Pii: S0003 4878(99)00028 9

PII: S0003-4878(99)00028-9

Ann. occup. Hyg., Vol. 43, No. 4, pp. 247±255, 1999 # 1999 British Occupational Hygiene Society Published by Elsevier Science Ltd. All rights reserved Printed in Great Britain. 0003±4878/99/$20.00 + 0.00

Determinants of Exposure to Inhalable Particulate, Wood Dust, Resin Acids, and Monoterpenes in a Lumber Mill Environment KAY TESCHKE,*} PAUL A. DEMERS,$ HUGH W. DAVIES,$ SUSAN M. KENNEDY,$ STEPHEN A. MARION* and VICTOR LEUNG$ *Department of Health Care and Epidemiology, University of British Columbia, Vancouver, BC, V6T 1Z3, Canada; $Occupational Hygiene Program, University of British Columbia, Vancouver, BC, V6T 1Z3, Canada In a lumber mill in the northern inland region of British Columbia, Canada, we measured inhalable particulate, resin acid, and monoterpene exposures, and estimated wood dust exposures. Potential determinants of exposure were documented concurrently, including weather conditions, tree species, wood conditions, jobs, tasks, equipment used, and certain control measures. Over 220 personal samples were taken for each contaminant. Geometric mean concentrations were 0.98 mg/m3 for inhalable particulate, 0.49 mg/m3 for estimated wood dust, 8.04 mg/m3 for total resin acids, and 1.11 mg/m3 for total monoterpenes. Multiple regression models for all contaminants indicated that spruce and pine produced higher exposures than alpine ®r or mixed tree species, cleaning up sawdust increased exposures, and personnel enclosure was an e€ective means of reducing exposures. Sawing wood in the primary breakdown areas of the mill was the main contributor to monoterpene exposures, so exposures were highest for the barker operator, the head rig operator, the canter operator, the board edgers, and a roving utility worker in the sawmill, and lowest in the planer mills (after kiln drying of the lumber) and yard. Cleaning up sawdust, planing kiln-dried lumber, and driving mobile equipment in the yard substantially increased exposures to both inhalable particulate and estimated wood dust. Jobs at the front end of the sawmill where primary breakdown of the logs takes place had lower exposures. Resin acid exposures followed a similar pattern, except that yard driving jobs did not increase exposures. # 1999 British Occupational Hygiene Society Published by Elsevier Science Ltd. All rights reserved. Keywords: wood dust; occupational exposure; cross-sectional studies; epidemiology

INTRODUCTION

In recent years, health and safety personnel in the primary manufacturing sectors of the forest industry have begun to focus attention on wood dust exposures, largely due to the designation of wood dust as a known human carcinogen by the International Agency for Research on Cancer (IARC, 1995). This has led to concern about other respiratory outcomes of wood dust exposure, including asthma, air¯ow obstruction, and mucous membrane symptoms (Demers et al., 1997). Because it is well established that exposure to dust from di€erent tree species can Received 18 September 1998; in ®nal form 3 February 1999. }Author to whom correspondence should be addressed. Tel.: +1-604-8222041; Fax: +1-604-8224994. 247

have di€erent health e€ects, there is also interest in investigating the natural components and contaminants of woods as a means of distinguishing their toxic potentials. In the summer of 1996, we conducted a cross-sectional study of the association between wood-related exposures and non-malignant respiratory outcomes in a lumber mill in the northern inland region of British Columbia, Canada. This mill processed spruce (Picea englemanii, Picea glauca, and a hybrid of the two), pine (Pinus contorta), and alpine ®r (Abies lasiocarpa, called `balsam' colloquially). The study included measurements of personal airborne exposures to inhalable particulate, resin acids, and monoterpenes. In addition, we collected information about weather conditions, tree species, wood condition, lumber mill equipment, jobs, tasks, venti-

248

K. Teschke et al.

lation and enclosures to allow us to identify factors which contributed to increasing exposure levels, as well as factors which e€ectively reduced exposures. These data were used to develop an exposure model which would allow estimation of exposures for epidemiological analyses. The model could also be used to guide the location and design of control measures. The results of the exposure measurements and the data analysis to model determinants of exposure are presented here. More descriptive details on the dust, monoterpene, and resin acid measurements are reported elsewhere (Demers et al., 1998).

