Commentary: variability in workplace exposures and the design of efficient measurement and control strategies.

95
Ann. occup. Hyg., Vol. 47, No. 2, pp. 95–99, 2003
© 2003 British Occupational Hygiene Society
Published by Oxford University Press
DOI: 10.1093/annhyg/meg021
Commentary
To celebrate the British Occupational Hygiene Society’s fiftieth anniversary this year, we are reproducing in
our on-line edition ‘classic papers’ from past issues of the Annals, with accompanying commentaries in the
print and on-line edition. For this issue, the classic paper we reproduce is Kromhout H, Symanski E,
Rappaport SM. (1993) A comprehensive evaluation of within- and between-worker components of occupa-
tional exposure to chemical agents. Ann Occup Hyg; 37: 253–70.
Variability in Workplace Exposures and the Design of
Efficient Measurement and Control Strategies
ALEX BURDORF1* and MARTIE VAN TONGEREN2
*Author to whom correspondence should be addressed.
Department of Public Health, Erasmus MC, University
Medical Center Rotterdam, PO Box 1738, 3000 DR
Rotterdam, The Netherlands. Fax: +31-10-4089449; e-mail:
burdorf@mgz.fgg.eur.nl
1Department of Public Health, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The
Netherlands; 2Centre for Occupational and Environmental Health, University of Manchester,
Manchester, UK
Received 8 November 2002; in final form 19 November 2002
Kromhout et al.’s (1993) well-cited publication presented detailed information on statistical
procedures to estimate the magnitude of exposure variability within and between workers,
drawing from a large database on chemical exposures throughout industry. It convincingly
demonstrated that the construct of homogeneous exposure groups often does not hold true and
suggested ways to improve measurement strategies. The authors hit a rich vein of research, and
many publications, not at least by the authors themselves, followed in the decade after publica-
tion. In recent years the principles of estimating the variation in exposure have been applied in
new methods for optimization of sampling strategies, for compliance testing, for quantifying
exposures in epidemiologic studies, and for identifying important sources of emissions and
suggesting strategies for controlling exposures. Many occupational hygienists across the globe
have adopted these new methods as powerful tools in their exposure assessment strategies.
Keywords: exposure assessment; variability
INTRODUCTION
An essential tool of occupational hygiene has always
been the assessment of exposures to potentially
hazardous chemicals in the workplace. The purpose
of occupational exposure assessment strategies can
be manifold, ranging from diagnostic monitoring to
reveal the sources and tasks that pose the largest
exposures in a workplace to characterizing the distri-
bution of exposures of workers at risk. One of the
first authoritative documents on evaluation of work-
place exposures was the well-known NIOSH Occu-
pational Exposure Sampling Strategy Manual from
1977 (Leidel et al., 1977). In this manual, emphasis
was placed on procedures to demonstrate that expos-
ure at the workplace would not exceed the threshold
limit value. A formal compliance test was advocated,
based on a limited number of exposure measurements
on employees believed to experience the highest
exposure (worst-case approach). Within a few years
this NIOSH Manual had become the most cited and
discussed publication in occupational hygiene.
A fundamental criticism of the NIOSH compliance
test was that decisions would only be reasonable if
the air concentration experienced by a worker was
more or less constant. It became clear that the worst-
case sampling strategy was surrounded with large
uncertainties, partly stemming from the inability to

96 A. Burdorf and M. van Tongeren
a priori select workers with the highest exposure.
Hence, attention shifted towards strategies where
personal monitoring was performed on a sample of
workers with similar exposure profiles (Burdorf ,
1993). Different techniques were developed to assign
workers to homogeneous exposure groups,
depending on information collected on work
processes, chemicals, jobs, tasks and actual layout of
the workplace (Corn and Esmen, 1979; Hawkins et
al., 1991).
