The Science oft he Total Environment, 127 (1992) 13-35 Elsevier Science Publishers B.V., Amsterdam

13

Nuclear installations and childhood cancer in the UK: mortality and incidence for 0-9-year-old children, 1971 - 1980* J o h n R. G o l d s m i t h

Epidemiology and Health Services Evaluation Unit, Occupational Epidemiolagy Section, Faculty of Health Sciences, Ben Gurion University of the Negev, P.O.B. 653, ~e:i Sheva, Israel ABSTRACT Data on cancer incidence and mortality in the vicinity of nuclear installations in England and Wales have been published, covering the period !959-1980. Several age classes and a number of cancer sites have been included. Systematic analysis has so far been limited to cancer mortality. This suggests that childhood leukemia is unusually frequent near some ty~s of installations and the excess is greater for years 0-9 than for later ages. In the earlier- decade, 1959-1969, there was questionable consistency of incidence reporting. This report uses the incidence and mortality data only for the period i97 I-1980; leukemia and non-leukemia cancer data for those 0-9 years are analyzed, and consistency of iricidence reporting is evaluated by comparing incidence and mortality. For comparison with reported data for grouped local authority areas (LAAs) near to 21 different installations, two sources of expected incidence are used. The first is b~,sed on regional data, and the second is based on a group of LAAs similar in location, urban/rural character and population size. Effectiveness of treatment of acute lymphocytic lev,kemia (ALL, the principal type among 0-9-year-old children) ha~ improved in recent decades and this has decreased the case fatality rate, and increased the relevance of incidence data. No significant differences in the mortality/incidence ratio were found when comparing areas near installations with their matched reference areas. Although some standardized incidence ratios (SIRs) are low in comparison areas, these are a reflection of the small numbers in the denominator. Low and high SIRs are observed in both installation and comparison areas. Excluding the populations near Sellafield and Dounreay, the installations of which include nuclear reprocessing and which are reported elsewhere, children living near one of a group of six installations have significantly increased leukemia incidence. One hundred sixty-five cases were observed for the combined populatiofi group, whereas 133.6 were expected on the basis of regional data and lal.9 on the basis of comparison areas. Tikc operations of these installations 'include nuclear research, nuclear fuel production, nuclear weapons manufacture, and radio-isotopic reagent manufacture and marketing. No evidence of any increase of leukemia incidence is found for populations surrounding nuclear power installations, nor in coastal communities possibly affected by Sellafield discharges. In comparing these data with leukemia mortality and with other types of cancer (non-leukemia) mortality and morbidity, an incidental finding was of significantly elevated non-leukemia *Supported in part by a contract with the European Regional Office of the World Health Organization; this report is not considered a report by or of the World Health Organization.

J.R. GOLDSMITH

mortality in the vicinity of a group of nuclear power installations, based on comparison Out not on regional data (24 observed cases, 25.7 expected on basis of regiom~ldata, 10.6 on the basis of comparison area data). Due to small number~.~,no specific cancer site can be found to account for this, and it may be an artifactual finding. These results contrast with the findings of Cook-Mozaffari et al.'s regression analysis of mortality which showed a positive, but not significant association for leukemia among residents near nuclear power plants. This study confirms Cook-Mozaffari et al.'s findings by mortality analysis for the group of installations involved in non-power producing nuclear technology. This study sheds no light on the possible basis for the association, which remains to be clarified. Initial steps have been taken by Gardner et al. using case-referent methods Key words: acute lymphatic leukemia (ALL); radiobielogy; cancer ~pidemiology; nuclear

power; nuclear fuel; nuclear weapons; pre-natal exposures

. .

INTRODUCTION

In November 1983, the F~ntish Independent Television network r,~ported on the apparent excess incidence of leukemia among children living near the nuclear installations at Sellafield on the Cumbrian coast. An expert Advisory Group was immediately appointed to make a report which it completed within a year, generally c o ~ r m i n g the TV broadcast [1]. A governmental Committee on Medical Aspects of Radiation in the Environment (COMARE) was appointed and has made three reports [ 2 - 4 ] . The Office of Population Censuses and Surveys (OPCS) has compiled and published a report on cancer incidence and mortality in the vicinity of nuclear installations in England and Wales [5] and several highly qualified teems have analyzed m~rtality data from the report [6,7]. Others have ~'eported on cancer in the vicinity of nuclear installations in Scotland [8.o]."At the request of the European Regional Office of the World Health Organization, a review of The usefulness of Epidemiological Data fox: Evaluation of Health Effects from Point Source Radiation has been prepared, summarizing much of the British experience and opinions. The Report has been widely circulated and many comments, corrections, and criticisms Have been received [10]. A revision has be~n prepared and distributed. However, one large set of available data has not been systematically analyzed, and that is the purpose of this report. These are the data for 1971-1980 on the incidence of childhood leukemia at ages 0 - 9 years [5]. The data will be presented in parallel with the corresponding mortality data. Both will be analyzed with respect to expected numbers based on both the regiona.l health statistics and on the basis of matched local areas thought to be similar in location, population, and urban/rural character. Both analytical reports of mortality indicate [6,7] that among those 0-24 years of age, living near certain kinds of installations, there appears to be a small, but significant excess of leukemia deaths. Analyses were based on the Standardized Mortality Ratios (SMRs), adjusting for age.

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

15

There are three reasons given as to why the massive British report on a problem which has engaged many of the country's best statisticians and ep~demio!ogb;~Lso,~er a period of nearly :~i~years~ ~eems to have only been partly analyzed" (1)

that during the 1959-1969 period, the cancer registration system- was being introduced and incidence data were likely to have been incomplete; (2) there seemed to have been some evidence or at least a suspicion that in communities near to nuclear installations, such cases of leukemia as did occur were more likely to be reported than if the case had occurred in some other community. (3) The third reason given for not completing the analyses originally intended for the OPCS data [5] was that a number of the SIRs or SMRs for the comparison areas were unusually low.

