Setting course for a translational pharmacology and a predictive toxicology based on the numerical probability of clinical relevance.

For a significant share of the chemicals, current bioassays mispredicted the outcomes in the reference methods they simulate. For any drug or chemical, and depending on the regulatory or corporate situation, three different approaches calculate the numerical probability by which agreement (or discrepancy) can be statistically expected between (1) the result of a predictive bioassay, and (2) the outcome on its reference method. If such concordance is expected with enough confidence based on a sufficient percentage probability, then specific results from that bioassay can be considered as correctly predictive. The statistical approaches analyzed in this article assist in valuable tasks, including (1) a better translation of the clinical relevance (or insignificance) of specific preclinical findings; (2) waiving unnecessary animal testing (or any other unpredictive testing; e.g., a given in vitro bioassay), and (3) in advancing only the most promising candidates in the pharmaceutical, pesticide, or chemical development process.

Applying Bayesian forecasting to predictive toxicology: The probability of innate carcinogenicity to humans of colorants synthesized from benzidine.

The preclinical identification of health hazards relies on the performance (the historic concordance to the respective gold standard) of regulatorily recommended bioassays. However, any testing with less than 100% sensitivity (or 100% specificity) can deliver false results (outcomes discordant to the respective gold standard). Conversely, the predictive values approach (a.k.a. Bayesian forecasting) weighs (1) the performance of the predictive bioassay (battery, or framework) with (2) the prevalence of -positivity to the respective gold standard- in the most representative category to which the test substance can be allocated. Thus, the predictive values approach (PVA) provides the numeric probability for the toxicity to humans of chemicals that, circumstantially, are evaluable only through nonclinical data. Consequently, the PVA improves the predictivity of nonclinical toxicology, and increases the impact of hazard identifications entirely based on preclinical data. This article aimed to introduce the PVA through a worked example. Due to their toxicological homogeneity and public health relevance, the superfamily of colorants synthesized from benzidine (BZ) or some mutagenic congeners was selected (colorings hereafter mentioned as BZ-related-colorants). Through the PVA, the numeric probability of innate carcinogenicity to humans of 259 BZ-related-colorants was either estimated from rodent carcinogenesis bioassays (RCBs) or predicted from methods alternative to the RCB. A discussion was provided on (1) some limitations and implications of the PVA, and (2) the probable significance of the predictive values figured here for 259 BZ-related-colorings.

The numerical probability of carcinogenicity to humans of some antimicrobials: Nitro-monoaromatics (including 5-nitrofurans and 5-nitroimidazoles), quinoxaline-1,4-dioxides (including carbadox), and chloramphenicol.

Many substances are already tested in the long-term rodent bioassay (RCB). Nonetheless, statements such as the following are common in the regulatory literature: "the significance of the carcinogenicity findings in rodents relative to the therapeutic use of drugs in humans is unknown." (U.S. FDA prescribing information for nitrofurantoin). In the absence of epidemiological data, chemicals carcinogenic in RCBs are typically classified as either possibly or probably carcinogenic to humans, particularly without the -numerical probability for the carcinogenicity to humans- (PPV) of the classified substance. Through the biostatistics-based and regulatorily pertinent -predictive values approach- (PVA), the present study investigated the PPV of several antimicrobials relevant to human or veterinary medicine. A combination of structure-activity relationship, mutagenicity, and tumor-related histopathology was used to resolve reliable and pertinent PPVs. For 62 specific antimicrobials (e.g., carbadox), a 97.9% (or more) probability of carcinogenicity to humans was estimated. For nitrofurantoin, a 99.9% probability of carcinogenicity to humans was reckoned. Therefore, a risk-benefit evaluation on the in-force authorization of nitrofurantoin for uncomplicated human urinary infections is needed. A discussion was provided on the involved mechanisms of carcinogenic action and some regulatory implications of the findings. Neither this study nor the PVA aimed to encourage indiscriminate animal testing but the contrary, to reduce unnecessary or redundant in vivo testing by powering the predictivity of nonclinical toxicology.

