Occupational and Environmental Medicine 1997;54:605-612

Risks of beryllium disease related to workprocesses at a metal, alloy, and oxide productionplant

Kathleen Kreiss, Margaret M Mroz, Boguang Zhen, Herbert Wiedemann, Barbara Barna

AbstractObjectives-To describe relative hazardsin sectors of the beryllium industry, riskfactors ofberyllium disease and sensitisa-tion related to work process were sought ina beryllium manufacturing plant produc-ing pure metal, oxide, alloys, and ceram-ics.Methods-All 646 active employees wereinterviewed; beryllium sensitisation wasascertained with the beryllium lym-phocyte proliferation blood test on 627employees; clinical evaluation and bron-choscopy were offered to people withabnormal test results; and industrial hy-giene measurements related to work proc-esses taken in 1984-93 were reviewed.Results-59 employees (9.4%) had abnor-mal blood tests, 47 of whom underwentbronchoscopy. 24 new cases of berylliumdisease were identified, resulting in aberyllium disease prevalence of 4.6%,including five known cases (29/632). Em-ployees who had worked in ceramics hadthe highest prevalence of beryllium dis-ease (9.0%). Employees in the pebble plant(producing beryllium metal) who hadbeen employed after 1983 also had in-creased risk, with a prevalence of beryl-lium disease of 6.4%, compared with 1.3%ofother workers hired in the same period,and a prevalence of abnormal blood testsof 19.2%. Logistic regression modellingconfirmed these two risk factors for beryl-lium disease related to work processes andthe dependence on time of the risk at thepebble plant. The pebble plant was notassociated with the highest gravimetricindustrial hygiene measurements avail-able since 1984.Conclusion-Further characterization ofexposures in beryllium metal productionmay be important to understanding howberyllium exposures confer high contem-porary risk ofberyllium disease.

(Occup Environ Med 1997;54:605-612)

Keywords: beryllium; occupational lung disease; lym-phocyte proliferation test; surveillance; exposure-response

Beryllium exposure leads to cell mediatedimmunological sensitisation in a small percent-age of workers exposed to beryllium aerosols,dusts, or fumes; of the sensitised workers, many

have granulomatous lung disease.' Preventionof beryllium disease depends on knowledge ofrisk factors which can be modified. Althoughinborn genetic factors are associated with riskof disease in those exposed to beryllium,4 thesecannot be changed in an existing workforceexposed to beryllium. In contrast, work relatedrisk factors offer the opportunity to lower riskof beryllium disease and to understand theexposure characteristics associated with highdisease rates. In our previous studies of plantworkforces exposed to beryllium, we foundrisks of beryllium sensitisation or diseaserelated to work processes in three plants repre-senting single sectors of the beryllium industry.These include machining of beryllium metal,'grinding, dicing, and drilling of berylliaceramics,' dry pressing, and research anddevelopment in a plant which manufacturedberyllia ceramics historically.' We report herethe results of epidemiological and exposuresurveillance in a plant which encompassesmost sectors of the beryllium industry inproduction of beryllium metal, alloys, andberyllium oxide from which ceramics weremade historically. We sought to describe risksof beryllium disease related to work processeswhich could provide opportunities for futurestudy of exposure variables conferring excessrisk. Understanding of qualitative and quanti-tative exposure-response relations is critical toprevention of disease in the many sectors of theberyllium industry.The plant opened in 1953 to produce

beryllium-copper alloy, which is cast and fabri-cated into tubes, wire, sheet, plates, and metalparts before shipment to other factories tobecome finished products. Beryllium metaloperations were developed in about 1957 inbuildings and under management which werelargely separate from alloy operations. Beryl-lium metal is produced from beryllium hydrox-ide through a chemical process. The two com-ponent areas involved in beryllium metalproduction are the pebble plant, which con-tains fluoride and reduction furnaces, andvacuum melting.' As the crystalline structure ofcast beryllium metal is unsuitable for manyapplications, the metal is partitioned intodiffering grades of powder and pressed intometal shapes. A machine shop grinds andfinishes many of the cast and pressed berylliumand alloy parts.The plant also reclaims scrap beryllium and

alloy metal and began production ofAlbemet, aberyllium-aluminum alloy, in 1990. Histori-

Occupational andEnvironmentalMedicine Division,National Jewish Centerfor Immunology andRespiratory Medicine,and Department ofPreventive Medicineand Biometrics,University of ColoradoHealth SciencesCenter, Denver,CO,USAK KreissMM MrozB Zhen

Cleveland Clinic,Cleveland, OH, USAH Wiedemann,B Barna

Correspondence to:Dr K Kreiss, Division ofRespiratory Disease Studies,National Institute forOccupational Safety andHealth, 1095 WillowdaleRoad, MS-234,Morgantown, WV 26505,USA.

Accepted 23 January 1997

605

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from

Kreiss, Mroz, Zhen, Wiedemann, Barna

Machining

Engineering Shipping or receivingAdministration Plant clean up Analytical labsSecurity Decontamination

Quality controlSales Jant ant Production controlMedical department Janitorial plant services Environmental control

Laundry

Figure Process classifications and their subgroups used in work history analyses for risks related to work processes and cumulative exposure estimates.

cally, the plant produced beryllium hydroxidefrom ore until 1972. Beryllium oxide forceramics has been produced since about 1958at the metal side of the plant. Beryllia ceramicproducts were produced at the plant until theoperation was moved to another factorybetween 1980 and 1984. Production of beryl-lium pebbles in metal production increasedcoincidentally during a similar period, dou-bling in 1979, declining to baseline, and thendoubling again in 1983-4.

