james etal2008

O

34 PROFESSIONAL SAFETY SEPTEMBER 2008 www.asse.org

Occupational HazardsOccupational Hazards

Worker Exposureto Secondhand Smoke

Evaluating a prediction modelBy Harry R. James, Lisa Barfield, Janice K. Britt and Robert C. James

OCCUPATIONAL EXPOSURE to secondhand smoke(SHS)—also known as environmental tobaccosmoke—still occurs. According to the U.S. SurgeonGeneral, “Approximately 30% of indoor workers inthe U.S. are not covered by smoke-free workplacepolicies” (DHHS, 2006). Occupational safety profes-sionals and agencies have investigated SHS exposuresto workers in the hospitality industries (e.g., restau-rants, lounges, casinos), as well as in offices or envi-ronments where smokingwas allowed (Trout, Decker,Mueller, et al., 1998; Repace & Homer, 2005; Repace,Hughes & Benowitz, 2006b).

For example, professionals investigating indoorair quality (IAQ) complaints have long known thatone must rule out the possible contribution of smok-ing to the environmental and health impacts beingevaluated (Hodgson, 1989). Given the potential mag-nitude and concern for SHS in occupational environ-ments (DHHS, 2006; CDC, 2007; Goodwin, 2007;WHO, 2007), SH&E professionals must continue tocharacterize SHS exposure levels in the workplace.

[This is particularly true given the fact that legalactions have been initiated by nonsmoking employ-ees against companies that fail to consider SHS expo-sures (e.g., see primer on legal actions againstCasinos found at www.wmitchell.edu/tobaccolaw/documents/casino.pdf). In fact, it is believed thatsuch litigation will only increase (Sweda, 2001)].

It is also conceivable that biomonitoring ofemployees could mitigate claims of respiratoryinjuries or cancers against employers if it were deter-mined that the worker’s personal smoking habit—off the job, and not the worksite—was a more likelycausal factor. The monitoring of tobacco smoke mayalso be of interest to SH&E professionals whenchemicals in SHS are routinely being monitored forin the workplace (e.g., carbon monoxide).

For the SH&E professional, quantitation of work-er exposure to SHS is a logical step toward under-standing the scope and depth of this issue.One potential tool for exposure assessment thatcould overcome the potential limitations of area orpersonal monitoring is biomonitoring. Biologicalmonitoring, when using a validated methodology(Lauwerys & Hoet, 2001; Sexton, Callahan & Bryan,1995), moves the measurement from the potential ofexposure (the opportunity for interaction betweenthe worker and the chemical) to a more direct meas-ure of dose (the specific quantity of the chemicalabsorbed systemically by the worker). This not onlyindividualizes each worker’s exposure, it also elimi-nates the question of whether areamonitoring is suf-ficient to capture each worker’s exposure. For SHSexposures, biomonitoring is performed via serum orurinary nicotine metabolite (cotinine) measure-ments. (See for example ACGIH’s Introduction tothe Biological Exposure Indices at www.acgih.org/products/beiintro.htm.)

Furthermore, biomonitoring for specific, exposedemployees might well reduce expenditures whencompared to wholesale monitoring of an entire site.Thus, the utility of determining SHS exposurethrough a simple, cost-efficient biomonitoring proto-col via urinalysis should be attractive to practitioners.

Short of an enforced smoking ban in all places ofbusiness (and all places a worker is expected to be as

Harry R. James is chief indoor air quality (IAQ) and industrial hygiene fieldinvestigator for TERRA Inc., a toxicology consulting firm in Tallahassee, FL. He performsinvestigations, monitors workers, writes site safety plans and manages remediation of

workplaces suspected of having IAQ problems and performs Phase I and II siteassessments. He holds a B.A. from the University of Utah and is pursuing a master’s

degree in occupational safety and health from Columbia Southern University.

Lisa Barfield, M.S., is a toxicologist with TERRA Inc. She holds an M.S. inInterdisciplinary Toxicology from University of Arkansas for Medical Sciences. Barfield

has 17 years’ experience in toxicological and environmental services, includingtoxicology, human health risk assessment, environmental regulatory analysis, product

stewardship, human health hazard assessment and product liability.

Janice K. Britt, Ph.D., is a toxicologist with and vice president of TERRA Inc. She hasmore than 15 years’ experience in the field of toxicology. Britt holds a B.S. in Zoology

and a Ph.D. in Toxicology from Texas A&M University. Her primary focus at TERRA is onthe human health impacts of various chemicals as well as conducting risk assessments.

Robert C. James, Ph.D., is a toxicologist with TERRA Inc. and an associate scientist withthe Center for Environmental and Human Toxicology at the University of Florida, wherehe conducts research and teaches graduate-level courses in toxicology and risk assess-

ment. James holds a B.S. in Chemistry and a Ph.D. in Pharmacology from the Universityof Utah. He is a chief author and editor of Principles of Toxicology: Environmental and

Industrial Applications (2nd ed.) and is a member of the Society of Toxicology, theSociety of Environmental Toxicology and Chemistry, and the Society of Risk Analysis.