Swerdlow has given a detailed analysis of the development and consistency of the cancer registration system in England and Wales [11 ]. He emphasizes that the registration although a national programme, is carried out through regional registries. In order to take suitable account of othe:,o primarily demographic variables, which may have affected cancer incidence and mortality, the OPCS study matched areas adjacent to nuclear installations with areas sinfilar in urban/rural character, population size, al~d region to use as a ba:.~i,s for expected numbers. Both the group of Local Authority Areas (LAAs) areas near to the installations (installation areas) and the comparison areas had their incidence and mortality data compared with regional incidence and mcxtality d~ta, to produce a Standardized Mortality Ratio (SMR) or a Standardized ln~idence Ratio, SIR (or Standardized Registration Ratio as they were designated in the OPCS repot0. The ratios of incidence and of mortality ratios for installation areas and for comparison areas were to be assessed according to the resulting Risk Ratios. While to look at all sites of cancer at a variety of distances from all nuclear installations in comparison with regional and comparison area experience is clearly an exploratory study of ecological data, one could define a specific hypothesis derived from previously available data, and use the OPCS data to test the hypothesis. That approach will be used here. FORMULATION OF THE HYPOTHESIS

Such an hypc,thesis is readily enough defined from the ITV programme which found elevated leukemia incidence in children under 10, living near Sellafield, and f~,'om the data derived from the Oxford Childhood Cancer

16

J.R.

GOLDSMITH

Study (OCCS) which found that leukemia in children under 10 years of age was the principal childhood cancer related to radiation during pregnancy [12]. Gardner et al. have shown that the excess ieukemia was restricted to children born in Sellafield and did not occur among children who moved 'to the community after birth [13,14]. This finding is ,consistent with the possibility that any radiation exposure which may be relevant occurred in the preconception, pre-natal, or peri-natal period. Doll has recently published an informative review of the Epidemiol0gy of Childhood Leukemia [15]. During the last 20 years, 'the prognosis of acute lymphatic leukemia, the principal childhood form of leukemia, has greatly improved, so that since 1955 there has been a dramatic drop in childhood leukemia mortality, despite an increase in incidence. Figure l (from Ref. 15) shows that the peak incidence is now at 3-4 years. Other reports had shown that choice of boundary conditions for areas

37 •t

~ -0'"

3O e-

== -v °D

e" O

0 0 0

/

d 0 0

,c =

20

O. q)

4.A

c c

10

.._/

5 0

----J~

1911 '!6

_1

'26

...11_

!

I

'46

'6~,

'86

Year Fig. 1. T:~nds in mortality and morbidity attributed to leukemia per million children per year

(England and Wales): Mortality 1981-1985 (O) anO Morbidity 1961-1965 to 1981-1934 (O) from Doll, 1989 (see Ref. 15).

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

17

near installations could greatly alter the ratios of observed to expected deaths or cases. For this reason, it was decided to examine only the largest area (and population) for each installation and its comparison area reported in the OPCS study, namely the area including all Local Authority Areas with at least 1/3 of their population living within l0 miles of the installation. Because cancer incidence reporting has been relatively uniform in recent years [6], it was decided to examine the OPCS data to test the hypothesis that leukemia incidence at ages 0 - 9 years during 1971-1980 was associated with residence near (as defined above) nuclear installations of various types. This report examines the ratio of incident cases to mortality for each area, as a test for consistent and complete reporting. The observed to expected ratios for both incidence and mortality for leukemia and for non-leukemic cancers will be evaluated when the expected numbers were: (l) derived from regional data and (2) when the expected numbers were derived from the experience in the matched comparison area. We shall consider consistency between results based on these different sources of expected numo,.,ko-o~to weigh against any results being based on a statistical artifact. METHODS

What we can excerpt from Cook-Mozaffari et al. [5] are the observed, numbers of cases and deaths and the SRRs [Standardized Incidence (registration) Ratios] and SMRs (Standardized Mortality Ratios) based on Regional health statistics for installation and for comparison areas. (In Appendix 1 a formal treatment of these relationships can be found.) The expected numbers can be estimated readily by dividing the observed numbers by the corresponding SMR or SRR (or SIR), expressed as a ratio. (That is instead of dividing by say 120 or 94, divide by 1.20 or 0.94.) The relative size of the age-adjusted populations of the installation and coroparison areas is proportional to the ratio of the expected numbers derived as indicated. Division of the observed numbers for the Comparison area by the expected number of that area and multiplication by the expected number of the Installation area gives the expected number derived from the comparison area experience adjusted for any differences in population size and distribution. In this case, since the expected values are really derived from the ex.perience of a population of nearly the same size as that which produced the obserced data, the relationship of Observed to Expected should be evaluated by the Binomial distribution, and not the Poisson, as was appropriate for the O/E based on the Regional expected numbers. The computing algorithms used are shown in Appendix 2.

18

J.R. GOLDSMITH

TYPES O F SITES

The OPCS data set deals with three different types of sites, treated as installation areas and their matched comparison areas. The first group are the oldest nuclear processing, nuclear research, nuclear weapons, nuclear reagent, and nuclear fuel installations, all of which, except one, were in operation before 1955. They include Springfields and Capenhurst, operated by British Nuclear Fuels plc (BNFL), Harwell operated by the United Kingdom Atomic Energy Authority (UKAEA), Aldermaston, operated by the Ministry of Defence (MOD) and Amersham, operated by Amersham International plc, a supplier of radioisotopes. The exception is Winfrith, an UKAEA installation which began operations in 1964. Although Sellafield and Dounreay were also operated by the UKAEA, they are not included in the summary for type of site because Sellafield was the site which first raised suspicion, and Dounreay is in Scotland, and not covered by the OPCS. Both of these installations are involved in nuclear fuel reprocessing. For information, data for Sellafield are tabulated. The second group are nuclear power installations operated by the Central '~Electricity Generating Board (CEGB), the first of which began operating in 1961. They include Berkeley and Oldbury, which are adjacent, Bradwell, Hinkley Point, Trawsfynydd, Dungeness, Sizewell and Wylfa. Together they are designated the CEGB installations. The third group are not nuclear installations at all but locations on the Irish sea which at one time were thought to be at risk of exposure to seaborne contaminants from SeUafield. For each of these coastal communities, a comparison area was also chosen. The coastal areas include Northern Cumbria, Mid Cumbria, Southern Cumbria and Northern Lancashire, Southern Lancashire and North Merseyside, Liverpool, Wirral Peninsula, and North Wales° When the data for these are summed, Mid Cumbria is omitted ~.s it includes the SeUafield area. Thus we can consider three different types, each including seven locations, naively, the Pre-1955 group plus Winfrith, (which will be designated 'Pre-1955' consistent with other reports) and Sellafield; 7 CEGB sites, and seven Coastal areas and a comparison area for each of the 21. No cases or deaths were observed during the study period for some sites, thus reducing the numbers of actual observations, especially for CEGB sites, which were usually located in remote places, with small populations. The data are tabulated in the OPCS microfiche by 5-year periods, of which we shall consider only the last 2, 1971-1975 and 1976-1980 taken together, for leukemia incidence and mortality and for other cancer incidence and mortality at ages 0-9 years. Thus for each of the 21 sites there are four possible types of outcome