MethodsPOPULATION

We invited all 655 employees on the plant ros-

ter during the screening period to participate inscreening for beryllium disease after presentinginformation sessions to each workshift. Fiveemployees had already had a diagnosis ofchronic beryllium disease. The management atthe plant required all current employees to beinterviewed during work time over a 13 monthperiod. The screening blood test for berylliumsensitisation was voluntary.

QUESTIONNAIREThe personnel department provided employ-ees with their work histories for review beforeconfidential interviews with non-plant person-nel. Trained interviewers administered an

abbreviated standardized respiratoryquestionnaire,6 supplemented with additionalmedical and work history questions (beforeblood test results were available) usually withintwo months of drawing blood. For data analy-sis of work history, jobs were classified by theindividual and grouped processes (fig 1).

BLOOD LYMPHOCYTE PROLIFERATION TESTAfter written informed consent, employeesgave a 40 ml blood sample collected inheparinised vacutainer tubes, half ofwhich wassent by overnight courier to one laboratory andhalf to another laboratory for the berylliumlymphocyte proliferation (BeLP) test.7 Split

blood samples were collected weekly over a 15month period from late February 1993 to June1994. Repeat testing of abnormal and indeter-minate results extended blood collection toOctober 1995.

CLINICAL EVALUATION

Participants with confirmed abnormal BeLPresults or with one abnormal result and an

indeterminate result during the study were

offered clinical evaluation after additionalinformed consent, including bronchoscopy forbronchoalveolar lavage BeLP test and trans-bronchial lung biopsy. People with an initialabnormal blood test that was not confirmed ontwo subsequent split tests were not referred forclinical evaluation but were offered repeatblood tests after the study. We definedberyllium sensitisation as a repeatable abnor-mal BeLP test. We defined beryllium disease asgranulomas on lung biopsy in the presence ofberyllium sensitisation or an abnormal bron-choalveolar lavage BeLP test.8

HISTORICAL INDUSTRIAL HYGIENE DATA

We reviewed all computerised industrial hy-giene measurements of beryllium mass con-

centration (n= 124 600) collected from 1980 to1993 without regard to particle size or

respirable fraction. We excluded 10 080 specialsamples without location codes and 8322 sam-ples that could not be classified as general area,full shift, breathing zone, or lapel samples, mostof which were taken between 1980 and 1984.General area samples (n=30 872) were col-lected with high volume samplers for 30minutes in the work area of specific processes.Continuous samples (n=59 360) were also col-lected, usually over a full workshift or occasion-ally over 24 hours. Process breathing zone

samples were collected for 1-15 minutes withhigh volume samplers (n=15 787). Beginningin 1990, personal lapel samples (n= 179) were

collected with a monitoring pump positionedon the worker's lapel for the entire work shiftfor 20 different jobs.

Alloy

Alloy metal productionArc furnaceMelting or casting

Alloy fabricationBillet preparationExtrusionRod/bar/tube fabricationStrip/plate fabrication

Beryllium metal

Beryllium metal productionPebble plantVacuum melting

Beryllium oxide productionBeryllium powder processes

Beryllium powder productionVacuum hot pressing

Beryllium special products

Other industrial processes

Ore processingCeramicsResource recoveryMaintenanceFurnace rebuilding

Administrative and professional

Research and development

Support services

Waste water treatment

606

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from

Beryllium disease surveillance

We estimated daily weighted average (DWA)exposures for a working day with formulasincorporating quarterly average general area,continuous area, and breathing zone measure-ments based on time studies for most jobs.5The DWA formulas were added or changedover the years as jobs changed and wereavailable only since 1984 for most productionjobs. In alloy areas, only jobs at the arc furnace,the melting-casting furnaces, and some strip-plate operations had DWA formulas. In beryl-lium metal areas, almost all jobs had DWA for-mulas. The DWA formulas were not availablefor jobs in quality control, engineering, decon-tamination, janitorial, production control, en-vironmental control, administration, security,waste water treatment, and the medical depart-ment. We recreated quarterly DWA estimatesfor individual jobs and job categories with theavailable formulas and industrial hygienemeasurements. If no quarterly measurementsexisted for a procedure code of a DWAformula, we used measurements from thenearest available quarter to complete theestimate. We created DWAs with existing gen-eral area and continuous area sample measure-ments for workers in alloy rod-bar-tube areas,engineering administration, offices, sales, andwaste water treatment. We could not createDWAs for quality control, decontamination,plant cleaning services, production control,environmental control, and security job titles.As these jobs involved tasks in all parts of theplant and in the administrative areas, weassumed that exposure was the average ofDWAs for all other jobs during the quarter. Weestimated cumulative beryllium exposure forstudy participants who started employmentafter 1983 by summing the quarterly DWAs fortheir job titles, weighted by days of employ-ment in the job title during the quarter,accounting for reported time away from work.The proportion of person-days for which wehad job specific DWAs was 90.5%, with theremaining 9.5% being the all plant averageDWA estimates. We calculated average beryl-lium exposure by dividing cumulative expo-sure, expressed in gglm'-days, by the totalnumber of days worked.The beryllium detection limit for all indus-

trial hygiene samples was 0.1 gg/m', but theindustrial hygiene database did not distinguishbetween samples below or at the detectionlimit. For the purpose of DWA analyses, werecoded all 0.1 measurements as 0.05 pg/m'.

