Regulatory Toxicology and Phannacology 36, 262-279 (2002)doi: 10. lOO6lrtph.2002. 1588

Scientific Criteria Used for the Development of Occupational ExposureLimits for Metals and Other Mining-Related Chemicals

Lynne T. Haberl and Andrew Maier7b.ricology Excellence for Risk Assessment, Cincinnati, Ohio

Received Man:h 29, 2002

sionals with an important tool for protecting workerhealth (discussed in Paustenbach, 1990; Paustenbach,1994). The importance of OELs is highlighted by thefact that many organizations around the world developthem. However, despite the general agreement on theneed for OELs, significant disparities can exist in thespecific OEL values established by different organiza-tions and in the methods used to derive them.

The general approach used by most organizationsincludes a review of the epidemiology and toxicologyinformation to identify potential hazards, selection ofcritical effects, and dose-response estimation to de-termine suitable exposure levels. Although this over-all framework is generally shared among organizationsthat derive OELs, there is significant disparity in themethods used. Some of this diversity reflects risk man-agement decisions, such as determining the appropri-ate levels of residual risk to the exposed population orweighing health-based limits with technical feasibilityor economic impact. The approaches that are used forOEL derivation are often chosen to fit the unique needsof the constituencies involved and the mission of the or-ganization. These science policy issues are outside the

nervous system; CSAF, chemical specific adjustment factor; CSTEE,Scientific Committee for Toxicity, Ecotoxicity, and the Environment;DECOS, Dutch Expert Committee on Occupational Standards; DFG,Deutsche Forschungsgemeinschaft; ESADDI, estimated safe and ade-quate daily dietary intake; EU, European Union; FAIR, Food, Agricul-ture and Fisheries Programme; HCN, Health Council of the Nether-lands; ICNCM, International Committee on Nickel Carcinogenesisin Man; illLH, immediately dangerous to life or health; !PCS, Inter-national Programme on Chemical Safety; LOAEL, lowest observedadverse effect level; MAK, maximale arbeitsplatz-konzentrationen;MMAD, mass median aerodynaInic diameter; MOAEL, minimal 0b-served adverse effect level; NOAEL, no observed adverse effect level;NOC, not otherwise classified; NRC, National Research Council; NTP,National Toxicology Program; OEL(s), occupational exposure limit(s);OSHA, United States Occupational Safety and Health Administra-tion; PBPK, physiologically based pharmacokinetics; POD, point ofdeparture; RDA, recommended daily allowance; Rro, reference con-centration; SCOEL, Scientific Committee on Occupational ExposureLimits; TLV-TWA, threshold limit value-time weighted average;TNO, Netherlands Organization for Applied Scientific Research; UF,uncertainty factors; U.8. EPA, United States Environmental Protec-tion Agency; WHO, World Health Organization.

The scientific approaches employed by selected in-ternationally recognized organizations in developingoccupational exposure limits (OELs) for metals andother mining-related chemicals were surveyed, anddifferences and commonalties were identified. Theanalysis identified an overriding need to increasetransparency in current OEL documentation. OELdocumentation should adhere to good risk characteri-zation principles and should identify (1) the methodol-ogy used and scientific judgments made; (2) the dataused as the basis for the OEL calculation; and (3)the uncertainties and overall confidence in the OELderivation. At least within a single organization, aconsistent approach should be used to derive OELs.Opportunities for harmonization of scientific criteriawere noted, including (1) consideration of severity inidentification of the point of departure; (2) definition ofthe minimum data set; (3) approaches for interspeciesextrapolation; (4) identification of default uncertaintyfactors for developing OELs; and (5) approaches forconsideration of speciation and essentiality of metals.Potential research approaches to provide the funda-mental data needed to address each individual scien-tific criterion are described. Increased harmonizationof scientific criteria will ultimately lead to OEL deriva-tion approaches rooted in the best science and willfacilitate greater pooling of resources among organi-zations that establish OELs and improved protectionof worker health. @ 2002 Elsevier Science (USA)

Key Words: occupational exposure limit; harmoniza-tion; metals.

INTRODUCTION

Little disagreement exists that occupational expo-sure limits (OELs)2 provide health and safety profes-

1 To whom correspondence should be addressed. Toxicology Excel-

lence for Risk Assessment 1757 Chase Ave. Cincinnati, OH 45223.Fax: (513) 542-7487. E-mail: Haber@tera.org.

2 Abbreviations used: ACGIH, American Conference ofGovernmen-

tal Industrial Hygienists; AOEL, acceptable operator exposure level;BEl, biological exposure indices; BMD, benchmark dose; CNS, central

262

~), ,0273-2300'02 $35.00@ 2002 Elsevier Science (USA)All rights reserved.

SCIENTIFIC CRITERIA FOR DELe 263

efforts to define aerodynamic diameter fractions for co-ordinating OEL derivation with workplace exposure as-sessment for particles exemplifies how success in har-monizing scientific criteria for OELs can be achieved(Vincent, 1999).

As an effort to evaluate the degree to which harmo-nization of approaches is taking place and to facilitatecommunication on scientific criteria for deriving OELs,this study evaluated science-based issues related to de-riving OELs for metals, metal compounds, and othersubstances associated with mining-related industries.The currently published OELs were evaluated for asmall sample of compounds (including silica and com-pounds of chromium, copper, lead, and manganese) asa means of highlighting opportunities for greater har-monization or even standardization of approaches andto emphasize areas in which scientific criteria need fur-ther development. Metals provide a good class of com-pounds for this effort because they are common in in-dustrial processes. Metals also present some uniquechallenges, such as consideration of speciation and es-sentiality, in addition to the occupational risk assess-ment considerations faced in evaluating other classesof chemicals.

METHODS

scope of this review. However, some of the diversity inapproaches reflects issues that have not been fully eval-uated scientifically and the absence of well -documentedmethods for OEL derivation. It is in these areas thatincreased communication and transparency regardingthe scientific basis for setting OEL values can serve toenhance the usefulness ofOELs in occupational hygienepractice.

The potential value of harmonizing approaches forderivation of OELs is being increasingly recognizedthrough trends toward workforce globalization andcollaboration among standard-setting bodies. Harmo-nization encourages understanding of methods used bydifferent organizations, increasing the acceptance of as-sessments derived by these various approaches, and,when appropriate, working toward convergence of sci-entific criteria. It is important to note that the conceptof harmonization is distinct from standardization anddoes not imply that identical approaches be used amongorganizations (Sonich-Mullin, 1997). It does, however,imply clear communication of the methods used by anorganization to aid in sharing of information (with asso-ciated time and cost savings). This would mean that dif-ferences between approaches of different organizationswould reflect clearly defined scientific policy differencesor differences in scientific judgment, and be identifiableas such.

A strong argument for harmonization of OELmethodologies can be made for many reasons, as sum-marized in Table 1. In light of these considerations, it isencouraging that the field of occupational risk assess-ment is already moving toward increased harmoniza-tion. Probably the clearest example of efforts in thisdirection is the development of standard approaches forsetting OELs within Europe. The Scientific Committeeon Occupational Exposure Limits (SCOEL, 1999) haspublished its methodology for deriving OELs within theEuropean Union (EU). Another recent EU-sponsoredeffort has focused on criteria for establishing OELsfor workers exposed to pesticides (FAIR, 2000). Recent

Since OELs are established by many organizationsthroughout the world, a case study approach was usedto highlight similarities and differences, as well asthe strengths of different approaches to OEL deriva-tion used by diverse groups. The case study chemicalswere selected based on importance (production volume,potential for exposure, inherent toxicity), as well asreadily apparent differences in the current OEL valuespublished by several internationally recognized organi-zations.

After finalizing case study selection, the OEL doc-umentation for each chemical was critically analyzed.OELs established based on toxicological considera-tions rather than economic or technical feasibility(i.e., health-based OELs) were the focus of the anal-ysis. OELs that were included in the review werethe American Conference of Governmental IndustrialHygienists (ACGlli) threshold limit values (TLVs),the Dutch Expert Committee on Occupational Stan-dards (DECOS) health-based OELs (a committee ofthe Health Council of the Netherlands), and theCommission for the Investigation of Health Hazardsof Chemical Compounds in the Work Area of theDeutsche Forschungsgemeinschaft (DFG) maximaleArbeitsplatz-Konzentrationen (MAKs). Note that somehealth-based OELs (e.g., the MAKs and the values setby the DECOS) may be further modified based on tech-nical or economic feasibility considerations in establish-ing final published limits. These organizations were se-lected as examples of OEL-setting bodies that receive a

TABLE 1Potential Benefits of Harmonization in Setting OELs

Increase the transparency of health-based OELs, making clear tothe occupational health practitioner the uses and limitations ofa particular OEL.

Enhance the degree of confidence in the process used to derive theOELs by emphasizing the communication of key scientific criteriathat are used to derive them.

Facilitate the pooling of resources among bodies that set OELs toincrease the coverage of OELs for the thousands of substancesfor which no guidance exists, and to decrease the time it takes toupdate OELs as new data become available.

Increase the provision for similar levels of worker health protectionin different parts of the world, by increasing the consistencyin the scientific criteria used as the basis for deriving OELs.

264 HABER AND MAIER

large degree of international attention and that estab-lish OELs based on health rather than risk manage-ment considerations. A second reason for the choice ofstudy organizations was that each group has publishedOELs for some, ifnot all, of the case study chemicals. Al-though the primary purpose in this study was to reviewa small sampling of compounds and organizations tohighlight opportunities for harmonization in scientificcriteria for OEL development, methods for a numberof other organizations in the analysis of the scientificissues involved are also cited.

