Researchers early on recognized that chronic beryllium
disease (CBD) occurred after both high and low levels
of exposure and hypothesized that the disease was immunologically
mediated (Sterner and Eisenbud 1951). Subsequent work
has confirmed the importance of cellular immunity to
beryllium in the pathogenesis of CBD (Rossman 2001).
The factors that determine why some individuals develop
cellular immunity to beryllium while others do not
still need to be elucidated. Medical screenings of
beryllium-exposed workers consistently demonstrate
that a larger percentage of individuals will have a
positive blood lymphocyte proliferation test to beryllium
(become sensitized) than will be diagnosed with CBD
(sensitization and granuloma in lung parenchyma) (Henneberger
et al. 2001; Kelleher et al. 2001; Kreiss et al. 1993a;
Stange et al. 2001). It is not known what proportion
of individuals who are sensitized to beryllium will
progress to develop CBD. Furthermore, there is a varied
clinical presentation of patients with CBD and variability
in its progression (Newman et al. 1996; Rossman et
al. 1999).
The current occupational air standard for beryllium,
first proposed in 1951, was based on the toxicity of
other metals such as arsenic, lead, and mercury and
modified to reflect beryllium’s lower atomic
weight and concern about its greater toxicity (Eisenbud
1982). Epidemiologic health outcome and exposure studies
were not used to develop the initial time-weighted
average permissible exposure level of 2 µg/m3.
Fifty years later, this remains the current air level
that Occupational Safety and Health Administration
(OSHA) enforces in the workplace.
Recent studies looking at beryllium disease and exposure
have either used a surrogate of exposure (i.e., months
of exposure, percent exposed to unfired beryllium oxide)
or calculated exposure metrics and found increased
disease with some parameters of increased exposure
(Henneberger et al. 2001; Kelleher et al. 2001; Kreiss
et al. 1993b, 1997; Viet et al. 2000). One study found
an exposure-response relationship for sensitization
with CBD but not for sensitization without CBD (Viet
et al. 2000). Other work has addressed the possibility
of particle size (McCawley et al. 2001), skin absorption
(Tinkel et al. 2003), and/or genetic susceptibility
(Saltini et al. 2001) as important factors that confound
a straightforward exposure-response relationship.
We investigated possible exposure-response relationships
separately for various measures of exposure, including
mean, peak, and cumulative metrics and differing chemical
and physical forms for the development of beryllium
sensitization and for the development of CBD.
We have also assessed whether the current OSHA (2005)
and Department of Energy (DOE 1999) permissible levels
were protective against the development of CBD and
sensitization.
The cohort was composed of workers from a beryllium
production facility in eastern Pennsylvania, which
operated from 1957 to 1978. The names of former workers
with at least 2 days of work up to 31 December 1969
who had previously been identified from personnel records
and matched with Social Security Administration Form
941 records by the National Institute for Occupational
Safety and Health (NIOSH), as part of a seven-company
mortality study, were obtained from NIOSH (Ward et
al. 1992). The last owner of the facility provided
the names of workers, social security numbers, demographic
information, and the last known address of all individuals
who began work from 1 January 1970 until the plant
closed in 1978.
Because this study was a cooperative effort with
NIOSH, addresses from the last income tax filing of
members of the cohort were obtained by NIOSH from the
Internal Revenue Service. NIOSH had previously ascertained
the vital status of the cohort as of 31 December 1988
using the Social Security Administration, the Internal
Revenue Service, post office cards mailed to the last
known address, the Department of Veterans Affairs,
the Health Care Finance Administration, and the National
Death Index (Ward et al. 1992).
We mailed the initial invitation to participate in
the medical screening program to the last known address
of all members of the cohort not known to be deceased
as of 31 December 1988. The mailing included a cover
letter about the study, a fact sheet about beryllium,
a one-page two-sided questionnaire, and a postage-paid
envelope. The questionnaire requested demographic information
and had questions about previous lung disease, smoking
history, and work history at the beryllium facility.
