1. Age-Time Patterns of Radiation-Related Cancer Risk
Donald A. Pierce
Dale L. Preston
RERF Hiroshima
Michael Vaeth
Aarhus University Denmark
INTERPRETERS: I WILL NOT BE SPEAKING VERBATIM FROM ANY PREPARED NOTES. I AM PROVIDING A NOTE PAGE FOR EACH SLIDE, ONLY ROUGHLY INDICATING WHAT I WILL BE SAYING. FOR SLIDES WITH A LARGE NUMBER OF WORDS I WILL BE SORT OF INTERMIXING THE READING OF THE SLIDE AND THE COMMENTS IN THESE NOTES. FOR SLIDES THAT ARE MAINLY GRAPHS I WILL BE SAYING MORE OR LESS WHAT IS IN THE NOTES.
GSF 2004

2. Genesis of Considerations Here
Initially, most thought an single exposure would cause a “wave” of excess cancer, vanishing after 20 years or so
But by about 1985, we found that age-specific cancer rates were elevated for most or all of lifetime
Why should this be? Insufficient attention was given to the implications, for both radiation-related cancer and carcinogenesis in general
We can learn much about radiation-related cancer, and carcinogenesis in general, from the age-time patterns of the excess cancer among A-bomb survivors.
It was initially thought that the A-bomb radiation would “cause” some extra cancers, and that eventually --- after these developed --- the effect would disappear. It turned out that nothing like this happened --- the brief radiation exposure elevated survivors’ cancer rates for all their remaining lifetime.
How could this be? The likely answer tells us important things about radiation-related cancer, and about cancer in general.
GSF 2004

3. 1980’s and Still-Common View of the Relative Risk
(ERR is the % increase in age-specific cancer rate)
The ERR is Excess Relative Risk, meaning as noted the % increase in age-specific cancer rates.
In this slide the ERR is shown as a function of ages during our follow-up, for 4 ages at exposure. The span of each line is the range of ages during our 1950 – 1997 follow-up, which differs with age at exposure.
The ERR here is constant in age, but depends on age at exposure (highest for those exposed as children). But understand that those exposed as children have quite small normal cancer rates during our follow-up since they are not very old, and thus a large % increase at young ages only would not mean a large number of extra cancers.
GSF 2004

4. But the Current Understanding is More as Shown Here
Much of so-called exposure-age effect was
due to variation with attained age
We are now realizing that, in contrast to the previous slide, the ERR actually decreases with age during the follow-up.
One consequence of this is that much of what used to be considered ERR variation with age at exposure, is actually variation with age (compare this slide with the previous one).
More importantly, even though the radiation-related cancer risk does last for all of lifetime, the effect when measured by the ERR is not constant, but decreases with age. The main point of the talk is to explain why it is plausible that:
(a) the effect lasts for all of lifetime, and
(b) but it decreases with age.
I should note that even though the ERR decreases with age, this only means that the excess cancer rates increase with age somewhat more slowly than the natural increase with age. The excess cancer rates, as opposed to the % increase, do increase with age.
GSF 2004

5. Reasons Why ERR Should Look Like This
Considering malignancy of a cell as due to accumulated mutations, suppose as a substantial idealization that:
The spontaneous rate of the next mutation in a cell depends arbitrarily on its mutational status, but not otherwise on age
A brief radiation exposure causes mutations, i. e. momentarily increasing all relevant mutation rates by a factor
A highly idealized stochastic model for radiation and cancer is useful in understanding what was just shown.
A cancer is caused by an accumulation of mutations in a (stem) cell. The rate of such mutations probably depends on the current state of the cell, for example some mutations impair repair mechanisms, causing the mutation rate to increase. As an idealization, suppose that mutation rates do not otherwise depend on age.
Suppose further that a brief radiation exposure momentarily increases all relevant mutation rates, by a factor depending on radiation dose but not on the current state of the cell or on age. This is an idealization, but not totally implausible, since radiation is a fairly general mutagen.
GSF 2004

6. Implication of This Cancer rate following exposure to dose d is
Aside from the idealized assumptions, this describes remarkably well the actual radiation cancer risks for A-bomb survivors
The age increase is about 2-3 days per mSv
The relation in the first bullet means that under the idealized assumptions the effect of an increment of radiation exposure is equivalent to increasing one’s “cancer age” by a dose-dependent increment. This implication is not surprising since assumption B is that exposure momentarily increases all mutation rates by a factor --- so what happens due to exposure would happen in some time period without exposure. However, the age-homogeneity of assumption A is required to derive this result.
The age increment equivalent to a mSv of radiation dose is about 2-3 days.
The remarkable thing is that, aside from the idealized assumptions, this “increase in cancer age” describes the excess risk for A-bomb survivors extremely well. With one minor adaptation, there seems to be no more accurate and comprehensive way to describe the radiation cancer risks.
GSF 2004

