Occupational Exposure to Carbon Nanotubes and Nanofibers
Occupational Exposure to Carbon Nanotubes and Nanofibers
Occupational Exposure to Carbon Nanotubes and Nanofibers
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
<strong>to</strong> a <strong>to</strong>xicant, <strong>and</strong> at some point, they may resultin irreversible, functional impairment of the lungs[NIOSH 1986]. If data are available, the BMRs canbe based on a biologically significant response thatis associated with, or expected <strong>to</strong> result in, a materialimpairment of health. However, there may beinsufficient data <strong>to</strong> determine a specific level thatis associated with a measurable adverse response.In that case, a statistical criterion may be used as aBMR for continuous data [Crump 1995].A statistical method (originally referred <strong>to</strong> as a“hybrid” method) is described by Crump [1995] <strong>to</strong>provide BMD(L) estimates from continuous datathat are equivalent <strong>to</strong> a 10% excess risk based ondicho<strong>to</strong>mous data, assuming that an abnormal orbiologically significant response is defined as theupper 99 th percentile of the unexposed (control)distribution. According <strong>to</strong> this method, “for a normaldistribution with constant variance, settingBMR = 0.1 <strong>and</strong> P O= 0.01 is equivalent <strong>to</strong> choosingthe BMD <strong>to</strong> be the dose that results in an increasein the mean equal <strong>to</strong> 1.1 times the st<strong>and</strong>arddeviation,” assuming a normal distribution withconstant variance [Crump 1995]. That is, if one assumesthat the probability of the specified adverseresponse in the unexposed population is the upper1% of a normal distribution of responses, then selectinga BMR of 1.1 st<strong>and</strong>ard deviations above thecontrol mean response is equivalent <strong>to</strong> a 10% BMDas estimated in dicho<strong>to</strong>mous data.In evaluating possible BMRs for the continuousdata of CNT in mice, earlier studies of chronicozone exposure in rats were examined <strong>to</strong> determineif a biologically-based BMR could be identifiedfor pulmonary fibrosis (measured as alveolarconnective tissue thickening) associated with abnormalpulmonary function [Chang et al. 1992;Costa et al. 1995; S<strong>to</strong>ckstill et al. 1995]. However,those rat findings did not appear <strong>to</strong> extrapolate well<strong>to</strong> the mice in Shvedova et al. [2005, 2008]. Thatis, the observed abnormal response in rats (associatedwith a persistent lung function deficit) was a36% increase in the control mean alveolar connectivetissue thickness [Chang et al. 1992; Costa et al.1995]; however, this amount of response occurredin up <strong>to</strong> 30% of the control (unexposed) mice inShvedova et al. [2005, 2008] (vs. 2.5% of controlsin Chang et al. [1992], in part due <strong>to</strong> the greatervariability in the alveolar tissue thickness in theunexposed mice. In addition, no data were foundof a biologically relevant BMR for the amount ofhydroxyproline in the lungs of rats or mice. In theabsence of an identified biological basis for a BMRfor the continuous response measures of alveolarconnective tissue thickening or the amount of hydroxyproline,NIOSH used the statistical criteriondescribed by Crump [1995], in which a BMR of1.1 st<strong>and</strong>ard deviations above the control meanresponse is equivalent <strong>to</strong> a 10% excess risk in thedicho<strong>to</strong>mous data, assuming the 99 th percentile ofthe distribution of control responses is abnormal orbiologically significant.That is, the BMR for the continuous data (alveolarconnective tissue thickness <strong>and</strong> hydroxyprolineamount) is defined as follows:Equation A–3:BMR = µ(d) – µ(0)where µ(d) is the mean response at the BMD (d);µ(0) is the control mean response; <strong>and</strong> BMR is thespecified number of st<strong>and</strong>ard deviations (SDs) (i.e.,1.1 in these analyses). Thus, the continuous databasedBMD is the dose associated with a 10% increasein the proportion of animals exposed at dosed with response greater than the 99 th percentile ofthe control mean response. The estimates of µ(d)<strong>and</strong> µ(0) are derived from the fitted dose-responsemodels (polynomial) (Section A.2.3.3).A.2.3.3 BMD Model FittingThe animal dose-response data were fit using thebenchmark modeling software (BMDS 2.1.2) [USEPA 2010]. The dicho<strong>to</strong>mous data were fit with amultistage (polynomial degree 2) model. This is theonly model that provided an adequate fit <strong>to</strong> the subchronicinhalation data, each of which [Ma-Hock etal. 2009; Pauluhn 2010] had only one dose betweenzero <strong>and</strong> 100% response for the endpoints evaluated(granuloma<strong>to</strong>us inflammation or alveolar septalthickening, his<strong>to</strong>pathology grade 1 or higher).104 NIOSH CIB 65 • <strong>Carbon</strong> <strong>Nanotubes</strong> <strong>and</strong> <strong>Nanofibers</strong>