Radiation and Environmental Biophysics

, Volume 46, Issue 4, pp 299–310

Are cancer risks associated with exposures to ionising radiation from internal emitters greater than those in the Japanese A-bomb survivors?

Authors

    • Department of Epidemiology and Public HealthImperial College London, Faculty of Medicine
  • Per Hall
    • Department of Medical Epidemiology and BiostatisticsKarolinska Institute
  • Monty W. Charles
    • School of Physics and AstronomyUniversity of Birmingham
Review

DOI: 10.1007/s00411-007-0122-3

Cite this article as:
Little, M.P., Hall, P. & Charles, M.W. Radiat Environ Biophys (2007) 46: 299. doi:10.1007/s00411-007-0122-3

Abstract

After ingestion or inhalation of radionuclides, internal organs of the human body will be exposed to ionising radiation. Current risk estimates of radiation-associated cancer from internal emitters are largely based on extrapolation of risk from high-dose externally exposed groups. Concerns have been expressed that extrapolated risk estimates from internal emitters are greatly underestimated, by factors of ten or more, thus implying a severe underestimation of the true risks. Therefore, data on cancer mortality and incidence in a number of groups who received exposure predominantly from internal emitters are examined and excess relative risks per Sv are compared with comparable (age at exposure, time since exposure, gender) matched subsets of the Japanese atomic bomb survivor cohort. Risks are examined separately for low LET and high LET internal emitters. There are eight studies informative for the effects of internal low LET radiation exposure and 12 studies informative for the effects of internal high LET radiation. For 11 of the 20 cancer endpoints (subgroups of particular study cohorts) examined in the low LET internal emitter studies, the best estimate of the excess relative risk is greater than the corresponding estimate in the Japanese atomic bomb survivors and for the other nine it is less. For four of these 20 studies, the relative risk is significantly (2-sided < 0.05) different from that in the Japanese atomic bomb survivors, in three cases greater than the atomic bomb survivor relative risk and in one case less. Considering only those six low LET studies/endpoints with 100 or more deaths or cases, for four out of six studies/endpoints the internal emitter risk is greater than that in the Japanese atomic bomb survivors. For seven of the 24 cancer endpoints examined in the high LET internal emitter studies the best estimate of the ERR in the internal emitter study is greater than the corresponding estimate in the Japanese atomic bomb survivors and for the other 17 it is less. For six studies, the relative risk is significantly (2-sided < 0.05) different from that in the Japanese atomic bomb survivors, in one case greater than the atomic bomb survivor relative risk and in five cases less. Considering only those eight high LET studies/endpoints with 100 or more deaths or cases, for five out of eight studies/endpoints the internal emitter risk is greater than that in the Japanese atomic bomb survivors. These results suggest that excess relative risks in the internal emitter studies do not appreciably differ from those in the Japanese atomic bomb survivors. However, there are substantial uncertainties in estimates of risks in the internal emitter studies, particularly in relation to lung cancer associated with radon daughter (alpha particle) exposure, so a measure of caution should be exercised in these conclusions.

Introduction

Cancer risk is the largest component of radiation-induced detriment for a general population [1]. Most standards-setting bodies rely on cancer risks derived from cohorts exposed at high doses and dose-rates, in particular the Japanese atomic bomb survivor Life Span Study (LSS) cohort [1, 2], although risk estimates for certain other sites (e.g., liver, bone cancer) have also been derived from studies of medical uses of alpha-emitting radionuclides [1]. Traditionally, assessments of cancer risk associated with internal emitters require extrapolation of risk from high-dose groups and use of dosimetric models, because direct evidence of low-dose and low-dose-rate risk from internal emitters is scarce, limited to certain medically, occupationally and environmentally exposed groups [24]. Conventionally this has been of most importance for high linear energy transfer (LET) radiation, such as 239Pu, where radiation weighting factors of 20 are recommended [1]. There has been considerable recent interest and controversy, associated with estimation of risks from internal emitters [5, 6], with claims that risks from internal emitters could be greatly underestimated, by “well over a factor of 10”, by current International Commission on Radiological Protection (ICRP) risk factors and dosimetric models [5]; enhancements of risk from 90Sr by factors of 300 have been derived [7], although with less justification [5]. In part as a consequence of this, there have been recent reviews of risks in groups who received high LET internal emitters, for example from medical exposures (e.g., patients given the diagnostic contrast medium Thorotrast or 224Ra to treat ankylosing spondylitis and other non-malignant conditions) or from occupational exposure (e.g., radium dial painters, underground miners) [810].

In this article, we systematically review the epidemiological literature on internal emitters. We shall concentrate on the main groups exposed to internal emitters and compare risks with comparable subsets of the latest Japanese atomic bomb survivors [1115]. This survey therefore extends a similar survey in relation to cancer after medical treatment for malignant and non-malignant disease [8, 16].