We attempted to contact everyone who did not return
the questionnaire. This included multiple phone contacts
or actual visits to the person’s home if telephone
contact was unsuccessful. Internet address searches
using search engines such as Yahoo! and Netscape were
performed to locate current mailing addresses of individuals
with returned mailings. In addition, we used the Social
Security Death Index (Ancestry.com 2005) to help determine
vital status of individuals. Local staff in the two
communities not only made visits to last known addresses
but also asked the long-term workers to assist in identifying
individuals who could not be located.
All individuals located, whether or not they participated
in the medical screening or completed a questionnaire,
received a subsequent mailing summarizing the results
of the screening and notification of federal legislation
passed in the fall of 2000 that provided compensation
for workers with CBD and coverage for medical costs
for follow-up of workers with beryllium sensitization
from this facility.
All individuals located had the opportunity to have
a blood lymphocyte proliferation test for beryllium
(BeLPT), a posterior-anterior chest radiograph, and
simple spirometry. Before the testing, we obtained
consent to conduct testing from the individual. In
addition, each participant completed a questionnaire
on other work exposures that might contribute to respiratory
deficiencies. This included other possible sources
of beryllium exposure as well as exposure to asbestos,
coal, and silica.
Medical testing was performed at two primary sites
in the community in eastern Pennsylvania. For individuals
who had moved to other parts of the country, medical
testing was performed in a location convenient to the
individual (i.e., personal physician, local medical
centers, etc.).
Blood was collected Monday through Thursday and shipped
for next morning delivery. All blood was processed
the next day and analyzed. All BeLPT was performed
at the University of Pennsylvania. Any individual with
a positive BeLPT test was offered a repeat test. If
an individual’s results were negative on the
repeat test, then the individual was offered the opportunity
to repeat the blood test 1 year later.
A panel of three “B” readers interpreted
all chest radiographs. One B reader was a radiologist
(J.A.), one a pulmonologist (J.E.P.), and one an internist
and occupational medicine physician (K.R.). At least
two B readers had to classify a radiograph with ≥ 1/0
profusion in order for a radiograph to be classified
as positive for parenchymal disease.
Any individual who had two positive BeLPTs and/or
a consensus chest radiograph reading of ≥ 1/0
for profusion was referred to the University of Pennsylvania
for follow-up testing, which consisted of a posterior-anterior
chest radiograph, a BeLPT, an electrocardiogram, a
complete medical history including respiratory symptoms
using a standardized collection instrument, and bronchoscopy
with both bronchial biopsy and lymphocyte testing of
lavage fluid for beryllium. All bronchoscopies were
performed by a single pulmonologist (M.R.).
Table
1
|
Whether or not an individual had CBD or beryllium
sensitization was decided by consensus by the internist/occupational
physician (K.R.) and pulmonologist (M.R.). Table 1
outlines the criteria used to categorize the medical
testing results. However, only individuals who had
bronchoscopy were used in the analysis describing the
predictive power of radiographs or BeLPT.
All individuals received a letter with the results
of their initial screening and, where applicable, a
letter with the results of the
follow-up testing. The Human Subject Review Boards of Emory University, Michigan
State University, the University of Cincinnati, and the University of Pennsylvania
approved this study.
Through discussions with long-term production and
management employees, we identified major changes in
the process and engineering/work practice controls.
Trends in the exposures over time were evaluated in
relation to dates of process changes and visually from
plots of the data to identify other time points at
which exposure measurements indicated a change in conditions.
Exposure had been monitored at the facility using
a method that combined the concentration at each task
performed by a worker, weighted by the duration in
the shift of that task; the products of concentration
and duration at all tasks performed as part of a job
were summed and divided by the duration to the shift.
This final value was called the daily weighted average
(DWA) exposure. Data accumulated over the operating
history of the plant were identified and computerized.
Using this information, a task exposure matrix (TEM)
and a job exposure matrix (JEM) were constructed (Chen
2001). Task-related exposure measurements were available
for two time periods, 1957-1962 and 1971-1976. Because
the data most closely followed a log-normal distribution,
the geometric mean was calculated for each task-year
combination. For years with no measurements, we estimated
exposures by interpolating between the previous and
subsequent values. For example, if measurements were
available for 1957, 1958, and 1959 but not for 1960,
the 1959 value was entered into the TEM. The plant
history was used to develop a strategy for imputing
values from 1963 to 1971. We used the mean of task
estimates for 1962 and 1971 for the period 1963-1969;
because of the engineering changes in 1970, the 1971
values were used for 1970. Estimates for 1976 were
used for the remaining years of plant operation, based
on employee interviews. For tasks never measured, the
task in the same work area most similar to the unmeasured
task was identified with the assistance of long-term
employees; the exposure value for the measured task
was entered into the TEM for the unmeasured task.
