Prediction of the Carcinogenic Potential of Human Pharmaceuticals Using Repeated Dose Toxicity Data and Their Pharmacological Properties

Abstract

In an exercise designed to reduce animal use, we analyzed the results of rat subchronic toxicity studies from 289 pharmaceutical compounds with the aim to predict the tumor outcome of carcinogenicity studies in this species. The results were obtained from the assessment reports available at the Medicines Evaluation Board of the Netherlands for 289 pharmaceutical compounds that had been shown to be non-genotoxic. One hundred forty-three of the 239 compounds not inducing putative preneoplastic lesions in the subchronic study did not induce tumors in the carcinogenicity study [true negatives (TNs)], whereas 96 compounds were categorized as false negatives (FNs) because tumors were observed in the carcinogenicity study. Of the remaining 50 compounds, 31 showed preneoplastic lesions in the subchronic study and tumors in the carcinogenicity study [true positives (TPs)], and 19 only showed preneoplastic lesions in subchronic studies but no tumors in the carcinogenicity study [false positives (FPs)]. In addition, we then re-assessed the prediction of the tumor outcome by integrating the pharmacological properties of these compounds. These pharmacological properties were evaluated with respect to the presence or absence of a direct or indirect proliferative action. We found support for the absence of cellular proliferation for 204 compounds (TN). For 67 compounds, the presence of cellular hyperplasia as evidence for proliferative action could be found (TP). Therefore, this approach resulted in an ability to predict non-carcinogens at a success rate of 92% and the ability to detect carcinogens at 98%. The combined evaluation of pharmacological and histopathological endpoints eventually led to only 18 unknown outcomes (17 categorized as FN and 1 as FP), thereby enhancing both the negative and positive predictivity of an evaluation based upon histopathological evaluation only. The data show the added value of a consideration of the pharmacological properties of compounds in relation to potential class effects, both in the negative and positive direction. A high negative and a high positive predictivity will both result in waiving the need for conducting 2-year rat carcinogenicity studies, if this is accepted by Regulatory Authorities, which will save large numbers of animals and reduce drug development costs and time.

Threshold of Toxicological Concern—An Update for Non-Genotoxic Carcinogens

“…Within this comparison, the differences between the NOELs were not higher than the well-established extrapolation factor of 2 for subchronic to chronic exposure (EFSA, 2012 ). The data support the hypothesis that there is no significant added value of chronic studies neither with respect to POD (Braakhuis et al, 2018 ) nor concerning the predictability of the effect (van der Laan et al, 2016 ; Woutersen et al, 2016 ) for non-genotoxic carcinogens. Nevertheless, we used the chronic dataset for threshold derivation simply, as it was larger than that for subchronic effects and covers a greater number of compounds.…”

“…A slight shift to higher values is observed for the neoplastic NOELs compared with non-neoplastic and overall NOEL values ( Figure 1 ). This finding raises the question of whether the data support the widely spread thesis that in rodent bioassay tumors occur at higher doses, or even only at the highest tested dose (HTD) (Anisimov et al, 2005 ; van der Laan et al, 2016 ). In the dataset, tumors occurred in 55% of the compounds at the LOEL of the study.…”

Threshold of Toxicological Concern—An Update for Non-Genotoxic Carcinogens

“…Within this comparison, the differences between the NOELs were not higher than the well-established extrapolation factor of 2 for subchronic to chronic exposure (EFSA, 2012 ). The data support the hypothesis that there is no significant added value of chronic studies neither with respect to POD (Braakhuis et al, 2018 ) nor concerning the predictability of the effect (van der Laan et al, 2016 ; Woutersen et al, 2016 ) for non-genotoxic carcinogens. Nevertheless, we used the chronic dataset for threshold derivation simply, as it was larger than that for subchronic effects and covers a greater number of compounds.…”

“…A slight shift to higher values is observed for the neoplastic NOELs compared with non-neoplastic and overall NOEL values ( Figure 1 ). This finding raises the question of whether the data support the widely spread thesis that in rodent bioassay tumors occur at higher doses, or even only at the highest tested dose (HTD) (Anisimov et al, 2005 ; van der Laan et al, 2016 ). In the dataset, tumors occurred in 55% of the compounds at the LOEL of the study.…”

EMA commentary on the ICH guideline for testing for carcinogenicity of pharmaceuticals

“…This group assisted also in the evaluation of the PhRMA dataset, which has been extended by similar data from FDA and the Japanese Pharmaceutical Manufacturers Association (JPMA), making a total of 298 compounds 12 . A renewed and extended evaluation of the EU data confirmed the outcome of this study 13 . These analyses have set the tone to highlight the pharmacology of the compounds as a primary factor.…”

ICH S1 prospective evaluation study: weight of evidence approach to predict outcome and value of 2-year rat carcinogenicity studies. A report from the regulatory authorities subgroup

“…The RND that initiated the PES included a description of WoE factors that should be addressed in a CAD ( ICH, 2013 ). These recommendations were informed by prior retrospective studies that identified pharmacological and toxicological attributes of pharmaceuticals that correlated with a negative or positive tumor outcome in 2-year rat studies ( Sistare et al, 2011 ; Van der Laan et al, 2016a ; Van der Laan et al, 2016b ). The PhRMA dataset ( Sistare et al, 2011 ), which formed the primary basis for the prospective evaluation study, consisted of 182 compounds, and an additional 76 compounds were later included from the IARC dataset.…”

