ESTIMATION OF MEDIAN HUMAN LETHAL RADIATION DOSE COMPUTED FROM DATA O N OCCUPANTS OF REINFORCED CONCRETE STRUCTURES IN NAGASAKI, JAPAN Sheldon G. Levin,* Robert W. Young,+ and Robert L. Stohlers neering Physics and Mathematics Division of the Oak Ridge National Laboratory (ORNL) to compute doses at locations inside the Chinzei and Shiroyama school buildings, which were reinforced concrete structures. The third factor was the use of detailed data on the location of occupants of the two reinforced concrete structures in Nagasaki at the time of detonation of the atomic bomb and on their medical condition after irradiation. Two reinforced concrete buildings in Nagasaki were selected for the study because: 1) they were structurally intact after the bombing; 2) they were close enough to ground zero (GZ) to have doses inside the building that spanned the expected lethal dose range; 3) details of occupants’ locations at the time of the bomb, as well as their ages, gender, and health status, were available; and 4) detailed drawings and photos of the buildings were available. The Dikewood Division of Kaman Sciences Corporation of Albuquerque, NM, using detailed location and injury status data from Urabe et al. and Oughterson et al. (1951), used floor plans to locate people inside the two buildings. ORNL then used their TORT code and the drawings to calculate the dose to each person in the building. Finally, based on the injury status data, a study cohort was selected for which the dose-response function (probit) calculations (Finney 197 1) were performed.
Abstract-This paper presents an estimate of the median lethal dose for humans exposed to total-body irradiation and not subsequently treated for radiation sickness. The median lethal dose was estimated from calculated doses to young adults who were inside two reinforced concrete buildings that remained standing in Nagasaki after the atomic detonation. The individuals in this study, none of whom have previously had calculated doses, were identified from a detailed survey done previously. Radiation dose to the bone marrow, which was taken as the critical radiation site, was calculated for each individual by the Engineering Physics and Mathematics Division of the Oak Ridge National Laboratory using a new three-dimensional discrete-ordinates radiation transport code that was developed and validated for this study using the latest site geometry, radiation yield, and spectra data. The study cohort consisted of 75 individuals who either survived >60 d or died between the second and 60th d postirradiation due to radiation injury, without burns or other serious injury. Median lethal dose estimates were calculated using both logarithmic (2.9 Gy) and linear (3.4 Gy) dose scales. Both calculations, which met statistical validity tests, support previous estimates of the median lethal dose based solely on human data, which cluster around 3 Gy. Health Phys. 63(5):522-531; 1992 Key words: atomic bomb; radiation effects; Nagasaki; mortality
INTRODUCTION THISSTUDY was undertaken to provide a new independent estimate of human median lethal radiation dose (LD50) using data not previously available. The reassessment of the prompt radiation field (Roesch 1987) was the first factor in making this estimate. The second factor was the development and use of a new three-dimensional, discrete-ordinates transport code called TORT (Rhoades and Childs 1987) by the Engi-
Three major sources of data have been used to estimate the LD50 for humans: radiation therapy data, radiation accident data, and survey data on the populations of Hiroshima and Nagasaki, Japan, irradiated by the atomic bombs that ended World War 11. Each of these types of data has advantages and drawbacks when used to estimate LD50.
*Technic0 Southwest Inc, Box 847, Los Alamos, NM 87544; Defense Nuclear Agency, 6801 Telegraph Road, Alexandria, VA 223 10-3398; Kaman Sciences Corp/Dikewood Division, 6400 Uptown Blvd. NE, Albuquerque, NM 871 10. (Manuscript received 23 September I99 1 ; revised manuscript received 14 February 1992, accepted 24 February 1992) 00 11-9078/92/$3.00/0 Copyright 0 1992 Health Physics Society
Therapeutic irradiation Patients have been treated with therapeutic, wholebody radiation of 3 or more Gy because they were suffering from fatal diseases. Advantages in using these 522
Lethal radiation and occupants of concrete structures in Japan 0 S. G.LEVINET AL.
