(FORTUNE Magazine) – I WAS GENENTECH'S chief medical officer when we began clinical-trial studies of Avastin in 1997. The hypothesis was that Avastin could stun a tumor--keep it from growing beyond the size of a BB pellet and keep the cancer from spreading. To me, as a former practicing oncologist, that was a beautiful idea. We successfully got through Phase I and Phase II trials with late-stage cancer patients, but then, in September 2002, Phase III results in late-stage breast cancer came back negative.

I remember the day so well. I felt sick. I couldn't eat. You know, your butt is on the line, and I felt accountable. The stock fell 10% that day. We had already spent well over $100 million developing Avastin. Financial analysts are tough customers, and at our winter 2003 investor meeting, they were asking, "Are you going to keep spending good money after bad? Does this failure say something about your ability to pick winners?" That second question made me ask myself whether I had overbelieved our hypothesis about Avastin. I questioned whether I was good at this.

We had to decide whether we should stay the course. We went back and focused on our results in another Avastin study, with colon-cancer patients. Phase II had been successful, and it appeared to be a great study. Art [Levinson, Genentech's CEO] and I agreed: We need to go on. In May 2003 we announced the Phase III results, showing that Avastin not only slowed tumor growth but also helped colon-cancer patients live longer. Genentech stock went up 45% that day. Avastin has turned out to be the most successful oncology product launch in U.S. history--$676 million in sales in its first full year [through February 2005]. It has changed Genentech and changed the way we think about what a negative trial means. Stopping a program is not necessarily the right thing to do, even if the world is criticizing you. Sure, you rely on statistical results, but don't forget to think about them..