Measurement strategies based on homogeneous
exposure groups have become very popular. This
concept is at the core of many workplace surveys,
especially in risk assessments for legal requirements
and epidemiological studies. The growing insight in
variability of exposure, and its impact on the homo-
geneity of occupational exposure groups, spurred
research into quantification of specific parameters of
exposure distributions within occupational groups. In
1993, Kromhout and colleagues published a compre-
hensive evaluation of determinants of exposure vari-
ability and convincingly demonstrated that many
occupational groups were not uniformly exposed as
was generally assumed by occupational hygienists
(Kromhout et al., 1993). Similarities in environ-
mental conditions, work environments, job tasks and
identifiable exposures may not always be sufficient
to assign workers to homogeneous exposure groups
without availability of quantitative exposure data.
The phenomenon of variability of exposure over time
and among persons needs to be understood since it
affects the basic elements of any measurement
strategy. It was exactly this message of Kromhout
and co-authors that attracted attention and their publi-
cation soon became a source of inspiration for many
researchers and practitioners alike. This paper
describes the results of this classic paper and analyses
its impact on newly developed measurement and
control strategies in occupational hygiene.
ANALYSIS OF EXPOSURE VARIABILITY
Although most analysis of variance techniques
stem from the early 1960s, their application in occu-
pational hygiene was not advocated until the late
1980s. Early examples were published by Samuals et
al. (1985), Spear et al. (1987) and Kromhout et al.
(1987). In his review of methods for exposure assess-
ment, Rappaport (1991) presented detailed guidance
on the estimation of the variance components of
exposure and their interpretation in defining exposure
groups. The analysis described exposure patterns in
31 groups of workers exposed to nine different
agents, with exposure information based on repeated
measurements on individual workers within each
group. He introduced the term ‘monomorphic group’
for a uniformly exposed group, defned as a group in
which 95% of the individual mean exposures lie
within a factor of 2 (Rappaport, 1991). On the basis
of the between-worker variance within a specified
group of workers, the ratio of the 97.5th percentile to
the 2.5th percentile (range ratio) was estimated. In 27
out of 31 groups this range ratio was >2, ranging from
2.6 to 6230. These results prompted Rappaport to
argue that occupational hygienists should move away
from compliance testing in worst-case sampling strat-
egies, since variability of exposure would make it
almost impossible to identify the most exposed
workers by walk-through surveys.
This publication was followed by a discussion on
statistical techniques to calculate the within- and
between-worker variance. In a letter to the editor,
Heederik et al. (1991) proposed a more appropriate
procedure and suggested expanding the analysis to
other occupational groups. This correspondence lead
to a collaboration and, subsequently, a more compre-
hensive evaluation in order to investigate the general-
izability of the conclusions. A much larger database
was constructed with exposure data from the UK,
The Netherlands, the USA, Sweden and China, and a
detailed analysis of exposure variability was con-
ducted with approximately 14000 measurements
obtained from more than 1500 workers in 165 occu-
pational groups, defined by job title and factory
(Kromhout et al., 1993). Exposures were measured
by personal sampling on at least two occasions,
which enabled the estimation of the within- and
between-worker components of variance. Of all
occupational groups, only 42 groups (25%) had 95%
of the individual mean exposures lying within a
factor 2, almost 30% of the groups had a more than
10-fold range, and 10% of the groups showed a range
of over 50-fold. Generally, the within-worker vari-
ability exceeded the between-worker variability, sug-
gesting even larger differences in exposure between
work shifts than among workers with the same job in
the same factory. The influence of the measurement
strategy was also evaluated, demonstrating that groups
with non-randomly chosen workers and workers
measured on non-randomly chosen days had signifi-
cantly lower between-worker variability than in a
random sampling strategy but the non-random
approach increased the day-to-day variability.
In addition, it was shown that production factors
had a clear impact on the within-worker variability,
but less on the between-worker variability. The largest
day-to-day variability was demonstrated among groups
working outdoors, those working without local
exhaust ventilation, groups with mobile workers, and
groups working with intermittent processes. A
regression model with environment (outdoors versus
indoors) and type of process (intermittent versus
continuous) explained 41% of the variability in the
within-worker component of variance. On the basis
of these results, the authors concluded that it seemed
impossible to predict which occupational groups are

Variability in workplace exposures 97
more or less homogeneously exposed and, therefore,
a priori assumptions about of homogeneity were not
possible. They strongly advocated using measure-
ment strategies with repeated measurements from the
same individuals (Kromhout et al., 1993).
