208 Health Physics separately estimated risks for leukemia, thyroid cancer, stomach cancer, and colon cancer. For the residual group of all solid cancers except thyroid, stomach, and colon, we subtracted the estimates for stomach and colon cancer from the estimate for the combination of all solid cancers excepting thyroid cancer and non-melanoma skin cancer (obtained using weights 0.7 and 0.3 for multiplicative and additive transfer, respectively). For thyroid cancer, the estimated ERR (ERR,<s) obtained above for thyroid cancer is directly applicable to the MI population (i.e., ERR, = ERR,ss). Thus, ERR, has a lognormal uncertainty distribution with GM = 0.0805 and GSD = 3.40 (mean = 0.170) (Fig. 5). Life tables Until this point, the narrative has concerned only estimates of ERR at specific ages. Projected lifetimerisk, however, is calculated as a differentially-weighted sum of age-specific absolute risks in which the weightsreflect the inverse relationship between the likelihood of reaching a given age and the numerical value of that age. For this purpose, we used a 1989-1991 life table for the U.S. (NCHS 1997) (Fig. 6) to adjust for competing, agespecific mortality in estimating cumulative baseline and radiation-related excess risk for exposure to a given radiation dose at a given age. This life table, based on sex- and age-specific mortality rates for the U.S. in 1989-1991, provides one-year survival data for persons alive at any given age during that period, Le., the proportion of persons of a given age that survived until the next year oflife. However,it is often used, as in Fig. 6, to show the average likelihood that a newborn person would survive until one, two, three, etc. years of age 100 a N a \ provided that the age-specific mortality rates observed in 1989-1991 also were to hold for every other calendar year (which, of course, may not be the case). If we multiply age-specific baseline rates by the life table survival probability for that age, and sum over the different ages, the “life-table-weighted sum” is an estimate of lifetime cancer risk, here also assuming that age-specific baseline cancer rates, as well as survival probabilities, do not change over time. With the same assumptions, we can calculate the lifetime excess cancer risk for each year following exposure to a given dose at a given age, as the life-table-weighted sum of estimated age-specific excess risks. Of course, in order to be exposed at a given age one must have survived until that age, so a modified life table is required, conditional on survival to the specified exposure age. As noted in the previous paragraph, life tables are not perfect, but they do provide a standard way of accounting for competing mortality risks when estimating future and lifetime risk associated with a particular exposure of interest. For computational convenience, we used a simple life table for projecting lifetime radiationrelated excess risk rather than a doubly-decrementedlife table which takes account of additional mortality due to radiation-related cancer. This likely resulted in a slight overestimation of excess cancer risks for most of the communities, and probably more so for risks associated with the higher-dose exposures experienced on Rongelap, Ailinginae, and Utrik in 1954. Similarly, in the absence of any MIlife tables, the use of standard U.S. life tables may have led to overestimation of both baseline and radiation-related cancer; however,the effect on relative risk and attributable risk is unlikely to be great considering the greater influences of uncertainty in radiation dose and year-to-year variation in population sizes by atoll. Calculation of projected lifetime baseline and excess risks In the present analysis, lifetime baseline risk is calculated from birth or from 1948, whichever occurred first. Thus, for someone born in 1950, the sex-specific life table survival probabilities in Fig. 6 would be used, - T 1 — 1 | 2 ° Males — — Females a o Percent Surviving L - g0 August 2010, Volume 99, Number 2 but for someone born in 1942, and known (or assumed) to have been alive in 1948, an adjusted life table would 20 - 0 4 pop 0 20 40 60 80 100 120 Age (years) Fig. 6. U.S. male and female life tables used in the preparation of this report. Drawn from data in NCHS (1997). be used, with probabilities of survival to ages one, two, ..., Six, each set equal to one and probabilities of survival until ages seven,eight, etc. (from Fig. 6), each divided by the Fig. 6 probability of survival to age six (i.e., survival until 1948). Denoting the adjusted life table probability to age a by L(a) and the age-specific baseline cancer rate by Byy(a), the lifetime baseline rate from age six is calculated as the life table-weighted sum over ages a