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

19

(leukemia or other type of cancer, mortality or morbidity'.~, each of which is reflected in an observed and two expected numbers, one based on regional data and the second on comparison areas. Since the expected numbers based on regional data have a much larger population denominator at risk than the installation LAAs, Poisson distributions are used to evaluate the likelihood that random processes have led to the observed value given the expected ones. For expected numbers based on comparison areas, the populations at risk are approximately equal for observed and expected, so the relationship between Observed and Expected numbers will be evaluated based on Binomial probabilities. In Tables 3 and 4, w~ present the area designator (NAME), the observed INCIDENCE or MORTALITY, for the installation LAAs and below INCIDENCE or M O R T A L I T Y for the comparison area, the SMR o r SIR based on Regional data (which are given in the microfiche [5]), the Expected number (EXPEC.r) based on the regional data (column 2/column 3) times 100, the SMR or SIR based on the comparison area incidence or mortality and regional data, and the adjusted Expected number based on the experience in the comparison area corrected for any differences in population size, age or sex distribution (EXPEC.c). [This is also designated as E ' ( C 0 of E'(Mi) as derived in Appet~dix 1], the one-tailed probability that a value at least as high as the observed would be found from a distribution derived from EXPEC.r, based on Poisson distributions (pr), and finally the onetailed probability that a value at least as high as the observ,ed wo~ld be found from a d~tribution based on the Binomial relationships derived from EXPEC.c (pc). When no cases or deaths are observed for the 10 years of the study. 0' is in the second column, and the others are blank. When no cases or deaths are observed in one of the two 5-year periods, this ;s indicated by a superscript a or b, and the expected number in the period wi~h observations is doubled. For convenience the probabilities for elevated O/E, 0.05 > P > 0.01 are indicated as *, and 0.01 > P are indicted by **. RESULTS

We first examine the ratio of mortality to incidence by location end time period. The results are shown in Tables 1 and 2. In only one pair are there significant differences between the ratios for installation and comparison areas. That is for Springfields, and the ratios are consistent with more incident cases in the comparison area. It can be seen that there is no undue excess of cases to deaths in recent y,ears (1971-1980). The overall rates are almost identical 0.6842 fc,l~instailauon areas and 0.6851 for comparison areas. Of course tb ,~ ratios are not very stable when the numbers of cases or deaths are small.

20

J.R. GOLDSMITH

Comparison areas have a higher ratio (more deaths per each ease) in the Pre-1955 areas and in (CEGB) Nuclear Power areas. In the Coastal areas, the cases of leukemia are proportionately fewer in the (so-called) installation areas. Thus the data in Tables 1 and 2 provide no support for rejecting the analysis of 1971-1980, 0 - 9 leukemia incidence data, due to inadequate or biassed reporting. This may not be the case for other cancer sites, ages and time periods. Tables 3-6 give the data for each location and for the summary by types TABLE 1

Ratio of mortality to incidence for malignancy (all malignancies (All), except brain malignancies and leukemia (Except B and L), leukemia (L)) years by grouped locations and time periods for installation and comparison areas, ages 0-.9 years Location

Installation areas All

Pre-55 except Sellafield 61-65 1.144 167/146 66-70 0.624 116/186 71-75 0.617 116/188 76-80 0.497 93/187

Total

0.696 492/707

All CEGB 61-65 0.846 33/39 66-70 0.696 32/46 71-75 1.031 33/32 76-80 0.367 11/30

Total

0.809 119/147

Comparison areas

Except B and L

L

All

Except B and L

L

1.149 77/67 0.540 47/87 0.605 46/76 0.316 25/79

1.451 74/51 0.889 48/54 0.684 52/76 0.575 46/80

1.355 191/141 0.631 128/203 0.701 117/167 O 4521 82/181

0.973 73/75 0.550 61/111 0.632 48/7.5 0.305 26/8:5

2.1i6 91/43 1.059 54/51 0.789 45/57 0.565 39/69

0.631 195/309

0.812 220/271

').749 518/692

0.599 208/347

1.041 229/220

0.813 13/16 0.500 10/20 0.867 13/15 0.400 6/15

1.273 14/11 1.000 17/17 I. 143 16/14 0.538 4/13

1.133 34/30 0.885 23/26 0.697 25/33 0.500 10/20

0.800 12/15 0.636 7/11 0.600 6/10 0.222 2/9

2.222 20/9 0.667 6/9 0.750 15/20 0,889 8/9

0.636 42/66

0.927 51/55

0.826 90/109

0.600 27/45

1.255 59/47

21

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

TABLE 2 Mortality to incidence ratios for leukemia, ages 0-9 years by individual and grouped location for installation (I) and Comparison (C) areas in recent time periods Location

A. Pre-1955 installations Sellafield

Time periods

~ C

Springfields

I C

Capenhurst

I C

Amersham

I C

Aldermaston

I C

Harwell

I C

Win frith

I

C

All except Sellafield

I C

B. Nuclear power installatio~ls All CEGB* I installations (Considered together C as numbers are small

1971-1975

!976-t980

1971-1980

0.750 3/4 0.500 !/2 0.643 9/14 0.333 7/21 ! .000 13/13 1.545 17/11 0.581 18/31 0.545 6/11 0.688 11/16 1.C~0 I1/11 0.500 1/2 1.333 4/3 1.000 2/2 0.750 3/4 0.692 54/78 0.787 48/61

0.000 0/4 0.000 0/0 0.786 11/ ! 4 0.556 10/18 0.500 8/16 0.778 7/9 0.586 17/29 0.520 13/25 0.421 8/19 0.563 9/'16 !.000 2/2 0.000 0/1 0.286 2/7 0.800 4/5 0.552 48/87 0.581 43/74

