Opportunities for Personalized Medicine in Cancer Trials

Delivering the right drug to the right patient is a challenge being addressed by researchers and clinicians in a variety of indications. But the challenge is most acute in oncology, due to the generally narrower therapeutic index of oncology drugs.

The therapeutic advantage of any drug may be lost due to the genetic variability in the metabolism and/or effect of the drug. In oncology, however, researchers must not only deal with somatic polymorphisms but also with polymorphisms, mutations, and genomic expression profiles peculiar to the tumor. Indeed, cancer is characterized by variations in genetic expression, proliferative rates, and survival characteristics of malignant cells compared to normal cells.

To some, personalized medicine is the same as pharmacogenomics. Pharmacogenomics is the science that examines the inherited variations in genes that dictate drug response and explores the ways these variations can be used to predict whether a patient will have a good response to a drug, a bad response or no response at all. However, one could take a broader view and look at phenotypic manifestations of the genetic variations and recognize that these can be adequate surrogates of actual genetic testing. For example, you could directly test for estrogen receptor expression biochemically or you could assay the gene expression for estrogen receptor expression.

Polymorphisms Important in Cancer

In general, personalized medicine takes variations between individuals into account in the design of treatment. We are already aware of polymorphisms in the pharmacokinetics of drugs due to variations in the expression of the enzymes responsible for the absorption, distribution, metabolism, and excretion of drugs and their metabolites. For example, there are numerous polymorphisms involved in the cytochrome P450 enzyme system that can affect the handling of drugs and create the potential for drug interactions. Other examples of enzymes and drugs that have influence in the cancer setting include: UGT1A1 (irinotecan), DPD (fluoropyrimidines), and TPMT variants (thioguanine, mercaptopurine). Enzyme activity can be assessed by analyzing the patient's somatic cells and does not require tumor samples.

Unlike other therapeutic areas however, in oncology this approach can be taken a step further to take into account variations between tumors by analyzing the expression profiles of the malignant cells themselves. In this way, personalized medicine may allow us to further optimize drug effects for both efficacy and toxicity outcomes. However, this often requires sampling of the tumor and sometimes creates problems because the procedure may involve a difficult biopsy or it may involve pain or inconvenience. Doing so in clinical trials may raise ethical concerns, since the results of the testing often do not directly benefit the patient.

Forecasting best treatment

Personalized medicine considerations are often crucial to trials. This is easily illustrated in the case of breast cancer human epidermal growth factor receptor 2 (HER2) status in trials of trastuzumab and lapatinib and in the case of estrogen/progesterone receptors for selective estrogen receptor modulators (SERMS) and aromatase inhibitors. Without the ability to enrich the subjects in these trials, the utility of these agents might still be in question.

We have reached a point in patient treatment assessment where single nucleotide polymorphisms (SNPs) are sometimes being replaced with genomic signatures using multiple gene expression assays. Since therapeutic response and toxicity to most chemotherapy agents are most likely multigenic traits, we cannot expect the detection of a SNP for a single genetic trait to predict efficacy or toxicity reliably under most circumstances. For example, just as intelligence is inherited through multiple genes, you would need to look at a multigene assay to predict intelligence. In cancer, too, you must look at a profile of genes to predict a relapse after definitive surgery.

Therefore, we are seeing a revolution in using multigene assays that are capable of representing the complex interplay among multiple polymorphic genes. This is evident by the now accepted Oncotype DX and MammaPrint assays that are being used to plan treatment in breast cancer.

The Oncotype DX test takes advantage of a reverse transcriptase-polymerase chain reaction assay of 21 prospectively selected genes in paraffin-embedded tumor tissue. The results are capable of determining the likelihood of distant recurrence in tamoxifen-treated, node-negative breast cancer patients. This leads to a recurrence score, which divides patients into low-, intermediate-, and high-risk groups.

The MammaPrint assay has been approved recently by the FDA for assessing the risk of distant metastasis for female breast cancer patients younger than age 61, with Stage I or II disease, tumor size less than or equal to 5.0 cm, and who are lymph node negative.

MammaPrint measures the level of gene expression for 70 genes in a biopsy of surgically resected breast cancer tissue. Unlike Oncotype DX, MammaPrint requires fresh tissue. The assay yields a score that indicates whether the patient is in a low-risk or high-risk category for metastatic recurrence of cancer.

Categorizing the cancer profile

The magic bullet to cure cancer probably does not exist, because cancer is not one disease with one causative mechanism. The same morphologic-appearing tumors may be driven by different pathways.

We now recognize that similar morphologic tumors are heterogeneous and can perhaps be better classified by their drivers. For example, patients who have triple negative breast cancer are now recognized by different expression profiles and are now classified as basal-like breast tumors (usually ER/PR- and HER2-), and there are separate profiles for "luminal A" (usually ER+/or PR+, HER2-), "luminal B" (usually ER+/or PR+, HER2+) and HER positive breast cancers (HER2+, ER/PR-). Although there is still some overlap, such classifications better enable clinicians to prescribe the most appropriate treatment for a patient: personalized medicine.