METHODS

Sampling design Personal air samples were collected over a onemonth period in the season expected to produce the highest exposure levels and respiratory symptoms, i.e., the dry period of the summer. Attempts were made to sample on at least 4 occasions all jobs in the production areas (a sawmill and 2 planer mills), as well as the wood kiln, yard, and carpentry shop (other maintenance areas were excluded). Jobs expected to have high and/or variable exposures were sampled more frequently. Individual workers within each job were randomly selected for two measurements each on randomly selected days. Processing of alpine ®r in the sawmill was done on di€erent days than processing of spruce and pine. Because employee complaints indicated that there may have been di€erences in exposures and/or respiratory symptoms according to the type of wood, air sampling in the sawmill was strati®ed according to the species being processed. Most air sampling was conducted during the day shift to be concurrent with lung function testing. Because clean-up jobs were performed only on graveyard shifts, they were sampled during that shift. Inhalable particulate Inhalable particulate was collected on 0.45 mm pore size, 25 mm diameter Te¯on ®lters (Costar, USA) mounted in 7-hole samplers (JS Holdings, UK) attached to the employee's lapel as close as possible to the breathing zone. Air was drawn through the ®lters for a full shift (7±8 h) at a ¯ow rate of 2.0 l/min, calibrated using an automated soap-®lm ¯ow meter (Gilibrator, USA) at the work site, before and after sampling. Filters were pre- and post-weighed (triplicate weighings) on a microbalance (Sartorius M3P-000-V001, Germany). Prior to pre-weighing, ®lters were equilibrated to a stable temperature (2020.58C) and relative humidity (502 5%) for 48 h. Prior to post-weighing, ®lters were dessicated for 3 days, then equilibrated to a stable temperature (2020.58C) and relative humidity (502 5%) for 48 h. Two ®eld blanks were analyzed for

every 10 samples to allow correction for humidity or other background e€ects on the ®lters. The mass detection limit was calculated as 3 times the standard deviation (SD) of 24 lab ®lter blanks, and was 0.013 mg.

Resin acids Abietic and pimaric acids, the most abundant resin acids naturally occurring in wood, were quanti®ed using a method developed in our laboratory, based on a technique originally developed to quantify the resin acid component of solder fumes (Pengelly et al., 1994; Demers et al., 1998). After post-weighing, the inhalable dust ®lters were extracted in dichloromethane. The resin acids were derivatized to their methyl esters, then quanti®ed using gas chromatography/mass spectrometry. The samples were analyzed in four batches, one for each week of sampling. Limits of detection were calculated for each batch as the mean plus 3 times the SD of the lowest concentration standard solution detected by the mass spectrometer. For pimaric acid, the mass limits of detection ranged from 0.05 to 0.50 mg, and for abietic acid, they ranged from 0.05 to 7.85 mg (the high limits of detection were for batch 2 only). Samples with exposure concentrations less than the detection limit were assigned a value of the detection limit divided by the square root of 2 (Hornung and Reed, 1990). For the purposes of the determinants of exposure model, a total resin acid concentration was calculated as the sum of abietic and pimeric acids.

Wood dust There were opportunities for airborne exposures to particulate other than wood dust (e.g., dust from soil in the yard), however the particulate sampling method used is non-speci®c. To consider wood dust separately, concentrations were estimated using resin acids as the basis, since these compounds are characteristic of wood, but not of other dusts present in the mill setting. The resin acid content of each particulate sample was calculated as a percent of particulate by weight (summary data presented in Demers et al., 1998). Because resin acid content varies between tree species, and within species according to tree age, tree part, growing environment, and storage conditions, the resin acid contents could not be used to calculate wood concentrations of individual samples. We therefore decided to average the resin acid content for groups of samples and use these averages as the basis for calculating wood dust concentrations, as follows. The occupational hygienists who performed the exposure measurements classi®ed the likelihood of exposure to wood versus other particulates in each job, based on proximity to dust-generating sources

Exposure to particulate, wood dust, resin acids, and monoterpenes in a lumber mill environment

and tasks. All jobs were assigned to one of four categories: 1. all particulate expected to be wood; 2. majority of particulate expected to be wood; 3. majority of particulate expected to be from nonwood sources; and 4. almost all particulate expected to be from nonwood sources. The mean resin acid content for each category was then used as an indicator of the relative wood dust content of the particulate in each group of jobs. In category 1, with the highest mean resin acid content, samples were assigned an estimated wood dust concentration equal to the inhalable particulate concentration (i.e., 100% of the particulate was assumed to be wood). For each of the other categories, an estimated wood dust concentration was calculated by weighting the inhalable particulate concentration by the mean resin acid content relative to that of category 1. Category 2, majority wood dust, had a weight of 70%; category 3, majority non-wood sources, had a weight of 40%; and category 4, almost all non-wood sources, had a weight of 10%. (Demers et al., 1998). Monoterpenes The volatile and odiferous component of softwood resins, the monoterpenes (including a-pinene, b-pinene, and D3-carene), was collected using SKC 575-003 passive samplers with 300 mg of Anasorb 727 as the sorbent (Eriksson et al., 1994; Demers et al., 1998). A sampler was attached to each participant's lapel for the duration of the shift (7±8 h). After sampling, the sampler was sealed, refrigerated, and returned to the laboratory within 7 days. Monoterpenes were desorbed in carbon disul®de, then quanti®ed using gas chromatography with ¯ame ionization detection. The retention-time peaks were 2.87 min for a-pinene, 3.34 min for b-pinene, 3.72 min for D3-carene, and 4.85 min for undecane. A number of peaks with retention times between 3.72 and 4.85 min, of comparable size to those of the three monoterpenes were also quanti®ed as `unidenti®ed wood volatiles'. Mass spectra of these peaks were all similar and all were assigned an approximately 70% match by the NIST-92 mass spectra library (National Institute of Standards and Technology, USA) to the monoterpenes pinene, carene, and phellandrene. Limits of detection were calculated as the mean of 27 lab blanks plus 3 times the SD. The mass limits of detection were 0.3 mg for a-pinene, 1.4 mg for b-pinene, 0.2 mg for D3-carene, and 0.9 mg for the unidenti®ed wood volatiles. Samples with exposure concentrations less than the detection limit were assigned a value of the detection limit divided by the square root of 2 (Hornung and Reed, 1990). For the purposes of the determinants