Methods

A Medline search was conducted on 15 February 2007 using the terms “internal” + “emitter” + “radiation” (49 references from 1990 onwards), “internal” + “emitter” (72 references from 1990 onwards) and “internal” + “radiation” + “epidemiology” (579 references from 1990 onwards). This was supplemented by searches of appropriate tables in the most recent UNSCEAR report [2]. Only those studies in which there was reliable ascertainment of cancer incidence or mortality were considered and in which the majority of organ-specific dose came from internal emitters. Only peer-reviewed papers in English were considered; abstracts and conference proceedings were not included. In general, studies were excluded if there was no reliable estimation of organ dose. In some cases (e.g., all malignancies other than stomach in the study of Holm et al. [17], non-haemopoietic malignancies in the study of Ron et al. [18] and the studies of Nekolla et al. [19, 20] and Wick et al. [21]) the only comparison group available was the appropriate national population mortality or incidence rates, otherwise, internal comparison groups were generally available. In the case of haemopoietic malignancies in the study of Ron et al. [18] and in the study of Dickman et al. [22], Poisson linear relative risk models were refitted to the data given in the published material. For the Techa River study of Krestinina et al. [3], only leukaemia was examined, since only for this endpoint is the dose predominantly (92%) from internal emitters (mainly 90Sr and 137Cs).

The latest versions of the cancer incidence [11, 12] and mortality [1315] datasets for the Japanese atomic bomb survivors LSS cohort were employed. Follow-up for the haemopoietic (leukaemia, lymphoma, multiple myeloma) cancer incidence data [11] was from October 1950 to December 1987 and follow-up for the solid cancer incidence data [12] was from January 1958 to December 1998. Mortality follow-up started for all cases in October 1950, ending in December 1990 for the haemopoietic mortality dataset of Pierce et al. [13], in December 1997 for the solid cancer mortality dataset of Preston et al. [14] and in December 2000 for the solid cancer and leukaemia mortality data of Preston et al. [15]. Most of these data [11, 13, 14] employed the older DS86 dosimetry system [23], the only exceptions being the most recent solid cancer incidence [12] and mortality [15] datasets, which use the most current DS02 dosimetry [15, 24]. The basis of comparison is the value of the excess relative risk (ERR) coefficient (ERR Sv−1). To make such comparisons, a simple linear relative risk model was fitted to various subsets of the Japanese data, in which it was assumed that the expected number of cases in stratum i and dose group d with average organ dose D (in Sv) is given by:
$$ PY_{{id}} \lambda _{i} {\left[ {1 + \alpha D} \right]} $$
(1)
where PYid is the number of person-years of follow-up in stratum i and dose group d.
The dose to the atomic bomb survivors is a mixture of gamma rays and neutrons, with most dose imparted by radiation with high energy (>2 MeV) [23, 24]; the Sv organ dose is derived using the ICRP recommended [1] weighting factor for high energy neutrons of 10, as employed in other analyses of these data [1115]. The background cancer rates λi for each stratum i as well as the ERR coefficient α were estimated by Poisson maximum-likelihood [25]. The stratification was defined by city, sex, time since exposure and age at exposure in the haemopoietic cancer incidence data [11] and by city, sex, attained age and age at exposure in the mortality data [1315]; this is identical to the stratification employed in the original analyses of these datasets [11, 1315]. In the solid cancer incidence data [12] stratification was by city, sex, attained age, age at exposure, time since exposure and distance from hypocentre (<3,000 m, >3,000 m) for analysis of cancers other than thyroid cancer; for thyroid cancer, stratification was by city, sex, attained age, age at exposure, time since exposure and membership of the Adult Health Study (AHS). Distance from the hypocentre could not be used as a stratifying variable for thyroid cancer because such model fits were subject to numerical instability. Stratification by AHS membership is critical for thyroid cancer. AHS membership is proportionally greater in higher dose groups [26]. Diagnosis of thyroid cancer is highly dependent on the efficiency of the cancer registration process, which is more thorough in the AHS, so that it is possible that the “true” magnitude of the dose-response curve in the Japanese dataset could be distorted by variations in the completeness of ascertainment between the various dose groups within the cohort. Again, this stratification is similar to that used by Preston et al. [12] in their analysis of this dataset. All bomb survivors with shielded kerma dose >4 Gy were excluded from the analysis because of possible errors in dose estimates at high doses. The group not in either city at the time of the bombings was also excluded, as in the most recent analyses of this data [1115]. Because of the well-known modifying effects of age at exposure, time since exposure and sex as modifiers of the relative risk of radiation-associated cancer [2], appropriate subsets of the Japanese data were selected so that the period of follow-up, the age at exposure and the sex corresponded as nearly as possible to those of the respective internal emitter dataset. Consequently, the possible biasing effects of these variables in the comparisons will be minimized. Therefore, in Tables 1 and 2 two ERR estimates are given: the first being that obtained from the internal emitter study and the second based on an analysis of a relevant subset from the LSS cohort, both given as ERR Sv−1. The authors compared ERR estimates for studies of cancer incidence with those derived from the LSS incidence datasets [11, 12] and those for cancer mortality with the ERR estimated from the LSS mortality data [1315]. We have divided the internal emitter studies into those associated with low LET exposure (Table 1) and those associated with high LET exposure (Table 2). Because one would expect a dose-rate reduction for the protracted low LET exposures that characterise all of the studies considered in Table 1, we multiply the risks in these studies by the central estimate of the dose and dose-rate effectiveness factor (DDREF) used by the BEIR VII committee [27], namely 1.5. There is no reduction associated with the protraction of dose associated with high LET radiation, as would be expected on the basis of radiobiological data [28], so for the studies given in Table 2 we leave the estimates derived from each study unchanged.
Table 1

Risk estimates for cancer incidence and mortality from studies of internal low LET radiation exposure, adjusted to high dose rate using a DDREF of 1.5 [27]

