Targeted Therapies; Who Detects the Target?

After decades in which anti-cancer drugs mainly were designed to inhibit tumor cell proliferation or to interfere with DNA integrity in order to induce apoptosis in a very general way, we recently have been witnessing the development of a new class of anticancer drugs specifically targeted at biological properties of tumor cells. Examples are Imatinib working against the Kit oncogene and Bcr-Abl tyrosine kinase, respectively, Trastuzumab against Her2Neu, and several drugs including Gefitinib and Erlotinib against EGFR [5,6,11]. In addition, drugs targeted against VEGF, SRC and other molecules have been registered or are in development [4,12]. The concept of targeted anticancer drug therapies is very appealing and preclinical studies in this area have been very promising. Yet, results in some clinical studies have been less exciting than had been expected on the basis of pre-clinical studies [8]. Two factors are relevant in this respect. The first is whether in the tumors of the patients receiving these therapies, the target molecule actually plays a critical role. One approach to determine this is trial and error. Just give the drug to any patient with a tumor type in which some activity of the drug has been reported. This approach has a number of drawbacks. First of all, treating a patient with a drug that has no chance of being effective because it does not match with the biology of that tumor, is a waste of time in which alternative drugs with a higher chance of response could have been administered to the patient. Second, costs of these targeted drugs have a major impact on healthcare budgets and hardly justify a trial and error approach [16]. The other approach is prediction of response to therapy, based on demonstrating certain biological characteristics of the target molecule in a given tumor. This frequently implies immunohistochemical analysis of gene expression in tumor tissue samples. This approach has proven to be of value in the case of c-kit and Her2Neu. Yet, standardization issues remain, both in terms of performing as well as interpreting the stainings. In this respect looking at DNA alterations like amplifications seems to be more robust, and FISH actually now is the gold standard for analyzing Her2Neu amplifications. The second factor is whether in addition to the status of the target gene, other major biological tumor characteristics determine clinical behavior in such a way that the effect of targeted therapy is futile. Diagnostic tests that look at the status of the target gene only – like FISH and immunohistochemistry – may give a false prediction of clinical outcome. Genome wide approaches like expression microarrays or array CGH could be more informative here, but this area is largely unexplored [18]. Also in diagnostic pathology we need to find new ways of anticipating this issue. Most tumor classifications today are based on the organ of origin and on similarity in microscopic phenotype. To some extend, expression of certain proteins measured by immunohistochemistry, is already used for classification purposes, especially in lymphomas. Yet, most of the time these are different proteins than those to which the drugs are pointed. Since drugs work best on tumors to whose biology they match, it seems justified to aim for tumor classifications that are based on prediction of response to (targeted) therapies. While this is a revolution in diagnostic pathology, it also holds a paradigm shift in the practice of randomized clinical trials (RCT). The variables in randomized clinical trials are no longer only (or mainly) the different drug regimens that are compared, but rather the combination of the diagnostic tests and drugs. In fact this increases complexity of the RCT in that not only methodological aspects of drug therapy (dosage, pharmacokinetics, etc.) matter, but also methodological aspects of the diagnostic test. Are we ready for this new situation? Which test should be used in combination with which drug? How do we overcome the issue of other biological tumor properties overruling the response prediction based on assessment of activity of the target molecule in a patient’s tumor. Standardization issues are important here and technical platforms, thresholds, and reproducibil-