034_044_JamesetlFeature_0908:Layout 1 8/12/2008 12:49 PM Page 34

www.asse.org SEPTEMBER 2008 PROFESSIONAL SAFETY 35

Given this endorsement, SH&Eprofessionals per-forming assessments of workforce SHS exposuremight employ these equations, believing they pro-vide reliable estimates of discrete SHS exposure forspecific individuals. However, Repace, Al-Delaimeyet al. (2006) concede that these equations tend to bet-ter predict group mean or median values and thatthey are less accurate for individual values as onemoves away from the central tendency value of thedistribution (Repace, Jinot, Bayard, et al., 1998).

Therefore, the purpose of this article is to performa simple validation assessment of the model and itsreported predictive utility. This was achieved byemploying a NIOSH study dataset that containscontemporaneous measurements of most end-points/outputs of the model’s equations. Thus, theNIOSH study provides the kind of data that wouldbe useful for determining how reliably one canextrapolate or predict SHS exposure that resulted inthe biomonitoring measurement one has taken.

The NIOSH Data Used for ValidationNIOSH conducted a health hazard evaluation

(HHE) of a group of nonsmoking employees ofBally’s Park Place Casino Hotel in Atlantic City, NJ(Trout & Decker, 1996; Trout, et al., 1998—hereafterreferred to as the NIOSH casino study). In this study,samples were collected during a 2-day exposureinterval. These included area respirable particulatemonitoring, area and personal monitoring of airnicotine, and biological measurements of environ-mental tobacco smoke (ETS) exposures consisting ofboth pre- and postshift serum and urine cotinine lev-els. (The casino workers also answered a question-naire that provided information on their estimatedduration of exposure to ETS while at home and atwork on the day that samples were taken, and for

the 3 days before sampleswere collected.)

This NIOSH dataset is thetype of SHS exposure sce-nario an SH&E professionalcould be faced with in a non-industrial setting. Repace, Al-Delaimy, et al. (2006) suggestthat by using their model onecan reliably move betweenair measurements of SHSexposure and biomonitoringmeasurements that reflect thenicotine levels associatedwith that exposure. Physicaland pharmacokinetic (PK)models are given that enableintercomparison of studies ofSHS exposure using atmos-pheric markers (CO, RSP andnicotine) and biomarkers(serum saliva, urine cotinineand hair nicotine). Accordingto Repace, Al-Delaimy, et al.,the dataset in the NIOSH

a function of his/her job), how can one determine therelative safety of the working population as it relatesto SHS? How does an epidemiologist or public healthofficial quantify exposure in a free-ranging populationto determine the relative contribution of SHS to healthimpacts? Not all people work in specific, controlledenvironments. For example, installers, repair person-nel and contractors may visit homes and private busi-nesseswhere smoking is allowed; thus, theymight notbe ideal subjects for traditional industrial hygieneevaluation and control methodologies. Mobile em-ployees or those who work outdoors and outsidedirect management control might themselves smoke,which could skew exposure data.

Abenefit to having a noninvasive biological mon-itoring program thatwould readily identify exposurepatterns to SHS would be the evaluation of ventila-tion systems in various settings. If nicotine and res-pirable suspended particles (RSP) concentrationswere to be accurately predicted by cotinine in urine,the engineer responsible for ventilation would haveexcellent data on the efficacy of his/her work with-out going through the expense of personal and areamonitoring. Job safety assessment would beenhanced with such exposure data as well. Workerhealth programs could be tailored to monitor forthose health effects seen in overexposed populations,and ruling out SHS contribution to the total environ-mental load of a worker’s exposure to other contam-inants would allow professionals to concentrate onother likely sources of exposure (e.g., PAHs, amines,aldehydes, RSP and other chemicals that come fromSHS) (DHHS, 2006). Currently, there is no biologicalmonitoring standard for SHS exposure.

A recent publication by Repace, Al-Delaimy andBernert (2006) presents a series of simple equationsthat the authors refer to as the Rosetta Stone calcula-tions (hereafter referred to as themodel or the model’s equations).These authors suggest that a reli-able mathematical relationshipfor predicting certain atmospher-ic markers of SHS from singlecotinine biomonitoring measure-ment now exists. Assuming theseequations could be shown as reli-ably predictive, the equationsmight be an effective tool forscreening or determining themagnitude of worker exposure toSHS, especially the urinary andsaliva cotinine as they are nonin-vasive measurements. Repace,Al-Delaimy, et al. state that usingthese equations, differentmarkersfor SHS exposure can be reason-ably estimated if just one surro-gatemarker ismeasured, and thatthese equations then permit“intercomparison of clinical andatmospheric studies of SHS forthe first time.”

