- Email: [email protected]
Childhood leukaemia mortality and population change in England and Wales 1969–1973
SIC. Sci. Med. Vol. 33, No. 4, pp. 435-W. 1991 Printed in Great Britain. All rights reserved
0277-9536191 63.00+ 0.00 Copyright 0 1991Pergamon Press plc
CHILDHOOD LEUKAEMIA MORTALITY AND POPULATION CHANGE IN ENGLAND AND WALES 1969-73 IAN LANGFORD School of Environmental
Sciences, University of East Anglia, Norwich NR4 7TJ, England
Abstract-Recent evidence from new towns in Great Britain suggests that childhood leukaemia mortality is associated with rapid population growth. It has been proposed that this may reflect patterns of
population mixing and exposure to infectious diseases which may predispose to the development of leukaemia in children. This study examines childhood leukaemia mortality for 1365 local authority areas of England and Wales for the period 1969 to 1973 with reference to population change between 1961 and 1971.-A significantly increased risk of childhood leukaemia mortality was found for O-14 year olds in areas which exnerienced more than a 50% increase in oooulation over the decade CR. R. I .408,95% C. I. 1.126-1.761). k cumulative sum analysis shows a threshold at approximately 50% population growth rather than a continuous relationship. A map of the data suggests that areas of significantly raised mortality compared to the national average are concentrated in and around the major conurbations of the area studied. Key worr&-childhood Wales
leukaemia, population change, viral infections, &eaves’ hypothesis, England and
INTRODUCTION
In recent times, there has been increasing interest in the hypothesis that some of the observed geographical distribution of childhood leukaemia may be explained by patterns of exposure to infectious diseases in the population. Possible links between infectious diseases and human leukaemia have been suggested since early this century [l], and there is an extensive literature on the role of viruses in causing leukaemia in the animal kingdom developed since the 1930s [2,3]. In humans, it is known that certain viruses play an important part in the development of lymphoid tumours, Epstein-Barr Virus being associated with Burkitt’s lymphoma [4] and HTLV-I with adult T cell leukaemias [5]. Kinlen (61 suggested that the excesses of childhood leukaemia observed around the Sellafield and Dounreay Nuclear Reprocessing Establishments may be associated with factors other than exposure to ionizing radiation from the facilities. Previous studies [7, 81 have failed to find a convincing connection between radioactive emissions from the reprocessing facilities and childhood leukaemia in respect of the quantity of radioactive substances discharged based on current dose-risk estimates [9, IO] and reliable pathways for the transfer of radioactive substances to humans [I 11. A recent study by Gardner et al. [12] found an elevated risk of leukaemia in children with fathers working at Sellafield and exposed to 10 mSv or more of radiation in the 6 months prior to conception. However, there is much doubt over the possible mechanisms by which paternal exposure prior to conception may influence leukaemia incidence in offspring, and the results 435
do not preclude the possibility that other factors including residential location and exposure to other leukaemogenic factors such as hydrocarbons may be of importance. Kinlen [6] specifically investigated the possibility that cases of childhood leukaemia around West Thurso, Scotland [13], an area surrounding the Dounreay Nuclear Reprocessing Establishment may be due to the unusual population mixing which occurred in settlements near the site. Dounreay was built in an isolated area, and experienced rapid population growth due to an influx of migrants finding employment at the plant. Postulating that leukaemia may result from a rare response to a widespread virus infection, Kinlen points out that herd immunity may be unusually low in an isolated area. Further, rapid inmigration would lead to a sudden rise in exposure to and dose received of a viral agent by populations not previously exposed to each other, and this might produce an increase in childhood leukaemia in the period following inmigration. Kinlen searched for other remote areas in Scotland which reported a similarly large increase in population to compare with West Thurso but without success. Indeed, there was only one rural local authority area that increased appreciably in population that was even moderately separated from a conurbation, namely Kircaldy D.C., an area which included the new town of Glenrothes. He found that during the period 1951 to 1967, when most of the population growth occurred, there were IO leukaemia deaths reported in the under 25 year age group compared to 3.60 expected (O/E ratio = 2.78, P < 0.001). However, in the period 1968 to 1985, when the population was relatively stable, there was
436
IAN
LANGFORD
a non-significant deficit in leukaemia cases, with 1 death being reported compared to 5.18 expected. The majority of the excess leukaemias in the earlier period were 6 leukaemias in the under 5 age group occurring between 1954 and 1959. Kinlen [14] then studied childhood leukaemia mortality in the &24 year age group in British new towns. He divided these into two categories, overspill and other, more rural new towns. The overspill new towns chosen by Kinlen were Basildon, Bracknell, Crawley, East Kilbride, Harlow, Hatfield, Hemel Hempstead, Stevenage and Welwyn Garden City. For rural new towns, he chose Glenrothes, Aycliffe, Corby, Cwmbran and Peterlee. All these new towns were designated by 1950, and experienced rapid population growth after this date. For the period 1946-65, he found a significant excess of leukaemia mortality in the rural new towns for children in the O-4 year age group (O/E = 2.75, P < O.OOOl), compared with a deficit in the 5-24 age group (O/E = 0.38, P < 0.05). There were nonsignificant deficits in both age groups for overspill new towns. For the whole time period studied (1946-1985) Kinlen found a significant deficit in the O-24 year age group in the overspill new towns (O/E = 0.83, P < 0.05) and a significant deficit in 5-24 year olds in the rural new towns (O/E = 0.47, P < 0.001). The excess of leukaemia mortality in O-4 year olds in the rural new towns remained significant over the whole time period (O/E = 1.99, P < 0.001). Kinlen also discovered that the density of pre-school children rose rapidly in all new towns in the 1950s but that in the rural new towns new residents had a 90% chance of originating from areas of lower population density. For overspill new towns, the reverse was true, with population density being less than that of the principal city of origin. New residents in rural new towns also came from more diverse places of origin. Kinlen concluded that these factors were consistent with the theory that childhood leukaemia is a rare response to a common but unrecognised infection. Increase in population density and diversity of contacts in rural new towns in the 1950s may have produced an epidemic of the hypothesised infection, leading to increased childhood leukaemia. However, this increased exposure may also have lead to immunizing doses being received by a large number of the population, hence the deficits of leukaemia in older age groups later on. Greaves [15] has hypothesised that childhood leukaemia, particularly the most prevalent form, (common acute lymphoblastic leukaemia or CALL), may be influenced in its development by the pattern of exposure to infections experienced by a child. Unusual patterns of exposure to a constellation of infections could be caused to a native population by a rapid influx of migrants to an area. This may cause proliferative stress in the immune system which in a number of cases could become leukaemia. Greaves [16] acknowledges that there are a wide range of known leukaemogenic factors (for a thorough review see Cartwright and Bernard [17]), but suggests these account for a minority of cases of CALL. He states that there may be a more common cause based on two independent spon-