DEVELOPMENT OF NEW MEASUREMENT
STRATEGIES
The work of Kromhout, Symanski and Rappaport
soon attracted attention and publications quickly
followed on components of exposure variability in
various working conditions. Within a year their
results were corroborated in studies among bakery
workers (Burdorf et al., 1994; Nieuwenhuijsen et al.,
1994), electric power workers (Loomis et al., 1994)
and sodium borate workers (Woskie et al., 1994).
The technique of partitioning exposure variability
into its main components became a method often
used in research (Symanski et al., 2001; Vinzents et
al., 2001) and an essential element in evaluation of
exposure data in different areas (Kromhout and
Vermeulen, 2001; Tielemans et al., 2002; Wild et al.,
2002). This new concept in assessment of occupa-
tional exposures was soon adopted into optimization
of sampling strategies, formalized procedures for
compliance testing, measurement strategies in
epidemiological studies, and strategies to identify
important sources of emissions.
Information on the expected variability in exposure
among subjects may guide towards an optimum
sampling scheme for exposure measurements. The
appropriate number of measurements depends on the
relative accuracy, study size and discriminatory
power. Formulae have been presented that combine
classical equations for determining the discrimina-
tory power of a survey with expressions for evalu-
ating the influence of exposure variability on the
precision of the average exposure (Armstrong et al.,
1992). The efficiency of increasing the number of
repeated measurements or increasing the number of
subjects, is partly determined by the variance ratio
(Werner and Attfield, 2000). In most surveys, costs
considerations will be incorporated in the decision on
the required efficiency of the sampling scheme
(Lemasters et al., 1996).
A new compliance testing procedure for agents
with chronic health effects has been developed that
accounts for within and between sources of vari-
ability (Rappaport et al., 1995). This procedure starts
with two shift-long measurements randomly collected
from each of 10 randomly chosen workers from an
occupational group. In the first step it is evaluated
whether the selected occupational group may be
regarded as a monomorphic group. For monomorphic
groups the probability of overexposure is assessed
and for non-monomorphic groups alternative grouping
should be attempted. For occupational groups with
unacceptable exposure levels, resampling is suggested
to increase the power of the compliance test. If it
appears that workers in the occupational group are
uniformly exposed to unacceptable levels, engin-
eering or administrative controls are commended.
For non-uniformly exposed workers in a group, inter-
ventions at individual level should be considered,
such as modifications of tasks and work practices
(Rappaport et al., 1995).
The consequences of exposure variability in epi-
demiological studies have primarily been explored in
the context of its effect on the exposure–response
relationship. It has been shown that in a study with
measurements on all individuals the variance ratio
(within-worker variance/between-worker variance) is
directly linked to the attenuation in the observed risk
estimate (Liu et al., 1978). Hence, the exposure–
response function depends on the degree to which the
exposure assessment is successful in providing precise
estimates of individual exposure levels. In epidemio-
logical surveys it is more common to monitor a
random sample of workers in each occupational
group under study. In the analysis of an exposure–
response relationship all subjects within the same
group will be assigned the same exposure level.
Kromhout and Kupper have developed mathematical
expressions that use estimates of variance compon-
ents of exposure (within- and between-worker and
between-exposure group) for estimating the group-
based attenuation and evaluating the effect of
different grouping strategies on observed associations
between exposure and health outcomes. These
formulae were first presented in a keynote lecture at
the Exposure Assessment Conference in Lyon, 1994,
and subsequently published (Kromhout et al., 1996).
