Overview

Estimates of radiation casualties in a nuclear war depend on assumptions made about the LD₅₀ value in humans. In the absence of direct evidence, this value has been deduced partly from animal data and partly from a few radiation accidents, many victims of which have been receiving extensive medical treatment. The LD₅₀ value thus deduced was very high, 600-rad bone marrow dose. The largest amount of data for humans—the 1945 inhabitants of Hiroshima and Nagasaki—has been rejected for a variety of reasons.

The recent reassessment of the dosimetry in the Japanese cities for long-term effects has provided an opportunity for the assessment of acute radiation effects as well. A survey carried out on a large number of people in Hiroshima, who were inside their houses during the explosion, contains information about dates of deaths at various distances from the hypocenter. It is suggested in the paper that this survey is highly suitable material for an estimate of radiation casualties under wartime conditions.

A detailed analysis of the mortality as a function of time of death and distance from the hypocenter has been carried out with the aim of proving that, after the first day, the mortality was due predominantly to radiation exposure. The distance at which 50 percent mortality occurred has been deduced from this analysis and found to be 892 ± 11 meters.

To convert this to an LD₅₀ one needs to know the intensity of the radiation field as a function of distance, the transmission factors for Japanese-style houses, and the organ factor. All these quantities have been the subject of detailed studies by a U.S.-Japan Workshop. Although the final values are yet to be agreed upon, it is unlikely that they will differ significantly from those presented so far. Using these data, a probit of mortality versus bone marrow dose was obtained, which showed that the bone marrow LD₅₀ for the Hiroshima survey was only 154 rads (220 rads at body surface). The slope of the line is several times smaller for humans than for animals.

The implications of these findings on the number of radiation casualties in a nuclear war are discussed.

In this paper the basis for calculations of casualties from acute effects of radiation in a nuclear war, which may result in death within 60 days after exposure, is discussed. The time of death after whole body exposure is a function of dose; the general trend of this function, compiled mainly from mammalian data,¹ is shown in Figure 1.

Figure 1

Time of occurrence of death from acute radiation effects. Note that both axes have logarithmic scales (1 Gy = 100 rad).

At very high doses death may occur within hours, but with decreasing dose, the time of death is extended to weeks. Down to a dose of the order of 1,000 rads mortality is 100 percent. At lower doses, where the hemopoietic syndrome is relevant, there is an increasing chance of survival. In this dose range the probability of death is a sigmoid function of the dose that reaches the bone marrow, and is best examined by probit analysis or by using special probability paper (Figure 2) with the dose plotted logarithmically. (This particular curve is the result of experiments with SAS/4 mice, carded out over many years by Lindop and Rotblat.² I will make frequent use of these results when comparing various effects in mice with those in humans.) The sigmoid curve is then transformed into a linear relation, yielding two characteristic values: the LD₅₀ (the dose that causes 50 percent mortality in the population exposed to it) and the slope. The remarkable steepness of the line means that estimates of radiation casualties are very sensitive to the LD₅₀. An error of 30 percent in the LD₅₀ can make all the difference between practically 100 percent survival and practically 100 percent mortality. It will be shown later that for humans the line is less steep, but the LD₅₀ is still the best statistic for an estimate of casualties.

Figure 2

Probability of death as a function of dose, for SAS/4 mice.

The problem is that while there are plenty of such data for animals, there are practically none for humans. Early data from a group of patients with cancer,³ which indicated an LD₅₀ in bone marrow of about 250 rads, were dismissed as not being applicable to the general population. The estimate of the LD₅₀ in humans is based mainly on the very small number of people exposed to radiation in accidents. Most of these victims received intensive medical treatment, including barrier nursing, antibiotics, platelet and red blood cell concentrates, and bone marrow transplants.⁴ Although it is well known that such treatment enables people to survive very high doses, nevertheless, it is being assumed that this does not affect the LD₅₀. In the United Kingdom an effective LD₅₀ of 600 rads to bone marrow—deduced mainly from the people exposed to radiation in accidents—is being used to estimate radiation casualties in a nuclear war.⁵

In Hiroshima and Nagasaki a large number of people were exposed to radiation under wartime conditions, but these data have not been used because of the alleged difficulty in separating mortalities caused by radiation from those caused by blast or heat.⁶ However, recent surveys carried out in Japan in connection with the reassessment of the dosimetry for long-term effects provided an opportunity for another look at the acute effects of radiation. The World Health Organization—which carried out a study of the effects of thermonuclear war⁷—has requested that two Japanese teams undertake such studies. These are still in progress, but the team directed by T. Ohkita has produced data which form the basis for this paper. I should stress that while the data are those of Ohkita and coworkers, they are not responsible for the analysis that I have carded out.