taneous mutations, event one occurring in ufero or shortly afterwards in which a B stem cell is transformed into a pre-leukaemic clone. This clone will have increased transit time in the B cell compartment and a finite life span. A similar clinically silent mutation event has been demonstrated in a strain of mouse [18]. During the lifespan of the pre-leukaemic clone, the affected child is at risk of event two where the clone transforms into a leukaemic cell in differentiation arrest. Greaves argues that lymphocyte precursors are particularly prone to mutation, this being a statistical risk dependent on the rate of proliferation of cells. The chances of mutation, in particular event two, will therefore be modulated by immune response and patterns of exposure to disease in infancy. Greaves suggests that delayed exposure to infections, associated with improved living standards could cause the necessary proliferative stress and explain various observed features of childhood leukaemia, including the peak of CALL in advanced western societies [ 191. The emphasis here is not on one particular viral agent causing leukaemia, but on particular histories of exposure to a variety of infections. However, unusual patterns of infection could potentially be experienced by any child whose parents have migrated to, or lived within an area of unusual or extreme population mixing. Hence, the degree of population change may be more significant than the initial remoteness of an area. This paper attempts to address some of these considerations by studying a wide geographical area, allowing comparisons over a range of population changes in urban and rural areas, expanding the area of study examined by Kinlen. A unique opportunity for this study is provided by the availability of unpublished data for the 1969-73 Registrar Generals’ Decennial Supplement. This provides leukaemia mortality data for all local authority areas in England and Wales on the most spatially disaggregated base that is available from regularly collected vital statistics data. By limiting the period of the study to 5 years around the 1971 Census, population figures from this Census should be reasonably accurate even for areas which experienced rapid population growth or decline. METHODS
AND MATERIALS
Data on childhood leukaemia mortality for children under 15 years were collected from the Registrar Generals’ Decennial Supplement on Mortality (Urban and Rural Supplement) [20] which records leukaemia deaths by geographical units for the period 1969 to 1973. The geographical units are the pre-1974 local authority areas, comprising London Boroughs, County Boroughs, Municipal Boroughs, and Urban and rural Districts. Population figures were taken from the 1971 Census, and poplation change calculated as the percentage change between 1961 and 1971 for the entire population of each area. Poisson P-values were calculated for each area compared to the national average to produce a map of areas differing significantly from the national averaee _~~_~ at the 0.05 sienificance level (Fin. . _ 1).
437
Childhood leukaemia mortality and population change
Local with
Authority Poisson
significant
at
areas p-values 0.05
levd
%CW
Fig. I. Excesses of childhood leukaemia mortality in England and Wales 1969-73.
The data were aggregated into eight categories based on population change between 1961 and 1971, ranging from areas which showed a population decrease of more than 10% through 10% increments to areas with a population increase of greater than 50%. These categories are of course arbitrary, but were chosen on the basis of the distribution of population changes, to provide as much information from the data as possible without losing too much statistical power due to small numbers of leukaemia deaths in each category. The categories were also chosen before inspection of the mortality data. Confidence intervals were calculated using a standard statistical technique, namely Taylor Series Confidence Intervals [2 11. In addition to the use of these categories, leukaemia mortality and population were calculated as cumulative sums from the area of highest population change. This allowed calculation of a cumulative rate which was then divided by the national average rate for males and females combined in the O-14 age range (Fig. 2). This approach removes the problem of defining artificial boundaries, and allows examination of whether any relationship between population
change and childhood leukaemia is a continuous function or shows a threshold effect. RESULTS
Figure 1 displays graphically the 38 local authority areas in England and Wales which showed an excess in childhood leukaemia mortality (O-14 years) during the study period, significant at the 0.05 level, calculated from Poisson P-values. This number is no more than would be expected by chance, but the distribution of areas of significantly elevated leukaemia mortality may be of importance. For example, Heasman et al. [22] found that childhood leukaemia clustered no more than would be expected by chance in Scotland, but one of the three significant clusters observed occurred near Thurso. Although no formal statistical interpretation has been applied so far, it can be seen visually that the majority of individual areas which show a significant excess of leukaemia mortality are concentrated in and around the main conurbations of the country, from London and the South East through the Midlands to the North West. There is a notable absence of areas showing a
IAN LANGFORD
438
Table I. Leukaemia rates per million for local authorities in England and Wales (1969-73) O-4 year olds
5-14 year olds
Percentage population change 1961-71
Males
Females
Males
Females
Less than - 10%
37.79
22.91 n = 19
23.24 R =39 22.51 n = 124 24.68 n = 57 19.17 n = 57 28.41 n = 52 15.88 n = I4 21.06 n = I5 4021 n = 26 23.37 n = 436
- 10% to 0% 0% to 10% 10% to 20% 20% to 30% 30% to 40% 40% to 50% Greater than 50% Total
” =33 30.82 n =a9 33.00 n = 76 35.62 n = 58 31.95 n = 32 31.87 n = I6 31.99 n = I3 40.04 n = 16
25.78 n =71 29.59 n =65 24.70 ” = 38 30.69 n =29 23.01 n = II 28.59 n=ll 37.08 n = 14
37.01 R =65 28.56 n = I66 24.12 n = II3 31.71 n=lOO 28.58 n = 55 28.54 n = 27 29.32 n = 22 36.31 n =25
33.29 n = 333
27.15 n = 258
29.07 n = 573
n = no. of deaths.