We completed the JEM by first calculating the geometric
mean exposure for each year in which at least one DWA
measurement was available. Exposure estimates for job-year
combinations without measurements were estimated based
on the plant history of engineering changes. In the
absence of information showing production or control
technology changes in years before or after measurement
data, the measurements were assumed valid and extended
to the empty cells in the JEM. Where increases or decreases
in exposure were justified from the plant history,
we used analysis of variance (ANOVA) to evaluate the
significance of the change in exposure. Where statistically
significant changes were identified, the new value
was entered into the cell of the JEM.
For 39 of the 130 job titles, no measurements were
available for the job in any year. For each of these
jobs, we used information from the long-term workers
to identify the job with tasks most similar to it with
measurements. The time-activity pattern needed for
the evaluation of exposure was developed and used to
calculate a DWA estimate of exposure using data in
the TEM.
The values were reviewed by a group of long-term
employees who represented experience in all production
areas of the facility, maintenance, and management.
They were specifically asked to review the relative
exposure values for production areas. For example,
the exposure estimated for the fluoride furnace operator
is slightly higher than the helper; this was confirmed
to be correct because the helper stood away from the
furnace and supplied materials to the perimeter only.
The involvement of the group of long-term employees
provided added confidence in our derived estimates.
The JEM and TEM were linked through the listing of
the tasks in each job taken from the DWA calculation
sheets. For jobs never sampled, the association was
through the time-activity information developed with
the help of long-term employees and, finally, put into
DWA format.
For every job title in the JEM, the chemical and
physical form of the exposure was listed. Chemical
forms included beryl ore, beryllium metal, beryllium
fluoride, beryllium hydroxide, and beryllium oxide;
physical forms included dust, fume, or mixed (dust
and fume). Individuals from the facility were assigned,
based on jobs worked, the number of months exposed
to three different chemical forms: nonsoluble beryllium
compounds (beryllium metal and oxide), soluble beryllium
compounds (beryllium fluoride and hydroxide), and mixed
chemical forms. Individuals were similarly assigned
to the number of months exposed to the three physical
forms: dust (beryllium metal, hydroxide, or oxide),
fume (beryllium fluoride), and mixed (mixed dust and
fume). This allowed us to evaluate any differences
in response due to very small particle size (fume)
or larger particle size (dust or mixed).
We used chi-square tests to compare the groups (definite
or probable disease vs. sensitized vs. no disease)
with respect to discrete outcomes. ANOVA was used to
compare the groups with respect to continuous outcomes
(age, cumulative, mean, and peak exposure levels).
For the three disease outcome group comparisons, a
screening p-value was set at 0.25, below which
the pairwise comparisons between groups (definite or
probable disease vs. no disease, definite or probable
disease vs. sensitized, sensitized vs. no disease)
were
further investigated. For the discrete outcomes, further chi-square tests were
performed on the resulting 2 k tables.
For the continuous outcomes, the linear contrasts for these pairwise comparisons
were examined in order to control for multiple comparisons. For ease of presentation,
we also used two-sample t-tests to examine pairwise comparisons of the
groups. These parametric tests were followed by the Wilcoxon rank-sum test,
a nonparametric test used to ameliorate the effects of violations of the assumptions
for the parametric tests (e.g., normal distribution).
We further explored exposure-response relationships
with logistic regression analysis after adjustment
for potential confounders (smoking, age, other beryllium
exposure). In addition to an analysis where only cases
with complete information were included, an analysis
was carried out after multiple imputations (Rubin 1987)
of cumulative and mean exposure values (missing on
65 of 574 individuals).
p-Values are presented as calculated. All
analyses were performed using SAS statistical software
(version 9.1; SAS Institute, Cary, NC). The results
of the spirometry testing are not reported in this
article.
