Co-developing a targeted therapeutic with a companion diagnostic can be crowned with spectacular success. But how early in development patients should be selected based on such a biomarker is a complex question. The story of Pfizer Inc.'s Xalkori (crizotinib) illustrates several of the issues that need to be considered during the development of such combinations.

Sometimes, to select or not to select is not the question, but a moot point. Among the data presented at the American Society of Clinical Oncology (ASCO) meeting were those showing that Xalkori is a promising agent in some cases of pediatric neuroblastoma. But study lead Yael Mosse of the Children's Hospital of Philadelphia, who presented the data, and her team did not in fact know the EML4-ALK status of the majority of patients that they treated.

"It speaks to the challenges we as pediatric oncologists have," Mosse told reporters at a press briefing. "We are not in a position to recommend biopsy after relapse," at least not for the time being.

Repeat biopsies may become more palatable, though, as clinicians are able to get more out of them. Pfizer Inc.'s Robert Sweetman predicted that at some point, individual tests will be replaced by those that test for whole panels of mutations. That approach is already in place at some academic centers, such as the Dana-Farber Cancer Institute. And last week, diagnostics company Foundation Medicine Inc. launched Foundation One, a test that looks for genetic alterations in more than 200 cancer-related genes.

Such panels will benefit more of the patients who are screened with them, because while any one cancer mutation may be rare, having at least one so-called "actionable" mutation is not.

Even where screening for a mutation is feasible, opinions differ on how early to start with it in the development of targeted drugs.

Later Selection?

At an ASCO panel, Anthony Tolcher of the South Texas Accelerated Research Therapeutics group, which specializes in conducting Phase I oncology trials, argued strongly against biomarker-based selection in Phase I trials.

Tolcher gave several reasons for his point of view. Limiting patients means that trials take longer to accrue. Screening takes time, and once a patient with a mutation is identified, that patient may already be on another Phase I trial. Even if developers have the right idea about how their drugs work, Tolcher said, "you can see that you will make progress go very slowly indeed." And time is of the essence for the patient population that takes part in Phase I oncology trials.

But the bigger problem may come when developers have the wrong idea of how their drug works.

Tolcher used Xalkori as an example of the pitfalls of selecting patients too early during development. Xalkori is now approved for the treatment of lung cancer patients with the EML4-ALK fusion. But originally, its developers thought it would target another oncogene, c-Met. It was during an early all-comers trial that its molecular effect was discovered, and if the original trial design had selected for patients with c-Met mutations, it is possible that its biggest strength would never have been discovered.

Or Early?

At least one of Xalkori's developers, however, disagrees on the moral of its story. Pfizer's Rob Sweetman pointed out that the company did, in fact, use molecular markers as early as Phase I – though not from the very beginning.

"It was a risk," he said. "And many naysayers said, 'it's a small subset and you're not going to be able to reliably detect them.'" But the decision turned out to be the right one.

Going into the clinic with Xalkori, the development team knew that the drug was both a c-Met and an ALK inhibitor. The company originally did a Phase I trial that was not stratified by biomarkers.

But a funny thing happened on the way to Phase II – or rather, two things happened about a dozen patients into the Phase I trial.

First, a paper in Nature described the EML4-ALK rearrangement that is Xalkori's target. And when the team tested two of their Phase I patients that had a significant response to Xalkori, both of them had that rearrangement.

"That was – well, not a red flag. A green flag, perhaps?" Sweetman laughed. "So we decided to turn the ship around in the middle of the course," changing the study protocol to select patients with the fusion gene.

Sweetman acknowledged that testing for biomarkers is more feasible in common cancers. Though EML4-ALK rearrangements drive only a small subset of lung cancers, lung cancer overall is such a big indication that there are enough cases to accrue at a reasonable rate. As Sweetman put it, "A small part of a big number is still a pretty big number."

But if biomarker-based accrual can delay Phase I trials, it can speed up Phase III trials. First, demonstrating success early on makes referring clinicians more motivated to test for the target. And the trials of well-targeted agents can be smaller. "If you are proposing a large difference" between the treatment groups, he said, "then you need a smaller trial."

Overall, Xalkori was approved just four and a half years after the fusion it targets was identified. For melanoma drug Zelboraf (vemurafenib, Roche AG), it took nine years; Tarceva (erlotinib, Astellas Pharma Inc. and Roche AG) was approved 26 years after its EGFR target was identified; and the first targeted therapy, Gleevec (imatinib, Novartis AG), was approved 40 years after the bcr-abl fusion gene was discovered. "Clearly," Sweetman said, "we're getting better at this."