249

of exposure model, a total monoterpene concentration was calculated as the sum of a-pinene, bpinene, D3-carene, and the unidenti®ed wood volatiles.

Determinants of exposure Information on potential determinants of exposure was gathered at several levels: for each sampling day, job, or individual sample. Weather information was recorded on each sampling day, in the morning at the start of the shift, at midday, and at the end of the day shift. Temperature and relative humidity were measured using a Psychro-Dyne (Cole Parmer, Chicago, USA). Both measures were averaged over the working shift. Precipitation was recorded as a dichotomous variable (1=yes, 0=no); these measures were summed over the working day, as the amount of rain. Wind speed was recorded as an ordinal variable (0=no wind, 1=little wind, 2=moderately windy, and 3=very windy); the readings over the work day were averaged. The species of tree processed (alpine ®r, spruce and pine, or mixed) and the level of production was recorded on every sampling day for each department of the mill, based on reports from production management personnel. For each job sampled, the following information was gathered: the mill department; up to four potential sources of dust exposure (including 27 pieces of equipment such as saws, planing machines, and cyclones; and two tasks, cleaning up sawdust and manual lumber handling); the distance from these sources; whether the sources were ventilated; whether the sources were enclosed; whether the job was performed in an operator's booth or vehicle cab; an estimate of the percentage enclosure provided by the booth or cab; the source of air for a booth or vehicle cab; whether the air was ®ltered; the nearest source of natural ventilation (none, open door, open window, outside air); and the distance to the natural ventilation source. For each individual sample, the worker was queried at the end of the measurement period about which jobs he performed on that day (in most cases, the employees worked at only one job, but in a few cases, they rotated through 2 or 3 jobs in a shift), the number of times he used compressed air for `blowing down' sawdust, whether he used diesel equipment, whether he used a chainsaw, the percentage of the shift he spent in a booth or cab, the species of trees processed in his area of the mill (alpine ®r, spruce and pine, mixed), the condition of the wood (green, kiln dried, chips, or mixed), and how many cigarettes he smoked during the shift. A new variable, % enclosure, was created by multiplying the % of time spent in a booth or cab by the % enclosure provided by the booth or cab.

250

K. Teschke et al.

Data analysis Descriptive statistics (counts for categorical data, and means, ranges, standard deviations, and frequency distributions for continuous data) were calculated for all available variables. The distributions of the exposure variables were positively skewed and approximately log-normal, so exposures were logtransformed (base e) prior to analysis to improve the eciency of the models and to ensure that predicted concentrations would be greater than zero. The number of compressed air blowdowns was also log-normally distributed, so was also log-transformed (base e). We examined the correlations between all independent variables (pearson r). Among pairs with r=0.60, only one was selected for further inclusion in the determinants of exposure models (the one more logically explained as associated with exposure, or the one more strongly associated with exposure in univariate analyses). In all cases, variables which were strongly correlated were logically related in the mill operations (e.g., job `cleanup/janitor' was strongly related to the task `cleaning up sawdust', job `grader' was strongly related to the task `manual lumber handling'). The e€ect of the variable that was retained for analysis would reasonably include the e€ect of the correlated variable. We initially examined univariate associations between each remaining variable and the log-transformed exposure concentrations. Since several variables suggested that exposures di€ered depending on whether the samples were taken indoors in the production areas, versus outdoors in the log and lumber yards, a new `indoors/outdoors' variable was created, and used as a nesting variable for all other variables, except tree species and wood condition. No other interactions were suggested a priori, and there were too many potential interactions to usefully test without prior hypotheses. A manual backward stepwise regression procedure was used to create the exposure models. All variables (except subject) with pR0.25 in univariate modeling were initially o€ered in the model. An exception was that variables representing hypothesized exposure sources were also required to have a positive coecient in univariate analyses. `Sources' with negative coecients were likely to be negatively correlated surrogates of other sources of exposure, and therefore better excluded from the model. Variables with the highest p-valuesr0.10 were eliminated one at a time, then the model was re®tted until all included variables had p < 0.10. A general linear least squares model ®tting procedure was used initially for all variables which could be treated as ®xed e€ects (using JMP, version 3.2, SAS Institute, Cary, NC, 1997). Variables remaining were then reevaluated by ®tting the model using generalized estimating equations (using the xtgee procedure in Stata with subject as the group variable, link=iden-

