Appendix B

Benchmark Dose Estimation

BELOW are examples of how to derive benchmark doses (BMDs) using data on 1,4-dichlorobenzene, botulinum, vinyl chloride, and aflatoxin.

DERIVING BENCHMARK DOSES FOR 1,4-DICHLOROBENZENE

In the examples provided below, the subcommittee used two 13-week (wk) toxicity studies and one 2-year (yr) study by the National Toxicology Program (NTP 1987) as the basis for deriving BMDs. Two 13-wk studies were done because the first study did not demonstrate a no-observed-adverse-effect level (NOAEL). In the first study, rats were administered 1,4-dichlorobenzene by gavage at doses of 300-1500 mg/kg (milligrams per kilogram body weight) per day (d), 5 d/wk. Because histologic changes of the kidney were observed in male rats in all dose groups, a second study was done at lower doses of 38-600 mg/kg/d. In the 2-yr study, 1,4-dichlorobenzene was administered by gavage 5 d/wk at 150 and 300 mg/kg/d for male rats and at 300 and 600 mg/kg/d for female rats.

The results of these studies are used below to illustrate the way to calculate BMDs for nonquantal response data, quantal data that are highly variable, and for carcinogenic effects (Table B-1). The tabulated benchmark dose (BMDp), which is a lower 95% confidence limit (CL)



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Appendix B Benchmark Dose Estimation BELOW are examples of how to derive benchmark doses (BMDs) using data on 1,4-dichlorobenzene, botulinum, vinyl chloride, and aflatoxin. DERIVING BENCHMARK DOSES FOR 1,4-DICHLOROBENZENE In the examples provided below, the subcommittee used two 13-week (wk) toxicity studies and one 2-year (yr) study by the National Toxicology Program (NTP 1987) as the basis for deriving BMDs. Two 13-wk studies were done because the first study did not demonstrate a no-observed-adverse-effect level (NOAEL). In the first study, rats were administered 1,4-dichlorobenzene by gavage at doses of 300-1500 mg/kg (milligrams per kilogram body weight) per day (d), 5 d/wk. Because histologic changes of the kidney were observed in male rats in all dose groups, a second study was done at lower doses of 38-600 mg/kg/d. In the 2-yr study, 1,4-dichlorobenzene was administered by gavage 5 d/wk at 150 and 300 mg/kg/d for male rats and at 300 and 600 mg/kg/d for female rats. The results of these studies are used below to illustrate the way to calculate BMDs for nonquantal response data, quantal data that are highly variable, and for carcinogenic effects (Table B-1). The tabulated benchmark dose (BMDp), which is a lower 95% confidence limit (CL)

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES TABLE B-1 Toxicity Data and Benchmark Doses for 1,4-Dichlorobenzene Hematology: Decreased Hematocrit in Male F-344/N Rats (13-wk study) Dose (mg/kg)   0 300 600 900 1200 1500 Avg. hematocrit (%)   50.1 ± 0.8 47.8 ± 0.9 47.3 ± 0.7 47.2 ± 0.4 47.6 ± 0.8 42.5 ± 0.5 (±SE) Sample size   9 9 9 9 5 2 Model-Free (Experimental Dose)     Normal Distribution, 3 SD Adverse Effect NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01 None 300 NA     112 334 3 Hematology: Decreased Hemoglobin Concentration in Male F-344/N Rats (13-wk study) Dose (mg/kg)   0 300 600 900 1200 1500 Hemoglobin (g/dL)   17.6 ± 0.2 16.4 ± 0.2 16.5 ± 0.2 16.5 ± 0.2 16.6 ± 0.3 15.3 ± 0.2 (±SE) Sample size   9 9 9 9 5 2 Model-Free (Experimental Dose)     Normal Distribution, 3 SD Adverse Effect NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01 None 300 NA     75 203 3 Hematology: Decreased RBC Count in Male F-344/N Rats (13-wk study) Dose (mg/kg)   0 300 600 900 1200 1500 RBC (1,000,000/mm3)   10.0 ± 0.12 9.5 ± 0.11 9.5 ± 0.12 9.7 ± 0.09 9.8 ± 0.17 8.8 ± 0.16 (±SE) Sample size   9 9 9 9 5 2 Model-Free (Experimental Dose)     Normal Distribution, 3 SD Adverse Effect NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01 None 300 NA     81 230 3