data were that: 1) the doses to the individuals were accurately known; 2) they did not have burns or lacerations that might affect survival; and 3) their age, gender, and physical condition were known. However, problems arise in interpreting the data because the patients were ill and were often given pretreatment with drugs or radiation before the whole-body therapy. For example, the Rider-Hasselback ( 1968) cases were given radiation treatment at the specific tumor site before being given whole-body radiation. Many, if not most, patients were treated with fluids, transfusions, and antibiotics after irradiation. These treatments were shown to affect the LD505 (Alpen and Sheline 1954; Perman et al. 1962; MacVittie et al. 1984). Lushbaugh et al. (1967) reported a mid-epigastric LD50 dose of 2.81 Gy using a linear probit and 3.16 Gy using a logarithmic (log) probit. Langham (1967) estimated the LD50 (mid-epigastric dose) at 2.86 Gy using a linear probit and 2.45 Gy using a log probit. These studies were based on therapeutic irradiations using I3'Cs or 6oCoat 1.94 X to 4.13 X C kg-' min-' (0.75 to 1.6 R min-'). Accident cases The radiation accident cases (Lushbaugh 1969)are advantageousbecause the accident victims were healthy prior to the accident; their ages and genders were known, as were the radiation sources and durations of exposure. However, the use of these data to estimate the LD50 is problematic because most accident victims were moving during exposure. Furthermore, interposed objects often resulted in nonuniform doses and partial bone-marrow shielding so that survival may have been increased. To estimate the LD50 for humans, Cronkite (1 982) used the data on the accidental irradiation (0.035 Gy h-I) of Marshall Island inhabitants by fallout during Pacific atomic bomb tests and extrapolated the results using animal research data. Cronkite arrived at an LD50 of 3.50 Gy in air, which corresponds to 2.45 Gy in the bone marrow. An analysis by Mole (1984) combined data from 20 Rider and Hasselback (1968) patients therapeutically irradiated (60Co)with seven subjects selected from the Vinca and Y 12 accidents (Hurst et al. 1959, 1961), who received mixed gamma and neutron radiation, to obtain an LD50 estimate of 4.50 Gy to the bone marrow of humans. Mole (1985) revised the estimates to 4.95.0 Gy in 1985. All the Rider and Hasselback cases were irradiated at about 3 Gy; therefore, the LD50 that Mole obtained was estimated by extrapolation. The slope of the dose/response curve was obtained as a typical value from several small animal studies. Ledney, D.; Steel, L. K.; Elliott, T. B.; Jackson, W. E. Trauma enhanced mortality in irradiated mice. Paper presented at the 35th Annual Meeting of the Radiation Research Society, 2 1-26 February, Atlanta, GA; 1987.
Hiroshima and Nagasaki victims Many attempts have been made to estimate human radiation lethality from data based on the Hiroshima and Nagasaki atomic bomb casualties (Oughterson et al. 1951; Masuyama 1952; Shirabe 1953; Oughterson and Warren 1956; Ohkita 1975; Hayakawa et al. 1986; Fujita et al. 1989). Surveys to estimate mortality in these populations have the advantage of large numbers of individuals on which to base estimates of LD50 and permit the use of average transmission factors for cases where shielding data are not available. However, surveys have the disadvantage of relying on the memories of survivors to establish their own position, shielding, and injuries as well as for nearby persons who were killed. Because no information is available on either the nonsurvivors, who were alone, or on entire groups that perished, it is likely that a smaller proportion of deaths is attributed to a given dose than was the actual population response. All mortality studies based on the Japanese experience also suffer from the fact that the entire at-risk population can never be known. Furthermore, the higher the dose, the fewer survivors there were, hence the data in a dose-response curve must necessarily be heavily weighted on the low-mortality side. Thus, any mortality estimates at high doses must be based on extrapolations from the lower-dose data available to the surveys. The earliest and possibly most extensive studies of the Hiroshima and Nagasaki victims were carried out shortly after the war under the direction of Col. A. W. Oughterson, who headed The Joint Commission for the Investigation of the Effects of the Atomic Bomb in Japan. The Joint Commission consisted of American and Japanese physicians and scientists who produced a six-volume description of the effects (Oughterson et al. 1951). The major results were later summarized in a book by Oughterson and Warren (1956). The Joint Commission conducted sample surveys of about 7,000 individuals in Hiroshima and Nagasaki that were used to estimate injury and mortality vs. distance for the populations of Hiroshima and Nagasaki. This first effort at predicting radiation lethality did not attempt to estimate doses to individuals; instead, individuals were placed in 500 m-wide bands radiating out from their best estimate of the location of GZ. They did not separate the shielded from the unshielded nonsurvivors, they did not know time-to-death, they could not distinguish between those who died of radiation and those who died from radiation combined with other injuries, and they did not break down mortality by age and gender. The Joint Commission survey (Oughterson and Warren 1956) yielded estimates of 1,290 m for the 50% lethal distance for both cities. Using the current dose vs. distance curves'' (Roesch 1987)leads to bone marrow LD50 doses of 0.46 Gy for Hiroshima and 0.83 Gy for 'I Kaul, D. C. An assessment of dosimetry system 1986 (DS86) components. Paper presented at the 23rd Annual Meeting of the NCRP. Washington, DC, 1987.