The 2-year rodent bioassay in drug and chemical carcinogenicity testing: Performance, utility, and configuration for cancer hazard identification.

For several intended uses of chemicals, the 2-year rodent bioassay (RCB) has been the benchmark method to screen the carcinogenicity to humans of substances, according to the hazard identification sphere. Despite the ongoing controversy around this traditional testing, the RCB is in force and being used by stakeholders. After assembling the RCB's ability to forecast the carcinogenicity to humans of substances, the current review aimed to provide a discussion on the RCB's (1) sensitivity and specificity; (2) utility; (3) configuration, and (4) provisional role in the regulatory policy. In general, RCBs conducted at maximum tolerated doses (MTDs) exhibited a functional ability to (1) not missing the great majority of human carcinogens, and to (2) not responding to the large majority of human non-carcinogens. There is citable evidence supporting the use of MTDs to render RCBs as sensitive as possible, particularly provided the ethically-justified small samples used in RCBs. The literature shows that rodent-specific mechanisms of chemical carcinogenesis contribute significant unspecificity to RCBs. Nonetheless, the paradox between a functional sensitivity and a significant unspecificity can be predictively resolved through the application of Bayesian forecasting. In terms of performance to forecast the carcinogenicity to humans of either genotoxic or non-genotoxic substances, 2-species-RCBs added no value over the rat-RCB. Nevertheless, there is preliminary evidence cautioning that 15% of the rodent carcinogens probably carcinogenic to humans could be missed if mouse-RCBs are indiscriminately discontinued. More than thirteen RCB-related issues relevant to regulatory pharmacology and toxicology were discussed and summarized in this review.

The numerical probability of carcinogenicity to humans of some pharmaceutical drugs: Alkylating agents, topoisomerase inhibitors or poisons, and DNA intercalators.

The nonclinical branch of regulatory pharmacology has traditionally relied on the sensitivity and specificity of regulatorily recommended bioassays. Nonetheless, any predictive testing (eg, safety pharmacology) with less than 100% sensitivity or 100% specificity is prone to deliver false positive or negative results (namely, outcomes discordant to the clinical gold standard). It was recently suggested that the statistics-based and regulatory pertinent "predictive values approach" (PVA) might help to reach a more predictive use of preclinical testing data. To resolve the associated probability of carcinogenicity to humans, the PVA was applied to 37 pharmaceuticals bearing inadequate epidemiological evidence of carcinogenicity, but identifiable as unequivocal mutagens. According to current knowledge, a 98.9% (or more) probability of carcinogenicity to humans was reckoned for those 37 genotoxic drugs. Accordingly, these pharmaceutical drugs might be either scientifically or regulatorily regarded as "carcinogenic to humans." In the USA, European Union, or Canada as examples, the great majority of these 37 pharmaceuticals are authorized for medical use in humans. From the results of the present appraisal, the following is suggested (1) for the pharmaceuticals listed in this report, to include significant carcinogenicity warnings in their prescribing information; (2) to conduct pharmacoepidemiology studies or risk-benefit analyses (if warranted), and (3) based on the respective risk-benefit analyses, to re-evaluate the authorization of hydralazine and phenoxybenzamine as antihypertensives, oxcarbazepine as an anticonvulsant, and phenazopyridine as a urinary tract antimicrobial or analgesic. For the four latter drugs (eg, phenoxybenzamine), a 99.5% probability of carcinogenicity to humans was estimated.

The 2-year rodent bioassay in drug and chemical carcinogenesis testing: Sensitivity, according to the framework of carcinogenic action.