TABLE 2Occupational Exposure Values for Case

Study Chemicals"

DECOS MAKCompound Acorn

ND"0.060.06CaCa

CaeCa

0.50.5_4

0.050.01

CaCa

1(1'10.1 (R)f

10.2

RESULTS0.05

0.2

O.3(R)

4(1)10 (1); 3 (R)

2(R)0.1 (R)

1010

0.3 (R)0.3 (R)

4 (I)4ffi4ffi

0.3 (R)CaCaCa

0.1 (R)'0.05 (R)0.05 (R)0.05 (R)0.1 (R)

-0.0750.0750.075

Case Study Summaries

Table 2 provides the current OEL values for thefive case study chemicals for the ACGIH, DECOS, andthe DFG (MAK values); Table 3 provides the basis forthe OELs for metals for the case study chemicals forwhich English-language documentation was availablefrom more than one organization. Because the valenceof the chromium (Cr) ion is important in determin-ing chromium biology and toxicology, the OELs forchromium differ greatly according to chemical specia-tion. OELs have been developed for Cr metal, Cr (Ill),Cr (IV), and Cr M) species. The TLV- TW A for Cr metaldust was adopted in 1981 (as described in ACGIH,1996). In a reevaluation in 1996, ACGIH concluded thatit was appropriate to retain the current value, since ex-posure over the previous 10 years did not yield reports ofadverse health effects (ACGIH, 1996). In support of thisdecision, the chosen value was supported by a NOAELof 3.1 mg/m3 (the highest exposure level) in a 4-weekinhalation study in rabbits (Johansson et aI., 1980). DE-COg reviewed the data for Cr metal and concluded thatthe data were not sufficient to derive a health-basedOEL, but did not disagree with the continued use of the0.5 mg/m3 limit in force at the time of the reevaluation(DECOS, 1998a). No MAK value has been establishedfor Cr metal, based on the assigned cancer classificationfor chromium and compounds. DFG's policy is not to de-velop MAR values for nonthreshold carcinogens (DFG,2001).

The TLV-TWA for Cr (III) was based on the absence ofchanges of pathophysiological significance at exposuresto Cr (Ill) salts below 0.5 mg/m3 in epidemiology or an-imal studies, and the lack of an association betweenCr (Ill) salts and cancer (ACGIH, 1996). Therefo~, theprevious TLV-TWA value was retained (ACGIH, 1996).The DECOS (1998a) relied on similar studies, but usedtwo animal studies (Johansson et ai., 1986a, 1986b)to quantitatively derive the OEL. These two reportsof the same subacute inhalation study in rabbits ex-posed to soluble Cr (III) compounds noted abnormalmacrophages (oblong, smooth) in alveoli at 0.6 mg/m3.This exposure concentration was selected as a MOAEL

3 The values of 3 shown here actually represent the square root of10 (3.16, or 1/2 of 10 on a log scale), rounded to one significant figure.This leads to the confusing result that 3 x 3 = 10

ChromiumMetalCr(lII)Cr(IV)Cr (VI) (soluble)Cr (VI) (insoluble)

CopperInorganic dustFume

LeadAnd its inorganic compounds

ManganeseInorganic

SilicaAmorphous diatomaceous

earth calcinedAmorphous diatomaceous

earth uncalcinedAmorphous fumeAmorphous fusedAmorphous gelAmorphous precipitatedAmorphous pyrogenicAmorphous quartz glassCrystalline CristobaliteCrystalline QuartzCrystalline TridymiteCrystalline Tripoli, as quartz

a All values are in mg/m3.b ND indicates that the data were evaluated but no OEL was de-

rived.. Ca indicates that no OEL was established based on the cancer

classification.d -indicates that no documentation exists.e (1) indicates that the OEL is based on the inhalable fraction.(R) indicates that the OEL is based on the respirable

fraction.6The ACGIH TLV-TWA includes quartz glass under silica,

amorphous-fused.

(minimal observed adverse effect level). To derive theOEL for soluble Cr (Ill), the MOAEL was adjusted by afactor of3 for extrapolation from a MOAEL to a NOAEL(no observed adverse effect level), and by a factor of 3for interspecies extrapolation, resulting in a compositeuncertainty factor of 10.3 For insoluble Cr (Ill) com-pounds, the data were regarded as insufficient to derivean OEL, but DECOS did not disagree with the contin-ued use of the 0.5 mg/m3limit in force at the time of thereevaluation (DECOS, 1998a). Based on the assignedcancer classification for chromium and chromium com-pounds, no MAK value has been established for Cr(Ill). Thus, in contrast to ACGIH and DECOS, which

.- -

'.

265SCIENTIFIC CRITERIA FOR OEI..

TABLE 3Derivation of OEL for Case Study Meta18-

POD OELOEL ba8is mg/m3 mg/m3 ur

Compound.organization

NS"Cr metal, AOOUI 0.5

Cr (ill). ACGm NS 0.5 Ie

101'Cr (m). DECOS 0.6 0.06

lfyl

IC0.5NS

0.050.05

NS0.4NA

0.010.20.06

288

61:

Mn. ACGIH

Mn,MAK1

0.250.20.5 <I

differences resulted in a composite uncertainty factorof 10.

The approaches used for developing OELs for hex-avalent chromium compounds also differed substan-tially among the three organizations. ACGIH (1996)derived different TLV-TWA values for soluble and in-soluble forms ofCr M), not otherwise classified (NOC).Separate TLV values have also been developed for anumber of specific chromates. The 1996 ACGIH docu-mentation for water soluble Cr M) (NOC) noted thatthe "TLV for soluble chromates will be maintained atcurrent value of 0.05 mg/m3 until an in-depth review iscompleted by the committee." Cr M) (NOC) was clas-sified as a confinned human carcinogen (A1). No othersupporting data were provided. For insoluble Cr M)compounds, the same TLV-TWA was recommended asfor zinc chromate until more dose-response informationbecomes available. DECOS (199&) considered Cr (VI)compounds carcinogenic to humans and assumed agenotoxic mode of action. Consistent with its methodol-ogy, DECOS (199&) calculated cancer risk estimates,but no health-based OFJ, ~~nded. Based onan evaluation of the carcin.nicity ofCr (VI), no MAKvalues have been derived for hexavalent chromium com-pounds (DFG, 1992). The OELs for Cr M) highlighta clear difference in approaches used by different or-ganizations for setting OELs for compounds of knowncarcinogenic potential, even when similar conclusionsregarding the carcinogenic potential of a chemical arereached (see further discussion of this issue in Seeleyet- al., 2001).

In summary, ACGIH, DECOS, and DFG consid-ered a similar array of data, but used widely vary-ing approaches to develop markedly different OELs forchromium and chromium compounds.

The TLV-TWA for copper was derived to protect fromirritation and systemic effects, and is based largelyon epidemiological data (ACGIH, 1991). Based on thechronology presented in the 1991 documentation, areevaluation in 1973 resulted in ACGIH increasing theTLV-TWA for copper fume from 0.1 to 0.2 mgtm3, basedon a letter reporting that employees exposed to cop-per fumes at levels up to 0.4 mglm3 experienced no illeffects (Luxon, 1972) and supported by other epidemi-ology studies. The ACGIH documentation did not pro-vide any other information on the exposure conditions,cohort size, or endpoints evaluated. Considering thisletter to identify a human NOAEL of 0.4 mg/m3, the cur-rent TLV provides a margin of exposure of 2 from a hu-man NOAEL for copper fume. The TLV for copper dustsand mists is 1 mgim3, a value that the documentationconsidered .should provide similar protection againstadverse health effects." The strength of the databaseand the rationale for the copper dust/mist TLV-TWA isless clear than for copper fume. Irritation is the criticalendpoint for copper dust, but details were not providedon the degree of irritation observed at specific exposure

did not consider Cr (III) a human carcinogen, the DFG(1992, 2001) based the MAK for Cr (ill) on the valuefor Cr (VI) (a Category 2 carcinogen, considered to becarcinogenic in humans based on animal studies sub-stantiated by epidemiological studies). Comparison ofthe documentation for the OELs derived by these threeorganizations does not specifically indicate the reasonfor the apparent discrepancy in the carcinogenicity as-sessments. However, it appears that ACGIH (1996) andDECOS (1998a) weighed the negative findings in epi-delniology studies more heavily, while the DFG (1992)may have considered the epidemiology data less con-clusive, after taking into account positive findings inintratracheal instillation studies with Cr (ill).

The DECOS (1998a) is the only one of the three or-ganizations studied that developed an OEL for Cr (IV).The OEL was based on a LOAEL (lowest observed ad.verse effect level) of 0.5 mg/m3 for effects on the lungin a chronic study in rats (Lee et aI., 1989). An uncer-tainty factor of 3 for extrapolation from a WAEL to aNOAEL and an uncertainty factor of 3 for interspecies

HABER AND MAIER266

identical conditions with the equipment generally usedin the United States and Sweden. (See also Vincent,1999, for a discussion of this issue.) So, 0.25 mg/m3,measured in Sweden in the study by Iregren (1990),is equivalent to 0.5 mg/m3 in total dust measured inGermany. Therefore, a MAK value of 0.5 mg/m3 was rec-ommended. Since the effects at this level were minimal,did not develop in all exposed persons, and were notdose-dependent, this concentration was used as a basisfor the MAK, without application of uncertainty factors.Manganese is currently on the working program forDECOS (personal communication from DECOS, 2001).The scientific bases for the manganese OELs of the dif-ferent organizations were very different, although theresulting values of the OEL do not vary greatly.

In addition to metal compounds, a variety of othersubstances are often associated with industrial pro-cesses related to metal mining, manufacturing, or pro-cessing. Differences in the derivation for OELs for non-metal compounds were also noted; the OEL values forsilica were used as a case study. All three organizationsincluded in this study differentiate among a variety offorms of silica on the basis of the degree to which amor-phous versus crystalline silica is present.

The ACGIH has derived TLV-TWA values for mul-tiple types of amorphous silica (ACGIH, 1991, 1996).The TLV-TWAs for uncalcined amorphous diatoma-ceous and amorphous precipitated silica were based onthe default values for inert dusts, presumably due to theapparent low intrinsic toxicity of these forms of silica.For calcined diatomaceous earth (which is formed in aprocess that generates crystalline silica), ACGIH rec-ommended that exposure be gauged against exposurelimits for the three crystalline forms of silica, reflect-ing greater concern for crystalline forms. A TLV-TWAhas also been established for amorphous silica fume.This value was based on studies that found abnormalX-ray findings in workers exposed to high (but unmea-sured) concentrations of silica fume. The single studythat included exposure measurements did not providesufficient data to identify a no-effect level. In this study,Corsi and Piazza (1970) reported that less than 1% ofworkers exposed for less than 5 years, and 14% of work-ers exposed for 10 or more years, had abnormal X-rayfindings. Based on silica content, respirable silica ex-posure in this study was estimated as 1.6 mg/m3. ATLV-TWA value of 2 mg/m3 was recommended for sil-ica fume in accordance with recommendations of sev-eral independent groups that had evaluated the samedata. For fused amorphous silica, few data exist ex-cept for several intratracheal instillation studies, whichdemonstrate that fused silica can induce effects similarto those seen with crystalline silica (as described below),although to a lesser degree. Due to evidence that fusedsilica can induce fibrosis, a TLV of 0.1 mg/m3 was rec-ommended. Precipitated silica (amorphous gel) did notincrease fibrosis in chronic inhalation studies in several

levels. Copper is currently listed on ACGIH's "Chemi-cal Substances Under Study" list (ACGIH, 2001). TheMAK for copper and compounds is 1 mg/m3 inhalabledust. According to the DFG methodology (DFG, 2001)the inhalable fraction is to be measured when the MAKis based on total dust. The MAK for copper fume is0.1 mg/m3 respirable. (The respirable fraction is to bemeasured when the M;AK value is based on adverse ef-fects associated with the level of fine dust.) No pub-lished documentation in English on the MAK value wasavailable for review. Copper is currently on the work-ing program for DECOS (personal communication fromDECOS, 2001).

