Is saccharin safe? Animal testing revisited.

The ability to predict harmful effects of chemicals, drugs, and food additives has historically depended on the process of extrapolating from animal studies. The current guidelines of the U.S. Food and Drug Administration recommend nine animalbased studies to assess the safety of additives used directly in food handling or preparation or at the table in amounts in excess of 1 part per million in the diet. With respect to the controversial food additive saccharin, there remains an unresolved debate about its future as a commercial product, despite a seemingly exhaustive list of studies that have been completed to date. The article by Takayama et al. (1) in this issue of the Journal gives us further pause and forces us to think about our present use of animal tests for risk assessment. The issue of the carcinogenicity of saccharin has been the subject of numerous animal and epidemiologic studies. Results from standard long-term rodent carcinogenicity tests indicate that sodium saccharin is a weak carcinogen and tumor promoter in the male rat bladder, particularly when it is administered in a two-stage, initiation–promotion protocol with specific bladder genotoxic carcinogens such as N-butyl-N-(4-hydroxybutyl)nitrosamine and N-[4-(5-nitro-2-furyl)-2-thiazolyl]formamide (2). Results in female rats and in mice have been equivocal (3), and several hypotheses have been presented for the specificity of saccharin to induce bladder cancer in male rats. It is widely believed that unique physiologic conditions in the male rat bladder, especially the presence of the protein a2m-globulin, high pH, and high [Na], result in the formation of precipitates in the bladder, causing irritation, hyperplasia, and ultimately tumors. In the absence of any conclusive epidemiologic data that saccharin is associated with increased tumors in humans, it has been argued that the mechanism for saccharin induction of bladder tumors in rats is specific to the physiology of the male rat bladder and that saccharin should be taken off the National Toxicology Program’s list of agents that are “reasonably anticipated to be carcinogen[s].” Thus, given these unique circumstances for this particular carcinogen, it might also be argued that the rodent bioassays of saccharin were irrelevant for human studies. Indeed, the practice of giving animals the highest possible amount of the substance under study—the maximum tolerated dose—long a requirement for such bioassays, has raised questions about the relevance of rodent bioassays in general (4). The study presented in this issue of the Journal by Takayama et al. (1) was undertaken “to determine the effects of long-term feeding of sodium saccharin to three species of nonhuman primates.” No effects were found. Sodium saccharin did not cause increased urothelial cell proliferation, bladder tumors, or the appearance of large crystals in the urine of the monkeys. How should/can these data be interpreted? Is the evidence persuasive that sodium saccharin is not carcinogenic for humans because it was without effect in nonhuman primates? One must look at this study carefully before judging whether it warrants primacy in the assessment of human risk from sodium saccharin. This study was begun in 1970, before any information had been elucidated about the mechanism of saccharin-induced bladder carcinogenesis in the rodent. We now recognize that, under the conditions of the experiment, it is unlikely this crystallization phenomenon would have been observed; i.e., the dose of sodium saccharin was insufficient, and there are important differences between the composition of primate urine and rodent urine. Moreover, the small number of animals, the multiplicity of species, and the low doses of saccharin greatly reduced the sensitivity of the study to subtle effects. With the benefit of hindsight, one may question whether there was a high probability that compelling information could have been gleaned from this study. The level of precision in the data, given the multiplicity of species and the small number of animals, was less than optimal, limiting the ability to extrapolate the results to humans. This latter issue also brings to mind how approaches to animal tests have changed over the last few decades since the beginning of this study. For example, the “three Rs” have become more of a driving force. The three Rs were first described in a book by Russell and Burch published in 1959 (5) and are defined as methods that refine existing animal protocols by minimizing pain and distress, reduce animal usage to the minimum necessary to produce statistically significant data, or replace whole animal methods. In a recent article by Balls et al. (6), the concepts originally presented by Russell and Burch were reaffirmed, and numerous recommendations were put forth. The concept of reduction includes increasing the level of precision, given the number of animals, as well as optimizing experimental design and statistical considerations. Adherence to these prin-