7. Simplest Evaluation of the Age-Increment Description
The left panel gives a rough idea of the radiation effect on cancer rates. Looking carefully, one can see that the excess cancer rate at a given dose increases with age, but not as fast as the cancer rates themselves.
The right panel shows that when (cancer) age is measured on the transformed scale age + beta*dose , the radiation effect vanishes and all survivors have the same cancer rates.
This display ignores age at exposure, which is taken up on the next slide.
The age increment removes the most basic evidence
of a radiation effect on cancer rates
GSF 2004

8. Same Results by Exposure Age
Age at exposure is equivalent to birth cohort. The dashed lines here show age-specific cancer rates for unexposed persons, which differ substantially by birth cohort (the young people in the cohort have higher age-specific cancer rates).
On the transformed age scale age + beta*dose , all the cohort members have the same cancer rates, even though these depend on birth cohort.
Effect seen here is birth cohort variation
in background cancer rates
GSF 2004

9. Implications for Relative Risk
The relation involves no assumption regarding some number of required mutations
But for whatever reason, during most of life natural cancer rates take form
Thus the RR could be expected to take form
Now we return to the important age-time pattern shown earlier for the ERR.
We can see, roughly, that due to the manner in which natural cancer rates increase with age, the effect of the “cancer age shift” is larger at young ages than old ages ---- that is, the ERR decreases with age roughly as 1/age .
ERR
GSF 2004

10. Age-Change Result vs Description
Theoretical result has no exposure age effect, an important issue with a variety of explanations
For this plot we have estimated the actual background rate age dependence, rather than taking it as a power of age.
Simplifications regarding birth cohort have been made, and dealing better with this would matter --- in fact, resulting in dependence of the ERR on age at exposure.
But the main point is that the most basic aspects of accumulation of mutations result in age-declining ERR.
GSF 2004

11. Several Refinements Will Explain the Modest Age-at-Exposure Effect
Slight improvement in characterizing birth cohort variations in background rates
Allowing that part of the birth cohort effect acts additively with radiation
Modest increase at young ages of mutation rates per unit time
That the model curve on the previous plot shows no dependence on age at exposure is not really very important. There are various ways that a somewhat less idealized formulation would introduce such effects.
These issues are a little too complicated to go into here.
GSF 2004

12. Effect of Slightly Higher Mutation Rates at Young Ages
Solid: predicted Dashed: observed
Mutation rate variation
I will show, however, that a modest relaxation of the age-homogeneity of background mutation rates introduces an age-at-exposure effect very much like that seen in the data.
Mutation rates for purposes of the development here have been per unit time, whereas they are more often thought of as per cell division. Thus more rapid cell division at young ages would result in what is shown here.
GSF 2004

13. More Immediate Effects
Those results are for well after end of exposure (following latent period between malignant cell and cancer)
Exposure could cause final required mutation and allowing for this we have (but with latent period smoothing)
This added factor can be important
GSF 2004

14. Typical Result for Underground Miners and BEIR VI Models
GSF 2004

15. Implications for Radiation & Cancer
In this way of thinking, radiation does not “induce” cancers, but contributes to the natural process
can explain very simply why naturally why risk persists for lifetime, but ERR decreases with age
The “cancer age increase” interpretation can be useful, for analysis and communication
Suggests substantial commonality between radiation-induced and spontaneous mutations
GSF 2004

16. Implications for Carcinogenesis in General of any Multi-Mutation Model
Accumulation of required mutations probably begins at a very early age
Very large numbers of cells would acquire the first several mutations, and thus this would be quite stable from person-to-person
Most of the person-to-person variation would be due to the long waiting time for the final one or two of required mutations
We can see on several grounds that any given cancer involves relevant mutations that happened at a very early age. Partly this is just ordinary stochastic reasoning, and partly because it is clear that even for exposure at an early age the radiation-induced mutation is seldom the first one.
In simulations, we find that nevertheless carcinogenic processes are very stable from one person to another until sometime in middle age (around 40 years old).
The main person-to-person variation is in the very long random waiting time (usually decades) for the final required mutation. This affects not only the age at which cancer occurs, but also whether one dies to other causes before it can occur.
GSF 2004

17. You Can Find Two Papers on This At
GSF 2004