Abstract: SH&E profes-sionals may be asked todetermine secondhandsmoke (SHS) exposuresas a part of workersafety or exposureassessments. This articlecritically examines arecently proposed setof equations for pre-dicting concentrationsof SHS markers in air. Itconcludes that suchmodels should not beemployed until theyare fully validated andasserts that use ofmeasured, population-specific, SHS concentra-tions should remain thepreferred practice injob safety assessmentfor this potential work-place hazard.

CotinineCotinine is a metabolite formed bythe body from nicotine, which is achemical in tobacco smoke andchewing tobacco. Levels of cotininein blood indicate the amount ofexposure a person has had totobacco smoke. A laboratory testcan measure cotinine in blood, urineor saliva.

Cotinine ExposureNicotine gets into people’s bodies

if they:•Smoke or chew tobacco. All

people who smoke have cotinine intheir bodies.

•Are exposed to secondhandtobacco smoke (also called environ-mental tobacco smoke).

•Are involved in tobacco produc-tion and handle tobacco.



nicotine) and personal air (nico-tine only) monitoring sampleswere conducted for approxi-mately 8 hours during eachworkshift, and the reported airconcentration results wereexpressed as 8-hour time-weighted averages (TWAs).Subject #8 was considered anactive smoker (due to highpreshift urine and serum coti-nine concentrations) and wassubsequently eliminated fromanalyses by the study authorsand from the present analysis.

NIOSH staff statisticallycompared pre- and postshiftserum cotinine levels in theremaining 28 participants whoreported SHS exposure outsideof their work to those whoreported no SHS exposure out-side of work (Trout, et al., 1998).No statistically significant dif-ference was observed, indicat-ing that the cotinine levelsmeasured in these individualswere not significantly influ-enced by nonwork SHS expo-sures. There also were nostatistically significant differ-ences inmeasurements betweenthe two shifts (Trout & Decker,1996), allowing an analysis to beconducted using combined datafrom both shifts. Table 1 pro-vides a summary of the NIOSHcasino study’s individual work-er nicotine/cotinine data. Table2 lists data from area nicotineand RSP air monitoring. Resultsindicated air nicotine concentra-tions from personal monitors (4to 15 µg/m3) and area monitors(6 to 16 µg/m3) to be compara-ble across their ranges.

Application of the Model’s Equationsto the NIOSH Study Data

Repace,Al-Delaimy, et al. (2006) proposed the fol-lowing equation to model the relationship betweenurine cotinine and air nicotine:

aU = KϕαρδrHN/δt Vu [Equation 1]Where:U = cotinine in urine, ng/mLK = conversion factor of 1,000 ng/µgϕ = nicotine to cotinine conversion efficiency by

liver enzymes, 0.78 (unitless)α = nicotine absorption efficiency, 0.71 (unitless)ρ = typical adult respiration rate for light activity,

1 m3/hourδr = renal cotinine clearance rate, 5.9 mL/min

casino study, which collected both SHS exposuremeasures and biomonitoring samples from a singleworkplace, should provide a means to evaluate thereliability of the model’s equations because all ofthese measurements were taken contemporaneouslyand need no qualification. One should be able to cal-culate the model’s equations forwards or backwardsto determine their predictive utility using this dataset.

Twenty-nine nonsmoking casino employees partic-ipated in the study. They were working one of two8-hour shifts. Participants provided pre- and postshifturine samples; 28 of 29 also provided pre- and post-shift serum samples for cotinine analysis. The urinecotinine concentrations were total cotinine (free coti-nine plus cotinine glucuronide). Area air (RSP and

Table 1Table 1

NIOSH Casino Study,Individual Measurements

034_044_JamesetlFeature_0908:Layout 1 8/12/2008 12:49 PM 6


plus cotinine that is conjugated with glucuronide. InRepace and Homer, the method for converting urinetotal cotinine to urine free cotinine is to divide urinecotinine by 1.5 (effectively multiplying total cotinineconcentrations by 0.67 to yield the free cotinine lev-els). As noted, in the unpublished report (Repace),urine total cotinine concentrations were multiplied

δt = total cotinine clearance rate, 64 mL/minVu = adult daily urine output, 1,300 mL/dayH = hours exposureN = air nicotine concentration, µg/m3

aFrom equation 4, p. 182 and Table 1, p. 184 ofRepace, Al-Delaimy, et al. (2006).

Solving for N from measured urine cotinine, theequation is as follows:

N = U* δt Vu/KϕαρδrH [Equation 2]

Using the default pharmacokinetic parametersand adult respiration rate for light activity, the equa-tion becomes:

N = U*25.46/H [Equation 3]

They then presented equations that encompassedthe above-modeled relationship between urine coti-nine and air nicotine, as well as the relationshipbetween air nicotine and othermarkers of SHS in air.The equations that are applicable to adult exposuresare shown in Table 3. For some conversion equations(e.g., carbon monoxide, polycyclic aromatic hydro-carbons, salivary cotinine), there is no correspondingmeasured data in the NIOSH casino study, so theseequations could not be evaluated. The remainingapplicable set of equations evaluated was: RSP fromair nicotine (Equation 4), serum cotinine from urinecotinine (Equation 5), and air nicotine from plasma(serum) cotinine (Equation 6).