CUMULATIVE INCIDENCEThe plant medical department supplied namesand dates of hire of 24 former employees diag-nosed with chronic beryllium disease beforethe cross sectional screening. The plant man-agement provided us with a count of uniqueemployees by year of hire, with which wecalculated a minimum cumulative incidence.

STATISTICAL ANALYSISThe five current employees carrying a diagno-sis of chronic beryllium disease were includedas cases in most analyses. We excluded themfrom the analyses of chest and dermatological

symptoms and workplace risk factors as theyanswered these questions after knowing theirdiagnosis. The prevalent cases were included inwork history analysis until their date of diagno-S1S.We analysed questionnaire and laboratory

results with PC SAS,' using x', Fisher's exacttest, Student's t test, and Wilcoxon's rank sumtest, as appropriate. We chose a probability of0.05 for significance. For logistic regression(PROC LOGISTIC in PC SAS), we used vari-ables from the univariate analyses that wereassociated with beryllium disease or abnormalBeLP test with a statistical significance of<0.25 and tested the assumption that the logitwas linear in the continuous variables. We usedstepwise selection with a P value criterion of<0.05 to determine which variables and possi-ble interaction terms should be included in thefinal model.Data on industrial hygiene were left cen-

sored due to the detection limit of 0.1 gg/m'and did not fit a normal or a log normal distri-bution. We used medians, ranges, and quartilesto describe these data. We compared groupswith Wilcoxon's rank sum test, x2 test, andFisher's exact tests. We compared the mean ofthe personal lapel samples for each job titlewithin a quarter to the corresponding quarterlyDWA estimates with Wilcoxon's sign rank testand Spearman's correlation coefficient.

ResultsDEMOGRAPHIC DATAOf the 655 employees invited to participate,646 (98.6%) completed the interview, the resthaving left employment, retired, or died beforeinterview. Most employees were male (85.1%)and non-hispanic white (94.7%). The meanage was 43.9 years, with a range of 24 to 61.Average employment at the plant was 17.6years, with a range from seven months to 38.3years. Employees changed work processes anaverage of 9.4 times throughout their employ-ment (range 1-50), but were employed on anaverage of only 4.4 different processes (range1-15).

BLOOD BERYLLIUM BeLP TEST RESULTS ANDCLINICAL CLASSIFICATIONExcluding the five employees with a diagnosisof beryllium disease, 627 (97.8% of thoseinterviewed) had blood drawn for the BeLPtest. Fifty people had a confirmed abnormalblood test from either one or two laboratoriesor had a combination of an abnormal bloodtest and indeterminate test (table 1). Fortyseven of the 50 underwent clinical evaluation.Twenty had granulomas on biopsy in the pres-ence of beryllium sensitisation, and four othershad abnormal bronchoalveolar lavage BeLPtests and no granulomas. Of these four people,one had a 29% lavage lymphocytosis withmultinucleated giant cells and no history ofcobalt exposure; one had 31% lymphocytesand fibrosis on biopsy; and the other two had19% and 46% savage lymphocytes. Three ofthe four had rebiopsy without obvious granulo-mas about one year after initial bronchoscopy,and the fourth had a granuloma on rebiopsy.

607

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from

Kreiss, Mroz, Zhen, Wiednemann, Barna

Table I Clinical evaluation results ofemployees with abnormal blood test results

Clinically Beryllium Sensitised to berylliumevaluated disease but without disease

Screening category n n n n

Confirmed abnormal in bothlaboratories 15 14 9 6*

Confirmed abnormal in onelaboratory 16 16 10 6

Initially abnormal with indeterminateretest 19 17 5 3

Intially abnormal with normal retests 9 0 - 4tTotal 59 47 24 19

* One of those classified as sensitised to beryllium but without disease did not undergo clinicalevaluation.t Six of nine employees were retested in the autumn of 1995, allowing four to meet the definitionof a case of sensitisation; disease has not been excluded.

The remaining 23 people did not haveberyllium disease documented by biopsy orlavage BeLP test abnormalities; however, threehad increased lavage lymphocyte proportionsof 22%, 27%, and 37%.Nine people had initial abnormal blood

tests, retested normal on two subsequent tests,and were not offered clinical evaluation. Six ofthe nine had follow up blood testing in theautumn of 1995, four of whom had abnormalBeLP test results confirming beryllium sensiti-sation.

PREVALENCE OF BERYLLIUM DISEASE,SENSITISATION, AND ABNORMAL BLOOD TESTThe prevalence of abnormal blood BeLP testswas 9.4% (59/627), excluding previously diag-nosed cases ofberyllium disease as they did nothave blood tests as part of this study. Theprevalence of beryllium sensitisation was 6.9%(43/627), again excluding the known cases ofdisease. The prevalence of beryllium disease,including previously diagnosed cases in thecurrent workforce, was 4.6% (29/632). Twelvepeople with abnormal blood tests did not haveclinical evaluation required for classification asberyllium disease.

