The village of Seascale is adjacent to the Sellafield nuclear reprocessing plant in west Cumbria, northwest England. Since 1950, the incidence of leukaemia/non-Hodgkin's lymphoma in children in Seascale has been over 10 times that expected from national rates.1 Concern that the excess could be caused by discharges of radioactive material from Sellafield has not been supported by detailed radiologic studies.2, 3 In 1990, Gardner et al.4 reported the results of a case-control study of leukaemias and lymphomas in young people born and diagnosed in west Cumbria. They concluded that fathers' exposure to external ionising radiation while employed at Sellafield was a risk factor for leukaemia/non-Hodgkin's lymphoma in children subsequently conceived and that this effect could explain the Seascale excess. Although subsequent epidemiologic research in humans did not confirm the findings of Gardner et al.,5, 6, 7 experimental studies in animals showed that paternal preconceptional irradiation (ppi) could increase the risk of cancer in offspring8, 9 and genetic studies in humans have suggested that parental preconceptional exposures may increase the risk.10 Alternatively, it has been suggested that the findings of Gardner et al. may have been due to chance or confounding with internal radionuclides4, 11 and that the excess in Seascale may have been due to a high level of population mixing.12, 13, 14
Our aim was to evaluate the evidence for an association between ppi and leukaemia/non-Hodgkin's lymphoma among children of Sellafield radiation workers, using a robust cohort study design, extending the temporal and geographic boundaries beyond those of Gardner et al.4 and allowing for the possible effects of population mixing and other demographic variables. In contrast to the study of Gardner et al.,4 the present study included all children born during 1950–1991 to mothers living in Cumbria, considering their risk of leukaemia/non-Hodgkin's lymphoma in relation to all monitored radiation exposures, both external and internal, of fathers working at Sellafield. Cases diagnosed throughout the United Kingdom were ascertained. The objectives were (i) to compare the incidence of leukaemia/non-Hodgkin's lymphoma among children whose fathers were Sellafield radiation workers to that among children whose fathers had never worked at Sellafield, both with and without adjustment for demographic characteristics, and (ii) to investigate whether ppi (either external or internal) was associated with an increased incidence of leukaemia/non-Hodgkin's lymphoma and to determine whether any such association could be explained by demographic characteristics and was confined to children born in Seascale.
MATERIAL AND METHODS
The present study was based on a database comprising registration details of all 274,170 live births to mothers domiciled in Cumbria between 1 January 1950 and 31 December 1991.15, 16 Fathers' occupations recorded at birth were assigned a social class. Parents' names were used to identify siblings, birth order and multiple births. Mother's addresses were grid-referenced and assigned to 1 of 171 census wards.14
Identification of Children of Sellafield Fathers
Births to fathers employed at Sellafield since 1947 were identified in a record linkage procedure described in detail elsewhere.15, 16
The radiation worker cohort comprised children conceived after the father had been exposed to external ionising radiation while employed at Sellafield (conception was assumed to have occurred 266 days before the child's date of birth). This cohort was compared to children whose fathers had never worked at Sellafield (the non-Sellafield cohort). The non-radiation worker cohort comprised children conceived after the father was employed at Sellafield but who did not have a preconceptional radiation dose and were, therefore, excluded from consideration of the dose-response; this cohort was compared to the non-Sellafield cohort (Fig. 1).
Ascertainment of Cases of Leukaemia/non-Hodgkin's Lymphoma
Cancer registrations for the cohort, recorded throughout the United Kingdom, were obtained from the Office for National Statistics, from 6 regional and national cancer registries and from scrutiny of death registrations, also obtained from the Office for National Statistics.14, 17 The diagnosis of all cases was reviewed centrally, from biologic specimens (33%), pathologic or postmortem reports (18%), clinical records (24%) or case registration information (25%). Only first malignancies were considered. Cases of leukaemia/non-Hodgkin's lymphoma were extracted.
We identified children born in Seascale and cases included in the case-control study of Gardner et al.4
Children were followed up until the end of 1991, death, emigration from the United Kingdom, diagnosis of cancer or age 25 years, whichever came first. Hence, the offspring-years at risk were calculated (Table I).
Table I. Radiation Worker and Non-Sellafield Cohorts: Numbers of Cases and Offspring-Years at Risk
Numbers of cases included in the study of Gardner et al.4 are given in parentheses.