Nagasaki. These values are too low to be considered valid estimates of the true human LD50. Two other early surveys by Masuyama (1952) and Shirabe (1953) were not random samples, nor were they extensive or detailed enough to provide valid estimates of the population mortality or injury. A survey of 3,2 15 Hiroshima people who had been between 600-1300 m from the hypocenter was conducted by Hayakawa et al. (1986) between 1969 and 1975. Hayakawa placed individuals who were in wood frame houses in 1OO-m-wide bands radiating outward from the hypocenter. Using the Hayakawa data on distance vs. mortality, excluding first-day mortality, the authors used the linear probit method to estimate the 50% lethal distance to be 904 m. Using the Kaul(l987) data that also appears in Roesch (1987), this translates to 3.0 Gy free-in-air inside wood-frame houses or 2.2 Gy to the bone marrow. If first-day mortality were included, the 50% lethal distance would be 1,032 m, which translates to an LD50 of 1.40 Gy, a decrease of almost 40%. This finding is consistent with the generally accepted view (Seigneur and Brennan 1966; Fanger and Lushbaugh 1967) that first-day mortality is due to causes other than radiation; hence, its inclusion would lower the LD50. Another factor highly likely to have caused lowering of the LD50 in the Hayakawa data is the inclusion of persons who had burns or blast injuries in addition to radiation. Several studies (Shirabe 1953; Oughterson and Warren 1956; Ohkita 1975) that used survey data in examining mortality in Hiroshima and Nagasaki showed that many, if not most, of the individuals in the 600- 1300 m range (from GZ) had more than one type of injury. Laboratory studies (Brooks et al. 1952; Alpen and Sheline 1954; Baxter et al. 1954) clearly showed that the LD50 in mice that were subject to both burns and radiation is lower than it is with mice subject to radiation alone. Ledney et al. (1987) recently showed that the LD50 for mice with combined wounding and radiation is lower than it is in mice with radiation alone. Therefore, we are led to conclude that mortality estimates based on the Hayakawa data tend to underestimate the true human LD50. Using the same Hayakawa survey data, Rotblat (1986) estimated the 50% lethal distance as 892 m. Using the dose vs. distance and the house-shielding data available in 1984, he estimated the 60-d median lethal dose (LD50/60 d) for humans as 1.54 Gy to the bone marrow. If the most recent transmission and weaponyield data (Kaul 1987) were applied to the 892 m, an LD50 of 2.35 Gy to the bone marrow would be obtained. Because Rotblat used the Hayakawa survey data that included many radiation victims with burns or blast injuries, Rotblat's estimated LD5O is most likely lower than the true LD50 for humans. The most recent use of Japanese atomic bomb survivors surveys to estimate the LD50 for humans is by Fujita et al. (1989) of the Radiation Effect Research Foundation. To arrive at estimates of the LD50, Fujita et al. used the new Roesch ( 1987) dose information and
November 1992, Volume 63, Number 5
the most complete location and mortality data from 20,000 RERF files. Their data were restricted to people who were inside wood-frame dwellings at the time of the Nagasaki bomb. They could not differentiate between victims who died with combined injuries and those with radiation only. Furthermore, they included people who died within 24 h, which markedly lowered the LD50, as shown by the Hayakawa data. Fujita et al. arrived at estimates of the LD50 that ranged from 2.10-2.54 Gy, depending on the method of estimation. Their estimate based on the linear probit was 2.16 Gy to the bone marrow. MATERIALS AND METHODS
Data sources The data on location, injury, and mortality of individuals who were in the Chinzei and Shiroyama schools at the time of the nuclear detonation have been obtained from several sources. The most useful data were contained in two reports by Urabe et al."* of the Imperial University, Tokyo. The location of each occupant of the building is shown on drawings and the name, gender, age, injury type, and some notes on survival are given. The data for first-day survivors are reproduced in Tables 1 and 2. Other reports and detailed building drawings by the U.S. Strategic Bombing Studies**++(U.S. SBS 1947) and by the Joint Commission (Oughterson et al. 1951) also showed the locations of the occupants and were used to cross-check the Urabe5I1data. Selection of cohort Three cases were deleted prior to selection of the cohort used for this study. After cross-checking all sources, doses could not be calculated for these cases because of disparity between the two sources of location data. It was necessary to restrict the cohort because the purpose of this study is to estimate an LD50 for healthy people due only to irradiation. Many people (Table 3) were killed instantly or died within the first day and could not be included in the study cohort because it is generally accepted that no one would have died within 24 h due to radiation alone in the dose range in these Urabe, M.; Obashi, S.; Ueda, H.; Hakamada, S.; Mikaido, S. Report of the inspection of casualties in the buildings of the Chinzei Middle School in the city of Nagasaki. Tokyo, Japan: Medical School of Imperial University; Report N-49; 1945a. Located in: Urabe, M. Collection of the reports on the investigation of the atomic bomb casualties. Tokyo: Science Council of Japan; Science Promotion Society; 1953: 497-504. 'Urabe, M.; Obashi, S.; Ueda, H.; Hakamada, S.; Mikaido, S. Report of the inspection of casualties in the buildings of the Shiroyama School in the city of Nagasaki. Tokyo, Japan: Medical School of Imperial University; Report N-50; 1945b. Located in: Urabe. M. Collection of the reports on the investigation of the atomic bomb casualties. Tokyo: Science Council of Japan; Science Promotion Society; 1953: 486-496. ** Gruel, A. E. Unpublished building plans. U S . Strategic bombing survey: Chinzei Middle School-Bldg I , Group 18, August 1946. tt Gruel, A. E. Unpublished building plans. U.S. Strategic bombing survey: Shiroyama School-Bldg 2, Group 16, August 1946.
Lethal radiation and occupants of concrete structures in Japan 0 S. G. LEVINET AL.
Table 1. Occwants of Chinzei school who survived at least 1 d after the atomic bomb detonation. Dose (Gy) Age
Died (D) or survived(S)
1 2 3 4 5 First floor
19 17 23 19 18
M M F M F
0.22 0.24 3.52 1.84 3.26
0.17 0.18 2.63 1.36 2.42
S S S S D
>60 >60 >60 >60 28
Minor cuts Minor cuts None Face wound None
11 12 13 16 17 92 101 102 103 104 112 115
22 23 26 18 18 15 19 19 19 19 15 15
F F M F F M M M M M M M
18.38 6.94 4.09 10.79 19.28 4.54 1.89 2.16 2.69 3.88 2.46 2.63
13.25 4.86 2.84 7.65 14.11 3.19 1.30 1.48 1.82 2.61 1.68 1.84
D D S D D S S D S D S D
? ? >60 45 ? >60 >60 31 >60 7 >60 ?
