• © 2008 by American Society of Clinical Oncology

Predicting Fluorouracil Toxicity: Can We Finally Do It?

Currently, no reliable markers of sensitivity or resistance to fluorouracil (FU) have been validated to permit their use as a standard of care for the management of patients with cancer, despite the large number of studies attempting to identify useful molecular predictors of response to treatment. This may be attributed to the complexity of the involved molecular events and/or incomplete understanding of the role of downstream cell signaling pathways in response to chemotherapy-induced stress. In addition and most important is the lack of a comprehensive, well-designed, standardized molecular approach that could evaluate all of these events.¹ To date, the majority of identified predictive markers of response to FU in patients with cancer have been at relatively upstream levels of drug metabolism (eg, dihydropyrimidine dehydrogenase [DPD] enzyme), drug target (eg, thymidylate synthase enzyme [TYMS]), and DNA repair pathways. The fact that molecular events might have overlapping or antagonistic function and could possibly be activated after chemotherapy-induced stress renders the use of a single marker of response unlikely in the prediction of outcome.¹ Because early detection of patients with cancer who are at risk of developing life-threatening toxicity to FU might allow dose reductions or selection of an alternative treatment regimen, various genotypic and phenotypic methods have been developed to predict toxicity and/or response. None, however, has proven to be reliable in prospective clinical studies, demonstrating a great margin of uncertainty.

In the study by Schwab et al² in the current issue of the Journal of Clinical Oncology, the authors have examined a few of the previously suggested markers of FU toxicity in a large, albeit diverse group of patients. The experimental design focused on an examination of only a few sequence variants in three selected genes (DPYD, the gene that encodes DPD enzyme; TYMS; and methylene tetrahydrofolate reductase [MTHFR]), coupled with nongenetic factors (sex and diet) previously suggested to be associated with severe toxicity from FU. Although the use of pharmacogenetic methods in the risk evaluation of patients undergoing chemotherapy is to be commended, without a comprehensive approach in which all of the key regulatory mechanisms and drug targets are also evaluated, there is a risk of oversimplification and misinterpretation. Given the nomogram proposed in the study by Schwab et al, ² to estimate the probability that a given patient will develop toxicity, the question arises whether the experimental approach to predict toxicity in such a diverse group of cancer patients (colon cancer, other gastrointestinal tumors, cancer of unknown origin, and breast cancer) can be expected to have reliable true predictive value with the limited number of selected genetic/nongenetic parameters.

Clinical studies, prospectively using genetic tests to predict FU toxicity, have raised numerous concerns because of the lack of inclusion of important factors that could contribute to FU toxicity. This has highlighted the necessity of a thorough re-evaluation of the parameters used (genetic/nongenetic) in screening tests to decrease the incidence of false positive or negative predictions. For example, on the catabolic side of FU metabolism, routine detection of DPD deficiency in patients with cancer likely to undergo FU therapy has been examined.^3-6 However, DPD deficiency has been observed in a relatively small percentage of patients with grade 3 to 4 FU toxicity,^3,7 leaving a large number of patients with an unexplained molecular basis of toxicity. Although the assessment of DPD activity is informative about the likelihood of DPD deficiency, and a few DPYD alleles (eg, DPYD*2A⁴ and DPYD*13⁸) have been associated with decreased DPD activity, DPD deficiency has also been detected in patients with wild-type DPYD.⁹ Scientific evidence supporting the role of epigenetic mechanisms, mainly methylation, in regulating DPD enzyme activity has been observed in a subset of patients who had no sequence abnormalities in their DPYD gene.⁹ In these patients (in contrast to the findings by Schwab et al²), the regulatory elements in the DPYD promoter region were methylated,⁹ suggesting that genetic and epigenetic mechanisms could act separately or in concert to downregulate DPD activity. Recently, it has been reported that individuals with mutations in genes downstream of DPD (dihydropyrimidinase [DHP]¹⁰ and beta-ureidopropionase [BUP1]^11,12) had altered activity of these two enzymes, which resulted in impairment of the uracil catabolic pathway.^10,12 This finding is supported by a recently developed uracil breath test¹³ that evaluates the integrity of the entire catabolic pathway (DPD, DHP, and BUP1 enzymes). Although the frequency of deficiency in DHP and BUP1 enzymes has been previously reported to be low,^6,14,15 recent studies suggest that both might occur more frequently than previously believed.^10,12