ASSOCIATIONS BETWEEN DEMOGRAPHIC DATAAND CHRONIC BERYLLIUM DISEASEWe compared the 24 new cases of berylliumdisease and the five previously diagnosed caseswith all employees with normal blood testresults. The 29 cases with beryllium diseasewere significantly younger than others at age ofdiagnosis (39.2 v 44.0 years, P=0.005). Theydid not differ significantly from the remainderof study participants in sex, race, ethnicity,cigarette smoking, pack years of smoking, ortime since first employment. Cases of beryl-lium disease had time since first employmentranging from 1.8 to 27.5 years. No significantdifferences existed in chest or dermatological(rash, ulcer, or delayed healing of cuts)symptoms between new cases of beryllium dis-ease and those with normal blood tests. Of thefive previously diagnosed cases of disease, onereported a usual cough and phlegm, fourreported wheezing occasionally apart fromcolds, and two reported having to walk moreslowly than people of their age on the levelbecause of breathlessness. Patients with beryl-lium disease were no more likely than normalpeople to report having been in an accident

or unusual incident that may have resulted inhigh beryllium exposures (58.3% v 53.9%,P=0.67).

ASSOCIATIONS BETWEEN WORK PROCESS ANDCASES OF BERYLLIUM DISEASEWe assessed possible beryllium exposure-response relations by comparing prevalences ofwork processes in cases and normal people andby comparing disease prevalence by workinggroups. The 29 disease cases were more likelythan those with normal blood tests to haveworked in ceramic fabrication (48.3% v24.1%, P=0.004). In assessing prevalence ofdisease among employees in different workprocesses, we controlled for overestimates ofrisk by removing employees who worked inceramics from the prevalence estimates forother processes (table 2). When we looked atgrouped processes, we found that cases workedin beryllium metal production (pebble plantand vacuum melting) more than normal people(51.7% v 34.2%, P=0.05). Two cases of beryl-lium disease had worked only for purchasing oraccounting. Another had worked only in alloymaintenance.As many employees without ceramic expo-

sure had started working after ceramics pro-duction stopped in the plant, we examinedwhether the increased risk of beryllium diseaseamong ceramic employees was attributable tosecular trends in exposures or latency. Whenwe restricted the analysis to all employees whobegan working before 1984 (last date ofceramic fabrication work), a significant differ-ence remained in the prevalence of disease,9.3% (14/150) in ceramics workers comparedwith a prevalence of 3.9% among workersnever in ceramics (P=0.025). Beryllium metalproduction was not associated with signifi-cantly increased prevalence, even when ceram-ics workers were removed from the analysis.

Conversely, we looked at all employees whobegan working after ceramics work was movedfrom the plant (n= 190) and found an 8%prevalence (4/50) in beryllium metal produc-tion compared with a 0.7% prevalence (1/140)among employees in all other processes(P=0.02). Looking at the two process areas inberyllium metal production, those employeesworking in the pebble plant had a 7.3% preva-lence (3/41) compared with a 1.3% prevalence(2/148) among other employees (P=0.07), andemployees working in vacuum melting had a13.3% prevalence (2/15) compared with 1.7%(3/175) among people in all other processes(P=0.05). Six employees had worked in boththe pebble plant and vacuum melting, ofwhomone had disease. The prevalences of berylliumdisease among all pebble plant workers andvacuum melting workers hired after 1983 were6.4% and 12.5%, respectively.

Logistic regression analysis reflected theunivariate analyses. We included 12 demo-graphic and workplace variables in the creationof our model: sex, race, smoking status, packyears of smoking, time since first employment,work in another beryllium facility, work aroundberyllium fluoride, pebble plant, vacuum melt-ing, alloy extrusion, ceramics, and shipping or

608

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from

Beryllium disease surveillance

Table 2 Adjusted prevalences of beryllium disease and abnormal blood BeLP test by workprocess

Process At risk (n) Beryllium disease* Abnormal blood test*

Ceramics 155 9.0 11.6Beryllium metal production 155 5.2 14.2Pebble plant 134 5.2 13.4Vacuum melting* 21 4.8 19.0Shipping or receiving* 17 5.9 5.9Furnace rebuild 21 0 4.8Office are4 65 3.1 7.7Beryllium powder processing* 47 2.1 8.5Maintenancet 104 2.9 11.5Engineering research anddevelopment 52 1.9 7.7

Alloy metal production 69 4.3 5.8Alloy melting or casting# 65 4.6 6.2Analytical laboratory 25 4.0 20.0Alloy arc furnace* 18 0 0

* Prevalences are based on all employees at risk and do not match prevalence comparisons in thetext, which were made with employees with normal blood tests.t All employees who ever worked in ceramics which have been removed from these groups.* All employees who have ever worked in ceramics or the pebble plant have been removed fromthese groups.

