How to Rationally Identify Promising Cancer Chemoprevention Agents?
Maarten C. Bosland · JNCI Journal of the National Cancer Institute · 2015
There is no doubt that risk of many cancers is to a large extent modifiable. Not smoking will prevent a large number of human cancers, and vaccination for HPV infection protects against cervical and oropharyngeal cancers. Landmark migrant studies showed that moving from Japan to the United States reduced the risk of stomach cancer while increasing the risk of cancers of the breast, colon, and prostate. Results of epidemiological studies are strongly suggestive that certain micronutrients and food consumption patterns may prevent several cancer types. However, the results of randomized phase III trials testing these notions have been mostly disappointing, and some studies have even indicated harm. Interfering with hormone action has provided a mechanism-based rationale for several phase III chemoprevention trials with antiestrogenic drugs for breast cancer and with 5α-reductase inhibitors for prostate cancer. While these studies have indicated efficacy of these agents, benefit may be restricted to certain cancer subtypes and side effects can occur. Antiestrogenic treatments are only indicated for women to prevent a second primary breast cancer and for women at high risk of a first breast cancer (1). However, antiestrogenic interventions are not generally used for the prevention of primary breast cancer (2,3), and the US Food and Drug Administration has not approved 5α-reductase inhibitors for the prevention of prostate cancer (4,5). Currently, there are no interventions with drugs or micronutrients for chemoprevention of primary cancers in the general population. It is thus not surprising that Potter recently declared cancer chemoprevention a failure (6). The National Cancer Institute (NCI) has invested considerable resources over several decades to support the systematic, mechanism-based identification of effective cancer chemopreventive agents. In this issue of the Journal, Dunn et al. (7) provide a partial update of the NCI’s chemopreventive drug development process, which follows a report published in 1996 (8). A new statistical approach is described designed to improve the identification of the most promising candidate chemoprevention agents. Following a selection process based on published data and other considerations, over 750 agents were evaluated by screening in mechanistic assays. Then, 552 of these were tested in up to six specific laboratory assays with morphologic endpoints, and 210 compounds were also tested for efficacy in at least one in vivo model of colon or mammary carcinogenesis. The in vivo azoxymethane-induced atypical colon crypt foci assay in rats was strongly predictive of efficacy in animal colon cancer models. A DMBA-treated mouse mammary organ culture model was modestly predictive of a protective outcome of mammary carcinogenesis. These results are not surprising because these two morphologic assays are models of the corresponding organ sites. The four other morphologic assays are cell-based models of respiratory tract cancer (human A427 lung cancer cells and rat primary tracheal epithelial cells) and squamous epithelial cancers (primary human neonatal foreskin cells and mouse JB4 epidermal cancer cells). The foreskin cell assay was modestly predictive of in vivo mammary cancer prevention, but the three other assays were not predictive of efficacy in either mammary or colon carcinogenesis models. However, by combining the results of all six assays and using an elegant statistical approach to analyze the data, the ability of the assays to predict a positive outcome in animal models of colon and breast cancer could be enhanced, sometimes substantially so. The authors propose that using this novel statistical approach will facilitate rational selection of promising agents for further study and may be useful in organ sites other than the colon or breast. Thus, the basic premise of the NCI chemopreventive drug development program is that these morphologic assays are predictive of efficacy not only in animal models of breast and colon cancer but also in models of other malignancies. Therefore, it is disappointing that no information was provided about whether any of the six morphological assays could predict negative or positive outcomes of animal efficacy studies in models of other important malignancies, such as prostate cancer. Such analyses would be particularly valuable for those efficacy studies in animal models that were successfully predictive of the outcome of randomized trials testing the same agents. The outcomes of both animal studies and chemoprevention trials are available for several of the agents tested by the NCI. For example, selenomethionine and α-tocopherol did not reduce prostate cancer in animal models and were negative in large randomized trials, whereas COX-2 inhibitors were preventive for colon carcinogenesis in animals and had the same activity in clinical studies. Given the null results obtained in several large and expensive randomized cancer chemoprevention studies, it would be important and timely to conduct a more comprehensive objective evaluation of the success of the NCI’s approach to rational identification and development of effective chemoprevention agents and to compare the results of such analysis with the use of epidemiologic results or secondary analyses of clinical trials to select seemingly promising compounds for clinical trials. Excellent examples of the effectiveness of the NCI’s approach to rational selection of chemopreventive agents are studies showing prevention of colon carcinogenesis in rats by a combination of the NSAID sulindac and the ornithine decarboxylase inhibitor DFMO (9,10) and a subsequent randomized trial showing the prevention of recurrence of sporadic colorectal adenomas by this drug combination with low toxicity (11). These findings are currently being extended in two Phase III trials (NCT01349881 and NCT01483144). COX inhibitors may also prevent colon polyps and cancer in patients with familial adenomatous polyposis (12,13), and a recommendation to use aspirin for the prevention of colon cancer has just been proposed (14). The pessimism of Potter (6) about cancer chemoprevention may be understandable. However, the modifiable nature of many human cancers and the effectiveness of anti-estrogens against breast cancer strongly suggest that chemoprevention of human cancer is possible. Furthermore, several promising agents have been identified on the basis of their considerable efficacy in preclinical models (15–19). More than 25 years of preclinical research were required to support the report by Dunn et al. (7), while over the same time period several very expensive Phase III clinical studies did not turn out as hypothesized. In retrospect, this suggests that identification and development of chemopreventive agents may have been more successful if funds had been more heavily devoted to preclinical research. It would seem to be time for the NCI to just to do that: invest in preclinical chemoprevention research in an intensive fashion and use existing and new data, new and existing preclinical models, and investigator expertise to: 1) accelerate identification of the preclinical assays and animal models that are most strongly predictive of outcomes (negative and positive) of randomized studies and 2) use a rational selection process to identify existing and newly discovered agents or repurposed drugs (and agent combinations) that have the greatest potential for clinical chemopreventive efficacy; these agents can then be fast-tracked into efficacy testing in clinical trials, recognizing that these are time consuming. Maarten Bosland receives funding from the National Institutes of Health (CA172169) and the Prevent Cancer Foundation. The funders had no role in the writing of the editorial nor the decision to submit it for publication. The author has no conflicts of interest to report. The author wishes to thank Drs. Alan M. Diamond, Peter H. Gann, and David L. McCormick for their critical comments.