For lead, only the ACGIH TLV-TWA documentationwas available for review (ACGIH, 1996). The TLV-TWAwas back-calculated from a blood lead level of9.5 JLg/dL.This blood lead concentration was selected to protectagainst developmental effects such as depressed intel-lectual development in children exposed to lead duringgestation. As discussed in the ACGIH documentation,the threshold for this effect remains uncertain. ACGIHconcluded that blood lead levels of 30 JLg/dL would notaffect a woman's ability to bear children who wouldsubsequently develop normally. However it was notedthat the TLV committee would reevaluate the value of30 JLg/dL as new data become available. Based on theratio of the calculated blood lead concentration result-ing from exposure at the level of the TLV (9.5 JLg/dL)and the biological exposure index (BEl) of30 JLg/dL, theTLV affords an approximate margin of safety of 3-fold.The lower blood lead level associated with the TLV ascompared to the BEl allows for other nonoccupationalexposures to occur without increasing blood lead levelsabove 30 JLg/dL. Although a MAK value has been de-veloped for lead (DFG, 2001), no English version of thedocumentation of this value was available for review.Lead is currently on the working program for DECOS(personal communication from DECOS, 2001).

The TLV-TWA (ACGIH, 1996) for elemental and in-organic manganese compounds is based on human epi-demiological studies that indicate the potential forlung and central nervous system effects (Roels et ai.,1987), and male reproductive effects at around 1 mg/m3(Lauwerys et ai., 1985). A TLV-TWA of 0.2 mg/m3 wasrecommended from these studies, since the thresholdfor effects on the CNS and the lungs is not known defini-tively and due to the progressive nature of these ef-fects. Manganese is currently listed on the ACGIH's"Chemical Substances Under Study" list. The MAKvalue (DFG, 1999) was based on the lowest averagemanganese concentration (approximately 0.25 mg/m3)shown to cause slight neurotoxic symptoms in ex-posed workers (lregren, 1990). The MAK documenta-tion (DFG, 1999) noted that, at the time of the Iregren(1990) study, the sampling equipment typically usedin Germany led to concentration measurements ap-proximately twice those that would be measured under

SCIENTIFIC CRITERIA FOR OELs 267

at the time that the TLV for tripoli was established(ACGIH, 1991). This TLV-TWA has not yet been up-dated to reflect the changes in the quartz TLV-TWA.It is noteworthy that cristobalite, tridymite, and tripoliare listed on the "Chemicals Under Study List" (ACGIH,2001), although they have not yet appeared on the no-tice of intended changes.

The DECOS developed a health-based OEL for thecrystalline silica polymorph, quartz (DECOS, 1992). Ina recent review of the carcinogenicity of crystallinesilica (DECOS, 1998b), it was concluded that "crys-talline silica inhaled in the form of quartz or cristobalitefrom occupational sources is carcinogenic to humans(Group 1)." The most well-supported mode of action fortumorigenic responses observed in animal studies wasconsidered to be epithelial proliferation associated withpersistent inflammation; however, direct genotoxicityor genotoxicity due to generation of reactive oxidantswas not ruled out. Based on this analysis, it seems likelythat no OEL would be developed if the DECOS were tore-evaluate crystalline silica in light of the 1998 review,since a potential genotoxic mode of action has not beenruled out. Based on its carcinogenic potential, no MAKvalues have been derived for the crystalline forms ofsilica (DFG, 2001).

These case studies demonstrate that even for "health-based" OELs, which presumably are not dependent onrisk management considerations, there are large dif-ferences in both the OEL values and the basis for thosevalues. A similar observation was made in a recent re-view of European OELs for a variety of carcinogeniccompounds (Seeley et al., 2001). Based on the case stud-ies, a central commonality and a number of differencesamong the approaches used by the different organiza-tions were identified. All of the organizations developedOELs by first conducting a critical evaluation of theexposure, epidemiology, and toxicology data and deter-mining the effects associated with the exposure of in-terest. All of the organizations also identified a criticaleffect level (also known as a point of departure, or POD)to the extent feasible based on the data and then ex-trapolated from that value to the OEL. Sometimes thesupporting documentation explicitly identified the PODand the study in which it was identified, while otherorganizations presented the supporting data, but useda significant amount of professional judgment in deter-mining the POD. In some of the latter cases, the docu-mentation was sufficiently complex that even an expe-rienced toxicologist-risk assessor found it challengingto identify the quantitative basis for the OEL. The or-ganizations also differed markedly in their approach toextrapolating from the POD to the OEL. Some organi-zations (e.g., DECOS for Cr (Ill)) applied specific un-certainty factors to perform this extrapolation, whileothers did not explicitly state the reason for the magni-tude of the difference between the POD and the recom-mended OEL. There was also a clear difference in the

268 HABER AND MAIER

approach used for carcinogenic compounds (see Seeleyet ai., 2001, for a more in-depth discussion of this area).The rest of this paper considers in detail these com-monalties and differences in the approaches used forthe development of OELs.

Database Considerations

When available, well-documented human studies ofthe appropriate duration will nearly always be pre-ferred for OEL derivation. This is likely to be partic-ularly true for metals and other substances used in themetal or mining-related industries, since long humanexperience with these industrial processes increasesthe likelihood for tracking of effects in exposed work-ers. For the case studies reviewed here, epidemiologydata played a key role in deriving the OELs for copper,lead, manganese, and crystalline silica. Occupationalepidemiology data can be particularly useful for the de-velopment of OELs, because the duration and condi-tions of exposure and the study population are similarto that for the population of interest.

Human data can be obtained from epidemiology stud-ies, case studies, or, in some cases, ethically conductedclinical exposures. Peer-reviewed studies are preferred.If a company report that has not been peer reviewedis considered, evidence that the data have been vali-dated (e.g., a quality assurance audit statement) is de-sirable. There are no hard-and-fast rules for determin-ing whether data are of adequate quality to include inOEL development. Considerable professional judgmentapplies, although higher standards would tend to be ap-plied to a study used as the basis for an OEL than astudy used as supporting data. It would be desirableto apply higher standards to studies used to derive in-ternationally recognized OELs, while a screening levelor preliminary OEL might require less rigorous docu-mentation. Guides have been developed to aid in thecollection of epidemiology data related to metals andtheir species (e.g., ICME, 1999; NiPERA, 2001).

Although epidemiology data can provide useful in-formation, these data need to be evaluated carefully todetermine whether the data are sufficient to establishcausality. Criteria for judging the adequacy of epidemio-logical studies are well recognized (Hill, 1965; U.S. EPA,1994; WHO, 2000). All of these criteria need to be evalu-ated in considering whether a chemical caused an effect;no one factor alone can definitively show causality, andthe reader is referred to the cited documents for a moredetailed discussion.

An issue related to database considerations that isoften faced in setting OELs is how to use so-called grayliterature, consisting of unpublished data. For example,the TLV-TWA value for copper fume was based in parton a letter reporting no adverse effects at exposures upto 0.4 mglm3, and ACGIH maintained the Cr (III) TLV-TWAin its 1996 review based on the absence of reported

TABLE 4ConsIderations In Evaluating Epidemiology Data"

Elements of a Well-Reported StudyA description of the objectives and research designA description of selection procedures for the study population and

comparison group. population characteristics, sources of the exposuredata, and the methods of data collection

A discussion of major confounding factors and data analysismethods

The criteria used for interpreting results

II Adapted and expanded from US. EPA. 1994 and WHO. 2000.

adverse effects at the previous TLV-TWA DevelopingOELs based on this type of data has the advantage ofmaking maximal use of the available human data, butit has the disadvantage of lack of transparency. Otherscientists reviewing the documentation cannot inde-pendently evaluate the conclusions, because minimalor no information is provided on the number of peopleexposed, endpoints evaluated, and exposure conditions(length of time exposed, type of monitoring done, rangeof exposure levels). At a minimum, organizations de-riving OELs and using such gray literature should pro-vide basic documentation of these elements and shouldweigh the quality of the data against standard criteria(see Table 4) when determining the confidence they willput in that data.

Gray literature, if it is of sufficient quality, can bea useful data source that is currently being used onlyminimally. If companies are already doing health mon-itoring of employees and exposure monitoring, much ofthe data for improving OEL development may alreadyexist in an unanalyzed form. Presumably the two majorimpediments to publishing such data are the cost of athorough statistical analysis as part of an epidemiologystudy and the lingering reluctance of some journals topublish negative studies. This suggests that it would beuseful to work with industrial hygiene and risk assess-ment journals in order to facilitate publication of in-dustrial hygiene reports in which no effects were noted.The goal would be to ensure that more of the sourcedata reach the open literature. Web-based publicationof data without statistical analyses may also be a meansfor getting the data into the open literature at a low costto the companies collecting the data. The public avail-ability of such data would increase the credibility ofOELs that are based on minimally documented indus-try reports concluding that experience with a certainexposure level indicates that exposure to this concen-tration is safe.

In the absence of adequate human data, the choiceof the best study(s) to use as the basis for an OELbecomes less clear. OELs are routinely established forsubstances that vary greatly in the strength of the un-derlying database, which may range from strong epi-demiology studies supported by animal studies to only

- I,. ,~ . . ~ "~~- ,-

~ ~-.,- -'-~-~~"c-::-'

SCIENTIFIC CRITERIA FOR OElA 269

for adequacy of the database or the use of confidenceratings (e.g., Dourson et al., 1992; U.S. EPA. 1994). Con-fidence ratings provide a means for characterizing thestrengths of the key study(s), as well as the strengthsand weaknesses of the entire database, and can be sup-plemented with a narrative describing such strengthsand weaknesses in more detail. Feron et ai. (1994), ina S11mmAry of OEL methods in the Netherlands, notedthat organizations should clearly state when the dataare inadequate to derive an OEL. The use of an un-certainty factor approach has been incorporated intosome of the proposed OEL methods and has been stan-dard practice in human health risk assessment for envi-ronmental exposures (reviewed in Dourson et al., 1996;Haber et al., 2001) (described in more detail below inthe Uncertainty Factor Section).