Estimating Urine Free CotinineConcentrations From Urine Total CotinineConcentrations

It was unclear whether Repace, Al-Delaimy, et al.(2006) used total or free urine cotinine as the inputvalue, and whether cotinine present in preshift sam-ples was to be subtracted from postshift samples.This is noted because an unpublished study of casinoworkers (Repace, 2006) that initiated this compara-tive analysis estimated free urinary cotinine levelsfrom the measured total cotinine using a 0.508 con-version factor. Thus, it was surprising that the articledescribing themodel did notmention the need to usefree or total cotinine values when estimating air nico-tine concentrations from urinary cotinine measure-ments. Moreover, in two other recent SHSstudies by the model’s authors (Repace &Homer, 2005; Repace, Hughes, et al., 2006)the equations were applied to estimateSHS exposures in nonsmoking individualsfrom urine cotinine. However, in all threestudies, different, additional proceduresfor applying the model’s equations arerevealed that are not presented in theRepace, Al-Delaimy, et al. publication.These additional procedures include:

1) The conversion of urine total coti-nine to free cotinine must be made, appar-ently because the equations were derivedbased on free cotinine (Repace & Homer,2005; Repace, 2006). Total cotinine meas-ured in urine is the sum of “free” cotinine

Table 2Table 2

NIOSH Casino Study,Area Measurements

Table 3Table 3

Rosetta Stone Conversion Equationsfor Adult SHS Exposure



average air nicotine concentration cannotbe accurately calculated from a single,postshift urinary cotinine measurement.

3) In Repace andHomer (2005), subjectswere instructed to avoid SHS as much aspossible for 5 days before their exposure inthe bar. In Repace (2006), subjects hadworked several days before their exam-ined shift exposure and were, therefore,more similar to the postshift exposure pat-tern seen in the NIOSH casino study.

4) Repace, Hughes, et al. (2006) was astudy similar to that of Repace and Homer(2005), in which urine cotinine was meas-ured in 8 volunteers before a 6-hour visit toan area bar, followed by two postvisit sam-ples at 2 and 12 hours. To obtain the urinecotinine concentration used to estimateexposure to SHS-RSP, background (e.g.,preshift) cotinine was subtracted from thepostshift concentrations and the two back-ground-corrected postshift samples werethen averaged. Repace, Hughes, et al.(2006) did not discuss the issue of free ver-sus total urinary cotinine in that study, andthemethod the study authors cited describ-ing the urine sample analysis (Bernert,Turner, Pirkle, et al., 1997) appears to be aserum cotinine method.

These disparatemethods are referred toas Method A (based on Repace & Homer,2005) and Method B (based on Repace,2006). In Method A, urine free cotinine iscalculated by dividing total urine cotinineby 1.5, and the equations are run on thedifference between pre- and postshift coti-nine concentrations. In Method B, urinefree cotinine is calculated by multiplyingtotal urine cotinine by 0.508, and applica-ble equations are applied to the postshiftconcentrations. These two methodologieswere then applied to the NIOSH casinostudy urine cotinine measurements to esti-mate the concentration of urine free coti-nine to be used in subsequent calculations.The estimates produced by bothMethodA

and Method B are shown in Table 4.The concentrations derived using Method A and

Method B differ considerably. Method A is a deter-mination on the net change of cotinine from pre- andpostshift—that is, contribution from preshift expo-sure is subtracted out. Method B does not include acorrection for contributions from preshift exposure,even though this should occur for individuals withknown exposures from shifts from previous days. Ifthis correction is not made, the air nicotine and RSPcalculated is greater than that from any one shiftbecause not all cotinine derived from metabolism iscleared in 1 day. Thus, postshift serum and urinaryconcentrations in workers should typically increasefor each day of exposure during the week before themeasurement is taken.

by 0.508 to calculate urine free cotinine levels. Thissingle difference in the application of just one of themodel’s equations can alter the final RSP/nicotinecalculations by 24 to 32%, depending on the correc-tion factor employed.

2) Repace andHomer (2005) took one preshift andtwo postshift samples (at 2 and 9 hours postexpo-sure). The two postshift samples were averaged,preshift concentrations were subtracted from thepostshift averages, and the calculations were appliedto the net difference. In contrast, in Repace (2006)only postshift urine samples were taken for cotinineanalysis, so the equations were applied to the uncor-rected postshift urine cotinine measurement. Thetiming of the postshift sample collection varied fromabout 6.5 to 23 hours after the shift end. Thus, theRepace and Homer report shows why a worker’s

Table 4Table 4

Conversion of NIOSH Casino StudyUrine Total Cotinine to Urine FreeCotinine, Methods A & B



measurements in the subset consisting of the 17NIOSH casino study workers for which such monitor-ing was performed (Table 6, p. 40). Based on the arith-metic averages, the model’s equations, when usingMethod A, overestimated air nicotine concentrationsby about 2 to 3-fold. When using Method B, the over-estimate was approximately 6 to 7-fold. For subjectswith airmonitoring data, all air nicotine estimateswereexaggerated regardless of the calculationmethod used.

