Page 324 - 20dynamics of cancer

15 Conclusions

Molecular technology promises to reveal the biochemical changes of can-
cer. With that promise has also come an implicit assumption: one will
understand cancer by enumerating the major biochemical changes in-
volved in progression and the linkages of biochemical processes into
networks that control cellular birth and death. But enumerating parts
and their connections is not enough.
Think about a large airplane. If you were on that plane, the flight
trajectory is what you would most care about. Could you predict the
flight trajectory if you knew all of the individual control systems and
their complex feedbacks? Probably not, because an inventory by itself
does not provide all of the rates at which changes occur. Even with all
of the rates for component processes, it would not be easy to work out
the trajectory.
One needs to link the parts to the outcome: how do particular changes
in components shift the plane’s trajectory? One ultimately assigns cau-
sality to parts by how changes in the parts affect changes in the outcome.
In a similar way, a genetic or environmental factor causes cancer to the
extent that it shifts the age-incidence curve—the trajectory of cancer. To
understand a particular type of cancer, we must understand the forces
that shape the age-incidence curve and the forces that shift the curve
from its normal pattern.
This book developed a synthesis between, on the one hand, the bio-
chemical processes that control cells and tissues, and, on the other hand,
the consequences for the age-incidence curve of cancer. There have, of
course, been many attempts to connect biochemistry to progression dy-
namics and incidence. Almost all attempts try to fit some model of
process to the observed pattern of incidence. They usually succeed:
most models can be fit to almost any reasonable pattern. The ease with
which different models can be fit to the same data means that one learns
relatively little from fitting.
In this book, I advocated two steps to move beyond facilely fitting
quantitative models of cellular processes to patterns of incidence. First,
breadth of analysis prevents one from uncritically accepting the first