The data come from a survey of people in Hiroshima (to my knowledge no such survey is as yet available for Nagasaki) who were shielded inside Japanese-style houses during the atom bomb explosion. The houses were at distances from the hypocenter that varied from less than 600 meters to 1,300 meters. There were a total of 1,216 people in the houses that were surveyed, of whom 451 died during the first day and 201 (26 percent of those surviving the first day) died during the following 2 months. The tabulated data give the number of people that died each day at various distances, in 100-meter intervals.

My thesis is that the deaths that occurred after the first day were predominantly due to radiation exposure and, therefore, that the data obtained from this survey are suitable for an estimate of the LD₅₀ in humans under conditions of a nuclear war. The evidence for this is based on an analysis of mortality as a function of time and distance, which shows that the observed mortality is in much better accord with radiation exposure than with other causes of death.

First, the time factor will be examined. Figure 3 shows the mortality in 4-day intervals as a function of time after the explosion. It is expressed as the percentage of the total number of people in the survey who died during 2 months, starting from the second day after the explosion. The histogram shows that there was initially an increase in mortality, which—after peaking at about 10 days—gradually decreased. This is not the result that would be expected for deaths from blast injuries or bums. A survey by Masuyama⁸ has shown that after the first day, the cumulative mortality—mostly in people caught in the open—was increasing according to an exponential law, with a half-value of 6 days. From Masuyama's curve, it can be calculated how the percent mortality, in 4-day intervals, would vary with time. As the curve in Figure 4 shows, this variation is quite different from the findings in the survey of the people in houses (histogram). By contrast, closer agreement is obtained with radiation exposures. In the absence of data from humans, data from animal experiments must be used. The histogram in Figure 5 shows the percent mortality observed in mice exposed to a range of doses on both sides of the LD₅₀. Here the time interval is 2 days instead of 4, because in small mammals death occurs over 30 days, instead of over 60 days as in larger mammals. The resemblance of the data to those from the Hiroshima survey (Figure 3) is quite good.

Figure 3

Percent mortality, in 4-day intervals, as a function of time after the Hiroshima explosion, starting from the second day.

Figure 4

Calculated percent mortality, in 4-day intervals, starting from the second day, for all victims of the Hiroshima explosion. The histogram is for the survey group (the same as in Figure 3).

Figure 5

Percent mortality in 2-day intervals, for SAS/4 mice, exposed to a range of doses on both sides of the LD₅₀.

Another way of looking at the time distribution is to calculate the mean survival time of a population exposed to a given dose. As shown in Figure 1, at high doses the time of death depends very much on the dose, but such dependence—albeit smaller—also occurs in the LD ₅₀ region. The lower line in Figure 6 shows the variation of the mean survival time, in days, as a function of dose, for mice. In order to compare the data obtained from mice with those from humans, the dose is expressed as the proportion of the LD₅₀. The upper line shows this dependence for the Hiroshima survey. Taking into account the difference in time of death, as explained above, the similarity between the results is striking.

Figure 6

Mean survival time as a function of dose expressed in terms of the LD₅₀. The upper line is for the Hiroshima survey group. (The doses at the relevant distances were taken from Figure 13.) The lower line is for SAS/4 mice.

Yet another time dependence of interest is the LD₅₀ calculated for a population surviving a given time. In Figure 7 the distance at which 50 percent of the exposed people died is plotted against the day in August 1945 from which the calculation of the mortality was started. For example, the first point (50 percent mortality distance = 1,022 meters) was calculated for all 1,216 people in the survey. For the second point, on August 7, the 50 percent mortality distance of 892 meters was obtained from the 765 people who survived after the first day, and so on. The notable feature of this graph is the very steep drop after the first day, after which the 50 percent mortality distance remains practically steady and then decreases gradually (indicating a gradual increase in the LD₅₀). The shape of the curve after the first day is as would be expected for radiation mortality. Indeed, the top graph, obtained from the data from mice, shows exactly the same behavior: the LD₅₀, calculated for consecutive days, changes little initially and then gradually increases.

Figure 7

Lower curve: Distance from the hypocenter at which a 50 percent mortality occurred in the people from the Hiroshima survey group who were alive on the date shown on the horizontal scale. Upper curve: Similar plot for SAS/4 mice, but with the vertical (more...)

The second evidence for the suitability of the survey data to calculate the LD₅₀ comes from the analysis of mortality versus distance. In a recent paper, Ohkita⁹ presented data (Figure 8) for the whole population in Hiroshima (both in the open and inside houses during the explosion) in terms of the daily mortality rate against distance at various time periods after the explosion. The earlier time periods show a two-component decrease, which Ohkita interprets to be due to the difference between radiation and other fatal casualties. The smaller slope must be due to the latter because the mortality extends beyond the distance at which the gamma rays from the bomb were significant. In Figure 9, line A is a reproduction of Ohkita's data for the period from 7 to 14 days after the bomb. Line B shows the data from the survey group. The notable difference between the two lines is to be expected, if it is assumed that line B gives the mortality predominantly due to radiation and that line A represents deaths from a mixture of radiation and other causes, with the latter being predominant.