None None Minor cuts None None Leg fracture Leg fracture Head wound Head, shoulder wounds None Head wound None
27 18 19 65 35 44 41 48 63 20 47
M M M M M M M M M M M
16.30 7.83 12.53 4.06 3.27 6.17 3.89 6.31 3.81 5.36 5.39
11.73 5.66 9.16 2.86 2.36 4.35 2.78 4.31 2.66 3.81 3.84
D D D S D D S S S D S
14 23 9 >60 ? 31 >60 >60 >60 26 >60
None Slight burn None Cut on head Minor contusion None Scratches Minor cuts Slight cuts Cuts-Hips and back Minor head wounds
26 30 38 24 29 24 27 23 24
F M M M M M M M M
7.51 4.68 12.36 28.30 28.07 35.82 10.61 13.52 12.62
5.29 3.27 8.90 20.46 20.32 26.05 7.80 9.88 9.06
D D D D D D D D D
3 25 17 27 3 4 13 2 1 -
Second Jloor 26 40 43 88 89 90 93 94 95 96 97 Deletions: 6 18 27 30 32 37 46 49 54
Reason: Burn Burn Burn Burn Burn Burn Burn Burn Burn
Burns-35% of body Burns-back, face, arms Burns-back, face, hands Burns-face, chest, hands Burns-50% of body Burns-50% of body Burns-face, hands, feet Entire body burned Burned
” ? = unknown.
buildings. The Urabe”I reports are clear about those who died within 24 h of the bombing, and they invariably gave the cause of death as “blast” or “crushed.” Fires in both buildings were started by the heat from the bomb. The fires did not trap people nor prevent them from leaving the buildings, but those who were unable to remove themselves promptly were burned, some severely. Tables 1 and 2, from Urabe summarize all of the injury and mortality information available on those people who survived at least 24 h and who were in the Chinzei and Shiroyama schools when the atomic bomb detonated over Nagasaki. Of those who survived at least 1 d, nine occupants of the Chinzei school and eight of the Shiroyama school had second-degree burns over 35% or more of the body and were deleted from the study cohort. Four of the nine significant burn injuries in the Chinzei school may have been flash burns because of the victims’ proximity to windows. There were no flash burns in the Shiroyama
school because the building was end-on to GZ and the end facing GZ had no windows. Five persons in the Shiroyama school who had serious large wounds were also deIeted from the cohort. The burn and blast cases were deleted because their chances of survival would have been markedly lowered due to the effects of combined injuries. The cohort that was finally selected for inclusion in the dose/mortality calculation is shown in Table 4. The study cohort is composed of 57 people who survived 24 h after the bomb but not longer than 60 d, plus 18 people who survived more than 60 d. The age and gender of that cohort is presented in Table 5. Estimates of doses to occupants of Chinzei and Shiroyama schools The doses used in this paper were calculated by Rhoades et al. ( 1992). They developed and validated a three-dimensional discrete-ordinates transport code
November 1992, Volume 63, Number 5
Table 2. Occupants of Shiroyama school who survived at least 1 d after the atomic bomb detonation. Dose (Gy) Case number
Died (D) or [email protected]
Second floor 12 18 19 36 58 62 69 75 76 95
24 24 22 20 17 21 20 23 17 17
F F F F F F F F F F
8.73 5.39 17.29 5.41 17.56 5.43 13.18 6.29 10.57 6.84
6.18 3.82 12.37 3.83 12.60 3.86 9.57 4.50 7.5 1 4.83
D D D D D D D D S S
23 7 6 7 ?
17 ? ? >60 >60
98 103 104 105 111 117 Thirdfloor 10 17 21 23 27 30
17 16 17 18 16 19
F F F F F F
7.95 5.68 8.69 5.82 6.69 10.56
5.61 4.02 6.09 4.14 4.73 7.56
D D D D
31 >60 >60 26 14 ?
23 21 23 23 22 22
M F F F F F
25.92 18.22 25.02 24.57 16.66 24.92
18.90 13.21 18.21 17.88 12.04 18.20
D D D D D D
11 7 7 7 7 7
31 32 38 41 42 43 44 46 47 48
22 20 21 23 21 20 20 21 19 21
F F F F F F F F F F
24.80 15.52 17.64 13.29 17.04 15.26 17.56 14.00 14.80 16.09
17.93 11.31 12.66 9.5 1 12.39 11.08 12.71 10.12 10.73 11.74
D D D D D D D D D D
51 52 53 54 56 65 86 88 90 91 94
23 20 20 23 18 20 16 16 19 16 18 18 15 19
F F F F F F F F F F M
17.81 19.46 17.04 22.40 18.92 18.02 15.81 16.59 12.91 13.0 1 19.08 16.88 18.64 8.78 18.77
12.91 14.22 12.27 16.25 13.82 13.08 11.55 1 1.97 9.29 9.35 13.85 12.16 13.47 6.36 13.66
D D D D D D D D D D D D D D D
F F M M
(Rhoades and Childs 1987; Rhoades and Childs 1988) in order to calculate doses inside the reinforced concrete buildings. Then, using the exterior fluences described in Roesch ( 1987), they calculated free-in-air kerma and doses to the bone marrow for each individual in the Chinzei and Shiroyama school buildings in Nagasaki, Japan. The uncertainty of estimates of doses to the bone marrow of the Chinzei and Shiroyama schools was calculated to be 30%, which included 20% source
10 10 10 10 10 7 5
10 7 ? 7 14 9 10 ?
