Personalized medicine (also referred to as individualized treatment) is a model of the way medicine will evolve through the use of specific treatments and therapeutics best suited to an individual's genotype. There has been much discussion of the scientific, social, economic, and ethical considerations of personalized medicine.1,2Cancer therapy presents a unique example of personalized medicine in which an individual's genotype is important in so far as it determines not only the therapeutic and toxic responses to a drug (the pharmacogenotype), but also the genotype of the tumor. Cancer therapy stands today at a confluence of two movements, one toward personalized medicine and the other toward the use of new molecularly targeted cancer drugs that exploit the tumor's molecular signature. This commentary focuses on the unique challenges of achieving the goal of personalized medicine for cancer. This encompasses both the prevention of preneoplastic or early cancer and the therapy of advanced cancer.
To achieve the goal of personalized medicine, it is necessary to develop drugs with defined molecular specificities and to have biomarker and imaging tests, preferably noninvasive to identify the molecular targets in a patient's tumor and the patient's pharmacogenotype. A related and sometimes linked objective of the same tests is to have a way to measure the effect of a drug on its molecular target. Clearly, an important question is how much drug is required to inhibit the molecular target in a tumor and whether there is a benefit to giving more drug. Alternatively, will such maneuvers only increase the chances of toxicity?3It is also desirable to have a test to assess early response so that a nonresponding patient can be spared unnecessary treatment and be moved to alternate therapies. If a noninvasive test is not available, there may be reluctance to take solid tumor biopsies, in part because of cost and the reluctance of Institutional Review Boards to subject patients to invasive procedures to obtain tumor tissue.
The limitations to achieving personalized medicine are our incomplete knowledge of tumor biology and the off-target effects of the drugs. For example, many protein kinase inhibitory anticancer drugs inhibit multiple kinases, which can lead to unexpected therapeutic opportunities.4This can also, of course, lead to unexpected toxicities. These drugs are sometimes referred to euphemistically as pan-inhibitors or, less endearingly, “dirty” drugs. There is also the challenge of identifying off-target toxicities that can limit or even curtail a drug's usefulness. The recent example of the unexpected cardiac toxicity of high doses of cyclooxygenase 2 inhibitors is an example.5
There are more insidious challenges to achieving the goal of personalized medicine stemming from the way cancer drugs are currently developed. The simplest of these challenges is that frequently when a drug first enters clinical trial, there is no validated biomarker or imaging test to select patients most likely to respond or for identifying response to treatment. “Validated” means that the test is standardized and has been shown to be reproducible across institutions, if not in patients then at least in animal models. All too frequently, the development of a biomarker or imaging test has been an afterthought and may become available only after the drug has been in trial for some time. An example is the lack of a validated biomarker for the epidermal growth factor receptor (EGFR) inhibitor gefitinib (Iressa), which was given early approval by the US Food and Drug Administration (FDA) for refractory non-small cell lung cancer but ultimately failed to demonstrate prolonged patient survival.6It is now known that there is a subset of lung cancer patients with an activating mutation of the EGFR kinase that confers sensitivity to gefitinib and other treatments.4It is a moot point whether knowledge of this mutation would have altered the course of development of gefitinib and its ultimate fate. An example in which a test for a biomarker has been of benefit is the use of a HER2/neu test for women with metastatic breast cancer receiving the HER2/neu monoclonal antibody trastuzumab (Herceptin). Currently, two immunohistochemistry test kits and one fluorescent in situ hybridization assay have FDA approval for the selection of patients to receive trastuzumab.
There has been much discussion whether defining subsets of responding patients is counter to the need of pharmaceutical companies to maximize the potential market for a drug. If there is limited confidence in the true molecular target of a drug, or a reliable test is not available, there may be a rational desire to treat all patients irrespective of whether or not they have the target. This does not mean that efforts should not be made to determine whether there is a molecular target-related effect during the clinical trial. Although subsetting patients may potentially decrease market share in a particular tumor type, it may actually increase the number of patients when the target is spread across many cancers. An example is Gleevec (imatinib mesylate), a drug developed for the treatment of chronic myelogenous leukemia with only 4,600 new patients a year. Gleevec is now also used to treat gastrointestinal stromal tumors (GIST) and generates more than $2 billion a year. It is also debatable whether treating large numbers of patients who are likely to fail on a drug is justifiable ethically, even if it is economically justifiable, rather than treating a smaller group that responds and takes the drug for a longer period of time. This is particularly the case when it is realized that the ability to recognize a small subset of responding patients is dramatically decreased by treating the entire population.7The current requirement for disease-specific drug trials means that responding patients must be identified by a separate trial in each disease group. The failure of EGFR mutations to predict drug response other than in non-small cell lung cancer suggests that our ability to predict response based solely on the presence of the target is at a rudimentary stage. A move to target rather than disease-based drug clinical development will require a paradigm shift in the way cancer drugs are developed.