Includes 1 case which Gardner et al.4 did not link to a Sellafield father.–3, 4Excludes cases classified by Gardner et al.4 as lymphoma but reclassified by our pathology review as rhabdomyosarcoma and malignant histiocytosis, respectively.
Measures of population movement used the places of birth of parents, recorded on birth registrations only from 1969 onwards. In addition, places of birth of parents of children born in Seascale during 1950–1968 were available, ascertained for a previous study.14
Places of birth of each child's parents were classified as inside or outside Cumbria, giving an indicator of parental migration. Additional characteristics considered were age, sex, time period of birth, birth order, singleton/multiple birth and paternal social class. Social class was unknown for 1,205 births (0.4%); place of birth of the mother and father was unknown for 1,128 (0.8%) and 7,033 (5%) births, respectively, after 1968.
A measure of community population mixing was calculated for each census ward for each 5-year time period from 1969 onwards. This measure was the proportion of parents (of babies born in the ward during the time period) who were born outside Cumbria and was standardised to have a range from 0 to 1.14 Before 1969, this measure was available only for Seascale ward.
Radiation Dosimetry Information
British Nuclear Fuels plc provided dosimetry data, captured as part of the routine monitoring of the workforce for regulatory purposes and described in detail elsewhere.15, 16
External Ionising Radiation Exposures.
Each employee's exposure, received anywhere within the nuclear industry, was monitored prospectively by film badges and summed over each calendar year to produce annual summary doses, from which the total dose up to the time of conception was estimated.16
Internal Ionising Radiation Exposures.
Urinalysis data were used to identify workers monitored for exposure to plutonium, polonium, fission products, enriched uranium, natural uranium and tritium before their children were conceived. However, analysis was restricted to internal plutonium, fission products and natural uranium (4,880, 2,165 and 667 children of fathers monitored preconceptionally for these radionuclides, respectively). As there were few births with preconceptional internal doses of polonium, fission products, enriched uranium or tritium (144, 2, 95 and 267 exposed births, respectively) and only 1 case of leukaemia/non-Hodgkin's lymphoma among these births, the risk in relation to these radionuclides could not be explored.
Plots of the incidence of leukaemia/non-Hodgkin's lymphoma by age showed that the incidence fell to an approximately constant rate at about age 7 years. Therefore, age was stratified into 0–6 and 7–24 years such that the younger age group included the childhood peak in incidence of leukaemia.18 An additional reason for choosing this age stratification was that the effect of population mixing was restricted to the younger age group14 and most marked when the cut point was age 7 years. The main findings are also reported for the age group 0–14 years, to facilitate comparison with other studies.19, 20 Time period of birth was stratified into 1950–1968 and 1969–1991, corresponding to the year when parents' places of birth became available. As no fathers had more than 1 child with leukaemia or lymphoma, no allowance was made for grouping of children within families. For rates, confidence intervals (CIs) were based on a quadratic approximation to the log likelihood. For rate ratios (RRs), profile likelihood CIs and 2-sided p values based on the likelihood ratio test statistic are reported.21
Comparison of Radiation Worker and Non-Sellafield Cohorts.
Rates of leukaemia/non-Hodgkin's lymphoma in the radiation worker and non-Sellafield cohorts were compared by estimating unadjusted RRs using Poisson regression.21 Adjusted RRs were then calculated using offsets estimated from the effects of the demographic variables in the non-Sellafield cohort,22 excluding births that had missing values of any of the relevant demographic variables. As community population mixing and parents' places of birth were routinely recorded only for births from 1969 onwards, adjusted RRs could be calculated only for the later time period.
Comparison of Non-radiation Worker and Non-Sellafield Cohorts.
Unadjusted RRs comparing the risk of leukaemia/non-Hodgkin's lymphoma in these cohorts were estimated as above. As there were only 2 cases in the non-radiation worker cohort, both born before 1969 when population mixing was unknown, adjusted RRs could not be calculated.
Evaluation of Dose-response Within Radiation Worker Cohort.
Poisson regression was used to estimate the unadjusted RR describing the trend in risk in relation to paternal preconceptional dose of external radiation (continuous variable). RRs describe the trend with dose measured in units of 100mSv (RR100mSv). The dose-response inside and outside Seascale was compared by adding an interaction term. The goodness of fit of the Poisson regression models was checked using Monte Carlo methods.23
For births during 1969–1991, adjustment for demographic variables was performed using offsets estimated, as before, from the non-Sellafield cohort.22 For births inside Seascale before 1969, 68% of parents' places of birth were known. Adjusted RRs were calculated, assuming that all unknown places of birth were either (i) inside Cumbria or (ii) outside Cumbria.