Cancer endpoint

Reference

Age at exposure (years)

Follow-up (years)

Endpoint

Observed deaths/cases

Observed deaths/cases in comparable A-bomb group

Mean dose (Sv)

Person-years

Excess relative risk at 1 Sv (and 95% CI), adjusted to high dose rate

Excess relative risk at 1 Sv in comparable A-bomb group (and 95% CI)

All solid

[42]

0 to >40

0 to >40

Mortality

889

10,071

0.63

582,750

1.22 (0.69, 2.00)a

0.48 (0.39, 0.58)b##

Oesophagus

[42]

0 to >40

0 to >40

Mortality

317

288

0.63

582,750

0.27 (−0.14, 0.99)a

0.71 (0.20, 1.41)c

Stomach

[17]

13 to 74 (mean 57)

1 to 28 (mean 15)

Incidence

58d

1,575

0.25

139,018

1.98 (1.14, 2.97)

0.25 (0.05, 0.49)e###

Stomach

[18]

0 to >60 (mean 46)

1 to 44 (mean 21)

Mortality

28

2,852

0.178

738,831

−0.25 (−3.00, 3.39)

0.30 (0.14, 0.48)f

Stomach

[42]

0 to >40

0 to >40

Mortality

150

2,852

0.63

582,750

1.43 (0.26, 5.24)a

0.30 (0.14, 0.48)f

Stomach

[18, 42]

0 to >60

0 to >40

Mortality

178

2,852

0.45

1,321,581

0.79 (−2.15, 3.73)a

0.30 (0.14, 0.48)f

Liver

[42]

0 to >40

0 to >40

Mortality

60

1,234

0.63

582,750

−0.12 (−0.62, 1.50)a

0.51 (0.26, 0.80)g

Lung

[42]

0 to >40

0 to >40

Mortality

130

1,260

0.63

582,750

2.64 (0.72, 13.25)a

0.79 (0.51, 1.11)c

Female breast

[42]

0 to >40

0 to >40

Mortality

61

269

0.63

582,750

1.64 (−0.08, 23.70)a

1.42 (0.75, 2.31)h

Bladder

[18]

0 to >60 (mean 46)

1 to 44 (mean 21)

Mortality

14

149

0.128

738,831

−0.35 (−5.50, 7.35)

1.09 (0.19, 2.46)i

Kidney

[17]

13 to 74 (mean 57)

1 to 28 (mean 15)

Incidence

37d

41

0.05

139,018

15.3 (1.8, 32.4)

−0.06 (<−0.06, 1.91)j#

Thyroid

[22]

0 to 74 (mean 43)

2 to 47 (mean 27)

Incidence

36k

398

0.94k

886,618

1.63 (−0.01, 11.16)k

1.20 (0.63, 1.99)l

Thyroid

[43]

0 to 14

6 to 12

Incidence

276

13

0.37m

NA

9.90 (3.00, 16.65)n

8.53 (<0, 49.16)o

NHL

[18]

0 to >60 (mean 46)

1 to 44 (mean 21)

Mortality

74

170

0.042p

738,831

0.90 (−44.83, 83.74)p

0.05 (<0, 0.70)q

Hodgkin’s disease

[18]

0 to >60 (mean 46)

1 to 44 (mean 21)

Mortality

12

21

0.042p

738,831

−1.50 (−157.0, 49.6)p

0.48 (<0, 3.96)q

Multiple myeloma

[17]

13 to 74 (mean 57)

1 to 28 (mean 15)

Incidence

10d

22

0.06

139,018

−1.75 (−13.75, 17.75)p

1.23 (<0, 6.01)r

Multiple myeloma

[18]

0 to >60 (mean 46)

1 to 44 (mean 21)

Mortality

28

49

0.042p

738,831

16.50 (−7.65, 85.73)p

0.60 (<0, 2.55)s

Leukaemiat

[44]

0 to 6

0 to 14

Incidence

421

12

0.0063p

NA

48.60 (13.17, 126.00)p

97.39 (<0, 5989)u

Leukaemiav

[45]

1 to 65

0 to >47

Incidence

53

192

0.34, 0.29pw

NA

3.75 (−0.45, 27.00)p

5.79 (3.95, 8.35)s

Leukaemiav

[3]

0 to >70

0 to >49

Mortality

49

284

0.30p

865,812

9.75 (2.70, 36.00)p

4.09 (2.90, 5.64)s

Leukaemiav

[18]

0 to >60 (mean 46)

1 to 44 (mean 21)

Mortality

82

284

0.042p

738,831

−1.50 (−5.60, 0.87)p

4.09 (2.90, 5.64)s###

Leukaemiav

[3, 18]

