PII: S0003-4878(00)00077-6
Ann. occup. Hyg., Vol. 45, No. 3, pp. 209–216, 2001
 2001 British Occupational Hygiene Society
Published by Elsevier Science Ltd. All rights reserved
Printed in Great Britain.
0003–4878/01/$20.00
Cohort Mortality Study of North American
Industrial Sand Workers. III. Estimation of Past
and Present Exposures to Respirable Crystalline
Silica
R. J. RANDO†*, R. SHI‡, J. M. HUGHES‡, H. WEILL§,
A. D. MCDONALD¶ and J. C. MCDONALD¶
†Department of Environmental Health, School of Public Health and Tropical Medicine, Tulane
University Medical Center, 1430 Tulane Avenue, New Orleans, LA 70112 USA; ‡Department of
Biostatistics, School of Public Health and Tropical Medicine, Tulane University Medical Center, 1430
Tulane Avenue, New Orleans, LA 70112 USA; §Department of Medicine, School of Medicine, Tulane
University Medical Center, 1430 Tulane Avenue, New Orleans, LA 70112 USA; ¶Department of
Occupational and Environmental Medicine, National Heart and Lung Institute, Imperial College
School of Medicine, Dovehouse Street, London SW3 6LY, UK
Background: Lung cancer and silicosis mortality were examined longitudinally and by a casereferent analysis in a cohort of workers selected from the North American industrial sand
industry. Date of hire in the case-referent sub-cohort extended as far back as the second
decade of the twentieth century.
Objective: The aim of this study component was to develop estimates of average and cumulative exposure to respirable crystalline silica for the 342 selected cases and referents.
Methods: Process and dust control histories were developed for each plant, and quantitative exposure data obtained from each of them and from a trade organization. An algorithm
was developed to convert historical exposures reported in particle count concentrations to
modern measures of mass concentration of respirable crystalline silica. Personal exposures
were adjusted for use of protective equipment based on frequency of use and type of protection.
Findings: Between 1974 and 1998, a total of 14 249 exposure measurements had been taken
using a cyclone and membrane filter and gave an overall geometric mean of 42 mg/m3. The
only exposure data identified earlier were based on approximately 500 samples collected
across the industry between 1947 and 1955 using the Greenburg–Smith impinger, with analysis by microscopy. These data were converted to modern measures using a factor of 1 mppcf
= 276 mg/m3 respirable dust and then adjusting for percentage silica. In general, the highest
exposures occurred in bagging and bulk-loading operations and the lowest in wet processing
of sand.
Conclusions: There has been a substantial decline in exposure levels in this industry over
time. The decline was rapid between the 1940s and 1970s and current exposures are, on
average, less than 50 mg/m3. The use of personal protective equipment was judged to have
had little impact on exposure before the 1970s.  2001 British Occupational Hygiene Society.
Published by Elsevier Science Ltd. All rights reserved
Keywords: silica; sand; mass-number conversion; exposure estimation
INTRODUCTION
Received 27 July 2000; in final form 28 September 2000.
*Author to whom correspondence should be addressed
The industrial sand industry provides high-purity silica sand and flour for glass manufacture, foundries,
pigment production, and oil drilling. This entails the
extraction of natural deposits of quartz mineral and
209
210
R. J. Rando et al.
processing to remove contaminants and grade the product. Raw sand is extracted in three ways: hard sandstone is quarried by drilling and blasting, friable sandstone is hydraulically mined using high-pressure
water cannon, and loose sand deposits are dredged
from ponds. The first step is typically wet reduction
using rod mills (chaser mills in the past). The material
is then sent to flotation tanks to remove clays and
other contaminants. These steps are collectively
known as “wet processing”, and are followed by a
drying operation. Steam coil dryers, common in the
past, have been replaced by rotary kiln dryers and,
more recently, fluidized bed dryers. Dried sand is segregated into size fractions by mechanical screening or
air separation, and then stored in bins or directly
bagged or loaded into bulk containers for transport
by rail and truck. In some plants, dried sand is fed to
a ball-milling operation to produce silica flour.
The silicosis hazard from industrial sand and other
crystalline silica-containing particulate has long been
recognized. In 1997, the International Agency for
Research on Cancer (IARC) concluded that crystalline silica was also carcinogenic to humans. As part
of a cohort mortality study of North American industrial sand workers (McDonald et al., 2001), a nested
case-referent study of deaths due to lung cancer and
silicosis was conducted (Hughes et al., 2001). This
report describes the reconstruction and estimation of
exposure to respirable crystalline silica for 342 workers (123 cases and 219 controls), spanning the time
period 1912 to 1994.