Point of Departure

As noted above, all of the OEL approaches evalu-ated the overall database in order to determine a PODfor development of the OEL. For this analysis we re-fer to the POD as the concentration to which uncer-tainty factors (explicit or implicit) are applied to derivethe OEL. None of the case study organizations definedthe appropriate basis for this POD, but this issue wasaddressed by SCOEL (1999), in tlle European Uniondocumentation for acceptable operator exposure levels(AOELs) (FAIR, 2000), and by HCN (2000). These lat-ter groups prefer to base OELs on a NOAEL. When aNOAEL is not available, a LOAEL can be used, takinginto account the additional extrapolation needed. TheU.S. EPA's IRIS defines a NOAEL as "the highest expo-sure level at which there are no statistically or biolog-ically significant increases in the frequency or severityof adverse effect between the exposed population andits appropriate control" (U.S. EPA, 2001). The Interna-tional Programme on Chemical Safety (IPCS, 1999) de-fines the NOAEL based on there being "no detectableadverse alteration of morphology, functional capacity,growth, development or life span of the target." Notethat the NOAEL can be defined by changes that are sta-tistically or biologically significant, and that increasesin frequency and severity of an effect are considered.One definition of adversity, used in the environmentalcontext, is available from the U.S. EPA (2001): "A bio-chemical change, functional impairment, or pathologiclesion that affects the performance of the whole or-ganism, or reduces an organism's ability to respond toan additional environmental challenge." In the occu-pational arena, the SCOEL (1999) developed a sever-ity scale for evaluating systemic and irritant effects(Table 5). Using this scale, the SCOEL defined adversityas beginning at concentrations that induce effects be-tween severity levels 2 and 3. While other approachesmight differ in some details, the approach developedby SCOEL is an important step toward enhancing

a few case reports or a single animal study. Manyorganizations do not define a minimal database for es-tablishing a risk value. Rather, a description of thetypes of data and their potential utility is more com-monly provided. Similarly, most organizations do notdefine a rigid hierarchy for selecting the type of study touse in determining the POD. Rather, it is more commonfor organizations to use a weight of evidence approachthat looks at all of the data together (e.g., SCOEL, 1999).In some settings, a clear data hierarchy provides thebenefit of enhancing consistency and transparency inthe OEL derivation process. For example, the method-ology for setting mLH (immediately dangerous to lifeor health) values (NIOSH, 2001) provides an exampleof a clearly defined data hierarchy system. On the otherhand, this level of rigidity may not be necessary as longas the OEL documentation clearly describes the extentof the database and provides a well-reasoned rationalefor selecting a particular subset of the data as the basisfor the OEL. Some organizations do have a minimumdatabase requirement. For example, the method usedby the Committee on Updating Occupational ExposureLimits of the Health Council of the Netherlands <HCN,2000) outlines minimum data requirements. Accordingto this method, data on acute toxicity and repeated-dosetoxicity are required. As a minimum, a multidose studyin a relevant species using a relevant route of admin-istration and evaluating an array of endpoints shouldbe available. The U.S. EPA methodology (1994) definesthe minimum database for deriving an inhalation refer-ence concentration (Rf(:) as an adequate human studyor a well-conducted subchronic animal inhalation studythat includes evaluation of the respiratory tract.

The rationale for a minimum data requirement isthat any value derived from fewer data would be toouncertain to provide meaningful guidance. On the otherhand, the absence of any OEL, even one based onlimited data, provides a severe handicap to the occu-pational health practitioner who often must make adecision in the field. It could be argued that the OELdocumentation itself provides the means to assess thedegree of confidence one should place on an OEL. How-ever, current approaches to OEL documentation mightbenefit from application of more direct strategies forcommunicating the strength of the database to the oc-cupational health practitioner. For example, the ex-tent of the database supporting the OELs for the fivecase study compounds reviewed varied greatly fromfairly robust epidemiology data for inorganic lead, man-ganese, and crystalline silica (as quartz) to limited hu-man and animal data for noncarcinogenic chromiumcompounds and amorphous silica. Yet none of the OELvalues provided explicit information on the degree ofuncertainty in the OEL value resulting from tlle in-ability of the available studies to detect potential ad-verse health effects. Possibilities for addressing this is-Rue include the use of well-defined uncertainty factors

270 HABER AND MAIER

TABLE 5Adversity Definitions Used by SCOEL"

Irritant effectsSystemic effects

1. No effects observed 1. No effects observed; no aware-ness of exposure

2. Very slight effects; awareness ofexposure

2. Compensatory effect orearly effect of dubious sig-nificance without adversehealth consequences

3. Early health impainnent(clear adverse effects)

4. Overt disease, possiblydeath

3. Slight irritant effects or nuisance(e.g. smell); easily tolerable

4. Significant irritation/nuisance,overt health effects; barelytolerable

5. Serious health effects (e.g. pul-monary edema); intolerable

0 Adapted from SCOEL (1999).

transparency and consistency by defining the effectsconsidered adverse for occupational risk assessment.

A relatively recent innovation in risk assessment fordetermining the POD is the use of the benchmark dose(BMD) approach, which fits a flexible mathematicalcurve to the experimental data to determine a dose cor-responding to a predetermined response level (Crump,1984; Dourson et al., 1985). A significant advantageof the BMD approach is that a BMD can be deter-mined even if an experimental study did not identifya NOAEL. In addition, the POD determined using thisapproach is not confined to the tested dose levels. Themathematical modeling approaches used are similar tothose used throughout the world for cancer modeling.Although none of the organizations included in the casestudies we evaluated specifically mentioned use of theBMD approach, recent guidelines for AOELs for pes-ticides (FAIR, 2000) do refer to this method. This sortof modeling approach was also used by the U.S. Occu-pational Safety and Health Administration (OSHA) inevaluating the risk of kidney dysfunction as part of thedevelopment of its cadmium standard (OSHA, 1993).BMD modeling may be particularly valuable for theevaluation of epidemiology data in the development ofOELs, because it allows the use of individual exposuredata, rather than "binning" of exposure levels. Unlikeanimal studies in which an exposure group representsanimals all exposed to the same concentration of chem-ical, each worker typically has a slightly different ex-posure history, resulting in a continuous range of expo-sures. Traditional risk assessment addresses this issueby grouping workers with similar exposure levels (e.g.,0-200,201-400,401-600, and 601-800 ppm/years) andusing the midpoint of the ranges to determine NOAELsand LOAELs. A problem with this approach, how-ever, is that the NOAELs and LOAELs are an artifactof the grouping, and different values could be ob-tained by different grouping (e.g., 0-300, 301~00, 601-

900 ppm/years). Conducting dose-response modeling ofthe individual data removes such artifacts by consider-ing each individual's exposure and response.

Although many OELs for metals have been devel-oped from epidemiology data, some (e.g., the OELsfor Cr (III» are based on a POD derived from animaldata. In such cases, it is necessary to extrapolate expo-sure levels from animals to humans. Ideally, such ex-trapolation takes into account interspecies differencesin the breathing rate and respiratory tract structureof experimental animals and humans. Several differ-ent approaches can be used for conducting such ex-trapolations. All of the case study organizations usedthe animal exposure level directly in developing theOEL. The AOEL methodology (FAIR, 2000) states thatextrapolation from an animal exposure concentrationdirectly to a human exposure concentration uses an im-plicit allometric scaling based on bwO' 75. This is becausealveolar ventilation rate scales across species using the0.75 power of body weight. Allometric scaling assumesthat the parent compound is the form responsible fora chemical's toxicity and that biotransformation detox-ifies the chemical rather than activating it. Allometricscaling may therefore be a reasonable initial approachfor metals, which generally are not metabolized andfor which clearance is the primary means of removalfrom the body, and thus a key determinant of tissuedose.

However, there are several other factors that do notscale directly with body weight, but which also affectthe tissue dose from inhalation exposure. The lung de-position of inhaled particles (and therefore the dose tothe lung) depends on the particle size and on the respi-ratory tract structure. Differences in particle size canbe important if the particle size in the animal study ismuch smaller than that found under occupational con-ditions. For example, the NTP (1996a, 1996b, 1996c)studies of nickel carcinogenesis used particle size dis-tributions with mass median aerodynamic diameters(MMADs) of 2-2.5 ILm. By contrast, the limited dataon occupational exposures to nickel compounds in re-fineries indicate that less that 10% of the mass is res-pirable, and the MMAD for those particles is on theorder of 4 ILm (discussed in greater detail in Haberet aI., 2000). This means that, for a given concentrationin air, a much larger dose was delivered to the lungin the animal study than would be delivered under oc-cupational exposure conditions. Differences in particlesize are addressed to some degree by differentiatingbetween OELs for respirable and inhalable fractions,as was done for the MAK OELs for copper, lead, man-ganese, and silica, and the ACGIH values for silica (seeTable 2). However, the precision of such an approachis low. The low precision (and potential lack of trans-parency) is evident in the development of the MAK OELfor manganese. Because the sampling equipment usedin Germany yields concentration measurements twice

SCIENTIFIC CRITERIA FOR OELs 271

oral exposure. Therefore, for developing an GEL fromoral data, sufficient inhalation data must be availableto show that the critical effect does not occur in the res-piratory tract. Route-to-route extrapolation can be ap-propriately performed in situations where the criticaleffect is a systemic one, such as neurological effects orkidney effects. Biological exposure indices can be par-ticularly useful for doing route-to-route extrapolation,since they allow exposure from different routes to beexpressed in consistent units. For example, the ACGIHTLV-TWA for lead is based on blood lead levels, andthe acceptable concentration was determined in stud-ies where both inhalation and oral exposure may haveoccurred. Use of the biological exposure index allows theintegration of total dose from all sources. Thus, biologi-cal indices are useful for some chemicals as a measureof internal doses to workers. As a result, organizationssuch as the ACGlli (2001) and DFG (2001) derive lev-els for assessing internal exposures to some chemicals.For example, the ACGlli (2001) has recommended bio-logical exposure indices (BEls) for water-soluble Cr (VI)and for lead, as well as for a variety of other metals.

those measured at identical workplaces with the equip-ment generally used in Sweden (where the study thatis the basis for the POD was conducted), the MAK OELis twice the minimal severity LOAEL used as the POD.Thus, the OEL was derived using an uncertainty factorof 1, even though it appears at first glance that a factorof 0.5 was used.

