Evaluation and Application of the RD50 for Determining Acceptable Exposure Levels of Airborne Sensory Irritants for the General Public

Background The RD50 (exposure concentration producing a 50% respiratory rate decrease) test evaluates airborne chemicals for sensory irritation and has become an American Society for Testing and Materials (ASTM) standard method. Past studies reported good correlations (R2) between RD50s and the occupational exposure limits, particularly threshold limit values (TLVs). Objective The main purpose of this study was to examine the relationship between RD50s and human sensory irritation responses in a quantitative manner, particularly for chemicals that produce burning sensation of the eyes, nose, or throat, based on lowest observed adverse effect levels (LOAELs) reported for human subjects. Methods We compared RD50s with LOAELs and acute reference exposure levels (RELs). RELs, developed by the California Environmental Protection Agency’s Office of Environmental Health Hazard Assessment, represent a level at which no adverse effects are anticipated after exposure. We collected RD50s from the published literature and evaluated them for consistency with ASTM procedures. We identified LOAELs for human irritation and found 25 chemicals with a corresponding RD50 in mice. Discussion We found the relationship between RD50s and LOAELs as log RD50 = 1.16 (log LOAEL) + 0.77 with an R2 value of 0.80. This strong correlation supports the use of the RD50 in establishing exposure limits for the public. We further identified 16 chemical irritants with both RD50s and corresponding acute RELs, and calculated the relationship as log RD50 = 0.71 (log REL) + 2.55 with an R2 value of 0.71. This relationship could be used to identify health protective values for the public to prevent respiratory or sensory irritation. Conclusion Consequently, we believe that the RD50 has benefits for use in setting protective levels for the health of both workers and the general population.

Although airborne chemicals can cause a number of harmful effects, the most common effect is sensory irritation (De Ceaurriz et al. 1981). Exposure to a sensory irritant may stimulate the trigeminal nerve endings and laryngeal receptors, eliciting any one or a combination of the following symptoms: burning sensation of the eyes, nose, or throat, as well as coughing sensations . Sensory irritation is also the most common end point for occupational exposure levels (OELs). For one specific OEL measure, threshold limit values (TLVs) [developed by the American Conference of Governmental Industrial Hygienists (ACGIH 2006)] are calculated based on sensory or pulmonary irritation for > 50% of the compounds. Kane et al. (1979) reported that approximately twothirds of the compounds for which they found a TLV acted as sensory irritants. A qualitative evaluation of sensory irritants indicated that sensory irritation responses in the mouse are predictive of responses in humans (Alarie 1973a).
In 1966, Alarie initially proposed the use of an animal test to evaluate the potency of airborne sensory irritants. The bioassay uses male Swiss-Webster mice to measure decreases in respiratory frequency resulting from exposure to a geometric series of concentrations of airborne irritants (Alarie 1966). The concentration inducing a 50% decrease in respiratory frequency is termed the RD 50 . From these measured RD 50 s, Alarie (1981b) ranked irritant potencies and found a good correlation (R 2 ) between RD 50 s and TLVs. The Alarie test evolved over the years and was adopted in 1984 as a standard test by the American Society for Testing and Materials (ASTM 2004). The "RD 50 test" or the "Standard Test Method for Estimating Sensory Irritancy of Airborne Chemicals" (ASTM 2004) quantitatively measures irritancy as indicated by the reflex inhibition of respiration in mice exposed to sensory irritants. For the test, four mice are first acclimatized to the chamber and are then simultaneously exposed to the airborne chemical. A sufficient number of groups are exposed to a geometric series of concentrations so that a concentrationresponse curve can be constructed from the analysis. The mice are placed in a body plethysmograph attached to an exposure chamber so that only the head is exposed to the test material. The plethysmographs are connected to pressure transducers, which sense changes created by inspiration and expiration. The amplified signals are transmitted to a polygraph recorder. The concentration of airborne irritant that produces an RD 50 is determined from the concentration-response curve constructed from the various data points obtained with a series of concentrations.
