And When ImClone's Drug Doesn't Work...

Physicians have greeted ImClone Systems Inc.'s (IMCL ) cancer drug, Erbitux, with a mixture of glee and grim realism. The drug dramatically shrinks colon tumors in some patients. But in others, it has little effect -- and no one knows why.

The situation may soon get more confusing. After decades of neglecting colon cancer, drugmakers are suddenly producing a slew of treatments that should hit the market in a few years. Right behind Erbitux is Genentech Inc.'s (DNA ) Avastin, which the Food & Drug Administration could approve by April. Novartis (NVS ) and Amgen Inc. (AMGN ) also have drugs in the pipeline. But what doctors really need is a toolbox of ultraprecise tests that can identify which medicines will work for which patients. It may take several more years for that wish to come true.

The problem boils down to this: Colon cancer is not just one disease. It's likely that there are many subtypes, each carrying its own set of "biomarkers." These might be mutated genes in some tumors, say, or abnormally high levels of certain proteins in the blood. The goal of oncologists is to identify all these signals, analyze how they relate to one another, and then develop tests to identify the various strains of the disease. "We know that patients are not all the same. What we're trying to do now is define the differences," says Dr. David R. Parkinson, head of clinical oncology at Amgen.

Despite its shortcomings, Erbitux is a step in the right direction. On the day it was approved, the FDA also O.K.'d a test that broadly indicates which patients might benefit from the drug. The test works by recognizing a protein called epithelial growth factor receptor (EGFR), which is found in abundance on the surface of many tumor cells. But this test is not perfect. The problem is that 77% of patients with colon cancer produce EGFR. On the basis of the test, they are all eligible for Erbitux -- yet only some of them show a significant response.

Much of the drug industry has been swept up in the hunt for biomarkers. The first place scientists typically look is in tumors themselves. After collecting tissue samples from patients, researchers run a battery of tests seeking abnormalities such as mutated strands of DNA. As the trials progress, they can study whether the abnormality correlates with a higher or lower success rate for a particular drug.

Collecting tumor samples isn't always feasible. That's why pharmaceutical companies are embracing new technologies that can detect proteins after they've been sloughed off from the tumor and have started circulating in the blood. By examining thousands of proteins at a time, "We may find some signature -- a pattern of proteins that define a fingerprint," says Dr. John Reed, CEO of the Burnham Institute, a cancer research center in La Jolla, Calif. That might guide doctors to the most effective treatment.

The search for more precise biomarkers has already produced some successes, including Novartis' Gleevec, which won approval in 2001 to treat a subset of patients with leukemia. A year later, Novartis won approval to market it to patients with a specific type of stomach tumor -- found with a test that searches for a telltale enzyme. In 2003, sales of Gleevec grew 83% to $1 billion. But such triumphs are still rare. "We have to get smarter about how we select drugs," says Dr. Leonard Saltz, a colon cancer specialist at Memorial Sloan-Kettering Cancer Center in New York. The coming flood of new cancer treatments makes that mission all the more urgent.

By Arlene Weintraub in Los Angeles

Before it's here, it's on the Bloomberg Terminal.