A total of 1,351 individuals were identified to have
worked at this facility. A summary of the participation
rate for this facility is shown in Table 2. Approximately
one-fourth (24.4%) of the cohort died before the medical
screening began, and another 10.8% could not be located.
Among the 875 individuals located, 160 (11.8%) indicated
either that they had worked for the company that owned
the facility but at a different location, or that they
had completed a job application and underwent a pre-employment
physical for work at the facility but had either not
been hired or had decided not to accept a job at that
plant.
Of the remaining 715 former employees, the participation
rate was 63.9% (457 of 715) for completion of all components
of the medical screening and 91.3% (653 of 715) for
completion of the questionnaire only. Five hundred
twenty-eight individuals (73.8%) completed at least
the blood and chest radiograph component. Reasons members
of the cohort gave for not participating included that
the individual a) had only worked for a short
time; b) felt he or she was too old and that
testing would not matter; c) did not have any
health problems; d) did not want to jeopardize
his or her current health insurance, especially with
no compensation available (at the time the individual
was contacted); and e) felt there was no effective
treatment for beryllium disease.
Table 3 compares the demographics of medical screening
participants with nonparticipants. Medical screening
took place from the fall of 1996 through the summer
of 2001. Participants were on average the same age
as the nonparticipants, the same sex and race, last
worked in a more recent year, and worked on the average
3.3 years longer. Participants were mainly male (91%),
and almost all white. Seventy percent had ever smoked
cigarettes.
Among the 577 individuals that were tested for beryllium,
110 were referred for follow-up testing at the Hospital
of the University of Pennsylvania (Table 4). In addition
to the 110 referred, 9 individuals from the facility
had previously been diagnosed at the University of
Pennsylvania with CBD. All 577 individuals, including
the 9 previously diagnosed with CBD, were categorized
per the criteria in Table 1. The results of this classification
are shown in Table 5. Of the cohort, 7.6% (44) had
probable or definite CBD, 2.1% (12) had possible CBD,
6.9% (40) were sensitized to beryllium, and 4.0% (23)
were possibly sensitized.
Table 6 shows the predictive power of having unrecognized
CBD documented by bronchoscopy based on the results
of the screening tests performed. Having two positive
BeLPTs and scarring on the chest radiograph, involving
either all zones or the lower zones only, had the highest
predictive value for the development of CBD (100%).
In descending order for the other combination of tests,
the predictive values for CBD were scarring on the
radiograph in all zones with negative BeLPT (75%),
positive BeLPT (48.3%), scarring in the upper zones
with negative BeLPT (40%), and scarring on the chest
radiograph just in the lower zones (7.7%).
There were 33 cases of definite/probable CBD among
production workers, 5 among clerical/office workers,
3 in engineers, 1 in a supervisor/inspector, 1 in a
laboratory worker, and 1 in an industrial hygiene technician.
There were 27 cases of sensitization among production
workers, 10 among clerical/office workers, 2 among
engineers, and 1 in a nurse.
Table 7 shows the occurrence of definite and probable
CBD and sensitization by first decade worked, last
decade worked, and duration of years worked. The mean
year of first exposure for definite/probable CBD was
1963, for sensitized cases it was 1965, and for the
normal group it was 1964. Further, the mean year last
exposed for definite/probable, sensitized, and normal
individuals was 1973, 1968, and 1971, respectively.
The mean duration of exposure for definite/probable,
sensitized, and normal individuals was 9.4 years, 2.7
years, and 8.7 years, respectively.
Tables 8-12 show the occurrence of definite and probable
CBD and sensitization by the peak, average, and cumulative
exposure metric, by chemical and physical form of beryllium
and the OSHA (2005) standard of 2 µg/m3 and
the DOE (1999) standard of 0.2 µg/m3.
Individuals who were sensitized had a lower total cumulative
and peak exposure (Table 8), lower nonsoluble cumulative
and average exposure (Table 11), and lower dust and
mixed exposure (Table 10). Individuals with CBD had
a lower soluble (Table 9) and fume exposure (Table
10). The mean beryllium exposure levels for the DWA
categories in Table 11 were 0, 1.23, and 8.95 µg/m3,
respectively and in Table 12 were 0.14, 1.19, and 4.76 µg/m3,
respectively.
