Prior to the advent of precision medicine, cancer diagnosis was based on tumor histopathology, and oncology drug development was focused mostly on chemotherapies, generally irrespective of molecular characteristics of the tumor. As molecular testing technologies like genetic sequencing advanced, the importance of molecular profiling of cancers became apparent and drug development evolved. “Researchers realized that two cancers could look the same under the microscope but have different genetic mutations that drove the cancer,” says E. Kelly Sullivan, PhD, Solutions Manager, N-of-One, a QIAGEN Company that provides molecular decision support. “This meant there could be different treatments for the same cancer type because the drugs would be targeting the effects of those different mutations.”
The focus turned toward creating targeted therapies and identifying diagnostic biomarkers to better select patients for whom a drug would be likely to work. And for the past two decades, the paradigm in the field has been “one biomarker, one drug,” leading to the development of many targeted therapies and companion diagnostics. In fact, by 2016, the U.S. Food and Drug Administration (FDA) had approved more than 80 molecularly targeted drugs for various cancer indications, many paired with specific companion diagnostics.
Blockbuster targeted therapies like trastuzumab (Herceptin) have emerged from the one biomarker, one drug paradigm, but as the complexity and heterogeneity of tumor biology has become better understood, the paradigm has become questioned. “Now, there is the thought of going beyond a single biomarker for a single drug,” Sullivan says.
The growing belief within the oncology space is that for some drugs or drug combinations, not one, but several, biomarkers may be needed to adequately identify patients who will respond to treatment. Members of the Office of Biotechnology Products for the Center for Drug Evaluation and Research at the FDA advocated in a 2017 Drug Resistance Updates article to “move away” from the one biomarker, one drug paradigm and develop ways to identify molecular signatures that can predict benefit from sequential or combined therapies.
Due to a lack of comprehensive understanding of the biomarkers that will predict a successful response or resistance in patients, drugs may fail to reach as many appropriate patients as possible. A positive precision diagnostic result may be predictive for some patients but may not predict sensitivity for all the patients who respond to the drug. For example, high PD-L1 expression predicts response to immune checkpoint inhibitors in some patients, but other patients who lack high PD-L1 expression have still responded well to immune checkpoint inhibitors. A requirement for only high PD-L1 expression would deny the drug to patients who could benefit. A recent study by Hellmann et al. demonstrated the benefit of nivolumab plus ipilimumab in non-small cell lung cancer patients with high tumor mutational burden, regardless of the expression level of PD-L1. Now, high tumor mutational burden has emerged as a surrogate biomarker that may independently predict response to these immune checkpoint inhibitors.
Lack of thorough biomarker evaluation for a drug may also result in a drug failing in clinical trials, not from a lack of efficacy, but from a lack of efficacy in a group with the wrong molecular profile. “It’s not that these drugs are necessarily ineffective,” explains Sullivan, “these drugs may not be getting to the right clinical trial populations.”
Getting to the right patients
“Comprehensive biomarker analysis is about narrowing—and expanding—the pool of eligible patients at the same time,” says
Sullivan. She explains that comprehensive biomarker analysis has the potential to unearth multiple molecular profiles for which a drug can work, thus expanding the treatable patient population without diluting efficacy. Such an analysis can also help revive drugs that have failed to show meaningful efficacy in Phase III trials. By going back and identifying biomarkers, companies may precisely select patients that will respond to the therapy and bring the drug back into clinical development.
Although biomarker analysis can revive drugs, the ideal time for biomarker analysis is early in the drug development process, that is, during preclinical or Phase I studies, and continuously throughout the drug life cycle, in order to best define the target patient population. “By the time a drug is entering a Phase III trial, the biomarker or biomarker rules should already be established,” maintains Sullivan. “Additional biomarker analysis at this stage can help gather as many of the gene variants that are predictive of a positive response in the drug in order to refine the molecular criteria on the drug or companion diagnostic label.”