On the anabolic side of FU metabolism, there is also complexity owing to both the multiple enzymatic steps (eg, orotate phosphoribosyl transferase^1,16 thymidine phosphorylase,¹ and uridine phosphorylase,¹⁷ and thymidine kinase^17,18) that must convert FU to FU nucleotides and the multiple sites of action (TYMS inhibition, incorporation into DNA and RNA). In the present article, the authors² have neglected other genes important for FU anabolic activity as well as other sites of action and have chosen to examine only selected mutations in TYMS, the variable number tandem repeats at the 5′UTR and the 6–base pair deletion at the 3′UTR of the TYMS gene. However, other mutations such as a C to G change in the second repeat of the TYMS enhancer region have also been reported to be associated with increased tumor response to FU. Lastly, because of the significant role of 5,10 methylene tetrahydrofolate in TYMS inhibition by fluoro-deoxy uridine monophosphate, assessment of the key genes regulating the folate pool in addition to methylene tetrahydrofolate reductase might also be important.

Although Schwab et al² have focused on the use of FU as monotherapy, currently FU monotherapy is not recommended for the treatment of most solid tumors for which it has historically shown only limited antitumor activity. The use of multidrug combinations of 5U/folinic acid with other drugs has become more standard. Thus in colorectal cancer (CRC), either irinotecan or oxaliplatin has been administered in combination with FU/folinic acid, resulting in significantly improved response rates approaching 40% to 50% with increased overall survival. New agents, such as the monoclonal antibodies cetuximab (an epidermal growth factor receptor inhibitor) and bevacizumab (a vascular endothelial growth factor inhibitor)¹⁹ have shown promising clinical benefit for patients with metastatic CRC.²⁰ Despite the current improvement in treatment outcome, more than 50% of patients with advanced CRC do not benefit from current treatment modalities,¹⁹ possibly owing in part to inability to predict drug resistance and/or host toxicity.

Although FU metabolism uses defined biologic pathways²¹ and the availability of high through put techniques permit the rapid detection of a large number of genetic and epigenetic variables, the current study by Schwab et al² does not fully address or explore these and other mechanisms potentially implicated in FU toxicity. Thus even though the current study of patients with cancer exposed to FU monotherapy is a large study in contrast to many of the previously reported smaller studies, this study is still far from being informative or comprehensive. Taken collectively, genetic tests proposed for the prediction of patients at risk of developing toxicity to FU remain underdeveloped, with a high percentage of false-negative predictions because of the absence of a comprehensive molecular approach that could account for all elements associated with FU toxicity (genetic, epigenetic, and nongenetic), including impairment of cell signaling pathways and/or DNA damage response, which may significantly influence the cellular response to FU.²²

Four important observations have significantly influenced the development of genetic screening tests that could predict patient toxicity or treatment outcome. The first is the clinical observation that chemotherapy-associated life-threatening toxicity and/or resistance cannot be satisfactorily explained by renal or liver dysfunction, age, sex, diet, lifestyle, or comedication. This has led to molecular studies investigating the genetic differences in drug metabolizing enzymes and drug targets with the conclusion that these could strongly impact treatment outcome.²³ Second, the highly diverse clinical response of histologically similar tumors to the same standard regimen of a chemotherapeutic agent has prompted many investigators to examine the mechanisms of tumor molecular pathogenesis. This in turn has led to the recent use of DNA and RNA microarrays for the molecular study of tumors and for the identification of potential new drug targets and predictors of sensitivity/resistance to treatment. Third is the observation that the development of resistance or toxicity is largely dependent on the overall effects of molecular events in key pathways of drug metabolizing enzymes, cell signaling pathways, and tumor suppressor genes. This has led to the concept that selective targeting of these specific molecular events in the tumor and normal cells could potentially result in increased efficacy and decreased toxicity of the administered drug.²³ Fourth, the recent use of multiple treatment modalities in cancer patients has further complicated the development of a straightforward predictive test.

A prospective comprehensive pharmacogenetic approach would be more suitable for the currently used combination chemotherapy regimens. This could potentially improve treatment outcome by permitting more rational selection of patients likely to benefit from chemotherapy and at the same time identification of those at risk of developing life-threatening toxicity. However, this is still not available for FU or its combination with other agents. Additional well-designed studies are still needed before the use of predictive molecular markers for FU toxicity can be recommended as the standard of care.