Reh11”~
Ri*k
cornpars
10
NatiD”.,
A”arl~
Fig. 2. Relative accumulating
risk of childhood leukaemia mortality from area of greatest population increase (see text for full explanation).
significant excess in the South West, in rural Wales there are only Llanfyllin R.D. and Montgomery U.D., and there are very few areas represented in East Anglia, Cumbria and Northumberland. This result is perhaps more significant because many of the small urban and rural districts have very small numbers of children aged O-14 within their boundaries, and even one case of leukaemia mortality would be sufficient to generate a significant P-value. However, the majority of significant excesses are contained in urban and rural areas with larger populations in more densely populated parts of the country. There also appears to be some clustering of the areas of significant excess around London, Birmingham, the North West around the Mersey. Only six areas showed a signifi-
cant deficit of childhood leukaemia mortality. These were Cardiff, Wallasey, and four London Boroughs, Hackney, Newham, Hounslow and Haringey. Again, this would appear to be a non-random result. Table 1 shows the leukaemia mortality rates per million for male and female children in the O-4 year and 5-14 year age groups for the eight categories of population change. As can be seen, there is consistently higher mortality in all four categories for areas with population increases of greater than 50% compared to other areas, although these do not reach statistical significance relative to the aggregate of areas with population change of less than 50% increase. The only exception to this is for males in both age groups in areas which lost more than 10% of their population, where higher mortality rates are also seen. The data were then combined (Table 2) to increase the number of leukaemia deaths in each category and increase statistical power. From Table 2, it can be seen that there is a statistically significant excess of leukaemia mortality for females (R.R. = 1.622, 95% C.I. 1.179-2.232) and males and females combined (R.R. = 1.408, 95% C.I. 1.126-l .761), though not for males alone when comparing the leukaemia mortality rates for areas experiencing over 50% population increase to all other areas. The final column in Table 2 shows the same results omitting areas which contained Kinlens’ more remote new towns. There were eight of these districts which contained 10 leukaemia deaths, and none of the areas fell into the ~50% population increase category. As can be seen, the removal of these areas has hardly any influence on the results,
Table 2. Leukaemia rates per million and relative risk estimates for areas with greater than 50% population change compared to other areas
Population change > 50% Population change ~50% Relative risk (*95% CL)
Males &14yr
Females 0-14yr
37.69 ” =41 30.23 n = 865 I .247 (0.912, 1.705)
39.07 n==40 24.08 n = 654 1.622 (1.179, 2.232)
Total o-14yr 38.36 n =8l 27.24 R = 1519 1.408 (1.126. 1.761)
*Excluding urban and rural areas containing Kinlens’ more remote new towns.
Total’ 0-14yr 38.36 n =81 27.25 n = 1509 I .408 (1.126, 1.761)
Childhood leukaemia mortality and population change that the effect described by Kinlen is not confined to these places. From Table 1, it can also be seen that there are excess leukaemia rates for males in areas which lost 10% or more of their total population between 1961 and 1971. This elevated risk is significant for male leukaemias in the combined O-14 year old age group (R.R. = 1.250, 95% C.I. 1.014-1.542) relative to the aggregate of other areas. Figure 2 is a graphical representation of the relative risk of leukaemia mortality in which areas successively accumulated from the highest population changes compared to the average national rate. Hence, each point represents the relative risk of leukaemia mortality in areas with higher population change (above and including the point) to the national average. Numbers on the graph represent multiple points, as the graph is necessarily condensed to a lower resolution as it contains 1365 individual data points. At the top of the graph there is a great deal of variation, as there are only a few areas included in the higher population change set, but this quickly settles down to a fairly constant relative risk around 1.4:l until the population increase reaches 50%. After this, the relative risk drops fairly consistently to around 1: 1 between 50% and 30% population increase. As the rate is cumulative, the relative risk will tend towards one as population change falls, and therefore the graph does not emphasize the elevated risk of leukaemia mortality at the lower end of the distribution. demonstrating
DISCUSSION
The geographical distribution of areas of significant excesses and deficits in childhood leukaemia mortality across England and Wales for the study period would suggest that there is more variability in large urban centres and their surrounding areas than in more remote rural areas. Areas of increased incidence seem to occur most frequently around London, the Midlands and Merseyside. With reference to population change, it may be that large urban centres and their surrounding areas experience a greater degree of population mixing than other areas, and hence the chance of a child being exposed to increased or unusual patterns of infection is greater. Population change is not, of course, a measure of migration but simply of the net loss or gain of population over a period. Areas which experience a net loss of population may in actual fact experience a greater inflow of migrants than those areas which experience a net increase in population. This could certainly be true of large conurbations, where a larger sector of the population may be transient than in smaller urban areas or rural districts. There were certainly large population movements associated with inner city renewal which was at its height in the 1960s and early 1970s [23]. From the evidence in this study, it could be tentatively suggested that degree of population mixing has apparently more influence over childhood leukaemia mortality rates than remoteness of an area. With reference to Merseyside, a number of other urban and rural districts in Lancashire and Cheshire also showed non-significant excesses of childhood leukaemia mortality. Birch et al. [24] found an increasing incidence in childhood acute
439
lymphoblastic leukaemia in Greater Manchester and the county of Lancashire, based on data from the Manchester Children’s Tumour Registry in O-14 year olds between 1954 and 1977. There are a number of factors which must be born in mind with reference to the results of this study. Firstly, mortality data has been used, which may vary from incidence data due to geographical variations in survival rates. However, one would suspect that treatment of leukaemia may be more effective around major urban centres, with better diagnostic and treatment facilities leading to greater survival rates than in more remote areas [25], whereas more significant excesses of childhood leukaemia mortality are found around these urban centres. Survival rates during the period studied [26] also suggest that some deaths occurring in the 5-14 year old group will have been incident in the O-4 year old age group. Discrepancies may also arise in the diagnosis of leukaemia due to death from other causes before leukaemia is diagnosed. Kneale and Stewart [27] note that there is loss of immunological competence in cancers of the reticula-endothelial system, including leukaemia, before the disease is recognizable. Death may therefore result from infectious diseases or, in the case of leukaemia, cerebral haemorrhage rather than the cancer itself [28,29]. Since the 1950’s, however, childhood mortality from infections has fallen rapidly, and with improved immunological detection techniques it is unlikely that infection deaths significantly mask the incidence of leukaemia nowadays [30,31]. Even at the period covered in this study, a childhood death from infection would have been treated as reason for further investigation in itself, especially in older children. It is possible that there could be migration of parents with children suffering from childhood leukaemia towards the main hospitals that treat the disease, leading to over-representation of these areas because last place of residence is recorded on death certificates [32]. The results on population change suggest that there is a consistent and significant excess of childhood leukaemias in areas which experienced high population change in the period 1961 to 1971. There is no way of telling from the present data exactly when the population increase occurred within this decade, but it must be remembered that the majority of childhood leukaemias have a latent period of between 2 and 4 years, and so there will be a time lag between contraction of the disease and diagnosis. This time period may even be longer when mortality data are considered, and so population change in the mid to late 1960s is of direct relevance to the period of mortalities considered here. From Fig. 2, it may be suggested that the effects of population increase on childhood leukaemia mortality have a threshold value somewhere between 30% and 50%, due to the relatively constant relative risk shown down to 50% increase, and the decline to unity at around 30%. As mentioned above, it is more plausible that population mixing from migration is affecting occurrence of leukaemia, and this could explain the increase in areas which lost population, mainly concentrated around the inner cities. There is no explanation of why this increase is only seen in males, however. The present results tend to support Kinlens’ research on
440
IAN LANGFORD
new towns and are also consistent with Greaves hypothesis that childhood leukaemia is related to patterns of exposure to infection rather than a single viral agent. Acknowledgement-This research was supported by an Economic and Social Research Council Postgraduate Training Award.