10 3 16 14 5 10
Injuries Slight face wound None None None None Face wound None None Minor cuts Lacerations, head and back None Minor cut None None None None None Scalp wound None None None Scalp and shoulder wounds Leg wound None None None Leg fracture Slight scalp wound Scalp wound None None Scalp contusion, fractured hand None None Leg fracture None Scalp laceration None None None None None None None None None None
uncertainty. Details of the dose calculations and uncertainty analysis appear in a companion paper in this journal (Rhoades et al. 1992). Dose/mortality calculations A dose-response function was calculated for percent mortality vs. dose to individuals. It was also calculated for the logarithm (base 10) of dose to individuals, using the probit (Finney 1971) method. Probit
Lethal radiation and occupants of concrete structures in Japan 0 S. G. LEVINET AL.
Table 2. Continued Dose (Gy) Case number
Died (D) or survived (S)
39 78 80 81 82 87 109 118 40
Burn Burn Burn Burn Burn Burn Burn Burn Blast
24 17 17 17 17 16 18 20 25
F F F F F F F M F
10.32 18.22 8.38 16.20 6.71 22.32 15.24 17.04 16.14
7.28 13.18 5.90 11.62 4.74 16.1 I 11.07 12.33 11.58
D D D D D D D D D
3 10 52+ 52+ 17 10 ? 1 3
lniuries Severe burns Extensive burns Severe burns Severe burns Extensive burns Extensive burns Extensive burns Extensive burns Ruptured abdomen Severe scalp wound Severe scalp laceration Severe back wound Glass wounds to total body
" ? = unknown.
Table 3. Postirradiation medical condition of occuvants of Chinzei and Shirovama schools. First-day mortality
2-60 d mortality
Alive >60 d
Total at risk
Totals do not include three cases deleted due to lack of location data.
Table 4. Survivor-mortalitybreakdown of radiation-only cohort used to estimate LD50. Survive >60 d
Mortality 2-60 d
Bone marrow dose (Gy)
0.15-1 1.26 3.61-15.53 -
(Finney 1971) is used to describe a graph of data that has dose or log dose on the x-axis and the Gaussian probability integral transformation on the y-axis. If the distribution of sensitivities of the individuals to radiation is Gaussian, then the data will form a straight line. Probit also refers to a type of nonlinear regression wherein the percent mortality is transformed to equivalent Gaussian deviates and a type of straight line is fitted to these deviates vs. dose or log dose. The resulting line is called the probit line. The linear probit line and the mortality data that have been grouped into intervals are presented in Fig. 1. The same mortality data points are presented on a log scale with the log probit line in Fig. 2. Probits also have been calculated using individual
doses to the bone marrow for the radiation-only cohort, and for that cohort supplemented by the burn cases. The doses for the five blast injury cases were very high (>13 Gy) and their inclusion did not change the LD50. Finally, probits were calculated using free-in-air kerma in place of doses to the bone marrow. All calculations used individual rather than grouped data and were done in original dose units and also using logarithms of dose to the base 10. The LD50, the 95% confidence interval for the LD50, the probit slope, and the chi-square values are presented in Table 6. Chi-square values were calculated using individual ungrouped observations and tested for significance. Chi-square was not significant at the p < 0.05 level for any dose-response parameters tested, indicating that the probit-fitted lines were a reasonable representation of the data in all cases. Therefore, the choice of whether to use the linear model or the log model must be based on other considerations. DISCUSSION Comparison of estimates The LD50 of 2.9 Gy to the bone marrow estimated by log probit in this study is surprisingly consistent with
November 1992, Volume 63, Number 5
Table 5. Age-gender distribution of radiation-only cohort. Age (Y)
21-25 3 1
Age (Y) 15
Combined Chinzei and Shiroyama School
99 1 I
.....I......... 1.....I .........
& 70 Q
L .........L ....-.I
Bone marrow dose (Gy) Fig. 1. Linear plot of dose-response data and probit line for
estimates made by other researchers, particularly considering the difference in methods used to arrive at their estimates. Langham (1967) estimated the LD50 at 2.86 Gy using the linear probit and 2.45 Gy using the log probit, based on total-body 6oCotherapeutically irradiated patients. Lushbaugh et al.3 (1967) estimate of 2.86 Gy was based on patients irradiated at a very low dose rate, with no irradiations greater than 3 Gy. Bond and Robertson (1957) estimated the LD50 at 2.65-2.70 Gy based on a combination of animal studies and the Marshall Island fallout and mortality data. Rotblat (1986), using the linear probit, estimated the LD50 at
Bone marrow dose (Gy) Fig. 2. Logarithmic plot of dose-response data and probit line
for radiation-only cohort.
1.54 Gy based on people in wood-frame dwellings in Hiroshima at the time of the bomb detonation. This estimate would be raised to 2.35-2.6 Gy if the newer (Kaul 1987) Nagasaki weapon yield and shielding factors were used.
