Trends in Wood Dust Inhalation Exposure in the UK, 1985-2005.

Ann. Occup. Hyg., Vol. 53, No. 7, pp. 657–667, 2009
� The Author 2009. Published by Oxford University Press
on behalf of the British Occupational Hygiene Society
doi:10.1093/annhyg/mep044
Trends in Wood Dust Inhalation Exposure in the UK,
1985–2005
KAREN S. GALEA1*, MARTIE VAN TONGEREN1,2,
ANNE J. SLEEUWENHOEK1, DAVID WHILE2,3, MAIRI GRAHAM1,
ANNETTE BOLTON2, HANS KROMHOUT4 and JOHN W. CHERRIE1
1Institute of Occupational Medicine, Research Avenue North, Riccarton, Edinburgh, UK; 2Centre of
Occupational and Environmental Health, University of Manchester, Manchester, UK; 3Centre for
Suicide Prevention, University of Manchester, Manchester, UK; 4Institute for Risk Assessment Sciences,
University of Utrecht, Utrecht, The Netherlands
Received 14 November 2008; in final form 21 April 2009; published online 14 July 2009
Objectives: Wood dust data held in the Health and Safety Executive (HSE) National Exposure
DataBase (NEDB) were reviewed to investigate the long-term changes in inhalation exposure
from 1985 to 2005. In addition, follow-up sampling measurements were obtained from selected
companies where exposure measurements had been collected prior to 1994, thereby providing
a follow-up period of at least 10 years, to determine whether changes in exposure levels had
occurred, with key staff being interviewed to identify factors that might be responsible for
any changes observed.
Methods: Analysis of the temporal trend in exposure concentrations was performed using
Linear Mixed Effect Models on the log-transformed NEDB data set and expressed as the rel-
ative annual change in concentration.
Results: For the NEDB wood dust data, an annual decline of geometric mean (GM) exposure
of 8.1% per year was found based on 1459 exposure measurements collected between 1985 and
2003. This trend was predominantly observed in data from inspection visits (measurements col-
lected on a mandatory basis by a Specialist HSE Inspector) (n 5 1009), while data from repre-
sentative surveys (measurements collected on a voluntary basis to provide information on
current practices and exposures) remained relatively stable. Ten follow-up surveys in individ-
ual workplaces in 2004–2005 resulted in 70 new measurements and for each of the companies
resurveyed, the GM of the wood dust exposure decreased between sampling surveys.
Conclusion: Analysis of the temporal trend in UK wood dust exposure concentrations re-
vealed declines of 8% per annum. Interviews with key long-serving employees and manage-
ment suggest that factors such as technological changes in production processes, response to
new legislation, and enforcement agency inspections, together with global economic trends,
could be linked to the downward trends observed.
Keywords: exposure; inhalation; time; trends; wood dust
INTRODUCTION
Wood dust is a general term covering a wide variety
of airborne wood dusts, which are produced during
the processing and handling of both hard and soft-
wood, chipboard, hardboard, and other composite
materials (HSE, 2003). Exposure to wood dust may
cause respiratory diseases and the International
Agency for Research on Cancer (IARC, 1995) has
classified wood dust as carcinogenic to humans based
on epidemiological evidence. In 2000–2003, it was
estimated that �384 000 UK workers were occupa-
tionally exposed to inhalable wood dust (Kauppinen
et al., 2006). It is estimated that �81 000 UK work-
ers and over 500 000 workers in the European Union
(EU) wood-processing industries may be exposed
to dust levels (any type of wood dust) exceeding
5 mg m�3 (Kauppinen et al., 2006).
An occupational exposure limit (OEL) of 5 mg m�3
(8-h time-weighted average) for inhalable hardwood
dust came into effect in the UK in April 1988; this
*Author to whom correspondence should be addressed.
Tel: þ44(0)870-850-5131; fax: þ44(0)870-850-5132;
e-mail: karen.galea@iom-world.org
The free full text version of this article can be found in the
online version of this issue.
657
by guest on M
ay 10, 2011
a
n
nhyg.oxfordjournals.org
D
ow
nloaded from

being replaced by a maximum exposure limit (MEL)
of the same value when the control of substances haz-
ardous to health (COSHH) Regulations came into
force in October 1989 (HMSO, 1988). A MEL of
the same value was also introduced in January 1997
for softwood. In 2005, the system of exposure limits
was changed in the UK, and currently there is a work-
place exposure limit of 5 mg m�3 for all wood dust
(HSE, 2005). It is therefore apparent that since
1988, the 8-h OEL, irrespective of type of exposure
limit in place, has remained a constant 5 mg m�3.