0 to >70

0 to >49

Mortality

131

284

0.18p

1604,643

−1.09 (−5.85, 3.67)p

4.09 (2.90, 5.64)s#

NHL Non-Hodgkin’s lymphoma

aBased on a dose-response analysis, restricted to the exposed group only

bBased on colon dose

cBased on lung dose

dRestricted to the period 10 or more years after treatment

eBased on stomach dose, >15 years age at exposure, <30 years since exposure

fBased on stomach dose

gBased on liver dose

hBased on female breast dose

iBased on urinary tract dose

jBased on bladder dose, >15 years age at exposure, <30 years since exposure

kAmong group without previous exposure to external radiation therapy of the neck, referred for a reason other than suspicion of thyroid cancer

lBased on thyroid dose

mMedian dose to all Belarus subjects

nBased on linear model fitted to cases and controls with dose < 1 Gy

oBased on thyroid dose, <15 years age at exposure, <20 years since exposure

pBone marrow dose

qBased on bone marrow dose, fitted to incidence data [11]

rBased on bone marrow dose, >15 years age at exposure, <30 years since exposure

sBased on bone marrow dose

tAcute leukaemia

uBased on bone marrow dose, acute radiogenic leukaemia (AML + ALL), <10 years age at exposure, <15 years since exposure

vLeukaemia excluding chronic lymphocytic leukaemia

wMean internal emitter bone marrow doses for cases, controls, respectively

#LSS and radiation therapy ERR statistically inconsistent with P < 0.05

##LSS and radiation therapy ERR statistically inconsistent with P < 0.01

###LSS and radiation therapy ERR statistically inconsistent with P < 0.001

Table 2

Risk estimates for cancer incidence and mortality from studies of internal high LET radiation exposure

Cancer endpoint

Reference

Age at exposure (years)

Follow-up (years)

Endpoint

Observed deaths/cases

Observed deaths/cases in comparable A-bomb group

Mean dose (Sv)

Person-years

Excess relative risk at 1 Sv (and 95% CI)

Excess relative risk at 1 Sv in comparable A-bomb group (and 95% CI)

Liver

[38]

<20 to 79

2 to >50

Incidence

136

1,142

249.6a

54,734

+∞ (0.17, + ∞)

0.59 (0.30, 0.94)b

Liver

[35]

NA

0 to >50 (mean 31.2)

Mortality

454

1,234

249.6a

NA

0.61 (0.19, 1.18)

0.51 (0.26, 0.80)b

Liver

[36]

<11 to >60

0 to 54

Mortality

79

1,234

137.3a

9,356

0.13 (0.08, 0.19)

0.51 (0.26, 0.80)b##

Liver

[37]

<20 to >60

5 to >40

Mortality

106

1,234

249.6a

44,648

0.04 (0.02, 0.08)

0.51 (0.26, 0.80)b###

Liver

[38]

<20 to 79

2 to >50

Mortality

22

1,234

249.6a

13,691

0.09 (0.00, 1.86)

0.51 (0.26, 0.80)b

Liver

[3538]

<11 to 79

0 to >50

Mortality

661

1,234

234.1

67,695

0.06 (0.03, 0.08)

0.51 (0.26, 0.80)b###

Lung

[32]

NA

5 to 35

Incidence

7,148

619

NA

NA

2.07 (0.65, 4.00)c

0.69 (0.13, 1.16)d

Lung

[33]

NA

5 to 30

Incidence

3,662

440

NA

NA

1.71 (0.00, 4.34)e

0.89 (0.42, 1.51)f

Lung

[31]

NA

5 to 30

Incidence

768

440

NA

NA

4.96 (1.24, 21.24)g

0.89 (0.42, 1.51)f#

Lung

[31, 33]

NA

5 to 30

Incidence

4,430

440

NA

NA

1.85 (−0.27, 3.97)g

0.89 (0.42, 1.51)f

Lung

[46]

15 to 60

0 to 55

Mortality: male

167

622

5.00h

52,546

0.24 (0.17, 0.34)i

0.40 (0.03, 0.86)j

Lung

[46]

15 to 60

0 to 55

Mortality: female

25

508

8.42h

17,476

0.95 (0.48, 1.95)i

1.40 (0.76, 2.2)j

Lung

[29]

15 to >60

0 to >35

Mortality

2,674

620

3.26k

888,906

0.30 (0.17, 0.51)l

0.37 (0.07, 0.74)m

Lung

[29]

15 to >60

0 to >35

Mortality

2,674

620

3.26k

888,906

0.38 (0.11, 1.29)n

0.37 (0.07, 0.74)m

Bone

[21]

NA

0 to >36

Incidence

4

18

100o

32,800

0.02 (0.00, 0.07)p

1.32 (<0, 6.37)q

Bone

[20]

0 to >70

0 to 56 (mean 25.2)

Incidence

56

18

612r

25,500

0.30 (0.23, 0.39)p

1.32 (<0, 6.37)q

Bone

[20, 21]

0 to >70

0 to 56

Incidence

60

18

323.9

58,300

0.06 (0.03, 0.10)p

1.32 (<0, 6.37)q

Bone

[35]

NA

0 to >50 (mean 31.2)

Mortality

4

27

99.8s

NA

0.02 (−0.01, 0.37)

1.29 (<0, 5.16)q

Bone

[37]

<20 to >60

5 to >40

Mortality

5

27

99.8s

44,648

0.07 (0.00, 3.59)

1.29 (<0, 5.16)q

Bone

[35, 37]

<20 to >60

0 to >50

Mortality

9

27

99.8s

44,648

0.02 (−0.16, 0.21)

1.29 (<0, 5.16)q

Bone

[38]

<20 to 79

2 to >50

Mortality

2

27

99.8s

13,691

+∞ (−0.01, + ∞)

1.29 (<0, 5.16)q

Female breast

[19]

0 to >70

0 to 56 (mean 25.2)