METHODS
Individual average levels of exposure to respirable
crystalline silica for all workers in the study were estimated by linking job histories to a job-exposure
matrix. A separate job-exposure matrix was
developed for each plant by taking into account existing exposure monitoring databases from the modern
era (post-1970) and from the distant past (1947–
1954). The historical data, reported in units of millions of particles per cubic foot (mppcf) were converted to modern units of respirable crystalline silica
concentration using a calculation algorithm. The
exposure estimates were also adjusted for use of respiratory protective devices since about the mid-1970s.
Details of the exposure reconstruction process follow.
Plant visits
Numerical and qualitative data on worker
exposures were first obtained from all possible
sources, including company documents, consultant
reports, publications and government databases. After
analysis of the data, each participating plant was visited by the industrial hygienist (RJR), who met with
plant health/safety personnel and, usually, a long-term
employee or retiree. A tour of each facility was made
with the aim of relating job titles and work areas to
exposure information, with special attention to temporal trends or changes that might have influenced
exposure. The history and current state of any respiratory protection program was also reviewed. These
inquiries were confined to the period beginning when
any of the cases or controls were first employed,
which ranged from 1912 to 1955.
Available data
Recent exposure measurements had generally used
a 10-mm nylon cyclone in series with a PVC membrane filter. After collection of the respirable dust
fraction, samples were analyzed for crystalline silica
content by quantitative x-ray diffraction spectroscopy
and for total particulate mass gravimetrically. Many
of the participating plants had moved from routine
laboratory analysis of silica content of collected dust
to estimates based on average percentage silica in
dust samples from a given area/department/task. In
each case, the modern exposure measurements were
presented in µg respirable crystalline silica per m3 of
workplace air. The application of the cyclone
sampler/x-ray diffraction technique began in the early
1970s; previously, samples had been collected using
Greenberg–Smith or midget impingers containing
alcohol/water solution, and the dust particles (⬍5 µm)
counted microscopically in units of mppcf.
Conversion of sample results
In order to utilize these data, conversion into modern measures was necessary. In other industries (e.g.
Vermont granite sheds), a conversion factor was
obtained from side-by-side samples collected by
impinger and by cyclone. However, no comparable
data of sufficient reliability could be identified for the
industrial sand industry, so a computational approach
using an existing sample database (Severns, 1979)
was taken. This was based on 14 samples collected
on membrane filters with a 10-mm nylon cyclone preseparator, in various processes in several industrial
sand plants, and analyzed by oil-immersion
microscopy, with particle counts reported in nine size
categories. While limited in scope, this was the only
information of this type that could be identified for
use in development of the conversion algorithm.
The computational algorithm entailed estimation of
the “true” environmental concentration of particles of
a given size by accounting for the sampling efficiency
of the cyclone, eC. The estimated concentration was
then converted to a mass concentration assuming
spherical silica particles of specific gravity 2.65. The
mass contribution of each size fraction was then totaled to derive the mass concentration of respirable
particulate in the sample. The collection/analysis
efficiency of the impinger/microscope, eI, was then
applied to the estimated environmental concentration
to determine the expected:observed concentration by
Cohort mortality study of North American industrial sand workers. III
impinger for the same sample. The number contributions in each size fraction were summed to determine the overall particle concentration expected from
an impinger sample of the same atmosphere. The conversion factor, equating the gravimetric respirable
concentration in units of µg/m3 to the impinger particle count in units of mppcf, was then:
µg/m3
=
mppcf
冢
p
6×0.0283
m3
ft3
冘
冣冘
nid3i
ni
eI,i
eC,i
In this equation, di is the geometric mean aerodynamic diameter and ni is the observed particle count
for the ith particle size interval in the microscopic
analysis of the sample. For this particular set of
samples, the values for particle size and collection
efficiencies used in the computational algorithm are
presented in Table 1.