In addition, as no method for converting “total”cotinine to “free” cotinine is described in Repace,Al-Delaimy, et al. (2006), anyone relying solely upon themodel will generate even higher overestimations

Three of the NIOSH casino study subjects hadpreshift urine concentrations that were higher thanpostshift concentrations. This is likely because ofsome preshift exposure to SHS that was higher thanthe exposure during the examined shift. Taking bothdifferences in the methods into consideration,Method B results, on average, in much higher urinecotinine concentrations when multiple exposureshave occurred, and so, higher estimated SHS expo-sure levels thanMethodA. By dividing total cotinineby 1.5 in Method A, the average estimated free coti-nine for a given total cotinine concentration isapproximately 1.3 times greater than the free coti-nine estimation forMethod B (multi-plying by 0.508).

Estimating AirNicotineConcentrations

The model sug-gests that air nico-tine concentration(in units of µg/m3)can be estimatedfrom urine free coti-nine (ng/mL) usingEquations 1, 2 and 3.Equation 3 was ap-plied to the NIOSHstudy urine free coti-nine estimated byboth Methods A andB. For Method A, airnicotine estimateswere only derivedfor those subjectswith an increase inurine cotinine pre- topostshift; those witha negative change inurine cotinine wereexcluded from theestimate.

The results areshown in Table 5.The mean estimatedair nicotine fromMethod A urine freecotinine was 26.2µg/m3 (8-hour TWA)and from Method Burine free cotininewas 67 µg/m3 (8hour TWA).

The predictiveaccuracy of modelEquation 3 was test-ed by comparingestimated nicotineconcentrations topersonal monitor

Table 5Table 5

Estimating NIOSH Casino Study Air NicotineConcentrations From Urine Free Cotinine Levels



The results of this compara-tive analysis are shown inTable 7. The predictive abilityof this equation was thenexamined by comparing esti-mated RSP concentrations tothe area RSP measurements(Table 8, p. 42). This compari-son indicates that mean RSPestimated from air nicotine,which was in turn estimatedfrom urine free cotinine levels,was 4.71 times greater usingMethod A and 12 times greaterusing Method B than thereported, measured mean forarea RSP. The mean estimatedRSP derived using measurednicotine concentrations via per-sonal monitors was 1.72 timesgreater than the measured RSP.

As occurred with the com-parisons between estimated andmeasured air nicotine, using themodel’s equations to estimatepersonal or area RSP exposuresfor SHS from the workers’ freeurine cotinine measurements(using either Method A or B)results in a significant overesti-mate of SHS exposure. Themore levels of estimationinvolved (RSP estimation fromurine cotinine versus RSP esti-mation from air nicotine), thegreater the degree of SHS over-estimation. All group and aver-age measurements, and almostevery individual measurement,were exaggerated when the

model’s equations were used to predict them. Forexample, the maximum estimated RSP exposure for aNIOSH casino study subject (from Method B urinefree cotinine) was 3,185 µg/m3, which is 35 times themaximum measured value of area RSP (90 µg/m3).The lowest estimated RSP exposure for a NIOSH casi-no study subject was 63 µg/m3, greater than 3 timesthe lowestmeasured air RSP concentration reported inthe NIOSH casino study (< 20 µg/m3).

The range of urine total cotinine measurements inthe NIOSH casino study subjects varied about 40 fold(range: 3.87 to 197 ng/mL) from exposures that variedonly about 2-fold for nicotine air concentrations andonly about 3-fold for air RSP concentrations. Thus,individual worker variations in the inhalation, absorp-tion andmetabolismof the respective roomair nicotineconcentrations, when extrapolated on an individualbasis, suggested a greater variation in roomair nicotineand RSP concentrations than actually existed.

This variation is important as the implied purposeof the model’s equations is to obviate the need to takeserum nicotine/cotinine and air nicotine and RSP

should they inadvertently apply these equations to atotal cotinine measurement. This comparison indi-cates that the equations cannot be used to reliablyestimate air nicotine exposures from urine cotinineconcentrations, either for a group or for a particularindividual. Application of this model to urine coti-nine measurements tends to result in a significantoverestimate of worker SHS exposure.

Estimating Air Nicotine ConcentrationsFrom RSP Concentrations

The model provides the following equation forestimating RSP in air from air nicotine:

RSP = 10x (nicotine concentration in air)[Equation 4] (Ux25.46/8x10)

RSP concentrations were estimated using:1) The estimated nicotine concentrations from the

NIOSH casino study based on estimated urine freecotinine measurements (Methods A and B);

2) The measured area nicotine concentrations inthe NIOSH casino study to derive an estimated RSPconcentration.