Table 3 Logistic regression model of variables predictive of beryllium disease

EstimatedIndependent variables coefficients OR (95% CIs) P value

Beryllium disease model:*Intercept -3.7172 - 0.0001Ceramics history 1.4790 4.39 (1.83 to 10.50) 0.0009Pebble plant 3.1561 23.48 (4.39 to 125.52) 0.0002Years since first employment -0.0016 1.00 (0.99 to 1.05) 0.9569Pebble plant x years since first employment -0.1764 0.84 (0.75 to 0.94) 0.0025

* This model cannot be used to accurately measure risk for pebble plant employment as time sincefirst employment is a continuous variable and cannot be interpreted at the high and low end of thecontinuum. This is an unavoidable problem when a continuous covariate is modelled linearly inthe logit. This does not affect the risk estimate for ceramics.

receiving. We did not include beryllium metalproduction (the grouped process of pebbleplant and vacuum melting) to find which ofthetwo processes was contributing to the increasedrisk. We also included possible interactionterms. Four variables were included in the finalmodel (table 3): ceramics, work in the pebbleplant, time since first employment, and aninteraction term between time since firstemployment and the pebble plant. The oddsratio (OR) of beryllium disease for peopleworking in ceramics was 4.4. Work in the peb-ble plant interacted with time since first expo-sure and reflects the previous analysis in whichberyllium metal production was a risk factorsince 1984. The model predicts that pebbleplant work conferred no additional risk for aperson with employment of 17.6 years but a5.7-fold risk for someone first employed in1985.

ASSOCIATIONS BETWEEN DEMOGRAPHIC DATA,WORK PROCESSES, AND ABNORMAL BLOOD TESTSWe compared the 35 employees with abnormalblood tests but without documented berylliumdisease with those employees with normalblood tests. People with abnormal blood testswere no different in age, sex, race, smoking sta-tus, pack-years of smoking, or time since firstemployment. There were also no significantdifferences in chest or skin symptoms, althoughthose with abnormal blood tests (but withoutdisease) were more likely than those with nor-mal blood tests to report having their jobchanged due to a rash (36.0% v 13.6%,P=0.007), and were more likely to have had arash within the month before interview (28% v

8.8%, P=0.008). No difference existed be-tween the two groups in reports ofhaving beenexposed to beryllium in an accident or unusualincident.

People with abnormal blood tests were morelikely than people with normal blood tests towork in the analytical laboratory (17.1% v5.5%, P=0.02). Restricting the analysis to peo-ple employed before 1984, the association withthe analytical laboratory remained (22.7% v6.5%, P=0.02). For people employed after1983, employees with abnormal blood testsonly were more likely to have worked in thepebble plant (46.2% v 20.5% among others,P=0.04); the prevalence of abnormal bloodtests among pebble plant workers, includingcases of beryllium disease, was 19.2% v 5.8%among others.We included 16 variables and possible inter-

action terms in the creation of our logisticregression model for abnormal blood BeLPtest (but without beryllium disease) and onlywork in the analytical laboratory entered themodel with an OR of 3.6.

HISTORICAL ENVIRONMENTAL MEASUREMENTS(1984-93)Full shift and continuous area samples (n=59360) had a median (max) beryllium concentra-tion of 0.6 (1290) [ig/m3. The 30 872 generalarea samples had a median (max) of 0.4 (2615)ig/m'. The 15 787 breathing zone samples hada median (max) of 1.4 (3750) gg/m'; andexceeded 5 gg/m' in 18.5% of samples and thepeak exposure limit of 25 gg/m' in 3.6% ofsamples. The 179 personal lapel samples had amedian (range) of 1.0 (0. 1-52.6) jg/M3.Median general area measurements in differ-

ent work areas ranged from 0.1 to 0.7 jig/M3with alloy arc furnace and alloy melting-castingareas having the highest median value and thearc furnace having the highest percentage(15.0%) of measurements over the 2 jg/M3standard. Median breathing zone samplesranged from 0.1 to 2.0 gg/m', with berylliumpowder and laundry areas having the highestmedian values and furnace rebuild having thehighest percentage (28.6%) of measurementsover the short term exposure limit of 5 jg/M3.

Quarterly DWA estimates were available for18 process areas (based on our groupings in fig1) for 40 quarters in 1984-93. Some processareas had multiple jobs with DWA estimates.Quarterly job specific DWAs ranged from 0.05to 63.11 gg/m', with alloy arc furnace workersand furnace rebuild workers having the highestmedian DWA estimates at 1.65 jg/m' and 1.63jg/mi', respectively. These two job descriptionsalso had the highest percentages of DWA esti-mates over the 2 jg/m' standard, 38% and 40%respectively.

Jobs characterized with personal lapel sam-ples had one to seven measurements per quar-ter in 1990-92, with beryllium oxide produc-tion, alloy melting and casting, and the arcfurnace having the highest median lapel samplemeasurements (3.80, 1.75, and 1.75 jig/M3,respectively) and the highest percentages ofsamples over the 2.0 jg/M3 standard (64.3%,47.2%, and 44.4%, respectively). Analysis of

609

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from

Kreiss, Mroz, Zhen, Wiedemann, Barna

Table 4 Pebble plant exposures from 1984 to 1993

Median >22ig/m3 or pertinentExposure assessment method n (6g/m3) standard* (%o) Range (,uglm9)General area measurement 13296 0.4 6.4 0.1-79.2Breathing zone measurement 1456 1.1 11.0 0.1-293.3Daily weighted average 223 0.7 4.0 0.1-7.9Lapel sample (1990-2) 40 0.9 27.5 0.2-19.0

* Beryllium standard for short term exposure limit for breathing zone measurements is 5 1ggm3.average personal lapel samples and corre-sponding quarterly DWA estimates for 103pairs of samples showed that lapel samplestended to be significantly higher than the DWAestimates with a median (range) difference of-0.19 (-27.2 to 47.8) jg/m' (P=0.06, Wilcox-an's sign rank test). The paired data had a cor-relation coefficient of 0.26.