REFERENCES 1 Ward G. The infective theory of acute leukaemia. Br. J. Child Dis. 14, 10, 1917. 2. Jarrett 0. Leukaemogenic viruses. In Leukaemia
3.
4.
5.
6.
7.
(Edited by Whittaker J. A. and Delamore I. W.), p. 24. Blackwell Scientific Publications, Oxford, 1987. BoraniC M. Exuerimental leukaemia. In Leukaemia (Edited by Whitiaker J. A. and Delamore I. W.), p. 64. Blackwell Scientific, Oxford, 1987. Epstein M. A., Morgan A. J., Finerty S. et al. Protection of cottontop tamarins against Epstein-Barr virusinduced malignant lymphoma by a prototype subunit vaccine. Nature 318, 287, 1985. Gallo R. C., Essex M. E. and Gross L. (editors). Human T-cell Leukaemia/Lymphoma Virus. Cold Spring Harbour Laboratory, New York, 1984. Kinlen L. T. Evidence for an infective cause of childhood leukaemia: comparison of a Scottish New Town with nuclear reprocessing sites in Britain. Lancer, 1323, 10th December 1988. Black D. Investigation of rhe Possible Increased Incidence of Cancer in West Cumbria: Report of the Independent Advisory Group. HMSO, London, 1984.
8. COMARE (Committee on Medical Aspects of Radiation in the Environment). Second Report. HMSO, London, 1988. 9. BIER (Committee on the Biological Effects of Ionising Radiations). The effects on populations of exposure to low levels of ionising radiation. National Academy Press, Washington, 1980. 10. UNSCEAR (United Nations Scientific Committee on the Effects of Atomic Radiation). Reporrs IOthe General Assembly. United Nations, New York, 1977, 1982, 1986. Il. Sumner D. Radiation Risks: an Evaluation. Tarragon Press, Glasgow, 1988. 12. Gardner M. J., Hall A. J., Snee M.P., Powell C. A., Downes S. and Terre11 J. D. Results of case-control study of leukaemia and lymphoma among young people near Sellafield nuclear plant in West Cumbria. Br. Med. 1. 300, 423, 1990.
13. Heasman M. A., Urquhart J. D., Black R. and Kemp I. W. Leukaemia in young persons in Scotland: a study of its geographical distribution and relationship to nuclear installations. Recognitions submitted to Dounreay Inquiry. Scottish Health Service Common Services Agency, Information Services Division, Edinburgh, 1986. 14. Kinlen L. T., Clarke K. and Hudson C. Evidence from population mixing in British New Towns 194685 of an infective basis for childhood leukaemia. Lancef 577, 8th September 1990.
15. Greaves M. F. Speculations on the cause of childhood acute lymphoblastic leukaemia. Leukaemia 2, 120, 1988. 16. Greaves M. F. Etiology of childhood acute lymphoblastic leukaemia: a soluble problem? In Acute Lymphoblasric Leukaemia (Edited by Gale R. P. and Hoeltzer D.) UCLA Symposia on Molecular and Cellular Biology New Series, Vol. 108. Wiley-Liss, New York, 1989. 17. Cartwright R. A. and Bernard S. M. Epidemiology. In Leukaemia (Edited by Whittaker J. A.-and Del&ore I. W.). D. 3. Blackwell. Oxford. 1987. 18 Langhon W. Y., Harris A. W., Cory S. and Adams J. M. The c-myc oncogene perturbs B lymphocyte development in Ep-myc transgenic mice. Cell 47, 11, 1986. 19. Parkin D. M., Stiller C. A., Draper G. J. and Bieber C. A. The international incidence of childhood cancer. Int. J. Cancer 42, 511, 1988. 20. Office of Population Censuses and Surveys. Area Mortality Decennial Supplement 1969-73, England and Wales: Urban Areas, Rural Areas Microfiches. HMSO. I ~
London, 1981. 21. Kleinbaum D. G., Kupper L. L. and Morgenstem
H.
Epidemiologic Research: Principles and Quantitative Methodr. Van Nostrand Reinhold, New York, 1982.
22. Heasman M. A., Urquhart J. D., Black R., Kemp I. W., Glass S. and Gray M. Leukaemia in young persons in Scotland: a study of its geographical distribution and relationship to nuclear installations. Hlfh Bull p. 147, May 1987. 23. Robson B. T. Those Inner Cities: Reconciling the Economic and Social Aims of Urban Policy. Clarendon Press, Oxford, 1988. 24. Birch J. M., Swindell R., Marsden H. B. and Morris Jones P. H. Childhood leukaemia in North West England 1954-77: epidemiology, incidence and survival. Br. J. Cancer 43, 324, 1980. 25. Swerdlow A. J. Cancer registration in England and Wales: some aspects relevant to interpretation of data. J. R. Sfafisf. A. 149, (2), 146, 1986. 26. Draper G. J. et al. Childhood Cancer in Britain: Incidence Survival and Mortality. Office of Population Censuses and Surveys Studies on Medical and Population Subjects No. 37. HMSO, London, 1982. 27. Kneale G. W. and Stewart A. M. Pre-cancers and liability to other diseases. Br. J. Cancer 37, 448, 1978. 28. Taylor D. Leukaemia: Towards Control. Office of Health Economics studies of current health problems No. 68. HMSO, London, 1980. 29. Greaves M. F. and Chan L. C. Is spontaneous mutation the major ‘cause’ of childhood acute lymphoblastic leukaemia? Br. J. Haematol. 64, I, 1986. 30. Greaves M. F. Subtypes of acute lymphoblastic leukaemia: implications for the pathogenesis and epidemiology of leukaemia. In Environmenfal Influences in rhe Parhogenesis
of Leukaemias
and Lymphomas
(Edited by Magrath I., O’Connor G. T. and Ramot B.), p. 129. Raven Press, New York, 1984. 31. Chessells J. M. The acute lymphoblastic leukaemias. In Leukaemia (Edited by Whittaker J. A. and Delamore I. W.). D. 331. Blackwell. Oxford. 1987. 32. Bentdam C. G. Migration and morbidity: implications for geographical studies of disease. Sot. Sci. Med. 26, 49, 1988.