Sensory irritation is a reflex reaction from stimulation of the trigeminal or laryngeal nerve endings (Boylstein et al. 1996). The sensory irritant response is mediated through binding to the trigeminal nerve receptors and appears to follow Michaelis-Menten receptor kinetics. Although the RD 50 concentration has been described as "intolerable" to humans, as indicated in the ASTM standard, "the test method will detect irritation effects at concentrations far below those at which pathological changes are observed" (Alarie 2000;ASTM 2004). Further, as demonstrated by Barrow et al. (1986), pathologically detectable responses are expected only after prolonged repeated exposure.
RD 50 s are a basis, at least partially, for a number of OELs by ACGIH (ACGIH 2006). The calculation methodology is based on Kane et al. (1979), who evaluated data from 11 sensory irritants and concluded that a level one-hundredth of the RD 50 would produce "minimal or no sensory irritation" in humans. The current suggestion of setting OELs at 0.03 RD 50 comes from Alarie (1981aAlarie ( , 1981b, because 0.03 RD 50 is halfway between 0.1 RD 50 and 0.01 RD 50 on a logarithmic scale. Alarie (1981a) reported a strong correlation (R 2 = 0.89) between 0.03 RD 50 and OELs for the 26 chemicals tested. Subsequently, both analyses, one using 41 chemicals (Alarie and Luo 1986) and most recently another using 89 chemicals (Schaper 1993), resulted in a lower but still strong correlation (R 2 = 0.78). Although most of the applications of the RD 50 have focused on OELs, Nielsen et al. (1995) found that protection against indoor sensory irritation effects could be achieved at a level of 0.025-0.25 of the OEL. Multiple studies show strong correlations between RD 50 s and OELs, supporting the continued use of the Alarie test for establishing OELs (Kane et al. 1979(Kane et al. , 1980Schaper 1993).
In this study we examined the relationship between RD 50 s and human sensory irritation responses in a quantitative manner, particularly for chemicals that produce burning sensation of the eyes, nose, or throat, based on lowest observed adverse effect levels (LOAELs) reported for human subjects. We also analyzed the relationship between RD 50 s and OELs for identified human sensory irritants. Finally, we evaluated the relationship between RD 50 s and acute reference exposure levels (RELs) developed to protect the public (Collins et al. 2004). RELs are defined as "[t]he concentration level at or below which no adverse health effects are anticipated for a specified exposure duration [1 hr for the acute RELs]. … RELs are based on the most sensitive, relevant, adverse health effect reported in the medical and toxicological literature." A strong correlation between RD 50 s and LOAELs, TLVs, and acute RELs will support the use of RD 50 s in establishing guidance levels to protect the public from sensory irritants.

Methods
LOAELs versus RD 50 s. Using published toxicologic studies of human subjects exposed to sensory irritants, we identified human LOAELs. Criteria for selecting human LOAELs required that the studies describe mild irritating effects (Alexeeff et al. 2002) resulting from acute inhalation exposure. Published human studies on hazardous air pollutants (HAPs) served as the primary sources of information for these chemicals (Alexeeff et al. 2002 terms included the chemical name, the type of LOAEL effects (e.g., irritation), route of exposure (inhalation), and exposure duration (acute). We also conducted online searches for additional non-HAP chemicals with an identified RD 50 . Further, we conducted manual searches from secondary sources through 2005. Five criteria were developed for inclusion of a study in the analysis: a) peerreviewed and published, well-conducted industry-sponsored studies or doctoral dissertations; b) inhalation exposure; c) discrete acute exposure; d) available LOAEL for a mild adverse health effect; and e) the original research. For each human study analyzed, information about the chemical, exposure time, end-point category (eye and/or respiratory irritation), and LOAELs were recorded. If multiple mild responses were reported at various dose levels for the same chemical and exposure time, then the lowest adverse effect level was considered the LOAEL.