Estimates of individual cumulative berylliumexposure based on quarterly job specific DWAswere available for 201 people beginningemployment after 1983 and before 1994. Indi-vidual median (range) cumulative exposureswas 1528.1 (1.5-5971.9) gig/m3-days for theworkforce after 1983. The median (range)average beryllium exposure of the workforcewas 1.0 (0.01 to 2.66) jig/m3.

EXPOSURES ASSOCIATED WITH HIGH RISK WORK

PROCESSESAs those employed in the pebble plant after1983 had excess beryllium disease and abnor-mal blood BeLP tests, we examined exposuremeasurements and estimates for pebble plantprocesses (table 4). Samples were coded as towhether or not a respirator was used while per-forming the work task. Of all breathing zonesamples in the pebble plant, 54.4% were codedas respirator samples compared with 57.0% ofsamples in all other areas combined. Of lapelsamples in the pebble plant, 43.2% were codedas respirator samples compared with 39.4% oflapel samples in other areas.

ASSOCIATIONS BETWEEN EXPOSURE, BERYLLIUM

DISEASE, AND ABNORMAL BLOOD BeLP TESTS

Cases of beryllium disease among employeesemployed since 1984 (n=5) had a mediancumulative beryllium exposure of 1635 pg/m3-days and a median average exposure of 1.3gg/m'. They did not differ significantly in theseexposure indices from normal people (mediancumulative exposure 1518 Pg/m3-days, P=0.47and median average exposure 1.0 jig/m',P=0.27). No significant difference existedbetween those with abnormal blood tests with-out disease (n= 13) and normal people forcumulative exposure (median for employeeswith abnormal blood tests=2153 gg/m3-days,P=0.32) or average beryllium exposure (me-dian for abnormal blood samples 1.3'g/in,P=0.22). The percentages of days of work withestimated DWAs (rather than job or locationspecific DWAs) for cases of beryllium disease,employees with abnormal blood tests, and nor-mal people were 0.5%, 3.6%, and 10.5%,respectively.

50 H-

en

(aI;(00L-

.0Ec

0F-

0

040 H

30 H

20

10 _-0

.0

00

4,

0

0 LI

5001953 1959

1000 1500 20001965 1973 1984 1993

Cumulative employment/year of hire

Figure 2 Cumulative number of cases of beryllium diseaseby cumulative number ofpeople employed at the year of hire.

CUMULATIVE INCIDENCE OF BERYLLIUM DISEASEIN THE PLANTThe minimum cumulative incidence of beryl-lium disease among the 2274 people everemployed at the plant since it opened in 1953 is2.3% (53/2274), with little secular variation inrate when cases are plotted by cumulativenumber of employees at the year of hire (fig 2).

DiscussionRISK OF BERYLLIUM DISEASEThis cross sectional study of an aging work-force, with historical experience in mostsegments of the beryllium extraction andproduction industry, showed the highest preva-lences of beryllium sensitisation and disease ofany large worker cohort which has beensystematically screened.1-3 Past ceramic work-ers in this workforce had high rates ofabnormalblood tests (11.6%) and beryllium disease(9.0%) compared with other workers. Theseprevalences were about double the rates foundamong ceramics workers in the modern plantto which ceramics work was relocated (5.9%for beryllium sensitisation and 4.4% for beryl-lium disease) and resembled the rates foundamong the machinist subgroup of ceramicsworkers at highest risk in that new plant(14.3% beryllium sensitised and 10.2% withberyllium disease).2 Among employees whostarted working after ceramics operations hadbeen moved out of the plant, pebble plantworkers in beryllium metal production hadhigh prevalences of abnormal BeLP tests(19.2%), beryllium sensitisation (14.9%), andberyllium disease (6.4%) compared withothers. The high rates of abnormal blood testsonly and sensitisation only in this group with10 years maximum employment may reflect anincomplete latency period for beryllium diseasearising after sensitisation. The possibility re-mains that the risk of disease among peoplesensitized to beryllium differs by work process.

Despite our findings of excess risk ofberyllium disease for ceramics and pebbleplant workers, employees in non-productionjobs had some risk of beryllium sensitisationand disease, as in other cohorts." Employeeswho had worked only in alloy operations also

610

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from

Beryllium disease surveillance

had some risk, which is of interest for the manyworkers exposed to beryllium-copper alloy inuser industries. However, alloy operations con-fer substantially less risk of beryllium diseasethan beryllium metal production and ceramicoperations, despite higher indices of exposureto beryllium in alloy production.The time dependence of the risk ofberyllium

disease in the pebble plant is of particularinterest in the light of a major United StatesDepartment of Energy research and develop-ment effort in the early 1980s to optimiseberyllium production at this plant. Pebble plantprocesses, which were formerly out of controlfrom an environmental point ofview, includingthe fluoride furnaces, reduction furnaces, andmelts crusher, were thought to have beenbrought under control in the first half of the1980s.5 These efforts coincided with a tran-sient doubling of 1979 pebble production in1980, repeated in 1983-4. Despite consider-able improvements motivated to achieve betterenvironmental control, the risk of berylliumdisease in the pebble plant increased amongberyllium metal production workers hired after1983, a finding which has also been found forworkers hired after 1979 (data not shown).This epidemiological finding suggests thatindices used to evaluate exposure of workershave been misleading with respect to the risk toworkers of beryllium disease.

