Incorporation and Exploration of Patient Selection Markers in Early Clinical Trials

Cancer is an extraordinarily heterogeneous disease. Even within the same histology, individual patient may respond differently to a given therapy. For MTAs designed to exploit the unique molecular vulnerability in tumor cells, the therapeutic window for a targeted agent or combination may only exist in a small subset of patients. Predictive markers of response or resistance are essential for optimized drug development and patient care, as exemplified in a few FDA approved agents (Table 3.3). Without predictive markers, agents such as Crizotinib and trastuzumab would be likely to fail in the clinical development. Retrospective identification of predictive markers such as EGFR activation mutation and KRAS mutation has also provided significant value for optimizing the treatment choice for patients with NSCLC and colon cancer.

However, identification of predictive markers remains the most challenge task in oncology. To date, some of the most commonly used agents in oncology practice (e.g., chemotherapy, VEGF targeting agents) still do not have validated patient selection markers (Table 3.3). With few exceptions, many investigational MTAs also do not have known predictive markers or strong candidates of predictive markers when entering into clinical evaluation.

Role of Biomarkers in the Different Stages of Drug Development 85

Table 3.3 Patient selection markers for FDA approved cancer drugs



Patient selection markers

Tamoxifen, Al

Breast cancer


Imatinib, dasatinib, nilotinib








HER2 amplification



EGFR activation mutation








Colon cancer

KRAS mutation (exclusion marker)

Head and Neck cancer












RCC, colon cancer, NSCLC, GBM


Sorafenib, Sunitinib, Pazopanib




RCC, Mantle cell lymphoma





Vironastat, romidepsin






The high failure rate of late stage and phase III trials pointed to the importance of biomarker and drug co-development. Efforts in predictive marker discovery, assay development and validation have been increasingly included as objectives of preclinical testing. From the clinical trial perspective, there is also an increasingly recognized need to incorporate predictive marker studies in early clinical trials.

86 Role of Biomarkers in Clinical Development of Cancer Therapies

Trial design for patient selection markers

Depend on the level of qualification for the candidate predictive markers, a variety of trial designs have been proposed and will be discussed in detail in another chapter of this book. In summary, the enrichment design (i.e., only the marker positive patients are enrolled) can be used if the marker-efficacy associated has been clearly defined in previous clinical experience or preclinical models. If there are several candidates of predictive markers or the predictive value of the marker is uncertain, a stratification design (enrolling both marker positive and marker negative patients but in separate marker cohorts) should be used to validate the marker(s) or prioritize among marker candidates. The enrichment design is more efficient for obtaining the proof of principle for the drug but is less robust than the stratification design for validating the marker. Both enrichment and stratification designs require validated assays and prescreening prior to study entry.

For many investigational agents (e.g., agents that target the IGF-1R, VEGF, BCL-2 or immunotherapy), despite extensive preclinical work the potential predictive markers remain elusive at the time of clinical trials, or even long after the FDA approval. Retrospective studies of the tumor or blood specimens would be the only option for the exploration of predictive markers. Sufficient specimen collection has proved to be critically important for the discovery of predictive markers in oncology and should be emphasized in all clinical trials of MTAs. A well-known example of retrospective predictive marker discovery is KRAS mutation as the negative predictor of activity with the anti-EGFR monoclonal antibody therapies.1012 Retrospective studies in patients with remarkable response to Erlotinib also led to the discovery of EGFR activation mutation,12,13 which was subsequently confirmed to be the predictive marker for Erlotinib and gefitinib. In a recently presented phase II trial in NSCLC of Erlotinib with or without MetMab, a monoclonal antibody against C-MET, while no improvement was observed in the overall population, a prospectively planned, retrospectively performed subset analysis revealed that the combination was significantly better than Erlotinib alone in patients with high C-MET IHC in tumors.14

In order for retrospective biomarker analysis to timely inform the drug development, it is important to ensure availability of validated

Role of Biomarkers in the Different Stages of Drug Development 87

assays and timely execution of the correlative studies such that biomarker hypothesis can be generated in time for the confirmatory trial. Ideally, quick turnaround of the biomarker data can streamline the hypothesis generating and hypothesis testing process, and as such, retrospective analysis on "all comers” and prospective enrichment design for the new identified marker candidates can be built in the same trial.

Use and exploration of predictive markers

• Prospective enrichment design

If marker is known (e.g. BRAFvm for vemurafenib)

Histology —»1 Marker +


• Prospective stratification design

If candidates of markers are available but not confirmed

Scientific and technical challenges of predictive markers

With the exceptions of ER, validated predictive markers for all signaling pathway targeting drugs are uniformly based on single gene alterations. Emerging data from cancer genome projects and biological studies reinforce the notion that most solid tumors are driven by more than one genetic changes as well as epigenetic perturbations, and that predictive markers for therapies of most solid tumors may have to rely on multi-analyte, high-dimensional molecular signatures.

An extensively studied and published platform prognostic or predictive marker studies is the gene-expression-based signatures. Well-developed and validated genomic signatures have the promise

88 Role of Biomarkers in Clinical Development of Cancer Therapies

to assist in personalized treatment decisions. An excellent example is the development of OncoType Dx for prognostic prediction of patients with breast cancer. In the context of MTA against key molecular targets and pathways, it is conceivable that gene expression signatures of pathway activations (e.g., MAPK or PI3K) may be potential candidates of predictive markers. However, development and validation of non-DNA based multi-analyte markers are technically and statistically challenging. In an analysis by Subramanian and Simon of 16 studies for gene expression-based prognostic signatures for non-small cell lung cancer,15 there was little evidence that any of the reported gene expression signatures were ready for clinical application. More importantly, there were serious problems in the design and analysis of many of the studies. Guidelines have now been proposed to aid the proper trial design, validated assays and unbiased analysis and validation.16

Another challenge is the longitudinal evolutions of the molecular and clonal profiles through the tumor progression, particularly under the selective pressure of various drug therapies. Therefore tumor tissues obtained from the time of diagnosis, may not reflect the molecular features at the time of new drug clinical trials. Exploration of predictive markers based on these distant tumor specimens can be misleading. However, fresh biopsies for current tumor tissues are costly and not always feasible in patients in advanced stage of cancer.

With the increasing realization of the important of predictive markers for drug development and personalized medicine, there is an enhanced awareness and investment in the biomarker studies, including the search of predictive marker hypotheses in preclinical development of new drugs. The rapidly evolving sequencing technologies have also made it possible to perform genetic characterizations with increasing efficiency and decreasing cost. As exemplified by the BATTLE trials, it is now feasible to screen a patient for potential clinical trials and therapeutic options through multiplexed platforms of genetic characterizations of multiple genes. Currently such "snapshot” approach to patient characterizations can be established for hot-spot mutations in many "actionable” genes. It can be envisioned that in the future, such approach can be extended to characterization ofwhole-exome sequencing or RAN- and proteinbased assays in a real time efficient manner.

Conclusions and Future Directions 89

< Prev   CONTENTS   Source   Next >