PEER REVIEW DRAFT - DO NOT CITE OR QUOTE «EPA United States Environmental Protection Agency EPA Document #740-Rl-8013 August 2019, DRAFT Office of Chemical Safety and Pollution Prevention Draft Risk Evaluation for 1-Bromopropane (w-Propyl Bromide) CASRN: 106-94-5 Supplemental Information on Human Health Benchmark Dose Modeling August 2019 ------- PEER REVIEW DRAFT - DO NOT CITE OR QUOTE TABLE OF CONTENTS TABLE OF CONTENTS 2 LIST OF TABLES 4 LIST OF FIGURES 9 ACKNOWLEDGEMENTS 12 1 INTRODUCTION 13 2 BENCHMARK DOSE MODELING OF NON-CANCER EFFECTS 13 2.1 Benchmark Dose Modeling of Non-Cancer Effects for Acute Exposures 13 2.1.1 Decreased Live Litter Size 13 2.1.2 Post implantation loss 18 2.2 Benchmark Dose Modeling of Non-Cancer Effects for Chronic Exposures 27 2.2.1 Increased Incidence of Vacuolization of Centrilobular Hepatocytes in Males 27 2.2.2 Increased Incidence of Vacuolization of Centrilobular Hepatocytes in Males 29 2.2.3 Increased Incidence of Vacuolization of Centrilobular Hepatocytes in Females 32 2.2.4 Increased Incidence of Renal Pelvic Mineralization in Males 35 2.2.5 Increased Incidence of Renal Pelvic Mineralization in Females 38 2.2.6 Decreased Seminal Vesicle Weight 40 2.2.6.1 Decreased Relative Seminal Vesicle Weight 41 2.2.6.2 Decreased Absolute Seminal Vesicle Weight 43 2.2.7 Decreased Percent Normal Sperm Morphology 46 2.2.8 Decreased Percent Motile Sperm 49 2.2.9 Decreased Left Cauda Epididymis Weight 51 2.2.10 Decreased Right Cauda Epididymis Weight 54 2.2.11 Increased Estrus Cycle Length 57 2.2.12 Decreased Antral Follical Count 59 2.2.13 Decreased Male and Female Fertility Index 59 2.2.14 Decreased Implantations Sites 62 2.2.15 Decreased Pup Body Weight 66 2.2.15.1 Decreased Body Weight in F1 Male Pups at PND 28 66 2.2.15.2 Decreased Body Weight in F2 Female Pups at PND 14 72 2.2.15.3 Decreased Body Weight in F2 Female Pups at PND 21 75 2.2.15.4 Decreased Body Weight in F2 Male Pups at PND 14 77 2.2.15.5 Decreased Body Weight in F2 Male Pups at PND 21 80 2.2.16 Decreased Brain Weight 83 2.2.16.1 Decreased Brain Weight in Fo Females 83 2.2.16.2 Decreased Brain Weight in Fo Males 85 2.2.16.3 Decreased Brain Weight in F1 Females as Adults 88 2.2.16.4 Decreased Brain Weight in F1 Males as Adults 90 2.2.16.5 Decreased Brain Weight in F2 Females at PND 21 92 2.2.16.6 Decreased Brain Weight in F2 Males at PND 21 95 2.2.17 Decreased Hang Time 98 3 BENCHMARK DOSE MODELING OF TUMORS 101 3.1 Lung Tumors in Female Mice 102 3.1.1 Summary of Multistage Model 105 3.1.1.1 Selected Frequentist Multistage - Multistage 1 Restricted; Extra Risk, BMR = 0.001 and 0.1, doses are in ppm 105 3.1.1.2 Selected Frequentist Multistage - Multistage 1 Restricted; Added Risk, BMR = 0.001 and 0.1, doses are in ppm 108 3.1.2 Summary of Frequentist Model Averaging 110 3.1.3 Summary of Bavesian Model Averaging Ill 3.1.3.1 Bayesian Model Averaging - Extra Risk, BMR = 0.001 and 0.1, doses are in ppm Ill 3.1.3.2 Bayesian Model Averaging - Added Risk, BMR = 0.001 and 0.1, doses are in ppm 112 ------- Table 2-19 Summary of BMD Modeling Results for Vacuolization of Centrilobular Hepatocytes in Female Fo Rats Following Inhalation Exposure to 1-BP in a Two-Generation Study 33 Table 2-20 BMD Modeling Results for Vacuolization of Centrilobular Hepatocytes in Female Rats Exposed to 1-BP Via Inhalation; BMR 10% Added Risk 34 Table 2-21 Incidence of Renal Pelvic Mineralization Selected for Dose-Response Modeling for 1-BP 36 Table 2-22 Summary of BMD Modeling Results for Renal Pelvic Mineralization in Male Fo Rats Following Inhalation Exposure to 1-BP in a Two-Generation Study 36 Table 2-23 BMD Modeling Results for Renal Pelvic Mineralization in Male Rats Exposed to 1-BP Via Inhalation; BMR 10% Added Risk 37 Table 2-24 Incidence of Renal Pelvic Mineralization Selected for Dose-Response Modeling for 1-BP 38 Table 2-25 Summary of BMD Modeling Results for Renal Pelvic Mineralization in Female Fo Rats Following Inhalation Exposure to 1-BP in a Two-Generation Study 39 Table 2-26 BMD Modeling Results for Renal Pelvic Mineralization in Female Rats Exposed to 1- BP Via Inhalation; BMR 10% Added Risk 39 Table 2-27 Relative Seminal Vesicle Weight Data Selected for Dose-Response Modeling for 1-BP 41 Table 2-28 Summary of BMD Modeling Results for Relative Seminal Vesicle Weight in Rats Exposed to 1-BP by Inhalation 41 Table 2-29 BMD Modeling Results for Relative Seminal Vesicle Weight; BMR = 1 Standard Deviation Change from Control Mean 42 Table 2-30 Absolute Seminal Vesicle Weight Data Selected for Dose-Response Modeling for 1-BP Table 2-31 Summary of BMD Modeling Results for Seminal Vesicle Absolute Weight in Rats Exposed to 1-BP by Inhalation 44 Table 2-32 BMD Modeling Results for Seminal Vesicle Absolute Weight; BMR = 1 Standard Deviation Change from Control Mean 45 Table 2-33 Sperm Morphology Data Selected for Dose-Response Modeling for 1-BP 46 Table 2-34 Summary of BMD Modeling Results for Sperm Morphology in the Fo Generation Exposed to 1-BP by Inhalation 47 Table 2-35 BMD Modeling Results for Sperm Morphology in Fo Rats Exposed to 1-BP by Inhalation; BMR = 1 Standard Deviation Change from Control Mean 48 Table 2-36 Sperm Motility Data Selected for Dose-Response Modeling for 1-BP 49 Table 2-37 Summary of BMD Modeling Results for Sperm Motility Fo Male Rats Following Inhalation Exposure to 1-BP 50 Table 2-38 Summary of BMD Modeling Results for Sperm Motility Fo Male Rats Following Inhalation Exposure to 1-BP with the Highest Dose Dropped 51 Table 2-39 Left Cauda Epididymis Absolute Weight Data Selected for Dose-Response Modeling for 1-BP 52 Table 2-40 Summary of BMD Modeling Results for Left Cauda Epididymis Absolute Weight Fo Male Rats Following Inhalation Exposure to 1-BP 52 Table 2-41 BMD Modeling Results for Left Cauda Epididymis Absolute Weight; BMR = 1 Standard Deviation Change from Control Mean 53 Table 2-42 Right Cauda Epididymis Absolute Weight Data Selected for Dose-Response Modeling for 1-BP 55 ------- Table 2-43 Summary of BMD Modeling Results for Right Cauda Epididymis Absolute Weight Fo Male Rats Following Inhalation Exposure to 1-BP 55 Table 2-44 BMD Modeling Results for Right Cauda Epididymis Absolute Weight; BMR = 1 Standard Deviation Change from Control Mean 56 Table 2-45 Estrus Cycle Length Data Selected for Dose-Response Modeling for 1-BP 58 Table 2-46 Summary of BMD Modeling Results for Estrus Cycle Length Fo Female Rats Following Inhalation Exposure to 1-BP 58 Table 2-47 Antral Follicle Count Data Selected for Dose-Response Modeling for 1-BP 59 Table 2-48 Summary of BMD Modeling Results for Antral Follical Count in Female Rats Following Inhalation Exposure to 1-BP 59 Table 2-49 Fertility Index Data Selected for Dose-Response Modeling for 1-BP 60 Table 2-50 Summary of BMD Modeling Results for Fertility Index of Fo Rats Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 60 Table 2-51 BMD Modeling Results for Fertility Index in Rats Exposed to 1-BP Via Inhalation BMR 10% Extra Risk 61 Table 2-52 Implantations Site Data Selected for Dose-Response Modeling for 1-BP 62 Table 2-53 Summary of BMD Modeling Results for Implantations Sites in Fo Rats Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 63 Table 2-54 BMD Modeling Results for Implantation Sites in Rats Exposed to 1-BP Via Inhalation in ppm BMR 1 Standard Deviation 64 Table 2-55 Pup Body Weight Data in Fi Males at PND 28 for Dose-Response Modeling 66 Table 2-56 Summary of BMD Modeling Results for Body Weight of Fi Male Rat Pups on PND 28 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 67 Table 2-57 BMD Modeling Results for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation BMR 5% Relative Deviation 68 Table 2-58 BMD Modeling Results