The long-term rodent bioassay (RCB) has been the gold-standard for the pre-marketing prediction of chemical and drug carcinogenicity to humans. Nonetheless, the validity of this toxicity test has remained elusive for several decades. In the quest to uncover the performance of the RCB, its sensitivity (SEN) was charted as the first step. This appraisal was based on (a) chemicals with sufficient epidemiological evidence of carcinogenicity, and (b) other substances with limited epidemiological evidence, or remarkable classifications of carcinogenicity based on mechanistic or pharmacological data. In the present study, chemicals evaluated for their carcinogenicity to humans in IARC Monographs volumes 1-123, U.S. EPA IRIS Assessments, and U.S. NTP RoC were considered. This investigation gathered additional evidence supporting that, in hazard identification, the RCB is unwarranted for mutagenic or direct-acting genotoxicants. However, for purposes of risk assessment or management, the RCB might be justified whenever there is a lack of reliable and/or comprehensive epidemiological data. The RCB exhibited a significantly different SEN for threshold-based human carcinogens compared to non-threshold-based ones. With threshold-based chemicals, to increase the SEN of the testing from 80% (rat-RCB) to 90%, the 2-species RCB might be warranted. Nevertheless, the resolve would depend on the viewpoint, and on the future analysis of the overall performance of the RCB. In terms of SEN, and cancer hazard identification, the comparison between the RCB and alternative methods (e.g. rasH2 mouse, Tg.AC mouse) is now enabled.

Drug excipients, food additives, and cosmetic ingredients probably not carcinogenic to humans reveal a functional specificity for the 2-year rodent bioassay.

Regarding carcinogenicity testing, the long-term rodent bioassay (RCB) has been the test required by most regulatory agencies across the world. Nonetheless, due to the lack of knowledge about its specificity, it has been argued that the RCB is unspecific or even invalid. Because of the substantial limitations of epidemiology to identify chemicals probably not carcinogenic to humans (PNCH), it has been very difficult to address the specificity of the RCB. Nevertheless, because mechanistic/pharmacological data are currently recognized as a valid stream of evidence for the identification of chemical hazards, the road is now open to gain insight into the specificity of the RCB. Based on sound mechanistic/pharmacological data that support the classification of chemicals as PNCH, 100 PNCH substances were gathered in this investigation. Contrary to what was previously forecast, in this study, the RCB exhibited a functional specificity that ranged from 83% to 91%, depending on the settings of the testing (2-species vs. rats only, and the nominal maximum tolerated dose). Other contributions of this work were: (a) enabling the comparison, in terms of specificity, between the RCB and the alternative methods that could replace it (eg, Tg.AC mouse, rasH2 mouse); (b) disclosing what the specificity is for alternative methods that were developed using the RCB as the reference standard; and (c) expanding the previous narrow (only seven substances) set of chemicals identified as not likely to be carcinogenic to humans by hazard identification programs.

The predictivity of the -alert performance- functionality of the OECD QSAR-Toolbox (c/w further issues on the predictivity of nonclinical testing).

The OECD QSAR-Toolbox can be considered a milestone in predictive toxicology. Because of the reliability of its supporting institutions (OECD and ECHA), its broadness in terms of feeder databases, and its predictive capacity, the QSAR-Toolbox is called to have a major role in regulatory toxicology. Recently, a novel functionality was built for the QSAR-Toolbox: the alert performance (AP). This prompted us to analyze the strengths, potentialities, and limitations of this new functionality, especially in the light of a pivotal framework recently discussed in the literature for the predictive use of nonclinical screening and testing. After meticulous analysis, and through some worked examples, high predictive capacity and applicability were found for the AP in both predictive and regulatory toxicology. For a specified chemical, the AP is useful in (a) anticipating its overall results in a given nonclinical test; (b) predicting its overall results regarding a selected toxicological endpoint in humans, and (c) facilitating post- to pre-test probabilities approaches that may support regulatory authorization for waiving the conduction of selected tests in laboratory animals. Furthermore, if a QSAR-Toolbox initiative is developed in or extended to pharmacology (e.g., safety pharmacology, drug abuse potential), it could represent another milestone; one that would give rise to the field of predictive pharmacology.