The Scientific Committee for Occupational Exposure
Limits of the EU has stated that exposure to wood dust
.0.5 mg m�3 induces pulmonary effects and there-
fore should be avoided (SCOEL, 2002).
We investigate the temporal changes in wood dust
exposure in the UK as part of a larger project exam-
ining long-term changes in inhalation exposure to
various hazardous substances in a number of indus-
tries within the UK (Creely et al., 2006).
MATERIALS AND METHODS
National Exposure DataBase wood dust data
set and data preparation
Wood dust data held in the National Exposure Data-
Base (NEDB), which was set up by the UK Health
and Safety Executive (HSE) in 1986 (Burns and
Beaumont, 1989), was made available for the analysis.
Several core data criteria were identified (Ritchie and
Cherrie, 2001; Tielemans et al., 2002) and applied to
select data to be included in the analysis, these being
economic activity (industry), sampling date, sampling
device, sample duration, analytical methods, mea-
sured concentration, and units used. In total, 1651
wood dust measurements from 168 locations or plants,
obtained between 1985 and 2003, were extracted from
the NEDB for the analysis. Occupation and industry
were coded using Standard Occupational Classifica-
tion 2000 (ONS, 2000a,b) and Standard Industrial
Classification 2003 (ONS, 2003), but these were
subsequently amalgamated into five industrial and
occupational categories (excluding unknown) for
analysis, as detailed in Table 1.
Data were then excluded from the analysis using
the following criteria, with steps (i), (ii), and (iii) be-
ing repeated as necessary:
(i) Data available for ,5 separate years for an in-
dustry or occupation;
(ii) Less than 50 measurements available across all
years for an industry/occupation;
(iii) Less than 50 measurements available for an indi-
vidual category of a confounding variable (wood
type—soft, hard, or mixed; visit type—inspection
or non-inspection; season—spring, summer, au-
tumn, or winter).
There were two types of survey: inspection surveys
are those undertaken by a Specialist Inspector as part
of the HSE inspection process; representative sur-
veys are traditionally undertaken to provide informa-
tion on the working practices and exposure levels
within an industry. Exposure measurements collected
during these representative surveys were collected
from companies on a voluntary basis P. Griffin (per-
sonal communication).
Table 1. Industry and occupation categories used in the wood dust analyses
# Industry description SIC codes n Yearsa
1 Manufacture of furniture 36.1; 36.12; 36.13; 36.14 849 16
2 Manufacture of wood products
(non-furniture)
20.4; 20.5; 36.3; 36.509; 297 10
3 Manufacture of non-wood products 15.209; 24.61; 26.15; 27.51; 28.73; 34.203; 34.3; 35.3;
36.15; 36.2; 36.5; 80.21
71 8
4 Timber industry/mills etc. 20.1 56 4
5 Manufacture of builder/construction
wood products
20.3; 45.45 359 14
6 Unknown 19 2
# Occupation SOC codes n Yearsa
1 Carpenter 5315; 5316; 5492; 5499; 8139; 9121 230 13
2 Cleaner 9233 22 9
3 Miscellaneous 1121; 3422; 5241; 5323; 5493; 6124; 8111; 8121; 8125;
8129; 8133; 8139; 8222; 9134; 9149
41 11
4 Sander/polisher 5323; 8121; 8139 349 15
5 Wood machinist 5315; 5412; 8121; 8125; 8139; 9121; 9149 1007 19
6 Unknown 2 1
SIC, Standard Industrial Classification; SOC, Standard Occupational Classification. Manufacture of non-wood products stands for
the manufacture of those products where the main raw material is not wood, but the product has some parts or elements which are
made of wood. n, number of measurements.
aTotal number of years for which data were available.
658 K. S. Galea et al.
by guest on M
ay 10, 2011
a
n
nhyg.oxfordjournals.org
D
ow
nloaded from

Repeat wood dust exposure monitoring surveys
The Institute of Occupational Medicine’s (IOM)
occupational hygiene consultancy login system and
the NEDB were used to identify companies suitable
for follow-up where wood dust surveys were carried
out prior to 1994 and yielded more than three expo-
sure measurements that fulfilled the core data crite-
ria, as well as further contextual information on the
tasks/processes monitored and the control measures
in place. Companies were contacted to establish their
willingness to participate in the study.