0.375 3/8 0.500 !/2 0.714 20/28# 0.436 17/39 0.724 22/29 i.200 24/20 0.583 35/60 0.528 19/36 0.543 19/35 0.741 20/27 0.750 3/4 1.000 4/4 0.444 4/9 C.777 7/9 0.618 102/165 0.674 91/135

I. 143 16/14 0.750 15/20

0.308 4/13 0.889 8/9

0.741 20/27 0.793 23/29

22 TABLE 2

J.R. GOLDSMITH

(Continued) Time periods

Location

1971-1975

1976-1980

1971-1980

C. Coastal sites N. Cumbria

I C

Mid Cumbria

l C

S. Cumb. and N. Lanc.

I C

S. Lanc. and N. Mersey

I C

Liverpool

I C

Wirral

I C

All except Mid Cumbria

I C

2.000 2/1 1.000 3/3 0.750 3/4 0.500 1/2 0.375 3/8 0.500 6/12 1.000 15/15 0.591 13/22 0.571 16/28 0.684 13/19 0.571 8/14 0.733 11/15 0.803 53/66 0.704 50/71

0.000 0/1 0.667 2/3 0.000 0/4 -0/0 0.500 3/6 0.100 1/10 0.857 12/14 0.625 10/16 0.818 9/11 0.714 10/14 0.429 6/14 0.636 7/11 0.717 33/46 0.630 34/54

1.000 2/2 0.833 5/6 0.375 ~/8 0.500 1/2 0.429 6/14 0.318 7/22 0.931 27/29 0.821 23/28 0.641 25/39 0.697 23/33 0.500 14/28 0.692 18/26 0.768 86/112 0.672 84/125

0.8101 123/158 0.7434 113/152

0.5822 85/146 0.6204 85/137

0.6842 208/304 0.6851 198/289

D. All locations I C

of installations. A summary for types is found in Table 7 which also gives the 'z' scores, which are graphed in Figs 2 and 3. A significant excess of leukemia incidence is found only for populations resident near the Pre-1955 installations, and this is the case for both types

23

I~UCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

TABLE 3 Incidence of leukemia, ages 0 - 9 years, in the vicinity of British nuclear installations during 1971-1980 Name a

Incidence

A. Pre-1955 installations Sellafield 8/2 Springfields 28/39 Capenhurst 29/20 Amersham 60/36 Aldermaston 35/27 Harwell 4/4 Winfrith 9/9 All except 165/135 Sellafield

SIRr

EXPEC.r

| E ' (Ci) j EXPEC.r

pr

pc

! 91.8 85.4 101.8 147.7 146.4 51.3 127.7

4.177 25.20 28.50 40.63 23.90 7.80 7.05

79.3b 130.0 69.4 78A 142.8 53.2 121.6

166 32.77 19.77 31.87 34.12 4. i 5 8.57

0.062 0.314 0.487 0.0026** 0.020* 0.952 0.277

0.029* 0.742 0.092 0.0016"* 0.461 0.638 0.550

123.5

133.61

96.1

141.91

0.005**

0.0~5"

82.81

5.41

0.890

C.781

143.4 16.81 94.68 5.79 . . 197.9 2.0! . . 88.9 b 0.92 i I 1.5 30.60

0.826 0.165

0.934 0.282

0.093

0.690

0.340 0.558

0.281 0.6~

B. Nuclear power installations Bradwell 4/5 61.26 6.53 lkrkeley-Oldbury 9,'!5 76.77 1!.72 Hinkley 9/6 i47.0 6.11 Trawsfynydd 0/0 . . . Dungeness 2/2 395.0 *b 1.01 Wylfa 0/0 . . . Sizewell 3/1 145.7 2.06 All CEGB 27/29 98.4 27.43 C. Coastal sites N. Cumbria 2/6 Mid Cumbria 8/2 S. Cumb. & N. Lanc. 1.4/22 S. Lanc. & N. Mersy. 29/38 Liverpool 39/33 Wirral 28i24 N. Wales 0 All except: 112/125 Mid Cumbria

SIRc

32.95 191.53

6.07 4.18

91.9 79.3 b

5.58 !.66

0,9~4 0.063

0.955 0.029*

93.42

14.99

120.12

18.00

0.636

0.776

97.28 29.81 103.77 37.58 91.36 30.65 . . .

116.71 97.75 81.66 . .

34.79 36.35 23.10 .

0.583 0.430 0.708

0.778 0.400 0.244

94.03 119.11

101.65

121.08

0.755

0.730

aFor definition of the column captions and contents, see the section Types of Sites.

24

J.R. C~LDSMITH

TABLE 4 Mortality of leukemia, ages 0 - 9 years, in the vicinity of British nuclear installations during 1971-1980 Name

Mottality

A. Pre-1955 installations Sellafield 3/1 Springfields 20/17 Capenhurst 21/24 Amersham 35/19 Aldermaston 19/20 Harwell 3/4 Winfrith 4/7 All except 102/91 Sellafield

SMRr EXPEC.r

{E'(Mi)} EXPEC.c

pr

pc

150.3 a 137.4 104.8 123.5 116.4 55.5 102.9

3.99 14.56 20.04 28.34 16.32 5.40 3.89

93.0 101.0 119.5 60.6 I ~3.1 137.3 a 171.3

0.99 14.71 23.95 17.18 24.99 3.71 6.61

0.323 0.!02 0.444 0.125 0.285 0.905 0.545

0.316 0.176 0.670 0.0054** 0.780 0.718 0.867

115.8

88.09

97.2

85.12

0.079

0.112

4.70

168.5

7.93

0.211

0.919

6.46 3.80 0.164 0.706 1.122 !. I 1 17.41

155.9 144.0 -247. I a --139.3

10.08 5.46 -0.87 --24.25

0.885 0.331 0.151 0.158 0.306 0.305 0.298

0.969 0.663 -0.449 --

B. Nuclear power installations Bradwell 4/7 170. I a Berkeley-Oldbury 4/9 61.9 Hinkley 5/5 265.3 a Trawsfynydd 1/0 609.7 a Dungeness 2/1 284.4 Wylfa 2/0 356. I a Sizewell 2/0 179.8 All. CEGB 20/23 114.9 C. Coastal sites N. Cumbria Mid Cumbria S Cumb. and N. Lane. S. Lane. and N. Mersey Live, pool Wirral N. Wales All exce!~t Mid Cumbria.