Figure 8

Average daily mortality rates for various periods as a function of distance from the hypocenter based on a 1946 survey in Hiroshima (data from T. Ohkita).

Figure 9

Line A: Data from Figure 8 for the interval from 7 to 14 days after the bomb. Line B: The same data for the group in the Hiroshima survey.

A similar but more direct result is obtained by plotting the probit of mortality found in the survey group against the distance from the hypo-center. Figure 10 shows the probit for mortality during day 1, and Figure 11 shows the probit for mortality during the subsequent 2 months. The slope of the latter is 2.2 times greater; therefore, I submit that Figure 11 represents a true regression line for radiation exposure in Hiroshima. The good fit enables the determination, with great accuracy, of the distance from the hypocenter at which there was a 50 percent mortality. This distance is 892 ± 11 meters.

Figure 10

Probability of death as a function of distance from the hypocenter for the people in the survey group who died on the first day. The bars denote —-1 standard deviation.

Figure 11

Probability of death as a function of distance from the hypocenter for people in the Hiroshima survey group who died from day 2 to 2 months after the explosion.

The next step is to convert this distance to dose, and here there is a snag. The necessary parameters for the conversion are: the variation of tissue kerma in air (a measure of the intensity of a radiation field, in rads) with distance; the transmission factor for buildings; and the organ factor, that is, the fraction of the dose that reaches the bone marrow. All these parameters have undergone considerable revision recently in the U.S.-Japan Joint Workshop for the Reassessment of Atomic Bomb Radiation Dosimetry. The last workshop meeting, held in Pasadena, California, in March 1985, was supposed to come up with final figures, but they will not be available until the end of 1986. However, the calculations yet to be made are not likely to bring significant changes. Therefore, I will use the most recent data available. The data by Kerr et al.¹⁰ from Oak Ridge

National Laboratory on tissue kerma are reproduced in Figure 12. It shows the different gamma-ray components, as well as the neutron component. The greatly reduced neutron contribution resulted in a large reduction of the transmission factors for Japanese-style houses.

Figure 12

Kerma versus distance for the various components of the radiation in Hiroshima (dam from Kerr et al.). Note that the doses are in grays (1 Gy = 100 rad).

By applying the appropriate values,¹¹ one can calculate the contribution of the various components to the LD₅₀. As is seen from Table 1, the LD₅₀ turns out to be 154 tads. (In this calculation the relative biological effectiveness of neutrons was assumed to be 1.)

Table 1

Calculation of the LD₅₀ for a distance of 892 meters.

Similar calculations for other distances establish the relationship between dose and distance (Figure 13). It fits excellently a straight line on a logarithmic scale of dose. By using this graph, the regression line can be redrawn to give the probit as a function of dose (Figure 14).

Figure 13

Bone marrow dose versus distance from hypocenter in the Hiroshima survey group.

Figure 14

Percent mortality versus bone marrow dose in the Hiroshima survey group.

Apart from the very low LD₅₀, another interesting feature is the small slope of the line obtained for humans, compared with that obtained for mice (Figure 2). The coefficient of variation, i.e., the ratio of the gradient of the probit line to the LD₅₀ value, is nearly 5 times smaller for humans than for mice. This coefficient depends on several factors, including the homogeneity of the population. A smaller coefficient is to be expected when a highly homogeneous population, like the purebred strain of mice, is compared with a highly heterogeneous population, like humans.

Before the LD₅₀ can be applied to an estimate of radiation casualties in a nuclear war, two more points must be considered. One is that the exposure to radiation in Hiroshima was practically instantaneous, while that from fallout is spread out over hours or days. Since there are no directly relevant data from humans, data from animal experiments must be used. From data presented in the literature¹² it can be inferred that, in larger mammals, if the same dose were delivered at a constant dose rate over 24 hours, the LD₅₀ would be increased by about 40 percent. However, in the case of fallout the dose rate is not constant; it decreases rapidly. Calculations show that for a fallout dose received in 24 hours, the LD₅₀ would be increased by about 10 percent.

The second point is that in fallout calculations, the dose at the surface of the body and not to the bone marrow is usually calculated, as was the LD₅₀ of 154 rads presented above. Therefore, this value must be divided by the organ factor, which probably lies between 0.75 and 0.8. This would give an LD₅₀ at the surface of the body of about 220 rads.