for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation BMR 5% Relative Deviation 70 Table 2-59 Pup Body Weight Data in F2 Females at PND 14 from Selected for Dose-Response Modeling 72 Table 2-59 Summary of BMD Modeling Results for Body Weight of F2 Female Rat Pups on PND 14 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 72 Table 2-60 BMD Modeling Results for Body Weight of F2 Female Rat Pups on PND 14 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study with Variances Fixed at Smallest, Pooled and Highest Values 74 Table 2-61 Pup Body Weight Data in F2 Females at PND 21 from Selected for Dose-Response Modeling 75 Table 2-62 Summary of BMD Modeling Results for Body Weight of F2 Females on PND 21 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 75 Table 2-63 BMD Modeling Results for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation BMR = 5% Relative Deviation 76 Table 2-64 Pup Body Weight Data in F2 Males at PND 14 from Selected for Dose-Response Modeling 77 Table 2-65 Summary of BMD Modeling Results for Body Weight of F2 Male Rat Pups on PND 14 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 78 Table 2-66 BMD Modeling Results for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR = 5% Relative Deviation 79 Table 2-67 Pup Body Weight Data in F2 Males at PND 21 80 ------- Table 2-68 Summary of BMD Modeling Results for Body Weight of F2 Male Rat Pups on PND 21 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 81 Table 2-69 BMD Modeling Results for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR = 5% Relative Deviation 81 Table 2-70 Brain Weight Data in Fo Females for Dose-Response Modeling 83 Table 2-71 Summary of BMD Modeling Results for Brain Weight of Fo Females Following Inhalation Exposure to 1-BP 83 Table 2-72 BMD Modeling Results for Brain Weight in Fo Female Rats Exposed to 1-BP Via Inhalation in ppm BMR = 1 Standard Deviation 84 Table 2-73 Brain Weight Data in Fo Males for Dose-Response Modeling 85 Table 2-74 Summary of BMD Modeling Results for Brain Weight of Fo Males Following Inhalation Exposure to 1-BP 86 Table 2-75 BMD Modeling Results for Brain Weight of Fo Male Rats Following Inhalation Exposure to 1-BP in a Two-Generation Study with Variances Fixed at Smallest, Pooled and Highest Values 87 Table 2-76 Brain Weight Data in Fi Females as Adults from Selected for Dose-Response Modeling 88 Table 2-77 Summary of BMD Modeling Results for Brain Weight of Fi Female Rats as Adults Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 88 Table 2-78 BMD Modeling Results for Brain Weight in Fi Female Rats as Adults Exposed to 1-BP Via Inhalation BMR = 1% Relative Deviation 89 Table 2-79 Brain Weight Data in Fi Males as Adults from Selected for Dose-Response Modeling 91 Table 2-80 Summary of BMD Modeling Results for Brain Weight of Fi Male Rats as Adults Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 91 Table 2-81 Brain Weight Data in F2 Females at PND 21 from Selected for Dose-Response Modeling 92 Table 2-82 Summary of BMD Modeling Results for Brain Weight of F2 Female Rats at PND 21 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 92 Table 2-83 BMD Modeling Results for Brain Weight in F2 Female Exposed to 1-BP Via Inhalation BMR = 1% Relative Deviation 93 Table 2-84 Brain Weight Data in F2 Males at PND 21 for Dose-Response Modeling 95 Table 2-85 Summary of BMD Modeling Results for Brain Weight of F2 Male Rats as Adults Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study 95 Table 2-86 BMD Modeling Results for Brain Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR = 1% Relative Deviation 96 Table 2-87 Hang Time from a Suspended Bar Data for Dose-Response Modeling for 1-BP 98 Table 2-88 Summary of BMD Modeling Results for Hang Time from a Suspended Bar; BMR = 1 std. dev. change from control mean 98 Table 2-89 BMD Modeling Results for Hang Time from a Suspended Bar; BMR = 1 Standard Deviation Change from Control Mean 99 Table 3-1 Incidence of Lung Tumors in Female Mice 102 Table 3-2 Summary of BMDS 3.0 modeling results for lung tumors in female mice exposed to 1- BP by inhalation for 2 years (NTP, 2011); BMRs = 10% and 0.1% extra and added risk, doses are in ppm 103 Table 3-3 Lung Tumors in Female Mice, Selected Frequentist Multistage - Multistage 1 Restricted; Extra Risk, BMR = 0.001 and 0.1 User Input 105 ------- Table 3-4 Lung Tumors in Female Mice, Selected Frequentist Multistage - Multistage 1 Restricted; Extra Risk, BMR = 0.001 and 0.1 Model Results 105 Table 3-5 Lung Tumors in Female Mice, Selected Frequentist Multistage - Multistage 1 Restricted; Added Risk, BMR = 0.001 and 0.1 User Input 108 Table 3-6 Lung Tumors in Female Mice, Selected Frequentist Multistage - Multistage 1 Restricted; Added Risk, BMR = 0.001 and 0.1 Model Results 108 Table 3-7 Lung Tumors in Female Mice, Summary of Frequentist Model Averaging 110 Table 3-8 Lung Tumors in Female Mice, Bayesian Model Averaging - Extra Risk, BMR = 0.001 and 0.1 User Inputs Ill Table 3-9 Lung Tumors in Female Mice, Bayesian Model Averaging - Extra Risk, BMR = 0.001 and 0.1 Model Results Ill Table 3-10 Lung Tumors in Female Mice, Bayesian Model Averaging - Added Risk, BMR = 0.001 and 0.1 User Inputs 112 Table 3-11 Lung Tumors in Female Mice, Bayesian Model Averaging - Added Risk, BMR = 0.001 and 0.1 Model Results 112 Table 3-12 Incidence of Large Intestine Adenomas in Female Rats 113 Table 3-13 Summary of BMDS 3.0 modeling results for large intestine adenomas in female rats exposed to 1-BP by inhalation for 2 years (NTP, 2011); BMRs = 10% and 0.1% extra and added risk, doses are in ppm 114 Table 3-14 Large Intestine Adenomas in Female Rats, Selected Frequentist Multistage - Multistage 1 Restricted; Extra Risk, BMR = 0.001 and 0.1 User Input 116 Table 3-15 Large Intestine Adenomas in Female Rats, Selected Frequentist Multistage - Multistage 1 Restricted; Extra Risk, BMR = 0.001 and 0.1 Model Results 116 Table 3-16 Large Intestine Adenomas in Female Rats, Selected Frequentist Multistage - Multistage 1 Restricted; Added Risk, BMR = 0.001 and 0.1 User Input 119 Table 3-17 Large Intestine Adenomas in Female Rats, Selected Frequentist Multistage - Multistage 1 Restricted; Added Risk, BMR = 0.001 and 0.1 Model Results 119 Table 3-18 Large Intestine Adenomas in Female Rats, Summary of Frequentist Model Averaging 120 Table 3-19 Large Intestine Adenomas in Female Rats, Bayesian Model Averaging - Extra Risk, BMR = 0.001 and 0.1 User Inputs 121 Table 3-20 Large Intestine Adenomas in Female Rats, Bayesian Model Averaging - Extra Risk, BMR = 0.001 and 0.1 Model Results 121 Table 3-21 Large Intestine Adenomas in Female Rats, Bayesian Model Averaging - Added Risk, BMR = 0.001 and 0.1 User Inputs 122 Table 3-22 Large Intestine Adenomas in Female Rats, Bayesian Model Averaging - Added Risk, BMR = 0.001 and 0.1 Model Results 122 Table 3-23 Incidence of Keratoacanthoma and Squamous Cell Carcinomas in Male Rats 123 Table 3-24 Summary of BMDS 3.0 modeling results for keratoacanthoma & squamous cell carcinomas in male rats exposed to 1-BP by inhalation for 2 years (NTP, 2011); BMRs = 10% and 0.1% extra and added risk, doses are in ppm 124 Table 3-25 Keratoacanthoma and Squamous Cell Carcinomas in Male Rats, Selected Frequentist Multistage - Multistage 1 Restricted; Extra Risk, BMR = 0.001 and 0.1 User Input 126 Table 3-26 Keratoacanthoma and Squamous Cell Carcinomas in Male Rats, Selected Frequentist Multistage - Multistage 1 Restricted; Extra Risk, BMR = 0.001 and 0.1 Model Results 126 Table 3-27 Keratoacanthoma and Squamous Cell Carcinomas in Male Rats, Selected Frequentist Multistage - Multistage 1 Restricted; Added Risk, BMR = 0.001 and 0.1 User Input 128 ------- Figure 2-13 Plot of Mean Response by Dose in ppm with Fitted Curve for Exponential (M4) Model with Constant Variance for Relative Seminal Vesicle Weight; BMR = 1 Standard Deviation Change from Control Mean 42 Figure 2-14 Plot of Mean Response by Dose in ppm with Fitted Curve for Hill Model with Constant Variance for Seminal Vesicle Absolute Weight; BMR = 1 Standard Deviation Change from Control Mean 45 Figure 2-15 Plot of Mean Response by Dose in ppm with Fitted Curve for Exponential (M2) Model with Constant Variance for Sperm Morphology in Fo Rats Exposed to 1-BP by Inhalation; BMR = 1 Standard Deviation Change from Control Mean 48 Figure 2-16 Plot of Mean Response by Dose in ppm with Fitted Curve for Polynomial 4° Model with Constant Variance for Left Cauda Epididymis Absolute Weight; BMR = 1 Standard Deviation Change from Control Mean 53 Figure 2-17 Plot of Mean Response by Dose in ppm with Fitted Curve for Polynomial 4° Model with Constant Variance for Right Cauda Epididymis Absolute Weight; BMR = 1 Standard Deviation Change from Control Mean 56 Figure 2-18 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (LogLogistic) for Fertility Index in Rats Exposed to 1-BP Via Inhalation in ppm BMR 10% Extra Risk 61 Figure 2-19 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Linear) for Implantation Sites in Rats Exposed to 1-BP Via Inhalation in ppm BMR 1 Standard Deviation. ...64 Figure 2-20 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Exponential (M2)) for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR 5% Relative Deviation 68 Figure 2-21 Plot of Mean Response by Dose with Fitted Curve for the Hill Model for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR 5% Relative Deviation 70 Figure 2-21 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Polynomial 2°) for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR = 5% Relative Deviation 76 Figure 2-22 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Polynomial 2°) for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR = 5% Relative Deviation 78 Figure 2-23 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Linear) for Pup Body Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR = 5% Relative Deviation. 81 Figure 2-24 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Linear) for Brain Weight in Fo Female Rats Exposed to 1-BP Via Inhalation in ppm BMR = 1 Standard Deviation 84 Figure 2-25 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Exponential (M2)) for Brain Weight in Fi Female Rats as Adults Exposed to 1-BP Via Inhalation in ppm BMR = 1% Relative Deviation 89 Figure 2-26 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Exponential (M2)) for Brain Weight in F2 Female Exposed to 1-BP Via Inhalation in ppm BMR = 1% Relative Deviation 93 Figure 2-27 Plot of Mean Response by Dose with Fitted Curve for the Selected Model (Power) for Brain Weight in Rats Exposed to 1-BP Via Inhalation in ppm BMR = 1% Relative Deviation 96 ------- Figure 2-28 Plot of Mean Response by Dose in ppm with Fitted Curve for Exponential (M4) Model with Modeled Variance for Hang Time from a Suspended Bar; BMR = 1 Standard Deviation Change from Control Mean 99 Figure 3-1 Plot of Results for Lung Tumors in Female Mice Frequentist Multistage Degree 1 Model with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL 107 Figure 3-2 Plot of Results for Large Intestine Adenomas in Female Rats Frequentist Multistage Degree 1 Model with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL 118 Figure 3-3 Plot of Results for Keratoacanthoma and Squamous Cell Carcinomas in Male Rats Frequentist Multistage Degree 1 Model with BMR of 10% Extra Risk for the BMD and 0.95 Lower Confidence Limit for the BMDL 127 ------- 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 means were well-modeled the variances are not well modeled by the non-homogeneous variance model (the non-homogeneous variance model was used because the BMDS test 2/>-value = 0.0130). To investigate the effect of the poor modeling of the variances on the BMDL, the models were run using the smallest dose standard deviation (2.21), highest (4.47) and pooled (3.54) for all dose levels and the results are summarized in Table 2-4. As shown in the last column of Table 2-4 the ratios BMDLs for the lowest to the highest variance for the two best fitting models the Linear and Exponential (M2) models are 1.15 and 1.20, respectively. Overall the adjustment of the variances from most-variable to least-variable for all of the models makes little difference on the BMDL. This is strong evidence that the poor variance modeling for the original data is not substantially impacting the BMDL estimates. It is reasonable to use the non-homogeneous Exponential M2 model for the original data because it has the lowest AIC of all the model choices for the original data and therefore a BMDL of 41 ppm (40.7 ppm rounded to two significant figures) was selected for this endpoint. Table 2-2 Summary of BMD Modeling Results for Reduced Litter Size in Fo Generation Exposed to 1-BP by Inhalation; BMRs of 1 Standard Deviation, and 5% and 1% Relative Deviation From Control Mean. Model" Goodness of fit BMD 1SD (ppm) BMDL 1SD (ppm) BMD 5RD (ppm) BMDL 5RD (ppm) BMD 1RD (ppm) BMDL 1RD (ppm) Basis for model selection />-value AIC Exponential (M2) Exponential (M3)b 0.533 291. 10 256 ^ 158 61.3 40.7 12.0 7.97 The Exponential (M2) model was selected based on lowest AIC from this set of models which have adequate /7-values, adequate fit by visual inspection and the BMDLs are < 4-fold apart considered sufficiently close. Power0 Polynomial 3od Polynomial 2oe Linear 0.433 291. 51 281 189 69.9 49.8 14.0 9.95 Hill 0.722 291. 96 178 error8 35.8 10.4 6.36 1.69 Exponential (M4) Exponential (M5)f 0.622 292. 08 181 69.4 40.4 17.8 7.48 3.23 a Modeled variance case presented (BMDS Test 2 p-value = 0.0130), selected model in bold; scaled residuals for selected model for doses 0,100,250, and 500 ppm were -0.16, -0.05, 0.66, -0.76, respectively. b For the Exponential (M3) model, the estimate of d was 1 (boundary). Hie models in this row reduced to the Exponential (M2) model. c For the Power model, the power parameter estimate was 1. The models in this row reduced to the Linear model. d For the Polynomial 3° model, the b3 coefficient estimates was 0 (boundary of parameters space). The models in this row reduced to the Polynomial 2° model. For the Polynomial 3° model, the b3 and b2 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Linear model. e For the Polynomial 2° model, the b2 coefficient estimate was 0 (boundary of parameters space). The models in this row reduced to the Linear model. f For the Exponential (M5) model, the estimate of d was 1 (boundary). The models in this row reduced to the Exponential (M4) model. g BMDL computation failed for this model. ------- Parameter Estimates Variable Estimate Default Initial Parameter Values lnalpha 10.4606 6.08025 rho -3.14328 -1.44632 a 14.4915 10.5312 b 0.000836398 0.00102437 c n/a 0 d n/a 1 Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 23 14.4 14.49 2.21 2.8 -0.1569 100 25 13.3 13.33 3.72 3.19 -0.04505 250 22 12.3 11.76 4.47 3.88 0.6554 500 11 8.3 9.54 4.1 5.4 -0.7614 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 -143.3786 5 296.7571 A2 -137.9879 8 291.9758 A3 -140.9173 6 1 293.8347 R ®53.5054 2 311.0108 2 -141.5475 4 291.095 Tests of Interest Test -2*log(Likelihood Ratio) Test df /j-valuc Test 1^ 31.03 6 <0.0001 Test 2 10.78 3 0.01297 Test 3 5.859 2 0.05343 Test 4 1.26 2 0.5325 ------- 58 59 60 Table 2-4 BMD Modeling Results for Reduced Litter Size in Fo Generation Following Inhalation Exposure of Parental Rats to Model" Smallest Standard Deviation Pooled Standard Deviation Largest Standard Deviation Ratio BMDLs Smallest to Largest Std Dev Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) />-value