Introduction

Reconsidering the utility of the National Toxicology Program 2-year rodent cancer bioassay

Our interest in understanding the 2-year rodent cancer bioassay data in the National Toxicology Program (NTP) database began as a limited inquiry as to why lung tumors were induced by 1-bromopropane via inhalation in only female mice and not in male mice, nor male and female rats. Prior to determining via statistical analysis a high degree of discordance across sex and rodent species in the 58 NTP inhalation studies for which tumor data were available up through November 2016, we had expected to find a high degree of concordance in both the development of tumors and the anatomical location of tumors. ¹ The discordance in tumor formation and tumor location in rats and mice via the inhalation route raised the question of whether the inhalation results were idiosyncratic or generalizable. Further evaluation of the NTP 2-year rodent cancer bioassays, where the chemical was administered in the feed, ² by gavage, in drinking water, dermally, or via intraperitoneal injection, ³ demonstrated that rats and mice also differ in tumor formation and tumor location by the additional routes of administration.

After establishing the degree of discordance between rats and mice in tumor development and location in the entire NTP database, we next attempted to understand the chemical structural factors underlying the induction of tumors across rodent species and sex. Toward this end, we recruited our longtime collaborator and Quantitative Structure Activity Relationship expert, Dr Rajni Garg, and her former student Dr Gene Ko. In this phase of the analysis of the NTP database, correlations were determined between and among the Ames mutagenicity, structural alerts of carcinogenicity, Hansch molecular parameters (ClogP, CMR, MgVol), tumor site concordance–multiplicity, and tumorigenicity rank. ⁴ A significant takeaway from this phase of the analysis was the robustness of structural alerts of carcinogenicity in predicting the tendency to induce rodent tumors.

Soon after completing the statistical analysis of the structural alerts and prior to its publication, ⁴ one of us (Smith) was meeting with the Risk Assessment Division of the United States Environmental Protection Agency (USEPA) on a different topic and became aware of the high degree of importance that USEPA placed on the results from the Oncologic™ (Oncologic) expert system. Thus, we next compared the carcinogenicity predictions by the Oncologic expert system with NTP 2-year rodent study tumorigenicity results. ⁵

At this juncture, we had now published five manuscripts ^{1 –5} heavily laden with statistical analysis and placed their voluminous accompanying Supplemental Materials online. This large amount of statistical analysis and interpretation was not readily amenable to review by third parties suggesting the need for a manuscript summarizing the important findings of the analysis of the database. ⁶ One of us (Smith) had been aware since the mid-1980s of the importance of cellular proliferation in benign prostatic hyperplasia under the tutelage of William A. Gardner, Jr, MD. Several years later, Smith was fortunate to take a short-course on carcinogenesis taught by Samuel M Cohen, MD, PhD, and became aware of the seminal work by his group on saccharin induction of bladder cancer in male rats. The background provided by Dr Cohen’s publications provided the knowledge base to facilitate an appreciation of the importance of the efforts by Drs Bruce Ames and Lois Gold illustrating the problem of mitogenesis caused by cytotoxicity in inducing mutagenesis in 2-year rodent cancer bioassays.

In the sixth paper in the series, ⁶ we described the “false-positive” conundrum in the NTP 2-year rodent cancer study database. Expert panels employed by NTP frequently use a weight-of-the-evidence approach to the genotoxicity evaluation of chemicals. This approach entails consideration of older historical genotoxicity results and those garnered from conduction of state-of-the-art genotoxicity assays conducted under Good Laboratory Practices. This protocol introduces an unintended error rate into the genotoxicity evaluation process due to the poor quality of many historical genotoxicity assays. ⁶ In the NTP database, we previously identified 180 chemicals whose current genotoxicity test results are negative but that induce at least one tumor in either rats or mice. ⁶ The large number of NTP studies demonstrating at least one tumor in either rats or mice despite the particular chemical testing negative in the Ames Salmonella mutagenicity test supports the idea that mitogenesis due to cytotoxicity is inducing mutagenesis via amplification of the background mutation rate as described by Cohen and Ellwein, ^7,8 Cohen et al., ⁹ Moolgavkar, ¹⁰ and Ames and Gold. ¹¹

A confluence of fortuitous circumstances not extant when Bruce Ames and Lois Gold in 1990 published their eminently readable and interesting paper titled “Too many rodent carcinogens: Mitogenesis increases mutagenesis” led us to revisit potential regulatory reform despite the past record of intransigence by NTP to properly consider the well-documented and important role of mitogenesis in mutagenesis. First, our recent analysis of the NTP database is in agreement with the many publications of Cohen et al., Moolgavkar and Knudson, and Ames and Gold, and recent studies by Tomasetti and colleagues. ^12,13 Second, the regulatory environment in the United States at least for the near future is tending toward deregulation. Third, NTP and USEPA are in the process of considering alternatives to chronic animal bioassays for the assessment of chemical cancer hazard. This serendipitous but probably transient set of circumstances motivated us to pursue an independent review of our analysis by a panel of experts in chemical carcinogenesis. The resultant commentary is included in this edition of Toxicology and Research Application as “The NTP 2-Year Bioassay: Controversies in Counting Rodent Tumors to Predict Human Cancer” by Sir Colin Leonard Berry, MD, Former Dean London Hospital Medical School; Samuel M Cohen, MD, PhD, University of Nebraska Medical Center; A. Wallace Hayes, PhD, University of South Florida at Tampa; and Norbert E. Kaminski, PhD, Michigan State University.

Now in possession of our six-paper analysis of the NTP database and an insightful commentary by the expert panel, we realized that a panoramic overview of the issue would be illustrative for the nonspecialist reader. This overview is expertly provided by Dr Bruce Ames in his Preface also found in this edition of Toxicology and Research Application. Our six-paper analysis of the NTP database, ^{1 –6} this introduction describing the natural history of the preparation of this special issue of the journal, Dr Ames’ Preface, and the expert panel commentary constituted the first phase of the overall analysis. Following the completion of the first phase of the analysis, we continued to examine further mechanistic issues associated with the NTP database resulting in the publication of four additional papers. ^{14 –17} We encourage the reader to examine these additional papers as they address some of the salient controversies in chemical hazard assessment.

Over the last several years, significant progress in clinical oncology has been made due to the increased specificity of antibody-based biological therapeutics and small molecule inhibitors. ¹⁸ The development of these new therapies was made possible via determination of gene products (proteins) made by specific tumor types or elucidation of specific metabolic and cellular reproductive pathways of the relevant tumors. Even within a specific histological subtype, the particular gene profile is sometimes important. In contrast with the increasing specificity employed in clinical oncology, NTP and USEPA are considering a set of very broad “key characteristics of carcinogens” (key characteristics) to be used in chemical carcinogen hazard assessment. ¹⁹

Currently, the false-positive rate for 2-year rodent studies is about 40%. That is, 40% of chemicals tested to date by NTP are not genotoxic but nonetheless induce at least one type of rodent tumor. ⁶ Employing the “key characteristics” has the potential to push the already high false-positive rate even higher. Over the last 30 years, NTP largely ignored the state-of-the-art understanding developed by Cohen and Ellwein, Moolgavkar, and Ames and Gold of mitogenesis inducing mutagenesis in chemical carcinogenesis as evaluated by the 2-year rodent cancer bioassay. It is essential that upcoming changes in the cancer hazard assessment of chemicals appropriately incorporate the current understanding of specific genetic and epigenetic changes that play a causal role in human cancer development.