Personal protection
The use of respiratory protective devices by workers was a potential confounder in the job exposure
matrix. Until recently, there was little evidence of
properly standardized or adequate respiratory protection and, in the past, only quarter-face masks with
resin-impregnated felt filters had been widely available in this industry. There are many reports in the
literature concerning the inadequacy of such respirators, both in terms of the penetration of dusts
through the filter media and of the propensity for
“face-seal” leakage. Furthermore, the use of these
devices in the study plants appeared to be sporadic
and there were few plant regulations requiring their
use and little, if any, enforcement. Because of these
factors, it was decided to ignore the earlier use of
respirators in estimating exposures.
Information regarding dates of well-designed and
enforced respiratory protection programs was
obtained during site visits and through discussions
211
with plant personnel. The National Industrial Sand
Association’s (NISA) Occupational Health Program
for Exposure to Crystalline Silica in the Industrial
Sand Industry (OHP) includes provisions for establishing and maintaining respiratory protection programs, instituted by some plants as early as the
mid-1970s.
For most of these, the exposure database included
information on respirator use during sampling periods. Thus it was usually possible to estimate frequency of use for a given job directly from the
recorded data. Where no information was available,
estimates of usage frequency were obtained from discussions with plant personnel and personal observation. Thus for a given job, the exposure estimate
was adjusted as follows:
C⬘ = C(1⫺f) + C
冉 冊
f
,
WPF
where C⬘ is the exposure concentration corrected for
respirator use, C is the estimated environmental
exposure concentration, f is the frequency of respirator use (0ⱕfⱕ1), and WPF is the workplace protection factor of the respirator. A WPF of 5 was used in
this calculation based on a literature survey of
reported protection factors for particulate hazards
(Harris et al., 1974; Smith et al., 1980; Toney and
Barnhart, 1976).
For those facilities where the exposure database
included information on respirator use, personal
exposures were adjusted for protection and frequency,
and exposure averages (C⬘) were computed from the
adjusted data. Where this information was not directly
available, an estimate was made for each job, based
on local inquiry.
Development and application of the exposure matrix
Job history profiles for cases and controls were collapsed into 244 unique job groupings distributed
across the participating plants. The exposure matrix
for a given job grouping was developed only for those
Table 1. Values used in development of the exposure concentration conversion factor
Microscopy particle
size categories (µm)
⬍0.3
0.31–0.42
0.43–0.6
0.61–0.85
0.86–1.2
1.21–1.70
1.71–2.4
2.41–3.4
3.4–4.8
a
Geometric mean of
range (µm)
Aerodynamic
equivalent diameter, di
(µm)
Cyclone penetration
efficiency,a eC,i
Impinger/microscope
sampling efficiency,b
eI,i
0.21
0.36
0.51
0.72
1.02
1.43
2.03
2.86
4.05
0.34
0.59
0.83
1.17
1.66
2.33
3.30
4.66
6.59
1
1
1
1
1
0.99
0.77
0.24
0.015
0
0
0
0
1
0.85
0.8
0.74
0.72
Taken from Caplan et al. (1977).
Taken from Drinker and Hatch (1936); adjusted for microscopic limit of resolution (苲1 µm).
b
212
R. J. Rando et al.
years where it appeared in job histories. In recent
times, exposures were corrected for respiratory protection. These exposure estimates were taken as the
arithmetic mean exposure calculated from the post1970 exposure monitoring database for an appropriate
time segment in which the exposures were judged to
be stable. This was determined by examination of the
timing of changes in process, work practices, and
controls and correlating their impact on exposure as
indicated from temporal trends in the exposure monitoring data. In some cases, no changes or trends were
identified and the mean of the entire database for that
job was used as the exposure estimate.
Exposures prior to the 1970s were handled in a
similar way, except that the converted historical data
were integrated into the estimates. The history of a
given job and facility served to identify probable
dates for any likely changes in exposure level. After
that time, the oldest exposure estimates from the current exposure database were taken as representative;
that is, exposure estimates derived from the data of
the 1970s and beyond were extrapolated back to the
point of the change. For periods before any recorded
change, plant- and job-specific historical dust
exposure data were used, after first being transformed
into modern exposure units using the conversion
algorithm. This specific approach was possible in
three plants. For the others, plant-specific historical
data were not available and geometric mean
exposures, averaged across all plants in an industrywide historical database and broken down by various
production areas, were calculated and assumed to be
representative of exposure. Both the plant-specific
and industry-wide estimates were then adjusted individually for the percentage silica prevalent in the
dusts from particular production areas.