tity, family=gaussian, and correlation=exchangeable; version 5.0, Stata Corp., College Station, TX, 1997), allowing us to take into account the possible correlation among repeated measures on subjects. Within-subject correlation could only reduce the signi®cance levels of associations found for the ®xed e€ects, therefore this step would only be expected to remove variables from the model. Variables with p r 0.10 according to the generalized estimating equation method were removed from the model (for the particulate models, only 1 variable was removed at this stage; for the monoterpene model, 2 variables were removed). Cook's D was used to check the ordinary least squares version of the ®nal model for in¯uential values, and residuals were plotted to look for patterns in the unexplained variance. RESULTS

The following descriptive information provides an overview of the mill. In the mid 1990s, the lumber mill received about 900,000 m3 of logs a year, mainly by truck. The logs were stored prior to processing in a large dry-land log yard, then debarked mechanically, sorted by size, and sent to one of three lines in the sawmill to make dimension lumber. Di€erent production characteristics of alpine ®r compared to spruce and pine required that it be processed in the sawmill on di€erent days. Most of the sawn lumber was dried to a moisture content of less than 20% in one of 5 dry kilns. Dried rough lumber was planed to a smooth ®nish in one of two planer mills, then graded and bundled in preparation for shipping. Except the cut-o€ saws and two of the primary breakdown saws, all of the saws and planing machines were enclosed, but only the trim saws and the planing machines had local exhaust ventilation. The log storage yard had a dirt surface; the lumber storage yard was paved. Bark waste was burned on site; wood waste was chipped and sold to pulp mills. 235 subjects participated in the cross-sectional study. Exposure measurements were made on 112 subjects in 37 jobs in 6 departments. Most of those who did not participate in the measurement study were maintenance, supervisory, oce or logging sta€ not designated for sampling. The number of samples per job ranged from 2 to 17, with a mean of 6.2. The number of samples per subject ranged from 1 to 5, with a mean of 2.0. A total of 229 ®lter (inhalable particulate, estimated wood dust, and resin acid) samples and 227 passive badge (monoterpene) samples were collected. Nine ®lter samples were excluded from analysis because of pump failure, large cross-shift drops in pump ¯ow, or damage to the ®lter cassette. Five passive samples were either lost or damaged during collection. Table 1 lists the arithmetic mean, geometric mean, geometric standard deviation, minimum, maximum, and percent of measurements less than

Exposure to particulate, wood dust, resin acids, and monoterpenes in a lumber mill environment

251

Table 1. Summary of inhalable particulate, estimated wood dust, total resin acid, and total monoterpene exposure concentrations Exposure 3

Inhalable particulate (mg/m ) Estimated wood dust (mg/m3) Total resin acids (mg/m3) Total monoterpenes (mg/m3)

N

AMa

GMa

GSDa

Minimum detected

%< LODb

Maximum

220 220 220 222

1.76 1.00 22.8 2.49

0.98 0.49 8.04 1.11

2.67 3.13 4.39 3.14

0.031 0.024 1.39 0.30

0 0 17.7 21.2

25.4 25.4 370.2 30.2

a

AM=arithmetic mean, GM=geometric mean, GSD=geometric standard deviation. %< LOD=percent of measurements less than limit of detection; for total resin acids, the LOD varied from 0.49 to 11.9 mg/m3 for samples in which both abietic and pimaric acid concentrations were below detection limits; for total monoterpenes, the LOD varied from 0.29 to 0.64 mg/m3 for samples in which a-pinene, b-pinene, D3-carene, and the unidenti®ed wood volatiles all had concentrations below detection limits. b

the detection limits for each of the exposures measured or estimated in this study. Correlations between the ®lter-based measurements (not logtransformed) were high: 0.78 between inhalable particulate and estimated wood dust; 0.60 between inhalable particulate and total resin acids; and 0.81 between estimated wood dust and total resin acids. Correlations between the total monoterpene measurements and the ®lter samples were negligible: ÿ0.02 for inhalable particulates; 0.08 for estimated wood dust; and 0.11 for total resin acids. The following variables were not o€ered in any of the multiple regression models either because they were not associated with exposure ( p>0.25), were strongly correlated with other selected variables, or were `sources' which were negatively associated with exposure: date, amount of rain, type of natural ventilation, distance from natural ventilation, booth or cab air source, number of cigarettes smoked, log drops, log cuto€ saws, planer feeder/tilt hoist, tilt hoist operator, chain height change, automatic lumber stackers, stacker operator, dry lumber o€bearers, lumber piles, manual lumber handling, lumber packaging machine, chip blowers, and road surface. The following variables were associated with at least one exposure in univariate analyses, and were o€ered in multiple regression models for the associated exposure, but were not entered in any

