In 2008, Dr. Charles Sawyers, currently the president of American Association for Cancer Research, wrote an article for the journal Nature entitled: ‘The Cancer Biomarker Problem.’ This excellent paper clearly explains what cancer biomarkers are, outlines the different categories of biomarkers, and emphasizes how important biomarkers are in the field of targeted therapies. Predictive biomarkers are indispensable tools that should direct the rational use of targeted drugs in cancer patients. There are additional types of biomarkers, including some that could help evaluate the course of cancer progression or help determine the effective dose of an investigational drug. But this post focuses on predictive biomarkers.
Most ‘old’ chemotherapeutic drugs, if successful, kill cancer cells (and some normal cells along the way) based on the common feature of many cancers: their rapid rates of division and growth. The common side effects of many chemotherapeutic drugs—nausea, anemia, and hair loss—are consequences of the drugs killing rapidly dividing normal cells such as intestinal cells, blood cell precursors in bone marrow, or hair follicle cells.
Targeted cancer drugs are selective, that is, each one actually targets a specific alteration or mutation in a cancer-causing gene that is present in a tumor, but not in other cells in the body. These are not blunt and toxic chemotherapeutic drugs, but a targeted drug only works if its target (ie, a certain mutated gene) is present in the cancers it is used against. The success of the first targeted drug, Gleevec (imatinib), was, in part, due to the fact that almost all patients with a chronic myelogenous leukemia (CML) for which Gleevec was developed, had a genetic aberration in their tumor cells known as the Philadelphia chromosome. The consequence of this aberration is abnormal activity of a gene called ABL. Gleevec inhibits activated ABL and, because most CML patients have this alteration in their tumors, they respond to Gleevec remarkably well.
But what if only 15% percent of patients with CML had the Philadelphia chromosome? In this case, if Gleevec trials recruited all patients regardless of their Philadelphia chromosome status, only a small minority (fewer than 15%) would have responded; in addition, it would have taken a long time (if it could be done at all) to reanalyze the results of any trial with Gleevec and connect the outcomes with the presence of the Philadelphia chromosome.
In most cancers, unlike in CML, the situation is different: tumors with the same names might have different genes mutated. For example, fewer than 5% of patients with adenocarcinoma of the lung have an alteration (chromosomal translocation) known as EML4-ALK. The abnormal ALK is successfully targeted with a specific new drug, Xalkori (crizotinib), producing tumor shrinkage. However, Xalkori is of no use to the other more than 95% of lung cancer patients. This is why patients are tested for the ALK abnormality in order to determine whether they qualify for Xalkori treatment.
This method of testing for mutations to guide treatment decisions is becoming an established practice in oncology, where many targeted therapies have so-called ‘companion diagnostics.’ The first test of this nature was developed back in 1998. The diagnostics company Dako launched the HercepTest, an immunohistochemistry assay used to identify patients with HER2-positive metastatic breast cancer, when it was discovered that only patients with HER2 amplification responded better to Genentech’s breast cancer therapy, Herceptin. Since then, a number of companion diagnostic tests were developed and eventually approved for use by the U.S. Food and Drug Administration (FDA). These tests can be found on the FDA website. Moreover, in 2011 the FDA issued a guidance for the development and use of companion diagnostics in 2011.
However, there is a serious problem with many clinical trials that explore new targeted drugs in various cancers. For example, a search of the clinical trials database at clinicaltrials.gov returns 347 active trials examining drugs that target a protein called mTOR in a variety of cancers. Indeed, mTOR is abnormally active in many cancers and it makes sense to explore it as a target. It is also well known that when mTOR is deregulated, a protein named s6 becomes modified in a way that is relatively easy to detect in a tumor biopsy. Yet, of the 347 trials using drugs that inhibit mTOR, only a handful include testing the s6 protein in tumors of patients who enroll in trials.
Should mTOR inhibitors be used in patients whose tumors do not have mTOR activated? The answer is: NO. They should be tested for the s6 biomarker first.
Even worse, the few patients in the trial who have abnormal mTOR activity might respond, but these responses will be ‘lost’ amongst the many cases that have no alterations in mTOR and have not benefited from the drug. Indeed, this has already happened in numerous clinical trials.
Dr. Sawyers wrote 5 years ago: “Despite much discussion on this topic, most clinical trials in patients with solid tumors do not include provisions to obtain additional tissue samples.” Apparently, in the last 5 years there has not been a radical shift in the design of clinical trials to include analysis of predictive biomarkers.
The discovery of tumor biomarkers, and the development and validation of diagnostic tests is not an easy matter. Clinical trials are expensive to run and the addition of diagnostic tests to choose the relevant patient population carries an additional cost burden. Obtaining biopsies for analysis of molecular alterations in tumors is also not always trivial. However, in the end, it is a far better choice than administering a targeted drug to patients who are unlikely to respond because their tumors do not have a mutation that a drug targets. If a patient enrolls in a clinical trial, which is something to be encouraged, she or he should ask: do I have a genetic alteration in my tumor that makes me a strong candidate for a good response? Or, in other words, does this trial involve a companion diagnostic test?