The current generation of rational cancer drugs promises to usher in a new era by targeting distinct genetic features of tumors. But the paradigm shift in treatment needs to be accompanied by a corresponding boost in diagnostics, experts say.

Cancer drugs have traditionally been evaluated on the basis of their effects on tumor size and patient survival. But researchers say those endpoints are too imprecise for testing targeted drugs.

“We need early returns and we need exit polling,” says Todd Golub, associate professor of pediatrics at Harvard University. “If you're using a kinase inhibitor, did you inhibit the kinase? If you didn't, you have no business waiting around to see whether the tumor shrinks.”

By their very nature, most targeted drugs have a low response rate. The drug Iressa, for example, dramatically improves survival in lung cancer—but in just 10% of patients. In April, a year after the drug's approval in the US, two groups independently found that its effect is dependent on a genetic mutation (Science 304, 1497–1500; 2004 and N. Engl. J. Med. 350, 2129–2139; 2004). “The lesson from Iressa is that you have to first determine who is responding,” says Brian Druker, chair of leukemia research at Oregon Health & Sciences University.

The lesson from Iressa is that you have to first determine who is responding. Brian Druker, chair of leukemia research at Oregon Health & Sciences University.

Because the drugs are often developed empirically, researchers may not know enough about the mechanism they target to develop diagnostics from the start. A few tests are beginning to emerge, says Charles Sawyers, professor of medicine at the University of California in Los Angeles, but they are still on the level of “Mom-and-Pop molecular assays” developed in small labs, he says. “That kind of stuff has to become mainstream.”

If querying tumor tissue for molecular markers is to become routine, collection and storage of biopsy samples will also require standardization. Most centers fix biopsy tissue in formaldehyde or embed it in paraffin, which degrades messenger RNA. Freezing tissue, another common storage technique, disrupts proteins. Researchers will need to either find noninvasive ways to take fresh biopsies, or develop assays that do not require tumor tissue, says George Demetri, director of the Center for Sarcoma and Bone Oncology at the Dana Farber Cancer Institute in Boston.

In the absence of standardized tissue-based assays, some researchers have turned to in vivo molecular imaging. Those methods do not reveal a drug's mechanism of action—and so cannot be used for predictive profiling—but they can detect a tumor's response to treatment within days rather than months.

Developing standardized diagnostics for research and clinical tests will take time, particularly as many of the assays are new. “Even now, the [US Food and Drug Administration] looks at our [imaging] data and says, we don't know what that means,” says Demetri. “It can take years to develop that validated data set. You want to be careful that you don't move it forward so fast that you get sloppy.”