Likewise for those time periods preceding the historical database (pre-1947), the exposure estimates
derived from the historical data were assumed to be
representative. For reported jobs in which no specific
sampling was done and no specific information on the
job’s tasks was available (for example, “labour” and
“utility”), exposure was assumed to be equal to the
overall mean of the exposure estimates for all production jobs in the facility.
For each subject in the case-control study, average
exposures to crystalline silica were then calculated by
integrating the unique job histories with the jobexposure matrix, enabling cumulative exposures to be
calculated using duration of employment/exposure.
Thus,
D=
冘
Ci⬘×ti
and
C̄⬘ =
D
冘
ti
where D is the cumulative exposure estimate, and
C̄⬘ is the estimated average exposure, with Ci⬘ the
exposure estimate over time period ti.
RESULTS
Historical descriptions of plants and processes from
the industrial sand industry were identified in several
trade reports (for example Rock Products, 1927; Sawyer, 1947; Mocine, 1948; Ayers and Crew, undated;
further information about these can be obtained from
the first author of the current paper). In the distant
past, certain common process changes dramatically
reduced exposure levels. In the modern era, controls
generally resulted in lesser reductions in exposure.
Table 2 is a listing of the more important and commonly-encountered process changes, broken down by
general process area, especially important in bagging
and bulk loading.
For the years before 1970, the only quantitative
exposure data discovered were those presented in four
unpublished reports on studies undertaken by Professor Theodore Hatch on behalf of the Industrial
Health Foundation of America, 1947 to 1955. These
reports, details of which can be obtained from the first
author, contained the results of approximately 400
samples from at least 19 sand-processing plants, three
of which were in the present study. The other plants
in the Hatch reports could not be identified, although
probably some also are in the current study. The
samples were collected with Greenburg–Smith
impingers and were quantitated by microscopy. The
industrial hygienist followed the workers around and
placed the impinger samplers “within a few inches of
the worker’s nose”; in areas where no specific dust
source could be identified, such as in break rooms,
area samples were collected.
Table 3 presents a summary of the data reported
by Hatch in his initial survey of 19 facilities in 1947.
As a limit of 5 mppcf was proposed at that time,
Table 3 suggests that there were excessive exposures
throughout the industry. Most notable were drilling
operations, screening, milling, and bagging and bulk
loading in box cars of both sand and silica flour. It
is also evident that there were large differences in
dust levels for similar processes between plants. It is
uncertain whether these were true differences or simply a reflection of the wide variability within plants
and the relatively small number of samples examined.
After about 1970, exposure databases were available for all plants in the study. For the most part,
these represented reliable data collected by modern,
state-of-the-art monitoring techniques. Most were the
result of personal sampling in the breathing zone of
workers, with the use of respiratory protection frequently recorded. Table 4 presents a summary of the
exposure data obtained from the companies since
1973, showing that plant-wide geometric mean
exposures to respirable crystalline silica ranged from
20 to 107 µg/m3 during this period. There were very
Cohort mortality study of North American industrial sand workers. III
213
Table 2. Common process changes and controls that reduced historical exposure to crystalline silica in the American
industrial sand industry
Industrial sand process
area/department
Significant process changes
Quarrying
Change from dry-drilling to wet-drilling
Replacement of secondary blasting/drilling with drop ball
Air-conditioning and filtration of heavy equipment cabs
Enclosure of primary crusher and construction of crusher operator control booth
Replacement of chaser mills with rod mills (as exposure was low and the material wet,
this common process change had little effect)