of the ®nal exposure models ( p r 0.10): week of sampling, day of week, average temperature, production level, booth or cab, sawmill, log yard, lumber yard, maintenance, logs, barkers, A5 debarker operator, model C edger operator, small board edgers, small board dropsorter, large board dropsorter, dropping lumber, trimmerman, grader, placing sticks to separate lumber for kiln drying, automatic lumber sorter attendant, packaging machine operator, yard labourer, cleanup/janitor, and dust-control cyclone. Tables 2a and 2b describe the ®nal determinants of exposure models for inhalable particulate, estimated wood dust, total resin acids, and total monoterpenes. Table 2a summarizes the ®t of the models. The residual within-subject correlations were low. Note that the generalized estimating equation model does not constrain these correlations to be positive; however, since negative correlations are implausible in this scenario, the negative correlations were interpreted as evidence that within-subject correlations were negligible. Predicted values were calculated using the predictors for each study measurement. Summary statistics indicate that the geometric means of the predicted values were the same as those for the raw data, but geometric standard deviations were somewhat smaller. Since the generalized estimating equations do not allow calculation of R2,

Table 2a. Fit of multiple regression models for inhalable particulate, estimated wood dust, total resin acid, and total monoterpene exposure, using generalized estimating equationsa

Number of observations Degrees of freedom chi2 p-value R2 b Within-subject correlation GMc of predicted values GSDc of predicted values Minimum predicted value Maximum predicted value a

Inhalable particulate

Estimated wood dust

Total resin acid

Total monoterpene

220 20 270.8 <0.00001 0.61 0.23 0.98 mg/m3 2.16 0.13 mg/m3 7.71 mg/m3

220 25 452.6 <0.00001 0.73 0.21 0.49 mg/m3 2.66 0.036 mg/m3 5.05 mg/m3

220 18 395.8 <0.00001 0.66 ÿ0.023 8.04 mg/m3 3.32 0.578 mg/m3 190 mg/m3

222 17 1078 <0.00001 0.80 ÿ0.17 1.11 mg/m3 2.78 0.46 mg/m3 14.8 mg/m3

xtgee procedure in Stata, with subject as the group variable, link=identity, family=gaussian, and correlation=exchangeable. b the proportion of variance explained (R2) could not be calculated for the xtgee model, therefore this represents the ordinary least squares R2 for the same model, without taking into account the within-subject correlation. c GM=geometric mean, GSD=geometric standard deviation.

Tree species (3 categories) Spruce and pine (N=43) Alpine ®r (N=56) Mixed (N=130) Wood condition (4 categories) Green (N=126) Kiln dried (N=85) Chips (N=4) Mixed (N=14) Outdoors, in the yardb (N=49) Average relative humidity (continuous, %; M=53.6, SD=16.1) Average wind speed (ordinal, 0±3; M=1.46, SD=0.51) Enclosure (continuous, %; M=66.6, SD=41.2) Chainsaw useT (N=6) Buckerj (N=4) Diesel equipment useT (N=47) Le tourneau driverJ (N=5) Log loaderJ (N=7) Sawmill outfeed forklift driverJ (N=4) Kiln utility forklift driverJ (N=7) Planer mill infeed forklift driverJ (N=5) Planer mill outfeed forklift driverJ (N=5) Yard equipment operatorJ (N=6) Indoors, in lumber production areasb (N=180) Average relative humidity (continuous, %; M=50.3, SD=11.2) Enclosure (continuous, %; M=26.7, SD=42.3) Compressed air blowdownsT (continuous, ln no.; GM=0.34, GSD=5.42) Cleaning up sawdustT (N=18) Chainsaw useT (N=24) Log cut-o€ operatorJ (N=17) Deckman/log deck attendantJ (N=4) A4 and unclassi®ed barkers operatorJ (N=4) Headrig and large board edgerE (N=17) Head sawyerJ (N=4)

Independent Variables (descriptive data)

Ð Ð Ð Ð ref ÿ0.041 ÿ0.68 ÿ0.0097 Ð Ð 0.93 0.95 Ð 0.85 Ð Ð Ð Ð ÿ3.01 ÿ0.012 ÿ0.0056 Ð 1.27 Ð ÿ0.49 ÿ0.69 Ð Ð ÿ0.71

ref ÿ0.37 ÿ0.60

(0.35)

(0.24) (0.40)

(0.21)

(0.57) (0.004) (0.002)

(0.39)

(0.44) (0.39)

(0.006) (0.19) (0.004)

(0.12) (0.15) ref 0.33 1.85 0.052 ref ÿ0.039 ÿ0.70 ÿ0.019 Ð 1.35 Ð 2.81 1.97 2.69 0.90 1.43 1.30 1.91 ÿ1.31 ÿ0.012 ÿ0.011 Ð 1.70 Ð Ð ÿ0.75 Ð Ð Ð

ref ÿ0.28 ÿ0.94

(0.37)