Baseline rates implied by the dose-response models were examined, corresponding to rates of leukaemia/non-Hodgkin's lymphoma in the absence of paternal preconceptional irradiation and, for the adjusted models, corresponding to rates at the lowest level of risk of the demographic variables.
Evaluation of Internal Dose Effect Within Radiation Worker Cohort.
Unadjusted and adjusted RRs were also estimated, comparing the risk among children of fathers monitored and not monitored for natural uranium, plutonium and fission products.
Workforce Consent and Ethical Approval
The planning of the project began in 1990. Ethical approval was obtained from the local research ethics committees. Consent to access birth, cancer and death registrations was obtained from the Office for National Statistics. Consent to access information from employer records was obtained from the workforce: following a detailed presentation of the study to the Company Joint Health and Safety Committee of British Nuclear Fuels plc (which includes representatives of management, trade unions and the workforce), written support for the project was given by the trade unions. Details of the project were then circulated throughout the workforce in the company newsletter by the company medical officer. Workers were given the opportunity to opt out of the study, and 4 men chose to do so. The workforce has been advised of developments in the study by briefings by the study team.
Comparison of Radiation Worker and Non-Sellafield Cohorts
The numbers of live births, offspring-years at risk and cases in these cohorts are shown in Figure 1 and Table I. The diagnoses of cases are summarised in Table II; 90% of the unspecified leukaemias were diagnosed before 1970.
Table II. Numbers of Cases of Leukaemia and Non-Hodgkin's Lymphoma, Diagnosed Under Age 25 Years, by Diagnostic Group, Among the Non-Sellafield and Radiation Worker Cohorts
Acute lymphoblastic leukaemia
Other specified leukaemia
Children of radiation workers had a higher risk of leukaemia/non-Hodgkin's lymphoma than children in the non-Sellafield cohort (for children aged 0–24 years, born 1950–1991, RR = 1.9, 95% CI 1.0–3.1, p = 0.05; Table IIIa). (For children aged 0–14 years, the corresponding RR = 1.8, 95% CI 0.9–3.3, p = 0.08.) The excess of leukaemia/non-Hodgkin's lymphoma was largely among children under 7 years, born in the earlier time period: children of radiation workers, aged 0–6 years and born outside Seascale, had almost a 2-fold risk during 1950–1968, whereas those born in Seascale had a highly significant 15-fold risk in both time periods.
Table III. Comparison of Radiation Worker and Non-Sellafield Cohorts: Rates of Leukaemia/Non-Hodgkin's Lymphoma (Number of Cases/100,000 Offspring-years) and Corresponding Unadjusted and Adjusted RRs Comparing Risk in the Radiation Worker and Non-Sellafield Cohorts
For children aged 0–6 years, those who were first-born (RR = 2.1, p = 0.02), those with both parents born outside Cumbria (RR = 2.1, p = 0.02) and those born in wards with higher levels of community population mixing (RR for trend = 7.5, p = 0.03) were at increased risk of leukaemia/non-Hodgkin's lymphoma (RRs adjusted for all other variables) (Table IIIb). No other demographic variables (sex, singleton/multiple births, social class) showed a significant association with risk either in a univariate model or after allowing for significant variables. As this model included parents' places of birth, recorded routinely only from 1969 onwards, modelling was restricted to children born during 1969–1991.
No demographic variables were significantly associated with the risk of leukaemia/non-Hodgkin's lymphoma in the older age group. Thus, RRs were estimated, adjusting for demographic variables only in the younger age group, born 1969–1991.
Adjustment slightly increased the RR comparing younger children of radiation workers and others (Table IIIb), reflecting significantly lower levels of population mixing among radiation workers outside Seascale compared to non-Sellafield fathers (p < 0.0001). For the 8% of children of radiation workers who were born in Seascale, where population mixing was high (in the top 5% of wards), the adjusted RR was nonsignificant and considerably lower than the unadjusted RR (for children aged 0–6 years, unadjusted RR = 15.4, 95% CI 2.5–50, p = 0.008 and adjusted RR = 4.6, 95% CI 0.7–15, p = 0.09; Table III).
