Revision 1
March 19, 1998
by Radiation, Science, and Health, Inc.,
Edited by J. Muckerheide

1.2.1
Japanese Atomic Bomb Survivors

1.2.1.1
Cancer/Leukemia

"A recent article by Shimizu, et al. (1992) concerning the effects of low level radiation in atomic bomb survivors concluded that analysis of dose response 'in the less-than-0.5 Sv region fails to indicate the presence of hormesis.' They did not observe any significant decrease in the relative risks (RR) of (a) leukemia, (b) all cancers except leukemia, (c) lung cancer, (d) thyroid cancer, or (e) non-cancer mortality. This conclusion is in agreement with the data shown for the three cancer groups (b,c,d), but appears inconsistent with the data presented for the RR of the leukemia and non-cancer mortality groups.

FIGURE 11.1

FIGURE 11.1 Dose-response analysis of atomic bomb survivors exposed to low-level radiation. The upper pair of relative risks of leukemia and non-cancer mortality show the best fit models of the authors to their data. The lower pair of relative risks shows the best-fit models of the author of this review to their data.

"The upper half of Figure 11.1 shows the data for these two groups as analyzed by the authors with a variety of models. The discussion of leukemia states that though the RR is less than 1 for the three groups with doses less than 0.1 Sv, since all had p>0.10 they did not differ statistically from unity and thus, were within the range of random variation. In clear contradiction to least square fits, the quadratic model for <0.5 Sv was considered to better fit the data than the linear-quadratic model for <0.5 Sv that demonstrated a RR of 0.78 at 0.11 Sv. The lower half of Figure 11.1 shows analysis of the data with models that provide a better fit. The five data points for leukemia are fitted by an empirical polynomial function. The RR for the 0.010 to 0.019, 0.020 to 0.049, and 0.050 to 0.099 Sv dose categories appear consistently related to one another, not varying randomly. The RR of 0.6 plotted at 0.075 Sv is 1.5 SD less than 1 (p<0.15). This study of atomic bomb survivors is in agreement with the decreased leukemia mortality seen in the nuclear shipyard worker study. In both studies the very low incidence of leukemia makes it difficult to obtain sufficient numbers for high statistical power.

"Desired statistical power is present, however, for mortality rates. In the upper half of Figure 11.1 the RR data for non-cancer mortality after low-level radiation are ignored and fitted with a threshold model derived from a prior study of survivors in the <4.0 Sv high-level dose range, assuming the threshold dose is 1.5. Though the mortality RR of 0.83 in the 0.200 to 0.499 Sv dose category is 3.2 SD below 1 (p=0.001) and is the most statistically significant data point of the entire study, nevertheless, this highly significant decreased RR is rejected with the statement, 'The RRs for the sub-groups within the low dose group (<0.5 Sv) when compared with the 0-Sv group did not differ and were close to unity.' If the only mathematical models used for analysis are those that a priori exclude a U-shaped dose-response relationship, it is not surprising that such analysis 'fails to indicate the presence of hormesis.' The lower half of Figure 11.1 fits a linear model down to, but no farther than, the non-cancer mortality RR of 0.83. This decreased mortality risk associated with acute low-level radiation is consistent with the highly significant (-16SD and -8SD) decreased standardized mortality rates observed in prolonged very low level exposures of the nuclear groups of shipyard workers. (Cameron 1992)."