0277-9536191 63.00+ 0.00 Copyright 0 1991Pergamon Press plc
CHILDHOOD LEUKAEMIA MORTALITY AND POPULATION CHANGE IN ENGLAND AND WALES 1969-73 IAN LANGFORD School of Environmental
Sciences, University of East Anglia, Norwich NR4 7TJ, England
Abstract-Recent evidence from new towns in Great Britain suggests that childhood leukaemia mortality is associated with rapid population growth. It has been proposed that this may reflect patterns of
population mixing and exposure to infectious diseases which may predispose to the development of leukaemia in children. This study examines childhood leukaemia mortality for 1365 local authority areas of England and Wales for the period 1969 to 1973 with reference to population change between 1961 and 1971.-A significantly increased risk of childhood leukaemia mortality was found for O-14 year olds in areas which exnerienced more than a 50% increase in oooulation over the decade CR. R. I .408,95% C. I. 1.126-1.761). k cumulative sum analysis shows a threshold at approximately 50% population growth rather than a continuous relationship. A map of the data suggests that areas of significantly raised mortality compared to the national average are concentrated in and around the major conurbations of the area studied. Key worr&-childhood Wales
leukaemia, population change, viral infections, &eaves’ hypothesis, England and
INTRODUCTION
In recent times, there has been increasing interest in the hypothesis that some of the observed geographical distribution of childhood leukaemia may be explained by patterns of exposure to infectious diseases in the population. Possible links between infectious diseases and human leukaemia have been suggested since early this century [l], and there is an extensive literature on the role of viruses in causing leukaemia in the animal kingdom developed since the 1930s [2,3]. In humans, it is known that certain viruses play an important part in the development of lymphoid tumours, Epstein-Barr Virus being associated with Burkitt’s lymphoma [4] and HTLV-I with adult T cell leukaemias [5]. Kinlen (61 suggested that the excesses of childhood leukaemia observed around the Sellafield and Dounreay Nuclear Reprocessing Establishments may be associated with factors other than exposure to ionizing radiation from the facilities. Previous studies [7, 81 have failed to find a convincing connection between radioactive emissions from the reprocessing facilities and childhood leukaemia in respect of the quantity of radioactive substances discharged based on current dose-risk estimates [9, IO] and reliable pathways for the transfer of radioactive substances to humans [I 11. A recent study by Gardner et al. [12] found an elevated risk of leukaemia in children with fathers working at Sellafield and exposed to 10 mSv or more of radiation in the 6 months prior to conception. However, there is much doubt over the possible mechanisms by which paternal exposure prior to conception may influence leukaemia incidence in offspring, and the results 435
do not preclude the possibility that other factors including residential location and exposure to other leukaemogenic factors such as hydrocarbons may be of importance. Kinlen [6] specifically investigated the possibility that cases of childhood leukaemia around West Thurso, Scotland [13], an area surrounding the Dounreay Nuclear Reprocessing Establishment may be due to the unusual population mixing which occurred in settlements near the site. Dounreay was built in an isolated area, and experienced rapid population growth due to an influx of migrants finding employment at the plant. Postulating that leukaemia may result from a rare response to a widespread virus infection, Kinlen points out that herd immunity may be unusually low in an isolated area. Further, rapid inmigration would lead to a sudden rise in exposure to and dose received of a viral agent by populations not previously exposed to each other, and this might produce an increase in childhood leukaemia in the period following inmigration. Kinlen searched for other remote areas in Scotland which reported a similarly large increase in population to compare with West Thurso but without success. Indeed, there was only one rural local authority area that increased appreciably in population that was even moderately separated from a conurbation, namely Kircaldy D.C., an area which included the new town of Glenrothes. He found that during the period 1951 to 1967, when most of the population growth occurred, there were IO leukaemia deaths reported in the under 25 year age group compared to 3.60 expected (O/E ratio = 2.78, P < 0.001). However, in the period 1968 to 1985, when the population was relatively stable, there was
436
IAN
LANGFORD
a non-significant deficit in leukaemia cases, with 1 death being reported compared to 5.18 expected. The majority of the excess leukaemias in the earlier period were 6 leukaemias in the under 5 age group occurring between 1954 and 1959. Kinlen [14] then studied childhood leukaemia mortality in the &24 year age group in British new towns. He divided these into two categories, overspill and other, more rural new towns. The overspill new towns chosen by Kinlen were Basildon, Bracknell, Crawley, East Kilbride, Harlow, Hatfield, Hemel Hempstead, Stevenage and Welwyn Garden City. For rural new towns, he chose Glenrothes, Aycliffe, Corby, Cwmbran and Peterlee. All these new towns were designated by 1950, and experienced rapid population growth after this date. For the period 1946-65, he found a significant excess of leukaemia mortality in the rural new towns for children in the O-4 year age group (O/E = 2.75, P < O.OOOl), compared with a deficit in the 5-24 age group (O/E = 0.38, P < 0.05). There were nonsignificant deficits in both age groups for overspill new towns. For the whole time period studied (1946-1985) Kinlen found a significant deficit in the O-24 year age group in the overspill new towns (O/E = 0.83, P < 0.05) and a significant deficit in 5-24 year olds in the rural new towns (O/E = 0.47, P < 0.001). The excess of leukaemia mortality in O-4 year olds in the rural new towns remained significant over the whole time period (O/E = 1.99, P < 0.001). Kinlen also discovered that the density of pre-school children rose rapidly in all new towns in the 1950s but that in the rural new towns new residents had a 90% chance of originating from areas of lower population density. For overspill new towns, the reverse was true, with population density being less than that of the principal city of origin. New residents in rural new towns also came from more diverse places of origin. Kinlen concluded that these factors were consistent with the theory that childhood leukaemia is a rare response to a common but unrecognised infection. Increase in population density and diversity of contacts in rural new towns in the 1950s may have produced an epidemic of the hypothesised infection, leading to increased childhood leukaemia. However, this increased exposure may also have lead to immunizing doses being received by a large number of the population, hence the deficits of leukaemia in older age groups later on. Greaves [15] has hypothesised that childhood leukaemia, particularly the most prevalent form, (common acute lymphoblastic leukaemia or CALL), may be influenced in its development by the pattern of exposure to infections experienced by a child. Unusual patterns of exposure to a constellation of infections could be caused to a native population by a rapid influx of migrants to an area. This may cause proliferative stress in the immune system which in a number of cases could become leukaemia. Greaves [16] acknowledges that there are a wide range of known leukaemogenic factors (for a thorough review see Cartwright and Bernard [17]), but suggests these account for a minority of cases of CALL. He states that there may be a more common cause based on two independent spon-