Comparison of Non-radiation worker and non-Sellafield Cohorts
Among children of non-radiation workers, there were 2 cases of leukaemia/non-Hodgkin's lymphoma, both born during 1950–1968, 1 born inside Seascale and diagnosed aged 0–6 years and 1 born outside Seascale and diagnosed aged 7–24 years. (The latter was included as a case in the study of Gardner et al.4) The rate in the non-radiation worker cohort was 9.2 (95% CI 2.3–37) cases/100,000 offspring-years, and the RR comparing this cohort with the non-Sellafield cohort for the entire time period, 1950–1991, was 2.2, 95% CI 0.4–6.8, p = 0.33 (RR = 1.2, 95% CI 0.1–5.2, p = 0.88 and RR = 16.0, 95% CI 0.9–71, p = 0.06 for non-Seascale and Seascale non-radiation workers, respectively). These RRs were higher for children born during 1950–1968, especially inside Seascale. In particular, for non-Seascale-born children aged 7–24 the RR was 2.9 (95% CI 0.2–13.3, p = 0.36) and for Seascale-born children aged 0–6 the RR was 43 (95% CI 2.4–197, p = 0.02).
Evaluation of Dose-response Within Radiation Worker Cohort
The paternal dose distribution among Sellafield radiation workers was strongly skewed (median = 29.2 mSv, range 0.01–826 mSv) and has been described in detail elsewhere.16 Children with external ppi of 50 mSv or more contributed 58,437 (37%) of the offspring-years at risk.
Table IV. Radiation Worker Cohort: Dose-Response for Child's Risk of Leukaemia/Non-Hodgkin's Lymphoma in Relation to Father's Total Preconceptional External Radiation Dose
The risk of leukaemia/non-Hodgkin's lymphoma increased significantly with increasing external ppi (for children aged 0–24 years, born 1950–1991, RR100mSv = 1.6, 95% CI 1.0–2.2, p = 0.05), this effect being determined largely by younger children born in earlier years (Table IVa). (For children aged 0–14 years, the corresponding RR100mSv = 1.7, 95% CI 1.0–2.3, p = 0.06.) This dose-response remained significant when dose categories were used (no cases in the lowest quartile of dose; RR comparing the highest quartile with the 2 middle quartiles = 3.1, p = 0.04). Inside Seascale, the dose-response was significant (for children aged 0–24 years, RR100mSv = 2.0, 95% CI 1.0–3.1, p = 0.04).4 Outside Seascale, the dose-response was elevated but not significant. There was no significant difference in the dose-response among children born inside and outside Seascale (p for interaction = 0.45 for children born 1950–1991, aged 0–24 years). As there was only 1 case of leukaemia/non-Hodgkin's lymphoma in children born in Seascale and aged 7–24 years, the relevant dose-response was not calculated; however, the ppi dose associated with this case was high (>100 mSv). When the 10 cases linked by Gardner et al.4 to a Sellafield father were excluded, the dose-response was nonsignificantly elevated (RR100mSv = 1.4, 95% CI 0.3–2.7, p = 0.56, based on 3 cases).
Among children aged 0–24 years and born during 1950–1991, the baseline rates corresponding to 0 dose in the dose-response models were 5.6, 3.6 and 22.9 cases/100,000 person-years for all children of radiation workers and those born outside and inside Seascale, respectively, compared to the mean rate of 4.3 cases/100,000 person-years in the non-Sellafield cohort.
All models were confirmed to be a good fit (p > 0.10).
Analysis Adjusted for Demographic Variables.
Adjustment was performed for all children aged 0–6 years, born 1969–1991 and for those born in Seascale 1950–1968 (Table IVb).
During 1969–1991, within the radiation worker cohort, first babies, those with both parents born outside Cumbria and those living at birth in areas of high population mixing tended to have lower ppi, suggesting that the children of high-dose fathers had a demographic profile that reduced their risk of leukaemia/non-Hodgkin's lymphoma under 7 years of age. Hence, adjustment increased the dose-response (for children aged 0–6 years, unadjusted RR100mSv = 1.2, 95% CI 0.2–2.7, p = 0.79; adjusted RR100mSv = 1.4, 95% CI 0.2–3.1, p = 0.60; Table IV). For children born in Seascale during 1950–1968, the dose-response was little changed by adjustment (adjusted RR = 2.0, 95% CI 0.9–3.4, p = 0.08 or RR = 2.3, 95% CI 1.0–3.7, p = 0.06, assuming missing parents' places of birth were all inside or all outside Cumbria, respectively).