(0.20)

(0.83) (0.78) (0.83) (0.48) (0.63) (0.74) (0.78) (0.61) (0.004) (0.002)

(0.53)

(0.006) (0.19) (0.007)

(0.21) (0.48) (0.26)

(0.12) (0.24

(SE)

b

b

(SE)

Wood dust model

Inhalable particulate model

Ð Ð Ð Ð ref Ð Ð ÿ0.009 ÿ1.24 Ð Ð Ð Ð Ð Ð Ð Ð Ð 0.69 ÿ0.013 ÿ0.0040 Ð 2.39 0.44 ÿ0.89 ÿ1.83 Ð Ð ÿ0.86

ref ÿ0.61 ÿ1.21

b

(0.45)

(0.25) (0.24) (0.33) (0.50)

(0.48) (0.006) (0.002)

(0.003) (0.43)

(0.19) (0.23)

(SE)

Total resin acid model

Ð Ð Ð Ð ref Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð Ð 1.38 Ð ÿ0.011 0.076 Ð Ð Ð Ð 1.20 0.60 0.58

ref ÿ0.62 ÿ0.68

b

(0.24) (0.23) (0.32)

(0.002) (0.024)

(0.19)

(0.12) (0.19)

(SE)

Total monoterpene model

Table 2b. Descriptive data, coecients, and standard errors for independent variables in multiple regression modelsa of inhalable particulate (mg/m3), wood dust (mg/m3), total resin acid (mg/m3), and total monoterpene (mg/m3) concentrations (all log-transformed, base e). All independent variables dichotomous except where otherwise noted.b=regression coecient; SE=standard error; N=number of measurements in dataset with this characteristic, for categorical variables; M=mean, SD=standard deviation, GM=geometric mean, GSD=geometric standard deviation, for ordinal and continuous variables; ref=reference category; Ð=variable not included in model

252 K. Teschke et al.

(0.20)

(0.20) (0.17) (0.11) (0.21) (0.19) (0.19)

(0.36)

xtgee procedure in Stata, with subject as the group variable. indoors vs outdoors is a single variable used as the basis for nesting all others, except tree species and wood condition; outdoors is the reference category T =task, J=job, E=equipment, D=department. b

a

(0.50) (0.58) (0.54)

(0.20) (0.31)

(0.16) (0.26)

Board edger operatorJ (N=4) Gang edger operatorJ (N=4) Roving utility workerJ (N=4) CanterE (N=16) Canter operatorJ (N=8) DescramblersE (N=34) TrimsawsE (N=29) Stacker labourerJ (N=16) Planer mill ID (N=34) Planer mill IID (N=39) Planing machinesE (N=25) Planer feeder operatorJ (N=8) Planer technicianJ (N=4) Scrap conveyorE (N=6) Carpentry machinesE (N=2) Intercept

ÿ0.97 Ð Ð Ð Ð Ð ÿ0.36 0.87 Ð Ð 0.70 0.69 Ð Ð Ð 3.90

(0.38)

Ð Ð Ð 0.55 Ð Ð Ð Ð Ð Ð 0.47 0.90 Ð Ð 1.50 1.83

(0.20)

(0.19) (0.29)

ÿ0.82 Ð Ð Ð Ð Ð Ð Ð Ð ÿ0.47 1.70 1.42 1.15 1.47 Ð 2.82

(0.23) (0.28) (0.39) (0.52) (0.39)

(0.47)

0.95 0.70 1.06 Ð 0.95 0.78 0.40 ÿ1.13 ÿ1.20 ÿ1.20 Ð Ð Ð Ð Ð ÿ0.077

(0.32) (0.35) (0.23)

Exposure to particulate, wood dust, resin acids, and monoterpenes in a lumber mill environment

253

a comparison of the geometric standard deviations of the raw data and the predicted values is a way of examining the proportion of variance explained. Alternatively, the models explained 61 to 80% of the ordinary least squares variance. Residuals were symmetrically distributed and no values were shown to be in¯uential (all Cook's DW 1.0). Table 2b lists the coecients and standard errors of the independent variables included in the models, as well as descriptive data about these variables. All the variables were modeled as ®xed e€ects, except within-subject correlation, included as a random e€ect. The variables included in the models are presented in logical groupings. For example, indoor jobs, equipment, and departments are listed in order of their appearance in the lumber mill production process. Only three factors had similar in¯uences on all four exposures. Spruce and pine produced higher exposures than alpine ®r which had higher exposures than mixed tree species (the latter result is not logical and may result from confounding related to the planer, kiln and yard locations where mixed species were present). Cleaning up sawdust (compressed air blowdowns for the monoterpenes) increased exposures. And personnel enclosure in booths or cabs (% enclosure) was an e€ective means of reducing exposures. Not surprisingly, there were many similarities between the models for the particulate-linked exposures (inhalable particulate, estimated wood dust, and total resin acids). Exposures decreased with increasing relative humidity, and also decreased in jobs at the log deck before the logs entered the mill. Exposures were substantially increased around the planing machines, especially for resin acids. Estimated wood dust and inhalable particulate exposures were also increased in many jobs or tasks involving mobile equipment use in the yard, but resin acid exposures were not. Monoterpene exposures were di€erently distributed in the mill: higher than average exposures were measured in many sawmill jobs and near sawmill machines at the beginning of the lumber sawing process (i.e., the headrig, canter, edgers, and trimsaws), whereas lower exposures were found in the two planer mills and the yard. DISCUSSION AND CONCLUSIONS