taneous mutations, event one occurring in ufero or shortly afterwards in which a B stem cell is transformed into a pre-leukaemic clone. This clone will have increased transit time in the B cell compartment and a finite life span. A similar clinically silent mutation event has been demonstrated in a strain of mouse [18]. During the lifespan of the pre-leukaemic clone, the affected child is at risk of event two where the clone transforms into a leukaemic cell in differentiation arrest. Greaves argues that lymphocyte precursors are particularly prone to mutation, this being a statistical risk dependent on the rate of proliferation of cells. The chances of mutation, in particular event two, will therefore be modulated by immune response and patterns of exposure to disease in infancy. Greaves suggests that delayed exposure to infections, associated with improved living standards could cause the necessary proliferative stress and explain various observed features of childhood leukaemia, including the peak of CALL in advanced western societies [ 191. The emphasis here is not on one particular viral agent causing leukaemia, but on particular histories of exposure to a variety of infections. However, unusual patterns of infection could potentially be experienced by any child whose parents have migrated to, or lived within an area of unusual or extreme population mixing. Hence, the degree of population change may be more significant than the initial remoteness of an area. This paper attempts to address some of these considerations by studying a wide geographical area, allowing comparisons over a range of population changes in urban and rural areas, expanding the area of study examined by Kinlen. A unique opportunity for this study is provided by the availability of unpublished data for the 1969-73 Registrar Generals’ Decennial Supplement. This provides leukaemia mortality data for all local authority areas in England and Wales on the most spatially disaggregated base that is available from regularly collected vital statistics data. By limiting the period of the study to 5 years around the 1971 Census, population figures from this Census should be reasonably accurate even for areas which experienced rapid population growth or decline. METHODS
AND MATERIALS
Data on childhood leukaemia mortality for children under 15 years were collected from the Registrar Generals’ Decennial Supplement on Mortality (Urban and Rural Supplement) [20] which records leukaemia deaths by geographical units for the period 1969 to 1973. The geographical units are the pre-1974 local authority areas, comprising London Boroughs, County Boroughs, Municipal Boroughs, and Urban and rural Districts. Population figures were taken from the 1971 Census, and poplation change calculated as the percentage change between 1961 and 1971 for the entire population of each area. Poisson P-values were calculated for each area compared to the national average to produce a map of areas differing significantly from the national averaee _~~_~ at the 0.05 sienificance level (Fin. . _ 1).
437
Childhood leukaemia mortality and population change
Local with
Authority Poisson
significant
at
areas p-values 0.05
levd
%CW
Fig. I. Excesses of childhood leukaemia mortality in England and Wales 1969-73.
The data were aggregated into eight categories based on population change between 1961 and 1971, ranging from areas which showed a population decrease of more than 10% through 10% increments to areas with a population increase of greater than 50%. These categories are of course arbitrary, but were chosen on the basis of the distribution of population changes, to provide as much information from the data as possible without losing too much statistical power due to small numbers of leukaemia deaths in each category. The categories were also chosen before inspection of the mortality data. Confidence intervals were calculated using a standard statistical technique, namely Taylor Series Confidence Intervals [2 11. In addition to the use of these categories, leukaemia mortality and population were calculated as cumulative sums from the area of highest population change. This allowed calculation of a cumulative rate which was then divided by the national average rate for males and females combined in the O-14 age range (Fig. 2). This approach removes the problem of defining artificial boundaries, and allows examination of whether any relationship between population
change and childhood leukaemia is a continuous function or shows a threshold effect. RESULTS
Figure 1 displays graphically the 38 local authority areas in England and Wales which showed an excess in childhood leukaemia mortality (O-14 years) during the study period, significant at the 0.05 level, calculated from Poisson P-values. This number is no more than would be expected by chance, but the distribution of areas of significantly elevated leukaemia mortality may be of importance. For example, Heasman et al. [22] found that childhood leukaemia clustered no more than would be expected by chance in Scotland, but one of the three significant clusters observed occurred near Thurso. Although no formal statistical interpretation has been applied so far, it can be seen visually that the majority of individual areas which show a significant excess of leukaemia mortality are concentrated in and around the main conurbations of the country, from London and the South East through the Midlands to the North West. There is a notable absence of areas showing a
IAN LANGFORD
438
Table I. Leukaemia rates per million for local authorities in England and Wales (1969-73) O-4 year olds
5-14 year olds
Percentage population change 1961-71
Males
Females
Males
Females
Less than - 10%
37.79
22.91 n = 19
23.24 R =39 22.51 n = 124 24.68 n = 57 19.17 n = 57 28.41 n = 52 15.88 n = I4 21.06 n = I5 4021 n = 26 23.37 n = 436
- 10% to 0% 0% to 10% 10% to 20% 20% to 30% 30% to 40% 40% to 50% Greater than 50% Total
” =33 30.82 n =a9 33.00 n = 76 35.62 n = 58 31.95 n = 32 31.87 n = I6 31.99 n = I3 40.04 n = 16
25.78 n =71 29.59 n =65 24.70 ” = 38 30.69 n =29 23.01 n = II 28.59 n=ll 37.08 n = 14
37.01 R =65 28.56 n = I66 24.12 n = II3 31.71 n=lOO 28.58 n = 55 28.54 n = 27 29.32 n = 22 36.31 n =25
33.29 n = 333
27.15 n = 258
29.07 n = 573
n = no. of deaths.
