and electron spin resonance. A grouping strategy based on job title resulted in a 2- and 3.5-fold reduction in expected attenuation of a hypothetical exposure-response relation for respirable dust and quartz exposure, respectively, compared to an individual based approach. Material worked on explained most of the between-worker variance in respirable dust and quartz exposure. However, for risk assessment in epidemiology, grouping workers based on the materials they work on is not practical. Microscopic characterization of dust samples showed large quantities of aluminum silicates and large quantities of smaller particles, resulting in a D 50 between 1 and 2 µm. For risk analysis, job title can be used to create exposure groups, although error is introduced by the heterogeneity of dust produced by different construction workers activities and by the nonuniformity of exposure groups. A grouping scheme based on materials worked on would be superior, for both exposure and risk assessment, but is not practical when assessing past exposure. In dust from construction sites, factors are present that are capable of influencing the toxicological potency.

MATERIAL AND METHODS

Exposure measurements were collected from construction workers subgroups that were included in an epidemiologic study on respiratory health effects. Workers performed the following tasks: concrete drilling, cutting recesses, cleaning of construction sites (sweeping), pointing, grinding mortar between bricks, inner wall construction, and demolition. Three workers (two carpenters, and one worker gluing concrete blocks), who experienced exposure generated by work of colleagues, were included as well. Concrete drilling workers were involved in drilling concrete with jackhammers and hammer drill; they also performed recess milling and sawing in either concrete or lime sandstone. Pointers were involved in filling joints with mortar or grinding mortar with hand-held grinders or jackhammers. Cleaners cleared or swept the work sites. The inner wall bricklayers used concrete blocks (a highly cellular material composed of quartzite, lime, and water). Demolition workers used jackhammers, drills, and excavators equipped with breakers for demolishing. Other tasks included welding, sawing, and clearing of rubble.

Personal respirable dust samples were collected from 34 construction workers on 1 to 3 different days in November and December 1999. Personal air sampling for respirable dust was conducted during full workdays (average duration of 61 / 2 hours), using Dewell-Higgins cyclones obtained from the Casella Group Ltd. (Bedford, UK), connected to Gilian Gilair5 portable pumps (Sensidyne Inc., Clearwater, Fla.) at a flow rate of 1.9 L/min. Of a total of 88 samples, 21 were not valid, due to irregularities or pump failure during measurements, leaving 67 samples for analysis. Five field blanks and 12 duplicate samples were included. After gravimetric determination of dust on the PVC filters (diameter 25 mm; pore size 0.2 µm), quartz was determined by infrared absorption spectrophotometry (IR) according to National Institute for Occupational Safety and Health (NIOSH) method 7602.

The limit of detection (LOD) was assessed as the average weight of the blank filters plus three times the standard deviation, and was 0.15 mg for respirable dust on the filters. Dust samples with dust levels below 0.15 mg (n = 5), were assigned a value of two-thirds of this limit divided by the average sampling volume (0.72 m3), which resulted in a limit of detection for dust measurements of 0.14 mg/m3 . The analytical limit of detection for α-quartz on the filters was 1.7 µg.

Based on four parallel stationary samples, the coefficient of variation (CV) was estimated to be 13% for respirable dust and 7% for respirable quartz. The percentage of quartz was calculated from the mass of quartz and the total amount of gravimetrically determined respirable dust on the filter. Samples with quartz levels below 1.7 µg (n = 4), were set at two-thirds of this value divided by the average sampling volume, which resulted in a limit of detection for quartz measurements of 1.6 µg/m3.

For a more detailed characterization of dust, six respirable dust samples, acquired during demolition, cutting recesses, grinding in both lime and cement based mortar, clearing rubble and pile top crushing, were studied with Scanning Electron Microscopy (SEM, Philips 515; Philips, Eindhoven, The Netherlands) and transmission electron microscopy (TEM, Philips CM12). Stationary sampling was performed on Nucleopore filters (polycarbonate) with a diameter of 25 mm and pore size of 0.2 µm, coated with carbon before sampling, in Dewell-Higgins cyclones, coupled with a vacuum pump (Becker Equipment Inc., Wuppertal, Germany) connected to a gas meter at a flow of 2 L/min. Before analysis, samples were coated with a thin layer of gold. Particle count, particle diameters and composition of the samples were estimated by TEM. The scanning electron microscope was used for analysis of morphology and was equipped with an elemental dispersive X-ray analyzer (EDAX; Philips) for chemical analysis. Hydroxyl radical generation in samples from drilling in concrete and pile top crushing was measured by electron spin resonance (ESR) spectroscopy. (18) Measurement of soluble transition metals was done in aqueous suspension with 0.08N HNO3 using inductively coupled plasma mass spectroscopy (IOPMS) after filtration through 0.2 µm filters.

Statistical Analysis

Distributions of dust and quartz exposure were examined to ascertain logarithmic distribution. Exposure levels were described for different jobs in terms of arithmetic and geometric means as well as the corresponding geometric standard deviations and ranges. Variance components were estimated using multiple linear mixed models. Job title was introduced as a fixed effect, while the worker identity was introduced as a random effect. The models have the following general form:

Y ijk =µ + β k + χ i(k) + ε j(ik) 

In this model, Y ijk represents the natural logarithm of the exposure concentration measured on the jth day of the ith worker in a group k; µ is the true underlying mean of log-transformed exposure averaged over all groups; β k is the fixed effect of group k; χ i(k) is the random effect of the i th worker in group k; and ε j(ik) is the random within-worker variation on day j for worker i in group k.