Among children aged 0–24 years and born during 1969–1991, the baseline rates corresponding to zero dose in the dose-response models were 1.3 and 2.6 cases/100,000 person-years for all children of radiation workers and those born inside Seascale, respectively, compared to the adjusted baseline rate of 2.7 cases/100,000 person-years in the non-Sellafield cohort. For children born in Seascale during 1950–1968, the baseline rate was 6.5 or 5.4 cases/100,000 person-years, assuming missing parents' places of birth were all inside or all outside Cumbria, respectively.
Evaluation of Internal Dose Effect Within Radiation Worker Cohort
Children whose fathers were monitored for natural uranium before conception had an increased risk of leukaemia/non-Hodgkin's lymphoma compared to other children of radiation workers. Three cases were exposed, 2 born in Seascale and 1 elsewhere, whereas 0.8 would be expected (for children aged 0–24 years, born 1950–1991, RR = 3.8, 95% CI 0.8–12.4, p = 0.08). This effect was little changed by adjustment for demographic factors (adjusted RR = 4.2, 95% CI 0.9–13.7, p = 0.06) but was partly confounded with external dose: fathers monitored for natural uranium tended to have much higher external ppi doses than other radiation workers (median dose = 75 and 27 mSv, respectively). After adjusting for external ppi, an almost 3-fold risk remained (adjusted RR = 2.9, 95% CI 0.6–9.8, p = 0.15).
There was no significantly increased risk for children whose fathers were monitored for plutonium or fission products before conception.
Comparison of Radiation Worker and Non-Sellafield Cohorts
We found a significant, nearly 2-fold, risk of leukaemia/non-Hodgkin's lymphoma in children of male Sellafield radiation workers compared to other children. This excess was driven by children under 7 years, born during 1950–1968, and was higher among children born in Seascale, who had a 15-fold risk in both time periods.
The risk of leukaemia/non-Hodgkin's lymphoma among young children is known to increase with population mixing, especially around the time of birth.13, 14, 24 Levels of population mixing in the non-Seascale radiation workers (known only for 1969–1991) were significantly lower than in the non-Sellafield cohort. However, population mixing was very high for Seascale radiation workers throughout the entire study period, 1950–1991, and explained much of the excess of leukaemia/non-Hodgkin's lymphoma in Seascale. Population mixing is a surrogate for the actual risk, postulated to be exposure to infections; had measures of this been available, more of the excess in Seascale might have been explained. The pattern of risk among children of non-radiation workers was similar. In particular, the rate of leukaemia/non-Hodgkin's lymphoma was very high among children born in Seascale to non-radiation workers, consistent with the very high levels of population mixing there.
Comparison with Other Studies
The power of a study to detect an effect depends on the number of subjects and the distribution of the exposure among them: if more subjects have high exposures, the study is more likely to detect an effect if one exists;25 if few subjects have high exposures, the study will have low statistical power, increasing the likelihood of a Type II statistical error, i.e., failure to detect a genuine effect. Studies of the Sellafield workforce are likely to have greater statistical power than those of other cohorts of nuclear workers since the Sellafield workforce has had the highest doses of any in the nuclear industry in Western Europe or North America, exposures to high doses elsewhere having been uncommon.26 Among the 8 case-control studies of leukaemia/non-Hodgkin's lymphoma in children of nuclear workers,4–7, 19, 27–29 that of the Health and Safety Executive (HSE) had the highest number of subjects (32) exposed to ppi ≥100 mSv, followed by the studies of Gardner et al.4 (in particular, that with area controls) and Draper et al.19 (both had 9 subjects, Table V). The studies of McLaughlin et al.5 in Ontario, Urquhart et al.6 in Scotland, Roman et al.28 in Berkshire, Kinlen et al.7 in Scotland and Sever et al.29 in the United States had 5, 1, 0, 3 and 4 subjects exposed to ppi ≥ 100 mSv, respectively. Our cohort study, with 1,737 subjects with ppi ≥ 100 mSv, had greater statistical power than the only other comparable study20 as it included 7% and 23% more live births following ppi of 50–99 mSv and ≥ 100 mSv, respectively (Table VI). Hence, the 4 other studies that included offspring of Sellafield fathers4, 19, 20, 27 are likely to have had the greatest statistical power to detect an effect, but none included more children of high-dose fathers than ours (Table V).
