Sirs

It is often assumed that cancer incidence continues to increase with age. This was assumed recently by Judith Campisi1, whose figure 1 shows cancer reaching 100% well before the end of a lifetime. Campisi then accepts that the number of senescent cells, which cannot proliferate, increases with age. Campisi has to invent a somewhat tortured explanation of the apparent paradox that senescent cells increase cancer incidence. The recent correspondence on the Campisi paper by Kamb2 seems to accept that such an explanation is required.

However, our analysis of extensive data (a quarter of the United States population) compiled by the National Cancer Institute Surveillance, Epidemiology and End Results (SEER) programme3 shows that cancer incidence peaks at about age 80 (at approximately 2.7% per year for males and 1.5% for females), and then seems to decrease towards 0% (ONLINE FIG. 1). If this is true, there is no paradox and no need for a tortured explanation.

Figure 1: Age-specific incidence of all Surveillance, Epidemiology and End Results cancers.
figure 1

Includes the beta function I(t) = (αt)k-1(1-βt) fit to the data, for males (α = 0.00076, β = 0.0092, k = 5.7) and females (α = 0.0062, β = 0.0092, k = 5.1). Error bars are ±1 standard error of the mean.

Full size image

Historically, there has been some controversy over the quality of the cancer data for the oldest people in the population, but modern cancer registries seem to be accurate, and other investigations in addition to our own work have reported this change in incidence4,5,6. ONLINE FIG. 1 is an analysis of more recent SEER data, which includes cancer incidence rates to older ages than have previously been available. The incidence change seems to be present for all individual cancers recorded by SEER, both male and female. Moreover, we have recently reported a similar change in incidence in a sufficiently large cohort (for data to be statistically significant) of mice that have been allowed to live their full natural lifetimes7.

We have found that the data can be fit by a beta function I(t) = (αt)k-1(1-βt), in which α, β and k are constants and t is age. We derived this equation by adding the factor (1-βt) to the well known Armitage–Doll8 multistage cancer model, and then searched for a possible biological meaning for this new factor. Cellular senescence increasing linearly with age is a plausible interpretation, as senescent cells lose proliferative ability and, so, cannot cause cancer9. This introduction of senescence so that increasing age reduces the pool of proliferating cells is similar to that of Campisi in her figure 4. But the data do not indicate that we have to introduce an ad hoc explanation to explain continuing cancer increases in spite of the reduction in proliferation.

We are studying whether the data still demand the extra factor when we use a mathematically exact multistage cancer model and allow for variations in individual susceptibility.