During the monitoring surveys, similar exposure
measurements to those collected during the previous
surveys were obtained. Wherever possible, the same
jobs or tasks as detailed in the original occupational
hygiene report were monitored and the same sam-
pling and analytical techniques as those used previ-
ously were applied. The majority of the inhalable
wood dust samples were obtained using a seven-hole
sampling head with the exception of one plant, W6,
where IOM sampling heads were used in 2004. The
measurements for this plant were reduced by
20% (Boffetta et al., 2003) to be comparable with
a seven-hole sampling head.
Key long-serving employees and management
were also interviewed using a semi-structured ques-
tionnaire to determine whether changes in processes,
technology, and control measures had occurred be-
tween the monitoring surveys and if so, the reasons
for these changes (Creely et al., 2006). The interview
included a combination of both closed and follow-up
open questions to allow the interviewee to respond as
fully as possible.
Statistical analysis
The exposure data were transformed using natural
logs prior to the analysis. Analyses were carried out
using Linear Mixed Effect Models assuming a hierar-
chical structure of the data using the Proc Mixed
procedure in SAS (version 8.02), with the log-
transformed wood dust exposure results as the depen-
dent variable. Plant was included in these models as
a random effect.
We considered three types of fixed effects in the
model. First, as the variable of main interest in tem-
poral trend analyses, year was included as a contin-
uous fixed effect. Next, we considered variables that
were not causally linked with any temporal change
in exposure. For example, season during which the
measurement was taken cannot be related to any re-
al change in exposure. However, if there was a shift
in sampling over time, for example, from predomi-
nantly measuring during the winter in the earlier
surveys to sampling in the summer during the later
survey, then this could lead to a perceived change in
exposure if the trend analyses were not corrected for
this variable. In this category of potentially con-
founding variables, the following variables were
available for analyses: industry, occupation, wood
type, survey type, and season. Finally, we consid-
ered fixed effects that could be causally related to
changes in exposure over time (exposure modifiers).
For example, the introduction of a local exhaust
ventilation (LEV) system is likely to result in a
change in exposure. However, including these vari-
ables in the model would potentially hide any real
change in exposure.
The Linear Mixed Effect Model was developed as
follows:
Step 1: Inclusion of ‘plant’ as random and ‘year’ as
fixed effect.
Step 2: Confounding variables were subsequently
considered for inclusion in the model using a system-
atic forward stepwise procedure. Variables were con-
sidered for inclusion in the model based on (i) impact
on the coefficient for year (.20%) and (ii) statisti-
cally significant improvement in the fit of the model.
Modelling of the main effects ended when there was
no further improvement in fit from systematically
testing the main effects.
Step 3: Inclusion of any exposure modifiers, de-
pending on any improvement of fit of the model.
The following model was used to describe the data
sets:
ln
�
EðikÞ
�
5 b0 þ b1 � Year þ b2 � Confounders
þ b3 � Modifiers þ vðiÞ þ ekðiÞ
for i 5 1, 2, . . . , ni plants and nk measurements
(nested within plant). v(i) represents the random ef-
fect of the i-th plant and ek(i) represents the random
effect of day (k) within plant (i). It was assumed that
v(i) and ek(i) are normally distributed with zero
means, and that these random effects are statistically
independent. It was further assumed that any two
measurements within the same plant have equal cor-
relation irrespective of the time interval between the
measurements, while measurements carried out in
different plants are uncorrelated (compound symme-
try covariance structure) (Peretz et al., 2002).
In addition, we also carried out stratified analyses
of the temporal trend by industry, occupational
group, and type of survey.
The temporal trends were expressed as the annual
change in geometric mean (GM) exposure using the
following expression:
% change per year5 100 � ðexp½b1� � 1Þ
RESULTS
NEDB wood dust data set
Descriptive statistics. After applying the data ex-
clusion criteria, results from 1459 measurements
Trends in wood dust inhalation exposure in the UK, 1985–2005 659
by guest on M
ay 10, 2011
a
n
nhyg.oxfordjournals.org
D
ow
nloaded from

from 153 plants were available for analyses (192
measurements were excluded). No details on the ac-
tual sampling duration were available, although all
the measurements were classified as long-term per-
sonal samples. The results ranged from ,0.01 (one
measurement) to 501.6 mg m�3 with a mean (arith-
metic mean) of 12.4 mg m�3, a GM of 5.2 mg m�3,
and a geometric standard deviation of 3.6. Table 2
describes the wood dust exposure by year. The years
with the most data were 1988 (n 5 321) and 2000
(n 5 345) whereas there was only one measurement
available for 1998. The highest average wood dust
level (in terms of GM) was observed in 1986
(GM5 15.2 mg m�3). There were no personal identi-
fiers in the database and so we were not able to esti-
mate within- and between-worker variances.
