Page 159 - 20dynamics of cancer

144 CHAPTER 8

relative, the faster the rate of progression. Those who progressed more
quickly appeared to have an inherited polygenic predisposition. Greater
polygenic predisposition was associated with lower age-specific acceler-
ation. I discuss various hypotheses about why such predisposition may
increase incidence and reduce acceleration.

8.1 Comparison between
Genotypes in Human Populations
Comparisons between sporadic and inherited cancers provide power-
ful support for multistage theory. With new genomic techniques, com-
parison of age-specific incidence between human groups with different
genotypes will become increasingly easy to accomplish. So, it is impor-
tant to have a clear sense of what has already been done and what can
be learned in the future.

RETINOBLASTOMA
Bilateral retinoblastoma, in which tumors develop in both eyes, is an
inherited disease. Most unilateral cases occur sporadically. Knudson
(1971) predicted that bilateral cases follow age-specific patterns consis-
tent with one inherited mutation (hit) and the need for only one somatic
hit to produce a tumor. By contrast, Knudson predicted that unilateral
cases require two somatic hits to form a tumor.
Figure 8.1 compares age-specific incidence of bilateral (inherited) and
unilateral (sporadic) cases. The typical measure of age-specific incidence
is the number of cases in an age group divided by the number of persons
at risk in that age group. However, given the small sample sizes and the
difficulty of measuring the base population that represents the number
of persons at risk, Knudson analyzed incidence as the number of cases
not yet diagnosed at a particular age divided by the total number of
cases eventually diagnosed, in other words, the fraction of cases not yet
diagnosed.
Knudson (1971) fit the bilateral cases to the model log(S) =−k 1 t,
where S is the fraction of cases not diagnosed, k 1 is a parameter used
to fit the data, and t is age at diagnosis. He fit the unilateral cases to
2
the model log(S) =−k 2 t , where k 2 is a parameter used to fit the data.
The figure shows a reasonable fit for both models, with k 1 = 1/30 and
k 2 = 4 × 10 −5 .