Incidence

28

834

2t

NA

1.25 (0.66, 2.03)p

1.62 (1.16, 2.18)u

NHL

[35]

NA

0 to >50 (mean 31.2)

Mortality

15

170

62.4v

NA

0.02 (0.00, 0.05)w

0.05 (<0, 0.70)x

Hodgkin’s disease

[35]

NA

0 to >50 (mean 31.2)

Mortality

2

21

62.4v

NA

0.00 (−0.01, 0.08)w

0.48 (<0, 3.96)x

Leukaemiav

[38]

<20 to 79

2 to >50

Incidence

28

192

62.4v

54,734

0.23 (0.05, 2.38)w

5.79 (3.95, 8.35)y###

Leukaemiavz

[35]

NA

0 to >50 (mean 31.2)

Mortality

42

160

62.4v

NA

0.06 (0.02, 0.09)w

5.25 (3.41, 7.86)xz###

Leukaemiax

[37]

<20 to >60

5 to >40

Mortality

6

284

62.4v

44,648

0.15 (0.00, 7.53)w

4.09 (2.90, 5.64)y

Leukaemiax

[38]

<20 to 79

2 to >50

Mortality

8

284

62.4v

13,691

0.25 (−0.01, 3.38)w

4.09 (2.90, 5.64)y#

Leukaemiax

[37, 38]

<20 to 79

2 to >50

Mortality

14

284

62.4v

58,339

0.24 (−1.31, 1.78)w

4.09 (2.90, 5.64)y###

NHL Non-Hodgkin’s lymphoma

aBased on a liver dose estimate of 0.40 Gy per year (0.22 Gy per year for the Japanese series [36]), assumed given over 31.2 years [35], and using ICRP [1] weighting factor of 20

bBased on liver dose

cComputed from estimate adjusted for dosimetric error in [32], assuming conversion factor of 0.13 WLM/y/100 Bq m3 [34] and assuming 30 year (5–35 years before exposure) mean occupancy of houses, and using mean of correction factors (17.2 mSv/WLM, 22.5 mSv/WLM) calculated by Birchall and James [30]

dBased on lung dose, <35 years since exposure

eComputed from estimate in [33], assuming conversion factor of 0.13 WLM/y/100 Bq m3 [34] and assuming 25 year (5–30 years before exposure) mean occupancy of houses, and using mean of correction factors (17.2 mSv/WLM, 22.5 mSv/WLM) calculated by Birchall and James [30]

fBased on lung dose, <30 years since exposure

gComputed from estimate adjusted for dosimetric error (1.5 error GSD) in [31], assuming conversion factor of 0.13 WLM/year 100 Bq m3 [34] and assuming 25 year (5–30 years before exposure) mean occupancy of houses, and using mean of correction factors (17.2 mSv/WLM, 22.5 mSv/WLM) calculated by Birchall and James [30]

h Mean monitored dose among workers monitored at radiochemical and plutonium plants, obtained by multiplying dose in Table 1 of [46] by the recommended ICRP [1] weighting factor of 20 and adding to this the mean external dose to all workers

iERR per Gy at age 60, obtained by dividing the ERR/Gy in Table 3 of [46] by the recommended ICRP [1] weighting factor of 20

jERR per Gy at age 60, for LSS cohort [14], aged between 15–60, taken from Table 7 of [46]

kComputed from mean cumulative WLM in Appendix D of BEIR VI [29], using mean of correction factors (17.2 mSv/WLM, 22.5 mSv/WLM) calculated by Birchall and James [30]

lComputed from random effects estimate in Appendix A of BEIR VI [29], using mean of correction factors (17.2 mSv/WLM, 22.5 mSv/WLM) calculated by Birchall and James [30]

mBased on lung dose, males, 15–60 years age at exposure

nComputed from two-stage estimate in Appendix A of BEIR VI [29], using mean of correction factors (17.2 mSv/WLM, 22.5 mSv/WLM) calculated by Birchall and James [30]

oComputed from estimate of bone surface dose of 5 Gy, and using ICRP [1] weighting factor of 20

pComputed from given observed and expected number of cases, using exact 95% Poisson CI [56]

qBased on skeletal dose

rComputed using ICRP [1] weighting factor of 20

sBased on a bone dose estimate of 0.16 Gy per year, assumed given over 31.2 years [35], and using ICRP [1] weighting factor of 20

tBased on a breast dose estimate of 0.1 Gy (39), and using ICRP [1] weighting factor of 20

uBased on breast dose

vBased on a bone marrow dose estimate of 0.1 Gy per year, assumed given over 31.2 years [35], and using ICRP [1] weighting factor of 20

wBone marrow dose

xBased on bone marrow dose, fitted to incidence data [11]