The structure of the respiratory tract is also impor-tant in determining interspecies differences in the res-piratory tract tissue dose. For example, the rat nasalpassages are much more convoluted than those of hu-mans, and rats are obligate nose-breathers, resulting inhigher upper respiratory tract deposition in rats thanin humans. The U.S. EPA (1994) has developed a par-ticle dosimetry model to aid in extrapolation from ani-mals to humans. The implications of the differences inparticle deposition in animals and humans can be ob-served using the Johansson et al. (1986a, 1986b) stud-ies that were the basis for the Cr (III) OELs developedby DECOS. DECOS used the exposure concentrationat the MOAEL of 0.6 mg/m3 as the POD, but dosimet-ric considerations conducted according to the U.S. EPAmethodology result in a MOAEL (human equivalentconcentration) of 0.07 mg/m3 for this effect on the pul-monary region of the lung. Details of this calculationare shown in Appendix A. The U.S. EPA (1994) method-ology also provides methods for calculating the humanequivalent concentration for gases, taking into accountwhether the chemical acts at the portal of entry, sys-temically, or both.

Other scientists have developed more sophisticatedmethods for extrapolating from animals to humans,although they are more time-intensive and require spe-cialized expertise. For example, the U.S. EPA (1994)particle dosimetry model cannot account for differencesin clearance from the lung, but Oberdorster and col-leagues (e.g., Hsieh et al., 1999; Yuet al., 2001) have in-cluded particle clearance in a model for inhaled nickeldosimetry and have applied it to occupational sce-narios. Physiologically based pharmacokinetic (PBPK)models can also be used to improve animal to hu-man extrapolation (reviewed in Clewell and Andersen,1994).

Application of Uncertainty Factors

Uncertainty factors (UF) are used routinely (eitherexplicitly or implied as part of a margin of safety) inestablishing OELs, as a means to account for uncer-tainty in extrapolating from the selected POD. How-ever, despite their wide application, the rationale fortheir use, areas of uncertainty considered, and the rec-ommended default values differ by organization. In ad-dition, the degree to which UFs are explicitly defined inOEL documentation varies widely. Table 6 provides ex-amples ofUF approaches proposed for occupational riskassessment. None of the three organizations includedin our evaluation of the case study chemicals appliesa defined set of default uncertainty factors, althoughDECOS uses the TNO (1996) recommendations as ageneral guide (personal communication from DECOS).Although the ACGIH and MAK likely make adjust-ments to OELs based on similar considerations, a sys-tematic approach for accomplishing this is not explicitin their OEL documentations. As a means of com-paring occupational and environmental health riskassessment approaches, default values used by U.S.EPA have also been included in Table 6.

Harmonization efforts in occupational risk assess-ment have not yielded specific recommended defaultvalues for uncertainty factors, even within an individ-ual organization. Instead, recognition has evolved thatthe selection of uncertainty factors will be done on acase-by-case basis (SCOEL, 1999). This represents astep forward in the development of scientific criteria,by indicating that, although the value of uncertaintyfactors will be chemical-specific, the magnitude andrationale for the UF should be described in the OEL

Extrapolations from Other Routes and fromBiological Indices

For some chemicals, there are insufficient data fromthe inhalation route to derive an OEL, or the oral datamay be of much higher quality. In such cases, it may beappropriate to conduct a route-to-route extrapolation.The key issue in doing such extrapolation is considera-tion of whether there are portal of entry effects. For ex-ample, route-to-route extrapolation is not appropriateif a chemical causes respiratory effects from inhalationexposure, gastrointestinal effects from oral exposure,or is subject to first-pass metabolism in the liver from

272 HABER AND MAIER

TABLE 8Proposed Default Uncertainty Factors

for OEL Derivation

Interspecies 3 I' 1-10 rlntraspecies 3 10 1-8 10Duration of exposured 1-10 3 6-10 10LOAEL to NOAEL -- 3 S-4 10Type of critical effect' 1 - - -Doee-reeponae curve' 1 - - -Database 1 1 - 10Muimum default 100 100 1200 ~

a U.S. EPA RflJ value was included as a comparison for occupationaland nonoccupational human health riak values. These default valuesare modified as appropriate based on the available data.

b The default value of 1 888Ume8 appropriate allometry has been

applied." Assuming appropriate dosimetric adjustments have been con-ducted.

d Refers to extrapolation from subacute to subchronic or subchronicto chronic studies.

. A dash indicates that a default value is not specified for this anaof uncertainty.

, This factor accounts for the biological significance of the observedeffect.

. The steeper the doee-reaponae curve the smaller the factor can be., The maximum default value is 3000 in recognition of the overlap

in coverage of the factors when four full areas of uncertainty exist.lffive full areas of uncertainty exist, an aft:: is not derived.

cupational and environmental risk assessment. It hasbeen argued that the default value for this factor shouldbe lower for the occupational setting, since the popula-tion of interest is composed of healthy working adultsand excludes potentially sensitive segments of the pop-ulation such as children or the elderly. As an example,in its recommended process for deriving health-basedOELs, TNO (1996) recommended a default value of 3for human variability, instead of the factor of 10 thathas been used traditionally for nonoccupational humanhealth risk assessment (for example, in U.S. EPA, 1994).The TNO (1996) methodology further notes that thisargument would not apply to occupational exposuresthat induce developmental toxicity. It also stands to rea-son that if factors other than age or "general health"drive differences in susceptibility, then a reduction inthe factor to cover human variability may not be ap-propriate. For example, if nonoccupational exposures(such as alcohol or smoking) or genetic predispositioninduce greater susceptibility, then reducing the defaultuncertainty factor based on age differences would notcapture the range of variability of the occupationallyexposed population. Overall, a reduction in the uncer-tainty factor for human variability seems reasonable formost cases, but the exceptions we note make a strongcase for using chemical-specific data to identify poten-tial populations at increased risk.

An argument related to decreasing the magnitude ofthe UF for human variability for metals has also re-cently been discussed in the context of nickel (C~E,2001). According to this argument, human variabilityfor some metals (inorganic nickel in this case) may beless than for nonmetal chemicals, because inorganicmetal compounds are generally less likely to be depen-dent on metabolism. However, as noted in the discus-sion of the UF for nickel in this document (CSTEE,2001), metabolism reflects only a subset of the factorsthat influence toxicokinetics. In light of this possibility,the data were not judged to be sufficient to support a re-duction of the UF. This evaluation of nickel highlightsthe utility of the CSAF approach for detennining theappropriate value ofUFs, since the CSAF approach pro-vides a framework for evaluating the degree to whichthe data for any particular metal would support move-ment away from default UF values.

Although movement toward refining the rationale forUF selection in occupational risk assessment is encour-aging, evaluation of the case studies highlights currentdifferences in the routine application ofUFs (Table 3).Our review represents just a small sampling of the OELvalues for metals derived by organizations around theworld, but it appears sufficient to highlight key points.For many of the OELs, the POD was not explicitly de-scribed, making it difficult to determine the magni-tude of any UFs that were considered. Even in caseswhere the POD was discemable, no explicit descrip-tion of the basis and rationale for the magnitude of the

documentation. Adopting this principle would increasethe transparency of currently available OELs.

The scientific basis for selection of default uncer-tainty factor values has been discussed in detail else-where for occupational risk assessment (FAIR, 2000,Haber et al., 2001; Naumann et al., 1995; Zielhuis andvan der Kreek, 1979), and therefore is not repeated indetail here. Newer developments, such as the currenteffort by the IPCS for deriving chemical-specific adjust-mentfactors (CSAF) (IPCS, 2001; Meeketal., 2001) forhuman health risk assessment for environmental expo-sures, maximize the use of chemical-specific data, whenavailable. Similar concepts would enhance the use ofchemical-specific data to derive UFs for OELs. Indeed,Naumann et al. (1997) proposed a similar approach forOEL development for pharmaceuticals, and Sweeneyet al. (2001) recently used a PBPK approach, coupledwith Monte Carlo analysis, to derive composite UFsfor ethylene glycol ether OELs. Thus, sophisticated ap-proaches are being used in the occupational arena andshould continue to be used, to the degree the data allow,for derivation of new OELs.

Although much of the scientific development in theapplication ofUFs in the environmental area appears tobe directly applicable to occupational risk assessment,extrapolation for human variability does differ for oc-

273SCIENTIFIC CRITERIA FOR DELe

of action include cytotoxicity followed by regenerativecell division and hormonally related events. Organiza-tions such as DECOS (1998a) and DFG (2001) considerdirect-acting genotoxic carcinogens to have no thresh-old for carcinogenicity, while nongenotoxic (or indirectgenotoxic) modes of action would result in a thresholdor a very shallow slope in the low-dose region. For ex-ample, if a chemical acts via cytotoxicity and regenera-tive cell division, doses below those causing cytotoxicitywould presumably not cause cancer. Some organiza-tions (FAIR, 2000) also distinguish direct DNA reac-tivity (forming DNA adducts) from indirect effects onDNA, such as chromosome aberrations resulting frominteraction with the proteins that drive cell division.

Although metals with positive results in genotoxicityassays are typically considered nonthreshold genotoxiccarcinogens, further research is needed in this area.Metals can induce genotoxicity through diverse mech-anisms, including through direct DNA reactivity or in-direct mechanisms, such as inhibition of DNA repair,or via the formation of reactive oxygen species leadingto DNA damage (Chang, 1996). The degree to whichthese alternative pathways can be demonstrated for aparticular metal will affect the type of dose-responsemodel (e.g. threshold or nonthreshold) used in deriVingthe OEL.

composite uncertainty factor was generally given. Theexception was the Cr (III) and Cr (IV) OELs recom-mended by DECOS (1998a), which applied a factor of3for extrapolation from a MOAEL or WAEL and a factorof 3 for interspecies extrapolation.

