Page 326 - 20dynamics of cancer

CONCLUSIONS 311

ical and cellular components to the quantitative processes that drive
progression and shape incidence curves.
Perhaps we should wait for all the molecular and cellular details, af-
ter which the nature of progression and the final outcome of incidence
may be clear. Unfortunately, enumeration will not work. The full list
of parts for our plane does not tell us how it flies. Measurements of
rate processes by which individual components work locally within the
broader system do not solve the problem. To understand cancer, we
would certainly like to know how a genetic variant of a DNA repair sys-
tem alters the somatic mutation rate. But, based on a compilation of
such rates, we would not be able to build a large, system-level model
that has generality, broad predictive power, and insight into causality.
Induction, ever attractive, does not work.
What does work? Simple comparative hypotheses that reveal causal-
ity and the design principles that determine outcome: the usual itera-
tive scientific cycle between, on the one hand, the genetic and physio-
logical variations in cells and tissues that define the causes and, on the
other hand, the rates at which cancer develops that define the conse-
quences.
Knudson (1971), one of the most cited papers in the history of cancer
research, provides a revealing sensor for current trends. Recent cita-
tions of Knudson’s paper reduce his work to an enumerative slogan and
ignore the powerful way in which Knudson himself analyzed causality
in cancer. Almost all recent citations of Knudson ascribe to him the
“two-hit theory”: for many genes, both alleles must be knocked out to
cause loss of function and progression toward cancer. However, the
two-hit theory was in fact raised several times during the 1960s, before
Knudson’s publication.
Knudson primarily contributed by figuring out a way to test theo-
ries of genetic causation in cancer (see also Ashley 1969a). He com-
pared age-specific incidence curves between inherited and noninherited
cases of retinoblastoma. The inherited cases had increased incidence
by an amount consistent with an advance of progression by one rate-
limiting step. This approach provided a method of analysis by which
one could use quantitative comparison of age-specific incidence between
two groups to infer underlying processes of progression. In this case,
the comparison pointed to a genetic mutation as a key rate-limiting step.