The prevalence of CBD and sensitization to beryllium
in former workers at this beryllium production facility
in eastern Pennsylvania was high: 7.6% with CBD, 6.9%
with sensitization, 2.1% with possible CBD, and 4.0%
with possible sensitization. This facility operated
from 1957 to 1978. Representative exposure estimates
for tasks ranged from 0.9 to 84.0 µg/m3 in
the 1960s, although most time-weighted averages were
below the OSHA (2005) standard of 2 µg/m3,
ranging from 1.1 to 2.5 µg/m3. Exposure
estimates in the 1970s were lower, with representative
tasks ranging from 0.5 to 16.7 µg/m3 and
time-weighted averages ranging from 0.7 to 3.5 µg/m3.
The 14.5% prevalence of CBD and sensitization in
the cohort we studied contrasts with overall prevalence
reports of 3.3% among nuclear workers from Rocky Flats
(Stange et al. 2001), 1.8-5.9% from beryllium ceramics
manufacturing (Kreiss et al. 1993b, 1996), and 4.6%
from a beryllium production facility (Kreiss et al.
1997). Our overall prevalence is similar to the prevalence
reports for more highly exposed subgroups from these
studies, such as machinists (Kreiss et al. 1996). Our
higher overall prevalence rate reflects both the level
and the widespread exposure to beryllium in the facility
we studied, where 11 definite/probable cases occurred
among nonproduction workers such as clerical, supervisory,
and engineering staff and 13 sensitization cases occurred
in clerical/office personnel. Our mean and range of
cumulative exposure, which was 199.25 µg-year/m3 (0.0-3970.61 µg-year/m3,
are appreciably higher than estimates reported in other
studies: 6.09 µg-year/m3 (0.15-10.64 µg-year/m3)
(Kelleher et al. 2001), 1.35 µg-year/m3 (estimated
range, 0-6.41 µg-year/m3) (Viet et
al. 2000), and no mean provided (estimated range, 0.9-41.2 µg-year/m3)
(Henneberger et al. 2001). An additional factor that
probably contributes to the higher prevalence of CBD
in our cohort is the long latency since last exposure,
which would have allowed a higher proportion of individuals
who were sensitized to progress on to CBD than in other
cohorts that have been studied (Newman et al. 2005).
Most previous prevalence studies of beryllium-exposed
workers have been of current employees (Henneberger
et al. 2001; Kelleher et al. 2001; Kreiss et al. 1996),
or they have included former workers (Stange et al.
2001) but have not presented the results separately
for current and former workers. One study similar to
ours only had formerly exposed individuals (Kreiss
et al. 1993b). This latter study, unlike ours, found
no individuals with sensitizations alone without CBD.
This would suggest that the higher prevalence of CBD
in our study population was not solely related to the
long latency since last exposure because we would have
expected a lower rate of sensitization alone without
CBD if increased prevalence of CBD was solely caused
by the long latency.
Despite the fact that there is an overall increase
of beryllium disease in working populations with higher
exposure to beryllium, investigators have been unable
to show a clear-cut exposure response between air concentrations
of beryllium and CBD or sensitization (Henneberger
et al. 2001; Kelleher et al. 2001; Viet et al. 2000).
This has led researchers to examine the possible role
of particulate size (Kelleher et al. 2001; McCawley
et al. 2001) and skin exposure (Tinkle et al. 2003).
We found no difference in duration of exposure for
individuals with CBD versus those who had no evidence
of beryllium disease, but we did find that those who
were sensitized had begun work later, last worked longer
ago, and had a shorter duration of exposure than did
those with CBD or those who tested normal (Table 7).
This difference for individuals with sensitization
was also true for cumulative and peak exposure (Table
8), cumulative mixed and cumulative and mean nonsoluble
exposure (Table 9), cumulative and mean dust, and cumulative
mixed exposure (Table 10). On the other hand, cumulative
and mean soluble and cumulative and mean soluble fume
exposures were lower for CBD (Table 10).