For RD 50 s, we first reviewed references identified from the database developed by Schaper (1993). We identified additional studies from Alarie et al. (2000). We also searched the scientific literature during the period 1992-2005 to identify newer published studies containing RD 50 s. For each identified study, we recorded information on the chemical, exposure time, species, and RD 50 . We reviewed the methodology used to attain each RD 50 for consistency with current ASTM methods (ASTM 2004); for this reason, we included studies with mice, but excluded studies with rats in this analysis.
In cases where both RD 50 s and human LOAELs were available for the same chemical, we log transformed and fit the data with a linear relationship using Microsoft Office Excel 2003 (Microsoft, Redmond, WA) and SAS version 9.1 (SAS Institute Inc., Cary, NC) for Windows. This procedure was similar to previous RD 50 comparisons (e.g., Alarie 1981b). When we found multiple LOAELs or RD 50 s for a single chemical, we considered each reported value in the analysis. Sensitivity analyses were conducted by evaluating the correlation generated from the regression of LOAELs with RD 50 value data sets, which varied by exposure time, or strain tested. We also conducted subanalyses using upper and lower respiratory tract effects.
RELs versus RD 50 s. As reported by Collins et al. (2004), the California Environmental Protection Agency (EPA) has developed 51 acute inhalation RELs. We evaluateds these RELs to identify those based on eye or respiratory irritation end points in humans, and compared with RD 50 s. Using Microsoft Office Excel 2003 (Microsoft) and SAS version 9.1 (SAS) for Windows, we log transformed and fit the data with a linear relationship.

TLVs versus RD 50 s.
For all RD 50 s used in the above analyses, we identified TLVs from ACGIH (2006). The TLVs included timeweighted averages, short-term exposure limits and ceilings. If the documentation reported more than one TLV value, we used the lowest, more protective value. A third comparison between RD 50 s and TLVs of identified human irritants, based on identification of a human LOAEL for irritation, was conducted using log-transformed data, fit with a linear relationship, and analyzed with Microsoft Office Excel 2003 (Microsoft) and SAS version 9.1 (SAS) for Windows.

Results
LOAELs versus RD 50 s. From our search, we identified 25 chemicals with 72 human acute irritation LOAELs from 49 studies ( Table 1). The adverse effects, exposure times, and information reflecting the quality of the study (e.g., placebo-control, blinding, subject selection, subject characteristics, exposure design, and data reporting) are indicated in Table 1. For the 25 chemicals identified, 63 RD 50 s were found in mice ( Table 2). The RD 50 s were based on seven mouse strains and exposure times ranging from 5 to 180 min. Figure 1 shows the correlation between RD 50 s and LOAELs for all RD 50 s identified in all strains of mice for the 25 chemicals, allowing for 198 comparisons. There is a strong overall correlation (R 2 = 0.80) between RD 50 s and human irritation LOAELs. When we conducted the analysis for Swiss-Webster mice only (Table 3), we were able to include 75 data points for 19 compounds, and the correlation decreased slightly (R 2 = 0.74). When we evaluated only the data for non-Swiss-Webster mice (Table 3), there was little change in the correlation (R 2 = 0.83). We conducted several subanalyses to consider the influence of the RD 50 study exposure duration. As indicated in Table  3 there was little influence on the R 2 . Thus, according to this analysis, the strain of mouse tested does not appear to affect this evaluation substantially. The equations do not change significantly, and the correlation is still significant for all analyses, validating the inclusion criteria used. As indicated in Table 3, we also considered several subanalyses to address the influence of the human LOAEL variability. Specifically, we considered the issue of LOAEL sensitivity, the type of irritation end point, study quality, and the duration of exposure for the human LOAEL. The only significant effect on the correlation was observed when considering human irritation end points of the lower respiratory tract; the poor R 2 appears to be attributed partly to the few number of data points (29) in the analysis.