EXPOSURE-RESPONSE OBSERVATIONSIndustrial hygiene measurements, available forthe period in which the pebble plant conferredexcess risk, did not document higher indices ofexposure than lower risk processes. Thisfinding contrasts with the only other surveil-lance study with usable historical measure-ments, in which the high risk process ofceramic machining was associated with higherindustrial hygiene measurements and DWAestimates than were low risk processes.2 In themodern ceramics plant, indices of exposure inmachining were similar to the same indices forthe modern pebble plant, as were the rates ofsensitisation (14.3 v 14.9) and disease (10.2 v6.4).

Risks of beryllium disease or sensitisationrelated to work processes either reflect expo-sure factors which confer biological risk oroccur by chance alone. The degree of excessrisk of beryllium disease conferred by ceramicsand the pebble plant are unlikely to haveoccurred by chance alone. For ceramics work-ers in this plant, the prevalence of disease iscredible, as their rates of disease and sensitisa-tion were similar to the prevalences of the highrisk subgroup in the plant to which ceramicsoperations moved. For pebble plant workers,the period of excess risk of having an abnormalblood test alone coincided with the period withexcess risk of disease, making chance a lessplausible explanation for the two outcomes inthe pebble plant.The historical measurements may misclas-

sify biologically pertinent exposure in at leastthree ways. Firstly, DWAs seem to be a poorestimate of personal exposure, as lapel samplescorrelated weakly with corresponding quarterly

DWA estimates, and lapel samples tended to behigher than the DWA estimates. Secondly, par-ticle size or other characteristic may be moreimportant contributors to risk than is gravimet-ric measurement without regard to respirabil-ity. Thirdly, the methods ofberyllium exposureassessment may poorly reflect actual exposuresfrom accidents, which historically often re-quired evacuations in the pebble plant. Wefound no evidence that respirator use was asso-ciated with jobs in the pebble plant, eliminatingpersonal protective equipment as a confounderof exposure characterisation.The dissociation of high disease risk in the

pebble plant from indices of exposure could beinterpreted as evidence against an exposure-response relation. We think that the limits ofthe historical exposure data make such aconclusion premature. Rather, risk related to awork process which can be currently studied, asin the pebble plant, provides an opportunity touse different methods of exposure assessmentwhich may be more pertinent to biological risk.Comparison of exposure variables in thepebble plant, such as respirable mass, particlenumber, surface area, and solubility, can bemade with low risk processes in this and otherplants. Understanding of characteristics ofsensitizing exposures for beryllium disease maybe paradigms for other hypersensitivity dis-eases, such as occupational asthma.

MISCLASSIFICATION OF SENSITISATIONThis surveillance study had the strength ofcomplete interview data on all active employeesand nearly complete lymphocyte proliferationdata on those employees not known to haveberyllium disease. However, the study had sev-eral limitations which have resulted in likelymisclassification of beryllium disease andsensitisation status. Misclassification of theseoutcomes of interest may have arisen in labora-tory tests and clinical evaluation practices.The two laboratories conducting BeLP tests

on split blood samples produced strikinglyinconsistent results, both between laboratoriesand within laboratories on repeat samples. Ourdefinition of beryllium sensitisation requiredtwo abnormal tests separated in time, for com-parability with previously published surveil-lance studies.' However, these abnormal testresults were often accompanied by normal testsin the same or different laboratory. Of the sub-set of abnormal cases which could not be con-firmed in two split repeat tests in the first 17months of the study, most of those retested atleast 16 months later had a second abnormaltest. Cases of beryllium disease occurred in thegroup with unconfirmed abnormal blood tests,and no difference in rate of beryllium diseaseexisted among those with abnormal testsconfirmed in one laboratory compared withthose with abnormal tests in both laboratories.Thus, our definition of beryllium sensitisationseems overly restrictive for epidemiologicalpurposes, and we are likely to have missed casesof beryllium disease among those people withabnormal tests not referred for clinical evalua-tion, many of whom with time met the studycriterion for diagnostic evaluation. Certainly,

61

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from

Kreiss, Mroz, Zhen, Wiedemann, Barna

the use oftwo laboratories led to the identifica-tion of more cases of beryllium sensitisationand disease than would have occurred witheither laboratory alone, a finding that limits thecomparison of rates found for this populationwith some previously published rates.' 3 Use ofeither laboratory alone (first blood test abnor-mal and confirmation on subsequent test)would have allowed identification of only46.5% of the cases of sensitisation from onelaboratory and 48.8% from the other.

In the face of laboratory insensitivity in find-ing beryllium sensitisation, we chose to presentresults for those ever having an abnormal bloodBeLP test. Although some employees with oneabnormal BeLP test among several may havehad false positive results, the theoretical likeli-hood of a false positive is about 0.06%; this isbased on having two stimulation indices(among six in a test) above the mean plus twoSDs for the peak stimulation index of unex-posed people.

UNDERESTIMATION OF BERYLLIUM DISEASEIt is likely that this study has underestimatedrates of beryllium disease because of threepractices in clinical evaluation. Firstly, not allpeople with abnormal blood tests underwentclinical evaluation, precluding any possibility ofmaking a diagnosis of beryllium disease.Secondly, the lavage BeLP test protocol usedby the clinical laboratory for this study usedthree beryllium concentrations on harvest day4, by contrast with three concentrations onharvest days 3 and 5 used in other publishedstudies.' As well as this change in protocol,less than three concentrations of berylliumwere used in most tests, which is likely to haveresulted in lowered sensitivity for detecting anabnormal lavage BeLP test.