yBased on bone marrow dose

zMyeloid leukaemia

#LSS and radiation therapy ERR statistically inconsistent with P < 0.05

##LSS and radiation therapy ERR statistically inconsistent with P < 0.01

###LSS and radiation therapy ERR statistically inconsistent with P < 0.001

In estimating the dose from alpha particles (the only source of radiation exposure for all studies considered in Table 2), we use the ICRP [1] recommended weighting factor of 20. In converting risk estimates per working level months (WLM) to risk per Sv associated with the BEIR VI [29] 11-cohort underground miner analysis, we also assumed the mean of the conversion factors (17.2 mSv/WLM, 22.5 mSv/WLM) derived by Birchall and James [30]. In converting the risks associated with the exposure rate (per Bq m−3) used in the domestic radon daughter exposure studies [3133], we also assumed the conversion factor of 0.13 WLM/y/100 Bq m−3 derived by Strom et al. [34] and assumed a mean occupancy of 30 years (5–35 years before case diagnosis) in the study of Darby et al. [32] and a mean occupancy of 25 years (5–30 years before case diagnosis) in the studies of Wang et al. [31] and Krewski et al. [33]. For the Thorotrast studies listed in Table 2 [3538], a common estimate of 31.2 years exposure of the respective organs was used, a figure derived from van Kaick et al. [35], combined with appropriate organ dose per year, which for all except the Japanese Thorotrast series [36] was also taken from Table 2 in [35]. For the 224Ra study of Nekolla et al. [19], a breast dose of 0.1 Gy = 2 Sv was employed, as indicated by Boice [39].

Also shown are the 95% confidence intervals (CI), which unless stated otherwise are likelihood-based (see below). All parameter estimates and CIs for the Japanese A-bomb survivors are estimated using AMFIT [40].

The last column of Tables 1 and 2 gives the P-values for the consistency of the Japanese and internal emitter datasets, calculated from the square of the normalized difference in ERR estimates:
$$ \chi ^{2} = [{\text{ERR}}_{j} - {\text{ERR}}_{m} ]^{2} /[\sigma ^{2}_{j} + \sigma ^{2}_{m} ] $$
(2)
where the estimated standard errors, σj, σm, for the Japanese and internal emitter data are computed from the 95% CI for the ERR of the Japanese data, (CIlj, CIuj) and from the 95% CI, (CIlm, CIum), for the internal emitter data, by means of the relation:
$$ \begin{aligned}{} & \left\{ \begin{aligned}{} & \sigma _{j} = [{\text{CI}}_{{uj}} - {\text{ERR}}_{j} ]/1.96 \\ & \sigma _{m} = [{\text{ERR}}_{m} - {\text{CI}}_{{lm}} ]/1.96 \\ \end{aligned} \right.\quad {\text{if}}\quad {\text{ERR}}_{j} < {\text{ERR}}_{m} \\ & \left\{ \begin{aligned}{} & \sigma _{j} = [{\text{ERR}}_{j} - {\text{CI}}_{{lj}} ]/1.96 \\ & \sigma _{m} = [{\text{CI}}_{{um}} - {\text{ERR}}_{m} ]/1.96 \\ \end{aligned} \right.\quad {\text{if}}\quad {\text{ERR}}_{j} \ge {\text{ERR}}_{m} \\ \end{aligned} $$
(3)
The P-values given in Tables 1 and 2 are derived from Eq. 2, which is assumed to have the distribution of a χ2 random variable with one degree of freedom (df) [25]. If the errors in the ERR estimates in the LSS and the medical data are both independently normally distributed, then Eq. 2 is equivalent to the likelihood-ratio test of the equality of the ERR. As the size of the samples gets sufficiently large this is, therefore, guaranteed to have the asymptotic distribution of a χ2 random variable with 1 df [41]. Alternatively, one may derive the distribution of this test statistic by considering, for simplicity, the case of normal random variables \( X_{j} \sim{}N(\mu _{j} ,\sigma ^{2}_{j} ) \) and \( X_{j} \sim{}N(\mu _{m} ,\sigma ^{2}_{m} ), \)\( Z = X_{j} - X_{m} \sim{}N(\mu _{j} - \mu _{m} ,\sigma ^{2}_{j} + \sigma ^{2}_{m} ), \) so that under the null hypothesis \( H_{0} :\mu _{j} = \mu _{m} , \) it is seen that \( Z^{2} = [X_{j} - X_{m} ]^{2} /[\sigma _{j} ^{2} + \sigma _{m} ^{2} ] \) has the distribution of a χ2 random variable with one df. Although this statistic behaves correctly asymptotically, for small samples the CIs are not approximately normal in the way that correct interpretation of this statistic requires; for this reason in those studies with five or fewer cases, we do not estimate this statistic.

In order to maximise statistical power, whenever possible (when the comparable group being considered in the Japanese atomic bomb survivors is identical) we derive joint estimates of internal emitter risks by weighting each ERR estimate by the inverse of the variance (computed in most cases by \( {\text{var}}_{m} = {\left( {[{\text{CI}}_{{um}} - {\text{CI}}_{{lm}} ]/[2 \times 1.96]} \right)}^{2} . \) This is the standard best linear unbiased estimator. The variance of this weighted estimator is easily seen to be given by \( 1/{\left[ {{\sum\limits_i {1/\text{var} _{{mi}} } }} \right]}. \) This is used to derive confidence intervals for this weighted estimate in the standard way.

Results

Internal low LET radiation exposure

There are eight studies informative for the effects of internal low LET radiation exposure [3, 17, 18, 22, 4245]. For 11 of the 20 cancer endpoints (subgroups of particular study cohorts) represented in Table 1 the best estimate of the ERR in the internal emitter study is greater than the corresponding estimate in the Japanese atomic bomb survivors. For four cancer sites, the risk is significantly (2-sided < 0.05) different from that in the Japanese atomic bomb survivors, in three cases (all solid: [42], stomach, kidney: [17]) greater than the atomic bomb survivor relative risk and in one case (leukaemia: [18]) less. Considering only those six studies/endpoints with 100 or more deaths or cases, for four out of six studies/endpoints (all solid, stomach and lung cancer: [42], thyroid cancer: [43]) the internal emitter risk is greater than that in the Japanese atomic bomb survivors; only for all solid cancer, in Bauer et al. [42], was this difference statistically significant. These numbers suggest that the relative risks in the low LET internal emitter studies do not appreciably differ from those in the Japanese atomic bomb survivors.