Replacement of steam-coil with rotary kiln and fluid-bed dryers
Enclosure and ventilation of screens
Improved screens that require less brushing
Installation of air filters and local exhaust hoods in operator booths
Replacement of burlap bags with paper bags
Replacement of bag-sewing with glueing
Improvements in bagging machine design and ventilation
Use of bagging station enclosures and air curtains
Surface coating of filled bags with oil or water
Use of automated palletizers and shrink wrappers
Replacement of box cars with hopper cars and trucks
Automation of loading process
Local exhaust ventilation of loader chutes and mechanisms
Paving and wetting of access roads and grounds
Improved dust collection on local exhaust systems
Upgrading of respiratory protection programs
Crushing
Wet processing
Drying
Screening
Milling
Bagging
Bulk loading
General
Table 3. Historical exposure data for 19 industrial sand plants (taken from Hatch, 1947)
Process area and operation
Quarry
Crushing
Wet processing
Drying
Screening
Milling
Bagging
Bulk loading
Drilling
Shovel
Primary
Secondary
Leaching and dewatering
Moist sand handling and storage
Sand washing
Rotary kiln
Steam coil
Drier screens
Grading screens
Personnel rooms
Grinding mill
Storage bins and upper levels
Personnel rooms
Bagging machines
Box cars (sand)
Hopper cars (sand)
Flat cars (sand)
Box cars (silica flour)
Mean of means—
plants [mppcf]
Range of means
[mppcf]
Number of
samples/number of
means—plants
7.8
1.0
4.6
2.8
1.0
1.3
1.5
3.5
4.6
84 (GM=39)
42 (GM=7.9)
1.6
7.8
19.7
1.4
20
67 (GM=20)
3.3
11
91 (GM=12)
0.5–16
0.3–1.7
1.5–15
1.4–4.8
0.4–1.8
0.3–3.3
0.3–4.7
0.7–12
0.4–14
2–>170
1.0–220
⬍2–3.8
0.6–40
1.0–77
0.6–2.6
5.0–60
1.8–350
1.5–5.0
0.3–5.0
0.8–180
9/4
5/3
18/8
9/4
11/4
17/9
28/10
24/10
33/9
31/12
38/6
5/–
25/9
23/9
12/5
25/8
22/9
3/2
3/2
5/2
GM: geometric mean of means.
wide variations across job groupings within a given
plant, with the highest levels generally seen in bagging and bulk-loading operations and the lowest
among administrative personnel.
Historical to modern dust conversion factors, computed according to the algorithm, are presented in
Table 5 for the various processes studied by Severns
(1979), with values ranging from 206 in an air sizing
operation to 364 around vibrators. A Kruskall–Wallis
ANOVA of the conversion factors across different
areas showed no important difference (P = 0.61). An
overall mean conversion factor of 276 µg/m3 per
mppcf was therefore applied to all historical dust
samples as follows:
214
R. J. Rando et al.
Table 4. Exposure data obtained from the plants since 1973
Location/plant
µg/m3—Respirable crystalline silica
Years
Number of samples
Illinois/1
Illinois/2
New Jersey/1
New Jersey/2aa
New Jersey/2ba
Ohio/1
Ohio/2
West Virginia and Pennsylvania/1
West Virginia and Pennsylvania/2
Quebec 1
1974–1996
1976–1994
1974–1996
1975–1996
1979–1991
1975–1997
1974–1997
1973–1996
1973–1996
1978–1998
1493
608
1420
1064
199
263
336
3478
5143
2565/444b
Total
1973–1998
14 249
Geometric mean
(geometric standard
deviation)
24
107
20
59
68
56
57
42
45
86
(9.2)
(7.5)
(9.2)
(6.2)
(3.0)
(6.2)
(6.8)
(7.1)
(5.1)
(4.0)
42 (6.5)
a
New Jersey Plant 2 is in two neighboring locations, requiring separate assessment.
Dust samples/silica samples.
b
Table 5. Historical–modern dust concentration conversion factors derived from the computational algorithm
Average count median diametera
(µm)
Plant/process area
Conversion factorb (µg/m3 per mppcf)
n
mean
s.d.
Drying
Screening
Air sizing
Vibrators
Bin room
Bulk loading
Bagging
0.46
0.45
0.45
0.38
0.56
0.36
0.31
4
1
2
2
1
1
3
292
252
206
364
276
344
297
128
–
3
5
–
–
103
Overall
0.42
14
276
59
a
Geometric mean particle size of respirable dust fraction collected with cyclone pre-separator.
Kruskall–Wallis ANOVA, P=0.61.
b
µg/m3 (respirable crystalline silica) =
[276]×[mppcf]×[% silica]
DISCUSSION
The percentage silica in the latter equation was
derived from the average sample assay reported in the
modern database for a given plant/area/job. In fact,
the silica assays reported by Hatch were often taken
from settled dust and appear extraordinarily high and
are possibly unreliable. On the other hand, silica
assays in the modern database are likely to be negatively biased, since high loadings derive primarily
from particles released from the sand as it is being
processed.