The equations were applied to industry-wide surveys
in order to study the effects of various sources of
exposure variability on choices among different
analysis strategies. In general, the individual-based
strategies will generate more precise, though biased,
estimates, while group-based strategies will result in
less precise but essentially unbiased estimates (Tiele-
mans et al., 1998). Several authors have used these
mathematical expressions to evaluate the effect of
exposure variability within and between occupational
groups on risk estimates in epidemiological studies
(Van Tongeren et al., 1997; Werner and Attfield,
2000).
The introduction of linear mixed-effects models in
standard statistical packages allows the simultaneous
estimation of the variance components and deter-
minants of the exposure (Peretz et al., 2002; Rappa-
port et al., 1999). A mixed-effects model combines
fixed and random effects into one model. A fixed
effect in such a model assumes that the differences in
exposure levels reflect true (constant) differences
among workplaces. The random effect evaluates
whether the variance within each workplace can

98 A. Burdorf and M. van Tongeren
partly explain the differences in average exposure
among the workplaces. Using these modelling tech-
niques it is possible to identify important deter-
minants of exposure, such as presence of ventilation,
type of production process, residence characteristics,
and time-activity patterns, while accounting for
random within- and between-worker variability. A
good illustration of this approach was recently
presented by Burstyn and colleagues (2000). A large
database on exposure among asphalt workers from 37
different sources in eight countries was constructed.
This database enabled the researchers to present three
models on the important determinants of bitumen fume,
bitumen vapour and polycyclic aromatic hydro-
carbon (PAH) exposure intensity among paving
workers. These statistical models explained 36–43%
of the total variability and revealed strong associa-
tions with various production factors, such as surface
dressing, oil gravel paving, and asphalt temperature
(Burstyn et al., 2000).
CONCLUSIONS
This article has focused on recent developments in
occupational hygiene that involve estimation of the
variability in exposure as a crucial concept in
exposure assessment strategies among occupational
groups. The classic papers of Rappaport (1991) and
Kromhout et al. (1993) have lead to further develop-
ments in exposure assessment strategies in the past
10 yr and has resulted in a greater understanding of
occupational exposure. The knowledge on estimation
procedures for exposure variability have been incor-
porated in new methods for optimization of sampling
strategies, for compliance testing, for quantifying
exposures in epidemiologic studies, and for iden-
tifying important sources of emissions and deter-
mining control strategies. Occupational hygienists
and epidemiologists worldwide have adopted these
new methods as powerful tools in their assessment
strategies.
REFERENCES
Armstrong BG, White E, Saracci R. (1992) Principles of expos-
ure measurement in epidemiology. New York: Oxford
University Press.
Burdorf A. (1993) Shifting concepts in in assessment of occu-
pational exposures. Ann Occup Hyg; 37: 447–50.
Burdorf A, Lillienberg L, Brisman J. (1994) Characterization
of exposure to inhalable flour dust in Swedish bakeries. Ann
Occup Hyg; 38: 67–78.
Burstyn I, Kromhout H, Kauppinen T et al. (2000) Statistical
modelling of the determinants of historical exposure to bitu-
men and polycyclic aromatic hydrocarbons among paving
workers. Ann Occup Hyg; 44: 43–56.
Corn M, Esmen NA. (1979) Workplace exposure zones for
classification of employee exposures to physical and chemi-
cal agents. Am Ind Hyg Assoc J; 40: 47–57.
Hawkins NC, Norwood SK, Rock JC. (1991) A strategy for
occupational exposure assessment. Akron, OH: American
Industrial Hygiene Association.
Heederik D, Kromhout H, Burema J. ( 1991) Letter to the editor
on Rappaport: Assessment of long-term exposures to toxic
substances in air. Ann Occup Hyg; 35: 671–4.
Kromhout H, Vermeulen R. (2001) Temporal, personal and
spatial variability in dermal exposure. Ann Occup Hyg; 45:
257–73.
Kromhout H, Oostendorp Y, Heederik D et al. (1987) Agree-
ment between qualitative exposure estimates and quantita-
tive exposure measurements. Am J Ind Med; 12: 551–62.
Kromhout H, Symanski E, Rappaport SM. (1993) A compre-
hensive evaluation of within- and between-worker compo-
nents of occupational exposure to chemical agents. Ann
Occup Hyg; 37: 253–70.