How many radiation casualties would result from such a low LD₅₀? In a recent paper, Lindop et al.¹³ investigated the sensitivity of radiation casualty estimates to the assumed value of the LD₅₀. For a single 1-megaton bomb over London, the number of fatalities was calculated for LD₅₀'s that varied from 300 to 800 tads and for protection factors (the ratios of the doses received in the open to those received inside buildings or in shelters) between 1 and 20. Although these calculations covered a large range of doses, the number of fatalities (N) can be expressed by the following simple empirical formula: N = 4 × 10⁶(PD)^-2/3, where P is the protection factor and D is the LD₅₀. According to this formula, a reduction of the LD₅₀ from 600 tads to 150 tads would increase the number of fatalities by a factor of 2.5. At an average protection factor of 5, this would mean an increase in the number of radiation deaths by more than half a million—just from one bomb.

However, as we pointed out in that paper,¹³ under wartime conditions, even exposure to sublethal doses could give rise to fatalities, because the suppression of the immune system would reduce the chance of recovery from other normally nonlethal injuries; indeed, the interactions may be synergistic. It has been suggested¹³ that any exposure above 100 rads should be considered a radiation injury. This would make the total fatalities, direct and indirect, less dependent on the LD₅₀.

In another paper presented in this volume, Greer and Rifkin listed several conditions that may impair the immune response. Apart from exposure to radiation, they include physical trauma, bums, and malnutrition. This last condition may explain the low LD₅₀ in Hiroshima, since there is evidence that the people in Hiroshima were undernourished both before and after the bomb.¹⁴ By the same token, the other conditions mentioned by Greer and Rifkin—if confirmed—would reduce the LD₅₀ in wartime, even without the malnutrition factor.

In conclusion, it must be stressed that although it is now fairly certain that the LD₅₀ in humans is considerably lower than was thought before, at least under wartime conditions, the actual values, and therefore the estimates of radiation casualties in a nuclear war, are still uncertain. While final calculations must be deferred until the new dosimetry has been firmly established, it is fair to conclude that estimates of radiation casualties previously thought to lie at the upper end of the range have now shifted to the region of probable.

Notes

• ¹ Bond, V. P., T. M. Fliedner, and J. O. Archambeau. 1965. Mammalian Radiation Lethality: A Disturbance in Cellular Kinetics. New York: Academic Press.
• ² Lindop, P. J., and J. Rotblat. 1960. Protection against acute effects of radiation by hypoxia. Nature 185:593-594 (and unpublished data from subsequent experiments). [PubMed: 14417189]
• ³ Lushbaugh, C. C. 1974. Human radiation tolerance. Pp. 475-522 in Space Radiation Biology and Related Topics, C. A. Tobias, editor; and P. Todd, editor. , eds. New York: Academic Press.
• ⁴ Hüibner, K. F., and S. A. Fry. 1980. The Medical Basis for Radiation Accident Preparedness. New York: Elsevier.
• ⁵ Martin, J. H. 1983. Human survival-radiation exposure levels. J. Soc. Radiol. Prot. 3:15-23.
• ⁶ Adams, G. E. 1984. Lethality from acute and protracted radiation exposure in man. Int. J. Rad. Biol. 46:209-217. [PubMed: 6333405]
• ⁷ World Health Organization. 1984. Effects of Nuclear War on Health and Health Services. Geneva: World Health Organization.
• ⁸ Masuyama, M. 1953. Statistical study of human casualties of the atomic bomb, especially of the death rate in the acute stage (quoted by T. Ohkita in Immediate Effects, 1985, Hiroshima ENUWAR Workshop).
• ⁹ Ohkita, T. 1985. Immediate Effects, in Lessons from Hiroshima and Nagasaki. Hiroshima ENUWAR Workshop.
• ¹⁰ Kerr, G. D., J. V. Pace, and W. H. Scott. 1983. Tissue kerma vs. distance relationship for initial nuclear radiation from the atomic bombs Hiroshima and Nagasaki. Pp. 57-103 in U.S.-Japan Joint Workshop for Reassessment of Atomic Bomb Radiation Dosimetry in Hiroshima and Nagasaki. Radiation Effects Research Foundation. February 1983.
• ¹¹ Ellett, W. H., and T. Maruyama. 1983. Shielding and organ dosimetry. Pp. 83-101 in U.S.-Japan Joint Workshop for Reassessment of Atomic Bomb Radiation Dosimetry in Hiroshima and Nagasaki. Radiation Effects Research Foundation. November 1983.
• ¹² Page, N. P. 1968. The effect of dose-protraction on radiation lethality of large animals. Pp. 12.1-12.23 in Proceedings of a Symposium on Dose Rate in Mammalian Radiation Biology. USAEC CONF 680410.
• ¹³ Lindop, P. J., J. Rotblat, and P. Webber. 1985. Radiation casualties in a nuclear war. Nature 313:345-346. [PubMed: 3969144]
• ¹⁴ Committee for the Compilation of Materials on Damage Caused by the Bombs in Hiroshima and Nagasaki. 1981. Hiroshima and Nagasaki: The Physical, Medical and Social Effects of the Atomic Bombing. Hutchinson: London.