“…The PhRMA dataset (without IARC data) was enlarged by data from FDA and JPMA to approximately 255 compounds ( Van der Laan et al, 2016a ). Another dataset of 289 compounds was analyzed later that year ( Van der Laan et al, 2016b ). In the study presented herein, these attributes were applied in a prospective manner to predict the outcome and potential value of 2-year rat studies that had not yet been completed.…”

Threshold of Toxicological Concern—An Update for Non-Genotoxic Carcinogens

“…Within this comparison, the differences between the NOELs were not higher than the well-established extrapolation factor of 2 for subchronic to chronic exposure (EFSA, 2012 ). The data support the hypothesis that there is no significant added value of chronic studies neither with respect to POD (Braakhuis et al, 2018 ) nor concerning the predictability of the effect (van der Laan et al, 2016 ; Woutersen et al, 2016 ) for non-genotoxic carcinogens. Nevertheless, we used the chronic dataset for threshold derivation simply, as it was larger than that for subchronic effects and covers a greater number of compounds.…”

“…A slight shift to higher values is observed for the neoplastic NOELs compared with non-neoplastic and overall NOEL values ( Figure 1 ). This finding raises the question of whether the data support the widely spread thesis that in rodent bioassay tumors occur at higher doses, or even only at the highest tested dose (HTD) (Anisimov et al, 2005 ; van der Laan et al, 2016 ). In the dataset, tumors occurred in 55% of the compounds at the LOEL of the study.…”

EMA commentary on the ICH guideline for testing for carcinogenicity of pharmaceuticals

“…This group assisted also in the evaluation of the PhRMA dataset, which has been extended by similar data from FDA and the Japanese Pharmaceutical Manufacturers Association (JPMA), making a total of 298 compounds 12 . A renewed and extended evaluation of the EU data confirmed the outcome of this study 13 . These analyses have set the tone to highlight the pharmacology of the compounds as a primary factor.…”

ICH S1 prospective evaluation study: weight of evidence approach to predict outcome and value of 2-year rat carcinogenicity studies. A report from the regulatory authorities subgroup

“…The RND that initiated the PES included a description of WoE factors that should be addressed in a CAD ( ICH, 2013 ). These recommendations were informed by prior retrospective studies that identified pharmacological and toxicological attributes of pharmaceuticals that correlated with a negative or positive tumor outcome in 2-year rat studies ( Sistare et al, 2011 ; Van der Laan et al, 2016a ; Van der Laan et al, 2016b ). The PhRMA dataset ( Sistare et al, 2011 ), which formed the primary basis for the prospective evaluation study, consisted of 182 compounds, and an additional 76 compounds were later included from the IARC dataset.…”

“…The PhRMA dataset (without IARC data) was enlarged by data from FDA and JPMA to approximately 255 compounds ( Van der Laan et al, 2016a ). Another dataset of 289 compounds was analyzed later that year ( Van der Laan et al, 2016b ). In the study presented herein, these attributes were applied in a prospective manner to predict the outcome and potential value of 2-year rat studies that had not yet been completed.…”

Threshold of Toxicological Concern—An Update for Non-Genotoxic Carcinogens

“…Within this comparison, the differences between the NOELs were not higher than the well-established extrapolation factor of 2 for subchronic to chronic exposure (EFSA, 2012 ). The data support the hypothesis that there is no significant added value of chronic studies neither with respect to POD (Braakhuis et al, 2018 ) nor concerning the predictability of the effect (van der Laan et al, 2016 ; Woutersen et al, 2016 ) for non-genotoxic carcinogens. Nevertheless, we used the chronic dataset for threshold derivation simply, as it was larger than that for subchronic effects and covers a greater number of compounds.…”

“…A slight shift to higher values is observed for the neoplastic NOELs compared with non-neoplastic and overall NOEL values ( Figure 1 ). This finding raises the question of whether the data support the widely spread thesis that in rodent bioassay tumors occur at higher doses, or even only at the highest tested dose (HTD) (Anisimov et al, 2005 ; van der Laan et al, 2016 ). In the dataset, tumors occurred in 55% of the compounds at the LOEL of the study.…”

EMA commentary on the ICH guideline for testing for carcinogenicity of pharmaceuticals

“…This group assisted also in the evaluation of the PhRMA dataset, which has been extended by similar data from FDA and the Japanese Pharmaceutical Manufacturers Association (JPMA), making a total of 298 compounds 12 . A renewed and extended evaluation of the EU data confirmed the outcome of this study 13 . These analyses have set the tone to highlight the pharmacology of the compounds as a primary factor.…”

ICH S1 prospective evaluation study: weight of evidence approach to predict outcome and value of 2-year rat carcinogenicity studies. A report from the regulatory authorities subgroup

“…The RND that initiated the PES included a description of WoE factors that should be addressed in a CAD ( ICH, 2013 ). These recommendations were informed by prior retrospective studies that identified pharmacological and toxicological attributes of pharmaceuticals that correlated with a negative or positive tumor outcome in 2-year rat studies ( Sistare et al, 2011 ; Van der Laan et al, 2016a ; Van der Laan et al, 2016b ). The PhRMA dataset ( Sistare et al, 2011 ), which formed the primary basis for the prospective evaluation study, consisted of 182 compounds, and an additional 76 compounds were later included from the IARC dataset.…”

“…The PhRMA dataset (without IARC data) was enlarged by data from FDA and JPMA to approximately 255 compounds ( Van der Laan et al, 2016a ). Another dataset of 289 compounds was analyzed later that year ( Van der Laan et al, 2016b ). In the study presented herein, these attributes were applied in a prospective manner to predict the outcome and potential value of 2-year rat studies that had not yet been completed.…”