Effect of burn and blast injuries on LD50 The 17 cases with both radiation and burn injuries are insufficient to calculate a median lethal dose represented by LDSO(R + B). However, if those 17 cases are added to the 75 radiation-only injuries ( R ) , the
Lethal radiation and occupants of concrete structures in Japan 0 S. G. LEVINET AL.
Table 6. Calculated probit parameters for Chinzei-Shiroyama data. 95% Confidence Interval
Degrees of LDlO LD90 x 2 freedom (GY) (GY)
probit model can predict deaths (mortality) at zero dose, whereas percent mortality must approach zero as dose gets very small in the log model.
and then calculate LD5O(R + B ) = 2.4 Gy. The estimated LD5O(R + B ) , for radiation plus burns, of 2.4 Gy to the bone marrow is 20% lower than the LD50 for radiation only. This result is consistent with the animal studies (Brooks et al. 1952; Alpen and Sheline 1954; Baxter et al. 1954; Ledney et al. 1987) that have shown a lower LD50 for combined injury. The confirmation found in this study is an important result and is analytically addressed here for the first time using human-combined injury data. Occupants of the buildings who received severe blast injuries also received very high radiation doses and died. For this reason, they do not affect the LD50 and have not been incorporated.
Effect of dose uncertainty on slope In the probit model, the slope of the probit line is the reciprocal of the standard deviation (S.D.);therefore, a low slope implies a large S.D. The true population S.D. is a measure of the variability in the sensitivity of individuals t o radiation. T h e calculated S.D. includes the true variability of the population (i.e., sensitivity to radiation) as well as the variability due to errors in dose estimates. Thus, the calculated S.D. increases as variability attributable to dose increases, although the true population S.D. is fixed. The calculated S.D. would also be increased if there were a mixture of more than one type of injury in the population, resulting in a bimodal frequency distribution with some deaths attributed to radiation and other deaths caused by flying debris. Morris and Jones (1989b) showed the effect of dosimetry errors on the LD90/LD10 and derived an adjustment for a special case. They also showed that the LD50 obtained from the log probit is least affected by dosimetry errors and, although their adjustment does not apply to our case, their LD50 conclusion argues for the use of the logarithmic transformation. The ratio of LD90:LD 10 for the log probit, using doses to the bone marrow, is 6.2 calculated from values in Table 6. The use of the ratio LD90:LDlO provides a standardized way of examining points on the distribution where “hardly any” and “almost all” of the subjects are affected. The ratio is a unitless measure that allows comparison of the variability of groups with a low LD50 to groups with a high LD50. For most animals other than humans, the LD90:LD 10 falls in the range of 1.53 (Jones 1981; Baverstock et al. 1985; Jones et al. 1986; Morris and Jones 1988, 1989a; Scott et al. 1988); however, it must be noted that in laboratory animal studies, the radiation doses are known to within 2-5%. Furthermore, the animals are most often a highly select, inbred strain housed under tightly controlled laboratory conditions which further reduces variability and increases slope, hence reducing the LD90:LD 10. The effects of dose uncertainty, the heterogeneity of people, and the possibility of combined injury cases tend to account for the differences in slopes and the LD90:LD 10 ratio obtained in this study.
Linear vs. log probit There is no clear choice between the use of the linear and the logarithmic probit model for the dose/ mortality data in this study on the basis of the chisquare goodness-of-fit criteria. The authors prefer the log model because the 10%and 90% mortality values that it predicts are more consistent with other studies and because its LD50 is least affected by the uncertainty in the dose measurements. Furthermore, the linear
Future efforts The data that were used in this study were verified using several sources of information. The calculation of doses to individuals used a state-of-the-art computing method, and the building dimensions and construction were cross-checked from several sources. We believe that it would be difficult to improve the LD50 estimates based on the information included in this study. However, the possibility of expanding this effort to include
LD50 Lower Upper (GY) (GY) (GY) Slope
Bone marrow Linear Logarithmic Linear (+ burn) Log(+ burn) Free-in-air Linear Logarithmic a
3.4 2.9 3.1
1.5 1.8 1.3
4.8 0.315“43 3.9 3.20’ 49 4.3 0.341” 57
73 73 90
-0.66 1.2 -0.62
7.5 7.4 6.9
3.6 3.37’ 56
6.7 0.225’44 5.5 3.22’ 49
-0.89 10.5 1.7 10.3
Units are Gaussian deviates per Gray. Units are Gaussian deviates per log Gray.
LD50 ( R ) = 2.9 Gy (for radiation only) is lowered to 2.8 Gy for the mixture LD5O(R,B). Finney’s (1971) mixture equation can then be used to estimate the LD5O(R + B ) with the quantities at hand: R,B/Total - RITotal LD5O(R,B) - LDSO(R)
(R + B)/Total LD5O(R + B )
Substitute calculated bone marrow doses of LD5O(R) = 2.9 Gy and LDSO(R,B) = 2.8 Gy and the proportions RITotal = 75/92, ( R + B)/Total = 17/92, and (R,B)/ Total = 92/92 = 1 in the previous equality statement to obtain: 1 - 75/92
17/92 LD5O(R + B )
additional buildings in Nagasaki and Hiroshima is currently being investigated.