Internal high LET radiation exposure

There are 12 studies informative for the effects of internal high LET radiation exposure [1921, 29, 3133, 3538, 46]. For 7 of the 24 cancer endpoints (subgroups of particular study cohorts) represented in Table 2 ,the best estimate of the ERR in the internal emitter study is greater than the corresponding estimate in the Japanese atomic bomb survivors. For six cancer sites the ratio is significantly (2-sided < 0.05) different from that in the Japanese atomic bomb survivors, greater in one case (lung [31]), and less than the atomic bomb survivor relative risk, in the other five cases (liver: [36, 37], leukaemia, incidence and mortality: [38], leukaemia mortality: [35]). Considering only those eight studies/endpoints with 100 or more deaths or cases, for five (liver cancer [35, 38], lung cancer [3133]) the internal emitter risk is greater than that in the Japanese atomic bomb survivors; only for lung cancer in Wang et al. [31] was this difference statistically significant. These numbers suggest that the relative risks in the high LET internal emitter studies do not appreciably differ from those in the Japanese atomic bomb survivors.

Discussion

The LSS is the principal source of data used to estimate risks of radiation-related cancer [1, 2, 27, 47]. However, these risk estimates have also been compared to and generally supported by data taken from a variety of other sources [2]. In particular, data from studies of patients exposed to ionising radiation for the treatment of cancer and non-neoplastic conditions have been frequently used; previous studies [8, 16], documented the generally lower relative risks in the medical series compared with the atomic bomb survivors. We have analysed data on a variety of malignant endpoints in various groups predominantly exposed via internal radiation emitters. In general, whether for low LET or high LET radiation, we find little evidence of systematic departures of risk estimates in these studies from those in the Japanese atomic bomb survivors. There is slightly more evidence in the larger studies (those with 100 or more deaths or cases) [29, 3133, 35, 37, 38, 4244, 46] for there being greater risks than those in the atomic bomb survivors. However, there are substantial dosimetric uncertainties in some of these studies [29, 3133] and other problems in certain other studies [42], that we discuss below.

There are various other studies where there may be appreciable internal emitter exposure which could not be used in this paper. In particular, the study of Sellafield plutonium workers [4] presumably contains some information on this route of exposure, although only 355 person Sv of the dose is due to internally deposited plutonium, compared with 959 person Sv of external dose. No separate information is provided on risk associated with internal emitter dose (results are only presented for internal emitter and external dose combined) so that this study cannot yet be used to estimate risks from internal emitters. A study of thyroid cancer risk in relation to 131I exposure from the Hanford plant also could not be employed, because of insufficient information in the published paper to facilitate calculation of relative risk [48]. There are many fundamentally ecological studies in relation to various groups exposed as a result of the Chernobyl accident, which we have not considered here. The possibilities for bias in ecological studies are well known [49].

The method used to compare risks between the internally exposed and the Japanese atomic bomb survivors is to assess the ERR coefficients. The underlying assumption is that the ERR per unit dose should be invariant between these studies. This may well not be correct. A number of different ways can be used to transport cancer risk between two populations with different underlying susceptibilities to cancer. As well as the multiplicative transfer of risks, assumed here, one could also assume an additive transfer of excess absolute risks (EAR), in which the radiation-induced excess cancer rates in the two populations are assumed to be identical. The data that are available suggest that there is no simple solution to the problem [50]. For example, there are weak indications that the ERRs of stomach cancer following radiation exposure may be more comparable than the EARs in populations with different background stomach cancer rates [50]. A comparison of breast cancer risks observed in the Japanese atomic bomb survivor incidence data with those in various medically exposed populations, many from North America and Europe, where underlying breast cancer rates are higher than in Japan, suggests that ERRs are rather higher in the LSS than those in the medically irradiated groups, but (time- and age-adjusted) EARs are more similar [51, 52]. The observation that gender differences in solid tumour ERR are generally offset by differences in gender-specific background cancer rates [50] might suggest that EARs are more alike than ERRs. In summary, these considerations suggest that in various circumstances relative or absolute transfers of risk between populations may be advocated or, indeed, the use of a hybrid approach such as that employed by Muirhead and Darby [53] and Little et al. [54]. The information given in many of the papers considered does not permit calculation of an EAR and for this reason we concentrated on comparison of ERRs.