It is clear, particularly with the newer types of cytostatic cancer drugs being developed that arrest the growth but do not cause shrinkage of the tumor, that cancer drugs will have to be given in combination. A rational way to develop molecularly targeted cancer drugs is to use combinations that attack different points in the same or parallel signaling pathways. A frustration for many clinical investigators wishing to move toward the goal of rational therapy is the reluctance of pharmaceutical companies to test their experimental agent in the clinic together with a competitor's experimental agent. A major step to overcoming this problem has been taken by the Cancer Therapy Evaluation Program (CTEP) of the National Cancer Institute as part of the Critical Molecular Pathways initiative that focuses on combining investigational agents in the EGFR and phosphatidylinositol-3-kinase signaling pathways. CTEP guidelines provide intellectual property protection to companies that test drugs in combination, and CTEP provides the framework for conducting these studies.8A number of CTEP-sponsored drug combination studies are under way.
Probably the most insidious bar to the development of personalized medicine is the increasing control of patients by pharmaceutical companies. This is partly for reasons of speed, but it may also be suspected that through the desire to limit access to the complete data set for a trial, companies are dividing trials between multiple institutions. This can mean that an academic institution has too few patients on a particular drug class to make the development of a biomarker or imaging test economically or practically worthwhile. Even when sufficient patients or groups of patients are available from several trials, there is the fear, rightly or wrongly, that if additional studies are added to the clinical trial, this will result in punishment by the company sponsor by withholding future clinical trials. Companies then de facto own access to the patients. This situation is inimical to the development of personalized medicine if the academic clinical researcher cannot have access to the very patients for whom it is necessary to develop these tests. It can be argued that this should be the patient's choice. The record does not suggest that trial sponsors have the desire, knowledge, or resources to carry the entire weight of the development of new selection and monitoring technologies. The difficulties that academic investigators encounter in getting access to patients on sponsored clinical trials could rapidly become the biggest block to the development of personalized medicine.
There is widespread frustration with the cost and time that it takes to bring a new cancer drug to clinical use. The fully capitalized cost for a successful drug can be over $800 million, and it takes 8 to 12 years from the laboratory bench to the pharmacy. Part of the problem is that drugs have to be tested against the standard of care for a disease, and this can change during the course of a trial, requiring new studies. Trials also have to be conducted for each cancer type and for each drug combination, which enormously slows the development process. To encourage the more rapid introduction of drugs into clinical trial, the FDA recently introduced the concept of phase 0 or microdose clinical trials at less than one-one hundredth of the dose calculated to yield a pharmacologic effect, to permit collection of human pharmacokinetic and bioavailability data earlier in the drug development process. These human data can then be combined with preclinical data to select the best candidates to advance to further, more expensive and extensive clinical development. It remains to be seen how popular this approach will be with companies, clinicians, and patients when there is no prospect of therapeutic benefit. It is not, however, much different from the early stages of a phase I clinical trials for patients who receive low doses of an investigational agent with little prospect of therapeutic benefit.
In summary, we are at one of the most exciting times in modern cancer therapy with the development of rational therapy and the move toward personalized treatment for cancer. The concepts to be tested are well established, the drugs are becoming available, and new emerging technologies in target identification, drug discovery, molecular markers, and imaging can make the goal a reality. There has, however, to be a paradigm shift in the way the drugs are developed, requiring a more enlightened relationship among pharmaceutical companies, the federal government, and academia to ensure that we deliver these advances as rapidly and safely as possible to the people who matter the most: patients.
ACKNOWLEDGMENTS
The helpful discussion and insights of my colleagues Dr. Daniel Karp and Dr. Razelle Kurzrock are gratefully acknowledged.