November 1992, Volume 63, Number 5 been authored by a contractor and subcontractor of the U.S. Government. Accordingly, the U.S. Government retains a nonexclusive, royalty-free license to publish or reproduce the published form of this contribution, or allow others to do so, for U.S. Government purposes.
SUMMARY This study analyzes dose and mortality data on people in the Chinzei and Shiroyama buildings in Nagasaki at the time the nuclear bomb was detonated. Estimates of the LD50 for humans made in this study are in reasonable agreement with values proposed by several prior investigators using other means. This analysis uses doses to the occupants computed by ORNL using a computer program that modeled radiation transmission through the reinforced concrete walls, ceiling, and windows of the two buildings in great detail. Data on each subject were carefully reviewed and excluded: a) if subject location was uncertain; b) if the subject was moderately to severely burned or injured by the blast so that the injury was a possible factor in survival; or c) if the subject died within 24 h of the nuclear detonation. After these exclusions, there remained 75 subjects that were included in the cohort used in the dose/mortality calculation. The LD50 estimate, based on a log probit calculation, was 2.9 Gy to the bone marrow with a 95% confidence interval of 1.8-3.9 Gy. Inclusion of the 17 cases having burn injuries lowers the bone marrow LD50 to 2.8 Gy from which we estimated LD50 for combined radiation and burns to all cases as 2.4 Gy. The population from which these data were drawn consisted primarily of healthy young Japanese persons on a wartime diet. The deaths in the study cohort were attributed to ionizing radiation delivered within 7 s. The radiation was mostly gamma but there was a 35% neutron component outside the buildings so that the doses due to neutrons and gammas were added and their sum was called “dose.” This study has a number of advantages not available in previous efforts to estimate the LD50 for humans: the doses received ranged from nonlethal to very lethal; the entire at-risk group was known (including age, gender, type of injury, and approximate time of death). It was possible to delete subjects who died the first day, presumably from causes other than radiation, and it was possible to separate subjects who had serious burn and blast injuries, which are known to affect mortality. Furthermore, these data are not clouded by the complication of pre- or postirradiation treatment. One problem this study has in common with all other Japanese atomic bomb studies is the uncertainty of dose. However, the doses received by people in this study have been calculated for each individual with greater accuracy than previously possible. Furthermore, the radiation-only cohort spans the nonlethal to verylethal range, and the LD50 estimated using a log probit is relatively insensitive to random dose errors.
Note-Research was sponsored by the Defense Nuclear Agency under RDT&E RMC Code B46620. The submitted manuscript has
REFERENCES Alpen, E. L.; Sheline, G. E. The combined effects of thermal burns and whole body x-irradiation on survival time and mortality. Ann. Surg. 140:113-118; 1954. Baverstock, K. F.; Papworth, D. G.; Townsend, K. M. S. Man’s sensitivity to bone marrow failure during whole body exposure to low LET ionizing radiation: Inferences to be drawn from animal experiments. Int. J. Radiat. Biol. 47:397-411; 1985. Baxter, H.; Drummond, J. A.; Stephens-Newsham, L. G.; Randall, R. G. Reduction of mortality in swine from combined total body radiation and thermal burns by streptomycin. Ann. Surg. 137:450-458; 1954. Bond, V. P.; Robertson, J. S. Vertebrate radiobiology (lethal actions and associated effects). Ann. Rev. Nucl. Sci. 7: 135162; 1957. Brooks, J. W.; Evans, E. I.; Ham, W. T.; Reid, R. D. The influence of external body radiation on mortality from thermal burns. Ann. Surg. 136533-539; 1952. Cronkite, E. P. The impact of estimates of human radiation tolerance upon radiation emergency management. In: Control of exposure of the public to ionizing radiation in the event of accident or attack. Bethesda, MD: National Council on Radiation Protection and Measurements; 1982: 21-27. U.S. Strategic Bombing Survey. Effects of the atomic bomb on Nagasaki, Japan; Volume 11. Washington, DC: U.S. Government Printing Office, Physical Damage Division; June 1947. Fanger, H.; Lushbaugh, C. C. Radiation death from cardiovascular shock following a criticality accident. Arch. Pathol. 83~446-460;1967. Finney, D. J. Probit analysis. Cambridge, MA: Cambridge University Press; 1971. Fujita, S.; Kato, H.; Schull, W. J. The LD50 associated with exposure to the atomic bombing of Hiroshima and Nagasaki. J. Radiat. Res. (Tokyo) 30:359-381; 1989. Hayakawa, N.; Munaka, M.; Kurihara, M.; Ohkita, T. The analysis of mortality rates of survivors exposed within Japanese wooden houses in Hiroshima by exposed distance. J. Hiroshima Med. Assoc. 39:126-129; 1986 (in Japanese). Hurst, G. S.; Richie, R. H.; Emerson, L. C. Accidental radiation excursion at the Oak Ridge Y12 plant, 111 determination of radiation doses. Health Phys. 2: 127-133; 1959. Hurst, G. S.; Richie, R. H.; Sanders, F. W.; Reinhardt, T. W.; Auxier, J. A.; Wagner, E. G.; Callihan, A. D.; Morgan, K. Z. Dosimetry investigation of the Jugoslav accident. Health Phys. 5 : 179-202; 1961. Jones, T. D. Hematologic syndrome in man modeled from mammalian lethality. Health Phys. 41:83-103; 1981. Jones, T. D.; Morris, M. D.; Wells, S. M.; Young, R. W. Animal mortality resulting from uniform exposures to photon radiations: Calculated LD50s and a compilation of experimental data. Oak Ridge, T N US. Department of Energy; ORNL-6338; 1986. Langham, W. H., ed. Radiobiological factors in manned space flight: Report of the space radiation study of the life sciences committee. Washington, DC: National Academy of Science, NRC; Publication 1487; 1967.