The proper use of personalized medicine in clinical trials requires the use of validated tools. These tools include biomarkers such as hormone receptors, genomic testing, molecular testing, RNA assays, and proteomic tests. The validation process for these tests is a complicated and rigorous process with which an oncology clinical trialist should be acquainted. Table 1 demonstrates some common polymorphisms that are important in oncology.

Finding the right niche

The potential benefits of using personalized medicine in clinical trials are enormous. Enriching the sample of subjects, for example, to have a greater likelihood of response or omitting patients with a high risk of toxicity is an obvious advantage that enables a trial to be done more safely with fewer patients and/or within a shorter time. This approach may enable a drug to be registered that otherwise would have been destined to fail if the appropriate patient population was not recognized. In addition, it could potentially rescue previously failed drugs if a newly discovered distinguishing polymorphism is recognized. By restricting the trial to a subgroup of optimal patients, time to registration can be shortened and a broader group of patients tested in the postmarketing period. The approach also allows for observation of long-term toxicity, which would not have been possible if short-term toxicity had led to discontinuation.

Recognizing population heterogeneity clarifies circumstances where more than one Phase I dose-finding study should be done. For example, a Phase I trial that blends patients who are slow metabolizers with fast metabolizers does not yield the appropriate dose for either population; whereas, a safe dose could be found for each population by running separate Phase I studies. Similar considerations apply when an SNP, genomic profile or surrogate is shown to correlate with therapeutic effect. For example, the pivotal trials for trastuzumab, which required over-expression of HER2, would have required more patients and shown less efficacy if all metastatic breast cancer patients had entered the trials.

The following are opportunities to use personalized medicine in clinical trials:

• Phase I trials. Perform dose escalation separately on cohorts with each relevant polymorphism, where AUC, Cmax or other relevant PK parameters are likely to correlate with toxicity. In this way, the dose for each population is maximized for efficacy while toxicity is limited.

• Phase II trials. Use this phase as an opportunity to see if enrichment is likely to improve the results of a Phase III trial enough to limit accessibility to this subset of the potential patient population. Other opportunities include:

–Test the hypothesis that a certain polymorphism might result in greater efficacy and/or less toxicity

–Use responders versus nonresponders or tolerant versus intolerant patients as test populations to uncover a genetic signature that might be of value in future trials

–Test a known relevant biomarker or imaging study to see if they are predictive of response

–Save specimens of somatic tissue and tumor tissue for future testing and back testing

–Attempt to salvage a failed trial by using a pharmacogenomic test that was not utilized in the original trial to enrich the population of interest

• Phase III trials. Use data from Phase II trials to develop enrichment criteria for Phase III trials. You can also:

–Confirm hypotheses generated from data mining the Phase II trial by testing it prospectively in Phase III

–Prescreen subjects to enrich the population who have favorable predictive pharmacogenomics and avoid those with unfavorable pharmacogenomics to allow the trial to be smaller, faster, and less expensive.

Conclusion

The future of personalized medicine will likely parallel the perfection of genome-wide mapping techniques. The large cooperative projects completing the comprehensive SNP and haplotype maps of the human and mouse genomes in the near future will likely accelerate the progress already made after the completion of the human genome project.

Ultimately, the ability to predict an individual's reaction to a drug before it is prescribed using personalized medicine will improve therapeutic outcomes and will increase both the physician's confidence in prescribing the drug and the patient's confidence in taking it. This, in turn, should encourage the development of new drugs—tested in a like manner—that will possess enhanced efficacy and safety.

Jeffrey Weisberg,* DO, is senior medical director for the Americas, oncology, i3 Research, Basking Ridge, NJ, email: jeffrey.weisberg@i3research.com . Mark Morrison , MD, PhD, is senior director and section medical director, oncology, the Americas, i3 Research.

*To whom all correspondence should be addressed.

Suggested reading

1. Genomics at FDA http://fda.gov/cder/genomics/default.htm.

2. Table of Valid Genomic Biomarkers in the Context of Approved Drug Labels http://www.fda.gov/cder/genomics/genomic_biomarkers_table.htm.

3. Pharmacogenetics, Pharmacogenomics and Personalized Medicine: Are We There Yet? http://www.asheducationbook.org/cgi/content/full/2006/1/111.

4. P. Lenehan et al., "Rescuing Drugs Through Personalized Medicine," Applied Clinical Trials, April 2005 Supplement, pp. 22–26.

5. C.R. Miller and H.L. McLeod, "Pharmacogenomics of Cancer Chemotherapy-Induced Toxicity," Supportive Oncology, 5 (1) 9–14 (January 2007).

6. The Personalized Medicine Coalition, "The Case For Personalized Medicine," http://www.personalizedmedicinecoalition.org/communications/TheCaseForPersonalizedMedicine_11_13.pdf.

7. R. Simon and S.J. Wang, "Use of Genomic Signatures in Therapeutics Development in Oncology and Other Diseases," Pharmacogenomics Journal, 6 (3) 166–173 (May/June 2006).