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Body Weight: Decreased Weight in Male F-344/N Rats (13-wk study) Dose (mg/kg) 0 300 600 900 1200 1500   Avg. weight gain (g)   187 ± 5 163 ± 8 142 ± 5 137 ± 6 114 ± 9 91 ± 22 (±SE) Sample size   9 10 10 9 5 2 Model-Free (Experimental Dose)     Normal Distribution, 3 SD Adverse Effect NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01 None 300 NA     87 232 3 Kidney: Renal Tubular Degeneration in Male F-344/N Rats 13-wk study) Dose (mg/kg)   0 37.5 75 150 300 600 Response   7/10 NR NR 5/10 3/10 9/10 Sample size   9 9 9 9 5 2 Severity increased at 300 mg/kg and 600 mg/kg; 150 mg/kg is considered a NOAEL Model-Free (Experimental Dose) Probit Model (ad hoc 95% LCL)     NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01 150 300 2     72 190 3 Liver: Hepatocellular Adenoma or Carcinoma in Male B6C3F1 Mice (2-yr study) Dose (mg/kg)   0 300 600       Response   17/50 22/49 40/50       Model-Free (Experimental Dose)     Probit Model (ad hoc 95% LCL) NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01 8 77 10     9 43 5 LCL, lower confidence limit; LML, linearized multistage model; LOAEL,lowest-observed-adverse-effect level; NOAEL, no-observed-adverse-effectlevel; NA, not applicable; NR, not reported; RBC, red blood cell;SD, standard deviation.

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES estimate of a dose corresponding to an excess risk of p, has not been adjusted for a discontinuous exposure regimen (dosing was done 5 d/wk). For the calculation of spacecraft water exposure guidelines (SWEGs), a duration conversion would have to be made along with other necessary conversions and adjustments. Nonquantal Response Data In the first 13-wk toxicity study, NTP reported hematotoxicity in male rats at all doses. Specifically, hematocrit, hemoglobin concentration, and red blood cell (RBC) count were decreased at all doses relative to the vehicle control. None of these hematologic changes was seen consistently in female rats. (Hematotoxicity was not mentioned in regard to the second 13-wk study.) Decreased body weight gain was observed at all doses in male rats, in parallel to hematotoxicity. No NOAEL was determined for decreased hematocrit, decreased hemoglobin concentration, decreased RBC count, or decreased body weight gain in male rats, and, thus, the lowest dose tested (300 mg/kg) was considered the lowest-observed-adverse-effect level (LOAEL). The data on these four parameters are used for illustrating the calculation of BMD for nonquantal response data. The data on each of the hematologic parameters and on body weight were modeled using the procedure of Kodell and West (1993) to calculate BMD. This method assumes a normal distribution for the observed end point, with a quadratic dose-response function for the average response. For the calculations here, a response was considered adversely low if it fell more than three standard deviations below the theoretical average response for the vehicle control. Because of substantially reduced survival at the two highest doses (1200 and 1500 mg/kg, 50% and 20% survival, respectively), data from those groups were not used for model fitting. Figures Figure B-1, Figure B-2, and Figure B-3 display each hematology data set plotted against the mean dose-response model estimated by maximum likelihood. Figure B-4 gives similar information for the body weight gain data. BMD01 and BMD10 for decreased hemoglobin concentration, decreased RBC count, and decreased body weight gain are all in close agreement. The BMD01 is 75, 81, and 87 mg/kg, respectively; the BMD10 is 203, 230, and 232 mg/kg, respectively. Each BMD10 is reason-

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-1: Hematocrit data modeled using the procedure of Kodell and West (1993). ably close to the corresponding LOAEL (300 mg/kg) for the toxic effect (Table B-1). For decreased hematocrit, the resulting BMD01 (112 mg/kg) and BMD10 (334 mg/kg) are higher than the corresponding BMDp for other hematotoxic effects and for decreased body weight gain. To determine the effect of including the two highest dose groups in the calculation of BMDp for hematocrit, the BMDp, was recalculated for all 6 dose groups. The resulting BMD01 was 250 mg/kg and the BMD10 was 746 mg/kg – both of which seem unrealistically high. The modeling procedure does not provide a goodness-of-fit test, but it seems likely that the model does not fit well when all dose groups are included. Certainly, the response at the highest dose appears to have too much influence on the predictions at low doses. It seems prudent,