The determinants models describe factors associated with lumber mill exposures and provide valuable information for the purpose of locating and selecting exposure controls. In general, lumber handling operations, such as grading, sorting, pulling, and stacking lumber, did not contribute to any of the contaminant exposures; exposures in these jobs are accounted for by the environmental conditions in the mill production areas that are included

254

K. Teschke et al.

in the ®nal model. Sawing green wood was the main contributor to monoterpene exposures. These wood volatiles were almost completely released by the sawmill cutting operations and kiln drying, so that planer mill exposures were as low as those in the yard where no cutting was taking place. Cleaning up sawdust, planing kiln-dried lumber, and driving in the yard increased exposures to particulates. The planing machines were enclosed, but these controls were not sucient to maintain exposures at low levels, likely in part because these enclosures were often opened. The models did indicate that personnel enclosures (booths in the production areas and cabs on mobile equipment in the yard) were an e€ective control mechanism, with the potential for reducing exposures by two- to six-fold, proportionate to the percent enclosure provided (including both the amount of time spent in the enclosure and the amount of physical enclosure provided). A limitation of this study is that since only one mill was included, the in¯uence of factors which are more likely to vary between than within mills, such as production level, log storage methods, local exhaust ventilation, and enclosure of saws, could not be observed. Because measurements were taken in only one season, temperatures were relatively stable (mean=16.98C, SD=3.28C), so it is not surprising that this element of weather did not enter the models. The models provide useful information for epidemiological purposes. They indicate clear di€erences in patterns of exposure between the particulate and volatile components of wood, so that investigations of exposure±response relationships using these models should be able to distinguish which are more likely to be the proximate risk factors for any observed morbidity. Even within the group of particulate contaminants, there are di€erences in the models, despite the strong correlations between the measured or estimated concentrations. The ultimate value of the models for the epidemiological investigation depends on the validity of the models in predicting exposures during any time periods at risk. Because of the small sample size, we chose not to set aside a portion of the dataset for testing the model. Another method for assessing the validity is to compare the models for consistency with other studies. We are currently in the process of analyzing several other sawmill datasets, including some with historical data. These will be used to look for similarities and di€erences in exposure determinants. A number of investigators have measured dust and monoterpene exposures in sawmills (see Demers et al., 1998; Teschke et al., 1994), but few have reported about exposure determinants. In a previous study of determinants of particulate levels in two western Canadian coastal sawmills, we examined fewer potential exposure determinants, but results were similar: personnel enclosure reduced

exposures approximately 3-fold; exposures in the planing mill were higher than in the sawmill; wood chips increased exposures; and exposures at the front end of the sawmill (log cut-o€ saw, barker, head saw) tended to be lower than in other areas of the mill (Teschke et al., 1994). The average levels of inhalable particulate measured in the current study were considerably higher than total particulate levels measured in the earlier study (where the geometric mean was 0.12 mg/m3, and the arithmetic mean was 0.51 mg/m3). This is likely due in part to the di€erent measurement technique. In side-by-side sampling of inhalable and total particulate in sawmills, we found that the 7-hole sampler used in this study captured on average 2.4 times more mass than a closedface total dust cassette (Davies et al., 1998). Other factors may have contributed to the higher levels observed in the current study mill, for example dry land storage of the logs rather than water-based storage, and a higher proportion of kiln-dried lumber processed in the planing mill. Few other studies have investigated determinants of dust exposure in sawmills. In a study of Finnish lumber mills, Kauppinen et al. (1984) measured the highest particulate exposures among cleanup personnel, similar to our results, however exposures in other jobs followed patterns much di€erent than in the western Canadian mills. Elevated exposures were observed in manual lumber sorting, packaging, and edging jobs, and low exposures were measured in the planer mill. Halpin et al. (1994) also found lower levels of inhalable particulate in the `dry' mill after kiln-drying than in the `green' mill area of a UK sawmill. It is not possible to suggest reasons for these di€erences, since no information is given on such factors as the pattern of personnel enclosure or cutting equipment used in the study mills. Eriksson (1996) measured monoterpene exposures in four Swedish sawmills and found the highest exposures among those sawing `trees' (assumed to be logs), edging boards, and cleaning up sawdust, whereas those sorting boards had lower exposures. This pattern is very similar to that found in our study, despite the fact that average monoterpene levels were 1 to 2 orders of magnitude lower in the Canadian mill. Di€erences in the overall exposure levels could be caused by di€erences in tree species, log storage conditions, or even factors in the trees' growing environment. Abietic and pimeric acids appear not to have been measured previously in sawmills. It is important to note that resin acids represent only one of many components that make up the complex matrix of wood and that to date have not been considered in industrial hygiene surveys. We were concerned about the high limits of detection for the resin acids in one of the four batches of samples. However, week of sampling (equivalent to batch) did not enter