Kromhout H, Tielemans E, Preller L, Heederik D. (1996) Esti-
mates of individual dose from current exposure measure-
ments. Occup Hyg; 3: 23–39.
Leidel NA, Busch KA, Lynch JR. (1977) Occupational expos-
ure sampling strategy manual. USDHEW, NIOSH Publica-
tion No. 77-173. Cincinnati, OH: National Institute of
Occupational Safety and Health.
Lemasters GK, Shukla R, Li YD et al. (1996) Balancing cost
and precision in exposure assessment studies. J Occup
Environ Med; 38: 39–45.
Liu K, Stamler J, Dyer A et al. (1978) Statistical methods to
assess and minimize the role of intra-individual variability in
obscuring the relationship between dietary lipids and serum
cholesterol. J Chron Dis; 31: 399–418.
Loomis DP, Peipins LA, Browning SR et al. (1994) Organiza-
tion and classification of work history data in industry-wide
studies: an application to the electric power industry. Am J
Ind Med; 26: 413–25.
Nieuwenhuijsen MJ, Sandiford CP, Lowson D et al. (1994)
Dust and flour aeroallergen exposure in flour mills and bak-
eries. Occup Environ Med; 51: 584–8.
Peretz C, Goren A, Smid T et al. (2002) Application of mixed-
effects models for exposure assessments. Ann Occup Hyg;
46: 69–77.
Rappaport SM. (1991) Assessment of long-term exposures to
toxic substances in air. Ann Occup Hyg; 35: 61–121.
Rappaport SM, Lyles RH, Kupper LL. (1995) An exposure-
assessments strategy accounting for within- and between-
worker sources of variability. Ann Occup Hyg; 39: 469–95.
Rappaport SM, Weaver M, Taylor D et al. (1999) Application
of mixed models to assess exposures monitored by construc-
tion workers during hot processes. Ann Occup Hyg; 43:
457–69.
Samuals SJ, Lemasters GK, Carson A. (1985) Statistical
methods for describing occupational exposure measure-
ments. Am Ind Hyg Assoc J; 46: 427–33.
Spear RC, Selvin S, Schulman J et al. (1987) Benzene exposure
in the petrolium refining industry. Appl Ind Hyg; 2: 15–63.
Symanski E, Sallsten G, Chan WY et al. (2001) Heterogeneity
in sources of exposure variability among groups of workers
exposed to inorganic mercury. Ann Occup Hyg; 45: 677–87.
Tielemans E, Kupper LL, Kromhout H et al. (1998) Individual-
based and group-based occupational exposure assessment:
some equations to evaluate different strategies. Ann Occup
Hyg; 42: 115–9.
Tielemans E, Marquart H, de Cock J et al. (2002) A proposal
for evaluation of exposure data. Ann Occup Hyg; 46: 287–
97.
Van Tongeren M, Gardiner K, Calvert I et al. (1997) Efficiency
of different grouping schemes for dust exposure in the
European carbon black respiratory morbidity study. Occup
Environ Med; 54: 714–9.
Vinzents PS, Schlunssen V, Feveile H et al. (2001) Variations
in exposure to inhalable wood dust in the Danish furniture
industry. Within- and between-worker and factory compon-

Variability in workplace exposures 99
ents estimated from passive dust sampling. Ann Occup Hyg;
45: 603–8.
Werner MA, Attfield MD. (2000) Effect of different grouping
strategies in developing estimates of personal exposures:
specificity versus precision. Appl Occup Environ Hyg; 15:
21–5.
Wild P, Sauleau EA, Bourgkard E et al. (2002) Combining
expert ratings and exposure measurements: a random effect
paradigm. Ann Occup Hyg; 46: 479–47.
Woskie SR, Shen P, Eisen EA et al. (1994) The real-time dust
exposures of sodium borate workers: examination of expos-
ure variability. Am Ind Hyg Assoc J; 55: 207–17.