The recent accelerated approval of Janssen Pharmaceuticals’ erdafitinib (Balversa) for metastatic urothelial carcinoma in conjunction with QIAGEN’s Therascreen FGFR RGQ RT-PCR Kit demonstrates how companion diagnostics can include more comprehensive variants to select patients; erdafitinib is indicated for patients with FGFR3 or FGFR2 genetic alterations. Another example of how companion diagnostics are evolving is the FDA approval of larger NGS diagnostic panels. The next step to improving drug targeting is an approval for a combination of biomarkers, or for several different possible biomarkers. “It may be more beneficial to have a combination of biomarkers to predict sensitivity to one drug or to a combination of drugs, especially to a targeted therapy in combination with an immunotherapy,” Sullivan explains. With a combination of biomarkers, for example, one biomarker could predict response for the targeted therapy, and the other could predict response for the immunotherapy.
A companion diagnostic with a combination of biomarkers moves beyond the one biomarker, one drug paradigm, but comprehensive diagnostics may eventually include various platforms in concert. “In order to identify the right combination of biomarkers, the search may need to extend beyond a single molecular testing technology,” Sullivan suggests. The testing could combine DNA sequencing, gene expression, individual protein or chromosomal tests, or surrogate biomarkers.
Biomarker analysis, done well
Analyses using the N-of-One knowledgebase provide support to pharmaceutical companies for the design of combination precision diagnostics. A co-occurring gene variant analysis can help identify genes in which mutations co-occur with mutations in a gene of interest and the cancer types in which they appear, eventually establishing a biomarker combination to be used as molecular criteria for targeted monotherapy or combination therapy trials, or for companion diagnostics.
Take, for example, the tumor suppressor gene ARID1A. Aberrations in ARID1A have been reported across several cancers, but there are no approved targeted therapies for aberrations in ARID1A. To investigate a potential therapeutic strategy for ARID1A, Sullivan and her N-of-One colleagues performed a co-occurring gene variant analysis and detected several altered genes that are significantly associated with altered ARID1A, which include PTEN, PIK3CA, CTNNB1, and BRCA2. “You might say, ARID1A shows up heavily with BRCA2, so a drug might be developed for ARID1A in combination with a PARP inhibitor to target the BRCA2 alterations,” she explains. Identification of populations with significant co-occurrence of mutations in two different pathways could open the possibility of a group of patients that could have a significant response to a combination therapy based on these two biomarkers. Without knowledge of these patient groups, the combination might not be explored.
A critical aspect of comprehensive biomarker analysis is establishing the real-world prevalence of a biomarker or biomarker profile because the profiles associated with a good treatment response in preclinical models or participants in a clinical trial may not accurately represent the profiles seen most often in the real world, where the drug will eventually be used.
“The N-of-One knowledgebase also reflects a microcosm of the testing patterns seen in the real world,” Sullivan points out. Analyses may reveal a type of cancer with a lower prevalence for a mutation, but a higher frequency of cases with the mutation due to testing practices or overall disease incidence. “Perhaps the drug should also be put into clinical trials for patients with that cancer type,” she continues.
The foundation of a reliable comprehensive biomarker analysis is the variant database, but not all databases are complete or current. “There are many reasons to be cautious about using a database,” says Sullivan. Crowd-sourced databases offer convenience, but they are often incomplete and may contain inaccurate information, and databases at commercial laboratories can have limited panels and knowledge of variants. As a result, trials may be designed based on incomplete or incorrect knowledge and curation. To develop a reliable companion diagnostic, it is critical to be certain of variants that are
“The N-of-One knowledgebase is quality controlled, and the variants are classified based on the scientific literature by PhD scientists, rather than by computer algorithms,” informs Sullivan. The knowledgebase represents more than 900 cancer types from more than 100,000 cases and contains over 100,000 variants from over 1800 genes and counting. With N-of-One, research questions are reviewed, data analyzed, and results presented in a flexible format. “When working with N-of-One, you’re not alone,” assures Sullivan. “Somebody’s bringing you through the process and helping you answer your research questions.”