Reh11”~
Ri*k
cornpars
10
NatiD”.,
A”arl~
Fig. 2. Relative accumulating
risk of childhood leukaemia mortality from area of greatest population increase (see text for full explanation).
significant excess in the South West, in rural Wales there are only Llanfyllin R.D. and Montgomery U.D., and there are very few areas represented in East Anglia, Cumbria and Northumberland. This result is perhaps more significant because many of the small urban and rural districts have very small numbers of children aged O-14 within their boundaries, and even one case of leukaemia mortality would be sufficient to generate a significant P-value. However, the majority of significant excesses are contained in urban and rural areas with larger populations in more densely populated parts of the country. There also appears to be some clustering of the areas of significant excess around London, Birmingham, the North West around the Mersey. Only six areas showed a signifi-
cant deficit of childhood leukaemia mortality. These were Cardiff, Wallasey, and four London Boroughs, Hackney, Newham, Hounslow and Haringey. Again, this would appear to be a non-random result. Table 1 shows the leukaemia mortality rates per million for male and female children in the O-4 year and 5-14 year age groups for the eight categories of population change. As can be seen, there is consistently higher mortality in all four categories for areas with population increases of greater than 50% compared to other areas, although these do not reach statistical significance relative to the aggregate of areas with population change of less than 50% increase. The only exception to this is for males in both age groups in areas which lost more than 10% of their population, where higher mortality rates are also seen. The data were then combined (Table 2) to increase the number of leukaemia deaths in each category and increase statistical power. From Table 2, it can be seen that there is a statistically significant excess of leukaemia mortality for females (R.R. = 1.622, 95% C.I. 1.179-2.232) and males and females combined (R.R. = 1.408, 95% C.I. 1.126-l .761), though not for males alone when comparing the leukaemia mortality rates for areas experiencing over 50% population increase to all other areas. The final column in Table 2 shows the same results omitting areas which contained Kinlens’ more remote new towns. There were eight of these districts which contained 10 leukaemia deaths, and none of the areas fell into the ~50% population increase category. As can be seen, the removal of these areas has hardly any influence on the results,
Table 2. Leukaemia rates per million and relative risk estimates for areas with greater than 50% population change compared to other areas
Population change > 50% Population change ~50% Relative risk (*95% CL)
Males &14yr
Females 0-14yr
37.69 ” =41 30.23 n = 865 I .247 (0.912, 1.705)
39.07 n==40 24.08 n = 654 1.622 (1.179, 2.232)
Total o-14yr 38.36 n =8l 27.24 R = 1519 1.408 (1.126. 1.761)
*Excluding urban and rural areas containing Kinlens’ more remote new towns.
Total’ 0-14yr 38.36 n =81 27.25 n = 1509 I .408 (1.126, 1.761)
Childhood leukaemia mortality and population change that the effect described by Kinlen is not confined to these places. From Table 1, it can also be seen that there are excess leukaemia rates for males in areas which lost 10% or more of their total population between 1961 and 1971. This elevated risk is significant for male leukaemias in the combined O-14 year old age group (R.R. = 1.250, 95% C.I. 1.014-1.542) relative to the aggregate of other areas. Figure 2 is a graphical representation of the relative risk of leukaemia mortality in which areas successively accumulated from the highest population changes compared to the average national rate. Hence, each point represents the relative risk of leukaemia mortality in areas with higher population change (above and including the point) to the national average. Numbers on the graph represent multiple points, as the graph is necessarily condensed to a lower resolution as it contains 1365 individual data points. At the top of the graph there is a great deal of variation, as there are only a few areas included in the higher population change set, but this quickly settles down to a fairly constant relative risk around 1.4:l until the population increase reaches 50%. After this, the relative risk drops fairly consistently to around 1: 1 between 50% and 30% population increase. As the rate is cumulative, the relative risk will tend towards one as population change falls, and therefore the graph does not emphasize the elevated risk of leukaemia mortality at the lower end of the distribution. demonstrating
DISCUSSION
The geographical distribution of areas of significant excesses and deficits in childhood leukaemia mortality across England and Wales for the study period would suggest that there is more variability in large urban centres and their surrounding areas than in more remote rural areas. Areas of increased incidence seem to occur most frequently around London, the Midlands and Merseyside. With reference to population change, it may be that large urban centres and their surrounding areas experience a greater degree of population mixing than other areas, and hence the chance of a child being exposed to increased or unusual patterns of infection is greater. Population change is not, of course, a measure of migration but simply of the net loss or gain of population over a period. Areas which experience a net loss of population may in actual fact experience a greater inflow of migrants than those areas which experience a net increase in population. This could certainly be true of large conurbations, where a larger sector of the population may be transient than in smaller urban areas or rural districts. There were certainly large population movements associated with inner city renewal which was at its height in the 1960s and early 1970s [23]. From the evidence in this study, it could be tentatively suggested that degree of population mixing has apparently more influence over childhood leukaemia mortality rates than remoteness of an area. With reference to Merseyside, a number of other urban and rural districts in Lancashire and Cheshire also showed non-significant excesses of childhood leukaemia mortality. Birch et al. [24] found an increasing incidence in childhood acute
439
lymphoblastic leukaemia in Greater Manchester and the county of Lancashire, based on data from the Manchester Children’s Tumour Registry in O-14 year olds between 1954 and 1977. There are a number of factors which must be born in mind with reference to the results of this study. Firstly, mortality data has been used, which may vary from incidence data due to geographical variations in survival rates. However, one would suspect that treatment of leukaemia may be more effective around major urban centres, with better diagnostic and treatment facilities leading to greater survival rates than in more remote areas [25], whereas more significant excesses of childhood leukaemia mortality are found around these urban centres. Survival rates during the period studied [26] also suggest that some deaths occurring in the 5-14 year old group will have been incident in the O-4 year old age group. Discrepancies may also arise in the diagnosis of leukaemia due to death from other causes before leukaemia is diagnosed. Kneale and Stewart [27] note that there is loss of immunological competence in cancers of the reticula-endothelial system, including leukaemia, before the disease is recognizable. Death may therefore result from infectious diseases or, in the case of leukaemia, cerebral haemorrhage rather than the cancer itself [28,29]. Since the 1950’s, however, childhood mortality from infections has fallen rapidly, and with improved immunological detection techniques it is unlikely that infection deaths significantly mask the incidence of leukaemia nowadays [30,31]. Even at the period covered in this study, a childhood death from infection would have been treated as reason for further investigation in itself, especially in older children. It is possible that there could be migration of parents with children suffering from childhood leukaemia towards the main hospitals that treat the disease, leading to over-representation of these areas because last place of residence is recorded on death certificates [32]. The results on population change suggest that there is a consistent and significant excess of childhood leukaemias in areas which experienced high population change in the period 1961 to 1971. There is no way of telling from the present data exactly when the population increase occurred within this decade, but it must be remembered that the majority of childhood leukaemias have a latent period of between 2 and 4 years, and so there will be a time lag between contraction of the disease and diagnosis. This time period may even be longer when mortality data are considered, and so population change in the mid to late 1960s is of direct relevance to the period of mortalities considered here. From Fig. 2, it may be suggested that the effects of population increase on childhood leukaemia mortality have a threshold value somewhere between 30% and 50%, due to the relatively constant relative risk shown down to 50% increase, and the decline to unity at around 30%. As mentioned above, it is more plausible that population mixing from migration is affecting occurrence of leukaemia, and this could explain the increase in areas which lost population, mainly concentrated around the inner cities. There is no explanation of why this increase is only seen in males, however. The present results tend to support Kinlens’ research on
440
IAN LANGFORD
new towns and are also consistent with Greaves hypothesis that childhood leukaemia is related to patterns of exposure to infection rather than a single viral agent. Acknowledgement-This research was supported by an Economic and Social Research Council Postgraduate Training Award.