AIC />-value AIC /j-valuc AIC Linear 0.279 213.92 63.5 53.5 0.605 288.69 63.5 49.2 0.729 326.11 63.5 46.6 1.15 Exponential (M2) 0.112 215.74 54.9 44.1 0.420 289.42 54.9 39.4 0.579 326.57 54.9 36.7 1.20 Exponential (M4) 0.112 215.74 54.9 42.6 0.420 289.42 54.9 34.4 0.579 326.57 54.9 29.1 1.46 Polynomial 3° 0.506 213.81 96.4 58.4 0.678^ 289.86 96.4 51.1 0.742 327.58 96.4 47.8 1.22 Polynomial 2° 0.393 214.09 105 57.4 0.593 289.97 105 50.8 0.672 327.65 105 47.6 1.21 Power 0.303 214.43 115 56.4 0.519 290.10 115 50.5 0.609 327.74 115 47.4 1.19 Exponential (M3) 0.239 214.75 127 56.1 0.461 290.23 127 42.6 0.559 327.82 127 38.7 1.45 Exponential (M5) 0.239 214.75 127 56.1 N/AbL 292.23 127 42.6 0.559 327.82 127 33.0 1.70 Hill N/Ab 216.43 115 56.4 N/Ab 292.10 116 50.3 N/Ab 329.74 116 47.2 1.19 61 62 a Constant variance case presented (BMDS Test 2 />-value = 1.000, BMDS Test 3 />-value = 1.000), no model was selected as a best-fitting model. b No available degrees of freedom to calculate a goodness of fit value. ------- 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 250 500 500 Number of Implantation Sites 14 16 14 14 15 16 12 16 16 14 18 16 16 15 15 15 12 18 16 16 15 15 17 14 15 13 15 17 16 16 11 15 12 18 18 12 12 Post Implantation Loss 1 Dam Weight at Study Week 0 (g) 152 165 166 158 168 143 148 177 154 153 179 171 180 170 165 157 164 162 159 160 151 141 179 150 153 175 146 161 167 165 166 162 157 153 158 166 167 146 164 155 161 158 ------- 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 Dose (ppm) Number of Implantation Sites Post Implantation Loss Dam Weight at Study Week 0 (g) 500 5 1 181 500 15 2 159 500 12 1 151 500 16 0 152 500 9 1 166 500 6 0 176 500 6 1 165 500 11 0 144 500 2 0 144 The application of nested dichotomous models to these data was possible because the incidence data for post-implantation loss were available for every litter, and preferable because they can account for intra-litter correlations and litter-specific covariates. A litter specific covariate that is potentially related to the endpoint of concern but is not itself impacted by dose is needed for this analysis. In this case, dam body weight measured at week 0 and the number of implantation sites were both used as covariates and the data was modeled separately in the same format for each. In this case, dam body weight measured at week 0 was selected as the preferred litter specific covariate because it was not affected at any dose and is potentially related to the implantation loss endpoint. Incidence of implantation loss presented a clear dose trend at lower doses but leveled off at the highest dose coincident with a reduction in implantation sites. The data were modeled with the all doses and the highest dose dropped for the purposes of this analysis because of the uncertainty associated with reduced sample size and improved model fit. The nested modeling was performed using the nested logistic and NCTR models contained in BMDS 2.7.0.4, as follows: • nested model for extra risk of 5% and 1%, using dam weight as a litter specific covariate, dropping the highest dose group (Table 2-6 and Table 2-7 and Figure 2-2 and Figure 2-3). • nested model for extra risk of 5% and 1%, using number of implantation sites as a litter specific covariate, dropping the highest dose group (Table 2-8 and Table 2-9 and Figure 2-4 and Figure 2-5). • nested model for extra risk of 5% and 1%, using dam weight as a litter specific covariate, including all dose groups (Table 2-10 and Table 2-11 and Figure 2-6 and Figure 2-7). ------- 184 185 186 187 188 189 190 191 192 193 194 195 196 Table 2-15 Incidence of Vacuolization of Centrilobular Hepatocytes Selected for Dose- Response Modeling for 1-BP Dose (ppm) Number of animals Incidence 0 15 0 100 15 0 200 15 0 400 15 3 800 15 6 The BMD modeling results for vacuolization of centrilobular hepatocytes are summarized in Table 2-16. The best fitting model was the LogLogistic based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit /rvalue (higher value indicates a better fit) and visual inspection. For the best fitting model a plot of the model is shown in Figure 2-9. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-17. Table 2-16 Summary of BMD Modeling Results for Vacuolization of Centrilobular Hepatocytes in Male Rats Following Inhalation Exposure to 1-1 :p Model3 Goodness of fit BMDlOPctAdd (ppm) BMDLioPctAdd (ppm) Basis for model selection />-value AIC Multistage 3° ( 0.955 38.189 ^ 346 226 Multistage 3° model was selected based on the lowest AIC from this set of models which have adequate />-valuc, adequate fit by visual inspection and the BMDLs are < 1.5-fold apart considered sufficiently close. Multistage 2° 0.898 39.202 289 198 LogProbit 0.951 39.678 345 225 Gamma 0.919 39.874 349 227 LogLogistic 0.903 ^40.003 349 224 Weibull 0.872 40.180 351 222 Probit 0.773 40.585 370 275 Logistic 0.662 41.195 382 290 a Selected model in bold; scaled residuals for selected model for doses 0, 100, 200,400, and 600 ppm were 0, -0.2, -0.56, 0.54, - 0.18, respectively. ------- 213 214 215 216 217 218 219 220 221 222 223 224 225 Table 2-18 Incidence of Vacuolization of Centrilobular Hepatocytes Selected for Dose- Response Modeling for 1-BP Dose (ppm) Number of animals Incidence 0 25 0 100 25 0 250 25 0 500 25 6 750 25 16 The BMD modeling results for vacuolization of centrilobular hepatocytes are summarized in Table 2-19. The best fitting model was the LogProbit based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit p-value (higher value indicates a better fit) and visual inspection. For the best fitting model a plot of the model is shown in Figure 2-10. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in . Table 2-19 Summary of BMD Modeling Results for Vacuolization of Centrilobular Hepatocytes in Female Fo Rats Following Inhalation Exposure to 1-BP in a Two- Generation Study Model3 Goodness of fit BMDlOPctAdd (ppm) BMDLioPctAdd (ppm) Basis for model selection />-value AIC LogProbit ( 0.988 64.438 ^ 415 322 LogProbit model was selected based on the lowest AIC from this set of models which have adequate /7-values (excluding Quantal-Linear), adequate fit by visual inspection and the BMDLs are 1.5-fold apart considered sufficiently close. Gamma 0.965 64.648 416 320 LogLogistic 0.945 ^ 64.843 415 320 Weibull 0.879 65.283 411 310 Probit 0.826 ^65.496 423 335 Logistic 0.661 66.491 431 347 Multistage 2° 0.410 68.583 279 228 Quantal-Linear 0.0134 80.285 153 109 a Selected model in bold; scaled residuals for selected model for doses 0, 100, 250, 500, and 750 ppm were 0, 0, -0.29, 0.19, -0.11, respectively. ------- 241 242 243 244 245 246 247 248 249 250 251 252 Table 2-21 Incidence of Renal Pelvic Mineralization Selected for Dose-Response Modeling for 1-BP Dose (ppm) Number of animals Incidence 0 25 1 100 25 0 250 25 1 500 25 2 750 25 6 The BMD modeling results for vacuolization of renal pelvic mineralization are summarized in Table 2-22. The best fitting model was the Multistage 3° based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit p-value (higher value indicates a better fit) and visual inspection. For the best fitting model a plot of the model is shown in Figure 2-11. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-23. Table 2-22 Summary of BMD Modeling Results for Renal Pelvic Mineralization in Male Fo Rats Following Inhalation Exposure to 1-BP in a Two-Generation Study Model" Goodness of fit BMDlOPctAdd (ppm) BMDLioPctAdd (ppm) Basis for model selection />-value AIC Multistage 3° 0.789 63.835 571 386 Multistage 3° model was selected based on the lowest AIC from this set of models which have adequate /7-values, adequate fit by visual inspection and the BMDLs are 1.5-fold apart considered sufficiently close. Multistage 2° { 0.668 64.258 ^ 527 368 Logistic 0.629 64.260 545 434 Probit 0.567 64.488 526 408 Weibull 0.603 65.825 581 375 LogLogistic 0.602 65.835 579 371 Gamma 0.597 ^65.856 575 371 LogProbit 0.597 65.894 577 355 Quantal-Linear 0.326 66.496 507 284 a Selected model in bold; scaled residuals for selected model for doses 0, 100, 250, 500, and 750 ppm were 0.6, -0.76, 0.26, -0.18, 0.07, respectively. ------- 326 327 328 329 330 331 332 333 334 335 336 337 338 339 Table 2-30 Absolute Seminal Vesicle Weight Data Selected for Dose-Response Modeling for 1-BP Dose (ppm) Number of animals Seminal Vesicle Absolute Weight (mg) Standard Deviation 0 8 1.88 0.27 200 9 1.38 0.26 400 9 1.27 0.25 800 9 1.00 0.36 Comparisons of model fits obtained are provided in Table 2-31. Models with homogeneous variance were used because the BMDS Test 2 /rvalue was 0.653. The best fitting model (Hill) was selected based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit p-value (higher value indicates a better fit) and visual inspection. The Hill model had an acceptable BMD to BMDL ratio of 2.5 and is indicated in bold. For the best fitting model a plot of the model is shown in Figure 2-14. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-32. Table 2-31 Summary of BMD Modeling Results for Seminal Vesicle Absolute Weight in Rats Exposed to 1-BP by Inhalation Model" Goodness of fit BMDisd (ppm) BMDLisd (ppm) Basis for model selection />-value AIC Hill ^ 0.429 -47.533 97.3 k. 38.4 The Hill model was selected based on the lowest AIC because the models with adequate goodness of fit />-valuc and adequate fit by visual inspection (including Hill and Exponetial M2 - M5, excluding Power, Polynomial and Linear) had BMDLs < 4-fold apart considered sufficiently close. Exponential (M4) Exponential (M5)b 0.337 -47.235 112 58.4 Exponential (M2) Exponential (M3)° 0.159 -46.484 219 152 Power"1 Polynomial 3oe Polynomial 2of Linear 0.0576 -44.450 299 222 a Constant variance case presented (BMDS Test 2 />-value = 0.653), selected model in bold; scaled residuals for selected model for doses 0,200,400, and 800 ppm were 0.07, -0.43, 0.61, -0.24, respectively. b For the Exponential (M5) model, the estimate of d was 1 (boundary). The models in this row reduced to the Exponential (M4) model. c For the Exponential (M3) model, the estimate of d was 1 (boundary). The models in this row reduced to the Exponential (M2) model. d For the Power model, the power parameter estimate was 1. The models in this row reduced to the Linear model. e For the Polynomial 3° model, the b3 coefficient estimates was 0 (boundary of parameters space). The models in this row reduced to the Polynomial 2° model. For the Polynomial 3° model, the b3 and b2 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Linear model. f For the Polynomial 2° model, the b2 coefficient estimate was 0 (boundary of parameters space). The models in this row reduced to the Linear model. ------- 355 356 357 358 359 360 361 362 363 364 Comparisons of model fits obtained are provided in Table 2-34. The best fitting model (Exponential (M2) with homogeneous variance because the BMDS Test 2 /;-value was 0.144) was selected based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit p-value (higher value indicates a better fit) and visual inspection. The best-fitting model is indicated in bold. For the best fitting model a plot of the model is shown in Figure 2-15. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-35. Table 2-34 Summary of BMD Modeling Results for Sperm Morphology in the Fo Generation Exposed to 1-BP by Inhalation Model3 Goodness of fit BMDisd (ppm) BMDLisd (ppm) Basis for model selection />-value AIC Exponential (M2) Exponential (M3)b 0.787 -401.21 472 327 The Exponential (M2) model was selected based on the lowest AIC from this set of models which have adequate /7-values, adequate fit by visual inspection and the BMDLs are < 1.5-fold apart considered sufficiently close. Power0 Polynomial 3od Polynomial 2oe Linear 0.780 -401.19 473 331 Exponential (M4) 0.534 -399.30 459 230 Hill N/Af -397.69 482 124 Exponential (M5) N/Af -397.69 1 463 112 a Constant variance case presented (BMDS Test 2 />-value = 0.144), selected model in bold; scaled residuals for selected model for doses 0,100,250, and 500 ppm were -0.05, 0.39, -0.53, 0.19, respectively. b For the Exponential (M3) model, the estimate of d was 1 (boundary). Hie models in this row reduced to the Exponential (M2) model. c For the Power model, the power parameter estimate was 1. The models in this row reduced to the Linear model. d For the Polynomial 3° model, the b3 coefficient estimates was 0 (boundary of parameters space). The models in this row reduced to the Polynomial 2° model. For the Polynomial 3° model, the b3 and b2 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Linear model. e For the Polynomial 2° model, the b2 coefficient estimate was 0 (boundary of parameters space). The models in this row reduced to the Linear model. f No available degrees of freedom to calculate a goodness of fit value. ------- 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 The BMD modeling results for sperm motility with non-homogeneous variance (BMDS test 2p- value = 0.0001749) are summarized in Table 2-37. Although the means are sufficiently fit for some models (e.g. the Polynomial 2° model has/rvalue of 0.516) the variances are not well modeled BMDS Test 3 p-value = 0.0426. This result suggests that due to the poor variance modeling for the data it is not reasonable to use BMDS for this endpoint. Instead the NOAEL of 250 ppm was used. Table 2-37 Summary of BMD Modeling Results for Sperm Motility Fo Male Rats Following Inhalation Exposure to 1-BP Model" Goodness of fit BMDisd (ppm) BMDLisd (ppm) Basis for model selection />-value AIC Polynomial 2° 0.516 657.83 386 346 Due to unacceptable fitting of the variances no model was selected. Power 0.334 659.73 399 313 Polynomial 3° 0.330 659.76 397 315 Exponential (M3) 0.324 659.80 402 317 Hill 0.139 661.73 400 323 Polynomial 4° 0.137 661.76 397 314 Exponential (M5) 0.133 661.80 402 317 Linear 0.00132 671.22 237 192 Exponential (M2) Exponential (M4)b 2.10E-04 675.10 226 178 a Modeled variance case presented (BMDS Test 2 />-value = 1.75E-04, BMDS Test 3 />-value = 0.0426), no model was selected as a best-fitting model. b For the Exponential (M4) model, the estimate of c was 0 (boundary). The models in this row reduced to the Exponential (M2) model. To investigate the effect of the poor modeling of the variances on the BMDL the observed standard deviations were considered and the standard deviation at the highest dose is much larger than at the other dose groups. The data set was investigated with the highest dose dropped. The model fits with non-homogeneous variance (BMDS test 2 p-value = 0.0966) are summarized in Table 2-38. Although the means are sufficiently fit for some models (e.g. the Polynomial 2° model has/rvalue of 0.676) the variances are not well modeled BMDS Test 3 /rvalue = 0.0426. ------- 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 Table 2-39 Left Cauda Epididymis Absolute Weight Data Selected for Dose-Response Modeling for 1-BP Dose (ppm) Number of animals Left Cauda Epididymis Weight (mg) Standard Deviation 0 25 0.3252 0.03673 100 25 0.3242 0.03149 250 25 0.3050 0.03556 500 23 0.2877 0.03170 750 22 0.2401 0.03529 The BMD modeling results for left cauda epididymis absolute weight with homogeneous variance (BMDS test 2/rvalue =0.911) are summarized in Table 2-40. The best fitting model (Polynomial 4°) was selected based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit p-value (higher value indicates a better fit) and visual inspection. The Polynomial 4° model had an acceptable BMD to BMDL ratio of 1.4 and is indicated in bold. For the best fitting model a plot of the model is shown in Figure 2-16. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-41. Table 2-40 Summary of BMD Modeling Results for Left Cauda Epididymis Absolute Weight Fo Male R .ats Following Inhalation Exposure to 1-BP Model3 Goodness of fit BMDisd (ppm) BMDLisd (ppm) Basis for model selection />-value AIC Polynomial 4° 0.622 ^ -714.88 438 ^ 313 The Polynomial 4° model was selected based on the lowest AIC from this set of models which have adequate /7-values (excluding Exponential M2 and M4), adequate fit by visual inspection and the BMDLs are < 1.5-fold apart considered sufficiently close. Polynomial 3° 0.565 | -714.69 440 316 Polynomial 2° 0.47 -714.32 437 315 Power 0.430 -714.14 444 317 Exponential (M3) 0.382 -713.91 446 320 Linear 0.133 -712.23 307 256 Hill 0.193 -712.14 444 317 Exponential (M5) 0.166 -711.91 446 320 Exponential (M2) 0.0636 -710.55 289 236 Exponential (M4) 0.0636 -710.55 289 235 a Constant variance case presented (BMDS Test 2 />-value = 0.911), selected model in bold; scaled residuals for selected model for doses 0,100,250, 500, and 750 ppm were -0.21, 0.64, -0.65, 0.26, -0.04, respectively. ------- 435 436 437 438 439 440 441 442 443 444 445 446 447 448 Table 2-42 Right Cauda Epididymis Absolute Weight Data Selected for Dose-Response Modeling for 1-1 :p Dose (ppm) Number of animals Left Cauda Epididymis Weight (mg) Standard Deviation 0 25 0.3327 0.03631 100 25 0.3311 0.04453 250 25 0.3053 0.04188 500 23 0.2912 0.05206 750 22 0.2405 0.04804 The BMD modeling results for right cauda epididymis absolute weight with homogeneous variance (BMDS test 2/rvalue =0.455) are summarized in Table 2-43. The best fitting model (Polynomial 4°) was selected based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit p-value (higher value indicates a better fit) and visual inspection. The Polynomial 4° model had an acceptable BMD to BMDL ratio of 1.4 and is indicated in bold. For the best fitting model a plot of the model is shown in Figure 2-17. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-44. Table 2-43 Summary of BMD Modeling Results for Right Cauda Epididymis Absolute Weight Fo Male R .ats Following Inhalation Exposure to 1-BP Model" Goodness of fit BMDisd (ppm) BMDLisd (ppm) Basis for model selection />-value AIC Polynomial 4° 0.493 -646.60 4851 338 The Polynomial 4° model was selected based on the lowest AIC from this set of models which have adequate /7-values, adequate fit by visual inspection and the BMDLs are < 1.5-fold apart considered sufficiently close. Polynomial 3° 0.442 -646.38 480 334 Linear 0.296 -646.32 371 303 Polynomial 2° 0.376 -646.06 472 327 Power 0.340 -645.86 474 323 Exponential (M3) 0.304 -645.63 473 317 Exponential (M2) 0.196 -645.33 350 277 Exponential (M4) 0.196 -645.33 350 270 Hill 0.142 -643.85 474 323 Exponential (M5) 0.123 -643.63 473 317 a Constant variance case presented (BMDS Test 2 />-value = 0.455), selected model in bold; scaled residuals for selected model for doses 0,100,250, 500, and 750 ppm were -0.09, 0.63, -0.9, 0.44, -0.08, respectively. ------- 461 462 463 464 465 466 467 468 469 470 471 Table 2-45 Estrus Cycle Length Data Selected for Dose-Response Modeling for 1-BP Dose (ppm) Number of animals Estrus cycle Length (days) Standard Deviation 0 25 4.2 0.49 100 25 4.5 1.05 250 25 4.7 0.9 500 23 5.5 2.17 750 22 5.6 1.79 The BMD modeling results for estrus cycle length with non-homogeneous variance (BMDS test 2 /;-value = <0.0001) are summarized in Table 2-46. The means are not adequately fit for any of the models as shown by the goodness of fit where the model with the highest p-value is 0.0065 for the Exponential M4 and M5 models (excluding the Hill model because a BMDL could not be calculated). This result suggests that due to the poor model fit to the data it is not reasonable to use BMDS for this endpoint. Instead the NOAEL of 250 ppm was used. Table 2-46 Summary of BMD Modeling Results for Estrus Cycle Length Fo Female Rats Following Inhalat ion Exposure to 1-BP Model3 Goodness of fit BMDisd (ppm) BMDLisd (ppm) Basis for model selection />-value AIC Hill 0.00656 160.04 145 errorb Due to inadequate fit of the models to the data means (shown by the goodness of fit p- value) no model was selected. Exponential (M4) Exponential (M5)° 0.00650 160.05 157 79.5 Power"1 Polynomial 4°e Polynomial 3of Polynomial 2°g Linear 0.00169 163.13 300 L 205 Exponential (M2) Exponential (M3)h 7.68E-04 164.81 344 244 a Modeled variance case presented (BMDS Test 2 />-value = <0.0001, BMDS Test 3 />-value = 0.506), no model was selected as a best-fitting model. b BMD or BMDL computation failed for this model. c For the Exponential (M5) model, the estimate of d was 1 (boundary). The models in this row reduced to the Exponential (M4) model. d For the Power model, the power parameter estimate was 1. Hie models in this row reduced to the Linear model. e For the Polynomial 4° model, the b4 and b3 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Polynomial 2° model. For the Polynomial 4° model, the b4, b3, and b2 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Linear model. f For the Polynomial 3° model, the b3 coefficient estimates was 0 (boundary of parameters space). The models in this row reduced to the Polynomial 2° model. For the Polynomial 3° model, the b3 and b2 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Linear model. g For the Polynomial 2° model, the b2 coefficient estimate was 0 (boundary of parameters space). The models in this row reduced to the Linear model. h For the Exponential (M3) model, the estimate of d was 1 (boundary). The models in this row reduced to the Exponential (M2) model. ------- 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 of males that did not sire a litter which is equal to the number of nongravid females. The incidence was used for modeling as a dichotomous endpoint. Table 2-49 Fertility Index Data Selected for Dose-Response Modeling for 1-BP Dose (ppm) Number of animals Fertility Index (%) Number Nongravid Females = Males that did not Sire a Litter 0 25 92 2 100 25 100 0 250 25 88 3 500 23 52 12 750 22 0 25 The BMD modeling results for the fertility index are summarized in Table 2-50. The best fitting models were the LogLogistic and Dichotomous-Hill based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit p-value (higher value indicates a better fit) and visual inspection. Dichotomous-Hill model slope parameter was at the boundary value of 18 which indicates some concern for using this model fit and so instead the LogLogistic model selected. The LogLogistic and Dichotomous-Hill models had nearly the same BMDLs with LogLogistic slightly lower (356 ppm) than Dichotomous-Hill (363 ppm). For the best fitting model a plot of the model is shown in Figure 2-18. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-51. Table 2-50 Summary of BMD Modeling Results for Fertility Index of Fo Rats Following Model" Goodness of fit BMDioPct (ppm) BMDLioPct (ppm) Basis for model selection />-value AIC LogLogistic 0.388 75.396 448 356 The LogLogistic model was selected based on the lowest AIC from this set of models which have adequate goodness of fit /7-value (excluding Quantal-Linear, Multistage 2°, Probit and Logistic) and adequate fit by visual inspection and the BMDLs are < 2-fold apart considered sufficiently close. The Dichotomous-Hill model had concern for the fit based on the slope parameter at the boundary and so instead the LogLogistic was selected. Dichoto mous -Hill 0.388 75.396 448 363 Multistage 4° 0.355 75.682 306 219 Weibull 0.253 77.024 361 252 Gamma 0.256 77.045 361 260 LogProbit 0.223 77.357 461 352 Multistage 3° 0.161 78.153 250 202 Logistic 0.0103 80.981 238 182 Probit 0.0031 82.358 208 159 Multistage 2° 0.0152 85.979 173 143 Quantal-Linear 0 106.73 68.4 52.1 ------- 527 528 529 530 531 532 533 534 535 536 The BMD modeling results for the number of implantations sites are summarized in Table 2-53. The best fitting models were the Linear and Power based on Akaike information criterion (AIC; lower values indicates a better fit), chi-square goodness of fit /;-value (higher value indicates a better fit) and visual inspection. Based on the parameter estimate for the Power model it reduced to the Linear, so the Linear model was selected. For the best fitting model a plot of the model is shown in Figure 2-19. The model version number, model form, benchmark dose calculation, parameter estimates and estimated values are shown below in Table 2-54. Table 2-53 Summary of BMD Modeling Results for Implantations Sites in Fo Rats Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Study Model" Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) BMDisd (ppm) P BMDLisd (ppm) Basis for model selection />-value AIC Linear Powerb 0.936 284.66 80.8 56.1 282 188 Linear and Power models were selected based on the lowest AIC from this set of models which have adequate p- values, adequate fit by visual inspection and the BMDLs are < 1.5- fold apart considered sufficiently close. Exponential (M2) 0.901 284.74 74.1 48.1 270 166 Exponential (M4) 0.901 284.74 74.1 37.3 270 138 Polynomial 3° 0.741 286.64 85.5 56.2 295 188 Polynomial 2° 0.724 286.66 ^84.3 56.1 289 188 Hill 0.715 286.67 80.6 55.8 282 195 Exponential (M3) 0.669 286.71 82.3 48.2 278 167 Exponential (M5) N/A° 288.71 82.3 48.2 278 167 a Modeled variance case presented (BMDS Test 2 />-value = 0.0493), selected model in bold; scaled residuals for selected model for doses 0,100,250, and 500 ppm were -0.17, -0.23,1, -1, respectively. b For the Power model, the power parameter estimate was 1. Hie models in this row reduced to the Linear model. c No available degrees of freedom to calculate a goodness of fit value. ------- Parameter Estimates Variable Estimate Default Initial Parameter Values lalpha 12.2915 2.51459 rho -3.77194 0 betaO 15.393 15.7286 betal -0.00952791 -0.01237 Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 23 15.3 15.4 2.53 2.69 -0.166 100 25 14.3 14.4 3.09 3.03 -0.231 250 22 13.8 13 4.23 3.69 1 500 11 9 10.6 4.54 5.41 -0.999 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 -140.289933 5 290.579865 A2 -136.366566 8 1 288.733132 A3 -138.26616 6 288.532319 fitted -138.332408 4 284.664816 R -151.740933 2 307.481866 Tests of Interest Test -2*log(Likelihood Ratio) Test df />-value Test 1 30.7487 6 <0.0001 Test 2 7.84673 3 0.04929 Test 3 3.79919 2 0.1496 Test 4 0.132497 2 0.9359 544 ------- 584 2-57. Also a plot of the Hill model is shown in Figure 2-21 and the model version number, model 585 form, benchmark dose calculation, parameter estimates and estimated values are shown below in 586 Table 2-57. 587 588 Table 2-56 Summary of BMD Modeling Results for Body Weight of Fi Male Rat Pups on 589 PND 28 Following Inhalation Exposure of Parental Rats to 1-BP in a Two-Generation Model" Goodness of fit BMD 1SD (ppm) BMDL 1SD (ppm) BMD 5RD (ppm) BMDL 5RD (ppm) Basis for model selection />-value AIC Exponential (M2) Exponential (M3)b 0.449 411.46 334.07 228.77 174 123 The Exponential (M2) model was selected based on the lowest AIC from this set of models which have adequate /7-values and adequate fit by visual inspection. The Hill model has the lowest BMDL and the BMDL is > 5-fold apart from other model BMDLs not considered sufficiently close, however the BMDL is > 4-fold from the lowest dose and BMD / BMDL ratio is 4-fold and the Exponential (M2) model is in line with the result from pup body weight decreases observed in this study at other time points. Power0 Polynomial 3od Polynomial 2oe Linear 0.406 411.66 345.22 242.64 a 183 133 Hill 0.578 412.17 234.74 85.21 92.2 23.2 Exponential (M4) Exponential (M5)f 0.512 412.29 238.92 95.80 101 36.8 a Constant variance case presented (BMDS Test 2 />-value = 0.785), selected model in bold; scaled residuals for selected model for doses 0,100,250, and 500 ppm were 0.77, -0.88, -0.17, 0.44, respectively. b For the Exponential (M3) model, the estimate of d was 1 (boundary). Hie models in this row reduced to the Exponential (M2) model. c For the Power model, the power parameter estimate was 1. The models in this row reduced to the Linear model. d For the Polynomial 3° model, the b3 coefficient estimates was 0 (boundary of parameters space). The models in this row reduced to the Polynomial 2° model. For the Polynomial 3° model, the b3 and b2 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Linear model. e For the Polynomial 2° model, the b2 coefficient estimate was 0 (boundary of parameters space). The models in this row reduced to the Linear model. f For the Exponential (M5) model, the estimate of d was 1 (boundary). The models in this row reduced to the Exponential (M4) model. 591 ------- Parameter Estimates Variable Estimate Default Initial Parameter Values lnalpha 4.19824 4.17769 rho n/a 0 a 86.7871 78.9392 b 0.000295534 0.000288601 c n/a 0 d n/a 1 Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 23 88.1 86.79 7.6 8.16 0.7717 100 24 82.8 84.26 7.74 8.16 -0.8765 250 21 80.3 80.61 9.04 1 8.16 -0.1719 500 10 76 74.87 9.45 8.16 0.4398 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 i -201.9297 5 413.8595 A2 -201.395 8 418.7901 A3 -201.9297 5 413.8595 R -210.4356 2 424.8712 2 -202.7313 3 411.4626 Tests of Interest Test -2*log(Likelihood Ratio) Test df /j-valuc Test 1 18.08 6 0.006033 Test 2 1.069 3 0.7845 Test 3 1.069 3 0.7845 Test 4 1.603 2 0.4486 599 ------- Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 23 88.1 88 7.6 8.09 0.0793 100 24 82.8 83.3 7.74 8.09 -0.299 250 21 80.3 79.6 9.04 8.09 0.398 500 10 76 76.6 9.45 8.09 -0.235 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 -201.929732 5 413.859464 A2 -201.39503 8 418.790061 A3 -201.929732 5 413.859464 fitted -202.084541 4 412.169082 R -210.435607 2 424.871213 Tests of Interest Test 2*log(Likelihood Ratio) Test df p-value Test 1 18.0812 6 0.006033 Test 2 1.0694 3 0.7845 Test 3 1.0694 3 0.7845 Test 4 0.309618 1 0.5779 606 ------- 623 624 625 To investigate the effect of the poor modeling of the variances on the BMDL, the models were run using the smallest dose standard deviation (2.29), highest (3.87) and pooled (2.89) for all dose levels and the modeling results are summarized in Table 2-61. ------- 626 627 Table 2-61 BMD Modeling Results for Body Weight of F2 Female Rat Pups on PND 14 Following Inhalation Exposure of Model3 Smallest Standard Deviation Pooled Standard Deviation Largest Standard Deviation Ratio BMDLs Smallest to Largest Std Dev Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) />-value AIC /j-valuc AIC />-value AIC Polynomial 3° 0.518 186.54 360 274 0.661 218.16 360 183 0.793 258.09 360 145 1.9 Polynomial 2° 0.318 187.51 304 199 0.485 218.78 304 260 0.667 258.44 304 140 1.4 Power 0.331 188.16 465 247 0.441 219.93 1 465 200 0.564 259.96 460 148 1.7 Exponential (M3) 0.331 188.16 473 249 0.441 219.93 470 202 0.564 259.96 473 143 1.7 Hill N/Ab 190.16 466 248 N/Ab 221.93 465 200 N/Ab 261.96 442 138 1.8 Exponential (M5) N/Ab 190.16 470 249 N/Ab 221.93 470 202 N/Ab 261.96 473 139 1.8 Linear 0.0533 191.08 193 146 0.154 221.07 193 138 0.348 259.74 193 127 1.1 Exponential (M2) 0.0443 191.45 188 139 0.137 221.31 188 131 0.325 259.88 188 119 1.2 Exponential (M4) 0.0443 191.45 188 131 0.137 221.31 188 115 0.325 259.88 188 90.2 1.5 a Constant variance case presented (BMDS Test 2 />-value = 1BMDS Test 3 />-value = 1.), no model was selected as a best-fitting model. b No available degrees of freedom to calculate a goodness of fit value. 