Nevertheless, the conversion factor of 276 used in
this paper is high compared to others reported previously. For example, Ayer et al. (1973) used a conversion factor of 110 for dusts from the Vermont
granite sheds, and the ACGIH (1981) recommended
a conversion factor of 167 for dusts in which the density and mass median diameter are unknown; Tomb
and Haney (1988) reported a range of 46 to 313 for
various mineral dusts; and Seixas et al. (1997) suggested factors ranging from 90 to 180 for the diatomaceous earth industry. Although a conversion factor
of 276 may be positively biased to some (unknown)
extent, the use of this conversion factor, which we
developed specifically for the industrial sand industry,
was deemed preferable in our judgement to using data
derived from other unrelated industries. In any case,
any bias in conversion should be internally consistent
and apply equally to cases and referents in the epidemiological analysis (Hughes et al., 2001).
For each of the 342 workers in the case-referent
study, annual exposure estimates were calculated by
coupling their job history for a given year with the
job-exposure matrix. Annual average exposure levels
Cohort mortality study of North American industrial sand workers. III
215
Fig. 1. Mean exposure levels and numbers of workers in the case-referent study present for each year in the study period.
ranged from approximately 400 to 500 µg/m3 in the
1930s and 1940s, falling to less than 50 µg/m3 in the
late 1980s (see Fig. 1). It must be remembered that
the data are representative only of job titles actually
recorded in the work histories of cases and referents
and, as such, are not necessarily representative of
industry-wide exposures for any given year. However, both the pattern of decline and estimated
exposure levels are quite similar to those reported
recently by Burgess (1998), in a parallel epidemiological study of silicosis and lung cancer in the British
potteries, with results comparable to our own.
Acknowledgements—This work was supported by an award
from the National Industrial Sand Association and the National
Mining Association. We are especially grateful to the
employees and management of the participating facilities who
provided invaluable assistance to the authors throughout this
study. The technical assistance of Dr Halet Poovey was
much appreciated.
REFERENCES
ACGIH. Threshold limit values for chemical substances and
physical agents and biological exposure indices. Akron
(OH): American Conference of Governmental Industrial
Hygienists, 1981.
Ayer HE, Dement JM, Busch KA, Ashe HB, Levadie BTH,
Burgess WA et al. A monumental study—reconstruction of
a 1920 granite shed. Am Ind Hyg Assoc J 1973;34:206–11.
Burgess GL. Development of an exposure matrix for respirable
crystalline silica in the British pottery industry. Ann Occup
Hyg 1998;42:209–17.
Caplan KJ, Doemeny LJ, Sorenson SD. Performance characteristics of the 10-mm respirable mass sampler: Part I—Monodisperse studies. Am Ind Hyg Assoc J 1977;38:83–95.
Drinker P., Hatch TF Industrial dust: hygienic significance,
measurement and control. New York: McGraw-Hill, 1936.
Harris HE, DeSieghardt WC, Burgess WA, Reist PC. Respirator usage and effectiveness in bituminous coal mining operations. Am Ind Hyg Assoc J 1974;35:159.
Hughes JM, Weill H, Rando RJ, Shi R, McDonald AD,
McDonald JC. Cohort mortality study of North American
industrial sand workers. II. Case-referent analysis of lung
cancer and silicosis deaths. Ann Occup Hyg
2001;45(3):201–7.
McDonald AD, McDonald JC, Rando RJ, Hughes JM, Weill
H. Cohort mortality study of North American industrial sand
workers. I. Mortality from lung cancer, silicosis and other
causes. Ann Occup Hyg 2001;45(3):193–9.
Seixas NS, Heyer NJ, Welp EAE, Checkoway H. Quantification of historical dust exposures in the diatomaceous earth
industry. Ann Occup Hyg 1997;41:591–604.
Severns NA. A profile of respirable free crystalline silica in
216
R. J. Rando et al.
the sand industry. Master’s Thesis, University of Pittsburgh,
Pittsburgh (PA), 1979.
Smith TJ, Ferrell WC, Varner MO, Putnam RD. Inhalation of
cadmium workers: effects of respirator usage. Am Ind Hyg
Assoc J 1980;41:624.
Tomb T, Haney R. Comparison of number and respirable mass
concentration determination. In: Advances in air sampling.
Chelsea (MI): Lewis, 1988. [American Conference of
Governmental Industrial Hygienists].
Toney CR, Barnhart WL Performance evaluation of respiratory
protective equipment used in paint spray operations. Cincinnati: HEW, 1976 [Publication no. (NIOSH) 76-177].