Lethal radiation and occupants of concrete structures in Japan 0 S. G. LEVINET AL.
Lushbaugh, C. C. Reflections on some recent progress in human radiation. In: Augenstein, L. G.; Mason, R.; Zelle, M., eds. Advances in radiobiology. New York Academic Press; Vol. 7; 1969. Lushbaugh, C. C.; Comas, F.; Hofstra, R. Clinical studies of radiation effects in man: A preliminary report of a retrospective search for dose response relationships in the prodromal syndrome. Radiat. Res. Suppl. 7:398-412; 1967. MacVittie, T. J.; Monroy, R. L.; Patchen, M. L.; Darden, J. H. Acute lethality and radiosensitivity of the canine hematopoietic system to cobalt-60 gamma and mixed neutron irradiation. In: Broerse, J. J.; Macvittie, T. G., eds. Response of different species to total body irradiation. Dordrecht, Netherlands: Martinez Nijhoff Publishers; 1984 113-129. Masuyama, M. Stochastic studies on the atomic bomb casualties. Bull. Math. Stat. 5:21-30; 1952. Mole, R. H. The LD50 for uniform low LET irradiation of man. Brit. J. Radiol. 57:355-369; 1984. Mole, R. H. The LD50 for uniform low LET irradiation of man, a postscript. Brit. J. Radiol. 58:98-99; 1985. Moms, M. D.; Jones, T. D. A comparison of dose-response models for death from hematological depression in different species. Int. J. Radiat. Biol. 53:439-456; 1988. Moms, M. D.; Jones, T. D. Hematopoietic death of unprotected man from photon irradiations: Statistical modeling from animal experiments. Int. J. Radiat. Biol. 55:445-46 1; 1989a. Morris, M. D, Jones, T. D. Some effects of radiation dosimetry errors on an estimated dose-response relationship. Health Phys. 56:219-222; 1989b. Ohkita, T. Acute effects. In: Review of thirty years study of Hiroshima and Nagasaki atomic bomb survivors. J. Radiat. Res. (Chiba) 16 (suppl.):44-66; 1975. Oughterson, A. W., Leroy, G. V.; Leibow, A. A.; Hammond, C.; Barnett, H. L.; Rosenbaum, J. D.; Schneider, B. A. Medical effects of atomic bombs. Oak Ridge, TN: U.S. AEC, Ofice of Technical Information, Technical Information Service; Rpt NP 304 1; Vol. 6; 1951.
Oughterson, A. W.; Warren, S. Medical effects of the atom bomb in Japan. New York: McGraw-Hill; 1956. Perman, V.; Cronkite, E. P.; Bond, V. P.; Sorenson, D. K. Regenerative ability of hematopoietic tissue following lethal x-irradiation in dogs. Blood 19:724-730; 1962. Rhoades, W. A.; Childs, R. L. The TORT three dimensional discrete ordinates neutron/photon transport code. Oak Ridge, TN: ORNL-6268; 1987. Rhoades, W. A.; Childs, R. L.; Ingersoll, D. T. Radiation exposure inside reinforced concrete buildings at Nagasaki. Oak Ridge, T N ORNL/TM-10569; 1988. Rhoades, W. A.; Childs, R. L.; Ingersoll, D. T. Radiation exposure inside reinforced concrete buildings at Nagasaki. Health Phys. 63510-521; 1992. Rider, W. D.; Hasselback, R. The symptomatic and hematological disturbance following total body irradiation of 300rad gamma-ray irradiation. Guidelines to Radiological Health. Environmental Health Series, Radiological Health. Washington, D C U.S. Department of Health Education and Welfare; Public Health Service Publication No. 999RH-3; 1968: 138-144. Roesch, W. C., ed. US-Japan joint reassessment of atomic bomb radiation dosimetry in Hiroshima and Nagasakifinal report. Vol. 1. Hiroshima, Japan: Radiation Effects Research Foundation; 1987. Rotblat, J. Acute mortality in nuclear war. In: Soloman, F.; Marston, R. Q., eds. The medical implications of nuclear war. Washington, DC: Institute of Medicine, National Academy of Sciences; 1986. Scott, B. R.; Hahn, F. F.; McClellan, R. 0.;Seiler, F. A. Risk estimators for radiation-induced bone marrow syndrome lethality in humans. Risk Analysis 8:393-400; 1988. Seigneur, L. J.; Brennan, J. T. Incapacitation in the monkey (Macaca mulatta) following exposure to a pulse of reactor radiations. Bethesda, MD: Armed Forces Radiobiology Research Institute; Scientific Report SR66-2; 1966. Shirabe, R. Medical survey of atomic bomb casualties. Military Surgeon 113:251-263; 1953.