George Sledge, MD, Professor of Medicine at Indiana University School of Medicine and current ASCO President, received the 2010 William L. McGuire Memorial Lecture Award at the San Antonio Breast Cancer Symposium, held December 8-12. In his presentation,¹ Dr. Sledge reflected on the wisdom of the award's namesake (see sidebar) and cast an eye to the future of breast cancer research. The scenario he envisioned would be exciting, were it not so worrisome.

"The standard approach to breast cancer is to look at it as a biologic problem amenable to laboratory investigation, but increasingly it is becoming clear that breast cancer is a math problem," Dr. Sledge maintained.

A decade ago, the human genome was deciphered at a cost of $3 billion. Today, society is at most 3 years away from "the thousand-dollar genome," he said. "Data are becoming so cheap that some worry that our ability to generate genomic data may soon outpace our ability to store it."

In the near future, the patient with cancer will arrive at her oncologist's office with a memory stick loaded with personal medical data. Every patient's cancer will be informative for tumor biology. "This will be incredibly liberating, but things will also get very complicated," he predicted.

The complications will be seen on the individual level-because tumor complexity will become ever-more apparent-and the population level-because clinical trials cannot possibly evaluate the combinations of drugs that will be necessary to target multiple pathways within a given tumor.

EML4-ALK: Example of the Impossible Future

"The recent triumph in the lung cancer world suggests this," he said. The existence of the EML4-ALK fusion gene as the key driver for a subset of non-small cell lung cancers led to "stunning" results for a targeted agent, crizotinib. "More than half the patients receiving four or five prior lines of therapy responded," he noted.

"This is a triumph for targeted therapy. The good news is that crizotinib meets an important unmet medical need, has a straightforward and biologically based biomarker, and produces a high response rate in heavily pretreated patients. Thus, there is a low number needed to treat, and it's also very safe," he noted. "The bad news is that the gene is present in just 5% of patients. Therefore, we must screen 25 individuals to find 1 ALK-positive patient, and not all such patients are trial-eligible, nor would all give informed consent. So we are talking about screening 50 patients to enter 1 on a trial, and this screening requires specialty labs, time, money, certification, and so forth."

Number Needed to Study Is Unfathomable

Taking the drug into the breast cancer setting, he noted that the incidence of ALK gene rearrangement in breast cancer is 2.4%. When multiplied by 40,000 patients with metastatic breast cancer, this totals only 960 patients a year, and when further multiplied by the 3% who enter clinical trials, this results in just 28 patients available for drug development research.

"It gets worse in a hurry," he added. "Most tumors are not as 'stupid' as ALK-positive lung cancer tumors, which occur mainly in nonsmokers with low mutational loads."

Tumors such as gliomas, on the other hand, are "smart" in that they activate multiple kinases. Optimal cell kill requires that all are inhibited simultaneously. This type of tumor is "probably present more often than we would wish," he pointed out.

"My sense of cancer biology is that tumors are sorting themselves out based on the number of drivers of invasion and metastasis they employ," Dr. Sledge said. "Smart" cancers will have multiple drivers that will require multiple inhibitors to derive sufficient clinical benefit.

In breast cancer, the "smartest" may be the triple-negative tumor, which involves DNA damage and repair issues and encourages an accumulation of mutations. Somatic rearrangements are common, as is message dysregulation, thus requiring that therapy target multiple points in the regulatory networks. In contrast, estrogen receptor- and progesterone receptor-positive, HER2-negative tumors have few inter- and intra-chromosomal rearrangements and might be less complicated to treat.

"In the future, what will happen when the next 10 triple-negative patients you see require 8 different combination regimens based on their whole-genome analysis?" he asked. Sufficiently evaluating the variety of treatments for just one tumor type will be a daunting challenge, as the "number needed to study" will be unfeasible, according to Dr. Sledge.

The number needed to study will be a critical concept when large numbers of genes are known to be driving invasion and metastases in a cancer type. At the same time, achieving this will be next to impossible, as calculations will be based on the fraction of patients who are biomarker-positive, accounting for some degree of assay inaccuracy, exclusions for trial ineligibility, and patient refusal. Clinical evaluation of a two-kinase inhibitor combination could theoretically require an "impossible" number needed to study, Dr. Sledge has determined.

This is a far more difficult venture than evaluating, for example, trastuzumab (Herceptin), for which investigators probably needed to screen an estimated 14 patients with metastatic breast cancer for every 1 patient entering the pivotal trial, according to Dr. Sledge's calculations. "Who would screen 154 patients to enter 1 on a clinical trial? This 'thought experiment' suggests there are complications awaiting us in the era of cheap genomics," Dr. Sledge commented. "We will be faced with large numbers needed to treat."

The concept of number needed to study has implications for clinical trials. Assays should be maximized for accuracy, preliminary pathology-based studies should be designed for "quick and dirty" characterization of tumors, exclusion criteria should be minimized, and trials should be made more user-friendly to enhance participation and regulatory approval, Dr. Sledge suggested.

Clinical Trial Implications of 'Genomic Chaos'

"The real problem is not a science problem anymore. It's that our current system is not designed to handle genomic chaos," Dr. Sledge explained. The current clinical trial system emphasizes the development of one new drug at a time and virtually never employs multiple biomarker-driven studies. "Biomarker development and analysis are of secondary importance, and our regulatory apparatus is antithetical to biologic reality," he said.

"We have next-generation gene sequencing. We need a next-generation clinical trials system based on personal genomics, with real-time bioinformatics," he emphasized.

With the huge requirement for trial participants in the targeted combination scenarios, there is a need for an extensive national health information technology system linking drug developers and laboratory scientists, greater collaboration among companies, and more appropriate trial designs, all intended to facilitate the identification of novel combinations. The informed consent process and regulatory apparatus must also be overhauled.

"None of this is easy, but I think all of it is necessary if we are to go down the path toward a cure for breast cancer," he concluded. ■

Reference

1. Sledge GW Jr: What would Bill do? Channeling your inner McGuire.  33rd Annual San Antonio Breast Cancer Symposium. William L. McGuire Memorial Lecture. Presented December 11, 2010.