Separate models were constructed for three measures of exposure:respirable dust, quartz and percentage of quartz.It is assumed that χ i(k) and ε j(ik) , which are mutually independent, are normally distributed with zero means and variances bw σ2 yk  and ww σ2 yik, respectively. Measurements on the same worker were assumed to be correlated (compound symmetry covariance structure). Variances are estimated as between-worker (bw σ2yk 2) and within-worker (ww σ2yik 2) variance components. To assess whether a more restrictive model could also describe the data when job was used as grouping scheme, variance components were assumed common for all jobs. The effect of pooling the variance components when job-title was used to group exposure data, was examined using the likelihood ratio test. The p-values for these tests were approximated by comparing –2 times the difference in log likelihoods between the models to a χ2 -distribution with 2*(number of jobs-1) degrees of freedom. A reduced model should be considered when the difference in –2 log likelihood between models is smaller than χ2 statistic for given p-value and degrees of freedom. For other exposure models the effect of pooling was not evaluated, because of the potential of a number of combinations: exposure samples comprised dust generated by more than one material, tool or worksite.

The hypothetical performance of a reduced exposure model was evaluated by percentage of explained between-worker variance. In addition, the attenuation ratio of the exposure-response relations (the hypothetical bias in the exposure-response relation towards zero due to non-differential misclassification) was calculated considering either a group-based strategy or an individual based strategy (model with worker only). For calculation of the attenuation ratio, the number of workers in each job-title category (n) and the number of repeated measurements (m) were averaged (n = 4.9, m = 1.97). Statistical analyses were performed with SAS statistical software (version 6.12, SAS Institute, Inc., Cary, N.C.). Statistical significance was reached at p < 0.05.
RESULTS

The mean respirable dust and mean respirable quartz concentrations were 2.4 mg/m3 and 0.40 mg/m3 , respectively. Hypothesis of log normal distribution of exposure data could not be rejected (Shapiro-Wilk statistic: 0.97; p = 0.1 and 0.98; p = 0.3 respectively). The ranges of exposure are large, in particular for quartz (GSD = 7.0). For respirable dust and percentage of quartz the geometric standard deviations were smaller (3.5 and 3.3, respectively) (Table I). The full-shift average exposure levels for respirable nuisance dust exceeded the Dutch maximum acceptable concentration (MAC) of 5 mg/m 3 in 15% (n = 10) of the measurements. Respirable quartz dust concentrations exceeded the Dutch MAC for respirable quartz (0.075 mg/m 3 ) in 58% (n = 39) of the measurements.

Comparison of the models with distinct variance components of each job title with the models with pooled between and within-worker variance components for job titles, showed that these were not statistically different for both dust, quartz and percentage of quartz (−2 log likelihood ratio test, χ2  statistic = 4.11, df = 12, p = 0.9; χ2 statistic = 4.03, df = 12, p = 0.9, and χ2 statistic = 11.3, df = 12, p = 0.5, respectively).

Introducing job title with pooled variance components, compared to the models with worker only, resulted in statistically significant different models (χ2 statistic = 15.8, df = 8, p = 0.046, and χ2 statistic = 29.4, df = 8, p = 0.0003 for dust and quartz, respectively).

When comparing respirable dust and quartz exposure models, which considered job title as grouping variable, the model for quartz explained more of the between-worker variance (84%) than the model for respirable dust (41%). For quartz, grouping by job title also resulted in less predicted attenuation of a hypothetical dose-response relation (from 14% to 4%). For respirable dust, the predicted attenuation dropped from 36% for the individual approach to 17% for the grouping approach (Table II).

Different exposure models for respirable dust did not alter the within-worker variability (Table II). The model with tools and materials explained 88% of the between-worker variance, opposed to 41% for the model with job-title only. For respirable quartz, an exposure model based on material or material and tools, resulted in a higher within-worker variance (ww σ2k = 1.5 and 1.2, opposed to 1.0 for the model with worker only). In these two models 96 and 93% of between-worker variance was explained compared to the model with worker only, resulting in low between-worker variances (bw σ2k = 0.12 and 0.22, respectively).

Six stationary respirable dust samples, taken for characterization of dust particles, ranged from 0.1 mg/m3 when clearing rubble in the open air to 4 mg/m3 when demolishing with hand held hammers. The size distribution of the samples was almost the same in all samples (Figure 1), and showed that the number of small particles was much larger than would be expected from the characteristics of the respirable dust convention. The mean diameter size (D50) was between 1 and 2 µm.

Morphological analysis (Table III) showed that most samples contained aggregates, some round particles, some soot particles, and fume and that every sample contained fibrous particles. The fibers either consisted of organic matter or of gypsum. X-ray analysis and X-ray mapping showed an abundance of aluminosilicates in all samples and quite a lot of quartz particles in samples from demolition, cutting recesses and grinding mortar (lime) (Table III). The other samples consisted of aluminosilicates in larger amounts than quartz. SEM and TEM images and X-ray analysis of a sample from demolition work, illustrate the morphology and composition of samples (Figure 2). Especially the amount of gypsum, fibers, calcium-rich (Ca), iron-rich (Fe), aluminum-rich (Al) particles differed among samples. Electron spin resonance analysis showed that hydroxyl radical activity was 13 and 22 times larger in samples from drilling in concrete and pile top crushing, compared to activity on blank filter. Concentrations of soluble aluminum were high in both samples (24 and 959 µg/L, respectively).