Table V. Results of Studies Comparing the Risk of Leukaemia/Non-Hodgkin's Lymphoma Among Children of Radiation Workers and Others
Roman et al.20 excluding children born in Cumbria before 1986 (numbers of live births as above)
Draper et al.19 excluding Gardner cases and controls
Studies Which Include the Gardner Cases.
Our study, like several other studies that considered the risk of leukaemia/non-Hodgkin's lymphoma among children of radiation workers, included the cases identified by Gardner et al.4 (Table Va). Our results are consistent with these studies in finding the risk of leukaemia/non-Hodgkin's lymphoma among children of radiation workers to be about twice that among other children.4, 19, 20, 27 This excess risk may be partly due to high levels or unusual patterns of population mixing among radiation workers, as suggested previously,2, 7, 19 though we found evidence of high levels of population mixing only among those living in Seascale. It is possible that the lower relative risk obtained by Gardner et al.4 using local controls (1.0, 95% CI 0.4–2.5) may be partly due to closer matching for population mixing, especially in the earlier time period, through matching on parish of birth.
Studies Which Exclude the Gardner Cases.
The 2 studies in the United Kingdom that excluded the Gardner et al.4 cases and had the greatest number of births exposed to ppi ≥ 100 mSv also indicated an excess risk among children of radiation workers (Table Vb).19, 20 In contrast, case-control studies of leukaemia in Ontario and of leukaemia/non-Hodgkin's lymphoma in the United States found no elevated relative risk for children of nuclear workers.5, 29 These low relative risks in North America may be a consequence of the few births exposed to ppi ≥ 100 mSv.
Dose-response Within Radiation Worker Cohort
The risk of leukaemia/non-Hodgkin's lymphoma increased significantly as total external ppi increased (RR100mSv = 1.6, 95% CI 1.0–2.2, for children born 1950–1991 and aged 0–24 years). This dose-response was determined largely by younger children born in earlier years. When the Gardner et al.4 cases were excluded, the dose-response remained elevated but nonsignificant. Adjustment for population mixing tended to increase the dose-response relation as, within the radiation worker cohort, children whose fathers had higher doses were more likely to have parents who were born in Cumbria and to be born in areas with lower population mixing. After allowing for the father's preconceptional radiation dose, population mixing and birth order, the baseline rate among children of radiation workers born during 1969–1991 was similar to that in the non-Sellafield cohort after adjusting for demographic variables, confirming that the combined effects of population mixing, birth order and radiation could explain statistically the excess of leukaemia/non-Hodgkin's lymphoma among children of radiation workers.
Comparison with Other Studies
Our study estimated the trend with dose (as a continuous variable) among children of radiation workers. Most previous studies compared children with ppi ≥ 100 mSv to those with no dose, the latter group including non-radiation workers and/or fathers who did not work in the nuclear industry (Table VI).4, 19, 20, 27 Children of radiation workers in general have a 2-fold higher risk of leukaemia/non-Hodgkin's lymphoma (Table V), partly due to higher levels of population mixing. Hence, it is likely that these other studies overestimated the dose-response through ascribing to ppi dose a risk due to population mixing. However, to facilitate comparison with these studies, we estimated the dose-response using their design.
Studies Which Include the Gardner Cases.
Our RRs comparing children with ppi ≥ 100 mSv and others were similar to those of Roman et al.20 but higher than the relative risks of Draper et al.19 and lower than those of Gardner et al.4 (Table VIa). The high relative risks reported by Gardner et al.4 appear to be largely due to underestimating the percentage of births at risk in west Cumbria before 1986 with ppi ≥ 100 mSv by factors of 2.6 and 1.6 using local and area controls, respectively, thus overestimating the relative risks by these factors. Draper et al.19 also estimated the trend with dose (as a continuous variable) among children of radiation workers and reported a nonsignificantly elevated relative risk for children aged 0–14 years similar to ours (Table VI).
Studies Which Exclude the Gardner Cases.
Roman et al.20 reported an RR of 4.3 (95% CI 0.5–41) comparing those with ppi ≥ 100 mSv and others. However, Draper et al.19 found a nonsignificantly decreasing risk with increasing ppi (Table VIb).
Dose-response Inside and Outside Seascale.