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-2: Hemoglobin concentration data modeled using the procedure of Kodell and West (1993). in this case, to confine BMD calculations to the four dose groups with good survival by eliminating the two highest dose groups. As Figures Figure B-1, Figure B-2, and Figure B-3 indicate, the quadratic models for mean hematology responses fitted to the four lowest dose groups are not monotone with respect to dose. One could fit monotonic dose-response models, but such models could be less steep and would not fit the data as well. The result, as with the inclusion of the highest doses, could be overestimation of BMDp. Highly Variable Quantal Data The second 13-wk study of 1,4-dichlorobenzene identified a NOAEL of 150 mg/kg in male rats, based on renal tubular degeneration, which

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-3: Red blood cell count modeled using the procedure of Kodell and West (1993). occurred at doses of 300 mg/kg and higher. (This also confirmed the results of the first study.) The data exhibited anerratic dose-response relationship: Degeneration was mild in 7 of 10 rats in the control group; mild to moderate in 5 of 10 rats in the 150-mg/kg group; moderate in 3 of 10 rats in the 300-mg/kg group; and moderate in 9 of 10 in the 600-mg/kg group (responses at 37.5 mg/kg and 75 mg/kg were not reported). Based on increased severity of renal tubular degeneration at 300 mg/kg, this dose was identified as a LOAEL. There was no increase in severity at 150 mg/kg, and because of the numerically lower incidence compared with the vehicle control, this was considered a NOAEL. A probit log-dose model was fitted to the incidence data, and a goodness-of-fit test indicated a marginally acceptable fit (p = .07) – the fitted

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-4: Body weight gain modeled using the procedure of Kodell and West (1993). model did not differ statistically from the data at the 5% significance level. However, because of the high variability in the data that describe the dose-response relationship as shown in Figure B-5, convergence of the maximum likelihood estimation procedure could not be achieved for calculating lower CLs on doses corresponding to the specified excess risks of 1% and 10% (BMD01 and BMD10). Therefore, the ad hoc method (outlined in Chapter 4 in the section on Experimental Variation) was used to derive factors by which to reduce central estimates of dose to calculate lower CLs. The resulting BMD01 is 72 mg/kg (449/6.2), which is about one-half of the NOAEL (150 mg/kg),

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-5: Renal tubular degeneration data modeled using a probit log-dose model. and the BMD10 is 190 mg/kg (494/2.6), which is about two-thirds of the LOAEL (300 mg/kg) (Table B-1). The dose-response relationship for these data are perhaps too erratic to be used as a basis for SWEGs (Figure B-5). However, the BMD01 and BMD10 are reasonably close to the corresponding values calculated for the three hematology end points and decreased body weight gain ( Table B-1). Carcinogenic Effects NTP found clear evidence of carcinogenicity in the 2-yr study. In male rats, there was an increased incidence of renal tubular cell adeno-

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES carcinomas. In male and female mice, there were increased incidences of hepatocellular adenomas and carcinomas. Because the U.S. Environmental Protection Agency (EPA) used hepatocellular adenomas and carcinomas in male mice to calculate a carcinogenic potency factor for 1,4-dichlorobenzene, the data are used here to illustrate the calculation of BMDs for carcinogenic effects. Two separate calculations were done, one using the (linearized) multistage model and the other using the probit log-dose model (Table B-1, Figures Figure B-6 and Figure B-7). The BMD01 values are in close agreement (multistage: 8 mg/kg; probit: 9 mg/kg). The BMD10 from the multistage model (77 mg/kg) is about 2-fold higher than the BMD10 from the probit model (43 mg/kg). The BMD01s for carcinogenic effects are almost a factor of 10 lower than the lowest BMD01 for noncancer effects. For the calculation of SWEGs for water aboard spacecraft, these BMD01s would need to be reduced further by an interspecies uncertainty factor and possibly by a factor to reflect the severity and rreversibility of cancer. If, for example, two factors of 10 were applied, then the resulting SWEG would be equivalent to a value that the National Aeronautics and Space Administration (NASA) would calculate as a spacecraft maximum allowable concentration (SMAC) for cancer, in that it would correspond to a linearly extrapolated excess risk of 10−4. It should be noted that the cancer BMD01 of 8 mg/kg produced by the multistage model can easily be related back to EPA's cancer potency factor of 2 × 10−2, assuming linearity of the dose-response curve at low doses. To define the potency factor, one must adjust 8 mg/kg to account for the discontinuous exposure regimen used in the study by multiplying by 5/7 (dosing was done 5 d/wk), and then divide by an interspecies surface area adjustment factor of approximately 13. This gives 0.44 mg/kg for a BMD01 that is adjusted for duration and species. Because 0.01 = 0.44 × potency, assuming linearity, then potency ≅ 2 × 10−2. Discussion It is interesting to note the ratios of LOAELS to NOAELs and BMD10s to BMD01s in Table B-1. All ratios lie between 2 and 10. If BMD01s are to be used instead of NOAELs and BMD10s are to be used instead of