The most serious shortcomings of the present study are in relation to the estimates of risks from internal high LET exposure. In particular and as noted in the Methods section, in considering all the studies relating to radon daughter exposure and lung cancer [29, 3133] a number of dosimetric and other factors had to be assumed. For example, to convert from risk estimates per WLM in the miner studies [29] to risk per Sv we assumed a mean dosimetric conversion factor derived by Birchall and James [30]. In converting the risks associated with the exposure rate (per Bq m−3) used in the domestic radon daughter exposure studies [3133], we also had to assume an additional conversion factor (to derive WLM/year from exposure rate in Bq m−3) as given by Strom et al. [34] and make additional assumptions on mean period of occupancy in these studies. These factors could well be incorrect, although most of these, in particular the dosimetric conversion factors [30], are unlikely to be substantially biased [9]. For the Thorotrast studies listed in Table 2 [3538] a common estimate of 31.2 years exposure of the respective organs was used, a figure derived from [35], combined with the appropriate organ dose per year, which for all except the Japanese series [36] was also taken from Table 2 in van Kaick et al. [35]. This reflects lack of information in all the studies apart from van Kaick et al. [35], but again, the estimates are unlikely to be much out, as many of the Thorotrast cohorts were exposed and assembled for follow-up at similar times. The risks in all these studies of internal high LET radiation exposure may change when more information, for example from microdosimetric re-evaluations, becomes available.

Another weakness of the present study is that for certain studies, in particular stomach cancer in [17], and various non-haemopoietic malignancies in [1821], the only comparison group available was the appropriate national population mortality or incidence rates. It is not clear to what extent national rates are representative of the expected cancer incidence or mortality in these populations.

Internal dose for the Techa River [3] and Semipalatinsk [42] cohorts was based on age-dependent group-level (village) internal dose estimates, so that these studies are, as presently constituted, fundamentally ecological ones in relation to their dosimetry. The possibilities for bias in ecological studies are well known [49]. Reinforcing this, the patterns of variation of risk for solid cancer in the Techa River cohort are unusual, with indications that ERR increases both with age at first exposure (2-sided P = 0.08) and with attained age (2-sided P = 0.03) [3]; these patterns are not observed for leukaemia, although there is a suggestion of an increase in ERR with increasing age at first exposure (2-sided P = 0.10). Likewise, the ERR was statistically significantly increased with increasing age at exposure (P < 0.0001) in the Semipalatinsk cohort [42]. Such patterns are the reverse of what is observed in the Japanese atomic bomb survivor cancer mortality data [1315] and in many other radiation-exposed groups [2]. However, the magnitude of such bias is likely to be lower for leukaemia in the Techa River cohort [3] than for solid cancers in this group and in the Semipalatinsk cohort [42], where environmental carcinogens (e.g., tobacco use) and occupational exposure other than ionising radiation are more likely to severely influence the risk.

Fundamental to our adjustment of risk from low LET radiation to estimate the effective high dose rate risk, is the value of DDREF used to adjust the risk estimates upwards, or downwards if the risk is negative. The value of DDREF that we use, 1.5, is the central estimate recommended by the BEIR VII committee [27]. BEIR VII estimated the 95% uncertainty range in this parameter to be 1.1–2.3 [27]. Clearly use of the larger value of DDREF of 2 recommended by ICRP [1] would elevate the effective high dose rate risks that we estimate further and equally, use of a value of 1 or even less, as implied by others [7], would reduce these risks. Generally, use of DDREF in the range 1.1–2.3 [27] would scarcely affect the conclusions of this paper. Even when using the upper extreme value of 2.3, only for one further endpoint (lung cancer in the Semipalatinsk study [42]) does the discrepancy in risk from that in the atomic bomb survivors become significant.

Arguably there is more that could be done to obtain pooled estimates of risk for certain endpoints, for example stomach, liver, lung, bone and leukaemia where there are multiple studies reporting risks (Tables 1, 2). We have conducted simple meta-analyses combining those studies that we judge similar enough with respect to years of follow-up and range of ages at exposure (Tables 1, 2). A more thorough analysis would require fitting models to the various datasets, ideally based on access to individual data and this is beyond the scope of the present paper.

As noted in the Introduction, there have been other reviews [8, 10] comparing the risks of internal emitters in medically exposed cohorts with those seen in the atomic bomb survivors. In general there is little evidence of discrepancy in risks between these groups [8, 10]. There have also been comparisons of risks associated with radon-daughter exposure in miners and among residentially exposed groups with those in the A-bomb data by various researchers [9, 10, 30] and more general comparisons of risks from alpha-particle emitters in occupationally-exposed groups [55]. Again, there is little evidence of discrepancy between the risks in these groups [9, 10, 30, 55]. Therefore, consideration of the totality of epidemiological evidence provides little evidence that cancer risks from internal emitters have been substantially underestimated, for example by the factors of 300 or more claimed by some groups [7], although until the dosimetric uncertainties in the alpha-particle-exposed cohorts considered here are resolved, there remain uncertainties in these risk estimates.

Acknowledgments

The authors are grateful for the detailed and helpful comments of the two referees. This report makes use of data obtained from the Radiation Effects Research Foundation (RERF) in Hiroshima and Nagasaki, Japan. RERF is a private, non-profit foundation funded by the Japanese Ministry of Health, Labour and Welfare (MHLW) and the U.S. Department of Energy (DOE), the latter through the National Academy of Sciences. The data include information obtained from the Hiroshima City, Hiroshima Prefecture, Nagasaki City and Nagasaki Prefecture Tumour Registries and the Hiroshima and Nagasaki Tissue Registries. The conclusions in this report are those of the authors and do not necessarily reflect the scientific judgment of RERF or its funding agencies. The work was funded in part by the UK Department of Health and by the European Commission under contract FI6R-CT-2003-508842 (RISC-RAD).

Copyright information

© Springer-Verlag 2007