In sum, we either found no exposure response or the
significant exposure responses we did find were in
the opposite direction than expected, with individuals
with CBD or sensitization having less estimated exposure
than those with no beryllium disease. The risk of CBD
compared with sensitization if a person’s mean
exposure was below the current DOE (1999) permissible
level of 0.2 µg/m3 was less than if
their mean level was > 0.2 µg/m3 but
below the current OSHA (2005) permissible exposure
level of 2 µg/m3 (Table 11). However,
only being exposed to beryllium less than either the
DOE or the OSHA time-weighted average did not protect
a worker from the development of CBD or sensitization.
There were only two people in the cohort whose highest
level of exposure was never above the 0.2 µg/m3 DOE
standard. CBD and sensitization occurred even if the
highest level of exposure was never greater than the
2 µg/m3 OSHA standard, and our data
would suggest that peak exposure levels > 0.2 µg/m3 were
as harmful as even higher peak exposure levels > 2 µg/m3 (Table
12).
A possible explanation for the failure to find an
association between increased beryllium disease and
sensitization and increased levels of exposure is that
this analysis did not consider the role of genetic
predisposition to both sensitization and disease. Because
the genetic marker glu69 on HLA-DPB1 has been associated
with 80-90% of cases of both CBD and sensitization,
a better control group for this analysis would be HLA-DPB1
glu69-positive individuals who did not have CBD or
sensitization. We have recently been funded to test
our population for this marker and thus will eventually
be able to determine the interaction of exposure and
genetic predisposition.
The finding of higher working lifetime beryllium
exposures in those with CBD compared with those who
are just sensitized suggests that the body burden of
beryllium might relate to the severity of disease in
those with a genetic predisposition. Our finding that
sensitized individuals compared with individuals with
CBD had a higher exposure to beryllium in a soluble
form and to fumes of beryllium supports this hypothesis
(in this facility soluble beryllium and fume is practically
equivalent, r = 0.94). Presumably the soluble
forms of beryllium would be more likely to be mobilized
and eliminated and result in a lower body burden of
beryllium compared with a similar exposure to insoluble
beryllium.
Individuals who have recently converted their PPD
(purified protein derivative) skin test for tuberculosis
to positive may, after treatment, revert to a negative
PPD (Tager et al. 1985). Thus, with decreasing or elimination
of the antigen, the cellular immune response (i.e.,
PPD reaction) may fade or be eliminated. Because the
PPD reaction is similar to BeLPT, this suggests that
a decreased immune response to beryllium may occur
in individuals with a lower body burden of beryllium
(i.e., antigen). Thus, a reduced immune response to
beryllium may account for the association of beryllium
sensitization with a lower body burden of insoluble
beryllium or predominantly soluble beryllium exposure
compared with individuals with CBD. An alternative
explanation that less soluble beryllium exposure is
confounded by elevated levels of other forms of beryllium
is not supported by analyzing potential correlations
between levels of exposures to the different forms
of beryllium.
Other researchers have suggested the importance of
skin exposures to the development of beryllium disease.
We have no data to directly address whether skin exposure
is of importance in the development of beryllium disease
in this cohort. However, others have hypothesized that
small particle size increases the likelihood of both
inhalation and skin absorption and exposure (McCawley
et al. 2001). Our data showed the opposite results
with reduced levels of exposure to fume, which would
be the smallest particle size form of exposure that
occurred in this facility, and CBD (Table 10).
A limitation of our study is the uncertainty in the
exposure estimates. The exposure metrics developed
for study participants were based on relatively sparse
data, with interpolation from measurement data for
years when no data were available. Major gaps in the
data were associated with the mid-1960s and from 1977
to 1981. Exposure estimates for the earlier time period
were based on measurements in preceding or succeeding
years; for the later time period, estimates for the
mid-1970s were extended into the later years. These
decisions were based on plant history and conversations
with long-term workers. All interpolation was accomplished
using preestablished rules and was independent of any
knowledge of disease status. The use of professional
judgment like this is often required in retrospective
exposure assessment studies. Because the exposure estimates
were created for jobs and tasks, without knowledge
of a work history or disease status, it is likely that
this misclassification would be nondifferential, attenuating
any ability to detect exposure-response relations (Checkoway
et al. 1991; Copeland et al. 1977).
