• © 2003 by American Society of Clinical Oncology

Molecular Diagnostics: Assays, Tissues, Progress, and Pitfalls

AFTER OVER 50 years of experience in oncologic therapeutics development, it is clear that the advancement of cancer therapeutics mandates an improved algorithm to replace the current selection of therapy based on histologic appearance and organ of origin of a particular cancer. Cancer is a genetic disease, and it is within the genes and their ultimate expression that lie the keys to the development of successful therapeutics. Myriad predictive and prognostic molecular diagnostic markers have been investigated in patients with colorectal cancer, with the goal of aiding in the selection of the best available therapeutic strategy or prognosticating the probable longevity of a particular patient. Thymidylate synthase (TS), p53, measures of proliferative rate, and the status of mismatch repair (MMR) enzymes have been some of the most heavily investigated parameters. Both an incompetent MMR complex and low levels of TS have been reproducibly associated with improved survival in patients with locally advanced and advanced colorectal cancer.^1–,⁷ TS is the only marker that has been consistently found to predict fluoropyrimidine responsiveness in patients with advanced colorectal cancer.^8–,¹² High levels of TS are associated with poor clinical response to fluoropyrimidine-based therapy in advanced disease patients, a scientifically rational finding given that TS is the target of the fluoropyrimidines and that preclinical evidence has overwhelmingly demonstrated that the major mechanism of fluoropyrimidine resistance is elevated intracellular levels of TS. In patients with locally advanced disease, no marker has been demonstrated to be unequivocally associated with fluoropyrimidine responsiveness. The literature is similarly unclear as to the predictive or prognostic value of p53 status.¹³

In this issue of the Journal, Gonen et al¹⁴ investigated the role of several molecular markers as potential predictors of clinical outcomes in 144 patients with hepatic-confined and resectable-advanced colorectal cancer randomly assigned to either systemic chemotherapy alone or to systemic plus intrahepatic artery chemotherapy following surgical resection. TS, p53, epidermal growth factor receptor, and several clinico-pathologic features were evaluated. Only TS and resection margin were found to be consistently and independently associated with clinical outcome, with TS overexpression being associated with a poor overall survival, progression-free survival, and hepatic progression-free survival. Of interest, patients with TS overexpressing cancers seemed to have a significantly longer survival with combined systemic plus intrahepatic therapy compared with those treated using systemic therapy alone (64 v 21 months).

Although the prognostic value of TS as reported in this work is consistent with previous work, the observed treatment interaction is less consistent with published preclinical and clinical data. In patients with advanced disease, preclinical and clinical evidence predicts that the patients with low levels of TS would enjoy the greatest benefit from fluoropyrimidine-based therapy. A possible explanation for the results of this report may be the presence of a threshold effect, where patients with low levels of TS receive a therapeutic benefit from the fluoropyrimidine levels associated with either systemic or intrahepatic therapy, but patients with high TS levels require the high fluoropyrimidine levels associated only with intrahepatic therapy to realize a benefit.

Several factors indicate that the Gonen et al¹⁴ results must be interpreted cautiously. Therapy by TS interactions in patients treated in the adjuvant setting have been inconsistent in the published literature. Various investigations have identified no interaction, positive interactions, and negative interactions.^5,7,15 These discrepancies may be explained by the fact the levels of TS in the primary tumor may not be reflective of TS levels in the metastatic deposit.^16,17 In studies of patients with advanced disease, where TS is assayed in the metastatic deposit, a setting more analogous to the patients included in the present report, patients with low TS levels treated with fluoropyrimidine-based therapies have consistently demonstrated higher response rates and improved outcomes.^8–,¹² This is in contrast to this report, in which patients with low TS levels seem to have received no benefit from the addition of intrahepatic therapy.

Technical aspects of the assay used also confound interpretation of the results. Although immunohistochemistry has been widely applied to the semiquantitative assessment of TS, conventional interpretation of the assay has focused on the intensity of cytoplasmic staining, which has been validated as an accurate reflection of TS levels in preclinical and clinical investigations.^18,19 The investigators of this report use an unconventional and unvalidated grading scheme for TS, which is focused on the assessment of the proportion of cells demonstrating nuclear staining, rather than evaluating cytoplasmic staining intensity. Multiple prior publications using immunohistochemistry or RT-PCR (reverse transcription-polymerase chain reaction) to evaluate TS levels generally find that approximately 40% to 60% of colorectal cancers will demonstrate high levels.^3–,¹² This is in contrast with this report, where fewer than 15% of patients were classified as having high TS.

A final potential limitation to the strength of conclusions that can be made based on the Gonen et al¹⁴ data is based on statistical considerations. Given the relatively small number of patients who demonstrated TS overexpression, the robustness of the observed interaction is questionable. Specifically, the interaction is based on a subset of only 21 high TS patients who were further subgrouped into 12 patients who received systemic therapy only and nine patients who received systemic plus intrahepatic therapy. Limited sample size is a common problem with marker studies, and it consistently limits the power and certainty of the conclusions. Finally, the lack of consistent findings for the various end points investigated is puzzling. The presence of a significant treatment by TS interaction for overall survival, but not for the more “sensitive” end points of disease-free survival and hepatic disease-free survival, is not easily explained.

Multiple relatively small retrospective series support the value of TS level as a predictor of response to fluoropyrimidine therapy in patients with advanced colorectal cancer. This is a clear step in the right direction, using prospectively collected patients, treated in a standard manner, with a sample size sufficient for limited multivariate analyses. However, the movement of this marker into clinical use will require a prospective study demonstrating that selection of therapy based on TS results in a measurable and clinically significant patient benefit. Although this report adds additional supportive data to the prognostic value of TS, it cannot be used to justify the selection of a particular therapeutic strategy in patients with resected, hepatic-only colorectal cancer.

What steps are necessary to speed the progress of moving prognostic/predictive markers into clinical practice? We indicate several. First, large, simple trials, with reduced eligibility criteria and data collection requirements, would allow larger trials to be feasibly conducted, thus enriching our databases. Second, careful consideration needs to be given to mandating tissue collection on clinical trials, to eliminate the potential bias introduced by subset analyses. Third, the evaluation of markers needs to be considered at the design stage of a trial, and consideration needs to be given to designing trials specifically aimed at marker questions.²⁰

The introduction of additional nonfluoropyrimidine agents with notable activity in colorectal cancer presents an opportunity, and an obligation, to identify and develop molecular diagnostics that will predict with a high degree of certainty that patients will derive the greatest benefit from each particular therapeutic strategy. The virtual explosion of high-throughput array technologies holds promise to rapidly and precisely enable the discovery of these diagnostics. The development and widespread use of such tools, and the accompanying creation of a new cancer taxonomy, will certainly lead to a revolution in cancer therapeutics. However, the foundation on which these discoveries will rest depends on the acquisition and characterization of tissue collections from large, well-characterized, and uniformly treated patient populations, evaluated using validated and standardized assays, subjected to robust statistical analyses.