SMRc

0.741

2/5 3/1

69.5 a 150.3 a

5.76 4.00

111.88 49.7 a

6.44 0.99

0.979 0.323

0.965 0.312

6/7

70.61

8.51

68.16

5.80

0.850

0.582

27/22 25/23 14/18 12/,o

154.16 107.05 73.54 150.54

17.51 23.35 19.04 7.97

117.17 110.95 91.22 116.67

20.52 25.91 17.37 9.30

0.021 0.393 0.903 0.110

0.169 0.549 0.732 0.281

86/84

106.85

80.49

102.80

82.74

0.284

0.401

aMeans that there are no deaths for one of the two 5-year periods. The expected number for the period with deaths has been doubled.

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

25

TABLE 5 Incidence of all cancer except leukemia, ages 0-9 years, in the vicinity of British nuclear installations during 1971-1980 Name

Incidence

A. Pre-1955 installations Sellafield 9/5 Springtields 49/53 Capenhorst 41/41 Amersham 53/72 Alde~maston 61/47 Harweil 15/9 Winfrith 7/11 All except 226/233 ~llafield

SIRr

EXPEC.r

11~5.9 103.1 97.4 ti3.11 163.0 !,22.8 62.3

7.77 47.52 42.10 63.77 37.42 ! 2.22 11.23

65.58 96.0 96.92 100.0 158.4 76.2 93.53

107.86

209.53

105.22

B. Nuclear power installations Bradwell 8/4 78.0 10.26 Berkeley-Oldbury ~6/7 85.9 18.63 Hinkley i0/8 102.9 9.71 Trawsfynydd 0/0 . . . Dungeness 0/1 . . . Wylfa 0/0 . . . Sizewell 1/4 61.0 a 3.28 A!l CEGB 35/24 80.7 43.35 C. Coastal sites N. Cumbria Mid C-mbria S. Cumb. and N. Lanc. S.Lanc. and N. Mersey Liverpool Wirral N~ Wales

12/12 9/5

SIRc

[ E'(Ci) ] pr EXPEC.c.

5.09 45.62 40.80 63.78 59.27 9.31 10.50

pc

0..375 0.434 0.588 0.924 0.0005** 0.248 0.930

0.213 0.360 0.491 0.844 0.437 0. ! 29 0.856

220.47

0.132

0.395

42.2 42.1 79.4 . . . 121.2 58.45

4.33 7.84 7.71 . . . 3.97 25.34

0.802 0.760 0.505

0.237 0.054 0.379

0.963 0.914

0.965 0.11 !

. . .

105.74 ! !591

! 1.35 7.76

98.18 65.58

i 1.14 5.09

0.462 0.374

0.428 0,223

26/28

93.95

27 67

82.86

22.93

0.650

0,322

58/55 63/50 39/44

114.33 i 22.4 92.49

50.73 51.47 42.17

979 ., 5 108.25 100.58

49.67 55.72 42.41

0.170 0.066 0.708

0,206 0,258 0,650

.

.

0.i49

0.181

All except 198/189 Mid Cumbria 107.95

.

.

183.42

.

98.39

.

.

180.17

~Means that there are no cases for one of the two 5-year periods. The expected number for the period with cases ha~ been doubled.

26

J R. GOLDSMITH

TABLE 6 Mortality of cancer other than leukemia, ages 0-9 years, in O~e vicinity of British nuclear installations during 1971-1980 Name

Mortality

A. Pre-1955 installations Sellafield 4/3 Springfields 22/32 Capenhurst 24/27 Amersham 26/31 Aldermaston 33/19 Harwell 6/6 Winfrith 4/5 All except Sellafield ! 15/120

SMRr

EXPEC.r

SMRc

[E'(Mi)}

pr

pc

EXPEC.c

98.3 79.6 91.3 71A 154.97 85.9 69.4

4.07 27.64 26.30 36.44 21.29 6.98 5.76

74.97 99.97 103.5 75.97 112.37 89.0 82.9

3.05 27.63 27.22 27.68 23.93 6.22 4.78

0.580 0.881 0.699 0.970 0.011" 0.697 0.826

0.508 0.795 0.675 0.594 0.131 0.635 0.745

128.34

93.85

120.45

0.891

0.637

B. Nuclear power installations Bradwell 6/2 103.5 5.80 Berkeley-Oldbury 9/2 94.2 9.55 Hinkley 6/4 212.6 a 5.54 Trawsfynydd 1/1 271.5 a 0.74 Dungeness 0/0 . . . Wylfa 1/1 97.5 a 2.05 Sizewell 1/0 105.2 1.90 All CEGB 24/10 93.5 25.66

72.0 a 41.4 a 141.3 319.0 a . 131.5 a 0 41.2

2.09' 1.98 3.91 1.18

0.522 0.614 0.477 0.519

0.161 0.033* 0.376 0.789

-10.59

0.874 0.853 0.658

0.800 -0.0127"

78.27 74.97

4.63 3.05

0.078 0.580

0.120 0.509

102.06

16.42

0.194

0.266

80.06 121.59 103.72 96.87

22.50 29.04 20.23 15.56

0.255 0.677 0.745 0.731

0.093 0.851 0.706 0.613

97.96

107.35

0.315

0.302

93.5

C. Coastal sites N. Cumbria 10/5 168.92 5.92 Mid Cumbria 4/3 98.33 4.07 S Cumb. and N. Lane. 20120 1 2 4 . 3 4 16.09 S.Lanc. and N. Mersey 32/25 113.85 28.11 Liverpool 22/26 92.13 23.88 Wirral 17/21 87.17 19.5 N. Wales 14/15 87.14 16.07 All except 115/112 Mid Cumbria 104.94 i09.59

.