RELs versus RD 50 s. From the 51 California acute RELs, we identified 16 that had irritation as their end point and a corresponding RD 50 .  Figure 3 shows the correlation of TLVs to RD 50 s with an R 2 value of 0.81. Thus, when focusing specifically on human irritants, the relationship between the TLV and RD 50 remains strong.

Conclusions
The focus of this paper is on the applicability of RD 50 s for human health risk assessment.
Exposure guidelines to protect workers and the public often focus on mild irritating signs or symptoms. For example, > 50% of the TLVs and > 60% of the California acute RELs based their end points on irritation (Collins et al. 2004). However, human studies from which to develop acute exposure guidance are not available for many of the hundreds of substances of concern, and therefore reliance on animal studies is necessary. The RD 50 test method is appealing because it generates data rapidly, requires minimal animal use, is low in cost, and is validated, calibrated, and standardized. The method was computerized, adding to the reproducibility of the results (Alarie 1998(Alarie , 2000Vijayaraghavan et al. 1994). The availability of RD 50 s in male mice for 89 chemicals (Schaper 1993), and their correlation with OELs suggests potential applicability to air exposure guidelines for the public. The result of this analysis quantitatively supports the applicability of RD 50 s in setting exposure guidelines for the public and workers.
We found a strong correlation between RD 50 s and human LOAELs, TLVs, and California RELs. Focusing on human studies where the subjects developed eye or respiratory irritation responses, we observed a strong correlation (R 2 = 0.80) between RD 50 s and LOAELs for 25 chemicals with irritating effects. The correlation remained close to 0.8 after conducting various subanalyses, indicating that the strains of mice or the RD 50 exposure time does not substantially affect the correlation. Previously, Nielsen et al. (1995) proposed an indoor air guideline for the public between 0.025 and 0.25 times the OEL, similar to 0.0008 and 0.008 times the RD 50 . In our analysis, the RD 50 to REL correlation can be expressed as REL = 0.00026 × Exposure times in the human studies varied from 1 to 480 min, and a subanalysis looking specifically at the effect of the duration of exposure made no significant change to the correlation. Further, subanalyses using LOAELs more closely associated with either upper respiratory or lower respiratory effects did not make a significant change to the correlations. Although the variability in the response rate, interindividual sensitivity, and differences in human study design, as described in Table 1, would be expected to have reduced the correlation with the RD 50 , specific factors were not identified in our subanalyses. Thus, we conclude that the irritating symptoms in humans correlate well with the RD 50 s of animals irrespective of the specific acute exposure duration. These results not only support the use of the RD 50 in setting guidelines for acutely irritating compounds, but also suggest that a concentration-time extrapolation for these effects appears unwarranted. This is consistent with the finding by Shusterman et al. (2006) that the human response to sensory irritants reached a plateau rapidly. Thus, the response appears to be influenced to a greater extent by the exposure concentration rather than the exposure time over the period of observation for most animal and human experiments considered in the present analysis, and over the periods of concern for the TLVs (15 min to 8 hr) and acute RELs (1 hr).
The results of this analysis are subject to several limitations. First, the number of available human studies limits the LOAEL data, and it is unlikely that human data will significantly increase in the future. The number of comparisons could increase as the numbers of RD 50 s increase for chemicals with human data. However, considering the robustness of the subanalyses, and the historical correlation of the RD 50 to the TLV, a significant change in the RD 50 to LOAEL correlation is unlikely after adding other sensory irritants in the analysis. Finally, we address issues raised by Bos et al. (1992Bos et al. ( , 2002Bos et al. ( , 2003. First, Bos et al. (2003) claimed that the RD 50 -OEL correlation is expected because Evaluation and application of the RD50 for public exposure levels Environmental Health Perspectives • VOLUME 115 | NUMBER 11 | November 2007  Table 3. Summary of linear least-squares regression analyses for various comparisons.   most OELs are based on animal data. Although many OELs are based on animal data, many are based on human data as well.