It also appears that there are differences in the useof UFs to account for intraspecies (human) variability.For example, an apparent UF of2 was used for the TLV-TWA for copper, while an apparent factor of3 was usedfor lead when the critical effect was based on a humanNOAEL. The UF values used for the copper and leadTLV-TWAs are generally consistent with intraspeciesfactors used in a series of health-based OELs in theUnited Kingdom as described by Fairhurst (1995) andare similar to the value of 3 recommended by HCN(2000) and TNO (1996). It is noteworthy that for chemi.cals such as lead, for which developmental toxicity is thecritical endpoint, exposure of pregnant workers wouldresult in exposure of the sensitive population (fetuses),which is the same sensitive population as has beenidentified for general population exposure (FAIR, 2000;TNO, 1996). For other OELs based on a POD from ani-mal data, such as the DECOS (1998a) OELs for Cr (III)and Cr (lV), there was no explicit notation of a factor toaccount for human variability. For manganese, the totalUFforthe TLV-TWA was a factorof5, but the documentauthors did not identify which portion of this factor wasintended to cover human variability and which portionwas intended to cover extrapolation from a LOAEL.The MAK value for manganese was based on a mini-mal human LOAEL and apparently did not incorporatea factor to account for human variability. Based on thislimited analysis, increased harmonization in the appli-cation ofUFs and better communication of the rationalefor their use is clearly needed.

OELs for Carcinogens

While a generally consistent paradigm is used bymost organizations for setting OELs for noncarcino-gens, dramatic differences are apparent in the ap-proaches used for carcinogens. Some organizations (e.g.,ACGIH) routinely establish quantitative OELs for car-cinogens, while others (e.g, DECOS, DFG) do not de-rive health-based OELs for nonthreshold carcinogens,although they do derive health-based OELs for carcino-gens with thresholds. In addition, carcinogen classifi-cation schemes vary widely among organizations thatestablish OELs.

All of the case study organizations consider mode ofaction to some degree in setting OELs for carcinogens.For example, all of the organizations evaluate whetherthe tumors induced by the chemical in animals are rele-vant to human carcinogenicity. All of the organizationsalso consider whether the chemical causes tumors bya genotoxic (by causing gene mutations) mode of actionor a nongenotoxic one. Examples ofnongenotoxic modes

Additional Issues of Specific Importance for Metals

An area of specific importance to metals is in deter-mining the degree to which data are available to developOEL recommendations for different forms of the metal.Issues of speciation include determining the degree towhich toxicological differences result from differencesin physical form (e.g., copper dust vs fume deposition),chemical form (e.g., the bioavailability ofCr (Ill) versusCr (VI)), or other unique toxicological properties of thecompound (amorphous versus crystalline silica).

F1gure 1 provides an exarnple of a decision processfor determining the degree to which an OEL can be de-veloped for different forms of metals. This frameworkcan also be used to improve the documentation for therationale for applying an OEL for one form of a metalto another form, or choosing not to apply the OEL. Thisanalysis includes an evaluation as to whether multipleforms exist and if the potential health hazards are likelyto differ on this basis. If the initial hazard identificationdoes not suggest that speciation is an im~rtant factor,then an evaluation of the adequacy of the available datafor the form (or forms) is made, followed by a decision tocollect data or derive an OEL from the existing data set.In the event that sufficient data are available for all rel-evant forms, separate OELs can be developed. It is morecommon, however, that sufficient data would be avail-able for only a subset of the forms. Two alternatives fordealing with this situation include (1) collecting addi-tional data for each forln of interest and (2) assuming

HABER AND MAIER274

~TMultipleFOOm?

T../ Suff1ciel1t. ~

~(I1each

~~"")Y

No

,t

y~

Decision process for considering speciation in deriving OELs.FIG. 1

The U.S. EPA RiD for Cr (III) is an example of theappropriate consideration of essentiality in the estab-lishment of oral risk values (U.S. EPA, 2001). A simi-lar concept would apply for OEL derivation, subject tothe caveats in the following paragraphs. Cr (III) actsas a cofactor for insulin and is required for maintain-ing normal glucose levels. The dietary requirement forCr (III) is not known, but an Estimated Safe and Ade-quate Daily Dietary Intake (ESADDI) of 50-200 JJ.g/dayhas been established (NRC, 1989). The Rffi for Cr (III)is 1.5 mg/kg-day, corresponding to 105,000 JJ.g/day for

~.....i~DoI~

similar toxicology based on a physical or chemical char-acteristic of interest (e.g., degree of crystallinity of sil-ica or the solubility of nickel compounds). If the latterapproach is followed, careful consideration needs to bemade of the characteristics of interest, to ensure thatthe surrogate choice is appropriate. A health-protectiveapproach in the absence of data is to assume that allcompounds are as toxic as the most toxic form.

Overall, this process is consistent with current ap-proaches, and the case studies suggested that the roleof speciation has typically been considered, at least tosome degree. However, it is critical that the evaluationbe sufficiently described to communicate the degree towhich the occupational hygiene practitioner can rely onan OEL for specific compounds of interest in the workarea.

Another special challenge for the derivation ofOELsfor metals and minerals is the issue of essentiality.For a number of metals, such as selenium, zinc, andCr (1lI), toxic effects can result from exposure to highdoses, but adverse health consequences can also re-sult from insufficient intake. This means that the netdose-response curve is the sum of the downward slop-ing dose-response curve for essentiality in the low-doseregion and the upward sloping dose-response curve fortoxicity at higher doses (see Fig. 2). In such cases, thechoice of uncertainty factor needs to consider the lowerportion of the dose-response curve.

Dose

FIG. 2. Assessment for essential elements. Hypothetical exam-ple of the impact of uncertainty factor selection on the probability ofan adverse response for an essential metal. EAR, estimated averagerequirement; RDA, recommended dietary allowance; UL, tolerableupper level intake; RiD, reference dose; LOAEL, lowest observed ad-verse effect level.

275SCIENTIFIC CRITERIA FOR OELs

ing theme in this analysis was the need to increase thedegree to which the selected approach was describedin the OEL documentation. Consistent with good riskcharacterization principles, OEL documentation should(1) be transparent with regard to methodology and sci-entific judgments; (2) clearly identify the data usedas the basis for the OEL calculation; (3) discuss thestrengths and weaknesses in the OEL derivation; and(4) use scientific approaches consistent with those forother OELs (at least within an organization). It will be-come increasingly important to adhere to these princi-ples of risk characterization as the field of occupationalrisk assessment moves away from default approachesand adopts tools that use chemical-specific data to themaximum extent possible. For example, application ofBMD modeling, dosimetric adjustments, CSAFs, andcancer mode of action summaries will require renewedeffort to make clear the scientific weighing of evidencethat was used in making decisions.

Recommendations are summarized in Table 7. Thefirst four recommendations can be implemented with-out additional research; the primary barriers to im-plementation are limited resources (for the develop-ment of improved documentation, or for developingmanuscripts on negative studies) and institutional in-ertia. This report found that even when organizationsreview similar data sets, differences in the resultinghealth-based OELs can occur. This situation raises asignificant problem for practicing occupational hygien-ists, who must often choose the appropriate OEL basedon limited or poor documentation of the scientific ra-tionale underlying the recommended value. Althoughsome differences in OELs can be explained by differ-ences in science policy (e.g., different approaches forsetting limits for carcinogens), other differences canbe addressed through increased information. The lastfive recommendations in Table 7 list specific areaswhere both harmonization and additional researchwould be beneficial. The first step in this harmonization

a 70-kg human. This Rffi is higher than the ESADDI,and is consistent with the ESADDI, since doses belowthe ESADDI may be of concern, while exposure to levelsabove the RID may also be of concern.

Two major caveats are important in applying the es-sentiality concept to OEL development. First, essential-ity refers to the systemic dose, not the dose to the por-tal of entry, the respiratory tract. Inhalation exposureto a metal fume or particulate could lead to respira-tory effects at atmospheric concentrations below thosethat would lead to systemic doses in the essentialityrange. For example, the OELs established for water-soluble Cr (Ill) by ACGIH (1996) and DECOS (1998a)are 0.5 and 0.06 mg/m3, respectively, based on lung ef-fects. Using the default air intake for occupational ex-posure of 10 m3/day and assuming the entire lung doseis absorbed, these OELs correspond to systemic doses of5000 and 600 JLg/day. Because the systemic dose at theDECOS OEL is close to the ESADDI, careful evaluationof the appropriateness of that OEL would be neededif the OEL were based on a systemic effect. However,since the OEL is based on a lung effect, a low systemicdose at the OEL is still consistent with the biology ofCr (Ill).

The second major caveat is that essentiality refersto the total dose from all sources. The development ofan OEL based on systemic effects of an essential el-ement would be a three-step process. First, the safetotal systemic dose is determined. Second, nonoccupa-tional exposure from sources such as the diet, drink-ing water, and ambient air is evaluated to determinethe background dose from these sources. Finally, thedifference between the safe systemic dose and the to-tal systemic background dose is determined in order toidentify the 0 EL. In other words, occupational exposureshould be considered to be on top of normal dietary in-take (and other background intake) of the metal. Thismay pose particular challenges if a narrow window ex-ists between essentiality (e.g., the recommended dailyallowance or RDA) and doses that begin to cause toxiceffects.

In summary, the key issue regarding essential metalsis that it is nonsensical to derive a safe exposure levelbased on systemic effects that are below the nutritionalrequirement. OEL development needs to take into ac-count both the total safe dose of essential metals andthe daily intake from dietary and other sources. How-ever, these issues only apply when the most sensitiveendpoint is a systemic effect.

TABLE 7Recommendations on Scientific Criteria for OELs

RECOMMENDA110NS AND CONCLUSIONS

A case study approach has been used to identify sim-ilarities and differences in OEL derivation for metals,metal compounds, and other chemicals related to min-ing operations, as well as general issues related to thedevelopment and documentation ofOELs. The overrid-

Improve transparency and completeness ofOEL documentation,including identification of strengths and weaknesses of theanalysis

Provide greater accessibility to "gray literature-Increase dissemination and publication of occupational studies

that evaluate several endpoints, but observe no effectDevelop approaches for characterizing the overall confidence in

OELsHarmonize the consideration of severity in the identification of the

point of departureHarmonize the definition of a lninimum data set for the

development of an OELHarmonize the approach for interspecies extrapolationHarmonize the default uncertainty factors used in developing OELsHarmonize the approach for consideration of speciation and

essentiality of metals

276 HABER AND MAIER

is clearer communication of the process used in address-ing these issues while deriving OELs. It is envisionedthat clear documentation of the approach used will helpcrystallize similarities and differences in OEL develop-ment. This can help to focus research addressing thelatter five points in the table and ultimately perhapslead to a greater scientific consensus.

