NEW YORK—Larry Norton, MD, is a gardener in the truest sense of the word. He is planting an idea (tumor self-metastasis) that could ultimately grow into a new way to understand and treat cancer. For decades, Dr. Norton, of Memorial Sloan-Kettering Cancer Center, accepted that cancer is a disease of cell division not unlike an acorn giving rise to an oak tree. But these days, his garden is full of weeds, and that has sprouted the theory that each tumor may be a collection of individual seeds that grow wildly in a bed—a product of cell migration and not necessarily unstoppable division.

"Each weed bed is not particularly big or aggressive or fast-growing," Dr. Norton said, but together they can form a large aggressive tumor. "And each weed grows independently, throwing seeds to the wind that may, in turn, end up in the neighbor's garden."

If Dr. Norton is right—and scientists at Memorial and other cancer centers around the world are testing this idea—it could explain why the current crop of medicines that target cell invasion are not curing cancer.

"I have spent most of my career developing therapies that target cell division," said Dr. Norton, who graduated medical school in 1972, a year after Nixon declared war on cancer. "But now I have come to believe that we may have been wrong in focusing our energies on cell division. It may not be a defining feature after all."

The Gompertzian curve

Dr. Norton is saying that biology and genetics may not be enough to explain the complex nature of tumorigenesis. In fact, he's using an old and powerful mathematical formula—the Gompertzian curve—that has been used to explain many events in nature.

He first used this model on the biological level to develop the idea of dose-dense therapy and now is using it on the molecular level to explain how tumors grow and metastasize.

In 1825, Benjamin Gompertz hypothesized that biological growth (such as population growth) follows a characteristic curve of rapid growth followed by slower growth, to an ultimate plateau. Applied to cancer, the Gompertzian curve suggests that malignant cells increase rapidly in number, then settle into slow growth up to a plateau. In this model, tumors can regress with treatment, but then, unless every single cancer cell is destroyed, the cancer cells naturally start increasing again in the rapid cycle mode.