A further limitation relates to the effect of nonparticipants
on study results. The overall participation rate was
high and nonparticipants were generally similar to
participants except their duration of exposure was
less. However, it is possible that the 11% of the total
cohort that did not participate had a lower rate of
CBD because asymptomatic individuals might be less
motivated to participate. On the other hand, the 24%
of the cohort who were deceased at the initiation of
our study and the 11% we could not locate might be
expected to have a higher prevalence of disease.
A third limitation of our study is that we used a
single laboratory for the blood lymphocyte testing
for beryllium in a one-time screening. It has recently
been reported that the use of a single laboratory results
in false negative results of 20-30% (Stange et al.
2004). Because radiographs were part of our screening,
we would expect the false negative rate for CBD to
be lower than the potential false negative rate for
sensitization. Because our cohort was no longer exposed
to beryllium, it is less likely that repeat screening
will identify additional cases of CBD or sensitization,
as has been shown in currently exposed cohorts (Newman
et al. 2001).
The participation level of individuals who warranted
more extensive testing after the initial screening
is another limitation of this study. Only 56 of the
110 (51%) individuals who screened positive by radiograph
or BeLPT elected to have a bronchoscopy. Participation
in more extensive testing was similar in those with
positive radiographs (47%) and those with abnormal
positive BeLPT only (57%). The lack of a biopsy and
broncholavage in half of the individuals who were positive
on the initial medical screening means we may have
misclassified individuals into the definite/probable
CBD and sensitization groups. This misclassification
would decrease the likelihood of finding an exposure-response
or other relationship with CBD or sensitization. To
minimize misclassification errors, we excluded cases
classified as possible CBD or possible sensitization
from both the disease and normal groups during analysis.
However, we included four individuals classified as
CBD because of a diagnosis at the University of Pennsylvania
before our study, although these individuals never
had evidence of sensitization in their bronchial lavage
fluid or blood. We are not aware of any reason how
misclassification could cause the inverse relationship
between exposure and disease that we found.
A final limitation is that multiple comparisons were
made in Tables 9 and 10. Adjustments for these multiple
comparisons can be made by tripling the p-value
reported, using the properties of the Bonferroni inequality.
If this were done, a number of the associations would
no longer be statistically significant in Tables 9
and 10. Given the consistent direction of the findings,
our conclusions concerning soluble and nonsoluble forms
of beryllium remain unchanged even if this adjustment
were made.
In conclusion, this cohort is a high-risk group for
the development of CBD and sensitization. The development
of beryllium disease has continued to occur years after
exposure has ceased. Former beryllium workers and their
health care providers need to be aware of this ongoing
risk. A combination of two positive BeLPTs and an abnormal
chest radiograph on the initial medical screening was
the best predictor of the presence of CBD. However,
there were individuals who had CBD with an abnormal
chest radiograph, involving all or just the upper lobes,
and negative BeLPT (Table 6).
We were unable to show an exposure-response relationship.
The inclusion of genetic data combined with exposure
data may better define which individuals in this cohort
are at a particularly high risk of development of CBD
and/or sensitization and may account for the absence
of the typical exposure-response seen with other environmental
or occupational toxins. We are currently performing
molecular typing of DRB1 and DPB1 alleles
on individuals with CBD and sensitization and a sample
of those who tested normal to investigate for a possible
gene-exposure relationship.
This cohort is a high-risk group for the development
of CBD and sensitization. The development of beryllium
disease has continued to occur years after exposure
has ceased. Former beryllium workers and their health
care providers must be kept aware of this ongoing risk.
The results of this study show that current occupational
health standards for beryllium do not provide adequate
protection against the development of CBD or sensitization.
Twenty-four percent of the workforce that was exposed
to beryllium below the current OSHA (2005) allowable
threshold limit value developed CBD or sensitization.
Similar levels of adverse outcomes (21%) were seen
in those exposed to beryllium below the time-weighted
average DOE (1999) guideline of 0.2 µg/m3.
Even the more protective time-weighted average of 0.02 µg/m3 proposed
by the American Conference of Governmental and Industrial
Hygienists (2005) did not eliminate adverse outcomes.
The identification of cases of CBD and sensitization
in this population at levels of cumulative
exposure lower than current standards or guidelines
underscores the need to more fully understand
the determinants of exposure (e.g., peak, physical/chemical
form) that may contribute to disease risk, so that
these may be included in standard setting.