Thirdly, the biopsy practices in the clinicalevaluations may have resulted in underestima-tion of disease. The average number ofbiopsiestaken in those undergoing bronchoscopy wasfour, which is considerably lower than the cus-tomary 8-12 biopsies taken in previously pub-lished work.'-3 In sarcoidosis, the rate of granu-lomatous abnormalities increases withincreasing number of biopsies,'0 and thisseemed to be the case in those undergoingbronchoscopy in this study (data not shown).Although we have documented that granulo-mas can precede abnormal lavage BeLP tests,'the authors from the National Jewish Centrehave invariably found granulomas in those withabnormal lavage BeLP tests, although a secondbronchoscopy has occasionally been necessary.Rebiopsy of three cases with abnormal lavageBeLP tests but no granulomas in this study didnot document granulomas, although all threecases had other evidence of pathologicalabnormality-such as lymphocytic alveolitis orfibrosis. Only future follow up will clarifywhether we have classified them correctly ashaving beryllium disease. Their exclusion fromanalyses did not change the risk factors whichwe have presented here.A final limitation of this cross sectional study

is a healthy worker effect, which would lead toan underestimation of rates of beryllium

disease. New cases of beryllium disease identi-fied in this surveillance effort had not come toclinical attention, and only five previouslydiagnosed cases were still working in the plant.We know that some cases of sensitisation toberyllium progress to subclinical berylliumdisease,' 2 and that some of these progress withthe passage of time to be clinical cases. Theyounger age of cases of beryllium disease com-pared with people with normal blood tests isbest explained by the possibility that clinicalcases of beryllium disease from this plant haveleft employment, and the remaining subclinicalcases are unrepresentative of the age distribu-tion of the plant population.

IMPLICATIONS FOR FUTURE SURVEILLANCEEfficacy is unknown for interventions afterearly identification of beryllium sensitisationand disease-such as removal from exposure orearly treatment. In this context, the main justi-fication of screening is scientific investigation ofrisk factors and ofnatural history. Understand-ing of risk factors can lead to effective primaryprevention. Understanding of natural historyof beryllium sensitisation, perhaps in concertwith genetic risk factors,4 can lead to consid-eration of intervention trials and appropriatepolicy for secondary prevention. This surveil-lance study has laid the groundwork for bothefforts in the production sectors of theberyllium industry and many client sectors ofthe industry. The most important priority forfurther investigation is exposure-risk relations,which will serve as the basis for primaryprevention throughout the beryllium industryand may have paradigmatic importance forother common lung hypersensitivity diseases,as well.

We thank the company's administrative officers for their visionin requesting this surveillance study; the plant and corporatemedical personnel for coordinating the interview schedules,blood drawing, clinical referrals, and transfer of clinical andlaboratory data; the plant and corporate industrial hygienists forhistorical environmental data; and members of the BerylliumIndustry Scientific Advisory Committee for their thoughtfulcomments on our work. This work was supported in part byNational Institute for Environmental Health Sciences grantK07 ES00214 and by a beryllium company.

1 Kreiss K, Mroz MM, Zhen B, Martyny JW, Newman LS.The epidemiology of beryllium sensitization and disease innuclear workers. Am Rev RespirDis 1993;148:985-91.

2 Kreiss K, Mroz MM, Newman LS, Martyny JW, Zhen B.Machining risk of beryllium disease and sensitization withmedian exposures below 2 ,ug/m'. Am a Ind Med1996;30:16-25.

3 Kreiss K, Wasserman SL, Mroz MM, Newman LS.Beryllium disease screening in the ceramics industry: bloodlymphocyte test performance and exposure-disease rela-tions. J7 Occup Med 1993;35:267-74.

4 Richeldi L, Sorrentino R, Saltini C. HLA-DPBI glutamate69: a genetic marker of beryllium disease. Science1993;262:242-4.

5 National Materials Advisory Board, Commission on Engi-neering and Technical Systems, National Research Coun-cil. Beryflium metal supply options. Washington, DC:National Academy Press, NMAB-452, 1989.

6 Ferris BG. Epidemiology standardization project. AmReview Respir Dis 1978-118: 1-120.

7 Mroz MM, Kreiss K, Lezotte DC, Campbell PA, NewmanIS. Re-examination of the blood lymphocyte transforma-tion test in the diagnosis of chronic beryllium disease. JAlleg Clin Immunol 1991;88:54-60.

8 Newman LS, Kreiss K, King TE, Seay S, Campbell PA.Pathologic and immunologic alterations in early stages ofberyllium disease: re-examination of disease definition andnatural history. Am Rev Respir Dis 1989;139: 1479-86.

9 SAS. SAS stat user's guide, release 6.03. Cary, NC: SAS,1988.

10 Roethe RA, Fuller PB, Byrd RB, Hafermann DR.Transbronchoscopic lung biopsy in sarcoidosis. Chest1980;77:400-2.

612

on 24 August 2018 by guest. P

rotected by copyright.http://oem

.bmj.com

/O

ccup Environ M

ed: first published as 10.1136/oem.54.8.605 on 1 A

ugust 1997. Dow

nloaded from