Table 3 shows the results of the wood dust meas-
urements by the various confounders and determi-
nants of exposure considered in the model. These
were industry sector, occupation, type of wood dust,
type of survey, season, and the potential determinant
of exposure being the presence of LEV.
There appears to be very little difference in the
mean exposure between the industries, with GMs
ranging from 5.0 mg m�3 in ‘manufacture of furni-
ture’ (Industry 1) to 7.7 mg m�3 in ‘manufacture of
non-wood products’ (Industry 3). Following exclu-
sions, only data from three occupations were avail-
able for analyses. Exposure levels were similar for
these three occupations, with the GM ranging from
4.8 mg m�3 for ‘wood machinists’ to 6.1 mg m�3
for ‘sanders/polishers’. The majority of the measure-
ments were for hardwood dust, with 36% mixed
wood dust measurements and only 6% for softwood
dust. The exposure levels appear to be lower in cases
where workers were exposed to a mixture of hard and
softwood (GM 5 3.6 mg m�3) compared to 7.8 mg
m�3 for exposure to softwood and 6.3 mg m�3 for
exposure to hardwood. The majority of the data
(n 5 1009; 69%) were collected during inspection
visits with 450 data points being collected during
representative surveys. The exposure levels found
during inspection visits and representative measure-
ment surveys were very similar (GM 5 5.5 mg m�3
and 4.5 mg m�3, respectively).
Most of the representative surveys were carried out
during the late 1980s, although a few measurements
were collected in a representative survey in 1996.
Data from inspection visits were available through-
out the period between 1985 and 2003. Exposure to
wood dust appeared to be higher during the autumn
(GM 5 6.8 mg m�3) and lowest in the winter (GM 5
4.0 mg m�3). For the majority of measurements, some
form of LEV was present, although the presence of
LEVappeared to have little impact on thewood dust ex-
posure levels in this data set, with the GMs being sim-
ilar whether LEV was present or not.
Linear Mixed Effect Models. After entering year in
the Linear Mixed Effect Model, only inclusion of the
variable related to the type of survey significantly im-
proved the model. Finally, the inclusion of the vari-
able indicating the presence of LEV was tested to
Table 2. Personal wood dust exposure (mg m�3) by year
Year Nsamples Nplants AM GM GSDtot GSDbp GSDwp Min Max
1985 24 4 12.0 6.4 2.8 1.5 2.6 1.1 110.0
1986 125 9 27.1 15.2 2.9 2.0 2.3 1.1 305.5
1987 120 11 12.5 8.0 2.5 1.8 2.0 1.2 83.0
1988 321 26 10.4 4.0 3.1 2.3 2.2 0.4 501.6
1989 125 17 11.3 5.3 3.4 2.4 2.3 0.3 120.0
1990 17 2 14.8 12.1 2.0 1.0 2.0 4.5 35.8
1991 104 14 15.5 10.0 2.6 1.9 2.1 1.3 77.0
1992 81 6 8.1 5.2 2.6 2.0 1.9 0.2 40.0
1993 46 4 11.7 6.5 3.1 2.0 2.4 0.9 65.7
1994 37 5 5.4 4.5 1.9 1.4 1.7 0.5 17.7
1995 20 3 10.6 5.0 3.4 1.9 2.8 0.7 75.7
1996 21 4 6.9 4.6 2.6 1.9 2.0 0.7 23.2
1997 16 2 18.5 11.6 2.7 1.0 2.7 2.1 95.0
1998 1 1 0.1 0.1 1.0 1.0 1.0 0.1 0.1
1999 13 4 4.0 2.5 2.9 2.2 2.0 0.5 12.0
2000 345 40 7.9 2.7 4.1 2.8 2.6 ,0.01 305.3
2001 27 4 48.6 11.5 5.5 4.8 1.9 1.2 449.0
2002 5 1 5.7 5.4 1.3 1.0 1.3 4.4 9.2
2003 11 2 8.5 5.9 2.7 1.0 2.7 2.2 21.5
Nsamples, number of measurements; Nplants, number of plants; AM, arithmetic mean; GSDtot, total geometric standard deviation;
GSDbp, between-plant geometric standard deviation; GSDwp, within-plant geometric standard deviation; Min, minimum
exposure; Max, maximum exposure.