Of the 24 substances we evaluated in our RD 50 -OEL correlation, the OEL for only one compound, n-pentyl acetate, relied on the RD 50 for its derivation, which was based solely on animal data. The strong correlation between RD 50 s and human LOAELs also addresses this concern. Second, Bos et al. (2002) reported the RD 50 s did not correlate well with histopathologic changes in the respiratory tract or with corrosivity, and therefore RD 50 s were inappropriate to evaluate respiratory tract irritation. However, the stated purpose of the ASTM standard is to evaluate sensory irritation potential, not histopathology or corrosivity. In our comparison of the RD 50 s with human irritation LOAELs, the correlation was strong with the inclusion of respiratory tract irritation end points in the analysis. Further, the risk assessment framework for occupational and public exposure levels addresses the concerns regarding the potential for other, more severe effects. In cases where other health effects occur at or below levels producing sensory irritation, exposure guidelines use the more sensitive adverse effect.
Third, Bos et al. (1992) raised concerns regarding the inconsistency of RD 50 s among strains and species. Although RD 50 s have been generated for various strains and species with varying test procedures, adhering to the ASTM standard method addresses this concern. Limiting the RD 50 test to those conducted in mice, or Swiss-Webster mice, and limiting the exposure time keeps the test to a more standardized method, although intrastrain variability was not a cause for concern in our subanalyses. Finally, we addressed the concern regarding time-concentration response curves (Bos et al. 1992), with separate subanalyses based on exposure time. These analyses show that time did not appear to be a factor in our analyses. Our presumption is that if the study adheres adequately to the ASTM standard method, experimental exposure time plays a minor role. It is also worth pointing out that all of the figures comparing RD 50 s to LOAELs, RELs, and TLVs are plotted on a log-log plot because of the wide range of values. Because of the nature of log-log plots, the correlation is higher compared with the same correlation using a nonlogarithmic scale.
The applicability of the RD 50 test to human health protection has been demonstrated in several analyses, but extrapolation of the test results to the general public would be improved with greater focus on the tail of the dose-response curve, to ensure protection of sensitive subpopulations. One solution would be for RD 50 studies to report sufficient information to calculate a benchmark dose (BMD) value, and not focus solely on the specific RD 50 value. A standardized BMD value could be calculated at the tail of the distribution, taking into account the slope of the dose-response curve. Alternatively, the test procedure could be refined to identify the "just detectable effect level," which is approximately a 12% decrease in the respiratory rate (Alarie 1998). Although some work has been done in this area (Boylstein et al. 1996), additional information is needed to better understand the tail of the dose-response curve and to address any concerns for spurious results from low exposure concentrations. The reported just detectable effect level of 12% appears to be close to the no observed effect level of the procedure. Use of this response rate in risk assessment is consistent with the recommendation by the U.S. EPA (2007) that the BMD for a continuous response may be set on statistical criteria of distinguishability from the control value, as well as on grounds of anticipated biological significance. A major benefit of focusing on the just detectable effect level would be to reduce potential animal suffering, and possibly animal usage.
In conclusion, the RD 50 test is a good starting point for setting exposure standards for acute airborne irritants. As noted by Alarie et al. (2000), the TLV may need to be < 0.03 RD 50 to prevent other toxic effects. Consequently, the literature should be adequately evaluated to determine that sensory irritation is likely the most sensitive adverse effect. The application of RD 50 s appears most useful when qualitative data are available indicating sensory irritation as the most sensitive adverse effect, but quantitative human data are lacking. The RD 50 has proven its usefulness with the ability to appropriately rank the potency of airborne chemicals as sensory irritants and help establish exposure limits. A strong correlation between RD 50 s and LOAELs provides further support for using RD 50 s in determining guidance levels to protect the general public from sensory irritants.