. 1.35

aMeans that there are no deaths for one of the two 5-year periods. The expected number for the period with deaths has been doubled.

of expected numbers (Table 3A). One hundred sixty-five cases were observed, compared to 133.6 expected based on the regional data, or 141.9 based on the comparison area data. This would indicate an increased relative risk of 23.5 or 16.3%, respectively. No significant excess in any group is found

27

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER iN THE UK

TABLE 7

Summary of results of study of 0-9-year-old children and cancer near British nuclear installations 1971-1980 Outcome

'z' for Regional

'z' for Comparison

pr

pc

2.164 -0.468 -0.599

0.005 0.588 0.755

0.015 0.682 0.730

1.216 =0.631 0.252

0.079 : f~.298 0.284

(~.I ~2 0.74! 0.401

0.265 i .224 0.912

0. i 32 0.9i4 0.149

0.395 0. i i0 . 0.181

-0.349 2.235 0.518

0.891 0.658 0.315

0.637 0.013 0.302

Leukemia incidence Pre-55 CEGB

Coastal

2.591 -0.146 -0.689

Leukemia mortality Pre-55 CEGB

Coastal

!.412 0.530 0.571

Other cancer incidence Pre-55 CEGB

Coastal

1.100 -1.368 !.039

Other cancer mortality Pre-55 CEGB

Coastal

- 1.230 -C.399 0.481

for leukemia mortality nor for ca~cer incidence other than leukemia. For the pre-1955 group, the pr for leukemia mortality is relatively low (0.079), and the pc is also low (0.112), (Table 4A). For individual locations, Amersham shows very significant excess leukemia incidence by both criteria (Table 3) (60/46.3, pr = 0.0026; 60/31.87, pc = 0.0016) and significant excess leukemia morality by the pc (comparison) criteria (Table 4) (35/17.2, pc = 0.0054). By regional criteria expected is 28.34, pc = 0.125. The elevated relative risk for leukemia incidence would then be 60/40.6 or 47.8% based on regional data, or 60/31.9, 88.3% on the basis of comparison data. For leukemia mortality in the vicinity of Amersham, Table 4 shows 35/28.34, 23.5% increased risk for regional data and 35/17.2, 103.5% increase for comparison data. Alderm~oston shows significantly low probability for leukemia incidence by the regional data criteria, (35/23.9, pr = 0. 020) but not by comparison. Aldermaston also has a significantly elevated incidence and mortality of non-leukemic cancer by regional criteria. Springfields, which had a substantially higher mortality to

28

J.R. GOLDSMITH

-2

It LEUK. INCID

LEUK.MORT

OTHERINOD

OTHERMORT

TYPE OF OUTOQMEj----PRE-lg68

~

POWER 8TA.

~

COASTAL

OOMPARED TO REGIONAL RATE8 Fig. 2. 'z' Score for excess childhood (0-9 years) cancer in local authority areas near various tyes of UK nuclear installations, 1971-1980. The expected values are based on regional data and statistical tests on Poisson statistics. The so-called 'Pre-1955' installations include nuclear weapons, nuclear research, and nuclear reproccssing installations. The so-called 'coastal" installations are not nuclear installations at all but sites on the coast of the Irish Sea which may have beet, me contaminated through sea water discharges.

incidence ratio for installation areas never achieves any significant deviation from expectation. One of the Coastal sites, Mid-Cumbria (which includes Sellafield) has significant excess incid~:nce by the pc criteria and nonsignificant excess by the pr criteria. The S. Lancashire and N. Mersey LAAs show significant excess leukemia mortality (Table 4C) (27/20.52, pr = 0.021). Since leukemia incidence is less than expected with both criteria for these LAAs, (29/29.81, pr = 0.583 and 29/34.79, pc = 0.778, in Table 3C and recalling the risk of thlse positives when multiple tests (N = 56) are perform~d, such a result for coastal sites is not outside the bounds of normal expectation. The apparent excess of non-leukemia cancer mortality for nuclear power installations as a group and for Berkeley-Oldbury (Table 6B) are based on scant data, for exar~Dle deaths only occurred in one of the 5-year periods in the comparison area, for Berkeley-Oldbury and in three of the other areas it only occurred in one of the 5-year groaps for installation areas. The wide variability both upward and downward of the SMR reflects this. If one were bound by a rule that data would be compiled only if at least one death occurred in each 5-year period, there would be no data for any site in this

29

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

2.6

Z

2.0 1.5

1.O o.6

IL |

o.o

/®i l

-0.5

-1.O

LEUK. INCID

LEUK.IVlORT

OTHERINCID

OTHERMORT

TYPE OF OUTOOME l

PRE-1055 ~

POWER STA.

~

OOASTAL

WITH REFERENCE TO ~4R-',R!SC÷,I AREAS Fig. 3. 'z' Score foe excesschildhood (0-9 years) cancer in local authority areas near various types of UK nuclear ins~,allations, 1971-1980. The expected values are based on comparison areas and statistical tests on Binomial statistics. TI~,¢so~aiicd 'Pr¢ 1955' installations include nuclear weapons, nuc|ear research, and nuclear reprocessing installations. The apparently significant excess for mortality from other types of cancer other than leukemia in the CEGB or power plant data set is based on small numbers and is, not consistent across sites.

Table. For each comparison based on regional data, (except Hinkley) there are more deaths expected than observed in this set. (That is the pr values are alll < 0.500.) The non-leukemia cancer mortality for all CEGB locations is 24, compared to 25.7 based on the regional data and 10.6 based on the compariso,a area data. Other cancer sites are examined from the microfiche [5]; two bone cancers in installation areas (none in comparison), 6 brain tumors, compared to 3 in comparison areas and 3 non-Hodgkin's lymphoma compared to 2 in the comparison areas are reported. The compafisgn area had two deaths from testieular cancer but none occurred in the installation areas. This leaves 13 cancer deaths at other sites than those tabulated in the installation areas, compalred to 3 in the comparison area. NO conclusion is justified on these findings; however, additional years of data are worth examining. CONCLUSIONS

Selecting an arbitrary cu~: off for 'LGw' SMRs or SIRs, there are !3 sites with SMR or SIR < 70, based on regional ~ata, compared to 17 < 70, based

30

J.R. GOLDSMITH

on comparison LAAs. Using data only for leukemia incidence and mortality, there are 4 SIRs and 5 SMRs < 70 based on regional data and only 3 SIRs and 3 SMRs based on comparison areas. Thus f--ore the point of view of leukemia, the hypothesis being tested in this report, the argument for ignoring the comparison areas because so frequently they gave low SIRs and SMRs loses its justification. It is in fact the small numbers on which many of the SIRs and SMRs are based which tends to make them, unstable and give both unusually high and unusually low values. The null hypothesis of no positive association for leukemia in the vicinity of nuclear installations should be rejected with respect to the set of Pre-55 installations, based on the consistent significant association with leukemia incidence and the positive findings (albeit.not conventionally significant) for mortality excess. The null hypothesis of no effect for leukemia is accepted for nuclear power installations and for coastal sites. Table 7 summarizes the results, and includes the 'z' score, which is the standard normal deviation, for each type of outcome for groups of sites. DISCUSSION