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-6: Multistage model of hepatocellular adenoma or carcinoma incidence. LOAELs in the calculation of SWEGs, then it would be better to use a BMD01 whenever possible, rather than starting with a BMD10 and, perhaps, having to reduce it by as much as 10-fold. Another ratio of interest for comparison between BMD01s and BMD10s is the ratio of the model-based central estimate of dose to its corresponding lower 95% CL. For renal tubular degeneration, because of the ad hoc way the lower CL were calculated, it is easy to see that the ratio for BMD01 is at least 2-fold higher than the ratio for BMD10 (449/72 = 6.2 versus 494/190 = 2.6). For hepatocellular tumors, the ratio for BMD01 is 3-fold higher than the ratio for BMD10 (e.g., multistage model: 56/8 = 7.0 versus 181/77 = 2.3). Likewise, for decreased hematocrit, the ratio for BMD01 is higher than the ratio for BMD10 (203/112 = 1.8 versus 398/334 = 1.2). However, for decreased hemoglobin concentration, decreased RBC count, and decreased body weight gain, the ratio for BMD01 is smaller than the corresponding ratio

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-7: Probit log-dose model of hepatocellular adenoma or carcinoma incidence for BMD10 (111/75 = 1.5 versus 332/203 = 1.6, 126/81 = 1.6 versus 491/230 = 2.1, and 134/87 = 1.5 versus 367/232 = 1.6, respectively). Thus, it is not possible to predict that the ratio of the model-based central estimate of dose to its corresponding lower 95% CL will always be larger for one value of p than the other. DERIVING BMDs FOR BOTULINUM TOXIN, VINYL CHLORIDE, AND AFLATOXIN Data on one toxic end point from botulinum toxin (lethality), vinyl chloride (liver tumor), and aflatoxin (liver tumor) are used to further illustrate the calculation of model-based BMD01s and BMD10s, and to

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES make comparisons to the corresponding model-free NOAELs and LOAELs (Table B-2). Each data set illustrates a particular type of dose-response relationship, and each highlights interesting characteristics of BMDpcalculations. A probit log-dose model was used to fit the botulinum toxin data in Table B-2 (Food Safety Council 1980), assuming a background response level of zero. Using all data points resulted in a significant lack of fit of the model (p < .05). Thus, the highest doses were eliminated one at a time until a satisfactory fit was achieved (p = .76) (Figure B-8). This resulted in the elimination of the three highest dose groups. The elimination of high dose groups is a generally accepted practice because the focus is on getting the best possible estimates in the low-dose region of interest. Whether all doses below 27 pg (picograms) were kept or discarded made little difference in the fit of the model and its predictions; thus, these doses were kept for the analysis. The resulting 1% and 10% benchmark doses are given in Table B-2 along with the NOAEL and LOAEL, for purposes of comparison. The LOAEL was established at 30 pg, because the response at this dose level was statistically different (p < .05) from the zero response at all lower doses, including the NOAEL of 27 pg. For the botulinum toxin data, the dose-response relationship is very steep, and very thresholdlike (Figure B-8). For this reason, the BMD01 and BMD10 are close, and they mimic their counterparts, the NOAEL and LOAEL, fairly well. For the vinyl chloride data in Table B-2 (Food Safety Council 1980), all dose groups except the 1 ppm (part per million) group were from the same experiment. The 1 ppm group was part of a later experiment conducted under the same conditions as the first. A probit log-dose model was used to fit these data, assuming a background response level of zero. An excellent fit was obtained (p = .79). The resulting 1% and 10% BMDs are given in Table B-2, along with corresponding NOAELs and LOAELs. Although 250 ppm is the lowest dose that is statistically different from control (p < .05), the nonzero response at 50 ppm makes it difficult to decide which of the two doses (50 ppm or 250 ppm) should be established as the LOAEL. The data and maximum-likelihood fitted model are shown in Figure B-9. The BMD01 and BMD10 are far apart for the vinyl chloride data, differing by a factor of 285. This is because of the one-hit nonthreshold shape of the dose-response curve. Excess risk is predicted all the way