660 K. S. Galea et al.
by guest on M
ay 10, 2011
a
n
nhyg.oxfordjournals.org
D
ow
nloaded from

determine if this improved the fit of the model; how-
ever, this was not the case. Table 4 provides informa-
tion on the modelling steps. The between-plant
variance decreased by 21% after the addition of year
and type of survey to the model; however, the within-
plant variance showed little change, due to the lim-
ited amount of longitudinal data within plant (only
6 of the 160 plants had data for .1 year). After in-
cluding the type of survey in the model, the effect
of year on the log-transformed wood dust exposure
was �0.085, which is equivalent to an annual de-
crease in the GM exposure of 8.1%.
Stratified analyses were also carried out by indus-
try, occupation, and visit type (Table 5). When ana-
lysed by industry, statistically significant trends in
exposure were observed for manufacture of furniture
(�8.9% per year), ‘manufacture of wood products
(non-furniture)’ (�8.8% per year), and ‘manufacture
Table 3. Personal wood dust exposure (mg m�3) by confounders and determinant of exposure
n AM GM GSDtot GSDbp GSDwp Min Max
Industry
1. ‘Manufacture of furniture’ 825 11.6 5.0 3.4 2.5 2.2 0.2 391.0
2. ‘Manufacturer of
wood products’ (non-furniture)
269 14.4 4.9 4.9 3.4 2.7 ,0.01 449.0
3. ‘Manufacturer of non-wood
products’
54 14.4 7.7 3.0 1.7 2.6 0.5 104.0
4. ‘Manufacturer of builder/construction
wood products’
311 12.3 5.7 3.2 2.4 2.2 0.1 501.6
Occupation
Carpenters 212 10.6 5.7 2.9 2.2 2.0 0.2 115.0
Sanders/polishers 335 12.7 6.1 3.3 2.3 2.3 0.1 258.0
Wood machinists 912 12.7 4.8 3.8 2.7 2.4 ,0.01 501.6
Wood dust
Mixed 524 8.7 3.6 3.8 2.7 2.5 ,0.01 305.3
Soft 82 13.3 7.8 3.0 2.3 2.0 0.2 95.0
Hard 853 14.5 6.3 3.3 2.5 2.1 0.2 501.6
Type of survey
Representative 450 10.9 4.5 3.2 2.3 2.3 0.3 501.6
Inspection 1009 13.0 5.5 3.7 2.8 2.3 ,0.01 449.0
Season
Winter 478 9.4 4.0 3.8 2.7 2.4 ,0.01 305.5
Spring 368 13.0 5.2 3.5 2.4 2.5 0.1 449.0
Summer 249 13.4 5.8 3.5 2.8 2.1 0.4 501.6
Autumn 364 14.9 6.8 3.3 2.2 2.4 0.2 391.0
LEV present
Yes 779 13.0 5.3 3.5 2.5 2.4 0.1 501.6
No 552 10.7 4.9 3.7 2.8 2.3 ,0.01 137.0
Unknown 128 15.9 6.3 3.5 2.4 2.4 0.3 449.0
n, number of measurements; AM, arithmetic mean; GSDtot, total geometric standard deviation; GSDbp, between-plant geometric
standard deviation; GSDwp, within-plant geometric standard deviation; Min, minimum exposure; Max, maximum exposure.
Table 4. Information on the mixed effects model during the various stages of model development
Modela �2RLL AIC Diff df P-value byear SE Annual trend (%) Between plant Within plant
0 4305.7 4309.7 — — — — — — 0.6185 0.9255
1 4289.5 4293.5 16.2 1 P , 0.001 �0.058 0.012 �5.7 0.5323 0.9226
2 4274.5 4278.5 15.0 1 P , 0.001 �0.085 0.013 �8.1 0.4887 0.9185
�2RLL, �2 Residual Log Likelihood estimate; AIC, Akaike information criterium (smaller is better); Diff, difference in �2RLL
associated with the inclusion of the additional variable (i.e. difference with the model in the previous row); Df, degrees of freedom
associated with inclusion of the additional model; P-value, P-value of log-likelihood ratio test to determine improvement in fit;
byear, parameter estimate for year; SE, standard error for b10; Between-plant, between-plant variance in exposure; Within-plant,
within-plant variance in exposure.
a0, no fixed effects included in the model; 1, year; 2, year and type of survey.
Trends in wood dust inhalation exposure in the UK, 1985–2005 661
by guest on M
ay 10, 2011
a
n
nhyg.oxfordjournals.org
D
ow
nloaded from