Table 6Table 6

Personal Monitor Nicotine ConcentrationsCompared to Estimated ConcentrationsFrom Urine Cotinine



Where:inhalation rate = assumed 1 m3/hour;exposure duration = for this analysis, assumed8-hour work shift.Although not stated clearly in Repace, Al-Delaimy,

et al. (2006), this equation should be applied to the dif-ference between a pre- and postshift serum concentra-tion for the most accurate results. This equation was,therefore, applied to theNIOSH casino study postshiftserum cotinine concentrations, both estimated andmeasured, to predict air nicotine concentrations. Theseestimated concentrations were then compared to themeasured nicotine concentrations taken from personalair monitoring samples from the NIOSH casino study.

measurements by simply taking a worker urine sam-ple and measuring its cotinine level. However, to doso will increase the variation in the estimates of SHSexposure, substantially increase the maximal estimateand, thereby, increase even the average estimate ofexposure. In short, there is an obvious potential for theSH&E professional or epidemiologist to overestimatean individual’s SHS exposure substantially when onesimply applies themodel’s equations to urine cotininemeasurements, for either the individual or the group.

Serum Cotinine ConcentrationsEstimated From Measured Urine CotinineConcentrations

The ability of the model topredict or estimate the serumcotinine concentrations fromone’s urine cotinine concentra-tions [Equation 5: serum coti-nine = urine free cotinine/6.5]was examined using theNIOSHcasino study data. This equationwas applied to preshift andpostshift urine free cotinine con-centrations as estimated byMethod B (urine total cotinine x0.508 = urine free cotinine).Based on group averages,Equation 5, when applied tourine free cotinine derived byMethod B, overestimates meas-ured serum cotinine by about1.5- to 1.6-fold. A review of indi-vidual preshift measurementsindicated that serum cotininewas overestimated in 18 of 27individuals; postshift measure-ments indicated serum cotininewas overestimated more fre-quently (24 of 27 individuals),and sometimes by more than 5-fold. This procedure also doesnot allow one to consider thepossibility that the last exposureinterval serum cotinine concen-trations might have been lowerthan prior exposure intervals,which in fact occurred in 5 sub-jects in the NIOSH casino study(subjects 1, 4, 13, 14 and 17).

Estimating Air NicotineConcentrations FromSerum Cotinine

The following equation de-scribes the relationship be-tween serum cotinine and airnicotine concentrations:

(serum cotinine) = 0.006 x(inhalation rate) x (exposureduration) x (air nicotine con-centration) [Equation 6]

Table 7Table 7

Estimation of RSP ConcentrationsFrom Estimated Nicotine Concentrations



ConclusionWhether SHS can be determined to be a health

hazard, either by itself or as a contributor to totalparticulate or chemical exposure, worker exposureto it will be laid at the feet of the professional whomakes no effort to characterize these exposures andaddress them accordingly. Worker exposure to car-cinogenic or debilitating agents can generally beeffectively monitored through personal or areameasurements of the offending agent.

This is the basis for deriving and setting safeworkplace exposure values such as threshold limitvalues as controls for chemical exposures. Some bio-logical monitoring of chemical metabolites has alsobeen accepted as a tool for ensuring that overexpo-sure does not occur (e.g., benzene exposure via phe-nol in urine and blood lead measurements). Controlof exposures deemed excessive by comparison to therecommended exposure guideline can then proceedusing traditional safety engineering principles.

In the present analysis, several compounds ormetabolites collected while monitoring workersexposed to SHS were evaluated by comparing thesemeasured values to predicted values for each expo-sure marker generated using a model proposed byRepace, Al-Delaimy, et al. (2006).

The comparative analyses, using contemporane-ous data generated byNIOSH of all themodel’s sug-gested surrogate marker compounds, revealed thatthis model (or more precisely, this collection of sim-ple, algebraic equations) cannot be used to accurate-ly estimate one surrogate marker of SHS exposurefrom the measurement of another surrogate markerof SHS exposure.

In fact, the comparative analysis indicated that theNIOSH casino study data could not be reasonablypredicted by themodel’s equations to any acceptabledegree of accuracy—neither for individual measure-ments nor for the statistical characterizations ofgroup exposures (e.g., mean values). Because of theseinaccuracies, which occur if one moves from a bio-marker chemical measured in a worker’s serum orurine to airborne concentrations of chemicals or viceversa, the authors strongly recommend that this

The results of these comparisons show that theestimated air nicotine concentrations derived usingthe model’s equations predicted concentrationshigher than those actually measured by NIOSHinvestigators. A comparative analysis (data notshown but available upon request), indicated thatmeasured nicotine concentrations from personalmonitoring samples averaged 4.1-fold lower thanthose estimated via the corresponding model equa-tions. The overestimation was compounded whenthe model’s air nicotine concentration was derivedfrom an estimated serum concentration that in turnwas derived from a urinary free cotinine concentra-tion. Air nicotine levels estimated in this manneraveraged 7-fold higher than measured values (com-parison not shown but available upon request).