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES TABLE B-2 Toxicity Data and Benchmark Doses for Botulinum Toxin, Vinyl Chloride, and Aflatoxin Botulinum Toxin: Lethality Dose (pg) 1 5 10 15 20 24 27 30 Response 0/10 0/10 0/30 0/30 0/30 0/30 0/30 4/30 Dose (pg) 34 37 40 45 50 55 60 65 Response 11/30 10/30 16/30 26/30 26/30 17/30 22/30 20/30 Model-Free (Experimental Dose)   Probit Log-Dose Model (95% LCL) NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01   27 30 1.1     22 28 1.3   Vinyl Chloride: Liver Tumor Dose (ppm)   0 1 50 250 500 2500 6000 Response   0/58 0/118 1/60 4/59 7/60 13/60 13/59 Model-Free (Experimental Dose)   Probit Log-Dose Model (95% LCL) NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01   1 or 50 50 or 250 50 or 5     0.9 257 285   Aflatoxin B1: Liver Tumor Dose (ppb)   0 1 5 15 50 100   Response   0/18 2/22 1/22 4/21 0/25 28/28   Model-Free (Experimental Dose)   Probit Log-Dose Model (95% LCL) NOAEL LOAEL LOAEL/NOAEL     BD01 BD10 BD10/BD01   150   300 2     72 190 3 LCL, lower confidence limit; LOAEL, lowest-observed-adverse effectlevel; NOAEL, no-observed-adverse-effect level. Source: Food SafetyCouncil (1980).

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-8: Probit log-dose model of botulinum toxin lethality data. down to zero dose (Figure B-9), so that the BMD01 is low. For these data, the BMD01 and BMD10 do not agree well with their counterparts, the NOAEL and LOAEL. If the NOAEL is taken to be 1 ppb (parts per billion), then there is good agreement with the BMD01 of 0.9 ppb, but the LOAEL of 50 ppb does not agree well with the BMD10 of 257 ppb. On the other hand, if the NOAEL is taken to be 50 ppb, then it does not agree well with the BMD01 of 0.9 ppb, but the LOAEL of 250 ppb agrees well with the BMD10 of 285 ppb. This case illustrates one advantage of using BMDp instead of NOAEL, that is, no subjective judgment is needed to determine precisely which experimental dose is a NOAEL. Rather, fitting a model to the dose-response data enables estimation of any dose as a BMDp.

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-9: Probit log-dose model of liver tumor data on vinyl chloride A probit log-dose model was fitted to the aflatoxin B1 data in Table B-2 (Food Safety Council 1980), without any assumption about background response. Thus, a nonzero background was allowed. A good fit was obtained (p = .30) (Figure B-10). The resulting BMDp estimates are given in Table B-2, along with the NOAEL (5 ppb) and LOAEL (15 ppb). The LOAEL was 15 ppb because that was the lowest dose for which the response was statistically different from the zero response at zero dose (p < .05). For the aflatoxin data, the BMD01 and BMD10 are roughly half the

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES Figure B-10: Probit log-dose model of liver tumor data on aflatoxin B1. corresponding NOAEL and LOAEL, respectively. The fitted dose-response is sigmoid, and flattens out below 15 ppb (Figure B-10). Because of the inversion of the observed responses at 1 ppb (9%) and 5 ppb (4.5%), the model predicts a nonzero background rate of approximately 5%, even though the observed background rate is zero. Although the data are quite variable at low doses, they are consistent with a nonzero background rate, and the model does provide a good fit to the data. It appears reasonable that the BMD01 and BMD10 are 2-fold lower than the corresponding NOAEL and LOAEL, because they reflect the variability of the dose-response relationship at low doses

OCR for page 131
METHODS FOR DEVELOPING SPACECRAFT WATER EXPOSURE GUIDELINES (the response at 1 ppb is numerically higher than is the response at 5 ppb). REFERENCES Food Safety Council. 1980. Proposed System for Food Safety Assessment. Washington, DC: Food Safety Council. Kodell, R.L., and R.W. West. 1993. Upper confidence limits on excess risk for quantitative responses . Risk Anal. 13:177-182. NTP (National Toxicology Program). 1987. Toxicology and Carcinogenesis Studies of 1,4-dichlorobenzene (CAS No. 106-46-7) in F344/N Rats and B6C3F1 Mice (Gavage Studies). Technical Report No. 319. Research Triangle Park, NC: U.S. Department of Health and Human Services.