Dr. Larry Norton is deputy physician-in-chief for Breast Cancer Programs at Memorial Hospital and medical director of the Evelyn H. Lauder Breast Center at Memorial Sloan-Kettering Cancer Center. He delivered the keynote address at the General Motors Cancer Research Foundation 2005 Annual Scientific Conference at NIH June 13-14. Co-author of the Norton-Simon Model, Dr. Norton's ideas have helped to shape the field of medical oncology.

Where did the Norton-Simon model come from?
The intellectual germ was Dr. Howard Skipper's idea of summarizing how cancers behaved when treated with drugs with a model, rather than using the voluminous dataset itself. I came to NIH at a pivotal time in cancer research, working alongside a legendary Medical Branch team and people such as Dr. Vince deVita, Dr. Paul Carbone, and Dr. Richard Simon, a brilliant biostatistician, who is now chief of the Biometric Research Branch in the Divison of Cancer Treatment and Diagnosis.

In a nutshell, we recognized the importance of what others had already noted: that cancers follow Gompertzian growth curves, meaning that "small things change in size relatively faster." About 1976 we saw this relationship applied not only to unimpeded tumor growth, but also to regression in response to treatment. To my knowledge no clinical exception has been found in the almost 30 years since.

How has this idea influenced cancer research and treatment?
There has been a steady evolution of these ideas as a consequence of clinical and laboratory experiments. We learned that the sequential use of different drugs worked better than alternating therapies, and that "dose density," using effective dose levels over a shorter period of time, both improved results and minimized toxicity, cutting the relapse rate from primary estrogen receptor-negative, node-positive breast cancer by more than half compared with where we were in the mid-1980s. With NCI support, Dr. Joan Massagué and I are now exploring the biochemical reasons for the phenomenon, based on his seminal observations concerning the molecular genetics of metastases. We are hypothesizing that cancers, in fact, seed themselves as well as distant sites, so that a malignant mass is, in effect, a collection of small Gompertzian tumors.

Where does fractal geometry come in?
Fractal geometry quantifies complex shapes that have fractional dimensions, unlike simple forms such as sheets and solids. According to this, small cellular masses are more dense than larger ones; hence, they have a higher fraction of dividing cells and grow relatively faster. The small tumors that make up the conglomerate each have their own growth curve - so the more self-seeding, the greater density, the faster growth rate, and the larger the eventual whole-mass size.

We have traditionally thought that cancers metastasize because they grow large and thereby generate distant-seeding mutants. But perhaps cancers are large because they self-seed, and self-seeding is a step toward distant seeding. By identifying the genes responsible for this process we hope to improve prognostication and therapy.