204 Health Physics August 2010, Volume 99, Number 2 the same for all organs, but separate internal doses were considered to be better for external as compared to stomach. Total site-specific radiation dose estimates accumulated over time varied substantially by geograph- Rongelap and Ailinginae were more precisely estimated than for those on Utrik, and both were judged to be more precise than the estimates of exposures on the midlatitude and southern atolls. Accordingly, the lognormal uncertainty distributions for estimated doses were described by geometric standard deviations (GSD) of 1.2 and 2.0 for external and internal exposures, respectively, on Rongelap and Ailinginae, 1.5 and 2.5 for exposures on Utrik, and 1.8 and 3.0 for exposures on the otheratolls. External and internal dose estimates were assumed to be highly correlated because they both depended strongly on estimated fallout deposition levels. In practice, it made very little difference to the calculation outcomes whether the correlation coefficient was as- estimated for bone marrow, thyroid gland, colon, and ical location (atoll) and year of birth. Fig. 3, drawn from Table 8 in Simon et al. (2010a), illustrates the 10-fold differences in total thyroid absorbed dose between Rongelap and Utrik and between Utrik and Kwajalein (the latter representative of the other mid-latitude islands and atolls), and the two-fold difference between Kwajalein and Majuro (representative of the other southern latitude communities). Fig. 3 also gives some idea of the overwhelming significance of fallout from the 1954 Castle series of tests (see Table 1 of Simon et al. 2010a), with a steep drop of 1-2 orders of magnitude in dose corresponding to birth dates before and after 1954. For risk projection purposes, colon dose estimates were used for organs other than bone marrow, thyroid, and stomach. The exposures associated with any one fallout event were considered to be continuous (and, in general, decreasing over time until 1970, after which they were considered to be negligible), as distinguished from the acute (i.e., near-instantaneous), direct external radiation exposures experienced by persons exposed to the Hiroshima and Nagasaki atomic bombings; the Hiroshima-Nagasaki experience forms the primary basis for current dose-response estimates of radiation-related cancerrisk (Preston et al. 2007; NRC 2006) on which the risk projections presented later in this report are based. Estimated dose varied by atoll, fallout event, calendar year, and age at exposure. Asdiscussed in the companion papers, the precision of the dose reconstruction data was ax Oo p T = o we T T Soro a ot TT Rongelap Island Community I << 2, Mean Dose (mGy) 10° 10° 1930 Utrik Community namin es Kwajalein residents sacminiemiese Majuro residents e “; , —————— 1935 1940 1945 117111 1950 ao Joy 1955 a i 1960 Birth Year Fig. 3. Estimated cumulative thyroid doses for four different communities, by year of birth, drawn from Table 8 in Simonetal. (2010b). Dose estimates for persons born in 1931 also pertain to persons born earlier. internal radiation sources. Moreover, exposures on sumed to be about 0.8 or 1.0, so perfect correlation was assumed for computational convenience. Also, each dose estimate was assumed to represent the mean of its lognormal uncertainty distribution, which implies that the median of that uncertainty distribution therefore equals the point estimate divided by exp[0.5 xX In°(GSD)]. For example, an estimated radiation dose to the thyroid gland, in 1954, of 0.01 Gy from external sources and 0.03 Gy from internal sources at a midlatitude atoll would be treated as the sum of two perfectly correlated lognormal random variables. A Monte Carlo simulation indicates that the uncertainty distribution of this sum is approximately lognormal with mean = 0.0390 Gy, GSD = 2.51, and geometric mean (GM) = 0.02558 Gy.** Estimation of baseline cancer rates In the absence of comprehensive cancer incidence data for the Marshall Islands, approximate tissuespecific, baseline cancer rates were calculated by age and sex, using incidence rates reported by the Surveillance, Epidemiology, and End Results (SEER) registry of the U.S. National CancerInstitute (NCDfor all ethnic groups combined (NCI 1997). These rates were adjusted to reflect the ratio of site-specific, age-standardized (world) rates for ethnic Hawaiians from the Hawaii Tumor Registry (which is a part of the SEER registry) to the corresponding age-standardized (world), or ASW,rates for the SEER registry as a whole. ASW rates are weighted averages of age-specific rates, with the weights determined by the estimated sex-specific age distribution of the entire world population, which is somewhat youngerthan those for most developed countries (Parkin et al. 2002). For example, the 1973-1998 SEER ** Here and elsewhere, results of intermediate calculations are given to greater than two significant digits as needed for subsequent calculations.