Calculating Serum & Urinary CotinineConcentrations From Measured Air NicotineConcentrations

As a final examination of the model’s predictiveutility, the equations were used to reverse-calculateindividual serum and urine cotinine concentrationsfrom personal nicotine air measurements taken inthe NIOSH casino study. Equation 3 was used toestimate the urinary cotinine concentration from themeasured air nicotine concentration, and Equation 6,once rearranged, was used to estimate the serumcotinine concentration from the measured air nico-tine concentration. These estimated values werethen compared to the measured concentrations(again these results are not shown in the interest ofspace but can be made available upon request).

Based on group averages, estimating a serumcotinine concentration from an air nicotine measure-ment resulted in a 4.4-fold underestimate of theserum cotinine. Estimating a urine cotinine concen-tration from an air nicotine measurement results inan 8-fold underestimate. Just as the model’s equa-tions overestimate air concentrations of nicotine andRSP from urine and serum cotinine measurements,the equations conversely underestimate urine andserum cotinine concentrations when using a meas-ured nicotine air concentration.

Table 8Table 8

Comparing the Average Estimated RSP Concentrationsto the Average Measured Area Monitoring ResultsFrom the NIOSH Casino Study



nicotine to cotinine and the processes for eliminatingcotinine from the blood compartment.

As an example of this, comparative analyses usingthe NIOSH casino study data showed that measuredRSP and nicotine exposure levels which varied nomore than 3-fold for the group, resulted in a greaterthan 40-fold variation in the magnitude of group uri-nary cotinine concentrations. No simple algebraicequation, which always multiplies a single constant

model or set of equations not be used in either theworkplace or in public health studies.

Aswritten, themodel’s equations tend to underes-timate the serum and urinary concentrations of coti-nine that an individual generates when s/hemetabolizes inhaled nicotine, and correspondinglyexaggerate inhaled nicotine and RSP levels whenbased on an individual’s or group’s serum/urinarycotinine measurements. The perceived utility of thismodel’s equations isindirect measure-ment of SHS (i.e.,to use the noninva-sive urinary cotininemeasurement to pre-dict SHS inhalationexposure).

However, use ofthismodel by SH&Epersonnel in theworkplace will onlyoverstate SHS expo-sure to a substantialdegree. Thus, safetyassessment is bestperformed by adirect measurementof accepted SHSmarkers.

The reason for theinaccuracy of themodel’s equationsrests largely in thefact that there is con-siderable interindi-vidual variation inthe metabolism ofnicotine to cotinine.Available data (Ben-owitz, 1996; Huk-kanen, Jacobs &Benowitz, 2005) onthe metabolism ofnicotine indicatesthat this variation istoo great to be able toconvert cotinine toair nicotine levelswith a single conver-sion factor as themodel’s equationsattempt to do.

Thus, the sugges-tion that a single,linear relationshipexists is incorrect andis an oversimplifica-tion of the biologicprocesses involvedin the absorptionof nicotine, the bio-transformation of

Table 9Table 9

Comparing Measured NIOSH Casino StudySerum Cotinine Levels to Serum Cotinine LevelsEstimated From Urine Free Cotinine (Method B)



predictions for more traditional industrial hygieneair monitoring methods for use in an exposureassessment. The poor predictive ability demonstrat-ed by this model is an example as to how proposingor using models that have not been properly vali-dated can easily lead to inaccurate safety assess-ments, improper attribution of morbidity ormortality in epidemiological studies, and flawedrisk management decisions. �

ReferencesBenowitz, N.L. (1996). Cotinine as a biomarker of environmen-

tal tobacco smoke exposure. Epidemiologic Reviews, 18(2), 188-204.Centers for Disease Control and Prevention (CDC). (2007,

May 25). Centers for Disease Control: Exposure to secondhandsmoke among students aged 13-15 years—worldwide, 2000-2007.Morbidity and Mortality Weekly Report, 56(20), 497-500.

Goodwin, R.D. (2007, May). Environmental tobacco smokeand the epidemic of asthma in children: The role of cigarette use.Annals of Allergy, Asthma & Immunology, 98(5), 447-454.

Hodgson, M.J. (1989). Environmental tobacco smoke and thesick building syndrome. Occupational Medicine: State of the ArtReviews, 4(4), 835-840.

Hukkanen, J., Jacobs, P. & Benowitz, N.L. (2005). Metabolismand disposition kinetics of nicotine. Pharmacological Reviews, 57,79-115.

Lauwerys, R.R. & Hoet, P. (2001). Industrial chemical exposure:Guidelines for biological monitoring (3rd ed.). Boca Raton, FL: Lewis.

Needham, L.L., Pirkle, J.L., Burse, V.W., et al. (1992). Casestudies of relationship between external dose and internal dose.Journal of the Experimental Analysis of Behavior [EpidemiologySupplement(1)], 209-221.

Repace, J.L. (2006, March 14). Report on the Empress HorsehoeCasino (unpublished). Bowie, MD: Repace Associates Inc.