We found a significant dose-response overall, which was significant for births inside Seascale (RR100mSv = 2.0, 95% CI 1.0–3.1, p = 0.04) and remained elevated but nonsignificant for births outside Seascale (RR100mSv = 1.5, 95% CI 0.7–2.3, p = 0.26). The findings of Roman et al.20 for births outside Cumbria are also consistent with an effect outside Seascale. In contrast, the HSE case-control study reported that the association between ppi and the child's risk of leukaemia/non-Hodgkin's lymphoma was confined to children born in Seascale.27 The difference between their findings and ours was partly due to our inclusion of an additional case diagnosed outside Seascale after the end of follow-up by HSE and partly to the distribution of offspring-years in the HSE study being based on a sample of 179 controls and differing from that of the 9,859 births to radiation workers in our study cohort by dose category and Seascale birth status (J. Hodgson, HSE, personal communication). These differences resulted in the HSE estimate of dose-response being higher inside Seascale and lower outside Seascale than ours. Further differences in the same direction are due to HSE using 4 ppi dose categories, whereas we treated ppi as a continuous variable. Hence, the possibility of a dose-response outside Seascale cannot be excluded.
Exposure To Internal Radionuclides
Children whose fathers were monitored for exposure to natural uranium were at increased risk of leukaemia/non-Hodgkin's lymphoma, even after adjustment for external ppi (adjusted RR = 2.9, 95% CI 0.6–9.8, p = 0.15). No other study has considered internal exposure to natural uranium in this context. Further studies are needed to investigate whether our finding is due to confounding of exposure to external radiation and internal uranium or to chance rather than a real effect.
Relative Contributions of Population Mixing and Radiation
In a previous study on this data set, we reported that “population mixing alone could account for the Seascale leukaemia and lymphoma cluster…[although] the possibility of additional risk factors in Seascale remains”.14 Our earlier study did not assess the possible effects of ppi but predicted 3 cases among children born in Seascale due to population mixing, with a CI that included 6, compared to 6 observed. We concluded that population mixing was a major risk factor for childhood leukaemia, which could explain the greater part of the Seascale excess, but did not exclude the possibility of other risk factors. Most cases among younger children of radiation workers are likely to be attributable to population mixing,14 which has almost certainly been the underlying factor drawing attention to the excess in Seascale. However, within the radiation worker cohort, the dose-response relationship was not explained by population mixing.
Ours is the largest and most comprehensive investigation of the risk of leukaemia/non-Hodgkin's lymphoma in children of the workforce with the highest occupational exposure to ionising radiation in Western Europe and North America. It considered a complete cohort of children born in Cumbria and followed up throughout the United Kingdom, thus avoiding the possibility of bias in ascertainment of cases or choice of controls. It included a quantitative estimate of population mixing, a major risk factor for leukaemia/non-Hodgkin's lymphoma.12, 13, 14
Children of radiation workers had a 2-fold increased risk of leukaemia/non-Hodgkin's lymphoma compared to other Cumbria-born children, partly due to high levels of population mixing among Seascale radiation workers. The risk of leukaemia/non-Hodgkin's lymphoma increased significantly as external ppi increased. Although this finding is not independent of that of Gardner et al.4 since most cases were included in both studies, our estimate of the effect of ppi was substantially lower than theirs. Among radiation workers, we found no evidence that ppi was confounded with population mixing. Although a substantial proportion of cases of leukaemia/non-Hodgkin's lymphoma among children of radiation workers are probably due to population mixing, some may be due to ppi and we cannot exclude the possibility of such an effect outside Seascale.
The implications of these findings for the current workforce of the nuclear industry must be viewed cautiously, in particular since current occupational exposures are very low compared to those during earlier decades.
We thank the staff and trade unions of British Nuclear Fuels plc and the United Kingdom Atomic Energy Authority for their cooperation with this study and the Office for National Statistics for providing us with birth registration information: Mr J. Hodgson of the HSE for advice; Drs M.M. Reid and P. Birch and Professor A. Malcolm for reviewing the diagnoses of cancer cases; Professor A.W. Craft and Dr M.M. Reid for comments on the manuscript; the staff of Westlakes Research Institute, in particular, Mr K. Binks, Mr L. Scott, Dr T. Riddell and Ms S. Jones for help with data linkage and dosimetry; and our colleagues in the Department of Child Health, University of Newcastle, for continuing support.