In many cases, there were scientifically defensible al-ternative approaches to address issues that were ex-plored, such as defining a minimum database, select-ing a point of departure, or making uncertainty factordecisions. For these areas, the primary barriers to har-monization are in validation of alternative approaches,education of the scientists who develop OELs, and addi-tional resources needed to implement some of these ap-proaches. The next few paragraphs briefly layout pos-sible approaches to research addressing some of theseremaining issues.

bution of that ratio, and the UF needed to cover a spec-ified proportion of this variability is then determined.This approach could be optimized for OEL derivation byexamining distributions of data for routes of exposureand effects common to workplace environments. As oneexample, the distribution of animal/human NOAEL ra-tios for inhalation studies for metals and metal com-pounds could be used to derive the default UF foranimal-to-human extrapolation. This general researchapproach could be used to derive default UFs to accountfor many areas of uncertainty relevant to OEL deriva-tion.

Overall, it is apparent that the potential value ofhar-monization of scientific methods for deriving OELs isbeing increasingly recognized. This is viewed as a fa-vorable trend for two specific reasons. First, increasedharmonization of default scientific approaches will im-prove the transparency in OEL derivation. Second, har-monization of approaches will tend to highlight thestrengths of methods used by diverse groups and thuswill favor the application of approaches viewed by awide audience of occupational risk assessors as beingrooted in the best science. This commonality in ap-proach might ultimately minimize the differences ob-served in health-based OELs and perhaps decrease p0-tential confusion for occupational health practitionersresponsible for worker health on a global basis. In-creased harmonization would also open the door to shar-ing of the resulting analyses. Pooling of resources isa critical goal, because decreased redundancy of workwould allow for the use of resources to expand the cover-age of OELs for substances for which no OEL guidanceis currently available.

APPENDIX A. CALCULATION OF THE HUMANEQUIVALENT CONCENTRATION

Database considerations: Defining a minimum dataset. A component of evaluating the adequacy of humandata is to ascertain the minimum database needed toplace high confidence in an OEL. A workable researchapproach would be to compile published human datafor metals and metal compounds having datasets richin various study types. To test the impact of using onlycertain study types (e.g., case studies, full epidemiol-ogy reports), one could develop preliminary OEL val-ues based on selected subsets of the data. By compar-ing OELs derived using various study types, it wouldbe possible to evaluate the level of data sophisticationneeded to derive a high quality OEL. The results of thisresearch approach would provide a further scientific ba-sis for determining minimum database requirements.

Interspecies extrapolation: Using state of the sciencemethods. Dosimetric adjustment methods capture im-portant interspecies differences in dose that are not di-rectly accounted for in current OELs. An ongoing inter-agency effort led by the EPA is developing dosimetry forthe oral, inhalation, and dermal routes, with particularattention to the portal of entry. Implementation of suchmethods will require validation of the approach, dis-semination of the approach, and dedication of resourcesfor the increased consideration of mode of action andmore time-intensive calculations needed to implementthe approach.

The DECOS OEL is based on the observation ofabnormal macrophages (oblong, smooth) in alveoli at0.6 mg/m3 in a subacute inhalation study in rabbits(Johansson et al., 1986a, 1986b). This effect in the alve-oli is considered a pulmonary effect. The study au-thors reported a mass median aerodynamic diameter(MMAD) of 1 JLm, but did not provide any measureof the breadth of the distribution, such as a geomet-ric standard deviation (GSD). Information on the par-ticle generation protocol can aid in the estimation ofthe breadth of the distribution, but this informationwas not reported. Therefore, a sensitivity analysis wasconducted using a range of GSD values. GSD valuesranging from 1 to 5 JLm were considered in the cal-culation, based on typical GSD values reported in ex-perimental animal studies, particularly those reportingan MMAD of 1 JLm. The regional deposited dose ratio(RDDR) was calculated using the RDDR program pro-vided by the U.S. EPA (1994). The RDDR represents theratio between the dose deposited in a given region of the

Uncertainty factors (UFs): Defining defaults. Re-search approaches used to validate default UF val-ues for application in the environmental arena (e.g.,Dourson et al., 1992) could be used successfully for theoccupational arena. One approach that has been usedto develop quantitative UF defaults is to determinethe ratio between the value of interest (e.g., the ani-mal NOAEL) and the surrogate value (e.g., the humanNOAEL) for a reasonably large group of sample chemi-cals. This sample is then used to characterize the distri-

277SCIENTIFIC CRITERIA FOR GELs

respiratory tract when animals are exposed to a givenconcentration of the particle in air and the dose tothe same respiratory tract region received by humansexposed to the same air concentration. The RDDR isnormalized by regional surface area. (Other dosimetrymethods, such as that used by Oberdorster and col-leagues, normalize based on tissue weight.) In additionto the particle size distribution, inputs to the calcula-tion of the RDDR include the animal and human bodyweight, the surface area of the respiratory tract regionof interest, and the minute volume4. Using this ap-proach, RDDR values for the pulmonary region of 0.123to 0.134 were calculated for male rabbits exposed to aparticle size distribution of with an MMAD of 1 JLm anda GSD of 1-5 JLm, indicating that the uncertainty in theGSD had a minimal impact for this set of conditions.An RDDR of 0.12 was chosen for this calculation. Thehuman equivalent concentration is calculated by multi-plying the RDDR and the NOAEL, LOAEL, or MOAEL,resulting in a MOAEL (HEC) of 0.072 mg/m3.

ACKNOWLEDGMENTS

This project was funded by the International Council on Mining andMetals (ICMM). although the conclusions are those of the authors.Useful comments provided by ICMM and its science advisory boardare gratefully acknowledged.

REFERENCES

American Conference of Governmental Industrial Hygienists(ACGIH) (1991). Documentation of Threshold Limit Values (TLVs)and Biological Exposure Indices (BEIs), 6th ed. American Confer-ence of Governmental Industrial Hygienists, Cincinnati, OH.

American Conference of Governmental Industrial Hygienists(ACGIH) (1996). Documentation of Threshold Limit Values (TLVs)and Biological Exposure Indices (BEIs), supplement to the 6thed. American Conference of Governmental Industrial Hygienists,Cincinnati, OH.

American Conference of Governmental Industrial Hygienists(ACGIH) (2000). Documentation of Threshold Limit Values (TLVs)and Biological Exposure Indices (BEIs), supplement to the 6thed. American Conference of Governmental Industrial Hygienists,Cincinnati, OH.

American Conference of Governmental Industrial Hygienists(ACGIH) (2001). Threshold Limit Values for Chemical SUb3tanceSand Physical Agents & Biological Exposure Indices. American Con-ference of Governmental Industrial Hygienists, Cincinnati, OH.

Chang, L. W. (1996). Toxicology of Metals. CRC Lewis, New York.

Clewell, H. J., and Andersen, M. E. (1994). Physiologically-basedpharmacokinetic modeling and bioactivation of xenobiotics. 7bxj.col. Ind. Health 10, 1-24.

Corsi, G., and Piazza, G. (1970). On the possibility of the appearanceof silicosis with personnel engaged in the production offerro-silicon.Med. Lavoro 61, 109-122.

Crump, K. S. (1984). A new method for determining allowable dailyintakes. Fund. Appl. 7bxicol. 4, 854--871.

Scientific Committee for Toxicity, Ecotoxicity, and Environment(CSTEE) (2001). Minutes of the 21st plenary meeting. InEuropean Union, Scientific Committee for Toxicity, Ecotoxi-city, and the Environment, BrU88eIs, Belgium. Available athttp://europa.eu.intlcomm/food/fs/ sclsctJout93 ~n.h tml.

Dutch Expert Committee on Occupational Standards (DECOS)(1992). Health-Based Recommended Occupational Exposure Lim-its for CrystaUine Forms of Silicon Dioxide (Free Silica). Ministryof Social Affairs and Employment, Dutch Expert Committee on 0c-cupational Standards, The Hague.

Dutch Expert Committee on Occupational Standards (DECOS)(1998a). Chromium and Its Inorganic Compounds. Publication No.199&'OI(R)WGD. Health Council of the Netherlands, Dutch ExpertCommittee on Occupational Standards, Rij8wijk.

Dutch Expert Committee on Occupational Standards (DECOS)(1998b). Quartz: Evaluation of the Carcinogenicity and Geno-toxicity. Publication No. 199&'02 WGD. Health Council of theNetherlands, Dutch Expert Committee on Occupational Standards,Rij8wijk.

Deutsche Forschungsgemeinschaft (DFG) (1991). Silica, amorphous.In Occupational 7bxicants: Critical Data Evaluation for MAK Val-ues and Classification of Carcinogens (D. Henschler, Ed.), Vol. 2,pp. 157-178. Wiley-VCH, Weinheim.

Deutsche Forschungsgemeinschaft (DFG) (1992). Chromium and iucompounds. In Occupational 7bxicants: Critical Data Evaluationfor MAE Values and Classification of Carcinogens, (D. HenBCbler,Ed.), Vol. 3, pp. 101-122. Wiley-VCR, Weinheim.

Deutsche Forschungsgemeinschaft (DFG) (1999). Manganese andits organic compounds. In Occupational 7bxicants: Critical DataEvaluation for MAE Values and Classification of Carcinogens,(G. Helmut, Ed.), Vol. 12, pp. 293-328. Wiley-VCH, Weinheim.

Deutsche Forschungsgemeinschaft (DFG) (2001). List of MAK andBAT Values 2tXXJ: Maximum Concentrations and Biological 7bler-ance Values at the Workplace. Deutsche Forschungsgemeinschafi;,Weinheim.

Dourson, M. L., Hertzberg, R. C., Hartung, R., and Blackburn, K(1985). Novel methods for the estimation of acceptable daily intake.7bxicol. Ind. Health 1, 23-33.

Dourson, M. L., Knauf, L. A., and Swartout, J. C. (1992). On refer-ence dose (RtD) and its underlying toxicity dsta base. Th.ricol. Ind.Health 8, 171-189.

Dourson, M. L., Felter, S. P., and Robinson, D. (1996). Evolution ofscience-based uncertainty factors in noncancer risk assessment.Regui. 7bxicol. Pharmacol. 24, 10S-120.

Food,Agriculture and Fisheries Programme (FAIR) (2000). Criteria toestablish health-based occupational exposure limits for pesticides,European Union: Food, AgricuUure and Fisheries Programme.

Fairhurst, S. (1995). The uncertainty factor in the setting of occupa-tional exposure standards. Ann. Occup. Hyg. 39, 375-385.

