Abstract
Prediction of the human carcinogenic risk of xenobiotic exposure is one of the most difficult challenges faced by toxicologists. The rodent lifetime bioassays (long-term carcinogenicity studies in rat and mice) have been used for the assessment of potential carcinogenic risk to humans for more than 40 years. There has been considerable debate and controversy over the years on the limitations and relevance of the bioassays. The International Conference on Harmonization’s (ICH’s) acknowledgment of the limited utility of conventional the 2-year mouse bioassay for pharmaceuticals in 1996 provided further encouragement to explore new approaches for assessing carcinogenic potential. The current ICH recommendation is to conduct one long-term rodent carcinogenicity study, typically in rats, plus another in vivo test, either another bioassay in a second rodent species (typically, mice) or a short- or medium-term rodent test (ICH, 1998, “Carcinogenicity: Testing for Carcinogenicity or Pharmaceuticals,” topic No. S1B). The rat carcinogenicity study is the longest and largest animal study in the battery of toxicology studies required for the registration of pharmaceuticals. Several attempts have been made to predict the outcome of carcinogenicity studies based on data generated in shorter term studies. 1,2 The implication is an elimination of the requirement to conduct the rat bioassay, leading to a substantial reduction in animal use and better use of resources for drug development.
Reddy et al 3 (this issue) present a retrospective analysis of the pharmaceutical industry’s experience with rat carcinogenicity studies and chronic toxicity studies. They conclude by offering a new pragmatic approach based on the idea that the absence of evidence for preneoplasia in any tissue in a rat chronic toxicity study can accurately predict the negative outcome of a 2-year carcinogenicity study in the species. Although the idea is not controversial as a concept, the difficulties arise with the definition of what constitutes “evidence for preneoplasia” and what study duration is necessary for accuracy.
The data set, supportive of the authors' conclusion, came from the internal archives of Merck & Co, Inc, and from freedom-of-information data publicly available on rat-based 6-month and 12-month chronic toxicology and 2-year carcinogenicity studies for pharmaceuticals. The derived database has several features that affect its representativeness to some degree—specifically, relatively small numbers of drugs and some data from old studies. The authors made every effort to maximize the robustness of the analysis. Inconsistencies may have occurred because information on individual drugs varied from a full data set (Merck compounds) to a secondary source of summarized data (Leadscope Database). That the majority of compounds are approved drugs does not appear to be a bias for the analysis.
The authors briefly present the method and criteria used to identify which histopathologic changes were indicative of preneoplasia in each case. Many pathologists would probably find a follow-up publication useful, discussing the details of the methodology and the few cases where difficult decisions were made. Indeed, it may be unclear to some pathologists how diagnoses such as multinucleated cells or tumors would fit within one of the three diagnostic categories (hyperplasia, cellular hypertrophy, and atypical cellular foci) indicative of preneoplasia, as defined by the authors. However, one could argue that this is an unnecessary complication for the purpose of the exercise, which is to decide whether a study is positive on the basis of the information in hand. The rationale for the decision for compounds in the Leadscope Database—that all incidences higher than the control values were considered positive—needs to be discussed given that it may create false-positive or false-negative results, as in those cases where the control values in the study are high or low, respectively.
The proposed method for application of the 6-month chronic toxicology data identified most noncarcinogens but mis-identified 5 compounds (Tumors were observed in the rat carcinogenicity study but no preneoplastic changes were found in the 6 month study). As expected, the negative predictivity improved when based on the 12-month study, with the authors considering the improvement marginal. Given the facts that the five false negatives are marketed drugs approved for non-life-threatening conditions and are associated with rat-specific mechanisms, the authors conclude that absence of preneoplasia in the whole animal in 6-month rat studies is a reliable predictor of negative tumor outcome in 2-year studies. They suggest that the pragmatic regulatory acceptance of such drugs as noncarcinogens in rats—specifically, drugs with no preneoplastic histopathology signal in any tissue in the 6-month chronic study—would make the rat carcinogenicity study unnecessary without compromising human safety. In this respect, another in vivo test in another rodent species is still required per the ICH. These five false-negative cases will likely cause much debate within the scientific and regulatory community. Let us hope that it will result in a consensus on whether they invalidate the entire hypothesis and proposal.
The authors went one step further in their database analysis. Consistent with others, they found that the predictivity of preneoplastic changes in 6- and 12-month chronic rat toxicology studies with tumor outcome in rat 2-year carcinogenicity studies is poor at the tissue level. They present data supportive of the more controversial hypothesis that evidence of preneoplasia in any tissue may serve as a sensitive predictor of tumor outcome in any tissue in the whole animal and not necessarily the same tissue. Applying this whole-animal approach to the database, they found that in chronic toxicology studies, most rat carcinogens in a long-term bioassay showed histopathologic signals for preneoplasia. The authors' interpretation is that preneoplastic changes at any site may be indicative of an increase in tumor incidence in that organ or in an organ at a distant site. Although the underlying pathobiology is well known in some specific cases—liver enzyme induction and thyroid tumors or mammary gland tumor and increased prolactin pituitary production, for example—other associations are observations with no biological explanation at this point in time. Uncovering the putative mechanism involved requires significant time and resources, as the authors acknowledge. However, this provocative suggestion should stimulate the generation of new ideas and lines of research that could lead to a greater understanding of the pathobiology of carcinogenesis. The authors propose that 2-year rat carcinogenicity studies be required for compounds with preneoplastic signals in chronic toxicity studies.
From a drug development perspective, the main contribution of this article is the proposal to eliminate the need for a 2-year rat carcinogenicity study when no preneoplastic signal has been identified in chronic toxicity studies. This work, in part, encouraged the Pharmaceutical Research and Manufacturing Association to sponsor a collaborative project that expanded the database with additional cases and parameters, such as genotoxicity and hormonal endpoints. The Association has shared its preliminary results of the initiative with key regulatory authorities. As such, making the details of the database analysis available for public discussion would be useful. A large scientific debate on the issue should help forge a consensus on the acceptance of the proposal.
The consequences of such a paradigm shift on the drug development process need to be fully evaluated. Currently, cancer study protocols and dose levels are typically discussed by sponsors with the Food and Drug Administration while clinical trials are ongoing. If this proposal is adopted, the dose levels and protocol endpoints in the protocols of chronic studies would likely be submitted to the Administration. There will also likely be situations in which the sponsor’s conclusion of “no evidence of preneoplastic changes in the chronic studies” will be challenged by regulatory authorities. Thus, there may be situations where these regulatory interactions cause delays in the drug development program that may not be compensated by the elimination of the rat bioassay.
Finally, although the proposal, if valid, would be a welcome improvement of current practices, it should be kept in mind that the ultimate goal is not to predict carcinogenesis in rats but to predict human carcinogenic risk. The authors should be congratulated for their work. It has the potential to greatly affect the drug development and regulatory approval process.