In interpreting these results, emphasis should be given to such consistency as there may be between the two different bases for derivation of expected numbers, to the consistency of incidence and mortality data and to the consistency between similar types of installations. Since among the Pre-1955 installations Amersham has the most consistent excess of leukemia both by incidence and mortality, and it is also the largest in terms of events and population there is a chance that the result of the group will be unduly influenced. The Fisher procedure is suited for that type of situation. When applied to the data in Table 3, the probability based on comparison data for the Pre-1955 group becomes 0.0105, rather than 0.015. Using these criteria there is thus no evidence for any excess leukemia risk in the vicinity of any of the CEGB installations, nor for all of them together. There is substantial evidence of excess leukemia incidence in the vicinity of Pre-1955 installations, and there is sufficient consistency so that the association may have causal relevance. Of all the installations, A,mersham International is the one with the greatest absolute excess. These conclusions differ from those of Cook-Mozaffari et al. [6] based on a multivariate logistic regression analysis and a larger area of presumed exposure. They found for a population group 0-24 years of age and for the period 1969-1978 a non-significant excess Relative Risk of mortality from leukemia of 1.15 for populations living near (at least 0.1% of the County District populations living with (10 miles) CEGB installations. For lymphoid leukemia, the R.R. was 1.20 (also non-significant). They found e:~:¢ess

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

31

re!~ive risks of mortality in populations living near 5 out of 7 CEGB ino sta~lations. Relative Risks were nearly identical for the Pre-1955 installations ~lthough in the latter group, significance was reached based on the larger populations and numbers of events. Ill the incidence study reported here, by neither of the criteria for expected numbers, was there any increase over expected for residents near CEGB plants. For leukemia mortality, by regional criteria there was a slight excess, but by the criteria of comparison LAAs there was a deficiency of deaths. Since the documented emissions from CEGB installations are low [1], the variety of technical procedures used at CEGB plants is small, and there is little requirement to handle radioactive materials frequently, it would seem that the risk of exposure ~br CEGB workers and local residents should be less than for the Pre-1955 group. The findings of this study seem to support that view, while the Cook-Mozaffari study does not. Finally, the positive findings, modest as they are, indicate that whatever the processes which initially lead to increased childhood leukemia among populations living in the vicinity of certain nuclear installations, seemed to continue during the period 1971-1980. Along with the data on th~ occurrence of excess cases near Dounreay during the same period, there is little room for complacency. The passage of time has not had a favourable overall effect on the associations with these processes. On the other hand, the relative risks are not very great. We do not yet know if the associations observed reflect a causal relationship between radiation exposure in the communities surrounding nuclear installations au~ ...I childhood leukemia. Case referc;nt studies now under way may shed some more light on this possibility. (See accompanying article for comment on Gardner's initial report.) ACKNOWLEDGMENT

Professor Geoffrey Howe, in reviewing the planned work, suggested the use of Binomial Distributions as a criterion for evaluation of the significance of the comparison area-installation area statistics. Professor Michael Quastel and Dr. Tsipporah Sorer made helpful criticisms. REFERENCES D. Black (Chairman), Investigation of the Possible Increased Incidence of Cancer in West Cmnbria. Report of An Independent Advisory Group. HMSO, London, 1984, 104 pp. M. Bobrow (Chairman), The implications of the new data on the releases, from Sellafield in the 1950s for the Conclusions of the Report of the Possible Increased Incidence of Cancer in West Cumbria. Committee on the Medical Aspec~s of Radiation in the Environment, COMARE First Report, HMSO, London, 1986, 4| pp.

32

J.R. GOLDSMITH

3 M. Bobrow (Ch~Jrman), Investigation of the possible increased incidence of leukemia in yoang people ,,ear the Dounreay Nuclear Establishment, Caithness, Scotland. COMARE gepol~ HMSO, London, 1988, 109 pp. 4 Bobrow (Chairman), Incidence of Childhood Cancer in the West Berkshire and North Hampshire Area, in which are situated the Atomic Weapons Research Establishment, Aldermaston and The Royal Ordinance Factory, Burghfield. COMARE, Third Report, HMSO, 1989. 5 P.J Cook-Mozaffari, F.L. Ashwood, T. Vincent D. Forman and M. Alderson, Cancer Incidence and Mortality in the Vicinity of Nuclear Installations, England and Wales, 1959-1980. Studies on Medical and Population Subjects No. 51, Office of Populations Censuses and Surveys, plus microfiche. HMSO, London, 1987, 279 pp. 6 D. Forman, P. Cook-Mozaffari, S. Darby, G. Davy, I. Straton, R. Doll and M. Pike, Cancer near nuclear installations. Nature, 329 (1987) 499-505. 7 P.J. Cook-Mozaffari, S.C. Darby, R. Doll, D. Forman, C. Hermon, M.C. Pike and T. Vincent, Geographical variation in mortality from leukemia and other cancers in England and Wales in relation to proximity to nuclear installations, 1969-1978. Br. J. Cancer, 59 (1989) 476-485. 8 E. Roman, V. Beral, L. Carpenter, A. Watson, C. Barton, H. Ryder and D.L. Aston, Childhood leukemia in the West Berkshire and Basingstoke and North Hampshire District Health Authorities in relation to nuclear establishments in the vicinity. Br. Med. J., 294 (1987) 597-602. 9 M.A. Heasman, I.W. Kemp, J.D. Urquart and R. Black, Childhood leukemia in Northem Scotland. Lancet, i (1986) 266. i0 J.R. Goldsmith, Usefulness of Epidemiological Data for Evaluation of Health Effects from Point Source Radiation: Epidemiological Aspects of Cancer Incidence and Mortality in the Vicinity of Nuclear Installations in the United Kingdom. A Draft Report prepared for the European Regional Office of the World Health Organization, Copenhagen. 11 A.J. Swerdlow, Cancer registration in England and Wales: some aspects relevant to interpretation of the data. J. R. Stat. Soc., Ser. A, 149 (1986) 146-60. 12 A.M. Stewart, J. Webb and D. Hewitt, Malignant disease in childhood and diagnostic radiation in utero. Lancet, ii (1956) 447. 13 M.J. Gardner, A.J. Hall, S. Downes ano J.D. Terrell, Follow-up study of children born elsewhere but attending schools in Seascale, West Cumbria (schools cohort). Br. Med. J., 295 (1987) 819-822. 14 M.J. Gardner, A.J. Hall, S. Downes and J.D. Terrell, Follow-up study ~,f ckfildren born to mothers resident in Seascale, West Cumbria (birth cohorti. B~:~Med. J., 29 ~ t~987) 82~-827. 15 R. Doll, The epidemiology of childhood leukemia. J. R. Stat. Soc., Ser. A, 152 (1989) 341-351. APPENDIX 1