Repace, J.L., Al-Delaimy, W.K. & Bernert, J.T. (2006).Correlating atmospheric and biological markers in studies of sec-ondhand tobacco smoke exposure and dose in children and adults(with erratum). Journal of Occupational and Environmental Medicine,48(2), 181-194.

Repace, J.L. & Homer, M. (2005). Laramie, Wyoming bar patronstudy (WYSAC Technical Report No. CHES-517). Laramie, WY:Wyoming Survey and Analysis Center, University of Wyoming.

Repace, J.L., Hughes, N. & Benowitz, E. (2006). Exposure tosecondhand smoke air pollution assessed from bar patrons’ uri-nary cotinine. Nicotine & Tobacco Research, 8(5), 701-711.

Repace, J.L., Jinot, J., Bayard, S., et al. (1998). Air nicotine andsaliva cotinine as indicators of workplace passive smoking expo-sure and risk. Risk Analysis, 18(1), 71-83.

Sexton, K., Callahan, M.A. & Bryan, E.F. (1995). Estimatingexposure and dose to characterize health risks: The role of humantissue monitoring in exposure assessment. Environmental HealthPerspectives, 103(Suppl 3), 13–29.

Sweda, E.L. (2001). Litigation on behalf of victims of exposureto environmental tobacco smoke: The experience from the USA.The European Journal of Public Health, 11(2), 201-205.

Trout, D. & Decker, J. (1996). Bally’s Park Place Casino Hotel,Atlantic City, NJ (HETA 95-0375-2590). Cincinnati, OH: NIOSH.

Trout, D., Decker, J., Mueller, C., et al. (1998). Exposure ofcasino employees to environmental tobacco smoke. Journal ofOccupational and Environmental Medicine, 40(3), 270-276.

U.S. Department of Health and Human Services (DHHS).(2006). The health consequences of involuntary exposure to tobaccosmoke: A report of the Surgeon General.Washington, DC: Author,Centers for Disease Control and Prevention, National Center forChronic Disease Prevention and Health Promotion, Office onSmoking and Health.

World Health Organization (WHO). (2007). Only 100%smoke-free environments adequately protect from dangers of sec-ondhand smoke [Press release]. Geneva, Switzerland: Author.Retrieved July 25, 2008, from http://www.who.int/mediacentre/news/releases/2007/pr26/en/index.html.

value by the measured concentration of one surro-gate marker of SHS exposure that has been meas-ured, can capture this 40-fold variation. Thus, onecannot use the model’s equations to calculate anyindividual’s exposure level or that of a group of indi-viduals. That is, the model lacks the requisite speci-ficity for application on a worker-by-worker basis.

For the equation to accurately predict an individ-ual’s RSP/nicotine exposure level, one would haveto know the rate of cotinine formation for each indi-vidual. For group exposures, one would have toknow how nicotine metabolism of each individualgroup member varied before the correct individualvalues and corresponding correct group mean val-ues could be calculated. Unfortunately, the model’sequations do not and cannot account for theinterindividual variations that occur in the metabo-lism of nicotine to cotinine. This is a key reason as towhy these equations lack predictive value.

Additional problems were identified for anySH&E professional using urinary cotinine measure-ments to predict exposures with these equations. Forexample, the inhalation exposure estimates derivedby these equations are easily skewed if the numberof individuals being sampled is not known to be rep-resentative of the larger group, especially if one ormore large urinary cotinine levels is recorded.

If one randomly selects two subsets of five indi-viduals in the NIOSH casino study, the predictedsubgroup mean RSP levels will likely differ. Thisoccurs because a small subset of workers may notcapture the full extent of variation in urinary coti-nine levels occurring in the cohort (or the generalpopulation) from which the subset was selected.Many of the pharmacokinetic or physiologic param-eters used in the model’s equations will also varyacross individuals, compounding the known prob-lem with interindividual variations in the rate ofnicotine metabolism. There does not appear to be aneasy way to resolve the many identified limitationsinherent to this model.

The present attempt to validate the model usingthe NIOSH casino study data revealed a high errorrate. Applying the model’s equations overstated theNIOSH-measured RSP levels in at least 27 of 28 indi-viduals (only 1 of 28 individuals yielded a urinarycotinine level that resulted in a calculated RSP levelthat arguably fell within the range measured byNIOSH investigators).

Thus, for this study, the proposed calculationsyield a 96% (or greater) error rate when estimatingthe RSP levels from the collected urinary cotininesamples. Forworkplace safety and riskmanagementdecisions, such a high error rate in an exposureassessment tool is unacceptable. This high error ratedemonstrates, as much as any line of evidence, thatthe use of this model would tend to grossly overes-timate nicotine/RSP levels when calculated frommeasured urinary cotinine levels.

This analysis stresses the need for SH&E profes-sionals to carefully consider the validity of anymodel that proposes to substitute exposure/dose


james etal2008

Documents

environmental toxicology

human toxicology

workers exposure

society of toxicology

exposure assessment

suf toxicology

field of toxicology

interdisciplinary toxicology