Feron, V. J., Hoeksema, C., Arts, J. H. E., Noordam, P. C., and Maas,C. I. (1994). A critical appraisal of the setting and implementationof occupational exposure limits in the Netherlands. Indoor Environ.3, 260-265.

Groth, D. H., Moorman, W. J., Lynch, D. W., Stettler, L. E., Wagner,W. D., and Hornung, R. W. (1979). In Proceedings of the Symposiumof Health Effects of Synthetic Silica Particulates, p. 118. Marbella,Spain, November 3-7.

Haber, L. T., Erdreich, L., Diamond, G. L., Maier, A. M., Ratney, R,Zhao, Q., and Dourson, M. L. (2000). Hazard identification and doseresponse of inhaled nickel-soluble salts. Regul. 7bxicol. Pharmacal.31, 210-230.

Haber, L. T., Dollarhide, J. S., Maier, A., and Dourson, M. L.(2001). Noncancer risk assessment: Principals and practice in

278 HABER AND MAIER

for phannaceutical active ingredients. Hum. &01. Risk Assess. 1,590-613.

National Institute for Occupational Safety and Health (NIOSH)(2001). www.cdc.gov/niosh/idlhintr.html.

Nickel Producers Environmental Research Association (NiPERA)(2001). Epidemiological guidelines for the collection and retentionof health information in nickel-expoeed workers. Presentation byN. Lightfoot, S. Williams, and G. Hall, at NiPERA Workshop onOccupational Exposure and Health Assessment Guidelines forNickel. Available from NiPERA, Durham, N.C.

National Research Council (NRC) (1989). Recommended Daily Al-lowances. National Academy Press, Washington, D.C.

National Toxicology Program (NTP) (1996a). 7b.ricology and Carcino-genesis Studies of Nickel Sul(lJte Hexahydrate (CAS No. 10101-97-0)in F344/N Rats and B6C3F1 Mice (inhalation studies). Washing-ton, DC, National Institutes of Health, 1-368.

National Toxicology Program (NTP) (l996b). 7b.ricology and Carcino-genesis Studies of Nickel Oxide (CAS No. 1313-99-1) inF344/NRatsand B6C3F1 Mice (Inhalation Studies). Washington, DC, NationalInstitutes of Health, 5-9.

National Toxicology Program (NTP) (1996c). To%icology and Car-cinogenesis Studies of Nickel Subsulfide (CAS No. 12035-72-2) inF344 / N Rats and B6C3F1 Mice (Inhalation Studies). Washington,DC, National Institutes of Health, 4-10.

Occupational Safety and Health Administration (OSHA) (1993). Oc-cupational exposure to cadmium. Fed. Reg. 57,42102. Available atwww.osha-slc.gov/PreambleiCa dmi ~ talC AD MIUM 6 . h tml.

Paustenbach, D. J. (1990). Occupational exposure limits: Their criti-cal role in preventive medicine and risk management. AIHAJ 51,A332-A336.

Paustenbach, D. J. (1994). Occupational exposure limits, phanna-cokinetics, and unusual work schedules. In Patty's Industrial Hy-giene (R. L. Harris, L. J. Cralley, and L. V. CraIley, Eds.), 3rd ed.,pp. 191-237. Wiley, New York.

Reuzel, P., Woutersen, R., and Bruntjes, J. (1987). Civo Insti-tutes TNO. Report No. V 86.347/240718. Degussa AG, Frankfurt,Germany.

Roels, H., Lauwerys, R., Buchet, J. P., Genet, P., Barban, M. J.,Hanotiau, I., de Fays, M., Bernard, A, and Stanescu, D. (1987).Epidemiological survey among workers exposed to manganese: ef-fects on lung, central nervous system, and some biological indices.Am. .lInd. Med. 11,307-327.

Scientific Committee on Occupational Exposure Limits (SCOEL)(1999). Methodology for the Derivation of Occupational ExposureLimits: Key Documentation. European Union, Scientific Commit-tee on Occupational Exposure Limits, Luxembourg.

Seeley, M. R., Tonner-Navarro, L. E., Beck, B. D., Deskin, R., Feron,V. J., Johanson, G., and Bolt, H. M. (2001). Procedures for healthrisk assessment in Europe. Regul. 7bxicol. Phormacol. 34.153-169.

Sonich-Mullin, C. (1997). Harmonization of approaches to the as-sessment of risk from exposure to chemicals. In EnvironmentalRisk Harmonization: Federal and State Approaches to Environ-mental Hazards in the USA (M- A Kamrin, Ed.), pp. 11-20. Wiley,New York.

Sweeney, L. M., Tyler, T. R., Kinnan, C. R., Corley, R. A, Reitz, R. H.,Paustenbach, D. J., Holson, J. F., Whorton, M. D., Thompson, K. M.,and Gargas, M. L. (2001). Proposed occupational exposure limits forselect ethylene glycol ethers using PBPK models and Monte CarloSimulations. 7bxicol. Sci. 62, 124-139.

Netherlands Organization for Applied Scientific Research (TNO)(1996). Methods for the Establishment of Health-Based Recom-mended Occupational Exposure Limits for Existing Substances.TNO Nustrition and Food Research Institute, The Netherlands.

U.S. Environmental Protection Agency (U.8.EPA) (1994). Methods forDerivation of Inhalation Reference Concentrations and Application

environmental and occupational settings. In JtJ.tty's 7bxicology(E. Bingham, B. Cohrssen, and C. H. Powell, Eds.), 5th ed.,pp. 169-232. Wiley, New York.

Health Council of the Netherlands (HCN) (2000). Committee onUpdating of Occupational Exposure Limits: Health-Based Re-a88e8Sment of Administrative Occupational Exposure LimitBoNo. 2000/150SH/OOO. Health Council of the Netherlands.

Hill, A B. (1965). The environment and disease: Association or cau-sation? Proc. Royal Soc. Med. 58, 295-300.

Hsieh, T. H., Yu, C. P., and Oberdorster, G. (1999). Modeling of depo-sition and clearance of inhaled Ni compounds in the human lung.Regul. 7bxicol. Pharmacol. 30, 18-28.

ICME (1999). Guide to Data Gathering Systems for RiskAssessment of Metals and Metal Compounds. Available atwww.icmm.com/htmVpubs_pubs.phpltop.

International Committee on Nickel Carcinogenesis in Man (ICNCM)(1990). Report of the international committee on nickel carcinogen-esis in man. Scand. .l Work Environ. Health 16, 1-82.

International Programme on Chemical Safety (IPCS) (1999). En-vironmental Health Criteria 210: Principals for the A8se88mentof Risks to Human Health from Exposure to Chemical& Interna-tional Programme on Chemical Safety, World Health Organization,Geneva, Switzerland.

International Programme on Chemical Safety (IPCS) (2000). Envi-ronmental Health Criteria No. 214: Human Exposure Assessment.International Programme on Chemical Safety, World Health Orga-nization, Geneva, Switzerland.

International Programme on Chemical Safety (IPCS), (2001). Guid-ance Document for the Use of Chemical-Specific AdjustmentFactors (CSAFs) for Interspecies Differences and Human Vari-ability in Dose/Concentration-Response Assessment. Available atwww.ipcsharmonize.org.

Iregren, A (1990). Psychological test performance in foundry work-ers exposed to low levels of manganese. Neurotoxicol. Teratol. 12,673-675.

Joh80880n, A, Lundborg, M., Hellstrom, P. A, Camner, P., Keyser,T. Ro, Kirton, S. E., and Natusch, D. F. (1980). Effect of iron, cobalt,and chromium dust on rabbit alveolar macrophages: a comparisonwith the effects of nickel dust. Env. Res. 21, 165-176.

Johansson, A., Wiernik, A, Jarstrand, C., and Camner, P. (1986&).Rabbit alveolar macrophages after inhalation ofhexa-and trivalentchromium. Env. Res. 39, 372-385.

Johansson, A., Robertson, B., Curstedt, T., and Camner, P. (l986b).Rabbit lung after inhalation ofhexa-and trivalent chromium. Env.Res. 41, 111}..119.

Lauwerys, R., Roels, H., Genet, P., Toussaint, G., Bouckaert, A, andDe Cooman, S. (1985). Fertility of male workers exposed to mercuryvapor or to manganese dust: a questionnaire study. Am. .l I nd. Moo.7,171-176.

Lee, K. P., Ulrich, C. E., Geil, R. G., and Trochimowicz, H. J. (1989).Inhalation toxicity of chromium dioxide dust to rats after two yearsexposure. Sci. 7btal Environ. 86, 83-108.

Luxon, S. G. (1972). Letter to TLV committee from industrial hygieneunit H. M. Factory Inspectorate, London, England, August 1.

Meek, B., Renwick, A, Ohanian, E., Dourson, M., Lake, B., Naumann,B., and Vu, V. (2001). Guidelines for application of compound spe-cific adjustment factors (CSAFs) in dose/concentration response as-sessment. Comments 7bxicol. 7,575--589.

Naumann, B. D., Weideman, P. A, Dixit, R., Grossman, S. J., Shen,C. F., and Sargent, E. v: (1997). Use of toxicokinetic an toxico-dynamic data to reduce uncertainties when setting occupationalexposure limits for pharmaceuticals. Hum. Ecol. Risk Assess. 3,555-565.

Naumann, B. D., and Weidman, P. A (1995). Scientific basis for un-certainty factors used to establish occupational exposure limits

SCIENTIFIC CRm:RIA FOR OEI.. 279

Mmiologicol Eviden« for Environmental Health RiM A ment.Guideline document, E~69.

Yu, C. P., Hsieh, T. H., Oller, A R., and Oberdo~r, G. (2001). Eval-uation of tJ1e human nickel retention model witJ1 workplace data.&gul. ~. p~ aa, 165-172.

Zielhus, R. L., and van der Kreek. F. W. (1979). 'n1e use ofa safety ra~r in setting health baaed permi88able levels foroccupational exposure. Int. Arch. Occup. Environ. Health 42,191-201.

of Inhalation Dosimetry. United States Environmental ProtectionAgency, Washington, DC.

U.S. Environmental Protection Agency (U.S. EPA) (2001). IRISDatabase. United States Environmental Protection Agency,Washington, DC, www.epa.gov/iris.

Vmc:ent, J. H. (1999). Occupational hygiene Ici~ and it. applicationin occupational health policy, at home and abroad. Occup. Med. 48,27-35.

World Health Organization (WHO) (2000). ElIOluation and U Be of Epi-