Based on the Regional da.:a the Standard Mortality Ra;.io (SMR~ or Standardized Incidence Ratio, also called the Standardized Registcafion Ratio (SIR) are the ratios of observed numbers of deaths (or cases), Mi and Ci or Mrf and Crf (depending on whether installation area deaths and cases or,

NUCLEAR INSTALLATIONS AND CHILDHOOD CANCER IN THE UK

33

refereaCe* area deaths and cases are meant) ~o expected numbers of deaths or cases, E(Mi), E(C;), E(Mrt) and E(Crf). Since we expect that the same standard rates are uscd for both installation and comparison areas, and in standardization we adjust the data to a common age distribution pattern, we can represent these ratios as: SIRi = Ci/(Pi x Aci) SMRi = Mi/(Pi x Ami) and SIRrf = Crf/(Prf x Acfi') SMRrf = Mrf/(Prf x Amrf) where the products in the denominators are the age-adjusted expected numbers, and where, for predetermined time periods, Ci is the number of new cancer cases in installation LAAs, and Cff is the number of new cases in the reference (or comparison) LAAs, and Mi and Mrf are the comparable numbers of mortalities or deaths in the installation and reference LAAs. Pi is the set of age-specific populations at risk in the installation LAAs, and Prf is the set of age-specific populations at risk for the reference LAAs. Aci is the set of standard age-specific rates for cancer incidence for the installation LAAs, which are identical to Acrf, the standard age specific incidence rates for the reference LAAs. Ami is the set of standard age-specific rates for cancer mortality for the installation LAAs, which are identical to Amrf, the standard age-specific mortality rates for the reference LAAs. It follows that the RRc, the Risk Ratio for installation compared to reference LAAs for incident cases is: RRc = Ci/Crf x Prf/Pi, and further for mortality RRm = Mi/Mff x Prf/Pi Since by design the reference LAAs are intended to have about the same population as the installation LAAs, the ratios of Ci/Crf and Mi/Mrf contain most of the information about relatively high or relatively low rates of cancer, and Prf/Pi reflects the ratios of the populations at risk. The data available in microfiche [5] do not include the age-specific populations at risk, nor expected numbers. *R~.fcrenee area, with the designation rf is used instead of comparison area, in order to avoid ambiguous notation if the designatioiJ ~'a~. c

34

J.R. GOLDSMITH

Since a common set of age-specific rates are used tbr both installation and comparison areas, and population struc:.ures are standardized, the ratios of the expected numbers for incidence and mortality (Pi x Aci)/(Prf x Acrf) and (Pi x Ami)/(Prf x Amrf) reflect only the ratios of sizes of the standardized populations at risk, Pi/Prf. We can use the experience in the comparison LAAs to derive a second set of expected numbers for the installation areas, E' (Ci) and E'(Mi). , n u s-.- E ' ( C i )

~.

= ,".-,, ," ,

x

,v~

x

Aci)/(Prf × Acrfi)

and E'(Mi) = Mrf x ((Pi x Ami)/(Prf x Amrf)) Since we are not provided with the age-specific population data for the incidence and reference areas, we cannot directly calculate E'(Ci) and E'(Mi). What we can calculate is the expected incidence and expected numbers of deaths, based on Regional data for both Installation and Reference areas, and since these are based on die same Regional age-specific data and the age structures of the two areas, the ratios of E(Ci) to E(Crf) and of E(Mi) to E(Mrf) gives us an unbiassed estimate of the ratio in each equation. Hence: E'(Ci) = Crf x (E(Ci)/E(Crf)'~, and E'(Mi) = Mrf x (E(Mi)/E(M~:f)). APPENDIX 2

A. To obtain SMR or SIR for 1971-1980 from data for 1971-1975 and 1976-1980, say from microfiche [5] or Table 2. Divide the incidence or mortality for the first period by its SMR or SIR expressed as a ratio, and add to it the similar figure for the second period. This gives the expected number for the two periods combined. Divide this into the sum of the observed values to obtain the 10-year SMR or SIR. These figures are of course based on the regional age-specific incidence or mortality. Thus, for example: SMR = (MI + M2)/I(MI/[SMRI x 100]) + (M2/[SMR2 x 100])1, where 1 and 2 refer to data for two successive 5-year perio¢is, 1971-1975 and 1976-1980, A computer program written by Mrs S. Beeser was used to compute the

NUCLEAR iNSTALLATIONS AND CHILDHOOD CANCER IN THE UK

35

probabilities for the cumulative Poisson used for estimation of pr probabilities. B. For the Binomial, we test whether a value at least as great as the observed number of cases (O) or deaths in the installation area would be expected by chance, given the binomial expansion of the sum of observed numbers for installation and corr~parison areas, and the expansion is of a probability based on the ratio of the expected numbers Ei to the sum of Ei and E'rf. Thus p = Ei/(Ei + E"rO, q = 1 - p, and the binomial expansion.n..is of: (p + q) raised to the power (Oi + OrO. Values of the cumulative binomial probability are read from tables for (Oi + Orfo up to 20, in Burington and May, (Handbook of Probability and Statist~,:s with Tables, Handbook Pub. Sandusky, 1958) with interpolation for p. Above that the normal approximation is used. z = (x-p)/SQR tip x q/(Oi + Or01, where z is the standard normal deviate, p is defined above, and x = OV(Oi + Or0.

Nuclear installations and childhood cancer in the UK: mortality and incidence for 0-9-year-old children, 1971-1980.

Data on cancer incidence and mortality in the vicinity of nuclear installations in England and Wales have been published, covering the period 1959-198...
2MB Sizes 0 Downloads 0 Views