REFERENCES 1 Ward G. The infective theory of acute leukaemia. Br. J. Child Dis. 14, 10, 1917. 2. Jarrett 0. Leukaemogenic viruses. In Leukaemia
3.
4.
5.
6.
7.
(Edited by Whittaker J. A. and Delamore I. W.), p. 24. Blackwell Scientific Publications, Oxford, 1987. BoraniC M. Exuerimental leukaemia. In Leukaemia (Edited by Whitiaker J. A. and Delamore I. W.), p. 64. Blackwell Scientific, Oxford, 1987. Epstein M. A., Morgan A. J., Finerty S. et al. Protection of cottontop tamarins against Epstein-Barr virusinduced malignant lymphoma by a prototype subunit vaccine. Nature 318, 287, 1985. Gallo R. C., Essex M. E. and Gross L. (editors). Human T-cell Leukaemia/Lymphoma Virus. Cold Spring Harbour Laboratory, New York, 1984. Kinlen L. T. Evidence for an infective cause of childhood leukaemia: comparison of a Scottish New Town with nuclear reprocessing sites in Britain. Lancer, 1323, 10th December 1988. Black D. Investigation of rhe Possible Increased Incidence of Cancer in West Cumbria: Report of the Independent Advisory Group. HMSO, London, 1984.
8. COMARE (Committee on Medical Aspects of Radiation in the Environment). Second Report. HMSO, London, 1988. 9. BIER (Committee on the Biological Effects of Ionising Radiations). The effects on populations of exposure to low levels of ionising radiation. National Academy Press, Washington, 1980. 10. UNSCEAR (United Nations Scientific Committee on the Effects of Atomic Radiation). Reporrs IOthe General Assembly. United Nations, New York, 1977, 1982, 1986. Il. Sumner D. Radiation Risks: an Evaluation. Tarragon Press, Glasgow, 1988. 12. Gardner M. J., Hall A. J., Snee M.P., Powell C. A., Downes S. and Terre11 J. D. Results of case-control study of leukaemia and lymphoma among young people near Sellafield nuclear plant in West Cumbria. Br. Med. 1. 300, 423, 1990.
13. Heasman M. A., Urquhart J. D., Black R. and Kemp I. W. Leukaemia in young persons in Scotland: a study of its geographical distribution and relationship to nuclear installations. Recognitions submitted to Dounreay Inquiry. Scottish Health Service Common Services Agency, Information Services Division, Edinburgh, 1986. 14. Kinlen L. T., Clarke K. and Hudson C. Evidence from population mixing in British New Towns 194685 of an infective basis for childhood leukaemia. Lancef 577, 8th September 1990.
15. Greaves M. F. Speculations on the cause of childhood acute lymphoblastic leukaemia. Leukaemia 2, 120, 1988. 16. Greaves M. F. Etiology of childhood acute lymphoblastic leukaemia: a soluble problem? In Acute Lymphoblasric Leukaemia (Edited by Gale R. P. and Hoeltzer D.) UCLA Symposia on Molecular and Cellular Biology New Series, Vol. 108. Wiley-Liss, New York, 1989. 17. Cartwright R. A. and Bernard S. M. Epidemiology. In Leukaemia (Edited by Whittaker J. A.-and Del&ore I. W.). D. 3. Blackwell. Oxford. 1987. 18 Langhon W. Y., Harris A. W., Cory S. and Adams J. M. The c-myc oncogene perturbs B lymphocyte development in Ep-myc transgenic mice. Cell 47, 11, 1986. 19. Parkin D. M., Stiller C. A., Draper G. J. and Bieber C. A. The international incidence of childhood cancer. Int. J. Cancer 42, 511, 1988. 20. Office of Population Censuses and Surveys. Area Mortality Decennial Supplement 1969-73, England and Wales: Urban Areas, Rural Areas Microfiches. HMSO. I ~
London, 1981. 21. Kleinbaum D. G., Kupper L. L. and Morgenstem
H.
Epidemiologic Research: Principles and Quantitative Methodr. Van Nostrand Reinhold, New York, 1982.
22. Heasman M. A., Urquhart J. D., Black R., Kemp I. W., Glass S. and Gray M. Leukaemia in young persons in Scotland: a study of its geographical distribution and relationship to nuclear installations. Hlfh Bull p. 147, May 1987. 23. Robson B. T. Those Inner Cities: Reconciling the Economic and Social Aims of Urban Policy. Clarendon Press, Oxford, 1988. 24. Birch J. M., Swindell R., Marsden H. B. and Morris Jones P. H. Childhood leukaemia in North West England 1954-77: epidemiology, incidence and survival. Br. J. Cancer 43, 324, 1980. 25. Swerdlow A. J. Cancer registration in England and Wales: some aspects relevant to interpretation of data. J. R. Sfafisf. A. 149, (2), 146, 1986. 26. Draper G. J. et al. Childhood Cancer in Britain: Incidence Survival and Mortality. Office of Population Censuses and Surveys Studies on Medical and Population Subjects No. 37. HMSO, London, 1982. 27. Kneale G. W. and Stewart A. M. Pre-cancers and liability to other diseases. Br. J. Cancer 37, 448, 1978. 28. Taylor D. Leukaemia: Towards Control. Office of Health Economics studies of current health problems No. 68. HMSO, London, 1980. 29. Greaves M. F. and Chan L. C. Is spontaneous mutation the major ‘cause’ of childhood acute lymphoblastic leukaemia? Br. J. Haematol. 64, I, 1986. 30. Greaves M. F. Subtypes of acute lymphoblastic leukaemia: implications for the pathogenesis and epidemiology of leukaemia. In Environmenfal Influences in rhe Parhogenesis
of Leukaemias
and Lymphomas
(Edited by Magrath I., O’Connor G. T. and Ramot B.), p. 129. Raven Press, New York, 1984. 31. Chessells J. M. The acute lymphoblastic leukaemias. In Leukaemia (Edited by Whittaker J. A. and Delamore I. W.). D. 331. Blackwell. Oxford. 1987. 32. Bentdam C. G. Migration and morbidity: implications for geographical studies of disease. Sot. Sci. Med. 26, 49, 1988.