628 ------- 678 679 Table 2-67 BMD Modeling Results for Pup Body Weight in Rats Exposed to 1-BP Via 680 Inhalation in ppm BMR = 5% Relative Deviation. Polynomial Model. (Version: 2.20; Date: 10/22/2014) The form of the response function is: Y[dose] = beta_0 + beta_l*dose + beta_2*doseA2 + ... A constant variance model is fit Benchmark Dose Computation. BMR = 5% Relative deviation BMD = 287.938 BMDL at the 95% confidence level = 135.688 Parameter Estimates Variable Estimate Default Initial Parameter Values alpha 10.1836 10.5942 rho n/a 0 betaO 28.9615 28.8658 betal 0 0 beta_2 -0.000017466 -0.000019675 Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 22 29.2 29^ 2.77 3.19 0.35 100 17 28.1 28.8 2.43 3.19 -0.887 250 15 28.4 27.9 3.65 3.19 0.643 500 16 24.5 24.6 4.14 3.19 -0.119 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 -115.551371 5 241.102743 A2 -112.600048 8 241.200097 A3 -115.551371 5 241.102743 fitted -116.227119 3 238.454239 R -125.255153 2 254.510306 ------- Benchmark Dose Computation. BMR = 5% Relative deviation BMD = 154.623 BMDL at the 95% confidence level = 116.114 Parameter Estimates Variable Estimate Default Initial Parameter Values alpha 30.4578 30.9275 rho n/a 0 betaO 49.5516 49.615 betal -0.0160234 -0.0160705 Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 22 49.5 49.6 5.14 5.52 -0.0439 100 17 46.9 47.9 5.03 5.52 -0.784 250 15 47.6 45.5 1 54 5.52 1.44 500 16 40.8 41.5 6.7 5.52 -0.536 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 ^-153.048201 5 316.096402 > A2 -152.146228 8 320.292456 A3 -153.048201 316.096402 fitted -154.572024 3 315.144048 R -163.858303 2 331.716606 Tests of Interest Test -2*log(Likelihood Ratio) Test df /j-valuc Test 1 23.4241 6 0.0006662 Test 2 1.80395 3 0.6141 Test 3 1.80395 3 0.6141 Test 4 3.04765 2 0.2179 705 ------- 743 744 745 746 747 748 749 750 751 752 753 754 The BMD modeling results for decreased brain weight in Fo males with non-homogeneous variance (BMDS test 2/rvalue = 0.000386) are summarized in Table 2-75. Although the variances are non-homogeneous and not well modeled for any of the non-homogeneous variance models the means were well-modeled (the highest /;-value is 0.618 for the Exponential (M4) model with non-homogeneous variances). Table 2-75 Summary of BMD Modeling Results for Brain Weight of Fo Males Following Inhalation Exposure to 1-BP Model3 Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) />-value AIC Exponential (M4) 0.618 -408.61 372 159 Hill 0.340 -406.66 354 107 Exponential (M5) 0.152 -405.52 115 102 Exponential (M2) Exponential (M3)b 0.0868 -405.00 636 453 Power0 Polynomial 4od Polynomial 2oe Linearf 0.0804 -404.83 644 463 Polynomial 3og 0.0804 -404.83 644 463 a Modeled variance case presented (BMDS Test 2p-value = 3.86E-04, BMDS Test 3 p-value = 5.66E-04), no model was selected as a best- fitting model. b For the Exponential (M3) model, the estimate of d was 1 (boundary). The models in this row reduced to the Exponential (M2) model. c For the Power model, the power parameter estimate was 1. The models in this row reduced to the Linear model. 11 For the Polynomial 4° model, the b4 and b3 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Polynomial 2° model. For the Polynomial 4° model, the b4, b3, and b2 coefficient estimates were 0 (boundary of parameters space). The models in this row reduced to the Linear model. e For the Polynomial 2° model, the b2 coefficient estimate was 0 (boundary of parameters space). The models in this row reduced to the Linear model. f The Linear model may appear equivalent to the Polynomial 3° model, however differences exist in digits not displayed in the table. s The Polynomial 3° model may appear equivalent to the Power model, however differences exist in digits not displayed in the table. This also applies to the Polynomial 4° model. This also applies to the Polynomial 2° model. This also applies to the Linear model. To investigate the effect of the poor modeling of the variances on the BMDL, the models were run using the smallest dose standard deviation (0.091), highest (0.177) and the pooled (0.0907) for all dose levels and the modeling results are summarized in Table 2-76. ------- 755 756 Table 2-76 BMD Modeling Results for Brain Weight of Fo Male Rats Following Inhalation Exposure to 1-BP in a Two- Model3 Smallest Standard Deviation Pooled Standard Deviation Largest Standard Deviation Ratio BMDLs Smallest to Largest Std Dev Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) Goodness of fit BMDsrd (ppm) BMDLsrd (ppm) />-value AIC />-value AIC /j-valuc AIC Exponential (M4) 0.0893 -477.73 375 164 0.108 -467.70 375 159 0.553 -303.82 375 78.7 2.1 Hill 0.0423 -476.44 289 106 0.0513 -466.35 1 289 106 0.315 -302.00 289 70.4 1.5 Exponential (M5) 0.0398 -476.34 246 104 0.0484 -466.26 246 103 0.309 -301.97 246 82.4 1.3 Exponential (M2) 0.0238 -475.11 669 515 0.0332 -465.43 669 510 0.503 -304.65 669 420 1.2 Exponential (M3) 0.0238 -475.11 669 515 0.0332 -465.43 669 510 0.503 -304.65 669 420 1.2 Power 0.0223 -474.96 674 523 0.0312 -465.29 674 518 0.496 -304.62 674 430 1.2 Polynomial 4° 0.0223 -474.96 674 ^523 0.0312 -465.29 674 518 0.496 -304.62 674 430 1.2 Polynomial 2° 0.0223 -474.96 674 523 0.0312 -465.29 674 518 0.496 -304.62 674 430 1.2 Linear 0.0223 -474.96 674 523 0.0312 -465.29 674 518 0.496 -304.62 674 430 1.2 Polynomial 3° 0.0223 -474.96 674 523 0.0312 -465.29 674 518 0.496 -304.62 674 430 1.2 757 758 a Constant variance case presented (BMDS Test 2 />-value = 1BMDS Test 3 />-value = 1.), no model was selected as a best-fitting model. ------- Parameter Estimates Variable Estimate Default Initial Parameter Values lnalpha -0.0282712 -1.99881 rho -15.3239 -8.92906 a 1.40066 1.33604 b 0.000120467 0.000129477 c n/a 0 d n/a 1 Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 22 1.4 1.4 0.06 0.07 -0.3121 100 17 1.39 1.38 0.09 0.08 0.3231 250 15 1.37 1.36 0.12 0.09 0.3377 500 15 1.31 1.32 0.1 0.12 -0.3236 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 131.2578 5 -252.5155 A2 134.8828 8 -253.7656 A3 133.1137 6 | -254.2275 R 126.819 2 -249.638 2 132.6574 4 -257.3148 Tests of Interest Test -2*log(Likelihood Ratio) Test df /j-valuc Test 1 16.13 6 0.01309 Test 2 7.25 3 0.06434 Test 3 3.538 2 0.1705 Test 4 0.9127 2 0.6336 827 ------- Table of Data and Estimated Values o 'Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 22 1.47 1.46 0.08 0.08 0.989 100 17 1.43 1.46 0.08 0.08 -1.62 250 15 1.47 1.46 0.06 0.08 0.522 500 16 1.36 1.36 0.1 0.08 -0.00000182 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 144.826466 5 -279.652932 A2 146.516124 8 -277.032248 A3 144.826466 5 -279.652932 fitted 142.841294 3 -279.682588 R 135.116612 2 -266.233223 Tests of Interest Test -2*log(Likelihood Ratio) Test df /j-valuc Test 1 22.799 6 \ 0.0008667 Test 2 3.37932 3 0.3368 Test 3 3.37932 3 0.3368 Test 4 3.97034 2 0.1374 ------- Parameter Estimates Variable Estimate Default Initial Parameter Values lnalpha -0.107405 0.415293 rho 1.46448 1.29675 a 26.8244 26.46 b 0.0174245 0.00510395 c 0.172048 0.15837 d n/a 1 Table of Data and Estimated Values of Interest Dose N Obs Mean Est Mean Obs Std Dev Est Std Dev Scaled Resid 0 5 25.2 26.82 15.25 10.54 -0.3447 10 5 23.8 23.27 7.53 9.5 0.1241 50 5 15.2 13.91 5.54 6.51 0.4434 200 5 5.2 5.3 3.42 3.21 -0.0668 1000 5 4.4 4.62 3.65 2.9 -0.1656 Likelihoods of Interest Model Log(likelihood) # Param's AIC A1 -62.64066 6 137.2813 A2 -54.60856 10 129.2171 A3 -56.01777 7 126.0355 R -73.64274 2 151.2855 4 -56.06343 5 122.1269 Tests of Interest Test -2*log(Likelihood Ratio) Test df /j-valuc Test 1 38.07 8 <0.0001 Test 2 16.06 4 0.002934 Test 3 2.818 3 0.4205 Test 6a 0.09133 2 0.9554 878 879 ------- |