Exposure to particulate, wood dust, resin acids, and monoterpenes in a lumber mill environment

the resin acid model, indicating that the detection limit did not in¯uence results. Current measurement methods for `wood dust' use ®lter capture with gravimetric analysis and therefore do not distinguish between dust components. Although we tried to estimate the wood component of the particulate, the models reported here suggest that our wood dust estimation procedure was not ideal. The estimated wood dust model predicts higher exposures for mobile equipment drivers than in other yard jobs, whereas the resin acid model predicts lower exposures in the completely enclosed cabs of mobile equipment. This suggests that our method for estimating wood dust exposure, which assumed constant wood dust proportions in a given area of the mill, may not have excluded particulate mass from soil in the mobile equipment jobs. We also considered visual inspection of the ®lters for particulate colour and morphology as a means of identifying wood particulate, but found that this method was very poorly correlated with resin acid levels (Demers et al., 1998). Other methods for distinguishing the wood component of dust in the future might include analysis for total organic carbon, or microscopic evaluation of the ®lter. In summary, these data provide guidance about the locations and jobs in the mill where exposures were highest, indicate tasks, tree species, weather, and wood conditions which contributed to exposure, and suggest e€ective control measures. Since this study included only one lumber mill in one season, it would be speculation to suggest that the results are generalizable, but a few other reports of particulate and monoterpene exposures indicate some consistency between studies. We hope that future investigations of sawmill environments will include detailed documentation of potential exposure determinants so that problems within the industry can be identi®ed and solutions found. AcknowledgementsÐThe authors thank all the mill employees who took part in the study, as well as the company, Canadian Forest Products, and the union, IWA± Canada, for their support throughout the study. J. A. Polonijo of the Workers' Compensation Board Laboratory

255

assisted in the development of the analysis methods for the resin acid samples. Funding for this study was provided in part by research grants from the Department of Labor and Industries of the State of Washington and the Workers' Compensation Board of British Columbia. REFERENCES Davies, H. W., Teschke, K. and Demers, P. A. (1998) A ®eld comparison of inhalable and thoracic size selective sampling techniques. Annals of Occupational Hygiene (Submitted). Demers, P.A., Teschke, K. Kennedy, S. M., Davies, H. and Leung, V. (1998) Exposure to dust, resin acids, and monoterpenes in a softwood lumber mill. American Industrial Hygiene Association Journal (Submitted). Demers, P. A., Teschke, K. and Kennedy, S. M. (1997) What to do about softwood? A review of respiratory e€ects and recommendations regarding exposure limits. American Journal of Industrial Medicine 31, 385±398. Eriksson, K., Levin, J.-O., RheÂn, M. and Lindahl, R. (1994) Evaluation of a di€usive sampler for air sampling of monoterpenes. Analyst 119, 85±88. Eriksson, K. (1996) Occupational exposure to terpenes in saw mills and joinery shops. Dissertation, Department of Environmental Technology and Work Sciences, Royal Institute of Technology, Stockholm, Sweden. Halpin, D. M. G., Graneck, B. J., Lacey, J., Nieuwenhuijsen, M. J., Williamson, P. A. M., Venables, K. M. and Newman Taylor, A. J. (1994) Respiratory symptoms, immunological responses, and aeroallergen concentrations at a sawmill. Occupational and Environmental Medicine 51, 165±172. Hornung, R. W. and Reed, L. D. (1990) Estimation of average concentration in the presence of non-detectable values. Applied Occupational and Environmental Hygiene 5, 46±51. IARC Working Group (1995) Wood dust and formaldehyde. In IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans, Vol. 62. IARC, Lyon, France. Kauppinen, V. T., Lindroos, L. and Makinen, R. (1984) Concentrations of wood dust measured in the workroom air at sawmills and plywood factories. Staub± Reinhalt Luft 44, 322±324. Pengelly, M. I., Foster, R. D., Groves, J. A., Ellwood, P. A., Turnbull, G. B. and Wagg, R. M. (1994) An investigation into the composition of solder fume. Annals of Occupational Hygiene 38, 753±763. Teschke, K., Hertzman, C. and Morrison, B. (1994) Level and distribution of employee exposures to total and respirable wood dust in two Canadian sawmills. American Industrial Hygiene Association Journal 55, 245±250.